technical document 6 (without appendices) · ITC In cooperation with FAO IKC N NIVR ... FPM-MAG US...

Kingdom of the NetherlandsMinistry of Economic AffairsMinistry of Foreign Affairs, DGISMinistry of Agriculture, Nature Management and Fisheries

USER REQUIREMENTS STUDY

FORREMOTE SENSING BASED SPATIAL INFORMATION

FORTHE SUSTAINABLE MANAGEMENT OF FORESTS

TECHNICAL DOCUMENT 6

REMOTE SENSING APPLICATIONS FOR FOREST MANAGEMENT

March 1999

ITC In cooperation with FAO IKC N NIVRIBN-DLO WAU DOFI NEOFokker Space BV NLR TNO-FEL Vissers DataManagement


FOR REMOTE SENSING BASED SPATIAL INFORMATIONFOR THE SUSTAINABLE MANAGEMENT OF FORESTS

TECHNICAL DOCUMENT 6


Prepared by

International Institute for Aerospace Survey and Earth Sciences [ITC]- Y. Hussin- W. Bijker

Wageningen Agricultural University [WAU]

- D. Hoekman- M. Vissers

National Aerospace Laboratory [NLR]- W. Looyen

March 1999

User Requirements Study: Technical Document 6 – Remote Sensing Applications for Forest Management Page i


FOR REMOTE SENSING-BASED SPATIAL INFORMATION FOR THE SUSTAINABLE MANAGEMENT OF FORESTS Preamble and acknowledgements This study originates from problems observed in relation to information availability for decision-making purposes insustainable forest management as a result of experiences gained in international programmes and processes over thepast decades worldwide. A solution to the problems observed in the supply of information was suggested through theFAME (Forest Assessment and Monitoring Environment) concept comprising an end-to-end forest assessment andmonitoring system. This is an integrated system, with functions for image data input, transmission, acquisition,processing, modeling and archiving, including the education and training required for these purposes. This FAMEconcept was developed by a number of institutes in the Netherlands and was positively received by, among others, theFAO and by the Instituto Nacional de Pesquisas Espaciais (INPE) from Brazil both expressing keen interest tocollaborate with the Netherlands in the further development of this concept. The study aims to address the following issues:- Assessment of requirements for spatial information in order to support sustainable forest management;- Preliminary evaluation of the extent to which these requirements for spatial information can be met by existing and

planned remote sensing systems;- Identification of the requirements for, and components of, an improved information supply mechanism in the form

of an “end-to-end” information system. Three Ministries of the Government of the Netherlands have sponsored the study: the Ministry of Economic Affairs,the Ministry of Foreign Affairs (Directorate General for International Cooperation – DGIS) and the Ministry ofAgriculture, Nature Management and Fisheries. The study was carried out by the International Institute for Aerospace Survey and Earth Sciences (ITC) of Enschede, theNetherlands, in cooperation with:- Food and Agriculture Organization of the United Nations (FAO), Rome, Italy- National Reference Centre for Nature Management (IKC Natuurbeheer), Wageningen, the Netherlands- Institute for Forest and Nature Research (IBN-DLO), Wageningen, the Netherlands- Wageningen Agricultural University (WAU), Wageningen, the Netherlands- DO Forestry International (DOFI), Bennekom, the Netherlands- Netherlands Geomatics and Earth Observation BV (NEO), Lelystad, the Netherlands- Netherlands Agency for Aerospace Programmes (NIVR), Delft, the Netherlands- National Aerospace Laboratory (NLR), Amsterdam, the Netherlands- Fokker Space BV, Leiden, the Netherlands- TNO-Physics and Electronics Laboratory (TNO-FEL), The Hague, the Netherlands- Vissers Datamanagement, Wageningen, the Netherlands

The study results have been incorporated in the “User Requirements Study – Final Report”. This Final Report is basedon the detailed study results contained in a number of technical documents as follows:1. International user identification and platform creation

[DOFI/J.M. Heering]2. Design of the user needs assessment study

[ITC/E. Westinga]3. Forest functions, management principles and information requirements

[IBN-DLO/H. Diemont]4. User needs assessment for spatial forest information; results and analysis

[ITC/E. Westinga; IKC N/H. Savenije]5. Spatial data requirements in sustainable forest management; A study in four tropical countries

[FAO/P. van Laake]6. Remote sensing applications for forest management

[ITC/Y. Hussin]7. User requirements versus existing capabilities

ITC/W. Bijker]8. Proceedings of the international workshop on the preliminary results of the user requirements study

[DOFI/R. Rowe & J.M. Heering]

This study would not have been possible without the collaboration and support of the user community. A special wordof acknowledgement and appreciation is therefore due to those who actively participated in the interactive questionnairesurvey, the country studies, the international workshop or other activities.

User Requirements Study: Technical Document 6 – Remote Sensing Applications for Forest Management Page ii

Abbreviations and acronyms

AfDB African Development BankAPR Airborne Profiler RecorderAsDB Asian Development BankAPFC Asian-Pacific Forestry CommissionATO African Timber OrganizationCCAB-AP Comisión Centro-Americana de Bosques, Areas ProtegidasCCRS Canadian Centre for Remote SensingCIFOR Center for International Forestry ResearchCILSS Comité Interafricain de la Lutte contre les Effets de la Sécheresse au SahelCOICA Coordinadora de las Organizaciones Indígenas de la Cuenca AmazónicaDGIS Directoraat Generaal Internationale Samenwerking / Directorate General for International

Cooperation (Ministry of Foreign Affairs, the Netherlands)DOFI DO Forestry InternationalECOWAS Economic Organization of West African StatesEFI European Forest InstituteELCI Environmental Liaison Center InternationalENVISAT Environmental SatelliteEOS-AM Earth Observation System – AMERS European Remote Sensing SatelliteEU European Union [Commission of the European Union]FAO Food and Agriculture Organization of the United NationsFAME Forest Assessment and Monitoring EnvironmentFPM-MAG US Forest Service Forest Pest Management - Method Application GroupFRA Forest Resources AccountingFSC Forest Stewardship CouncilGEF Global Environment FacilityGFIS Global Forest Information ServiceGFW Global Forest WatchGILS Global Information Locator ServiceGIS Geographical Information SystemGO Government OrganizationGPF Global Programme on Forests (UNDP)GRID Global Resource Information DatabaseIADB Inter-American Development BankIBN-DLO Instituut voor Bos- en Natuurbeheer (Institute for Forest and Nature Management, Mininstry of

Agriculture, Nature Management and Fisheries, the Netherlands)IBRD International Bank for Reconstruction and DevelopmentICRAF International Center for Research in Agro-ForestryIFF International Forum on ForestsIGAD Inter-Governmental Authority on Drought Control and Development in Eastern AfricaIGO Inter-Governmental OrganizationIIASA International Institute for Applied Systems AnalysisIKC-N National Reference Center for Nature ManagementINPE Instituto Nacional de Pesquisas EspaciaisIPF/CSD Intergovernmental Panel on Forests / Commission on Sustainable DevelopmentIRS Indian Remote Sensing SatelliteITC International Institute for Aerospace Survey and Earth SciencesITTO International Timber Trade OrganizationIUCN The World Conservation Union / International Union for the Conservation of NatureIUFRO International Union of Forestry Research OrganizationsJERS Japan Earth Resources SatelliteLANDSAT Land SatelliteLIF Laser Induced FluorescenceLISS Linear Imaging Self-ScanningNEO Netherlands Geomatics and Earth ObservationNFAP National Forestry Action Plannfp national forest programmeNGO Non-Governmental OrganizationNIVR Netherlands’ Agency for Aerospace Programmes

User Requirements Study: Technical Document 6 – Remote Sensing Applications for Forest Management Page iii

NLR National Aerospace LaboratoryNOAA AVHRR National Oceanic and Atmospheric Administration – Advanced Very High Resolution

RadiometerNTFP Non-timber Forest ProductsPAN SPOT Panchromatic modeRADARSAT Radar satelliteRESPAS Remote Sensing Processing and Archiving SystemSADC Southern African Development CommunitySAR Synthetic Aperture RadarSFM Sustainable Forest ManagementSPOT Système Pour l’Observation de la TerreTCA Tractado de Cooperación AmazónicaTFAP Tropical Forestry Action PlanTM LANDSAT Thematic MapperTNO-FEL TNO – Physics and Electronics LaboratoryUNCED United Nations Commission on Environment and DevelopmentUNDP United Nations Development ProgrammeUNEP United Nations Environment ProgrammeURS User Requirements StudyUSDA-MSCO2 United States Department of Agriculture Multi-Spectral Camera O2WAU Wageningen Agricultural UniversityWCFSD World Commission on Forests and Sustainable DevelopmentWCMC World Conservation Monitoring CenterWFI World Forestry InstituteWRI World Resources InstituteWWF Worldwide Fund for NatureXS SPOT Multispectral mode

User Requirements Study: Technical Document 6 – Remote Sensing Applications for Forest Management Page iv

Final Report: Executive summary

BackgroundInterest in sustainable forest management, which has been observed globally over the past two decades, hasincreased the need for proper decision-support systems. The effectiveness of decision-support systems forsustainable forest management is jeopardized by problems related to the availability of relevant informationwithin operational constraints, such as timeliness, affordability and accessibility. Problems of this naturehave been under discussion for a very long time and were confirmed in Agenda 21, chapters 11 and 40 andAnnex III (UNCED, 1992). The problems were also addressed in various FAO activities (nfp, FRA), in theregional hearings of the World Commission on Forests and Sustainable Development (WCFSD), and theEuropean Union studies and more recently by the Intergovernmental Panel on Forests (IPF) and theIntergovernmental Forum on Forests (IFF).

A solution to the problems observed in the supply of information was suggested through the FAME (ForestAssessment and Monitoring Environment) concept comprising an end-to-end forest assessment andmonitoring system. This is an integrated system, with functions for image data input, transmission,acquisition, processing, modelling and archiving, including the education and training required for thesepurposes. One of the features of FAME is the capability of receiving the data required for decision-makingautonomously and in a standardized way. This FAME concept was developed by a number of institutes inthe Netherlands and was positively received by, among others, the FAO and by the Instituto Nacional dePesquisas Espaciais (INPE) from Brazil both expressing keen interest in collaborating with the Netherlandsin the further development of this concept.

The Netherlands’ Ministry of Economic Affairs, the Ministry of Foreign Affairs (Directorate General forInternational Cooperation – DGIS) and the Ministry of Agriculture, Nature Management and Fisheriessubsequently decided to sponsor a study to assess the need for remote sensing-based spatial information ingeneral and possibly the need for a dedicated “end-to-end system”. They commissioned the InternationalInstitute for Aerospace Survey and Earth Sciences (ITC), in cooperation with a number of otherorganizations including the FAO, IKC Natuurbeheer, NLR, DOFI and others, to carry out a study into userrequirements for remote sensing-based spatial information for the sustainable management of forests. Thisdocument presents the findings, conclusions and recommendations resulting from this study, which wascarried out in the period May 1997 – February 1999.

ObjectivesThe User Requirements Study focused on:- Assessment of requirements for spatial information in order to support sustainable forest management;- Preliminary evaluation of the extent to which these spatial information requirements can be met by

existing and planned remote sensing systems;- Identification of the requirements for, and components of, an improved information supply mechanism

in the form of an “end-to-end” information system. Context The User Requirements Study was carried out in the following context:- The main point of departure for this study was the overall objective of sustainable forest management.

For the purpose of this study defined as “managing forest resources and associated lands to meet thesocial, economic, ecological, cultural and spiritual needs of present and future generations” (FSC,1996)

- This study focused on remote sensing options for data and derived information supply. It is fullyunderstood that spatial data obtained through terrestrial surveys are in many cases indispensable. Forsome specific purposes, terrestrial surveys are (still) the only way of ultimately providing theinformation required for example on socio-economic parameters.

- This study concentrated explicitly on information requirements as an essential prerequisite to achievingsustainable forest management. It is, however, fully recognized that other factors of an institutional,legal, and political nature may be equally important.

User Requirements Study: Technical Document 6 – Remote Sensing Applications for Forest Management Page v

Study design A first step in the study was the definition of forest functions (production, conservation, protection andconversion) in order to specify management principles and information requirements from a theoreticalperspective. This exercise accentuated the multi-functionality of forests and forest management, implyingthat information requirements will also be of a comprehensive nature. A classification and identification was subsequently carried out of users of spatial information for sustainableforest management (differentiating between the level of administrative operation, type of organization andtype of decision-making authority). Actual user information needs were obtained through:- A literature study of the most relevant documentation on the subject;- An interactive questionnaire survey among 224 stakeholders/users and- Detailed country studies in Brazil, Nepal, Malaysia and Costa Rica (a study planned in Cameroon could

unfortunately not be carried out for organizational and security reasons). An inventory was made of the existing and planned remote sensing-based information systems. Apreliminary evaluation was carried out of the capabilities of these systems vis-à-vis the identifiedrequirements in order to assess the extent to which the spatial information requirements, obtained through theuser needs study, can possibly be met. The entire study process was supported by the active involvement of the users including the creation of aninternational user network by means of participation in, and presentation of, the study at regional andworldwide meetings of a variety of forestry organizations. The preliminary results of the study were presented at an international workshop in November 1998 at ITCin Enschede, the Netherlands, in which 57 experts from twenty countries and a variety of both national andinternational organizations participated. Information needs The study has revealed a substantial and urgent global need for spatial data and information on forests. Theneed for information is particularly observed at a local level (including forest communities and forestmanagement units) and at sub-national levels (including provincial and state authorities). All themes thatrequire spatial information are relevant, irrespective of level. This applies to land and forest cover, forestfunction allocation, forest types, forest health, bio-diversity, biomass for carbon sequestration, forestproducts and stand parameters. Forest fire themes mentioned are fire detection, fire damage and fire hazard.Furthermore, land tenure, forest dependent communities and socio-economic parameters are indicated. Siteinformation needs include topography, hydrology, soils and geomorphology. Most important are the themesof land and forest cover, and forest degradation. The information requirements in all cases refer to both“state” and “change” parameters. At local and sub-national levels, nearly all themes have a high priority. At global, supranational and nationallevels, a specific priority was observed for information on bio-diversity and carbon sequestration. Information is currently derived from a variety of sources. These include secondary sources, such as lineagencies (for site information such as topography, soils, hydrology and land tenure) and primary sourcessuch as terrestrial surveys (for forest products, bio-diversity, stand parameters and forest health, growth anddevelopment) and remote sensing, both aerial photography and satellite imagery (for land and forestcover/composition, forest types and fire damage) or combinations of these sources. The study has further revealed a considerable discrepancy between the demand for and supply ofinformation. The interactive questionnaire survey indicated that 80% of the respondents need more data andinformation, either more recent, more specific or more detailed.

User Requirements Study: Technical Document 6 – Remote Sensing Applications for Forest Management Page vi

The identified user needs reflect existing and short-term information needs as currently perceived by theusers. Within the various user communities a considerable difference exists with respect to the experiencewith and awareness of the usability of remote sensing derived information, resulting in a wide range ofperceptions. Moreover information needs are not static. They will change as new technologies are developedand users become aware of application possibilities. They will also change in conjunction with changingpolicies and the development and redefinition/refinement of the concept of sustainable forest management.

Capabilities of existing and planned (satellite-based) remote sensing technology The user needs, formulated in forestry terms, were translated into user requirements, using technical terms,as demands to an information system. Capabilities of existing remote sensing-based information systemswere subsequently compared with (i.e. evaluated against) these technical requirements. Thiscomparison/evaluation has revealed that LANDSAT TM, SPOT XS and aerial photography are currentlypotentially the most suitable options available to meet the priority information requirements for land andforest cover, forest types and fire damage assessment, although substantial limitations do still exist. Theseoptical satellite systems and aerial photography are able to provide some information on the requisite scale.LANDSAT TM and SPOT XS are most suitable for smaller scales (1:50,000 and smaller), although SPOTXS might supply information on 1:25,000 scale. Aerial photography is more suitable when more detail(larger scale) is required. NOAA-AVHRR is the best option at the moment for forest fire assessment onsmall scales (1:1,000,000 and smaller). All the systems mentioned above, however, lack the ability to penetrate cloud covers. Although not claimedas a major problem in the questionnaire survey and country studies (but this might be due to a lack ofexperience in the use of imagery), this problem is frequently mentioned in literature and was confirmed atthe international workshop. Satellites equipped with radar sensors, such as ERS-1 and RADARSAT-1, donot have this problem but both lack the ability to accurately distinguish between essential land cover classes. From the point of view of affordability, aerial photography is exceptionally expensive if acquired for the solepurpose of forestry, while high resolution optical satellite remote sensing imagery is prohibitively expensivefor many users. Satellite-based radar data and low resolution optical data from NOAA-AVHRR appear to bethe least expensive but both lack the ability to meet the requirements. The evaluation reveals that there is neither now - nor likely to be in the immediate future - a remote sensing-based information system that can meet all information requirements identified in the study as being essentialfor forest management. Furthermore, the evaluation concludes that the information needs can only partiallybe satisfied through satellite remote sensing. Aerial photography, terrestrial survey, and secondary datasources are equally important. This will remain the case even when satellite imagery with higher spatialresolution becomes available. Constraints in information supply The study has revealed a considerable discrepancy between the demand for data and information for certainthemes and the ability of existing systems to meet that demand. This indicates a considerable under-utilization of existing data sources, caused by a variety of constraints of e.g. a political, institutional,operational and technical nature. The study shows that accessibility to and affordability of existing data andinformation are the major constraints mentioned by all users. Elimination of these two constraints wouldsatisfy the first priority needs for information on land and forest cover, and forest degradation. Lack of user-friendly technology, inadequate data quality, cloud cover and lack of standardization inmethods for data collection and analysis, are also mentioned as constraints of relatively minor importance.There is, however, no proof at present that imagery exists of all geographical areas of interest. This meansthat availability problems may still exist in addition to the usability, timeliness, accessibility, frequency(number of images per year) and continuity problems. Further analysis has revealed weak policies and organizations, and inadequate staffing and staff capabilities,which prevent proper management and operation of spatial information systems at a national level andbelow. At a national level, there is a lack of an information strategy for data dissemination and decentralized

User Requirements Study: Technical Document 6 – Remote Sensing Applications for Forest Management Page vii

data management and use (i.e. an information infrastructure). There is no system that accommodates andintegrates data of a diverse nature, such as remote sensing data, terrestrial observations, and socio-economicdata, from a variety of sources and of different qualities. The distribution of (remotely sensed) data is theweakest part of the chain, from data acquisition to delivery of the information to the desk of the user. Thereare no international protocols to facilitate open exchange of data between users at all levels. The FAME concept revisited At the start of the study, the FAME concept was taken as departure point. This concept is to serveoperational forest monitoring and includes all components necessary to enhance sustainable forestmanagement worldwide: the development of an end-to-end system, comprising simple PC-based receivingstations, data processing and archiving functions, optionally integrated with a GIS, the associated educationand training, as well as a satellite with a dedicated sensor, covering the world’s entire surface. In view of the findings of the study, this concept has been revisited and elaborated resulting in the followingspecific characteristics which have to be taken into account for the further development of the concept and,more specifically, the end-to-end information system which is part of the concept:- In sustainable forest management information is required on a large number of themes that cannot or

cannot entirely be derived from satellite remote sensing, e.g. socio-economic and political informationand NTFP, bio-diversity, degradation, health and stand parameters. This implies that the system will usea variety of space-borne, airborne, terrestrial and existing secondary sources.

- Different users require different themes of information originating from different sources (usability).Depending on the user, the timeliness, frequency, affordability and accessibility criteria will differ.

- The accessibility of data is regarded as being essential: both spatial data to the users, as well as terrestrialdata from users to higher levels (aggregation). This encompasses physical distribution as well as thepolicy and institutional framework for decentralized data management.

- Human and institutional capacity building is necessary to enable users to incorporate spatial informationinto decision-making processes.

At a national level, an information strategy that will provide the policy and institutional framework in whichit is to operate, including research, training and education, must support the end-to-end information system.Internationally an end-to-end information system is to be supported by international protocols for theexchange of data. The proposed end-to-end information system and the environment in which it is to operatecan be schematically presented as follows:

R E S E A R C H&

D E V E L O P M E N T

I N T E R N A T I O N A L

N A T I O N A L

I N F O R M A T I O N S T R A T E G Y

C A P A C I T Y B U I L D I N G

D I S T R I B U T I O N

S P A C E B O R N E

A I R B O R N E

T E R R E S T R I A L

S E C O N D A R Y

P R O C E S S I N G I N T E G R A T I O N

N A T I O N A L

S U B - N A T I O N A L

L O C A L

U S E R S

P R O T O C O L S F O R D A T A E X C H A N G E

A C Q U I S I T I O N

User Requirements Study: Technical Document 6 – Remote Sensing Applications for Forest Management Page viii

The implementation of an end-to-end information system of this kind has been referred to as an informationinfrastructure. A number of major components in the information infrastructure can be identified, i.e. datageneration, capture, transmission, reception and distribution, data processing, data integration and capacitybuilding:

- Data generation and capture: For several themes in sustainable forest management there is currentlysynoptic spatial data available, in particular data on forest cover, forest types and for fire damageassessment. It still has to be established whether there are currently adequate spatial data available tomeet the requirements of operational forest monitoring systems for all geographical areas of interest.Although the necessity was not established during the study, a dedicated satellite sensor might addressthe gaps in synoptic spatial data supply on forest degradation, forest production and bio-diversity.However, the study revealed that in order to improve information supply in the short and medium terms,priority should be given to a better and more widespread use of existing technologies rather than to theestablishment of a new dedicated satellite sensor system. The latter might become an option in theinformation infrastructure in the longer term.

- Data transmission, reception and distribution: To improve accessibility to synoptic spatial data, dataacquisition (transmission and reception) and distribution mechanisms are essential and they shouldtherefore be part of the information infrastructure. Another important functionality that thesemechanisms should support is the flow of locally generated data (such as terrestrial observations)upwards to users at national and international levels. This component only refers to physical mechanismsfor data distribution; policy and institutional aspects are included in the information strategy.

- Data processing: Many users of spatial information will not have adequate capacity (human, equipmentand software) to handle raw spatial data. Geometric and radiometric correction, geo-referencing, re-sampling and primary classification are examples of basic operations on spatial data that require expertknowledge and specific hardware and software. The spatial data will have to be processed to a degreethat suits the purpose of the user. Consequently, the degree to which data must be prepared is variable,between users and purposes. The processing could take place either by specialized agencies (databrokers) that directly communicate with the users, or it could be incorporated as automated procedures inthe physical systems supplied to the user.

- Data integration: The users make use of a multitude of data from different sources, of which synopticspatial data is only one. All of these have a potential function in decision-making and they shouldtherefore be appropriately integrated with the spatial data in order to arrive at proper conclusions. Theintegration should be transparent to the user.

- Capacity building: Institutional capacity at both policy and technical levels has to be increased tointegrate the information system into operational procedures for planning, management and monitoringand to ensure its sustainability within an organizational infrastructure. Staff qualities at all levels need tobe compatible with these requirements, requiring large staff development inputs of an organizational andtechnical nature. Creation of awareness and formal education and training needs have to be identifiedand corresponding programmes set up. National and international linkages of institutes engaged incapacity building should be further developed and optimized.

RecommendationsThe User Requirements Study has clearly revealed a substantial and urgent need for improved informationsupply, in terms of quantity, quality, detail and recentness. It has also made clear that substantial relevantdata and information already exist but that these are not accessible for various reasons. Efforts shouldtherefore be made to make these data and information available through the establishment of mechanisms forinternational data exchange and information strategies. Where required this needs to be supported by thedevelopment of new technology to make existing data and information available. Development oftechnology to meet the information requirements that cannot currently be met by existing sensors has nopriority for the majority of users at this stage.

User Requirements Study: Technical Document 6 – Remote Sensing Applications for Forest Management Page ix

The study has further created an interactive process and network of persons and organizations at all levelsfrom the world forestry community which share the same concerns about information supply.

In order to make use of the momentum created by the study, the following recommendations are proposedfor appropriate follow-up action, both nationally and internationally:

1. Put a supranational mechanism or mechanisms in place through an international protocol for theimproved exchange of data and information between users without impeding existing informationsupplies.

Currently there are no mechanisms for the exchange of (spatial) data between users, within and betweenlevels. The distribution of synoptic spatial data (e.g. from satellite imagery) and the aggregation of localand national data into data sets for use at regional and international levels should be included in themechanism.

2. National governments should formulate and implement an information strategy for decentralized datamanagement for sustainable forest management.

Creating awareness plays an important role in improving the data supply. Political leaders and thegeneral public should become more aware about the need for a sustainable use of the forests and the rolethat information plays in planning, management, assessment and monitoring. This increased awarenessshould result in commitment that can resolve part of the bottlenecks encountered in the distribution ofexisting data, by permitting general use of spatial data and by providing sufficient financial resources toacquire and process data.

Many countries have policies and procedures that prohibit or limit the distribution or use of spatial data.The information strategy should provide the framework in which this flow of data is possible, whilesimultaneously safeguarding other national interests.

3. The developed information strategies should be implemented. This could be done by validating andfurther tailoring the revisited FAME concept through pilot projects. These should encompass thecomplete flow of information from data generation and capture to the presentation of information tothe user.

Pilot projects, selected in countries with varying levels of remote sensing-based technology experience,will serve to develop and validate a physical implementation of an information infrastructure for alimited number of users (with different profiles) in a relatively short period. This will require refinementand prioritization of the user needs. Initially these will make use of existing technology and products. Forthat purpose the possibilities and limitations of existing and planned systems (availability, usability,frequency etc.) will be assessed in detail.

Specific user-friendly technology may have to be developed for the users, providing integration of thedifferent data sources. For satellite imagery, local receiving stations could be installed and reduced pricesfor images should be negotiated. Existing data sets should become more accessible. Capacity buildingmaterial will have to be developed.

Pilot projects in technologically more advanced countries could address the necessity of developingtechnology or data products that are not currently available. In such cases detailed requirements shouldbe specified, guided by the findings from the pilot projects concerned. This also applies to the actualdevelopment of new technology and/or data products, aiming at covering the identified gaps ininformation supply. Even when initiated at present, such development would only give results in thelonger term.

User Requirements Study: Technical Document 6 – Remote Sensing Applications for Forest Management Page x

4. Capacity building modules have to be developed and applied for the different categories of users of theinformation systems.

The users of the end-to-end system will require training in the operation of the system. Theirorganizations may have to adjust their working procedures and decision-making to incorporate theinformation infrastructure. The different management operations (planning, monitoring etc) requiredifferent kinds of training.

In view of the clear perspective that emerged on the desirability of an operational end-to-end system and itsmain characteristics for supporting sustainable forest management at local and national levels, it is of keyimportance to maintain the current global interest in the revisited FAME concept and its implementation.The many (inter) national initiatives in the field of spatial data management and sustainable forestmanagement and the technological developments can provide additional inputs or components to the end-to-end system during its further development.

User Requirements Study: Technical Document 6 – Remote Sensing Applications for Forest Management Page xi



Contents Page

Preamble and acknowledgements i

Abbreviations and acronyms ii

Final Report: Executive Summary iv

Contents xi

Summary xiii

1. Introduction 11.1 Remote Sensing applications in forestry 11.2 Background and objectives 11.3 Forest Resources and Remote Sensing 2

2. Remote Sensing Sensor Systems 42.1 Aerial Photography 42.2 Scanning Sensor System 62.3 Radar Sensor System 9

2.3.1 Fundamentals of Radar System 172.4 Videographic Sensor System 41

2.4.1 Aerial Videography 412.5 Lidar (Laser) Sensor System 46

2.5.1 Introduction 462.5.2 Laser Principles 472.5.3 Laser energy creation 472.5.4 Laser in remote sensing 48

2.6 Literature cited 49

3. Remote Sensing sensor systems applications in forestry 533.1 The applications of Aerial Photography in Forestry 53

3.1.1 Mapping (qualitative): with reasonable accuracy 75% and more 533.1.2 Measurements and estimation (quantitative): with reasonable

accuracy 75% and more 533.2 The applications of scanning imaging system in forestry 54

3.2.1 Mapping (qualitative): with reasonable accuracy 75% and more 543.2.2 Measurements and estimation (quantitative): with reasonable

accuracy 75% and more 543.3 The application of radar system in Forestry 55

3.3.1 Review of literatures of Radar applications in Forestry 563.3.1.1 Applications of mapping, detecting and monitoring forest resources 563.3.1.2 Forest stand parameters 62

3.3.2 Literature cited 813.4 The applications of Videography in Forestry 973.5 Application of Laser (Lidar) in Forestry 105

3.5.1 The applications of Bathymeter Laser System 1053.5.2 Flunrencence laser System 1083.5.3 Literature Cited 109

3.6 Combined use of optical and radar satellite images for forestry applications 111

User Requirements Study: Technical Document 6 – Remote Sensing Applications for Forest Management Page xii

4. Current and future satellites 1144.1 Current satellites 1144.2 Satellites of the Next Decade 1144.3 The future of Satellite Remote Sensing 114

5. Optimal radar airborne and satellite systems for high-resolution forest observation 1235.1 Systematics approach to determine optimal radar satellite configurations 1235.2 New airborne radar systems for high-resolution forest observation 1245.3 References 125

6. Assessment of the role of new developments for high-resolution spaceborne sensor systems 1346.1 Introduction 1346.2 Local ground receiving stations 134

6.2.1 Ground receiving stations in general 1346.2.2 Local ground receiving stations 1346.2.3 The advantages of local ground receiving stations for a dedicated forestry mission 136

6.3 Internet technology 1366.4 Conclusions 137

Appendices

Appendix 1 : Summary of the applications of remote sensing in forestryAppendix 2 : The future of remote sensing

User Requirements Study: Technical Document 6 – Remote Sensing Applications for Forest Management Page xiii

Summary

To meet the various information requirements in forest management different data sources, like field survey,aerial photography and satellite imagery is used, depending on the level of detail required and the extension ofthe area under study.

Before aerial photography was used for forest management purposes, information was generally obtained bymeans of field surveys, identifying and measuring forest types and stands. This is still by far the most accurateand detailed way of measurement, although the lack of geographical positioning systems did not allow accuratelocation of the forests classified. The method is, however very elaborate, time consuming and expensive, and itis nowadays used predominantly for research purposes and for intensive sustainable production purposes.

The traditional aerial photograph resulting from different film types was and still is an important remotesensing tool. Knowledge of photogrammetry and photography is essential for its proper use. For many decadesthe use of aerial photographic data has been accepted by many forest institutions as a tool in various forestactivities, such as planning, mapping, inventory, harvesting, area determination, road lay-out, registration ofdeclined and dead trees etc. on a local, regional or national scale.

For the purpose of consistently and repeatedly monitor forests over larger areas, it is desirable to use remotesensing data and automated image analysis techniques. Several types of remote sensing data, including aerialphotography, multi-spectral scanner(MSS), radar and laser data have been used by forest agencies to detect,identify, classify, evaluate and measure various forest cover types and their changes. Over the past decadestremendous progress has been made in demonstrating the potentials and limitations for identifying and mappingvarious earth surface features using optical remote sensing data. For large areas, satellite imagery has beenshown effective for forest classification, and consequently mapping. It is emphasised that one of the advantagesof the use of remote sensing in forest survey is the relative short time in which most of the required informationcan be obtained.

Gradually other types of remote sensing tools were developed with which forest object properties wereregistered from the air or from space. The new technologies, integrating satellite imagery, analyticalphotogrammetry and geo-information systems (GIS) offer new possibilities, especially for generalinterpretation and mapping and will be a challenge for future research and application. The analoguephotographic data of aerial photographs as well as the satellite scanning data can be digitised and used formulti-spectral or multi-temporal classification and corrections, geometrical or radiometrical. Scanningtechniques are also applicable in aeroplanes.

Nowadays the products of this aerospace technology are considered to be superior to and a replacement of the"old fashioned" analogue aerial photography. However, this technology is additional and complementary to theaerial photography. Sometimes the products are used alone, but in most cases a combination with aerialphotographs is applied. Also fieldwork is and remains essential when applying remote sensing techniques.

Various factors can be mentioned to explain why in managed forests the operational application of remotesensing in the estimation of a number of stand parameters, is relatively low. Foresters are in generalconservative, in the beginning they were reserved in applying aerial photography and nowadays other remotesensing techniques are not embraced whole-heartily. There is a hesitation to take risks when departing fromtraditional data sources. Lack of knowledge of access to data of the specialised technology is an other reasonfor the limited application.

User Requirements Study: Technical Document 6 – Remote Sensing Applications for Forest Management Page xiv

Overview of remote sensing application opportunties for forest management

Type/sub-type Application Scale/Resolu Frequency Cost Limitation

Aerial photography- Panchromatic +++ ++ ++ -- -- - True Color +++ ++ ++ --- -- - Color Infrared +++ ++ ++ --- -- - B&W Infrared ++ ++ ++ --- --

Scanning: Air- MSS +++ ++ ++ - -- - Hyperspectral +++ ++ ++ -- --

Scanning: Space ++ + + - -

Radar: Air ++ ++ ++ -- --

Radar: Space + + + - --(current satellits)

Lidar/laser: Air + ++ ++ -- --

Videography: Air + ++ ++ -- --

Conclusions

An assessment has been made of the use for mapping (qualitative) and measuring (quantitative) with 75%accuracy of various sensor systems for different purposes in forest management.

The results of this assessment are presented in Appendix 1 and have been used for the evaluation against theinformation requirements as summarised in the previous page. For the purpose of this evaluation anassessment was also made of the current and future satellite sensor systems (Chapter 4) and finally anassessment of the use of ground receiving stations and the use of internet for improvement of theaccessibility (Chapter 6).

The key issue in all remote sensing missions is how to get the data/information to the users. There are severalways of doing that. Main aspect to be considered here is the autonomy of users. Users want to beindependent, they do not want to be controlled, they simply want control over what to get. The only way ofachieving that is to give the local user the tools to receive data and the tools to extract the informationlocally. A local, low cost ground station seems the only possibility at this particular moment.

All of the remote sensing work which utilizes the optical portions of the electromagnetic spectrum hasexperienced two kinds of limitations. First, if cloud cover is present, data cannot be obtained using sensorsoperating in those wavelength regions. Second, the spectral regions sampled do not always provide sufficientinformation to differentiate between various forest cover types of interest.

User Requirements Study: Technical Document 6 – Remote Sensing Applications for Forest Management Page 1.

1. INTRODUCTION

1.1 Remote sensing applications in forestry

This publication presents a literature review of the use of remote sensing for measuring, estimating ordescribing forest characteristics and mapping forest cover types. It includes an inventory of remote sensingapplications in forestry. The publication starts with an introduction to remote sensing, followed by a chapteron remote sensing sensor systems. The following chapters give an overview of the applications for forestryper type of sensor (“type” defined by the portion of the electromagnetic spectrum that is used and the way ofrecording (digital or analogue), not by the platform). Another chapter will describe the current and futuresatellite system.

1.2 Background and Objectives

This workpackage is deal with the inventory of remote sensing applications in forestry. The objective was todefine the capabilities and limitations (or state of the art) of remote sensi sensor systems in different forestryapplications and planned developments.

Forests cover large areas of the global land surface. It is well known that they have a marked influence on ourplanet's environmental quality. Also for many developed and developing countries, forests represent animportant income source for their economies. Forests are not left untouched and trees are felled to provide rawmaterial for forest industries -in this way contributing to the growth of the national economy-and for the needof local communities.

A combination of extensive illegal felling and rapid legitimate (but mismanaged) exploitation could in a shortterm bring to an end the beneficial effects of the forest themselves, which have taken a very long time todevelop. Reduction of forest area and deterioration in composition of the forests are prevented in developedcountries, but still continue in most developing countries. In arid and semi-arid zones where moisture can be alimiting factor, the forest does not recover from destruction resulting from over-grazing, shifting cultivation andfire. With an increasing labour force in developing countries where an extensive farming system is practised, itcan be anticipated that a consequent destruction of the forests will continue proportionally. This will lead tocritical damage such as deterioration of watersheds, increased erosion, uncontrolled run-off, flooding andpossible subsequent drought in the low lands and depletion of crops. As stated above the mentioned problemsare caused by lack of proper internal forest management but also by a too great pressure on forests because ofan over-population and by a lack of effective legislation.

It seems clear that the problem of the conservation of forests embraces questions of a political and a practicalnature and its solution sometimes requires international co-operation. The United Nation agencies, such asFAO and UNESCO, are concerned with the future of these forests and initially they have proposed accordinglya World Forest Appraisal. Already in 1975 during an UNEP (United Nations Environmental Programme)meeting it was suggested to have a system for monitoring the remaining tropical forests. The co-operation ofthe countries concerned is required for a practical realisation of such a program. The public should be madeaware of the importance of science and technology which should be applied to the collection of information onthe dynamic processes in tropical and temperate forests. A tropical and sub-tropical forest cover programme forthe determination of quantitative and qualitative changes, periodic assessments of forested areas and changes inthe forest cover was submitted at the FAO and started in 1975.


1.3 Forest Resources and Remote Sensing

To consistently and repeatedly monitor forests over large areas, it is desirable to use remote sensing data andautomated image analysis techniques. Several types of remote sensing data, including aerial photography,Multispectral Scanner (MSS), Radar and Laser data have been used by forest agencies to detect, identify,classify, evaluate and measure various forest cover types and their changes. Over the past decades tremendousprogress has been made in demonstrating the potentials and limitations for identifying and mapping variousearth surface features using optical remote sensing data, that is with wavelength ranging from 0.35 to 1 mm(Figure 1).

Figure 1. The electromagnetic spectrum (after Trevett, 1986)

However, their limitations trigger to continual improvement and interest for sensors that can obtain data atwavelengths beyond the optical portion of the spectrum. The atmosphere is essentially opaque for wavelengthsbeyond 12 microns all the way to almost 5 mm. Although beyond 5 mm the atmosphere itself becomes totallytransparent. The microwave portion of the electromagnetic spectrum covers such range because it haswavelength ranging from 1 mm to 25 cm. These longer wavelengths have capability to penetrate clouds, fogand rain.

Different data sources, like land survey, aerial photography and satellite imagery can be used depending on thelevel of detail required and the extension of the area under study. For large areas, satellite imagery has beenshown effective for forest classification, and consequently mapping. It is emphasised that one of the advantagesof the use of remote sensing in forest survey is the relative short time in which most of the wanted forestinformation can be obtained.The traditional aerial photograph resulting from different film types was and still is an important remotesensing tool. Knowledge of photogrammetry and photography is essential for its proper use. Already duringmany decades the use of aerial photographic data has been accepted by many forest institutions as a tool invarious forest activities, such as planning, mapping, inventory, harvesting, area determination, road lay-out,registration of declined and dead trees etc. on a local, regional or national scale.


Gradually other types of remote sensing tools were developed with which forest object properties wereregistered from the air or from space. The new technologies, integrating satellite imagery, analyticalphotogrammetry and geo-information system (GIS) offered new possibilities, especially for generalinterpretation and mapping and will be a challenge for future research. The analogue photographic data ofaerial photographs as well as the satellite scanning data can be digitized and used for multispectral ormultitemporal classification and corrections, geometrical or radiometrical. Scanning techniques are alsoapplicable in aeroplanes (Konecny, 1991).

Nowadays the products of this aerospace technology are considered to be superior to and a replacement of the"old fashioned" analogue aerial photography. However, this technology is additional and complementary to theaerial photography. Sometimes the products are used alone, but in most cases a combination with aerialphotographs is applied. Also fieldwork is and remains essential when applying remote sensing techniques.

Various factors can be mentioned to explain why in managed forests the operational application of remotesensing in the estimation of a number of stand parameters, is relatively low. Foresters are in generalconservative, in the beginning they were reserved in applying aerial photography and nowadays other remotesensing techniques are not embraced whole-heartly. There is a hesitation to take risks when departing fromtraditional data sources. Lack of knowledge of access to data of the specialized technology is another reason forthe limited application.

Training and experience in sampling design and statistics in addition to familiarity with remote sensingpotentials and the value and the analysis of their data are required. Answers must be known as to which usefuldata can be obtained from remote sensing imagery and which is the relationship between the required forestryand the remote sensing data and how strong is the correlation. Also knowledge on the costs of the remotesensing equipment and of the data interpretation of the software to be applied is essential.


2. REMOTE SENSING SENSOR SYSTEMS

2.1 Aerial Photography

The primary remote sensing sensor systems are: photographic, scanning, radar, lidar and videography. Thephotographic sensor system instantaneously collects an image of a certain area on surface of the earth. The lensof an aerial camera creates an image of the scene at the focal plane. The mechanism of opening the shutter atselected intervals will allow the electromagnetic energy (e.g. visible light) to enter the camera and interact withfilm emulsion to create the negative image. The principles of the photographic system with its displacementsbecause of the central projection, is schematically shown in Figure 2. The output of this sensor is printed ineither black and white or colour photographs. Because the output is in analogue format, the term photo scale isused instead of spatial resolution. In the case of aerial photography the scale is defined by the camera focallength (f) and the altitude of the aircraft(Z):f/Z.

Aerial photographs are taken with forward and side lap e.g. 60% and 20% or 80% and 40% respectively (thelatter for steep terrain). This gives the possibility to study the registered objects stereoscopically and suppliesoverlapping points for triangulation and mapping purposes.

Because of the central projection, aerial photographs show different types of displacements, which mayinterfere with measurements and proper position of objects. On the other hand, relief displacements, whichoccur because of the central projection, enable heights of objects, elevation of points to be measured, andcontour lines to be drawn when aerial photographs are viewed stereoscopically. The true position of objects,points, contourlines and planimetric detail is only obtained when the stereophotos have proper relative positionswhich they had during exposure in flight and absolute(scale and levelling) orientation. Stereophotos orientedunder a stereoscope according to the line connecting the principal and transferred principal points, showincorrect positions of detail because of resulting model deformation. Stabilized mounts and integrated globalpositioning systems(GPS) of the cameras during the flight minimize the amount of longitudinal tilt and lateraltilt of the cameras and give them a nearly vertical position.Because of the orthogonal projection, maps and orthophotos show objects and points in their true orthogonalpositions. Orthophotos were produced from aerial photographs in photogrammetric stereo instruments.Nowadays they can also be compiled from digitized photos by computers. Together with so called stereomatesin which y-parallax is introduced, stereoscopy is obtained. In this way the orthophotos can be used for detailedheight measurements under a mirror stereoscope and for plot location in the field.

Depending on the scale, the photographic coverage varies from 20 ha to 200 km². The net coverage depends onthe amount of forward and side lap. Picture format usually is 23x23 cm. Although the focal length of camerasusually is 152 mm, for low altitude large scale photography a longer focal length of 205 mm or 300 mm ispreferred in order to have a better view of the terrain in between trees. This will facilitate a better landing of thefloating mark at the foot of the trees depending on the mutual distance between them. white infrared, truecolour and colour infrared. The spectral range varies from 0.4-0.9 m. In general panchromatic and colour filmsare sensitve to the range 0.4 m(blue) to 0.7 m(red) of the visible part of the electromagnetic spectrum. Thesensitivity of infrared and colour infrared films add an extension to approximately 0.9 micrometr andincludes some reflected infrared. Filters are in use such as Wratten 12(minus blue) for black and white infraredand colour infrared. Wratten 25A(red) and Wratten 89B(infrared) filters are in use to accentuate the separationof trees and stands of conifers and broadleaves on panchromatic film. It is possible to print black and whiteinfrared photos from colour infrared film.


Figure 2. Photografic Sensor System

Figure 3. Aerial photography


The smallest ground detail registered in a photograph that can be detected and measured is the definition ofresolution. Ground resolution varies between 0.1 and 2 m depending on several factors. Table 1 according toAldrich, 1979 shows some data:

Table 1. Ground resolution, photographic scale and film type to measure certain selected basic resourceparameters

_____________________________________________________________________ Ground Film type Smallest scale Platform

Parameter Resolution (1) for Measurement (2)1:

______________________________________________________________________Stand size 0.3 BW 9500 MAP, LAPTree height 0.3 BW 9500 MAP, LAPCrown diameter 0.3 BW 9500 MAP, LAPCrown area 0.3 BW 9500 MAP, LAPNumber of trees 0.3 BW 9500 MAP, LAPDead, declined trees 0.3 CIR 6400 LAPCover type 3.0 CIR 92000 HAP

BW 125000 HAPSpecies composition 0.1 CIR 1500 LAP______________________________________________________________________(1) BW - Panchromatic(Panatomic B410); IR - Infrared (Eastman Kodak Infrared Aerographic 2424); CIR - Colour Infrared(Aerochrome Infrared 2443)(2) LAP - Low altitude photography; 150 -3660 m

MAP - Medium altitude photography; 3660-9200 mHAP - High altitude photography; 9200-19800 m.

The resolving power of metric camera lenses is in the order of 30 to 40 line pairs per millimetre using a highcontrast black and white bar chart. In the majority of the combinations camera-film, it is the film that restrictsthe ground resolution. It must be realised that contrast ratio in forests and its surroundings is low, which willreduce resolution.

Interpretation and mapping devices have optics of which the magnification allows detection and measurementof fine details. Optical magnification of a typical pocket stereoscope is 1.5 time, of mirror stereoscope 2.5 to 6times and of zoom stereoscope is 2.5 to 30 times.

2.2 Scanning Sensor System

Multispectral scanning system produces segmented images sequentially arranged in a multispectral fashion.Like the photographic sensor system it is a passive remote sensing system. It uses the natural solar opticalenergy reflected from objects to create its own image. The reflected or emitted energy is detected and electricsensors generate an electric signal that corresponds to the reflected energy variations of the ground scene andrecords it as digital numbers. Digital numbers(DN) refer to the quantized and calibrated values for individualpixels i.e. they refer to transformed magnitudes of the received signals -geometrically and radiometrically-encoded into a discrete number of steps prior to being distributed to users in the form of computer compatibledigital tapes(CCT) of limited scenes. Formally the pixels were presented only in units of in-band radiance or inspectral radiance.


The system consists of four major parts: optics or photo cells to receive and focus the energy, filters to dividethe energy according to its wavelengths, detectors to measure the amount of reflection, and recording systemto save the reflected energy in a digital image format. Figure 3.1 is a typical multispectral scanner system.The area on the ground from which the energy is reflected is called the instantaneous viewing area. Theadjective instantaneous emphasizes the fact that, as the scanner mirror moves, a different portion of theearth's surface will come into view (Lindenlaub, 1976). The spatial or the ground resolution element of thisimage is represented by the instantaneous viewing area. An image produced by a scanner consists of a groupof picture elements(pixels) that can be treated one at a time of a group together in a picture-like format. Theimage is arranged in rows and columns and the scale of grey of the image is formed by the relativebrightness( or darkness) of each pixel. Satellite imagery systems used for forest registration andmeasurements as presented by Kalensky, et al., (1991) are given in Table 3.

Multispectral scanning system (MSS) was developed to operate from aircraft or satellite. The cross-trackmultispectral scanner systems have been developed and used since the 1960's. One of the earlier systemdeveloped was the Daedalus NASA U-2 aircraft scanner system (Sabins, 1987). As can be seen in Table 3 thesensors of Landsat-5 MSS and SPOT-HVR(Saspectral bands and two near infrared bands. The two visiblebands for Landsat MSS: 0.50-0.60 and 0.60-0.70 micrometer and for SPOT: 0.5-0.59 micrometer and 0.61-0.68 micrometer for vegetation reflectance optimums and maximum chlorophyll absorption respectively. Thetwo near infrared bands for Landsat MSS are: 0.70-0.80 micrometer and 0.80-1.10 micrometer and for SPOT:0.79-0.89 micrometer. The latter for contrast between living and declined and dead vegetation. Landsat TM hasextra bands: one in the visible range: 0.45-0.52 micrometer, two in the mid-infrared range: 1.55-1.75 micrometerand 2.08-2.23 �m and one in the thermal infrared: 10.4-12.50 micrometer tellite Probatoire d'Observationde la Terre-High Resolution Visibility) record in two visible and one infrared spectral band.

Figure 4. Multispectral scanning system


Whatever system is used for the registration of forest objects, one should realise that only approximately 10%of the solar energy that reaches the earth is reflected by forest and used for the image forming. Satellites recordradiance measurements which, depending on the field of view of the sensor, give so called pixels (single pictureelements). Landsat MSS provides pixels with a size of nearly 80 m by 80 m. The spatial resolution is 80 m. Theground area of a single image is 34000 km² or approximately 185 km by 185 km. Landsat TM provides asmaller pixel size: 30 m by 30 m and SPOT HVR even 20 m by 20 m (10 m by 10 m for panchromatic band -0.5 - 0.73 micrometer).

Table 3. Satellite systems for forest mapping and monitoring and measurements (after Kalensky et al., 1991).

_____________________________________________________________________Remote LANDSAT SPOTsensing ____________________________________________________________systems MSS TM HRV______________________________________________________________________Ground 80 m 30 m(VIS, IR) 10 m(PAN)resolution 120 m(TIR) 20 m(SPECTRAL)______________________________________________________________________Numbers of 2 Visible 3 Visible 1 PanchromaticSpectral 2 Near-IR 1 Near-IR 2 Visiblebands 2 Mid-IR 1 Near-IR

1 Thermal-IR______________________________________________________________________Spectral 0.5-1.1�m 0.45-12.5�m 0.5-0.89�range_______________________________________________________________________Ground area 34 000 km² 34 000 km² 3 600 km²_______________________________________________________________________Frequency 16 days 16 days 26 daysof repetitive (or better)coverage_______________________________________________________________________Compatible 1: 200 000 1: 50 000 1:50 000mapping to to to scales 1: 1000 000 1: 1000 000 1: 500 000_______________________________________________________________________Comparable Natural forest Forest types As for TMground Plantations Forest damages plus standclasses Forest burns Logging sites heights

Logging sites Reforestation stratificationWoodlands Roads, riversetc. etc.

________________________________________________________________________Main Low cost 7 Spectral 10 m/20 m

bandsadvantages Archives 30 m resolution resolution

since 1972 Medium costs StereocapabilityCoveragerepeatability

_________________________________________________________________________


2.3 Radar Sensor System

All of the remote sensing work which utilizes the optical portions of the electromagnetic spectrum hasexperienced two kinds of limitations. First, if cloud cover is present, data can not be obtained using sensorsoperating in those wavelength regions. Second, the spectral regions sampled do not always provide sufficientinformation to differentiate between various forest cover types of interest.Radar data, which utilizes the microwave portion of the spectrum can provide important additional informationof terrain surfaces and vegetation canopies. Longer radar waves can penetrate vegetation canopy more deeplythan optical wavelengths. This additional information may be able to facilitate differentiation between forestcover types that optical and infrared sensor systems are not able to accomplish. Moreover, in principle thephysical basis for microwave remote sensing is different from optical sensing so far as interaction withvegetation is concerned. Optical sensors depend on differential scattering and absorption caused by thechlorophyll and leaf structure as well as leaf area and leaf area contents. Microwave sensors respond to thelarger scale structure of the canopy, plant water content, plant part size distributions, and in the case of longerwavelength sensors, surface soil conditions. Several studies have concluded that radar data allows identificationof additional cover types that the TM optical and infrared data can not accomplish.Radar is an acronym for radio detection and ranging. It is an active remote sensing system. A radar systemilluminates the terrain with its own (e.g. man made) electromagnetic energy from radio to microwave ranges ofthe spectrum. The system detects the energy returning from the terrain (called radar return), and then recordsthis energy as an image. The radar antenna transmits the pulse of energy and receives the return from the target.An electric switch, or duplexer, prevents interference between transmitted and received pulses. Pulses of energytransmitted from the antenna illuminate strips of terrain in what is referred to as the look direction or rangedirection. The look direction is oriented perpendicular to the azimuth direction (e.g., the aircraft or spacecraftpath). The combination of range resolution and azimuth resolution determines the dimensions of the groundresolution cell, which in turn determines the spatial resolution of a radar image.In Figure 5 the main elements of side looking aircraft radar (SLAR) are depicted.Aircraft radar system started to be operational in the 1960's with short wavelength (e.g. Ka-band 0.83 cm). Adecade later RADAM (Radar Amazon) was the most significant survey project done by aircraft radar system,operate in the X-band (i.e. 3 cm) to map the forest of the Brazilian Amazon basin. Later several aircraft radarsystems were developed, especially after the development of the Synthetic Apertyure Radar(SAR) system suchas Westinghouse AN/APQ-97, Motorola, Goodyear GEMS, NEDSLAR.


Figure 5. The main elements of SLAR (after Trevett, 1986)

Since June of 1978, with the launching of the Seasat Radar Satellite SIR-A (1981) (Shuttle Imagery Radar-A)and SIR-B (1984), a great deal of interest has been sparked in both the scientific and engineering communitiesto further investigate the capabilities, technologies and applications of radar.Today, we are at another turning point for a major development of remote sensing techniques. All theadvantages of radar systems, the results of all interest in radar data and the research work that has been done inthe last decade, have led to several programs to launch unmanned radar satellites this decade. In this decade,there is a wide variety of radar data available from satellites, including various frequencies, polarizations, andincidence angles. The European ERS-1 and ERS-2, the Japanese JERS-1, and the Canadian Radarsat-1satellites already are in space. Experimental SIR-C missions 1 and 2 already accomplished in 1995 to help theU.S. space program SAR system in the space station. In addition, AIRSAR, Radarsat-2, SIR-C 3rd mission,and the European ENVISAT will guarantee a decade of continuous satellite radar data acquisition.


Table 4. Radar satellite systems for forest mapping, monitoring and measurements


Moreover, Spot Image is planning to have radar sensor on board Spot by the end of this Table 4 gives data ofradar satellite systems for forest mapping, monitoring and measurements. Before directly getting to the issue ofthe forest stand estimations from radar backscatter, and since radar is a relatively new system to be used in thisfield, it is necessary to introduce the fundamentals of the radar system.

2.3.1 Fundamentals of Radar System

RADAR is an acronym for "radio detection and ranging" . It is an instrument in the microwave system thattransmits a microwave signal and then receive its reflection as the basis for forming digital or pictorialimages of earth's surface. The "detection" is done by sensing the reflected energy, while "ranging" capabilityis accomplished by measuring the time delay from the time a signal is transmitted toward the terrain until itsecho is received, through this radar can accurately measure the distance from the antenna to features on theground. Another unique capability of radar being an active sensor, is its ability to detect frequency andpolarization. Frequency is the number of wavelengths that pass a point per unit time, while polarization is thedirection of orientation in which electrical field vector of electromagnetic radiation vibrates (Sabins, 1987).Because the sensor transmits a signal of known properties, it is possible to compare the received signal withthe transmitted signal. Figure 6 illustrates the basic procedure of SLAR system.

Imaging radar systems fall into two different categories: real aperture, with which the term SLAR is oftenassociated, although RAR is now more commonly used, and synthetic aperture radar known by the acronymSAR.

Imaging radar as presented in Figure 6 includes several components: a transmitter, a receiver, antenna array anda recorder. A transmitter is used to transmits repetitive pulses of microwave energy at a given frequency. Areceiver accepts the reflected signal as received by the antenna, filters and amplifies it as required. The antennaarray transmits a narrow beam of microwave energy, it is composed of wave guides, devices that control thepropagation of electromagnetic wave so that the waves follow a path defined by the physical structure of theguide. Usually same antenna is used both to transmit the radar signal and to receive echo from the terrain. Therecorder records and/or displays the signal as an image.


Figure 6. Block diagram of a radar imaging system (Sabins, 1987)

Real-Aperture Side-Looking Airborne Radar (SLAR)

Microwave energy transmitted from an antenna in a very short bursts or pulse. These energy pulses are emittedover a time period in the order of microseconds (10-6 sec.). In Figure 7 the propagation of one pulse is shown byindicating the wavefront locations at successive increments of time. By electronically measuring the return timesignal echoes, the angle, or the distance between the transmitter and reflecting objects, maybe determined. If weare using pulses t sec wide, the smallest resoluble distance between transmitter and object or in the slant rangedirection to the aircraft is:

SR = ct/2

where; SR is the Slant Range, c is the speed of light (3 x 108 m/sec) and t is the time between pulse transmissionand echo reception. The factor of 2 in the above equation means that time is measured for the pulse to travel boththe distance to and from the target, or twice the range (Trevett, 1986).

SLAR systems produce continuous strips of imagery representing very large areas located adjacent to the aircraftflight. An image is produced through the forward motion of the aircraft synchronized to the motion of the filmrecorder.

The ground resolution cell size of a SLAR system is controlled by two independent sensing system parameters:pulse length and antenna beam width. The pulse length of the radar signal is determined by the length of time theantenna bursts its energy. As shown in Figure 8, the signal pulse length dictates the spatial resolution in thedirection of energy propagation. This direction is referred to as the range direction. The antenna beam widthdetermines the resolution cell size in the along track or so-called azimuth direction. In fact the antenna isdesigned to give a fan-shaped beam, it is broad in the range direction to allow a reasonably wide swath to beobtained, but narrow in the azimuth direction to permit good along track resolution.


Instalclick

Instalclick

Instalclick

Instalclick

Figure 7 Side-looking airborne radar (SLAR) system operation (after Lilesand, 1986)

The angular beamwith of the antenna in a plane is directly proportional to the wavelength of the transmitted

pulses, and inversely proportional to the length of the antenna, l. That is :

B = � l

where is the operation wavelength and l is the length of the antenna.

Consider this to be the azimuth beam width of the antenna in the SLAR system. The corresponding azimuth

resolution is:

ra = ( � l) * R

where : R is the slant range to the target

Then, employ the typical values of = 0.2 m, R = 1000 m, with the azimuth resolution of say 30 m, then the

antenna length required is 6.7 m long - which is large to imagine to be mounted on an aircraft.

The antenna beam width can be controlled by one of two ways : (1) by controlling the physical length of the

antenna; and (2) by synthesizing an effective length of the antenna. Those systems wherein the beam width is

controlled by the length of the antenna are called brute force, real-aperture, or non coherent radar. The antenna


in brute force system must be many wavelengths long for the antenna beam width to be narrow. Moreover, the

antenna length requirements of brute force systems show problems when detailed resolutions are sought.

The use of longer wavelength and quality improvement of the imagery requires increasing the antenna length;

and obviously there is a physical limit on the length of the antenna that can be accomodated on the side of the

aircraft using real aperture radar (RAR). To overcome this constraint, side-looking airborne synthetic aperture

radar or SAR has been developed and is now widely used. The basic concept of SAR or coherent radar is to

record the change in amplitude and frequency of the reflected signals, as the antenna (fixed on the aircraft)

moves forward in the azimuth or flight direction.

A change in the observed frequency of electromagnetic waves and sound waves caused by the relative motion

between relative motion between source and observer is termed a Doppler frequency shift or Doppler effect.

This can be illustrated by the change in tone of sound waves caused by the whistle of a train, as it passes by the

observer. Although the SAR antenna is shorter than the RAR, nevertheless it has a wider field of view or beam

swath. A target therefore enter into the beam and as the platform moves past the target, and will remain in the

field of view for a period of time before passing out of the beam. The SAR like the RAR is also side-looking

and the beams increase in width across the observed area with distance from the platform. The result is that a

target at a distance from the platform will remain in the field of view longer than a target closer to the platform.

The effect of the synthetic aperture antenna with a target can be seen in Figure 8.

Figure 8 The interaction of a target with an airborne SAR on its recording as a hologram on film (after Trevett,

1986)


Instalclick

Radar System Parameters

Wavelength

It is interesting to know that wavelength and frequency are related in free space according to the equation below:

ƒ (MHz) = 300 � (m)

At wavelength of 1 µm for example the corresponding frequency is 300 x 1012 or 300 THz (Forster, 1992). To

assist in the understanding the range of units Table 5 summaries the metric prefixes that will be encountered

when dealing with electromagnetic radiation.

Table 5. Metric Prefixes (after Forster, 1992)

PREFIX NAME MEANING

f femto x 10 -12

n nano x 10 -9

µ micro x 10 -6

m milli x 10 -3

k kilo x 10 3

M mega x 10 6

G giga x 10 9

T tera x 10 12

The electromagnetic spectrum shows how much wider the microwave section as compared with the visible

segment. Moreover, it can be seen how each of the microwave bands vary in their width. Radar wavebands are

basically radio waves in higher frequencies as shown in Figure 2.5. Within the UHF and SHF bands there are

certain bands which can be used for remote sensing purposes without causing interference with and from other

military transmission bands. These bands have been allocated by the International Telecommunication Union.

The one commonly referred to in radar remote sensing was first introduced during World War II and employed

a lettering code for specific bands. Figure 9 presented these different microwave bands.


Figure 9. The microwave bands as elements of wavelength and frequency (after Trevett, 1986)

The radar image is a representation of the amount of energy backscattered from a surface, the returned energy

vary with the degree of roughness of surface (Trevett, 1986). The wavelength used will also affect the

representation of this roughness. Short wavelengths generally give more pronounced roughness effect.

However, in areas with excessive rough texture the shorter wavelength may give bright to almost non

interpretable image, whereas in more smooth areas it may be useful.

The selection of wavelength should be related to the degree of penetration required, moreover, one major

component should also be considered, the dielectric constant. The reflectivity of the target is related to its

moisture content or the magnitude of dielectric constant. The long wavelengths, have greater degree of

penetration.

Since antenna design is a big factor that influence energy source and wavelength of radar, it is not common for

SLAR system to have multi-frequency capability. Most radars tend to be designed for a specific frequency or

wavelength. However, it is envisioned to achieve this multi-frequency capability in the future. Most radar

systems tend to be designed for specific frequency or wavelength and it is essential to understand the respective

attributes of the different wavelengths in order to be able to interpret the image.

Polarization

Polarization as also defined in section 2 is the direction of orientation in which th electrical field vector or

electromagnetic radiation vibrates. Electromagnetic energy propagates through the air and in any medium, as a

combination of an electric and a magnetic fields (Forster, 1993). These vibrate at right angles to each other in a

regular manner (Figure 10).


Figure 10. Electromagnetic wave in two polarization (after Trevett, 1986)

Most imaging radars use an antenna which generates a horizontal waveform, this is referred to as horizontal

polarization. Upon reaching the target a portion of the returned energy may still retain same polarization as

transmitted signal, however, some surfaces tend to depolarize a portion of microwave energy. This condition in

which a signal is depolarized can be further a function of surface roughness. Through the switching mechanism

in the recording system, alternating between two receiver channels would permit both the horizontal and the

vertical returns to be recorded. Notations HH is used to denote horizontal send and horizontal received, while

HV being horizontal send and vertical receive, in a like manner VV and VH are also used. HH and VV

represent like polarization, while HV and VH are known as cross-polarization return. HV and VH tend to

produce similar results, therefore they are not used simultaneously. The like or cross polarization is an important

factor when considering the orientation of the ground targets or their geometric properties.

Radar Imaging Geometry

The nature of an imaging radar to illuminate a scene from an oblique point of view gives rise to specific

geometric characteristics. It is therefore essential to understand some of the basic factors in radar geometry and

the construction of the image before further radar image analysis. Figure 11 demonstrates the elements of SLAR

geometry.

Whereas, the depression angle is the angle from the horizontal to the mid line of the scan, note that this is not

necessarily the mid-point across the swath on the image. In the near range portion of the illuminated range

swath the depression angle is relatively steep, and in the far-range portion it is relatively shallow. The incidence

angle is defined as the angle between the radar beam and a line perpendicular to the ground surface (Sabins,

1987). It is an alternative term in defining the angular relationship between the radar beam and the ground

surface. The look angle of radar is defined as an angle between geographic North and the direction in which the

radar beam is pointing, that is, perpendicular to the flight direction of the platform. Look directions

perpendicular to topographic alignment will tend to maximize radar shadow, whereas look direction parallel to

topographic orientation will tend to minimize shadow. On the other hand, look direction is the direction of the

look, or direction in which the antenna is pointing when transmitting and receiving pulses (Collwell, 1983). The

exact values of the look angle vary with the design for specific radar systems, but some broad generalizations are

possible concerning the effects of varied look angles. First, at steeper depression angles a radar signal


illuminates a smaller area than same signal at shallow depression angles. Secondly, the slant-range geometry of

a radar means that all landscapes are viewed oblique. As a result, the image tends to record reflections from the

sides of the features. This obliqueness and the degree to which sides are viewed rather than the tops of the

features, varies with look angle. In some landscapes, the oblique view may be very different from the overhead

view to which we are accustomed in the use of other remotely sensed imagery. Nadir is the direct vertical or

plumb point under the platform (Trevett, 1986).

Figure 11. The main elements of SLAR geometry (after Trevett, 1986)

Image distortions inherent in the slant-range representation of the radar are controlled to some extent by the

depression angle, or by incidence angle of the radar beam and the radar look direction. Some of this image

distortions associated with imaging geometry will be discussed in subsequent sections.

Slant-Range and Ground-Range

As shown in Figure 12. the slant-range R is the radial distance from the nadir point of the aircraft or ground track

to the target. It is the plane where direct measurements between the antenna and the target maybe projected, that

is, time intervals between transmission of a radar signal and reception of its return signals from each illuminated

ground surface. The slant-range image plane extends from the antenna to the far range limit of the swath. It is

the natural radar coordinate in the direction perpendicular in the direction perpendicular to the flight line.

On the other hand, the ground-range Rg is determined by the horizontal ground distance from the nadir of the

radar to the point on the ground or extends to the far range portion. It is a factor of the height of the platform

above the ground. In a ground-range display the image scale in near range and far range remains identical.


However, in a slant-range display the projection of various ground points and distances onto the slant-range

plane is subject to scale changes from the near range relative to the far range. This compression in radar imaging

geometry is called slant-range distortion.

Figure 12. Slant range and ground range (after Trevett, 1986)

Foreshortening

The compression mentioned in the preceding section which resulted in the projection of terrain onto the slant-

range plane is called foreshortening. This distortion is due to the occurrence of relief in the target area, since the

amount of foreshortening is dependent on the time during which the slope is illuminated by the radar. As

illustrated in Figure 13, the distance a-b or the bottom and top of a hill, projected onto the normal plane would be

a-b", however, projected onto the slant-range image format it becomes a'-b' which it will be seen is shorter than

the true projected distance a-b". Based on this figure, slopes facing the platform will be shortened relative to

their true projected distance.

Figure 13. Foreshortening (after Trevett, 1986)


For steep slopes, foreshortening is at a maximum when the slope is orthogonal to the incident radar beam, or

when the local incidence angle is zero and the base, the slope and the top are imaged simultaneously and occupy

same position in the slant-range plane. In Table 6 we can see that with increasing incidence angle the

foreshortening effects are reduced. Only where the incidence angle equals 90o the slope foreshortening effect is

eliminated.

Table 6. Percentage foreshortening against incidence angle (after Trevett, 1986)

__________________________________________________________

Angle of incidence Foreshortening

(degrees) (%)

90 0.0

80 1.5

70 6.0

60 13.4

50 23.4

40 35.7

30 50.0

20 65.8

10 82.6

0 100.0

__________________________________________________________

A slope facing the radar will receive all the energy which would usually be spread over the ground projected

area. On the back slopes the grazing angle diminishes so the backscatter is reduced, therefore slopes facing

away from the radar show a weaker return. The overall effect is that in hilly place slopes facing the radar will be

brighter than normal and reverse slopes darker, the effect will increase with the severity of the terrain.

Layover

When radar has a steep angle and the terrain has extreme relief the foreshortening effect can increase to the point

where the higher point is imaged in advance of the base. This range dependent displacement of elevated features

give rise to radar layover effects. By extending the Figure 13 layover is demonstrated in Figure 14.


Figure 14. Radar layover (after Trevett, 1986)

Figure 15 Radar layover as a function of depression angle (after Trevett, 1986)

In terms of the slant range position, upstanding features such as mountain summits, ridge crests or the top of the

buildings, may lie at shorter slant-range distance than their bases. The reason for this displacement is that the

reflected energy from the upper portion of the feature is received before the return from its lower portion. Radar

layover therefore tends to occur mainly in the near range of the swath and will decrease in effect towards the far

range (Figure 15). If the angle of slope facing the radar is less than the angle of the incident radar beam, then,

layover exists. If the two angles are equal, summit and the base occupy the same slant-range position and their

position on the radar image coincide. In the case where the slope angle exceeds the angle of the incident radar

beam, layover does not occur.

Moreover, the amount of layover is influenced by three factors (Sabins, 1987):

1. Height of targets. Taller targets are displaced more than sorter targets.

2. Relationship between the radar depression angle and slope angle.

3. Location of targets. For targets of the same height, those located in the near range are displaced

more on the image than are those in the far range because depression angle is steeper in the near

range.

Because of the complex factors that cause layover, it is not practical to correct this distortion. Layover can be

minimized by acquiring images of shallow depression angles, but radar shadows may excessive for terrain with

high topographic relief.


Radar Shadow

Slopes facing the radar beam produces bright return, but, on the other side, it produces radar shadow. The radar

shadow is an area of no data or total black. Shadowing effects occur in the down-range direction. They are good

indicator for determining look direction on the image. Radar shadow is one of the factors which produces the

representation of relief on the radar image.

The principles of radar shadow is shown in Figure 16. In this case the beam hits the forward base of the hill

slope a to produce projected point a', similarly all the slope is up to the peak b projected as b'. No signal is

received at the backslope base c, no signal reaches the ground until the point d produces the next return after b.

Between b and d there is no record of a return signal and the time lapse record appears as no response or black.

In Figure 17, the relationship of slant-range, depression angle and slope angles is shown. As depression angle

decreases, as across the swath, so will the radar shadow of the same slope increase. If the two angles are equal,

the radar beam will "graze" the back slope, no detailed record of the back slope will occur, but neither will there

be a radar shadow.

Figure 16. Principles of radar shadow (after Trevett, 1986)

Figure 17. Relationships of radar shadow geometry and depression angle (after Trevett, 1986)


Radar Target Parameters

For interpreters who are concerned with visual discrimination of radar images, the degree to which they can

interpret an image depends upon whether they can identify a change in tone or texture due to system parameters

or being related to changes in vegetation or surface material. Aside from his/her knowledge of the area and the

parameters being investigated, it is desirable to understand those factors which influence radar backscattering of

the target. The radar energy sent to the target is reflected or scattered by the target. Some of the energy is

absorbed, some reflected away, some diffused within the target and some eventually returns back to the antenna

to be recorded and measured to produce the eventual image. Six target parameters will be discussed in the

succeeding sections, they are: surface roughness, complex dielectric constant, surface scattering, volume

scattering, point target and bragg resonance.

Surface Roughness

Surface roughness is the terrain property that most strongly influences the strength of radar returns. It is refer as

the statistical variations of the random component of the surface height relative to a reference surface (Ulaby,

1981). It is measured in centimeters and determined by textural features comparable to the size of radar

wavelength, such as leaves and twigs of vegetation, and sand, gravel and cobble particles. Therefore, a clear

distinction should be made between surface roughness and topographic relief. Since topographic relief features

such as hills, mountains, valleys and canyons which are measured in meters or even hundreds of meters and are

expressed in the radar image by highlights and shadows.

Figure 18 presents the typical description of the manner in which surfaces affect the signal return to the antenna

and are measured using radar. For smooth surface as in (a) the energy reflected away from the antenna and no

return signal is recorded thereby providing a blank or black image, whereas Snell's law stated that it is where the

angle of reflection is equal and opposite to the angle of incidence (Sabins, 1987). With an increase in surface

roughness, (b) and (c), the amount of energy reflected is reduced and there is an increase in the amount of signal

returned to the antenna known as backscattered component. The greater the amount of energy returned, the

brighter the signal is shown on the image. Radar imagery is therefore a measure of backscatter component, and

it is related to target roughness, radar is mainly a representation of surface roughness.


Instalclick

Instalclick

Instalclick

Instalclick

Instalclick

Instalclick

Figure 18. Reflection of microwaves from the surface (after Sabins, 1987)

By defining upper and lower values of h for surfaces of intermediate roughness, a modification of Rayleigh

criterion was created. This states that a surface is smooth if;

h < � 25 sin

where h is the vertical relief, is the radar wavelength and is the depression angle. Rough criterion was also

derived which states that a surface is rough if (Sabins, 1987);

h < � 4.4 sin

As discussed earlier, the depression angle effects the and rough criteria. Figure 19 illustrates the radar return

from smooth and rough surface as a function of depression angle.

It should be noted that the aforementioned roughness criteria were established with respect to horizontal plane.

They do not consider large scale slope or topography. Small angular changes at incidence angles smaller than

approximately 20o to 25o, result in relatively large changes in return intensity. Thus, different surfaces could be


Instalclick

Instalclick

Instalclick

Instalclick

Instalclick

Instalclick

Instalclick

Instalclick

Instalclick

Instalclick

Instalclick

Instalclick

Instalclick

Instalclick

Instalclick

Instalclick

Instalclick

discriminated and classified based on their angular signature in a similar way as spectral signature and

polarization can be used for classification. Table 7 shows the typical roughness of different radar system.

Figure 19 Mode of Surface roughness criteria and return intensity for SEASAT images (after Sabins, 1987)

Table 7. Surface roughness for typical radar systems (Sabins, 1987)

Roughness Aircraft Ka-band, cm Aircraft X-band, cm SIR-A L-band, cmcategory ( = 0.86 cm, = 40°) ( = 3 cm, = 40°) ( = 23.5 cm, = 40°) ( = 23.5 cm, = 70°)---------------------------------------------------------------------------------------------------------------------------------------Smooth h < 0.05 h < 0.19 h < 1.46 h < 1.00

Intermediate h = 0.05 to 0.30 h = 0.19 to 1.06 h = 1.46 to 8.35 h = 1.00 to 5.68

Rough h > 0.30 h > 1.06 h > 8.35 h > 5.68

Table 8. Response of different values of vertical relief (h) at different radar wavelengths Radar deperession angle ( ) is 40° (Sabins, 1987)

Roughness Aircraft Ka-band, cm Aircraft X-band, cm SIR-A L-band, cmcategory ( = 0.86 cm, = 40°) ( = 3 cm, = 40°) ( = 23.5 cm, = 40°) ( = 23.5 cm, = 70°)---------------------------------------------------------------------------------------------------------------------------------------0.05 Smooth Smooth Smooth Smooth

0.10 Intermediate Smooth Smooth Smooth 0.50 Rough Intermediate Smooth Smooth

1.50 Rough Rough Intermediate Intermediate

10 Rough Rough Rough Rough


Complex Dielectric Constant

Microwave reflectivity is a function of complex dielectric constant. The complex dielectric constant is a

measure of the electrical properties of surface materials. The dielectric constant ε of a medium consists of a real

part ε', referred to as permittivity, and a complex imaginary part jε", a loss factor, referred to as conductivity

(Trevett, 1986), giving:

ε = ε' - jε"

Both properties permittivity and conductivity, are strongly dependent on the moisture or liquid water content of a

medium. Material with high dielectric constant normally represents a strongly reflective surface. Therefore, the

difference in the radar return amplitude for two surfaces of equal roughness and material composition is an

indication of the difference in the dielectric properties, or in case of soils, their soil moisture content.

Surface Scattering

Surface scattering is directly related to the roughness of the dielectric interface, or discontinuity, at the

(homogenous) surface itself (Collwell, 1983). If the medium being imaged at microwave is homogenous then

reflection of the incident radiation will occur from its surface.

Suppose the surface being imaged is perfectly smooth - in other words it acts like a mirror, then the vertical

incident energy will reflect back to the energy source giving bright image, however, for incidence off nadir

nearly all the energy will reflect away from the radar. The latter is called specular reflection, this appear dark on

radar imagery. Suppose now the surface is not perfectly smooth but is rough. Microwave incident upon a rough

surface are scattered in many directions, therefore relatively more energy is reflected back towards the radar.

This is known as diffuse reflection. Thus, this region on the ground would appear brighter in the radar image

(as in Figure 18).

Volume Scattering

Volume scattering is related to multiple scattering processes within a material, such as the vegetation canopy of a

corn field or a forest. Figure 20 is a simple illustration of how tree cover and grasses act as volume scatterers.


Figure 20. Examples of volume scatteres (after Trevett, 1986)

The cover may be all trees, as in forested area, which may be of different species with the variation in leaf form

and size, or grasses and bushes with variations in form, stalk size, leaf and angle, fruiting and a variable soil

surface. Some of the energy will be backscattered from the vegetated surface, but some, depending on the

characteristics of radar system used and the target material, will penetrate the target and be backscattered from

surfaces within the vegetation. Volume scattering is therefore dependent upon inhomogeneous nature of the

target surface and the physical properties of the target, such as, the leaf size, direction, density, height, presence

of lower vegetation, etc., together with the characteristics of the radar used such as wavelength and related

effective penetration depth.

Point Targets

A point target is a discrete target with a simple configuration which gives a clear radar return, whose strength is

disproportionate to its size (Trevett, 1986). One of common point target is the dihedral corner reflector - a point

target situation results from two plane surfaces interesting at 900 and lying orthogonal to the radar incident beam

(Figure 21). Common forms of dihedral configurations are man-made features, such as transmission towers,

railroad tracks, or the side of buildings and adjacent smooth and level ground surface. Another type of point

target is trihedral corner reflector, this is formed by the intersection of three mutually perpendicular plane

surfaces, such reflections are even stronger.

Point targets of the corner reflector types are commonly used to identify known fixed points in an area by radar

in order to perform precise calibration measurements. Such targets can however, occur naturally and are best

seen in urban areas where buildings can act as trihedral or dihedral corner reflectors giving rise to intense bright

spots or an image so typical of urban areas.

Point targets are examples of the objects below the size of the resolution cell of the radar used that can dominate

the return from that cell to make a clearly identified point. It can even give a strength of signal that will extend

into surrounding cells.


Figure 21. Typical constructed point targets (after Trevett, 1986)

Bragg Resonance Effects

If a microwave signal is incident to a target or area that contains periods surface features comparable to the size

of the wavelength of the radar, resonance effects may occur under certain conditions. Just as sea surface is a

special form of rough surface, it is for example in constant motion but may yield a uniform pattern related to

climatic conditions, primarily wind strength and direction. Certain crop lands, notably cereals, can exhibit

similar wind induced effects. These resonance effects from backscatter of regular wave pattern can cause

abnormally high radar returns. The effect is known as Bragg resonance or Bragg scattering (Figure 22).

Figure 22. Bragg resonance (after Trevett, 1986)

The Speckle

The interpretation of radar imagery relies heavily upon the discrimination and identification of units by its

roughness or speckle factor. The radar image is composed of succession of resolution cells and since the cell

will comprise many scatterers at random ranges from the sensor, the reflected or backscattered signals combine

randomly giving rise to a complex interference phenomenon. The apparent reflectivity of a resolution cell will

vary between the accumulation of all the amplitudes and zero. The return signal strength in a radar imaging

system is subject to fluctuations observed and averaged over a resolution cell or pixel which produce the speckle

in an image. For a homogeneous target area some signals may return at full strength to give bright pixel while

fluctuations may lead to less return strength in adjacent cells to result in a darker pixel. Excessive speckle can


result in an uninterpretable image and there are number of ways in which speckle can be reduced in the

processing of the image, such as using low-pass filter.

Resolution in Radar System

The definition of radar resolution is best related to the minimum distance apart of the two radar reflectors need to

have to produce two separate signals on the image. Hence, the term is not related to the actual size of the target.

In radar, there is a comparable resolution cell, this is an area over which the signal is averaged and recorded.

Targets smaller than this cell cannot be discriminated, one small target of unique strength can dominate the

response such that the signal return virtually complete from the target. In optical systems bright light reflectors

are often used as targets (usually mirrors); in radar, corner reflector targets have similar effect and can appear as

bright spots on the image, although the targets used may be smaller than the resolution cell. Corner reflector is

the cavity formed by two or three smooth planar surfaces intersecting at right angles. Electromagnetic waves

entering a corner reflector are reflected directly back toward the sensor.

Resolution and scale are nevertheless closely related. In optical terms the smaller the scale the more targets have

to be aggregated into a cell and thereby the resolution is reduced. In radar much depends upon the ability of a

system to record the vast amount of signal data being returned from an imaged area. As a result, a relationship

between the resolution cell and the data that can be recorded, which in turn will determine the effective image

scales that can be used for interpretation.

There is further determinant in resolution or the ability to discriminate targets, and that concerns the recording of

the tonal values. Two targets maybe on the correct distance apart to provide separate signals but if the strength

of those return signals and those of the background are the same, then the targets will not be distinguishable.

Gray level discrimination is equally important in a radar system, this can be a function of the instrumentation's

ability to measure the strength of the returned signals with sufficient variation to contrast a well segmented gray

scale, and of the size of the resolution cell which is averaging responses to produce an average tone (Trevett

1986). Thus, a system may be capable of very fine individual level discrimination but the resolution cell is so

large that the averaging negates the value of this fine recording.

There is therefore a balance between the actual dimensions of the resolution cell and its gray level recording. As

radar imagery is essentially a representation of surface texture, there is a further factor involved in the grey level

resolution, it will be affected by the texture element size of the terrain.

Further studies of using SLAR/SAR within the Forest Stand

Intensity of the radar return or backscatter within the forest stand, for both aircraft and satellite systems, is

determined by the following properties:

a. Radar system parameters which include the wavelength, depression angle and/or incidence angle and

polarization.


b. Forest stand characteristics which comprised the dielectric properties, surface roughness and forest

structure or orientation.

The following sections discuss in detail the above mentioned properties.

For better understanding of electromagnetic wave interaction with a forest scene, Ulaby et., al. (1981) called for

formulation of all possible radar backscattering components, as follows;

δ°(Θ) = C(H,B,Θ,φ)δ°C + [1-C(H,B,Θ,φ)](δ°S+δ°CS+δ°ST)

where:

δ°(Θ) is the radar return at incidence angle Θ

δ°C is the canopy backscatter

δ°S is the soil backscatter

δ°CS is the soil canopy, canopy-soil reflection

δ°ST is the soil-trunk reflection (corner reflection)

C(H,B,Θ,φ) is the spatial parameter 0<C<1

Θ is the radar incidence angle

φ is the radar azimuth

The spatial parameter (or spatially heterogeneous distribution function) is a function of tree height (H) and basal

area (B), and the effect of spatial heterogeneity depends on radar incidence angle (Θ) and radar azimuth (φ).

Because of complexities of the forest structure various modelling works (Hoekman, 1990; Mougin et. al., 1990;

Lang, 1990; Brown et. al (1990) were done to understand the relationship of radar backscatter and forest stand

characteristics.

Figure 23 Backscattering within the forest stands (after Ulaby et., al, 1981)


Radar Backsctatter from Forest Stand as a function of Wavelength

Sieber (1985) compared the X and L band characteristics of coniferous and deciduous trees. He found that L-

band radar results in higher cross-section on coniferous forest as compared with mixed deciduous forest which

has significantly lower value. This is because at long L-band wavelength the deciduous leaves are not the

significant scattering centers. Moreover, the waves penetrate the leaves independently of the polarization and

are scattered by branches or twigs, while the needles of the conifers act as scattering centers by building a

dielectric 'coating' around the branch axis or at its tip. On the other hand, X-band showed the highest radar

cross-section on deciduous forest areas, while coniferous forests have mean backscattering value similar to

agricultural fields. This is so because X-band waves are reflected by the leaves of the deciduous trees' crown.

INSCAT (Invetigation's Navigation Scaterrometer) system which was used in this study to demonstrate that

height variations and radar shadows are less apparent in L-band data over deciduous forest than they are in the

equivalent X-band images. Furthermore, Hoekman (1987) reported that X-band contribution of scatterers of the

understorey and forest floor is primarily determined by canopy architecture (e. g. factors like crown closure,

crown shapes, etc.) and angle of measurement.

With C-band according to Wu (1986) the radar backscatter coefficients of deciduous forest are higher than those

of the coniferous forests. A similar backscatter behaviour of the species on both X-band and C-band was

expected by Sanden (1991) since in both wavebands, foliage volume scattering is dominant. Volume scattering

is further explained in the coming section.

It was concluded by Sieber (1985) that L-band measurements are very useful not only for identifying land use

classes but also differentiating between class members, especially with forest, e.g. deciduous and coniferous

trees. Moreover X-band radar can provide information about the tree height, tree-stand density and orientation.

Hoekman (1990) after studying the radar's potential use in classifying vegetation concluded that because of the

presumably strong relationship between C-band radar backscatter level and morphologic (species dependent)

properties, it is the most appropriate for species classification. While L-band, itself or in addition to the other

bands, appeared to be useful for differentiation of some broad vegetation classes, such as the

coniferous/deciduous, forest/non-forest and flooded/non-flooded forest class

discrimination. Furthermore, Hoekman (1990) cited various applications for radar of shorter wavelengths (C-

and X-bands), namely;

a. forest inventory, i.e. delineation of stands, area determination and forest species/type classification,

b. monitoring of phenologic development and physiologic processes,

c. detection of disease or stress conditions, and

d. monitoring of forest development.

Radar Backscatter from Forest Stand as a function of Polarization

The intensity of the different polarization responses varies with surface roughness, moisture content, or

vegetative cover (Ford and Wickland, 1985). It is influenced by the structure of the forest vegetation, including


tree heights, diameters, spacing, orientation, and geometry (Sader, 1987). With L-band, radar sensors act as

"filters" in detecting the dielectric geometry of the natural targets, orientation of electrical-field vector provides

an "analysis" of tree structure, which itself might be an indicator of tree type (Sieber, 1985).

Hoffer et., al (1985) using X-band found out that hardwoods had every high backscatter (and therefore light

tone) on HH polarization as compared to coniferous forest. With HV polarized data, the deciduous and

coniferous forest could not be differentiated. Conifer stands and pasture areas have very similar results in tone

in both HH and HV and could only be differentiated on the basis of their texture or speckle characteristics. In

general, it appeared that the like-polarized data (HH) had a wider range of tonal contrast and allowed better

discrimination than the cross-polarized data.

At L-band radar images deciduous trees showed higher response in VV polarization than HH polarization

(Sieber (1985), Sader (1987)), since the orientation and structure of the branches influences the backscattering

processes. For pine dominant sites, the HH and HV polarization response was higher than VV, hence, HV

response was higher than both HH and VV (Sader, 1987). This higher response of HV in pine dominant sites

may indicate the effect of higher biomass, height, and dbh and non homogenous structure resulting from addition

of substantial hardwood component in the pine-hardwood sites. This finding conforms with a study by Hussin

(1990), where he showed that there is a strong relationship between HV-polarized radar backscatter and several

forest stand parameters of pine (i. e. age, DBH, basal area, height, stand volume, number of trees per acre, and

stand biomass). Sader (1987) also showed that for areas of small diameter, low height and biomass, pine and

soft-hardwood species on very poorly drained soils showed high DN value in all three (HH,VV,HV)

polarizations. This relatively high HH and VV response may have been influenced by high soil moisture

content, while the higher HV suggested strong backscatter from the standing biomass content.

Sader (1987) in his conclusion cited four factors believed to be important in influencing SAR polarization

response using L-band synthetic aperture radar. They are as follows:

a. difference in canopy characteristics (branching patterns, crown weight, and areas between the hardwood

and other forest composition groups),

b. variability in species composition and stand structure (heights, stem diameters, and density) within and

between major forest composition groups,

c. the distribution and amount of exposed wet soils and ground litter in some of the lower biomass and

partially open forest sites, and

d. angular dependencies of SAR data collected from approximately 15 to 60 degree off nadir.

The results of the study by Knowlton and Hoffer (1981) suggested the usefulness of dual-polarized (HH and

HV) X-band airborne SAR system for forest cover mapping. Deciduous and coniferous forest cover were easily

separated in HH image due to distinctive tonal differences, but very difficult to separate on the HV image.

Deciduous forest cover has a distinctive light tone in HH image but it has darker tone in HV image. But

opposite condition was observed in coniferous forest, since it showed dark tone in HH image and lighter tone in

HV image. Moreover, dense deciduous forest stands located in ravines are easily identified on both polarization


because of topographical pattern being highlighted by the response of deciduous stands and partially highlighted

by the slopes acting as angular reflectors. These patterns are more distinct on the HH image than HV.

Mougin et., al. (1990) studied the attenuation measurements performed on cylindrical-shaped forest components.

As indicated in the polarization effects on coniferous trees, the angular behaviour or the loss factor strongly

depends upon the needle orientation on the branch. Species with relatively horizontal needle orientation, e.g.

Douglas fir (twig-needle angle +/- 70°) reveal higher attenuation HH waves. Also, the attenuation properties for

HH and VV polarization are highly dependent on the twig diameter, i. e. twigs with small diameter showed

higher attenuation in VV waves than HH, the attenuation decreases with the increasing twig diameter.

Radar Backscatter from Forest Stand as a function of Incidence Angle

It is a general assumption that the attenuation of microwave energy will increase with and increasing incidence

angle. This is based on the fact that microwaves, at higher incidence angles, will have to travel longer distances

to reach the surface target, consequently will lose the energy. Simonett and Davis (1983) stated that the

magnitude of the normalized backscattering echo generally increases with decreasing incidence angle, but as a

surface becomes very rough it becomes independent of the incidence angle. As the incidence angle increases,

like-polarized backscatter from a surface shows increasing dependence on the complex dielectric constant, and

decreasing dependence on the surface roughness. Moreover incidence angle effects on SAR data may have been

masked to a certain extent by the variation in backscatter from differences in species composition and structure.

Mueller et., al.(1987) investigated the capacities of SIR-B data (L-band, HH) for identification of coniferous and

deciduous forest in Florida using 28°, 45° and 58° incidence angle. They found out that an image taken at 28°

incidence angle provided the maximum amount of differentiation between the general classes of forest cover

types. Same imagery can be used to detect standing water beneath forest vegetation. On the same degree of

incidence angle Lee and Hoffer (1990) concluded that forest types, ground conditions, and relative amount of

understorey could be best defined.

With multi-polarized L-band radar, Sader (1987) found that the best results for correlating hardwood biomass

were achieved between 34-41° incidence angle of HV polarization. Cimino et. al. (1990) tested a relative image

brightness of multiple incidence angle SIR-B L-band radar data to classify vegetation types in the Cordon la

Grasa region of Argentina. With this more recent study, forest tree species discrimination was possible using

data acquired at incidence angles between 33° and 54°, but they are indistinguishable at 59°. This image

brightness was probably related to differences in canopy structure, canopy closure, and ground cover. Better

overall classification result of forest cover was obtained by Hussin (1990) also from SIR-B L-band radar data at

incidence angles from 35-45° as compared to 25-35°, and lowest classification performance obtained at 45-55°

incidence angle.


Radar Backscatter from Forest Stand as a function of Forest Structure

Le Toan et. al. (1980) showed that HH polarized L-band measurements over pine-tree areas have resulted in a

good seprabability from other land-use classes, whereas the backscattered represented by the mean-gray level in

the L-band imagery of pine forest areas depends mainly on the age of the trees. The intensity of the

backscattered signals increases with increasing age.

Christensen (1990) concluded that microwave backscatter at L-band is sensitive to changes in stand structure and

biomass associated by secondary forest succession in loblolly pine stands. Radar backscatter may level off in

more mature stands as a consequence of changes in stand structure which counterbalance each other, for

example, while average pine tree biomass increases throughout the life of the stand, pine density is constantly

decreasing.

A modelling work was performed by Lang and Chauhan (1990) using P-, L- and C-SAR frequencies to test radar

sensitivity to forest and stand parameters. The sensitivity itself as explained is a function of frequency, type of

forest, season, weather conditions, etc. The modelling suggests that medium to large incident angle C-band

radar returns are independent of the ground conditions. In the P- and L-band the tree trunks combined with the

ground surface produce a strong direct-reflected backscatter.

Lee and Hoffer (1990) studied the SIR-B backscatter relationships with forest stand characteristics in Florida.

They concluded that the differences in radar backscattering among forest cover types are apparently caused by

the combined effects of tree morphology, tree size and stand density, and ground conditions. Stand parameters

describing the physical characteristics of the forest (Tree height, DBH, basal area and biomass) were positively

correlated with all the three angles (28°, 45° and 58°) of SIR-B radar backscatter. Tree height showed the

highest correlation, while stand density has negative correlation. This correlation between the stand parameter

and SIR-B backscatter was found more distinct in an early age group of pine stands (1-9 years). But the radar

backscatter was not strongly influenced by the stand parameters when the crown are closed. Hussin (1990)

showed that with use of regression models the forest stand parameters such as number of trees per acre, DBH,

basal area, volume, age and biomass, can be estimated with reasonable accuracy using L-band HV polarized

radar data.

The Effect of Terrain Characteristics on Radar Backscatter from the Forest Stand

A slope facing the radar will receive all the energy which would usually be spread over the ground projected

area. On the back slopes the grazing angle diminishes so the backscatter is reduced, therefore slopes facing

away from the radar show a weaker return. The overall effect is that in hilly place slopes facing the radar will be

brighter than normal and reverse slopes darker, the effect will increase with the severity of the terrain.

The Side Looking Radar was successfully used in lieu of optical photography for reconnaissance of the Darien

Province of Panama and parts of Northwest Colombia (Viskne et. al., 1970). They used AN/APQ-97 side

looking radar with K-band. Since K-band does not penetrate vegetative cover, it made it possible to evaluate the

various vegetation forms on the basis of their reflectance characteristics. However, in the high relief zones of the


subject area, the varying tones of grey that reflect changes in vegetation were sometimes obscured by the

"shadow effect", while the steep surfaces directly exposed to the sensor show up much brighter than the

surrounding area. An additional difficulty encountered in delineating vegetation strata in the high-relief areas

was due to the heterogenous nature of the vegetation types found. Distinct boundaries between Evergreen Rain

Forest and the Mixed Semi-deciduous and Evergreen Forest could not be discerned on the imagery since these

two zones blend into each other gradually over a distance of from several hundred meters to several kilometers.

A solution was made by averaging the transitional zone for boundary placement. Some larger-scale aerial

photographs and local spot photography aided in delineation of these difficult areas.

A research was conducted to determine the usefulness of like-polarized with Ka-band AN/APQ-97 SLAR

imagery for evaluating wildland vegetation resources in northern Sierra Nevada Mountains of California (Daus

and Lauer, 1971). Tone and texture were the useful characteristics of the SLAR imagery in analyzing the

wildland vegetation resources, whereas vegetation was the main factor governing the texture of the imagery and

the combination of slope and aspect was the main factor governing tone. In areas which were nearly flat and

level, timber stands consistently were differentiated from everything else on the SLAR imagery because of

relatively coarse texture, however, slight differences in topographic relief or changes in slant range often caused

two nearly identical timber stands to appear quite different on the imagery.

Moreover, Sicco Smit (1975) with RADAM images in Colombia studied the use of radar images for forestry

applications. From the terrain physiography, the occurrence of specific vegetation classes were deduced. With

the use of additional information from aerial photographs, ground truth and knowledge of existing vegetation

characteristics, a significant improvement in the delineation and differentiation of vegetation types were made.

2.4 Videographic Sensor System

“Videography” is defined as: “ The science dealing with organic, electronic, or mechanical recording andplayback of information “ or “ the technology, process and art of producing information in analogue ordigital form”.

2.4.1 Aerial Videography

Aerial videography is the use of a video recorder in conjunction with an aeroplane or helicopter. Aerialvideography is basically just videography in the air. It has many uses and is easy to do. Just recentlyscientists have realised the full potential of this marvellous invention when it is used in conjunction withGPS locators, computers, and data maps.Videography offers a number of distinct advantages over photographic remote sensing as explained asbelow:

1. Immediate imagery availability: For practical applications this is the first and most advantage. Thispermits rapid turnaround for applications where the value of the information is short-lived (e.g.agriculture), or where delays in scheduling have significant time penalties (e.g., disaster operations likefires, flooding, storm damages).

2 Live video images: The operator can view ‘live’ imagery on a monitor in the aircraft concurrent withacquisition, making the video acquisition process less subject to error that of photography. The imagedisplay allows camera exposure setting to be adjusted interactively.

3 Notation of pertinent information: The audio tack permits notation of pertinent information, which maybe of use in future, analysis, directly onto the video tape while in flight.


4 Ability of using in GIS: A straightforward pathway is provided from the videotape to digital format forcomputerised image processing and incorporation into a geographic information system (GIS).

5 Low cost: The price of video camera and videotape is much lower than the cost of film and processing.This can result in significant savings if large areas are to be covered.

6 Higher sensitivity: Video cameras have higher light sensitivity than film cameras allowing usableimagery to be acquired under less than ideal weather conditions.

7 Easy to use since video cameras are small and light, they are easy to mount on aerial platform. Unlikelarge-format aerial survey cameras, which have to be mounted on specially adapted aircraft, videocameras can be easily bolted to any light plane or the undercarriage of a helicopter.

8 Flexibility in weather conditions: While all aerial surveys require good weather conditions with highvisibility and low wind speed, and are therefore best carried out in the summer when, in any case, thesun’s angle is higher in the sky. However, a low sun angle can reveal additional details in the terrain, andthe winter atmosphere can give great clarity. Videography is flexible enough to give good results in theseconditions.

9 Ability to collect spectral data in very narrow bands (5 to 12nm) in the visible to near infrared (NIR).10 Ability to collect data in the mid-infrared (1.35 to 2.50µm) water absorption regions.11 Data redundancy in which images are acquired every 1/25 second producing multiple views of a target or

scene.12 Wide spectral range of modern silicon detectors13 Linking of video data with ground control points.

Aerial Videography System Components

In order to understand aerial videography systems (single, multispectral), it is necessary to look at itscomponents, which are explained below: Figure 24 shows the components of a typical aerial video system.For each of these components, a variety of alternative equipment exists.

Figure 24. Typical aerial video equipment configuration (Meisner, 1986)

The Video Cameras

Conventional Video Cameras

The three varieties of conventional video cameras are illustrated schematically in Figure 2. The simplestvariety is the black and white, single tube model. The camera lens forms an image on the faceplate of thevideo tube, which is electronically scanned to generate a video signal. The two variables in black and whitecameras are the tube size and tube type.

The size parameter refers to the diameter of the video tube, which in turn defines the size of the image areaon the faceplate. The tube size is roughly analogues to the film size in photographic cameras, although due tothe fixed nature of the video scanning process constant number of scan lines per frame), a larger video tubewill not necessarily provide a proportional increase in spatial resolution.


A variety of tube types are available for video cameras, each with particular sensitivity and spectral attributesbut they are not interchangeable like film types in a photographic system. Generally filters are used tomodify the spectral sensitivity of the camera. Therefore, black &white video cameras used in remote sensingare usually equipped with visible /near-infrared sensitive tubes to allow sensing in a wide variety ofwavelengths.

Colour video cameras come in two varieties: single-tube models use a striped filter applied directly to thefaceplate to provide the primary colours; three tube models use optical beam separation to form images ineach primary colour. Most colour video remote sensing has been done using professional grade- three tubecameras. These cameras are fairly compacting built to take abuse, and provide very high quality video. Solidstate detectors are alternative to video tubes which are becoming increasingly used, These sensors consist ofan array of photodetectors and readout electronics etched onto a silicon chip using integrated circuittechnology. A number of acronyms are used for these sensors, referring to the microcircuit technology used:CCD (charge couple device); CID (charge injection device); and MOS(metal on silicon). The density of thedetector arrays has steadily increased to current levels of about 400X500 detectors. Solid cameras offeradvantages of compactness, shock resistance, low power consumption, and perfect geometry. The silicondetectors used in solid state cameras have a spectral response of0.4-1µm, making them useful for remotesensing applications.

Figure 25. Basic varieties of conventional video cameras (Meisner, 1986)

Multispectral Video

In addition to the standard video a number of cameras have been developed specifically for aerialapplications. These systems are designed to generate multispectral imagery using video technology. Theycan be thought of as video equivalents to multispectral photography, with a major difference being that acolour additive viewer. Multispectral video cameras can also be thought of as low cost multispectralscanners.

Three fundamentally different approaches have been used in the design of the multispectral video systems(Figure 26).

The simplest approach uses an array of standard black & white cameras. The cameras are electronicallysynchronised and are carefully mounted to provide equivalent fields of view (generally at a focus distance ofinfinity). Different filters are placed over the camera lenses to provide multispectral images. The principals


advantage of this type of system its simplicity. Disadvantages are the inability to image at close range due toparallax, problems in using zoom lenses since the three lenses must be set the same time, and the bulkinessof the system.

A second approach uses a single sensor and a spinning filter wheel. The filter wheel rotation is synchronisedwith the video scan, such that sequential series of spectral images is generated. This approach requires nogeometric alignment, since only a single lens and sensor are used. Similarly, close range focusing and use ofzoom lenses present no problems. However, the time delay between spectral image exposure is a problem inaerial video, since image motion will cause spectral misalignment, which must be removed prior to analysis.Also, the nonsimultaneous nature of the spectral imagery means that a colour composite cannot be generatedwithout using some form image memory. Accordingly, this type of camera is best used in conjunction withan image processing system, since the system can be used to realign the spectral images and to store them toallow display of a colour composite image.

The third type of multispectral camera uses an internal optical beam-splitter to form multispectral imagesusing a single camera lens. This unit contains three sensors, filtered for green, red and near infrared portionsof the spectrum. This design approach combines the single-lens advantages of the filter wheel camera withthe multiple sensor advantages of the camera array. This is the only system of the three, which provides animmediate display of false colour imagery, greatly facilitating acquisition and fieldwork, and simplifyinglaboratory analysis. Disadvantages of this approach are the limitation to three spectral channels and difficultyin changing spectral bands.

Figure 26. Three design approaches to multispectral video cameras (Meisner, 1986)

Video Recorders

Although a wide variety of equivalents are available for this component for remote sensing applications,machines should be selected for their “special effects” features, referring to fast and slow motion and stillframe capabilities. Of these, the still frame (or freeze frame) is the most important, since a single frame mustbe displayed during interpretation. The still frame must be free from noise bars and hold steady on thescreen. Generally the video cassette recorder specification do not make it clear whether a still frame consistsof just one field or a true two-field frame. A still field is preferable when image motion is present; a stillframe is better if there is still image motion.


Video Monitor

This component displays the video image on a CRT screen.

Camera Mount

The mount was designed to hold the camera in a vertical position over an aircraft camera hole; and handlecorrections for tip and tilt, and drift rotation.

Video Signals

While not a hardware component of the aerial videography system, it is important to understand some basiccharacteristics of the video signal. The video image is made up a horizontal scan lines, scanned left to rightand top to bottom. The scanning pattern is referred to as a raster. Within the scan, the voltage level of thesignal is varied in proportion to the image brightness.

Power

The aerial videography system requires 12 volts DC to operate. Most aircraft have 24-28vDC electricalpower and require the use of a 24 vDC to 12-vDC-power converter.

Caption Generator

The caption generator links the video camera, recorder, and an aircraft navigation system (LORAN) orglobal positioning system (GPS) unit. The caption generator allows the user to overlay the date, time, latitudeand longitude co-ordinates, and true heading of the aircraft onto the video frames when using LORAN.Altitude and GPS time may also be displayed when using GPS. The overlayed information helps the userlocate and interpret the video imagery after the flight.

Interpretation Equipment

In addition to the portable equipment used during image acquisition and field analysis, some additionalhardware can be useful when interpreting the data in the laboratory. High quality, non portable videocassetterecorders can provide a steadier image than compact portable units, and may be worth obtaining forinterpretation use.A digital freeze frame unit that also called a frame grabber provides the ultimate still frame. Such deviceconverts a frame of imagery to digital data, stores it in computer memory, and regenerates a video imagefrom the stored data. A number of computer based systems are designed specifically for takingmeasurements from video images. Most of these “image analysers “ are designed microscopy or industrialquality control applications. Some microcomputer-based image processing systems have video inputavailable. These systems allow video imagery to be analysed with Landsat-type multispectral imageprocessing software.

Acquisition of Aerial Video

While acquisition of aerial video is similar to working with small format photography, a number ofsignificant differences exist. These differences will be discussed under two general headings; missionplanning and camera mounting considerations.


Mission Planning

Flight planning for aerial photography frequently begins with the choice of photographic scale, since userswill often know what scale will be suitable for a given application. But in video scale consideration arecomplicated first by the lack of a physical image on which to base the scale computation. The dimensions ofthe video format cannot be directly measured, and are only approximately known (solid state sensor arraysare an exception to this). Furthermore, while photographic interpretation is frequently performed on theoriginal film product, at “constant scale”, video interpretation is done at the much larger scale of the displaymonitor. Finally the scale figure is less consistent than in photographic systems, since the resolution is lessdirectly linked to the image format size. Each of these factors makes the scale parameter less meaningful invideo images.

In effect, video has more in common with other electronic sensors such as multispectral scanners. In thesesystems, the scale of the hardcopy or display is not important unless it is the limiting factor on resolution. Ofgrater importance is the size of the image pixels on the ground. The pixel size is an inexact value in videodata due to analogue storage of the signal, but can be expressed as the number of resolvable elements acrossthe scan line.When deciding the scale of video imagery to be acquired, it is important to also consider the correspondingwidth of coverage. This is the difficult side effect of the low resolution of video. If a large scale or high-resolution image is required, the resulting width of coverage will be narrow. While this presents no problemwhen a narrow linear feature is to be covered or a transect sample is needed, it can be a severe problem whenfull coverage of an area is required.

In addition to the expense of additional flight lines, narrow coverage can cause great problems in orientationduring interpretation. The problem is worse than in the case of small format photography since the use of adisplay screen makes it impossible to construct a mosaic on which a larger number of landmarks will appear.This is an important point to keep in mind when planning a video flight: the width of coverage should bemade large enough to insure that recognisable features would appear each frame. If large format photographyor an orthophoto base is available for the study area, this will not be severe limitation. In many cases,however, this factor will be the major determinant of image scale will work for the application at hand.

Camera Mount

The blur problem, due to aircraft vibration, can be handled by proper mount design. The use of fairlymassive camera /mount set-up will reduce the amount of vibration. This factor is important when using anextremely compact, lightweight camera.

2.5 LIDAR (LASER) SENSOR SYSTEM

2.5.1 Introduction

The Laser considered as one of the most adaptable tools ever created. laser are competing the computer andcommunication by only replacing the electronics with photo-electronics; Surgeon have been done verydelicate work with the laser in the technique of what is called microsurgery; Environmental iat have beendetermined several case of weather pollution using the laser energy; and finally nowadays scientists areplanning to use the laser from the space battle station to destroy missiles in the program of the space ware.Therefore it is not surprise us when we now that laser has an important role in remote sensing fields.Certainly since the laser has the capability of doing an effective analysis at a distance so it can add a newdimension to remote sensing. Moreover the big range of applications that laser dealing with are making thisnew dimension revolutionise remote sensing. Best example to show how this tool is doing a very good jobwhen it estimate forest tree height using the airborne laser sensor.


2.5.2 Laser Principles

The word "laser" stand for Light amplification by stimulated emission of radiation.It is the device for producing light by emission of energy stored in a molecular or atomic system whenstimulated by an input signal. It used the energy of the electromagnetic spectrum EMS in three differentareas Ultraviolet, Visible Light, Near Infrared and Far Infrared. Table 9 show the frequencies, wave lengthsand the amount of energy used.

Table 9. Show laser wavelength, frequencies, and energy. (Adapted from Johnson, 1970).

Spectral region Wavelength Frequencies EnergyMicron Hz Joules

Ultraviolet 0.3-0.4 10 6.6x10Visible light 0.5-0.7 6x10 4 x 10Infrared NIR 0.8-1.1 3x10 2 x 10Infrared FIR 10 3x10 2 x 10

2.5.3 Laser energy creation

Laser basically created using a device such as the one shown in Figure 27. In this device the light pumpedinto suitable medium within a plan-mirror resonator. Laser action arises as soon as the population inversion(e.g. the level of energy that the light reach by which the laser created) is sufficient to overcome the opticallosses of the system. So the energyof the light is stimulated by the mirrors which reflects it continuouslyuntil the creation of the laser.

Figure 27. Schematic of flashlamp-pumped laser (after Measures 1984).


2.5.4 Laser in remote sensing

Laser sensor system used for the remote sensing purposes is called "Lidar" which is stand for Light detectionand ranging. It is radar like system. It is active remote sensing system, Day and Night operation (For daytime precaution must be taken to eliminate sunlight from the returns. Sharp filter can be used for thispurpose) (Mulders 1987). Lidar used the laser energy in two types pulses or continuous waves. Thefunctional elements and manner of operation of most remote sensing Lidar systems are schematicallyillustrated in Figure 28. The pulses or the waves of the laser energy in the lidar system are emitted by a laserand directed through some appropriate output toward the target of interest. "The function of the output devicecan be threefold-to improve the beam collimation, provide spatial filtering, and block the transmission of anyunwanted broadband radiation, including the emission and the arises from some laser" (Measures 1984). Thescattered energy collected by the receiver optics is passed through some form of spectrum analyzer on itsway to the photo detector. "The spectrum analyzer serves to select the observation wavelength interval andthereby discriminate against background radiation at other wavelength. The choice of photo detection systemis often dictated by the spectral region of interest, which in turn is determined by the kind of application andthe type of laser employed" (Measures 1984).

Lidar Way of operation : can be operated in profiling or manning mode (Lillesand & Kiefer 1987). Goodexample of the profiling system is shown in Figure 28, and typical example of the scanning lidar system isshown in Figure 29.

Types of lidar

Three different type of aircraft lidar system: (Mulders 1987).

1. Laser profiler or altimeter ( i.e. Bathymeter ): In which the laser emits either a pulse or acontinuous wave of light that directed straight down to the earth below the aircraft; in this way a profile ofthe terrain can be obtained.

2. Raman and Fluorescence Lidar: In which the radiation from the laser is used to excite thematerials of interest; the results of this system are recorded and unique for each type of material.

3. Reconnaissance or Mapping Lidar: In which the laser beam swep across a strip of terraineither directly beneath, or to the side of the aircraft and perpendicular to the flight line, the scattered energyis recorded in a way very similar to the scanning system.Because the first and the second Lidar system are the most popular in a lot of applications, some emphasisewill be given to them.

1. Bathymeter

The principle of the operation and measurement of this system basically is by sending the pulses or wave tothe target from the laser output unit which hit the object and reflected or returns as an echo which detectedby the receiver. The time interval between the initial pulse or wave and it's return is measured and convertedto a distance. Light velocity is known and the time is measured, therefore :

Distance = Velocity x Time(e.g. 1 ns time = 30 cm distance)

Typical example can express the above idea of the measurement is shown in Figure 30, (Aldred and Bonnor1985).


2. Fluorescence

It is the sensor in that the process in which the materials that capable of converting absorbed radiant energyat one wavelength to emitted radiation of a longer wavelength without converting first the absorbed intothermal energy (Suits 1983). The reemission usually happen shortly after excitation by the original energysource because the atom or molecular can no longer absorb radiation, therefore it is going to contributeenergy back to the radiation field either randomly or coherently through stimulated emission (Measures1984). Absorbing radiant energy causes electron orbiting in low-energy orbits of an atom or molecule tochange to higher orbits and it is often called laser-induced fluorescence (LIF). It is used to measure thecharacteristics of the materials because different materials have different ability of absorption and reemissionthe radiant energy of the laser.

3. Laser Beam

Laser beam produced in most of the remote sensing laser lidar system in a very narrow coherent beam (i.e.0.1 mRad).

2.6 Literature Cited

Aerial Videography &Land-Cover Mapping, http://www.snr.uvm.edu/airvideo.html

Aldred, A. H. and G. M. Bonnor.1985. "Application of Airborne Laser to Forest Surveys". InformationReport PI-X-51. Petawawa National Forestry Institute. Canadian Forest Survey. 62 pp.

Aldrich, R.C. 1979. Sampling for multiresource information--can aerial photographs help and at what costs?Forest Resource Inventories, Workshop Proceedings, Vol.I. Colorado State Univ.Fort Collins. July 23-26:393-405.

Anonymous 1, Videography, http://globike.com/vidgrafyO.html

Anonymous 2, Using Videography to Evaluate Impacts to Ecological Resources,http://www.ead.anl.gov/~web/newead/prgprj/proj/video/video.html

Brown, L. M., Veck, N. J. and Miller, R J. (1990). Preliminary Results from a High Resolution Forest ImagingModel For SAR. IGARSS '90, Vol. II, College Park, Maryland, p. 1199-1201.

Cimino, J., Brandani, A., Casey, D, Rabassa, J. and Wall, S. D. (1986). Multiple Incidence Angle SIR-BExperiment Over Argentina: Mapping of forest Units. IEEE Transactions on Geoscience and Remote Sensing,GE-24(4):498-509.

Christensen, N. L. Jr., Kasischke, E. S. and Dobson, M. C. (1990). SAR-Derived Estimates of AbovegroundBiomass in Forested Landscapes. IGARSS '90, Vol. II, College Park, Maryland, pp. 1209-1212.

Colwell, R. N. (1983). Manual of Remote Sensing, Vols. I and II. American Society of Photogrammetry, FallsChurch, Vancouver.

Daus, S. J. and Lauer, D. T. (1971). Testing the Usefulness of Side Looking Airborne for Evaluating ForestVegetation Resources, Final Report. The University of Kansas.

Ford, J. P. and Wickland, D. E. (1985). Forest Descrimination with Multipolarization Imaging Radar. IGARSS'85. Vol 1, pp. 462-465.

Forster, B. C. (1992). Radar Remote Sensing. Paper presented in course developed for BHP Engineering asConsultants to AIDAB on the Philippine Remote Sensing Project. University of the Philippines.


Haralick, R. M., Shanmugan, K. and Distein, I. (1973). Textural features for image classification. IEEETransactions on Systems, Man and Cybernetics. Vol. SMC-3, pp. 610-621.

Hoekman, D. H. (1985). Radar Backscattering of Forest Stands. International Journal of Remote Sensing, Vol 6.pp. 325-343.

Hoekman, D. H. (1990). Radar Remote Sensing Data for Applications in Forestry, Ph.D. Thesis, WagenigenAgricultural University, The Netherlands.

Hoffer, R. M., Davidson, S. E. Mueller, P. W. and Lozano-Garcia, D. F. (1985) A Comparison of X- and L-BandRadar Data for Discriminating Forest Cover Types. Pecora 10 Proceedings, Remote Sensing in Forest andRange Resource Management, Colorado State University, Fort Collins, Colorado, pp.439-440.

Howard, J. A. (1991). Remote Sensing of Forest Resources, Theory and application. Chapman and Hall.

Hussin, Y. A. (1990). The Effects of Polarization and Incidence Angle on Radar Backscatter from Forest Cover.Ph.D. Thesis, Colorado State University.

Inkster, D. R., Lowry, R. T., Thompson, M. D. (1980). Optimum Radar Resolution Studies for Land Use andForestry Applications. Proceedings of the 14th International Symposium of Remote Sensing of Environment,Vol. 2, pp. 865-882.

Johnson,C.M.1970."Laser Radar". From: Skolnik.M.I.1970."Radar Handbook ".McGrow-Hill BookCompany.(37-1)-(37-69) pp.

Kalensky, Z.D., P.G. Reichert and K.D. Singh, 1991. Forest mapping and monitoring in developingcountries based on remote sensing. In Fernerkundung in der Forstwirtschaft (Stand andEntwicklungen). Hrsg. Oesten, Kuntz/Gross:230-258. ISBN 3-87907-233-7. Wichmann Verlag, Karlsruhe.

Knowlton, D. J. and Hoffer, R. M. (1981). Radar Imagery for forest Cover Mapping. Machine Processing ofRemotely Sensed Data Symposium, Vol. 4, pp. 626-632.

Land Cover Classification of the Philipines. (1988). Department of Environment and Natural Resources - AsianDevelopment Bank - Swedish Space Corporation. DENR, Quezon City, Philippines.

Lang, R. H. and Chauhan, N. (1990). Radar Sensitivity to Forest and Stand Parameters. IGARSS '90, Vol. II,College Park, Maryland, p. 481.

Le Toan, T., Shahin, A. and Riom, J. (1980). Application of Digitalized Radar Images to Pine Forest Inventory:First Results. Proceedings of the 14th International Symposium on Remote Sensing of Environment. pp. 493-944.

Lee, K. and Hoffer, R. M. (1990). Shuttle Imaging Radar B (SIR-B) Backscatter Relationships with ForestStand Characteristics. ACSM-ASPRS Convention, Image Processing/Remote Sensing, Technical Papers. pp.270-281.

Lilesand, T. M. and Kiefer, R. W. (1986). Remote Sensing and Image Interpretation. John Wiley and Sons.

Lindenlaub, J.C., 1976. Multispectral scanners. Minicourse Study Guide, Fundamentals of RemoteSensing.

Measures,R.M.1984."Laser Remote Sensing: Fundamentals and Applications". John Wiley & Sons. 510 pp.

Meisner, D., E., 1986, Fundamentals of Airborne Video Remote Sensing, Remote Sensing of Environment,Vol.19, pp. 63-79.


Mueller, P. W., Lozano-Garcia, D. F. and Hoffer, R. M. (1985). Interpretation of Forest Cover on Microwaveand Optical Satellite Imagery. Pecora 10 Proceedings, Remote Sensing in Forest and Range ResourceManagement, Colorado State University, Fort Collins, Colorado, pp.578-592.

Mulders,M.A.1987. "Remote Sensing In Soil Sciences". Elsevier Publishing Company. 314-315 pp.

Mougin, E., Lopes, A. and Le Toan, T. (1990). Microwave Propagation At X-Band In Cylindrical-ShapedForest components: Attenuation Observations. IEEE Transactions on Geoscience and Remote Sensing,28(1):60-69.

Nakayama, Y., Kimura, H. and Mukai, Y. (1988). A Study of the Accuracy of Land Cover Classification fromSAR Images. Asian-Pacific Remote Sensing Journal, Vol. 1, No. 1. pp. 79-86.

Sabins, F. F. (1987). Remote Sensing Principles and Interpretation, Second Ed. W. H. Freeman and Company,New York.

Sader, S. A. (1987). Forest Biomass, Canopy Structure, and Species Composition Relationships withMultipolarization L-Band Synthetic Aperture Radar Data.Photogrammetric Engineering and Remote Sensing,53(2):193-202.

Sanden, J. J. van der. (1991). Microwave Remote Sensing for Forest Monitoring Purposes. IGARSS '91, Vol. 3,pp. 1507-1510.

Sicco Smith, G. (1975). Will the road to the green hell be paved with SLAR?. ITC Journal, Vol 2, pp. 245-266.

Sieber, A. J. (1985). Forest signatures in Imaging and Non-imaging Microwave Scatterometer Data. ESAJournal, Vol. 9, pp. 431-448.

Simonett, D. S. and Davis, R. E. (1983). Image Analysis-Active Microwave. Manual of Remote Sensing,Second Edition, Vol I, American Society of Photogrammetry, pp. 1125-1181.

Stellingwerf D. and Y. A. Hussin. 1997. Measurements and Estimations of Forest Stand Parameters UsingRemote Sensing. VSP Publishinh BV. Utrecht, The Netherlands. 272 pp.

Suits, G. 1983. "The Nature of Electromagnetic Radiation". from: N. R. Colwell. 1983. Manual of RemoteSensing. ASPRS. 37-60 pp.

The Integrated Land and Water Information System, ILWIS 1.4, User's Manual International Institute forAerospace Survey and Earth Sciences, The Netherlands.

Trevett, J. W. (1986). Imaging Radar for Resources Surveys. Chapman and Hall Ltd. 313 pp.

Ulaby, F. T., Moore, R. K. and Fung, A. K. (1981). Microwave Remote Sensing,Vol I and Vol. II. Addison-Wesley, Reading, Mass.

Umali, Ricardo T. (1978). Landsat Assisted Forest Inventory of the Philippine Island. Proceeding of theTwelfth International Symposium on Remote Sensing of Environment, Vol. 2. Ann Arbor, Michigan. pp.1401-1405.

Umali, Ricardo T. (1985). Remote Sensing for Natural Resources Management in the Philippines.Development and Applications of Remote Sensing for Planning, Management and Decision-Making. SeminarProceedings. UNDP/ESCAP Regional Remote Sensing Programme, Beijing. pp. 59-70.

Viskne, A., Liston, T. C. and Sapp, C. (1970). SLR Reconnaissance of Panama. Photogrammetric Engineering,36(3): 253-259.


Wu, S. T. (1986). Preliminary Report on Measurements of Forest Canopies with C-Band Radar Scatterometer atNASA/NSTL. IEEE Transactions on Geoscience and Remote Sensing, 24(6):894-899.

Wu, S. T. and Sader, S. A. (1987). Multipolarization SAR Data for Surface Feature Delineation and ForestVegetation Characterization. IEEE Transactions on Geoscience and Remote Sensing, 25(1):67-76.


3. REMOTE SENSING SENSOR SYSTEMS APPLICATIONS IN FORESTRY

3.1 The Applications of Aerial Photography in Forestry

The applications of the Aerial Photography in Forestry have exist some decades ago. Moreover, it has been

used as an operational system for forest inventory in most parts of the world. For this reason we are not

going to elaborate on these applications. Therefore, we will summarize these applications as follow:

3.1.1 Mapping (qualitative): with reasonable accuracy 75% and more

• Forest cover types• Identify individual species• Species composition• Forest fire detection• Forest fire hazard• Detecting forest trees health (vigor and stress)• Forest trees diseases and insects infestation• Forest trees under air, soil and water pollution• Assessment of wind damage and other sever climatic condition• Detecting deforestation and forest degradation• Forest monitoring:• Some of the above• Logging activities• Reforestation and afforestation• Timber harvesting planning• Forest roads planning• Forest inventory• Forest management• Assessing slope failure and soil erosion• Assessing and managing forest recreation resources• Assessing and managing wildlife habitat

3.1.2 Measurements and estimation (quantitative): with reasonable accuracy 75% and more

• Forest cover area measurement• Number of trees• Tree height measurement• Crown cover measurement• Crown closure measurement• Crown diameter measurement• DBH estimation• Age estimation• Site estimation• Timber volume estimation• Thinning volume estimation• Basal area estimation• Annual Growth estimation• Basal area growth estimation


• Biomass estimation• Stand size• Dead, declined trees

3.2 The Application of Scanning Imaging System in Forestry

The applications of the multispectral scanner satellite or aircraft images in Forestry have exist three decadesago. Moreover, it has been used as an operational system for forest inventory in many countries. Therefore,we will summarise these applications as follow:

Depending on the spatial and spectral resolution (Air or Space and number of spectral bands):

3.2.1 Mapping (qualitative): with reasonable accuracy 75% and more

• Forest cover types• Identify individual species• Forest fire detection• Forest fire hazard• Detecting forest trees health (vigor and stress)• Forest trees diseases and insects infestation• Forest trees under air, soil and water pollution• Assessment of wind damage and other sever climatic condition• Detecting deforestation and forest degradation• Forest monitoring:• Some of the above• Logging activities• Reforestation and afforestation• Timber harvesting planning• Forest roads planning• Assessing slope failure and soil erosion• Assessing and managing forest recreation resources• Assessing and managing wildlife habitat

3.2.2 Measurements and estimation (quantitative): with reasonable accuracy 75% and more

• Forest cover area measurement• Tree height estimation• Crown cover estimation• DBH estimation• Age estimation• Timber volume estimation• Basal area estimation• Biomass estimation


3.3 The Application of Radar System in Forestry

This is a summary of the applications of SAR data in forestry. Good part of the review done by Hoffer et al.,1995 was used as a basis for this review. The following applications were considered in this review: Landcover types mapping, forest cover types mapping, detecting deforestation and reforestation, detecting burnedareas, mapping and monitoring forest changes, mapping forest wetlands, forest stand parameters, biomassestimation, forest inventory, and stress conditions (insects, diseases).

Land cover types mapping:

- Radar data can be used to map vegetation cover in tropical rain forest areas in relatively flat terrain.- Radar has shown some potential capability for mapping at the forest cover types.- Tree plantations are detectable on X or L, and P band. HH or VV polarised imagery.- Some tropical crop types can be identified with X band, HH polarised imagery.- Major roads and railroads, canals and water bodies can be mapped if not covered by forest, using C, X, L

P-band imagery.

Forest cover types mapping

- Radar can detect different forest cover types depending on the wavelengths, polarizations and incidence angles. In general P, L, S, X, and C proved to beable to differentiate forest cover types. This ability increase as the number of bands (e.g., polarizations, incidence angles, and wavelengths) as sources of data increase.

Detecting deforestation and reforestation :

- Detection and mapping of clearcut is possible with X, C, S, Land P-band, HH polariseddata, but rather mixed results were reported with C and X-band data, particularly if there was regenerationpresent.

Mapping burn forest:

- A few cases involving the mapping of burned areas have been documented using L- and C-band data, but relatively little work has been done in this application area.

Forests with water beneath the canopy:

- Mapping flooded forests has been accomplished accurately and effectively using Ka, X, C and Lwavelengths, with both satellite and aircraft SAR imagery.

- Distinction between marsh vegetation and flooded forests has been accomplished using X and L band imagery.

Mapping mangrove forest:

- It is possible to accurately map mangrove forests using Kit. C, X and L band imagery obtained from both aircraft and satellite platforms.


Forest stand parameters:

- There is a good relationship between most of the forest stand parameters and radar backscatter of cross polarised and L and P-band (e.g. DBH, Height, Basal Area, Density, and timber volume).

Biomass:

- There is a strong correlation between biomass and radar backscatter in monotypic forest conditions using L and P bands cross polarisation, with both aircraft and satellite SAR imagery.

Forest inventory data:

- Forest stand characteristics (e.g. tree height, basal area, DBH, volume ) can be estimated using radar backscatter of L and P-bands both aircraft and satellite SAR imagery.

Stress conditions:

- Identification of plant stress has been indicated using X and C- band imagery with aircraft data on a few occasions.

3.3.1 Review of Literatures of Radar applications in Forestry:

3.3.1.1 Applications of mapping, detecting and monitoring forest resources

Land and forest cover types mapping

Radar data can be used to map forest and other types cover in relatively flat terrain:

References: Cimino et al 1986, Drieman et al. 1989b, Freeman et al., 1992; Hoekman et al., 1992. Hoffer etal., 1985b. Hoffer et al., 1987b, Hoffer et al.,1988c; Hoffer et al.,1994; Hoffer and Hussin, 1989, Hussin,1990; Hussin and Hoffer, 1992; Hussin and Shaker, 1996; Lecki et al., 1994; Lozano et al, 1993; Mueller etal., 1988. Mueller et al., 1989; Thompson and Dams, 1990; Viksne et al., 1970; Werle, 1989b

· P-band using both HH and VV polarisation together provides the best results when compared to C and L bands. Clear differences are detectable between open water, reeds. sedge and swamp forest in swamp areas. Flooded areas are distinguishable from dry areas. Separation is possible between cohune palms. bajo. broadleaf upland. bare soil. farmland. and new/recent clear-cuts.

· P, L and C-bands were found to be complimentary in a study based in the Netherlands. L-band is better than C-band for discriminating between forest and non-forest. L-band also provides the best results when classifying 11 forest stand types.

· P and L bands are generally better for discriminating forest from non-forest but less useful for distinguishing forest types, especially at the species level.

· L-band. HH polarized data at a 50 incidence angle shows that capabilities exist to identify commercial logging sites. newly established rural settlements. cattle ranching areas and grassland successions. Separation of agricultural areas within tropical forest regions are not identified as easily.


· L-band, HH polarized data, using a combination of 3 incidence angles, was used to qualitatively and quantitatively map three age classes of slash pine, cypress or cypress/tupelo swamps, clearcuts, agricultural cropland and pasture. and water achieving an overall accuracy of 85%.

· The use of Landsat TM data in combination with three incidence angles of SIR-B (L-band. HH polarized) satellite data resulted in a more accurate cover type map than could be obtained using either the Landsat or the radar data alone.

· C-band. VV polarized data is often better (but not always) at discriminating forest types when compared to HV polarized data of a boreal forest in Canada. HV data is better at delineating seismic lines. roads and other forest disturbances.

· X-band. HH polarized data provides good results for discriminating primary and secondary forest. Cover types which have been identified include: mangrove, Nypa palm swamp, beach forest, peat swamp forest, trill dipterocarp forest, high scrub forest, tea tree and wallum cover, eucalyptus forest mimosoid legume and palm forest. Another study shows the capability to successfully map 30 forest vegetation units in Columbia onto 1:50.000 and 1:25.000 scale maps.

· K-band data is capable of detecting drainage features because or the distinctive reflectance of the water and the banks of the water bodies.

· Multiple incidence angle views should be collected as separation ot the land cover types is dependent on the vegetation characteristics.

· The effects of reflection, orientation and aspect can significantly impact the results and should be carefully considered in data interpretation.

· The cost effectiveness or using radar imagery was determined to be 5 times faster than using aerial photographs.

· Due to the inherent speckle in radar data the utilisation of such data for computer classification of forest or other land cover types is effective only if the radar data has first been "smoothed" using a low pass digital filter or through use of a contextual classification algorithm. A geometric filter provides high classification performance, but a sigma filter did a better job of retaining linear features and edge boundaries.

Radar has shown some capability as well as limitations for mapping at forest cover type and specieslevel.

References: Ahern et al., 1993a.; Arunarwati et al., 1997, Cimino et al, 1986;. Furley 1986; Hoffer et al.,1986; Hoffer et al., 1989b. Hoffer, et al, 1994; Hussin and Hoffer 1990, Hussin and Shaker, 1996;Knowlton and Hoffer, 1981; Marwat et al.1997, Thompson et al., 90.Van der senden 1991.

· Incidence angle significantly influences the capability to separate forest and other cover types.

· L-band, HH polarized with multiple incidence angles is capable ot` discriminating certain species depending on the incidence angle. In one study two species of forest, lenga and nire. were distinguished at 33 and 54 degree but not at 59 degrees.

· Like polarized (HH and VV) L - band aircraft data were found to be best (as compared to cross-polarized data) for discriminating between pine and swamp forest land. However, the cross-polarized imagery provided greater contrast between forest and non-forest than did the like-poiarized imagery.

· C-band. HH polarized data is not capable of providing detailed species composition information.


· Deciduous and coniferous forests are easily separated on ,K-band HH polarized imagery (deciduous has a much higher backscatter whereas they are difficult to separate on HV polarized data.

· X-band. HH polarized data was successfully used to map (1:1,000,000 scale six major vegetation units in a study which covered the country of Brazil. Savanna pioneer formation (seral communities), dense tropical forest, open tropical forest zones, (transitional areas) and anthropicallv disturbed areas were identified based on the tone and texture of the imagery.

· K-band imagery was used to produce maps of` Panama forests including categories of evergreen rainforest. mixed semi-deciduous and rain forest. sub-montane forest. palm forest. swamps with low trees. swamp with tall trees. mangrove swamps. marshes. cultivation. lumbering clearings. and settlements. A general knowledge of the vegetation associations of the area is essential to the successful interpretation of the imagery.

· The capability of distinguishing at the species level is rare because of the high species diversity typical in tropical forest. Therefore individual species are not normally distinguishable.

Tree plantations are detectable on radar imagery.

References: Dellwing, et al., 1978; Hoffer et al, 1987b, Hussin and Hoffer 1989; Thompson et al., 1990,Werle, 1989b.

· Plantations of slash pine were distinguishable from natural stands on L-band HH or VV polarized aircraft data but not with HV or VH polarisations.

· L-band. HH polarized data is capable of identifying Palm plantations due to its unusually high radar backscatter as compared to the surrounding lowland forests and transmigration areas.

· X-band, HH polarized data has shown that tree plantations can be detected relatively easily, if the plantations are large. have a distinctive pattern of planting or are located within primary or secondary forest areas. Examples of plantations which have been mapped include: pine. banana eucalyptus. oil palm. coconut palm and rubber. (Del78, Tho90) Some tropical crop types can be identified with A; band. HH polarized imagery.

Detecting tropical agriculture crops

· L-band. HH polarized data has been snows to be capable of identifying certain crop types banana. pasture, bamboo, cocoa, ricer because of their distinctive radar signature. pattern or moisture conditions.

Major roads, railroads, and canals can be mapped, if not covered by forest, using X or L band, HHpolarized imagery.

References: Hoffer et al 1985; Hussin and Hoffer 1989; Hussin and Shantha, 1994; Thompson et al, 1990;Werle 1989b.

· L-band. HH polarized data detected logging roads as long as they weren’ t covered by the forest canopy.


Detecting and monitoring forest land cover chnages

Radar has shown some capability to detect changes with P, L, C and X band.

References: Arunarwati et al., 1997; Driemen et al., 1989a; Hoffer and Hussin 1989b; Hoffer et al. 1990b,Hoffer et al., 1990d, Hussin and Shaker, 1996; Hussin et al, 1996; Marwat et al. 1997; Pantamina 1996.Quiñones, 1995; Hussin and Shaker, 1995; Hussin and Sta Maria, 1994; Hussin and Hoffer, 1992; Shaker andHussin 1995; Van der Senden 1991.

· L-band. HH polarized satellite data from Seasat (1978 23° incidence angle) and SIR-B (1984, 28° incidence angle) were used to define and map areas of deforestation. reforestation. and four "no-change" classes (forested swampland, old age stands of pine. medium age stands of slash pine, and open water) with an overall accuracy of approximately 84%.

· C-band. HH polarized data is capable of clearly differentiating between major forest types and wetlands and allows for the detection of changes caused by forest harvesting varvin groundwater levels in a study of a forest region in Canada. The author notes that only one study to date has used a multitemporal approach to studying forest vegetation backscatter characteristics.

· C-band. HH and VV polarized data and X-band. HH polarized data are capable of monitoring events which affect the degree of crown closure E.g.. logging, thinning, windthrow em.), indicated by a study of a Dutch forest. Backscatter values are primarily dependent on incidence angle and tree species. Identification of individual deciduous species will be difficult. The images must be calibrated accurately because ot` the small backscatter range of deciduous torest stands. Low incidence angles will be advantageous fbr detecting canopy disturbances. Narrow incidence angles are better. however classification performance will be reduced.

Detecting deforestation and reforestation

Detection and mapping of clearcuts is possible with L, S and X-band, but rather mixed results werereported with C and X-band dated particularly if there was regeneration present.

References: Ahren and Driemen 1988; Ahren et al., 1993; Hoffer et al., 1988a. Hoffer et al., 1989b; Hoffer etal., 1990b; Hoffer et al, 1990d; Kneppeck and Ahren 1989; Leckie et al., 1994; Lee and Hoffer, 1988;Lillesand and Kiefer 1994. Thompson et al., 1990; Werle 1989. Leckie et al 1984. Werle 1989b.

· L and P band are better than the shorter wavelength bands for identifying clearcuts and other major forest depletions.

· L-band. HH polarized data at steep incidence angles (23-28°) was effective for mapping areas of deforestation and reforestation.

· L-band. HH polarized data at 50c incidence angle was found to be useful for detection and possibly monitoring of forest conversion and depletion areas on a regional scale 1:250.000). Commercial logging sites. newly established rural settlements. cattle ranching areas and grassland successions are identifiable. Recently cleared patches of areas larger than 10-15 hectares are easily distinguished. Areas cleared for agriculture are not as easily identified due to the backscatter of cultivated and abandoned farmland being similar to the surrounding forest.

· L-band. HH polarized SEASAT data was not effective for mapping clearcut and forest regeneration sites in a mountainous area in Canada, due to the combination of small incidence angle (23 degree) and mountainous terrain.


· S-band, HH polarized, with a 44 incidence angle was able to detect clearcuts in a Canadian forest whereas. C-band. VV polarized data, with a 23° incidence angle was not. Canadian clearcut areas generally have a rough surface and are difficult to separate from forested areas with incidence angles less than 30 degree. Clearcut areas in the Amazon are typically used for pasture and therefore have a relatively smooth grass surface resulting in a distinctive backscatter as compared to the surrounding forest.

· C-band. HH polarized data was determined to be unacceptable for use in clearcut mapping in a study of Canadian forest. The foresters required clearcut mapping accuracies of 15 to 20m whereas an average difference of 30m in the boundary placement was found in the study. Major sources of errors were from not being able to clearly distinguish the edges of the cut blocks because: 1) the look direction was near parallel to the edge of the cut. 2) a strong, long shadow existed near the edge. 3) a strong return was received because of corner reflector effect between ground and vertical trees at the cut edge, 4) regrowth or residual vegetation was near the edge, and 5) geometric distortions caused by processing effects.

· C-band. VV and VH polarized imagery often confused ground vegetation in clear cuts with regeneration sites. wetlands. and pastures due to their backscatter similarities. VH data provided better clear-cut identification as compared to VV.

· C-band. HH polarized data results are inconsistent. There was confusion between clearcut and other non-forested areas (e.g.. wetlands. regeneration. cropland. pasture land). Vegetation on the clearcut is a significant factor.

· C-band. VV and HV polarized imagery of a Canadian regenerating forest burn area was not capable of distinguishing different densities of conifer regeneration due to variation in signal returns of deciduous brush. dead standing timber. slope and aspect. Detection was somewhat easier on the VV polarized data.

· X-band. HH polarized multi-temporal data were used to identify recent clearcuts and regeneration sites.

· X band HH polarized data has been used to identity selective logging using image texture tone or by associated features such as haul roads. Undisturbed first cycle logged forest and second cycle logged forest is identified. Clearings for agriculture as small as 25 ha can be mapped.

· X-band, HH polarized data was shown to be not capable of separating 1, 5, 10, and 15 year old cutover and regeneration sites.

· Shadows and edge effects due to extreme differences in the height of vegetation in adjacent areas helped delineate the boundaries of clearcut. Such edge effects are much more apparent on HV than on HH polarized X-band data.

Mapping burned forest

A few cases involving the mapping of burned areas have been documented using L- and C-band data,but relatively little work has been done in this application area.

References: Ahren et al. 1993a; Kasischke et al 1992; Lee and Hoffer 1990a

· L-band, HH polarized satellite data at 28° and 45° incidence angles was shown to be correlated with the amount of understory vegetation present due to prescribed burning.

· C-band. VV polarized data distinguished one year old burns in Black Spruce stands. The ability to detect burns can change significantly depending on moisture conditions. topography. and other characteristics of the area.

· Clapping of burned areas is not well studied or understood.


Flooded forests

Mapping flooded forests has been accomplished accurately and effectively using Ka, X,C, and L wavelengths, with both satellite and aircraft SAR imagery.

References: Colwell 1983; Hess et al., 1990; Hoffer et al 1986; Hoffer et al 1987b; Hoffer et al 1994. Hussinand Hoffer 1990. Imhof et al., 1986; MacDonald et al. 1980; Imhof et al 1987; Leckie et al., 1994; Lewisand MacDonald 1972; Ormsby et al; 1985. Place 1985; Werle, 1989; Richard et al., 1987, Wu and Sader1987.

· The following radar specifications have been most successful in mapping flooding beneath forest canopies: L band. like polarisation (HH & VV) and incidence angles Of 20° - 30°.

· Cross polarization is not effective in mapping flooding beneath a forest canopy.

· Though it is possible under specific circumstances to detect flooding beneath forest canopy at large incidence angles (30° and above), this ability is dependent on the morphologic characteristics of the forest canopy such as density of tree crown, leaf orientation, and percentage of forest crown closure. These factors can cause absorption of the radar signal before it reaches the underlying water at larger incidence angles such as 58°. Since it is always possible at angles between 20° and 30°, these angles are advised.

· Ka and C bands are usable for mapping defoliated flooded forests but not over foliated forests. No indication of polarization or incidence angle was given with Ka and C band studies.

· X band does not appear to be effective for distinction between flooded foliated forest and non-flooded forests.

Distinction between marsh vegetation and flooded forests has been accomplished using X and L bandimagery. (Hess1990, Ormsby 1985)

· X band. HV polarisation is good for distinction between marsh vegetation and looded forests.

· L- hand. AH polarisation is best for discrimination between swamp and marsh vegetation.

Mapping mangrove forest

It is possible to accurately map mangrove forests using L, C, X and Ka-band imagery obtained fromboth aircraft and satellite platforms.

References: Ahren et al. 1993a; Dellwig, et al.,1978;. DeMolina et al., 1973; Gelnett et al., 1978;MacDonald et al., 1980; Imhoff et al., 1986a; Imhoff 1988; Lewis and McDonald, 1972d; Thompson, et al.,1990; Zuhair 1998; Zuhair et al., 1998.

· The greatest correlation with the L band, HH polarisation imagery occurred at incidence angles of 26° at high tide and 58° at low tide. This phenomenon may be explained by the fact that at larger incidence angles the vertical tree stems present a larger surface area to the radar thus creating a brighter overall response eaten in the absence of greater amounts of standing water (low tide) Two levels of image enhancement were applied to the data.


· The greatest correlation with the X and C band data occurred using HH polarisation.No incidence angles were mentioned. I; and X band. HH polarized data have successfully mappedmangrove forests because of its distinctive reflectance from the dense interlocking network of moistvegetation.

· Mangroves are readily distinguishable on K-band radar imagery because of the strong return signal and characteristic texture compared with other tropical vegetation.

· Mapping mangroves is easily and accurately accomplished in the near middle range of incidence angles of Ka-band aircraft radar data.

3.3.1.2 Forest stand parameters

Because of the importance of forest stand parameters and biomass we will elaborate on this part beforegiving some conclusions.

Description of forest stand parameters

For the purpose of this study, the description of forest stand parameters (Age of stand, DBH, Height, BasalArea, Stand Density) are adapted from B.Huch et al 1982.

Age of stand/trees:

The age of a tree is the length of time that has elapsed since the germination of the seed or the budding of thesprout. Even-aged stands are stands in which all the trees are essentially of the same age. But a standoriginating through natural reproduction, which generally takes 1 to 15 years or more, contains trees ofvarious ages. There are several conceptions of the age of such a stand. A simpler concept, however, withmany advantages, is to consider the age of the stand to be the average age of dominant and co-dominant trees(i.e. the largest individuals.).

Diameter at breast height (DBH/d):

Diameter measurement is important because it is one of the directly measurable dimensions from which treecross-sectional area, surface area, and volume can be computed. Diameter at breast height is a standardposition in standing trees at 1.3 m. above the ground level. Diameter should be qualified as outside bark(o.b) or inside bark (i.b). However, when this designation is omitted from breast height measurement (DBHor d) as it is often is, the measurement is assumed to be outside bark.

Height of stands/trees:

Total height of tree is the distance along the axis of the tree stem between the ground and the tip of the tree.The following are the important methods proposed to obtain height of a stand.

1. Measure and average the heights of all of the trees, or a sample of the trees, regardless of their size orrelative position in the stand.

2. Measure and average the heights of the dominant trees, or of the dominant and co-dominant trees.3. Measure and average the height of a fixed number of largest trees.

Basal Area:

The cross-sectional areas of planes cutting the stem of a tree normal to the longitudinal axis of the stem takenat breast height is called the basal area. The total basal area of all trees, or of specified classes of trees, perunit area is a useful characteristic of a forest stand. For example, basal area is directly related to standvolume and is a good measure of stand density.


Stand density:

It is a quantitative measurement a stand in terms of square feet of basal area, number of trees, or volume peracre. It reflects the degree of crowding of stems within the area.

Biomass:

Newbould, (1967) defines biomass as the total amount of green plants including wood, bark, stump, roots,live branches, and dead branches. Biomass of a tree consists of aboveground and belowground parts. In thisstudy only the aboveground biomass of a tree is considered.

Relationships between radar backscatter and forest stand parameters

Forest stand parameters, which are the biophysical characteristics of a stand, are the most important factors,which are used in estimating above ground biomass of forest stands. For a given species, standing biomassis mainly a function of DBH, tree height, and stand density, which are themselves dependent on tree age,forestry practices, and also environmental and genetic factors (Kashischke et.al 1990). Therefore, in order toexploit the use of radar backscatter that determine above ground biomass of forest stands, there must be agood relationship between radar backscatter and those biophysical forest characteristics.

In the past, several studies have been conducted focusing on the relationship between radar backscatter andforest stand parameters. WU (1984) reported that the digital classes (i.e. young pine, pine plantation andnatural pine of 5,15, and 30 years old, respectively) were highly correlated with radar backscatter in Alabamatest site. Le Toan et al 1992 made an experiment using P-, L and C- band SAR data and showed that for L-band and the highest correlation has been observed with basal area for all polarization and related that cross-polarization data were more correlated to basal area and other forest parameters than HH and VV data. Thebest correlation coefficient at L-band was between HV and basal area. At P-band the result has shown thatvery high correlation coefficients have been observed with height, trunk, biomass, DBH, basal area in thedecreasing order; and cross-polarized returns have higher correlation’s coefficients with forest parameters.However, the statistical analysis was incomplete due to absence of biomass characterization. (A.Beaudoin etal., 1994). Beaudoin et al 1994 have presented (Table. 2) the results of linear regression analysis betweenbackscattering coefficients of forest stands (24 stands + 9 clear-cuts) and eachforest parameter, including biomass as derived from allometric equations.

Table1. Correlation coefficients from linear regression analysis between backscatter and forest parameters atP and L band (8-46 year-old: 24 plots, 9 clear-cuts: 1 mean value). Boldface characters indicate thefour highest correlations’ for a given configuration. (Adapted from Beaudoin et al,1994)


From the table it can be explained that for a given frequency, higher correlation was found at HV, followedby HH and VV whereas for the general forest parameters, higher correlation was found with stand height,DBH, tree and stand basal area. For the biomass, it was found that correlation is generally higher than withthe general forest parameters, except for the needle biomass. In particular, correlations with trunk, branchand total above ground biomass are high and close to each other.

Hussin 1990 using L-band multipolarized multiple incidence angle aircraft SAR data indicated that there wasa strong statistically significant relationship between HV-polarized radar backscatter and several forest standparameters (e.g. the correlation coefficients comparing radar backscatter with age, DBH, Basal area, height,cords/acre, trees/acre, and stand biomass were 0.93, 0.94, 0.90, 0.83,0.91, 0.81 and 0.97 respectively).Hussin et al 1992 have also shown poor relationships between SIR B L-band multiple incidence angle HH-polarized data and forest stand parameters.

Le Toan, et al 1991 to establish relationships between forest parameters and measurements retrieve from P-,L and C-band SAR data have shown that both the radar backscatter intensity and co-polarization phasedifference are found strongly correlated to forest biomass and other main forest parameters including height,age, DBH, and basal area at P band. The dynamic range of the radar backscatter intensity corresponding to alarge range of trunk biomass was also found decreasing from cross- to co-polarization (Figure 1). Hussin etal 1991 has also shown a strong positive relationship between L-band HV polarized radar backscatter usingaircraft data and slash pine stand parameters (i.e. biomass, height and basal area).

Figure 1. P-band relative mean backscatter as a function of stands trunk biomass for a)HH, b)VV and c) VH polarization, at incidence angle from 35° to 50°.

Finally after analyzing the existence of good relationships between these forest stand parameters and radarbackscatter, above ground biomass can be estimated using one or more of the estimated forest biomassparameters through the use of different forms of equations (allometric equations).

Above-Ground Biomass Estimation of Forest Stands

“It is known that radar signals specially those which are operating in the longer wavelengths i.e. L- or P-band do have the potential of penetrating the surface of the canopy and hence interacting with the differentbiomass component of forest stands. Thanks to this penetrating capacity, radar remote sensing systems canoffer better capabilities for estimating forest biomass and other structural parameters than optical remotesensing systems” (Van der Sanden. 1997).

Recent efforts to use synthetic aperture radar (SAR) to measure aboveground biomass have focused almostentirely on comparisons between backscatter intensity and ground-based biomass measures. Most of theexperiments have been concerned with developing regression relationships between radar backscatter and


stand-level biomass for various tree species, many of which were grown in plantation settings (Imhoff.1995).

Model development

In order to develop a model to estimate the biomass of forest stands, relationships should be determinedbetween radar backscatter and stand parameters. These relationships will help in developing regressionmodels to predict the forest stand biomass from the radar backscatters. Different studies (Hussin et al, 1991;Le Toan et al., 1992; Dobson et al., 1994) have shown good correlation between radar backscatter and aboveground biomass.

To develop the model therefore, we need to have at least two types of data sets. One of the data sets is usedto develop the models; i.e. to identify the structural form of the equation and provide estimates of where asthe other one is used to validate the models. To estimate the stand biomass Hussin et al 1991,Sader198-, Wu198-have used “Total-tree Multi-product Cruise Program “ which contains various regression equationsdeveloped by US-forests Service- southern Forest Experimental Station and the school of Forestry at theUniversity of Georgia.

Model Specification

Assuming that average stand biomass (BIO) is directly related to the cubic volume which in turn is directlyrelated to basal area per ha (BA) and average stand height, Hussin et al 1991 have specified a form equationwhich appeared as: ∧Mode l: BA = f (Radar) ∧ HT = f (Radar) (2)

∧ ∧ ∧BIO = f (BA,HT)

This assumes that both BA and HT can be estimated using radar backscatter. This form of equation can beestimated well, if we get current unbiased information of basal area and/or height. After knowing that thereis some sort of relationships between radar backscatter and biomass of stands, it is possible to model theexisting relationship. Hussin et al (1991); Rignot et al (1994) have fitted a third order polynomial responsefunction to explore all possible linear combinations of variables to find the “best” combination ofindependent variables that best explains the variance of the dependent variables.

While developing the model to estimate biomass as a function of radar backscatter, it can happen thatdependent variables may vary over the range of data set. To stabilize this some form of transformation maybe needed. Hussin et al (1992) has used Box-Cox transformation method to stabilize the variance of themodel.

In the past, different techniques to estimate a set of coefficients associated with the system of equating havebeen developed. Several of the more common estimation procedures include Limited Information MaximumLikelihood (LIML), Zellner’s seemingly unrelated least squares procedure, Two stage Least Squares (2SLS),and Three-Stage Least Squares (3SLS) to name a few (Theil, 1971) cited in Hussin (1990).

For the study (Hussin, 1990), the coefficients associated with the structural equation identified in thecombination screening process were estimated simultaneously using a matrix programming language, toobtain the 3 SLS estimates. Based upon this, Hussin et al 1991 developed two sets of equations forestimating biomass, depending on the type of information available. The reduced forms of the system ofequations are:


Model 1 ∧BIO-1/2 = 0.6914 – 0.0074 (RADAR) (3)

+ 2.1903 (RADAR) 2

Model 2 ∧BIO-1/2 = 0.39889- 0.00772 (RADAR) + 4.7649 × 10-6 (RADAR) 2 – 9.6503 × 10-8 (4)

(RADAR) 3 + 4.3410 (BA) –1/2

Where RADAR is a digital number representing the amount of radar backscatter and BA is the basal area perha.

Using the data (Biomass Vs Radar backscatter) at C-, L- and P_ band Rignot et al 1994 derived regressioncurves relating the logarithm of δHV, δHH, or δVV at on frequency, i.e. of the form Log B= Σi =0

i=2 ai δxy

where B is the stand biomass, δxy is expressed in decibels, x’s and y’s are H’s or V’s, and ai’s are thecoefficients of the polynomial. Biomass is then predicted from the radar measurements and compared toinventory estimates.

Kassichke, et al 1995 presented biomass prediction equations as calculated after (G.L Baskerville, “Use oflogarithmic regression in the estimation of plant biomass “ Can J. For. Resource, Vol 2. Pp. 49-53, 1972)

Log biomass = b + a1 δC-HH + a2δ C-VV + a3δC –VH

+ a4δL- HH + a5δL-VV + a6δL-VH (5) + a7δP-HH + a8δP-VV + a9δP-VH

Where b is the regression constant and a1 to a9 are the regression coefficients.

Application of the Model

It is known that the use of conventional forest inventory techniques to estimate the above ground biomass offorest stand is time consuming and costly. In this method, one could randomly sample a representativenumber of N stands to estimate the average biomass per unit area using the following equation: (Hussin,1990).

∧ − Y = Y ΣAi

∧Where: Y = estimate of the total biomass ΣAi = summation of the area of ith forest stand

− Y = average biomass per unit area

Due to above-mentioned reasons, the use of remote sensing techniques such as radar data is of vital and costefficient estimate of biomass.


Hussin, (1990) in the research provided some insight on statistical properties of three estimators forestimating total biomass. Given a simple random sample of n out of N stands where each stand has an equalchance of being selected, an estimate of the average biomass per unit are Yi, is: − n − Y 1 = 1 Σ Y i (6)

n

− −Where Yi is the average biomass per unit area of the ith stand. The variance for Y1 is:

∧ − − − V(Y1) = ( N-n) Σ (Yi - Y1 ) (7)

N n (n-1)

Next if we define AiYi to be the total biomass of the ith stand, dividing by A, the total number of acres in thepopulation, we obtain an unbiased estimator of the population mean µ:

− − Y2 =( N ) Σ ( AiYI ) (8) A n

The variance is given by: − V(Y2) = ( N - n ) ( 1 ) Σ (AiYi - Ai Y2)

2

n A2 n – 1 (9)

−The estimator Y2 requires knowledge of the total number of acres in the population. When A is unknown, itmust be estimated from the sample data. One way to accomplish this is to multiply the average stand size ΣAi / n by the number of stands in the population, N. replacing A by its estimator, one obtains a ratioestimator of the average biomass per unit area:

Y3 = ΣAi Yi (10) Σ Ai

The estimated variance for Y3 has the form of the variance of the ratio estimator:

− V(Y2) = ( N - n ) ( 1 ) Σ (Ai (Yi - Y3)

2

n n A2 n – 1 (11)

In estimated total biomass of stands, one can use previously developed regression models for predictingbiomass. However, because of the bias associated with the sampling designing the estimate of averagebiomass will two main error components of various sources. The first component is due to the randomselection of stands and the other one is due to the regression model used to estimate biomass. Thus thevariance of y is given by

− − −V (Y) =V (YD) + V ( Y R ) (12)

− −


Where V (YD ) is the variance due to the sample design, and (YR ) the variance due to the error of theregression model used in estimating biomass.

Before one can estimate the variance associated with the regression model, one need to take intoconsideration the type of the transformation, if any, used in developing the biomass equation. If for example,the inverse of the square root of biomass was used as the dependent variable, the variance of T = Y –1/2 canbe obtained by applying the taylor expansion series to ;

δT2 = (f′ ( Y) δY) 2 (13)

Where f ′ (y) is the first derivative of T with respect to y and δY is the standard deviation of theuntransformed y . (Hussin 1990).

Finally a computer simulation was conducted to compare the statistical properties of the three estimators of yusing the two models developed in Section 4.1. The population used in the simulation consisted of N= 500stands randomly selected, with replacements, from 55 stands used in section Hussin, (1990) to develop andvalidate the models for estimating biomass as a function of radar backscatter. The hypothetical populationcovered 32,698.3 hectares with a total biomass of 4,440,566.4 metric tons, or 135.8 tons/ha. For each of thetwo models, 1000 estimates of the mean and variance of y1, y2, and y3 were obtained using sample sizes n=25, 50,100 and 200 stands randomly selected without replacement.

Above ground Biomass Estimation

The key questions concerning the possibility of inferring forest biomass from SAR data are the validitydomains of the inversion algorithms regarding the radar parameters, frequency, polarization, incidence angle,the forest type and growth stage, and the environment (Le Toan, et al., 1992).

The following are some of the prior works done by several researchers on the potential uses and applicationof radar backscatter for estimating aboveground biomass and biomass components in different forest regionand forest type.

Temperate Forest

Plantation Forest

Coniferous Forest

Several researches conducted on the use of radar backscatter for biomass estimation have been focusingmainly on plantation of coniferous forests mainly of pine species. Le Toan, (1991) explained that in order tooptimize the results obtained from the experiment, the test site had been selected to meet the followingcriteria: flat topography, homogeneous and large stands, presenting an optimum range of forest parameters ofinterest. It was also mentioned that monospecific stands would suit better as to increase the understanding ofthe mechanisms involving microwave interactions with vegetation canopies.

Wu, (1987) presented the techniques and the potential utility of multipolarization. Synthetic Aperture Radar(SAR) data for biomass estimation of pine plantation. In this experiment three channel of SAR data, onefrom the shuttle Imaging Radar SIR-A the other two from the aircraft SAR, were acquired over the BaldwinCounty, Alabama, study area. The SIR-A Data were acquired with HH polarization and the aircraft SARdata with VV and HH polarization. Linear regression techniques were used to estimate the pine plantationbiomass.


The result indicated that multipolarization data were highly related to pine-plantation biomass and therebysuggested their potential application in estimating pine-plantation total-tree biomass. Besides, it has shownthat the results of the aircraft SAR data were more reliable than those of the SIR-A data, even though they allindicated the potential on total biomass estimation.

Dobson, et al 1992 made two independent experiments to examine the dependence of radar backscatter onaboveground biomass of mono species conifer forests using polarimetric airborne SAR data at P-, L- and C-bands. The test sites were the plantations of maritime pine near Landes, France ranging in age from 8 to 46years with aboveground biomass between 5 and 105 tons/ha and the loblolly pine stands established onabandoned agricultural fields near Duke, NC range from 4 to 90 years and extend the range of above groundbiomass to 560 tons ha-1 for the older stands. Both sites were even-aged monospecies stands. For bothforests, the stands to be used in that study were imaged by the SAR over an equivalent range for angle ofincidence 40 < θ < 50.

The result of the experiment showed that for P-band δ was found to increase linearly with biomass andapproach saturation at a level of 100-200 tons/ha. For L-band, δ also increased linearly with biomass butapproaches saturation over the region from 60 to 100 tons/ha (2) For C- band, δ was found to increase withbiomass, but the available dynamic range was small (≅ 5 dB) and there was much dispersion in the data. Asa result, δ was found to be weakly correlated to total aboveground biomass (r >0.5) (Table2).

Table 2. Regression Results (Adapted from Dobson et al , 1992)

Figure 2. Calibrated L-band backscatter as a Figure 3. Calibrated C-band backscatterfunction of the log of total aboveground biomass (tons/ha) as a function of the log of total aboveof maritime pine and loblolly pine ground biomass of maritime pine and

loblolly pine (adapted from Dobsonet al, 1992)


The linear dependence of δ on biomass tends to saturate at biomass levels, which scale with wavelength.The results provide strong evidence that retrieval algorithms for aboveground biomass using longwavelength SAR data should be robust within structurally similar forests (i.e. excurrent conifer forests).

Le Toan et al 1992 made an experiment to demonstrate the use of radar to retrieve forest biomass in thelandes forest in Southwestern France test site. The test site selected for the experiment was an area of 7 km x10 km comprising quasiuniform large homogeneous canopy stands of mean area of 25 ha (~300m x 500m)rectangular in shape and delimited by fire protection tracks or access tracks. The mean backscatteringcoefficients in (m2 / m2) had been measured at P- and L- bands for HH, HV, and VV polarisation’s. Ingeneral, the correlation coefficients are higher with P-band compared to L-band. HH and VV yield similarrelationship with a slight advantage for HH.

The experimental result showed obvious usefulness of L- and especially P-band radar for retrieving forestbiomass. The best correlation coefficient was obtained between P-band and HV and the trunk biomass (r =0.95). From the results of correlation coefficient, it was not obvious to conclude about the particular effect ofa given forest parameter, since these could provide useful indications to initiate the modelling phase.

Hussin et al (1991) conducted a research to develop a mathematical model for predicting average standbiomass for slash pine plantations as a function of radar backscatters in Jacksonville, FL, test site.

The experiment came out with two models (look Model 1&2). These systems of equations accounted forapproximately 97% and 98% in biomass. However, the equation for predicting basal area in model 1accounted for only 83% of the observed variability in spite of the high R associated with the biomassequation in Model 1. To evaluate the models, the systems of equations were used to estimate averagebiomass per hectare using the validation data set. The statistical properties of the equations were evaluatedin terms of the percent bias and percent sampling error after retransforming the variables back into theiroriginal form. Both models generated predicted values that matched well with the observed values for theindependent set of data. The average percent bias associated with estimating biomass was less than 1%, i.e.0.98% for model 1 and 0.80% for model 2, with percent sampling errors of 3.27 % and 2.54%, respectively.As one would expect, the system of equations (Model 2) in which information on basal area was availablefrom the field data provided better estimates of average stand biomass than the system of equations (Model1) in which basal area was estimated using the SAR data.

The results of this study indicated that there was a strong positive relationship between L-band and HVpolarized radar backscatter using aircraft data and slash pine biomass (Figure 4). Regression equationsdeveloped in this study accounted for 98% of the variability observed in average stand biomass. Whenapplied to an independent data set, the biomass equations had an average bias of less than 1% with a standarderror of approximately 3%. The system of equations presented in this study, when combined with anappropriate inventory design could provide efficient estimates of the total biomass on a regional scale.(Hussin et al, 1991).

Reich et al, (1993) evaluate several estimates that could be used in designing a regional forest inventory toestimate average stand biomass using L-band radar data. Using two biomass models developed by Hussin etal (Model 1 and 2) three estimators were compared in a simulation study using sample sizes of 25, 50 100and 200 stands. Model 1 estimates average stand biomass as a function of the amount of radar backscatterand basal area per hectare. Because of differences in the biases associated with the two biomass model(Hussin et al, 1991), it was not possible to choose a single estimator that could be used with both models. Itwas suggested that a simple average of unweighted stand biomass was recommended for estimating averagestand biomass when using Model 1, while a weighted ratio of biomass to stand size was recommended foruse with Model 2.


Figure 4. Relationship between measured and estimated biomass per hectare as a function of HV radar backscatter.

Finally, Reich et al (1993) concluded that the decision to use model 1 or model 2 to estimate biomass woulddepend primarily on the cost associated with obtaining ground information on basal area per hectare.Otherwise, model 1 should be used to estimate biomass.

Recently, Beaudoin et al (1994) explained the higher correlation of P-band SAR images to the forest aboveground biomass and further explained the understanding of the observations, using theoretical modellingapplied to calibrated SAR data to explain the radar backscatter from the forest canopy on the test siteexplained by Le toan et al (1992).

In the study, three main scattering mechanisms identified crown volume scattering, trunk-ground interactionand crown-ground interaction, which are dependent on the polarization and forest growth stage. Themodelling and simulation studies had shown that the HH return was found to be physically related to trunkbiomass, where as VV and especially HV returns were found to be tightly linked to crown biomass andunaffected directly by environmental parameters such as soil moisture, ground local slope and the presenceof an understorey bush layer that show the possibility of retrieving total aboveground biomass from VV orparticularly HV P- band SAR data.


Figure 5. Main contributions to total backscattering coefficient at P- band with 45° incidence angle, as a function of aboveground biomass.

In the study (Kasischke, et. al 1995) a multifrequency, multipolarization airborne SAR data set was utilizedto examine the relationship between radar backscatter and aboveground biomass in loblolly pine forests andits potential to estimate biomass. The study uses a step wise multiple-linear regression approach, using all theradar channels as independent variables, and the log of biomass component as the dependent variables. Theresult of this regression analysis produced equations with high coefficients of linear correlation (r = 0.93 andhigher) and low standard errors of the regression equation (se = 0.15 - 0.23) for estimating total stand, boleand total stem biomass. However, when the predicted biomass estimates were expressed in arithmetic termsand compared to actual values, low levels of accuracy was found. Finally, a second, two-step, method wasdeveloped using total stem biomass being estimated from the radar image intensity values and this reducedthe coefficient of variation to lower ranges for all biomass components. The study concluded that from theanalysis the image intensity signatures recorded on SAR imagery have the potential to be used as a basis forestimation of aboveground biomass in pine forests, for total stand biomass levels up to 35 - 40 kg m-2.

Mixed Stands

Rauste et al 1994, studied the potential of radar based tree biomass estimation using polarimteric SAR datafrom the Freiburg test site of MAESTRO 1 campaign (organized jointly by the Joint Research Centre of theEuropean Communities and the European Space Agency) and scatterometer data from a test site in Finland.The result of this study indicated that in P- band, the maximum correlation’s, which were found near thelinear HV polarization, were up to 0.75. In the Finnish test site, a strong negative correlation (correlationcoefficient –0.65) existed between the pine biomass and X-VV Backscatter. When the combination of X andC bands (measured by the Hutschat scattrometre) was used, a correlation coefficient of 0.81 was obtained.

This study has shown that the highest correlation’s between tree biomass and backscatter found in theFreiburg test site were similar but slightly lower than those found by Hussin et al (1991) and Le Toan et al(1992). Some of the factors for this slight reduction mentioned were topography, type of stand, range ofbiomass and relatively low accuracy of the data used.


Figure 6. Schematic representation of the radar backscatter as a function of tree biomass at P, L, C and X band. (Adapted from Rauste et al, 1994)

In the analysis of the Freiburg SAR data, a positive correlation between P-HV backscatter and tree biomasswas found in the biomass range of 0-50 kg m-2. In L- band, there was a positive correlation at 0-7 kg m-2 afterwhich the backscatter decreased slowly with increasing biomass. In C- band, three was a weak positivecorrelation in the range 0-3.5 kg m-2 after which the backscatter decreased more rapidly than in L- band.

In the analysis of the scatterometer data, the C-bank backscatter was found to be almost independent ofbiomass. In X band, there was a positive correlation at 0-25 kg m-2 after which the backscatter decreasedrapidly with increasing biomass.The C band of the SAR abroad the ERS-1 satellite doesn’t seem to be very promising for tree biomassestimation.

Israelsson et al 1994 made a study in order to evaluate the potential of estimating forest bole volumes usingradar remote sensing on Flevoland test site in Netherlands for MAESTRO 1 campaign. The forest area is amixture of coniferous and deciduous forest types of tree species as popular, oak, ash and a number of verydense plantations of pine and spruce. The results of this study shows that at P-band, theory and experimentboth show the greatest sensitivity to bole volume of the three frequencies. A saturation level for P- bandbackscatter has been found theoretically at around 200-m3 ha-1 for poplar stands. At L band (λ= 24cm), thewave is partly penetrating the canopy layer giving a smaller sensitivity to bole volume. The model result ofthis experiment also indicated that a backscatter saturation at lower bole volumes than at P- band (λ =5.6cm), the backscattered energy had been shown to exhibit poor correlation with bole volume, as thesources of scattering were located in the crown layer (leaves, needles, twigs, small branches) of the canopy.Results from the model also show that less sensitivity to bole volume is expected at smaller incidence angles.

Natural Forests

For various reasons, most of the previous works on the estimation of aboveground biomass using radar datawere conducted dominantly on the temperate coniferous species. However, some studies have alreadyconducted in some natural stands in the region. Below are some findings by different scientists on this issueof aboveground biomass estimation using radar data.

Rignot et al 1994 examined the relationship between radar backscatter and aboveground biomass in a naturalforest setting, with mixture of coniferous and deciduous tree species, strongly varying environmentalconditions, and level to moderate topography. In the study Airborne SAR data gathered by the NASA/JPLthree frequency, polarimetric, radar system in winter, spring, and summer over the Bonanza CreekExperimental Forest, near Fairbanks, AK, were compared to estimate of whole tree above-ground drybiomass from 21 forest stands and two clear cuts.


The result of this study (Rignot et al 1994) suggested that SARs operating at long radar wavelengths havepotential for mapping above ground biomass of boreal forests in interior Alaska. The error in predictingbiomass from the radar was about 20% of the actual biomass at p-band. L band signals do not perform aswell because of more pronounced dependence on tree species (for instance alder trees), but the difference inperformance between the two frequencies remains small. Prediction of forest biomass from the radar inmonospecies forest plantations was 10% better, but this was an expected result. Data dispersion was moresignificant for natural forest ecosystem than for forest plantations because uncertainties in estimating above-ground biomes were larger in natural forests due to spatial variations in tree species, age, density, height, andDBH; and also because the radar signal, interact with a canopy of spatially varying three-dimensionalstructure.

Dobson et al 1995 made a study to show the accuracy of multifrquency, polarimetric SIR-c/ X-SAR data inestimating biophysical properties of forest structures. The result showed that high accuracy retrievals of totalbiomass were possible using SIR-C/X-SAR. Figure 7. While simple single frequency and polarizationretrievals were limited by saturation effects, the multi-step approach used herein showed retrieval wereaccurate up to at least 25 kg m2 for the forest communities studied.

Figure 7. Accuracy of SIR-C SAR-derived estimates of total dry.

Pullianen et al 1994 studied the backscatter properties of boreal forest using empirical airborne andspaceborne C- and X- bands radar data in forests of Finland. The result obtained showed that radar responseto the forest stem volume (directly related to forest biomass ) was relatively low at both C- and X- bands.However, Pullianien et al 1996 investigated the seasonal changes of the C- band backscattering properties ofboreal forests by employing multitemporal ERS-1 SAR data (C- band) and a semi-emperical backscatteringmodel. The result demonstrated that the response of radar backscatter to forest stem volume was foundchange drastically depending on weather and seasonal conditions. (Figure 8). The inversion techniquedeveloped for stem volume estimation using multitemporal ERS-1 SAR data appears to work satisfactorily.Previous studies have shown poor results in the estimation of forest biomass (stem volume) using C-bandbackscattering coefficient data exclusively. The result in the experiment showed a significant improvement.


Figure 8. Observed response of ERS-1 SAR to stem volume Mantsala test site.

Tropical Forest

The use of synthetic Aperture Radar (SAR) for monitoring and analysing tropical forest is of interest for twoprimary reasons. First, the microwave region of the spectrum is unaffected by the persistent cloud cover sotypical of humid tropics, and second, there is a possibility of deriving information about forest structure(height, stem diameter and frequency, basal area, canopy roughness, and aboveground biomass) from SARbackscatters measurements (Sader et al, 1990).

Based upon previous researches conducted focusing aboveground estimation of stands using backscatter, itcan be said that there has been comparatively lower number of researches conducted in the tropical forest.Different reasons might be accounted for this lower number of studies. The studies by Freeman et al, (1997)and Rignot et al (1997), illustrate a common problem for investigations into relationships between remotesensing measurements and tropical forest structure, namely the difficulty of obtaining the required set ofground referenced data as land cover maps which are seldomly available and typically require detailed andregular updating to be reliable.

Despite that problem, several studies have been conducted in the tropical forest region for estimation ofbiomass using radar data. The breaking up of a forest into botanical subunits may also prove difficult intropical forest where hundreds of tree species are mixed together (Rignot et al 1995). Below are somefindings of studies in mixed forest of tropical region.

Rignot, et al 1995, using P band HH, VV and VV polarization data combined estimate total abovegroundwoody biomass in the tropical rainforest of Manu, in Peru, where forest biomass ranges from 4 kg m-2 inyoung forest succession up to kg m-2 in old, undisturbed floodplain stands. The result of this experimentshowed that at circular polarization, P band data did not estimate forest biomass correctly because large HH-VV phase difference in palm forest yield large values of backscatter and >100% over predicted woodybiomass. This example shows the polarimetric phase differences were not well correlated with forestbiomass. Better radar predictions were obtained using the P band HH, HV, and VV polarization combinedwith no phase information.

The result consistent with the fact that HV signals were correlated with branch biomass, which might ormight not be a good indicator of total biomass depending on forest type and biomass level; where as HHpolarization was correlated with site biomass which was a more direct indicator of total biomass, especiallyat the high biomass levels.


Hoekman et al (1996) using C-, L- and P-band backscatter measurements in San Jose del Guaviare,Columbia have found that the correlation between backscaatter and biomass or basal area with backscatterand biomass or basal area was low for C-band and high for L-and P- band. It was further explained that thislow correlation between biomass and basal area with backscatter in C-band could be expected on the basis ofphysical considerations. At that wavelength radar senses the upper parts of the canopy only whereas the highcorrelation at L-and P_ band was an indication for a certain relationship between biomass and basal area andthe level of backscatter. The study found out that even though the biomass estimation may not be veryaccurate (errors may be in the order of 20-30%), for the large relative range of biomass values (several ordersof magnitude) such an error was still acceptably small).

In a brief summary one could state that for tropical forests for longer wavelength, i.e. L- and P-band, top soilmoisture content is a main factor for the intermediate range of biomass, i.e 10-20 tons ha-1 (fresh weight) andvegetation structure and soil surface state are the main factors for the high biomass range. (Hoekman et al,1996). It is expected that the precision of the radar will increase if structural differences between forest typesare accounted for during the inversion of radar data. (Rignot et al 1995).

Freeman et al (1992) also using multi-frequency, multipolarization SAR data have concluded that crosssection due to volume scattering appear to increase with biomass until a plateau is reached. However,vegetation with low biomass might have high cross section due to double bounce scattering.

In the hilly area, the spatial variability in radar backscatter is controlled by surface topography. That is,forest biomass is overestimated in the hills facing away from the radar (Rignot et al, 1995). In estimatingabove ground biomass of forest stands over the hilly terrain, one should also account the impact oftopography and employ digital elevation model before data inversion process is conducted. Most of thestudies made on the interaction of radar waves with forest stand characteristics were limited to relatively flatareas rather than mountainous areas with significant differences in elevation.

In a study made by Wu, (1990) a digital terrain elevation data set was coregistered with radar data forassessing mountainous tropical forest stands characteristics using both raw and topographically corrected L-band polarimetric radar data acquired over the tropical forests of Costa Rica. The result of this analysis using18 out of 81 plots for two test sites indicated that per-plot bole volume and tree volume are related tosynthetic Aperture Radar (SAR) data.

Structural variations also have an effect in estimation above ground biomass in tropical forest. Imhoff 1995made a study using NASA/JPL AIRSAR P-, L- and C- band quad polarization configuration at incidenceangles of 20°, 40° and 60° to study the effects of canopy geometry on radar backscatter from tropical andsubtropical broadleaf forests. By employing MIMICS models it was demonstrated that structural variationmight have a substantial effect on P-, L- and C-band quad polarization backscatter of forest stands with equalbiomass. A structural descriptor made by the ratio to increase with the consolidation of vegetation volumetended to increase with the consolidation of vegetation surfaces into fewer and larger parts. The lack of suchknowledge has impeded an explanation, in real physical terms, of mechanism by which SAR backscatter waspositively correlated with biomass.

Luckman et al 1997 made a study to quantify the relationship between radar backscatter and the biomassdensity of regenerating tropical forest using data acquired in the Tapajo’s region of the Amazon basin inBrazil. In that study the response of three space borne SAR instruments were examined and found out thatC-band SAR system was not suitable for monitoring biomass density in regenerating tropical forests but itmay be used only to discriminate between forested and non forested areas in dry conditions (Figure 10.)Whereas L-band system, either HH or HV polarization, may be used to derive information about biomassdensity up to a limit of approximately 60 tons/ha which is of below those quoted for coniferous forest (LeToan et al, 1992; Dobson et al., 1992) and seen to be in close agreement with other studies of broadleafevergreen forest such as Imhoff (1995), which estimated threshold values of 20 tons/ha for C-band and 40tons/ha for L-band SAR under similar conditions. The differences between the responses of the coniferousand tropical forest types are most likely to be due to the differences in canopy morphology and structure(Luckeman et al 1997).


Figure 9. Backscattering coefficient versus biomass Figure 10. Backscattering coefficient versusDensity for all L-band images studied. Sigmoid biomass density for all C-bandCurves have been fitted to the data by least images studied. Sigmoid curvessquares. have been fitted to the dry-season

ERS-1 data by least squares.

Summary

According to Evans et al. 1995 “the dependence of microwave backscatter on total above-ground biomasshas been documented in monospecific pine forests found in the southeastern U.S.and France (Hussin et al.,1991; Dobson et al., 1992; Kasischke et al., 1994a; LeToan et al., 1992), mixed deciduous and coniferousforests of Maine, northern Michigan, and Alaska (Ranson et al., 1994; Dobson et al., 1994; Harrell et al.,1995; Rignot et al., 1994), and coniferous forests of the Pacific Northwest (Moghaddam et al., 1994). Thesestudies all show the same results: (1) the sensitivity of microwave backscatter to biomass variations saturatesafter a certain level is reached; and (2) the biomass dependence of microwave backscatter varies as afunction of radar wavelength and polarization. In summary, the saturation point is higher for longerwavelengths, and the HV polarization is most sensitive and VV the least”.

A conclusion drawn by some scientists is these single-frequency saturation levels represent the upper limit ofSAR's ability to monitor changes or differences in aboveground biomass in forests (Waring et al., 1994).However, this conclusion overlooks several important considerations. Microwave backscatter is correlatedwith total biomass and various components of biomass (e.g., branch biomass, needle biomass, bole biomass)or other physical characteristics (e.g., tree height, basal area) (Dobson et al., 1995c; Hussin et al., 1991;Kasischke et al., 1994a). This should not be surprising, since we know that different biomass components oftrees are closely correlated. Since different radar frequencies and polarization combinations are sensitive todifferent layers of a forest canopy, it should be possible to use multiple channels of SAR data to estimatetotal above-ground biomass. Recent research supports this hypothesis.

Kasischke et al., (1994a) used a two-stage approach to estimate biomass of southern pine forests using JPLAIRSAR data. In step one, total branch biomass was estimated as a function of several different radarfrequencies/polarizations. Total biomass was then estimated from branch biomass based upon allometricequations, and resulted in a relative error on the order of 20% for biomass levels up to 400 t ha-1. Ranson etal., (1994) used a ratio of P-band HV (PHV) and C-band HV (CHV) to estimate total biomass (up to 250 tha-1) in mixed coniferous/deciduous forests in Maine. This technique was applied to SIR-C/X-SAR LHVand CHV data to estimate boreal forest biomass up to 200 tons/ha within +20 tons/ha (Ranson and Sun,1995). Finally, Dobson et al., (1995c) used a multi-step, semi-empirical approach to estimate abovegroundbiomass from a combination of channels from SIR-C data collected over a mixed coniferous/deciduousforest in northern Michigan. In this approach, different SAR frequency/polarization combinations were usedto estimate canopy-layer biomass, total height and total basal area, which were then used to estimate totalbiomass. Biomass estimates up to 250 t ha-1 with an uncertainty on the order of 16 t ha-1 were achieved.


Biomass conclusion

There is a strong correlation between biomass and radar backscatter using L and P bands, HV andVH polarisation, with both aircraft and satellite SAR imagery.

References: Christensen et al., 1990; Hoekman et. al., 1990; Hoffer et al., 1986; Hoffer et al, 1987b; Hussinet al., 1990; Hussin et al., 1991; Leckie et al., 1994; Lee and Hoffer, 1990a Lee and Hoffer, 1990b, Ransonand Sun, 1992; Reich and Hussin 1993; Stone and woodwell, 1988; Wu and Sader, 1987a; Wu, 1987b; Wu,1989; Dobson et al., 1991; Le Toan et al. 1991; Le Toan et al., 1991, Freeman et al. 1996; Van der Senden,1997; Hoekman et al, 1996; Freeman et al 1992; Luckeman et al 1997; Rignot et al, 1995; Rignot et al, 1997;Freeman et al, 1997; Pullianen et al 1994; Rauste et al, 1994; Kasischke, et. al 1995; Imhoff. 1995; Dobsonet al., 1994; Beaudoin et al,1994; Israelsson et al 1994; Sader et al, 1990; Dobson et al 1995; Wu, 1990;Ranson et al., 1994; Kasischke et al., 1994a; Moghaddam et al., 1994; Harrell et al., 1995; Waring et al.,1994; Dobson et al., 1995c; Ranson and Sun, 1995;

· In order to get efficient estimate of aboveground biomass of forest stands; the system of equationspresented by several researchers should be combined with an appropriate inventory design.

· Regression equations developed to estimate aboveground biomass forest stands should account for higher percentage of variability and lower percentage of bias and standard error.

· Inverse formulations seeking to retrieve estimate of above ground biomass is dependent upon the tree and/or forest structure.

· Over hilly terrain, correction of radar data-as wells as- the insitu biomass data – for surface topography is needed before data inversion.

· The L-band and P-band response are more appropriate than the C-band for estimation of different levels of biomass up to a certain threshold.

· The utility of radar backscatter for aboveground biomass estimation is going to be dependent on wavelength, frequency, incidence angle, backscatter mechanism of forests, topography, tree geometry, soil moisture content, etc.

· The sensitivity of microwave backscatter to biomass variations saturates after a certain level is reached.

· The biomass dependence of microwave backscatter varies as a function of radar wavelength and polarization.

· The biomass saturation point for longer wavelengths are larger than for shorter wavelengths, and “cross polarized” signals are more sensitive to biomass than “like polarized” signals.

· It is possible to use multiple channels of radar data to estimate aboveground biomass.

· The sensitivity of microwave backscatter to biomass variations saturates after a certain level is reached.

· The biomass dependence of microwave backscatter varies primarily as a function of radar wavelength and polarization and is also known to be dependent forest structural type, topography, incidence angle, weather and seasonal conditions etc.

· The biomass saturation point for longer wavelengths are larger than for shorter wavelength and ‘cross polarized’ signals are more sensitive to biomass than ‘like’polarized signals.

· It is possible to use multiple channels of radar data to estimate aboveground biomass.


· Aboveground biomass estimation is possible where net biomass is less than a frequency dependent saturation threshold.

· Aboveground biomass estimation is found more successful in mono-species forests such as pine plantations.

· In comparisons between P, L and C band data the P band HV polarisation showed the strongest correlation between radar backscatter and biomass incidence angle not mentioned). The L band HV polarisation (incidence angles between 30° - 40° is the next best combination) which also shows strong correlation between backscatter and biomass.

· Ratioing of C, L and P bands is also an effective way to map total biomass (aircraft data).

· Because of mixed species and variable ace classes in tropical crest regions. L band. HV and VH polarisation data are poorly correlated with forest biomass, although one study indicated that L- band HV and VH polarised aircraft at incidence angles of 33° - 55° data are related to bole and branch volumes.

· HV polarized. L-band data had better correlation with biomass and other stand parameters than any of the other polarisation. (HH or VV).

· L- band HH polarised satellite data does show a correlation between backscatter and biomass. using 343 stands of slash pine, correlation were found between biomass and backscatter at 28°, 45°. and 58° incidence angles (r: = 0.51-0.69).

· Like polarisation (HH and VV) does not show a strong correlation between backscatter and biomass, although it was found that ratios of VV/HH or VH/HH displayed significant correlation to biomass at incidence angles of 34-41 degrees.

· Young stands of slash pine (<17 years) showed good correlation between backscatter and biomass or other stand parameters, whereas older stands (18-48 year old plantations) had insignificant correlation (L-band. HV&VH polarised data). Above 100 tons ha the relationship between biomass and radar backscatter becomes asymptotic. This can be explained by attenuation of` the radar signal by the canopy in the higher biomass forests.

· L-band HH polarised SIR-B radar data combined with optical data such as TM imagery have produced higher correlation (em.. r = 0.90) than with either radar or TM data alone.

· Poor relationships exist between backscatter and biomass at X and C band.

Inventory Data

There is a strong correlation between forest stand parameters (e.g. tree height, DBH, trees/acre, basalarea, and age) and radar backscatter using L and X bands VH and (in some cases) HH polarisation,with both aircraft and satellite SAR imagery.

References: Colwell, 1983; Hussin and Hoffer, 1989; Hussin, et al., 1991; Hussin and Hoffer, 1992; Lee andHoffer, 1990a; Lee and Hoffer, 1990b; Hoffer and Hussin, 1989; Ford and Wickland 1985; Sieber, et al.,1987; Riom et al., 1980; Wu, 1987.

· To studies were found that used P-band imagery to study stand characteristics.

· Tree height DBH and trees/acre using 343 stands of slash pine were strongly correlated to radar backscatter of L-band HH polarised satellite data at incidence angles of 28°, 45° and 58° (r2 = 0.51-0.73).


· Basal area shows the highest correlation to radar backscatter using L-band, VH polarisation at incidence angles of 45°. HH or VV polarisation had very poor correlation with BA.

· With L-band aircraft data higher correlation between backscatter and tree height. DBH basal area or age were obtained with the HV polarisation than the HH or VV polarisation. A combination of all three polarisation gave the highest correlation. The HH polarised data was the most poorly correlated. An incidence angle of 45° gave the highest correlation with all stand parameters.

· Tree height shows correlation using X and L-band data with VH and HH polarisation. VV polarisation showed no correlation. The combined use ot VH/HH ratios significantly improved the tree height estimates. The studies by Lee and Hoffer, 1990 a & b were the only articles that mentioned incidence angle, that being 45° for maximum correlation. (r2 = 0.73)

· Age was strongly correlated ( r = 0.80**) to L-band HH polarised satellite data at incidence angles of 28°, 45o, and 58°.

· Older pine stands could be effectively identified and distinguished from mediumaged stands due to the higher backscatter at 58° (versus 45° or 28° ) incidence angle using L-band, HH polarised SIR-B data.

Stress conditions

Identification of plant stress has been indicated using X and C-band imagery with aircraft data on a fewoccasions but has not been proven definitively.References: Ulaby et al., 1981; Werle 1989.

· Statistical analysis in one study showed that in some tree species and broad categories of tree ages and areasof diseased woodland could be identified.

Optimum System Parameters

According to Evans et al. 1995, “the SAR parameters which define the utility of a specific system forecological applications are its frequency, polarization, angle, resolution and sampling frequency. Thecommon frequencies used today include P-band, L-band, S- band, C-band and X-band. Radar typicallytransmit horizontally or vertically-polarized microwave energy and can receive either polarization, resultingin four linear polarizations - HH, HV, VH, and VV. Today's spaceborne SAR systems usually have a fixedcenter angle between 20 degrees and 50 degrees with images covering a few degrees from near edge to faredge. Future systems will operate in a SCANSAR mode, with image swaths covering a 20 degrees to 50degrees range in angles. Today's spaceborne SARs have fairly fine resolution (20 to 40 m), narrow swathwidths (60 to 100 km), and long sampling frequencies (20 to 40 days). SCANSAR systems will have theability to cover wide areas at lower resolutions (up to 500-km swaths with 100 to 200 m resolution) andhigher sampling frequencies (every 2 to 4 days)”.

Utility of Existing/Planned SAR Systems

Evans et al. (1995) reported that five spaceborne imaging radar systems are now in operation or weredeployed during the last years: ERS-1, ERS-2 JERS-1, Radarsat and SIR-C/X-SAR. The ERS-1 or ERS-2SAR is a C- band VV (vertical transmit polarization/vertical receive polarization) launched in the summer of1991. This system has a 25 m resolution and 100 km swath. The orbit of this system is tailored such that ithas a 35-day exact repeat orbit during the northern hemisphere summer and fall (which means it can imagethe same ground location every 18 days or so), and a 3- day exact repeat orbit during the winter and spring inorder to obtain frequent coverage of the polar ice cap. Coverage is limited to locations where groundreceiving stations are installed. The JERS-1 SAR is an L-band HH system launched during the summer of1992. It has a 18 meter resolution and a 75-km swath. The exact repeat orbit on this system is 44 days.


Onboard recording provides global access. The Canadian RADARSAT is consist of a C-band SAR with HH-polarization and has a variety of modes for resolution/swath width. The SCANSAR mode can yield swathwidths up to 500 km with a spatial resolution of 100 m. This wide swath mode will allow imaging of thesame geographic location once every 2 to 3 days. The SIR-C/X-SAR system was flown onboard NASA'sSpace Shuttle on two ten day missions in April and October of 1994. This system consisted of a C- and L-band SAR system that was fully polarimetric (i.e., it collected HH, HV, VH, and VV imagery) and an X-band SAR which collected VV data. The resolution of this system ranged between 10 and 40 m, and itcollected image swaths between 15 and 90 km wide. The ground coverage of this system was limited inorder to image specific test sites during its two missions.”

Three spaceborne SAR systems are planned for the future:, RADARSAT-2, LightSAR and ASAR. TheAdvanced SAR or ASAR, will be deployed on the planned European ENVISAT and consists of a dual-polarized C-band SAR. It will have both co-polarized channels (HH and VV), but not cross-polarized (HV).The planned "wide swath" mode has a 400- km swath width and 100-m resolution.

3.3.2 Literature Cited

Ahern, F. J., J. A. Drieman. 1988. Assessment of Clear-Cut Mapping Accuracy With C-BandSAR. Proceedings IGARSS '88 Symposium, Edinburgh, Scotland. pp.1335-1338.

Ahern, F.J., D.G. Leckie, and D. Werle. 1993a. Applications of Radarsat SAR Data in ForestedEnviromnents. Canadian Joumal of Remote Sensing: 19(4):330-337.

Ahern, F. J, D. G. Leckie, and J. A. Drieman. 1993b. Seasonal Changes in Relative C-Band Backscatter ofNorthem Forest Cover Types. IEEE Trans. Geoscience and Remote Sensing. 31(3):669-680.

Ahern, F.J. 1993c. The Information Content of Multipolarized, Multi-Season C-SAR Data of the BorealForest: A Preliminary Assessment. Proc. 16th Canadian Svmp. Remote Sensing. Pp. 387-392.

Ahmad, W.Y., F. MacCallum, R.Y. Dams, S.B. Mokhtar, H.G. Cheah. 1988. Landsat MSS, SPOT and SARfor Tropical Forestry Applications in Malaysia.

Ahmad, Z., and J.A. Richards. 1989. Multiple Incidence Angle SIR-B Forest Observations. IEEETransactions on Geoscience and Remote Sensing , 27(5):586-591.

Ali, A.E. 1987. Topographic Mapping of Vegetation Features from SEASAT SAR Imagery:An Experiment from the Firth of Tay. Scottish Forest. 41(4):273-281.

Aselmann, I. and P. J. Crutzen (1989). Global distribution of natural freshwater wetlands and rice paddies,their net primary productivity, seasonality, and possible methane emissions. J. Atmos. Chem., vol. 8, pp.307-358.

Avery, T.E. and G. L. Berlin. 1985. Interpretation of Aerial Photographs (4th ed.) BurgessPublishing, Co., Minneapolis, MN. Pp. 168-212.

Bacastow, R. B., C. D. Keeling, and T. P. Whorf (1985). Seasonal amplitude increase in atmospheric CO2concentration at Mauna Loa, Hawaii, 1959-1982. J. Geophys. Res., vol. 90, pp. 10,529-10,540.

Bartlett, K. B. and R. C. Harriss (1993). Review and assessment of methane emissions from wetlands.Chemosphere, vol. 26, pp. 261-320.

Beaudein, A.T, Le Toan, S. Groze, E. Nezry, A. Lopes, 1994.’Retrieval of forest biomass from SAR data’.International Journal of Remote Sensing, vol.15, no.414, pp2796.


Beaudoin, A., T. LeToan, S. Goze, E. Nezry, A. Lopes, E. Mougin, C. C. Hsu, H. C. Han, J. A. Kong, and R.T. Shin (1994). Retrieval of forest biomass from SAR data. Int. J. Rem. Sens., vol. 15, pp. 2777-2796.

Beaudoin, A., T. Le Toan, E. Mougin, A. Lopes, D.H. Hoekman, and M. Wooding. 1991.Interpretation of the Microwave Backscatter From Pine Forests. Proc. Int. Geoscience andRemote Sensing Symp '91. Pp. 1511-1514.

Bonan, G. B. (1991a). Atmosphere-biosphere exchange of carbon dioxide in boreal forests. J. Geophys. Res.,vol. 96, pp. 7301-7312.

Bonan, G. B. (1991b). Seasonal and annual carbon fluxes in a boreal forest landscape. J. Geophys. Res., vol.96, pp. 17329-17338.

Bonan, G. B. and H. H. Shugart (1989). Environmental factors and ecological processes in boreal forests.Ann. Rev. Ecol. Syst., vol. 20, pp. 1-28.

Brisco, B., F.T. Ulaby, and M.C. Dobson. 1983. Spacebome SAR Data for Land-Cover Classification andChange Detection. Proc. Int. Geoscience and Remote Sensing Sym. '83. Pp. 1.1-1.8.

Brown, L. M., Veck, N. J. and Miller, R J. (1990). Preliminary Results from a High Resolution Forest ImagingModel For SAR. IGARSS '90, Vol. II, College Park, Maryland, p. 1199-1201.

Brown, R. J., B. Brisco, F. Ahem, S. Yatabe, and J. Drieman. 1992. Preliminary ERS-1 Assessment forCanadian Agriculture and Forestry Applications. Proc. Ist ERS-1 Symp., Cannes, France. pp.611-616.

Bryan, L. M., and J. Clark. 1984. Potentials for Change Detection Using SEASAT Synthetic Aperture RadarData. Remote Sensing of Environinent. 16:107-124.

Bullock, B. L. 1987. Radar Applications in Remote Sensing: An Airbome Remote Sensing Case History.Proc. 21st Int. Symp. on Remote Sensing of Envirownent5 Ann Arbor, MI. pp.617-619

Ceteno, R. L. 1973. Radar and Its Application on Natural Resources Research in Venezuela.Ist Panainerican Symp on Remote Sensing, Panaina City. pp.217-229.

Christensen, D.J., E.S. Kasischke, and M.C. Dobson. 1990. SAR-Derived Estimates of Above GroundBiomass in Forested Landscapes. Proc. Int. Geoscience and Remote Sensing Symp. '90. Pp. 1209-1212.

Chauhan, N. R., R. H. Lang, and K. J. Ranson (1991). Radar modeling of a boreal forest. IEEE Trans.Geosci. Rem. Sens., vol. 29, pp. 627-638.

Cimino, J., A. Brandani, D. J. Casey, J. Rabassa, and S. D. Wall. 1986. Multiple Incidence Angle SIR-BExperiment Over Argentina: Mapping of Forest Units. IEEE Trans. Geoscience and Remote Sensing. GE-24(4):498-510.

Colwell, R. N. (editor-in-chief). 1983. Manual of Remote Sensing. Vols. 1 & 2. American Society ofPhotogrammetry, Falls Church, VA.

Crandall, C. J. 1969. Radar Mapping in Panama. Photogrammetric Engineering. 35(6):641646.

Dams, R.V., D. Flett, M.D. Thompson, and M. Lieberman. 1987. SAR (STAR-1) Image Analysis for CostaRican Tropical Forestry Applications. SELPER Conference, Bogota, Colombia.

Dams, R.V., R.J. Hall, and F.J. Ahem. 1990. Forest Cutover Mapping from X-Band SAR. Canadian J.Remote Sensing. 16(2):2-7.


D'Arrigo, R., G. C. Jacoby, and I. Y. Fung (1987). Boreal forests and atmosphere biosphere exchange ofcarbon dioxide. Nature, vol. 329, pp. 321-323.

Daus, S. J. and Lauer, D. T. (1971). Testing the Usefulness of Side Looking Airborne for Evaluating ForestVegetation Resources, Final Report. The University of Kansas.

Dellwig, L. F., J. E. Bare, and R. H. Gelnett. 1978. SLAR--For Clear as Well as Cloudy Weather. Proc. Int.Symp. Remote Sensing for Observation and Inventory of Earth Resources and the Endangered Environment.pp. 1527-1546.

de Grandi, G. F., G. C. Lemoine, H. de Groof, C. Lavalle, and A. J. Sieber (1994). Fully polarimetricclassification of the Black Forest MAESTRO 1 AIRSAR data. Int. J. Rem. Sens., vol. 15, pp. 2755-2775.

de Molina, I., F. L. Mosquera, C. Molina, and D. Deagostini. 1973. SLAR in the Mapping of HumidTropical Forests of Colombia. First Panamerican Symp. on Remote Sensors. Panama. pp.231-272.

De Wulf, RR, R.E Groosens, B.P De rooves, and F.C Borry, 1990. “Extraction of forest Stand parametersfrom Panchromatic and multispectral spot-1 data”. International Remote Sensing, Vol 11,no.9, pp.1571-1588.

Dixon, R. K., S. Brown, R. A. Houghton, A. M. Solomon, M. C. Trexler, and J. Wisniewski (1994). Carbonpools and flux of global forest ecosystems. Science, vol. 263, pp. 185-190.

Dobson, M.C., F.T Ulaby, T. Le Toan, A. Beaudoin, E.E. Kasischke and M. Christensen, 1992.“Dependence of radar backscatter on conifer forest Biomass”. IEEE Transactions on Geoscience and RemoteSensing, Vol. 30, No.2, pp. 412-415.

Dobson, MC., Fit. Kendra, E.Li, Y.C. Lin, A. Nashashibi, Kamal Sarabandi, P. Siqueira, 1995.’Estimation offorest biophysical characteristics in Northern Michigan with SIR-C/X-SAR.’IEEE Transactions onGeoscience and Remote Sensing, vol.33, no.4, pp 893.

Dobson, M. C., L. E. Pierce, K. Sarabandi, F. T. Ulaby, and T. Sharik (1992a). Preliminary analysis of ERS-1 SAR for forest ecosystem studies. IEEE Trans. Geosci. Rem. Sens., vol. 30, No. 2, pp. 203-211.

Dobson, M. C., F. T. Ulaby, and L. E. Pierce (1995a). Land-cover classification and estimation of terrainattributes using synthetic aperture radar. Rem. Sens. Environ., vol. 51, No. 1, pp. 199-214.

Dobson, M. C., L. E. Pierce, and F. T. Ulaby (1995b). Knowledge-based land-cover classification usingERS-1/JERS-1 SAR composites. IEEE Trans. Geosci. Rem. Sens., (in press).

Dobson, M. C., F. T. Ulaby, L. E. Pierce, K. Bergen, and K. Sarabandi (1995c). Land-cover classificationand biomass estimation with SIR-C/X-SAR. IEEE Trans. Geosci. Rem. Sens. (in press).

Domik. G.. F. Leberl. and J. Cimino. 1988. Dependence of Image Grey scale texture on Topography inSIR-B Images. Int. J. Remote Sensing:101-1O8.

Drieman, J. A. (1994). Forest cover typing and clearcut mapping in Newfoundland with C- band SAR. Can.J. Rem. Sens., vol. 20, pp. 11-16.

Drieman. J.A.. D.G. LecKie. and F.J. .Ahern. .989a. Multi-Temporal C-SAR for Forest Typing in EasternOntario. Proc. Int. Geoscience and Remote Sensing Symp. 89. Pp. 1376-1378.

Drieman. J. A.. F. J. Ahern. and I. G.Corns. 1989b. Visual Interpretation Results of Multipolarization C-SARImagery of Alberta Boreal Forest. Proc. Int. Geoscience and Remote Sensing Symp. 89. pp. 1401-1405.


Drieman. J.A. 1992. Clearcut Clapping Keith C-Band SAR. Newfoundland Forest Service RDDPProject. Dept. Forestry and Agriculture. Corner Brook. Newfoundland. Canada. 4 p.

Durden, S. L., J. J. Van Zyl, and H. A. Zebker (1989). Modeling and observation of the radar polarizationsignature of forested areas. IEEE Trans. Geosci. Rem. Sens., vol. 27, pp. 290-301.

Durden. S. L.. J. D. Klein. and H. A. Zebker. 1991. Polarimetric Radar Measurements of Forested Area NearMt. Shasta. IEEE Transactions on Geoscience and Remote Sensing. 29(3):444-450.

Evans. D. Lo T. G. Farr. J. J. van Sol. and H. A. Zebker. 1988. Radar Polarimetry: Analysis Tools andApplications. IEEE Transactions on Geoscience and Remote Sensing. 26(6):774-789.

Foody, G. M., M. B. McCulloch, and W. B. Yates (1994). Crop classification from C-band polarimetricradar data. Int. J. Rem. Sens., vol. 15, pp. 2871-2885.

Ford. J. P.. J. B. Cimino. and C. Elachi. 1983. Space Shuttle Columbia Views the World With ImagingRadar: the SIR-A Experiment. JPL Publication 82-95, Jet Propulsion Laboratory, Pasadena CA. 179 pp.

Ford. J.P.. J.B. Cimino. B. Holt. and M.R. Ruzek. 1986. Shuttle Imaging Radar Views the Earth FromChallenger: The SIR-B Experiment. JPL Publication 86-10. NASA Jet Propulsion Laboratory. Pasadena CA.135 pp.

Ford. J. P.. and D. J. Casey. 1988. Shuttle Radar Mapping With Diverse Incidence Angles in the Rain Forestof Borneo. Int. J. Remote Sensing 9(5):927-943.

Ford, J. P. and Wickland, D. E. (1985). Forest Descrimination with Multipolarization Imaging Radar. IGARSS'85. Vol 1, pp. 462-465.

Forster, B. C. (1992). Radar Remote Sensing. Paper presented in course developed for BHP Engineering asConsultants to AIDAB on the Philippine Remote Sensing Project. University of the Philippines.

Freeman, A., S. Durden and R. Zimmerman, 1992. “Mapping Sub-tropical vegetation using multi-frequency,multipolarization SAR data, In: Proceedings of IGARESS’92, International space year. Space RemoteSensing Housten, 26-29 May 1992, pp 1686-1689.

Freeman, T. 1996.’What is Imaging Radar? http://Southport;jpl.nasa.gov./desc/imaging radar3.html.

Fung, I., J. John, J. Lerner, E. Matthews, M. Prather, L. Steele, and P. J. Fraser (1991). Three-dimensionalmodel synthesis of the global methane cycle. J. Geophys. Res., vol. 96, pp. 13,033-13,065.

Furlev. P. A. 1986. Radar Survevs For Resource Evaluation in Brazil: An Illustration From Rondonia. In I. J.Eden. and J. T. Parrv (eds.). Remote Sensing and Tropical Land Management. John Wiley and Sons. NewYork. pp. 79-99.

Gelnett. R. H.. L. F. Dellwi~, and J. E. Bare. 1978. Increased Visibility from the Invisible, A Comparison ofRadar and LANDSAT in Tropical Environments. Proc. Itch Int. Symp. on Remote Sensing ot Environrnent.pp.2205-2216.

Hardy. N. E.. J. C. Coiner. W. O. Lockrnan. 1971. Vegetation Mapping With Side-Looking AirborneRadar:Yellowstone National Park. Conk on Prorogation Limitations in Remote Sensing pp.11.1 - 11.19.

Harrell, P., L. L. Bourgeau-Chavez, E. S. Kasischke, N. H. F. French, and N. L. Christensen (1995).Sensitivity of ERS-1 and JERS-1 radar data to biomass and stand structure in Alaskan boreal forest. Rem.Sens. Environ. (in press).


Haralick, R. M., Shanmugan, K. and Distein, I. (1973). Textural features for image classification. IEEETransactions on Systems, Man and Cybernetics.Vol. SMC-3, pp. 610-621.

Heiska K., SILT. Hallikainen. and J. Hyyppa. 1992. Backscattering Behaviour of Forests at C-and X-BandUnder Winter Conditions. Proc. Int. Geoscience and Remote Sensing Svmp. 92. Pp. 536-538.

Hess, L. L. and J. M. Melack (1994). Mapping wetland hydrology with synthetic aperture radar. Int. J. Ecol.Environ. Sci., vol. 20, pp. 74-81.

Hess, L. L., J. M. Melack, and D.S. Simonett (1990). Radar detection of flooding beneath the forest canopy:a review. Int. J. Rem. Sens., vol. 5, pp. 1313-1325.

Hess, L. L., J. M. Melack, S. Filoso, and Y. Wang (1995). Delineation of inundated area and vegetationalong the Amazon floodplain with the SIR-C synthetic aperture radar. IEEE Trans. Geosci. Rem. Sens. (inpress).

Hirosawa. H.. A. Matsuzaka. and O. Kobayashi. 1989. Measurement of Microwave Backscatter From a(cypress With and Without Leaves. IEEE Transactions on Geoscience and Remote Sensing. 27(6):698-701.

Hoekman, D. H. (1985). Radar Backscattering of Forest Stands. International Journal of Remote Sensing, Vol 6.pp. 325-343.

Hoekman. D.H. 1987. Measurements of the Backscatter and Attenuation Properties of Forest Stands at X-.C-, and L-Band. Remote Sensing of Environment. 23:397-416.

Hoekman. D. H. 1990. Radar Remote Sensing Data For Applications in Forestry. Ph.D. Thesis, WageningenAgricultural University, The Netherlands. 279 pp.

Hoekman, D. H., J. J. van der Sanden. and G. J. Rijckenberg. 1992. Analysis of Forest Radar BackscatterUsing Airborne Polarimetric Data and Multi-Temporal ERS-I SAR Data. Proc. Int. Geoscience and RemoteSensing Symp. '92. pp. 1209-1211.

Hoekman, D.H., J.J.Van der Sanden and W.Bijker, 1996. Radar remote sensing of tropical rain forests: TheAIRSAR-93 campaign in Guyana and Columbia. Delft (BCRC, Netherlands Remote Sensing Board), BCRSreport no. 96-02,59 p.

Hoffer. R.M. 1984. Microwave and Optical Remote Sensing of Forest Vegetation. The SIRB ScienceInvestigations Plan. Jet Propulsion Lab. Pub. 84-3, Pasadena CA Pp. 4.71 -4.73.

Hoffer. R.M., S.E. Davidson. P.W. Mueller and D.F. Lozano-Garcia. 1985a. A Comparison of X- andL-Band Radar Data for Discriminating Forest Cover Types. Proceedings of the Tenth William T. PecoraMemorial Remote Sensing Symposiurn, Ft. Collins. CO. Pp. 439-440.

Hoffer. R.M., P.W. Mueller and D.F. Lozano-Garcia. t985b. Multiple Incidence Angle Shuttle ImagingRadar Data for Discriminating Forest Cover Types. Proceedings of the Fall Convention of the AmericanSociety of Photogrammetry and Remote Sensing, Indianapolis. IN. Pp. 476-485.

Hoffer. R.M.. D.F. Lozano-Garcia. D.D. Gillespie. P.Crueller and M.J. Ruzek. 1986a. Analysis of MultipleIncidence Angie SIR-B Data for Determining Forest Stand Characteristics. Proceedings or` the SecondSpaceborne Imagine Radar Symposium. Jet Propulsion Laboratory . Pasadena. CA. Pp. 159- 164.

Hoffer. R. M. D.F. Lozano-Garcia and D.D. Gillespie. 1986b. Mappine Forest Cover With SIR-B Data.Technical Papers of the Fall Convention of the American Society for Photograrnmetry and Remote Sensing,Anchorage. OK. Pp. 407-415.


Hoffer. R.M. and D.F. Lozano-Garcia. 1987a. Definition of Forest Stand Characteristics Based onMulti-lncidence Angle SIR-B Data. Technical Papers of the Annual Convention of the American Society forPhotogrammetry and Remote Sensing, Baltimore. MD.

Hoffer. R.M. D.F. Lozano-Garcia and D.D. Gillespie. 1987b. Characterizine Forest Stands WithMulti-incidence Angle and Multi-Polarized SAR Data. Advances in Space Research,

Hoffer. RM. and K.S. Lee. 1988a. Change Detection Techniques to Assess Deforestation Using Seasat andSIR-B Satellite Radar Data. Invited paper. Joint Technical Sessions of IUFRO (International Union of ForestResearch Oreanization and ISPRS (International Society for Photogrammetry and Remote Sensing), Kyoto,Japan. 8 pp.

Hoffer. R.M. ress of the International Societv for Photo~ramrnetrv and Remote Sensing, Kyoto, Japand D.F.Lozano-Garcia 1988b. Information Content of Combined Radar and Multispectral Scanner Data Obtainedfrom Satellite Altitude. 16th International Congan. 10 pp.

Hoffer. R..M. and D.F. Lozano-Garcia. 1988c. Spaceborne Radar for Mapping Forested Wetlands. TheInternational Forested Wetlands Resource: Identification and Inventory Abstracts Baton Rouge. Louisiana.pp. 21-22.

Hoffer. R.M. and Y.A. Hussin. 1989a. Per-point and Per-field Contextual Classification of Multipolarizationand Multiple Incidence Angle Aircraft L-Band Radar Data. Proceedings of the Fall Convention of theAmerican Society for Photogrammetry and Remote Sensing, Cleveland. OH. Pp. 154-170.

Hoffer. R.M. and K.S. Lee. 1989b. Forest Change Classification Using SEASAT and SIR-B Satellite SARData. Proceedings of IGARSS ' 89 Symposium. Vancouver. B.C.. Canada. Vol. 3:1372-1375.

IIoffer. Rut. 1990a. Remote Sensing for Forest Assessment and Global Deforestation. BAAS 1990 AnnualMeeting Abstracts. American Association of the Advancement of Science. Washington. DC. p. 90.

Hoffer. R.M.. 1990b. Satellite Radar for Monitoring Resources. Space Conference of the Americas, SanJose. Costa Rica. Sol. I. pp. 232-238.

Hoofer. R. M. and K.S. Lee. 1990c. Satellite Remote Sensing for Assessing Forest Resources. PlenarySession. Proceedings of the 5th Australasian Remote Sensing Conference, Perth. Australia. Pp. 88-97

Hoffer. R.M. and K.S. Lee. 1990d. Change Detection of Forest Cover Using Satellite Radar Data. In:Protecting Natural Resources With Remote Sensing, Proceedings of the Third Forest Service RemoteSensing Applications Conference. American Society for Photogrammetry and Remote Sensing. Bethesda.MD. Pp. 349-355.

Hoffer. R.M.. Y.A. Hussin. K.S. Lee. D. F. Lozano. and P.W. Mueller. 1994. "Microwave and OptimalRemote Sensing of Forest Vegetation." Final Report: California Institute of Technology, Jet PropulsionLaboratory, Pasadena. CA. 30 pp.

Holecz. F.. E. Meier. and D. Nuesch. 1992. Classification or' SAR Images In Areas with Severe TerrainVariations. Proc. Int. Geoscience and Remote Sensing Symp. 92. pp. 1123-1125.

Holmes. G. M. 1992. Monitoring Vegetation in the Future: Radar. Botanical Journal of the Linnean Society.180:93-109

Houghton, R. A. (1987). Biotic changes consistent with the increased seasonal amplitude of atmosphericCO2 concentration. J. Geophys. Res., vol. 92, pp. 4223-4230.

Houghton, R. A., R. D. Boone, J. M. Melillo, C. A. Palm, G. M. Woodwell, N. Myers, B. Moore, and D. L.Skole (1985). Net flux of carbon dioxide from tropical forests in 1980. Nature, vol. 316, pp. 617-620.


Howard, J. A. (1991). Remote Sensing of Forest Resources, Theory and application.Chapman and Hall.

Hsu, C. C., H. C. Han, R. T. Shin, J. A. Kong, A. Beaudoin, and T. LeToan (1994). Radiative transfer theoryfor polarimetric remote sensing of pine forest at P band. Int. J. Rem. Sens., vol. 15, pp. 2943-2954.

Huch, B., C.I. Miller, and T.W. Beers, 1982. Forest Mensutation. New York etc. (John Wiley & sons, Inc),3rd edition, 402 p.

Hussin, Y. A, R.M. Reich and R.M. Hoffer, 1992. ‘Effects of Polarization on radar Backscatter in Relation toSlash Pine Stand Biomass Using Aircraft and SIR-B Data’. Reprinted from: ISPRS 1992: Vol. 29: Part B7,commission VII, pp. 661-667

Hussin. Y.A. and R.M. Hoffer. 1989. Relationship Between Multipolarized Radar Backscatter and SlashPine Stand Parameters. Proceedings of the Fall Convention of the American Society for Photogrammetry andRemote Sensing, Cleveland, OH. Pp. 1 84-196.

Hussin. Y. An R. M. Hoffer. 1990a. Swamp Forest Differentiation with Multipolarized and MultipleIncidence Angle L-band Radar Data. Proceedings IGARSS '90 Symposium, Washington, D.C., pp.573-576.

Hussin, Y.A. and R.M. Hoffer. 1990b. Effect of Young Slash Pine Bedding Direction on Backscatter ofMultipolarized and Multiple Incidence Angle Radar Data. Technical Pagers. 1990 ACSM-ASPRS AnnualConvention, American Society for Photogrammetry and Remote Sensing, Bethesda. MD. 4:198-207.

Hussin. MA.. R.M. Reich. and R.M. Hoffer. 1992. Effect of Polarization on Radar Backscatter in Relation toSlash Pine Stand Biomass Using Aircrarft and SIR-B Data. Proceedings of the ASPRS-ACSM/RT 92National Convention on Mapping and Monitoring Global Change. Washington. DC. Vol: V: 532-541.

Hussin, Y. A, R.M Reich and R.M. Hoffer. 1991. ‘Estimating slash pine biomass using radar backscatter:IEEE Trans Geoscience and Remote sensing, 29, 427-431.

Hussin, Y.A, 1990. ‘The Effects of Polarization and Incidence angle on Radar backscatter from forestCover’. PhD Dissertation. Colorado State University, 187 pp.

Imhoff, M.L. 1995.’ A theoretical analysis of the effect of forest structure on synthetic aperture radarbackscatter and the remote sensing of biomass’. IEEE Transactions on Geoscience and Remote Sensing, Vol.33, no. 2,pp.341-352.

Imhoff. M. L.. M. Story. C. Vermillion. F. Khan. and F. Polcvn. 1986. Forest Canopy Characterization andVegetation Penetration Assessment with Space-Borne Radar. IEEE Transactions on Geoscience and RemoteSensing. 24(4):535-542.

Imhoff. Ivy. L.. C. Vermillion. .M. H. Story. A. M. Choudhurv. A. Gafoor. and F. Polcyn. 1987. MonsoonFlood Boundar! Delineation and Damage Assessment using Space Borne Imaging Radar and LANDSATData. Photogrammetric Engineering and Remote Sensing 53(4):405-413

Imhoff M L. 1988a. Flood Boundary Delineation Through Clouds and Vegetation Using L Band SpaceborneRadar: A Potential O'er Tool Disease Vector Control Programs.

Imhoff. M.L. and D. B. Gesch. 1988b. The Derivation or Sub-Canopy Surface Terrain odels of CoastalForests Using Synthetic Aperture Radar. Proceedings of IGARSS 88 Svmposium, Edinburgh. Scotland.pp.613-617.

Inkster, D. R., Lowry, R. T., Thompson, M. D. (1980). Optimum Radar Resolution Studies for Land Use andForestry Applications. Proceedings of the 14th International Symposium of Remote Sensing of Environment,Vol. 2, pp. 865-882.


Israelson, J. Askene and R. Sylvander, 1994.’Potential of SAR for forest bole volume estimation.International Journal of Remote Sensing, vol.15, no.14, pp.2809-2826.

Jano. A.P. 1990. Forest Depletion Monitoring with C-Band Synthetic Aperture Radar. Proc. 3rd ForestService Remote Sensing Application Conference. Pp. 299-306.

Kasischke. E.. L. Borgeau-Chavez N. French, P. Harrell.and N. Christensen. 1992. Initial Observations onUsing SAR to Monitor Wildfire Scars in Boreal Forests. Int. J. Remote Sensing. 13(18): 3495-3501.

Kasischke, E.S., N.L. Christensen, 1990.’Connecting forest ecosystem and microwave backscattermodels.’International Journal of Remote Sensing, vol.11, no.7, pp 1277-1298.

Kasischke, E.S., 1995.’Correlating radar backscatter with components of biomass in loblolly pine forests’.IEEE Transaction Geoscience and Remote Sensing, vol.33, pp 643.

Kasischke, E. S. and N. H. F. French (1994). Locating and estimating the areal extent of wildfires in Alaskanboreal forests using multiple season AVHRR NDVI composite data. Rem. Sens. Environ., vol. 50 (in press).

Kasischke, E. S., L. L. Bourgeau-Chavez, N. H. F. French, P. Harrell, and N. L. Christensen, Jr. (1992).Initial observations on the use of SAR imagery to monitor wildfires in boreal forests. Intern. J. Rem. Sens.,vol. 13, pp. 3495-3501.

Kasischke, E. S., N. L. Christensen, Jr., and L. L. Bourgeau-Chavez (1994a). Correlating radar backscatterwith components of biomass in loblolly pine forests. IEEE Trans. Geosci. Rem. Sens., vol. 32 (in press).

Kasischke, E. S., N. L. Christensen, Jr., E. Haney, and L. L. Bourgeau-Chavez (1994b). Observations on thesensitivity of ERS-1 SAR image intensity to changes in aboveground biomass in young loblolly pine forests.Int. J. Rem. Sens., vol. 15, pp. 3-16.

Kasischke, E. S., L. L. Bourgeau-Chavez, and N. H. F. French (1994c). Observations of variations in ERS-1SAR image intensity associated with forest fires in Alaska. IEEE Trans. Geosci. Rem. Sens., vol. 32, pp.206-210.

Kasischke, E. S., L. Morrissey, J. B. Way, N. H. F. French, L. L. Bourgeau-Chavez, E. Rignot, J. A. Stearn,and G. P. Livingston (1995a). Monitoring seasonal variations in boreal ecosystems using multi-temporalspaceborne SAR data. Can. J. Rem. Sens. (in press).

Kasischke, E. S., N. L. Christensen, Jr., and B. J. Stocks (1995b). Fire, global warming and the mass balanceof carbon in boreal forests. Ecol. Appl. (in press).

Kneppeck. I. D. & F. J. Ahern 1989. Stratification of Regeneration Burned Forest in Alberta UsingThematic Mapper and C-SAR Images. Proceedings IGARSS 89 / 12th Canadian Symposium of RemoteSensing, Vancouver. Canada Vol. 3. pp. 1391-1396.

Knowlton, D.J. and R.M. Hoffer. 1981. Radar Imagery for Forest Cover Mapping. Proceedings of theSeventh International Symposium on Machine Processing of Remotely Sensed Data. Purdue University.West Lafayette. IN. Vol 4. Pp. 626-632.

Kremmers,T. 1995. ‘ERS-1 SAR Imagery and GIS Techniques for Forest Monitoring in Central Europe- AFeasibility Study. http://www.metta.fil/cgi-glimpse/mfs/02/d4 pap114.htm? 12#mfs.

Kuga, Y., M. W. Whitt, K. C. McDonald, and F. T. Ulaby (1990). Scattering models for distributed targets,in Radar Polarimetry for Geoscience Application (F. T. Ulaby and C. Elachi, eds.), Artech, Dedham, MA.

Kwok, R., E. Rignot, J. B. Way, A. Freeman, and J. Holt (1994). Polarization signatures of frozen andthawed forests of varying environmental state. IEEE Trans. Geosci. Rem. Sens., vol. 32, pp. 371-381.


Lamb, A. D.. N. H. Mönnig, and D. W. van der Zel. 1986. Side-looking Radar - A Remote SensingApplication in South African Forestry Plantations. South African Forestry Journal. 139:54-59.

Land Cover Classification of the Philipines. (1988). Department of Environment and Natural Resources - AsianDevelopment Bank - Swedish Space Corporation. DENR, Quezon City, Philippines.

Lang, R. H. and Chauhan, N. (1990). Radar Sensitivity to Forest and Stand Parameters. IGARSS '90, Vol. II,College Park, Maryland, p. 481.

Lang, R. H., N. S. Chauhan, J. K. Ranson, and O. Kilic (1994). Modeling P-band SAR returns from a redpine stand. Rem. Sens. Environ., vol. 47, pp. 132-141.

Leberl. W.F., G. Domik. J. Taggam. and M. Kobric};. 1986. Radar Stereomapping Techniques andApplication to SIR-B Images of Mt. Shasta. FETE Transactions on Geoscience and Remote Sensing24(4):473- 481.

Leckie. O.G. 1983. Preliminary Results of an Examination of C-Band Synthetic Aperture Radar for ForestryApplications. Proc. 8th Canadian Symp. on Remote Sensing. Pp. 15l – 164.

Leckie. D.G.. KJ. Ransom. and G. Sun. 1994. Chapter 8 "Forestry Applications.'' In: Manual of RemoteSensing: Radar Volumes American Society for Photogrammetry and Remote Sensing. Bethesda.

Lee. K.S. and R.M. Hoffer. 1988. Assessment of Forest Cover Changes Using Multidate SpacebomeImaging Radar. Technical Papers of the 1988 Annual Convention of the American Congress on Surveyingand Mapping and the American Society for Photogrammetry and Remote Sensing, St. Louis. MO, Vol.4:198-211.

Lee. K.S. and R.M. Hoffer. 1989. Speckle Noise Filtering of Satellite Radar Data. Proceedings of ImageProcessing 89 Symposium. Reno. Nevada. 213-222.

Lee. K. S. and R. M. Hoffer. 1990a. Shuttle Imaging Radar (SIR-B) Backscatter Relationships With ForestStand Characteristics. Technical Papers. 1900 ACSM-ASPRS Annual Convention, American Society forPhotogrammetry and Remote Sensing, Bethesda, MD. Vol. 4:270-282.

Lee. K.S. and R.M. Hoffer. 1990b. Analysis of Combined SIR-B and TM Data for Assessing ForestBiomass. Proceeding of the 10th Annual International Geoscience and Remote Sensing Symposium, Vol. II:1227-1230

Lemoine, G.G., D.H. Hoekman, H.J.C. van Leeuwen. 1991. Separability of Agricultural Fields and ForestStands Using Multi-Frequency Polarimetric SAR Data of the Felvoland Site. IEEE Transactions onGeoscience and Remote Sensing.

Lemoine, G. G., G. F. de Grandi, and A. J. Sieber (1994). Polarimetric contrast classification of agriculturalfields using MAESTRO 1 AIRSAR data. Int. J. Rem. Sens., vol. 15, pp. 2851-2869.

Le Toan T., A. Beaudoin, J. Riom and D. Guyon, 1992.’Relating forest biomass to SAR data’. IEEEtransactions on Geoscience and Remote Sensing, vol.30, no.2, pp403-411.

Le Toan, T, A. Beaudon, J. Riom, D. Guyan, 1991.’ Relating forest parameters to SAR data. ‘In proceedingsof IGARSS’91. Pp689-693.

Le Toan, T., Shahin, A. and Riom, J. (1980). Application of Digitalized Radar Images to Pine Forest Inventory:First Results. Proceedings of the 14th InternationalSymposium on Remote Sensing of Environment. pp. 493-944.


Levin, S. B., J. R. Everett, J. van Roessel. 1973. Resource Mapping in Latin America by Combined Use ofRadar and Satellite Imagery. Ist Panamerican Symp. on Remote Sensing. Panama Citv. pp.65-76.

Lewis. A. J.. and H. C. MacDonald. 1972. Mapping of Mangrove and Perpendicular-Oriented Shell Reefs inSoutheastern Panama With Side-Looking Radar. Photogrammetria. 28: 187-199.

Lillesand T.M. and R. W. Keifer, 1994. Remote Sensing and image interpretation. New York, etc. (Johnwilley & sons, Inc.), 3rd edition, 750 p.

Liu J., T Xuyan and X Jinkai. 1986. .Application of Shuttle Imaging Radar Data For Land UseInvestigations. Remote Sensing of Environment. 19:291-301

Lodwick G. D. and E. R. Sudibjo. 1988. Base Tapping From Radar imagery Proceedings of IGARSS '.88Symposium. Edinburgh. Scotland. pp.583.

Lopes, A., E. Nezry, R. Touzi, and H. Laur (1993). Structure detection and statistical adaptive filtering inSAR images. Int. J. Rem. Sens., vol. 14, pp.1735-1758.

Lowry. R.T. P. van Eck and R.V. Dams. 1986. SAR Imagery for Forest Management. Proceedings IGARSS '86. Zurich. pp. 901 -905.

Lozano-Garcia. D.F. and Roger M. Hoffer. 1987. Uso de Radar Aeroespaciai para el Mapeo de ZonasForestales. Land and Resource Evaluation for National Planning in the Tropics. 'ADA Forest Venice GeneralTechnical Report W0-39. Washington. DC. P p .228-233.

Lozano-Garcia D.F. and R.M. Hoffer. 1993. Synergistic Effects of Combined Landsat-TM md SIR-B Datafor Forest Resources Assessments International Journal of Remote Sensing, 14(14):2677-2694.

Luvall, J. C. and H. R. Holbo (1989). Measurements of short-term thermal responses of coniferous forestcanopies using thermal scanner data. Rem. Sens. Environ., vol. 27, pp. 1-10.

McDonald, K. C., M. C. Dobson, and F. T. Ulaby (1990). Using MIMICS to model L-band multiangle andmultitemporal backscatter for a walnut orchard. IEEE Trans. Geosci. Rem. Sens., vol. 28, pp. 477-491.

McDonald, K.C., M. C. Dobson, and F. T. Ulaby (1991). Modeling multifrequency diurnal backscatter froma walnut orchard. IEEE Trans. Geosci. Rem. Sens., vol. 29, pp. 852-863.

McDonald, K. C. and F. T. Ulaby (1993). Radiative transfer modeling of discontinuous tree canopies atmicrowave frequencies. Int. J. Rem. Sens., vol. 14, No. 11, pp. 2097-2128.

MacDonald. H.C.. W.P. Waite and J.S. Demarcke. 1980. Use of SESAT Radar Imagery for the Detection ofStanding Water Beneath Forest Vegetation. D. Amer. Soc. Photogram. Ann. Tech.MtP., Niagara Falls. MY.

McFarlane, N. A., G. J. Boer, J.-P. Blanchet, and M. Lazare (1992). The Canadian Climate Center secondgeneration general circulation model and its equilibrium climate. Journal of Climate, vol. 5, pp. 1013-1044.

Melack, J. M., L. L. Hess, and S. Sippel (1994). Remote sensing of lakes and floodplains in the AmazonBasin. Rem. Sens. Rev., vol. 10, pp. 127-142.

Mehta, N.C. 1984. Use of Multi-Frequency Multi-Polarization, Multi-Angle Airborne Radars for ClassDiscrimination in a Southern Temperate Forest. Proceedings IGARSS '84. Strasbourg, France. pp. 149-154.

Mercer. J.B.. and S.C. Griffiths, 1993. Operational Topographic Mapping From Airborne SAR Data. Int.Symp. Operationalization of Remote Sensing. Pp. 77-88.


Moghaddan, M., S. Durden, and H. Zebker (1994). Radar measurements of forested areas during OTTER.Rem. Sens. Environ., vol. 47, pp. 154-166.

Moghaddan. M. S. Durden. H. Zebker. and J. Klein. 1992. Radar Measurement of Forested Areas DuringOTTER. IEEE Transactions on Geoscience and Remote Sensing. Pp. 1135-1137.

Morrissey, L. A., G. P. Livingston, and S. L. Durden (1994). Use of SAR in regional methane exchangestudies. Int. J. Rem. Sens., vol. 15, pp. 1337-1342.

Morrissey, L. A., S. L. Durden, G. P. Livingston, J. A. Stearn, and L. S. Guild (1995). Differentiatingmethane source areas with multitemporal ERS-1 SAR data within arctic environments. IEEE Trans. Geosci.Rem. Sens.

Mougin, E., Lopes, A. and Le Toan, T. (1990). Microwave Propagation At X-BandIn Cylindrical-Shaped Forestcomponents: Attenuation Observations. IEEE Transactions on Geoscience and Remote Sensing, 28(1):60-69.

Moulton. J.E. and D.R. Peddle. 1989. Assessment of the Information Content of C-Band SAR in ForestMonitoring Activities. Proceedings IGARSS '89/th Canadian Symposium on Remote Sensing. Vancouver.Canada. Vol I. pp. 1406-1409.

Mueller P.W., R.M. Hoffer, and D.F. Lozano-Garcia. 1985a. Interpretation of Forest Cover on Microwaveand Optical Satellite Imagery. Proceedings of the Tenth William T. Pecora Memorial Remote SensingSymposium.. Ft. Collins. CO. Pp. 578-592.

Mueller. P.W. and R.M. Hoffer. 1985b. Interpretation of Satellite and Aircraft L-Band Synthetic ApertureRadar Imagery. Proceedings of the Fall Convention or the American Society of Photogrammetry and RemoteSensing, Indianapolis. Id'. Pp. 465-475.

Mueller. P.W. and R.M. Hoffer. 1988. Spatial Filtering of Shuttle Imaging Radar-B Data. Technical Papersof the 1988 Annual Convention of the American Congress on Surveving and Madding and the AmericanSociety for Photogrammetry and Remote Sensing, St. Louis. MO. Vol. 3:114-127.

Mueller. P.W. and RM. Hoffer. 1989. Low-Passing Spatial Filtering of Satellite Radar Data.Photogrammetric Engineering and Remote Sensing. 55(6):887-895.

Nakayama, Y., Kimura, H. and Mukai, Y. (1988). A Study of the Accuracy of Land Cover Classification fromSAR Images. Asian-Pacific Remote Sensing Journal, Vol. 1, No. 1. pp. 79-86.

NASA., 1995. http://Southport,jpl.nasa.gov/nrc/chapter2.html#RTFToC7

Ormsby. J. P.. B. J. Blanchard. and A. J. Blachard. 1985. Detection of Lowland Flooding losing ActiveMicrowave Systems. Photogrammetric Engineering and Remote Sensing. 51(3):317-318.

Paris. J.F.. and S.L. Justin. 1990. Quantitative Estimation of Standing Biomass from L-BandMultipolarizaiton Data. Proc. Int. Geoscience and Remote Sensing Symp. 90. Pp. 147-150.

Pierce, L. E., F. T. Ulaby, K. Sarabandi, and M. C. Dobson (1994). Knowledge-based classification ofpolarimetric SAR images. IEEE Trans. Geosci. Rem. Sens., vol. 31, pp. 1081-1086.

Place J. L. 1985. Mapping of Forested Wetland: Use of SEASAT Radar Images to ComplementConventional Sources. Professional Geographer. 37(4):463-469.

Pope, K. O., E. J. Sheffner, K. J. Linthicum, C. L. Bailey, T. M. Logan, E. S. Kasischke, K. Birney, A. R.Njogu, and C. R. Roberts (1992). Identification of Central Kenyan Rift Valley fever virus vector habitatswith Landsat TM and evaluation of their flooding status with airborne imaging radar. Rem. Sens. Environ.,vol. 40, pp. 185-196.


Pope, K. O., J. M. Rey-Benayas, and J. F. Paris (1994). Radar remote sensing of forest and wetlandecosystems in the Central American tropics. Rem. Sens. Environ., vol. 48, pp. 205-219.

Post, W. M. (1993). Uncertainties in the terrestrial carbon cycle, inVegetation Dynamics and Global Change.(A. M. Solomon and H. H. Shugart, eds). Chapman and Hall, London, pp. 116-132.

Prati. C. and F. Rocca. 1990. Limits to the Resolution of Elevation Maps From Stereo SAR Images. Int. J.Remote Sensing. 11 (12):2215-2235.

Pulliainen, J. T., 1994. Investigation on the backscattering properties of finish Boreal forests at C- and X-band: A semi-emperical modelling approach. IEEE Transactions on Geoscience and Remote Sensing, vol.32. No5. PP 1041-1050.

Pulliainen, J.T. 1996. ’Seasonal Dynamics of C-band Backscatter of boreal forests with applications tobiomass and soil moisture estimation '. IEEE Transactions on Geoscience and Remote Sensing, vol. 34 PP7587.

Ranson, K.J.. and G. Sun. 1992. Mapping Biomass for a Northern Forest Ecosystem Using Multi-FrequencySAR Data. Proc. Int. Geoscience and Remote Sensing 92. Pp. 1220-1222.

Ranson, J. K. and G. Sun (1994). Northern forest classification using temporal multifrequency andmultipolarimetric SAR images. Rem. Sens. Environ., vol. 47, pp. 142-153.

Ranson, K. J. and G. Sun (1994). Mapping biomass of a northern forest using multifrequency SAR data.IEEE Trans. Geosci. Rem. Sens., vol. 32, pp. 388-396.

Ranson, K.J., S. Saatchi, and G. Sun. Boreal forest ecosystem characterization with SIR-C/X-SAR. IEEETrans. Geosci. Rem. Sens.

Rauste, T.H., K.H. J. Pulliainen and M. Mallikainen, 1994. ‘ Radar-based forest biomass estimation.‘International Journal of remote Sensing, vol.15, no.14, pp.2797-2808.

Reich, R.M., Y.A, Hussin, 1993.’Estimating average stand biomass for a regional forest inventory usingradar backscatter’. ITC Journal No.1, pp.82-87.

Rejmankova, E., D. R. Roberts, R. E. Harbach, J. Pecor, E. L. Peyton, S. Manguin, R. Krieg, J. Polanco, andL. Legters (1993). Environmental and regional determinants of anopheles larval distribution in northernBelize. Environ. Entomology, vol. 22, pp. 978-992.

Rejmankova, E., K. O. Pope, M. D. Pohl, and J. M. Rey-Benayas (1995). Freshwater wetland plantcommunities of northern Belize: Implications for paleoecological studies of Maya wetland agriculture.Biotropica (in press).

Richards, J. A., G-Q. Sun, and D. S. Simonett (1987). L-band radar backscatter modeling of forest stands.IEEE Trans. Geosci. Rem. Sens., vol. 25, pp. 487-498.

Richards, J.A., P.W. Woodgate, and A.K. Skidmore. 1987. An Explanation of Enhanced RadarBackscattering from Flooded Forests. Int. J. Remote Sensing, 8(7)(2) pp. 1093-1100.

Richards. J.A. and Z. Ahmed. 1988. Association of Radar Backscatter with Biophysical Characteristics ofAustralian Forests. Proc. Int. Geoscience and Remote Sensing Symp.'88. Pp. 1361-1365.

Ruzek. M. and R.,M. Hoffer. 1985. Seasat and SIR-B: A Comparison or a Forest Test Site. (Abstract)Proceedings of the 1985 International Geoscience and Remote Sensing Symposium. Amherst. MA. p. 579


Rignot, E. and J. B. Way (1994). Monitoring freeze-thaw cycles along north-south Alaskan transects usingERS-1 SAR. Rem. Sens. Environ., vol. 49, pp. 131-137.

Rignot, E., J. B. Way, K. McDonald, L. Viereck, C. Williams, P. Adams, C. Payne, W. Wood, and J. Shi(1994a). Monitoring of environmental conditions in taiga forests using ERS-1 SAR. Rem. Sens. Environ.,vol. 49, pp. 145-154.

Rignot, E., J. B. Way, C. Williams, and L. Viereck (1994b). Radar estimates of aboveground biomass inboreal forests of interior Alaska. IEEE Trans. Geosci. Rem. Sens., vol. 32, pp. 1117-1124.

Rignot, E., C. Williams, J. B. Way, and L. Viereck (1994c). Mapping of forest types in Alaskan borealforests using SAR imagery. IEEE Trans. Geosci. Rem. Sens., vol. 32, pp. 1051-1059.

Rignot, E., JJ. Van zyl and R. Zimmermann, 1994. “ Inference of Forest Biomass Using P-band Circularpolarized radar Signals,” In: Proceedings of IGARSS ‘ 94. Surface and Atmosphere Remote Sensing,Technology Data Analysis and Interpretation, Pasadena, August 8-12 1994, pp. 2467-2469.

Rignot, E.J., R. Zimmermann and JJ. Van Zyl, 1995. Spaceborne applications of P band Imaging radars formeasuring forest biomass. IEEE Transactions on Geoscience and Remote Sensing, vol.33, no. 5, pp.1162-1169.

Sabins, F.F., Jr. 1987. Remote Sensing PrinciDles and Interpretation (2nd ed.) W.H. Freeman & Co., N.Y.Chapter 6, "Radar Images," Pp. 177-233.

Sader, S. A. (1987). Forest Biomass, Canopy Structure, and Species Composition Relationships withMultipolarization L-Band Synthetic Aperture Radar Data. Photogrammetric Engineering and Remote Sensing,53(2):193-202.

Sader, S.A., R.B. Waide, W. T. Lawrence and A.T. Joyce, 1989. ‘Tropical forest biomass and successionalage class relationship to a vegetation index derived from landsat TM data. Remote Sensing of environment,Vol. 47, No.2, pp 109-119.

Sakohi, A, 1993. ‘Spot satellite response and tree sub-sampling approach for a cost-effective woody biomassestimation in Tunisia. MSc thesis.

Sanden, J. J. van der. (1991). Microwave Remote Sensing for Forest Monitoring Purposes. IGARSS '91, Vol. 3,pp. 1507-1510.

Schlentner, R. E. and K. Van Cleve (1985). Relationships between CO2 evolution from soil, substratetemperature and substrate moisture in four mature forest types in interior Alaska. Can. J. For. Res., vol. 15,pp. 97-106.

Schlesinger, M. E. and J. F. B. Mitchell (1987). Climate model simulations of the equilibrium climaticresponse to increased carbon dioxide. Reviews of Geophysics, vol. 25, pp. 760-798.

Sicco Smit, G. 1975. Will the Road to the Green Hell Be Paved with SLAR?: A Case Study of TropicalRain Forest Type Mapping in Colombia. ITC Joumal. 75(2):245-266.

Sicco Smit, G. 1988. A Practical Application of Radar Imagery for Tropical Rain Forest VegetationMapping. ITC Workshop on Synthetic Aperture Radar System and Its Application in Natural Resources.Enschede, Netherlands pp.249-263.

Sicco Smit, G. S. 1992. A Good, A Better or the Best Remote Sensing Application System For TropicalRain Forest Survey in the Amazon Region. ITC Workwhop on Synthetic Aperture Radar System and ItsApplication in Natural Resources. Enschede, Netherlands pp.760-762.


Sieber, A. J. (1985). Forest signatures in Imaging and Non-imaging Microwave Scatterometer Data. ESAJournal, Vol. 9, pp. 431-448.

Sieber, A., et al. 1987. The Global Forest Ecosystem as Viewed by ERS-1, SIR-C and EOS. IGARSS '87 -Remote Sensing: Understanding the Earth as a System, Intemational Geoscience and Remote SensingSociety, Ann Arbor, MI. IEEE 87 CH2434-9, Vol. II:967-974.

Simonett, D. S. and Davis, R. E. (1983). Image Analysis-Active Microwave. Manual of Remote Sensing,Second Edition, Vol I, American Society of Photogrammetry, pp. 1125-1181.

Smith, T. M., W. P. Cramer, R. K. Dixon, R. Leemans, R. P. Neilson, and A. M. Solomon (1993). The globalterrestrial carbon cycle. Water, Air, Soil Pol., vol. 70, pp. 19-37.

Solomon, A. M. and W. Cramer (1993). Biospheric implications of global environmental change, inVegetation Dynamics and Global Change. A. M. Solomon and H. H. Shugart, eds.), Chapman and Hall,London, pp. 25-52.

Sokeland, A. D., H. De Groof, J. Conway and A. J. Sieber. 1992. Trees Project: A Case Study on the Use ofMulti-Temporal ERS-1 SAR Data and SPOT Data For Monitoring Tropical Ecosystems. Proc. lst ERS-1Symp. pp.671-677.

Stennis Space Center. 1995. "Imaging Radar Sensor Profiles." (Booklet). Office of Space Access andTechnology, Commercial Remote Sensing Program Office, Stennis Space Center, MS.

Stone, A. T., and G. M. Woodwell. 1988. Shuttle Imaging Radar-A Analysis of Land Use in Amazonia. Int.J. Remote Sensing. 9(1):95-105.

Subotic, N., J. Gorinan, S. DeGraaf, D.R. Sheen, E.S. Kasischke, and D. Beverstock. 1990. Classification ofForested Terrain Using Multifrequency, Polarimetric SAR Data. Proc. Int. Geoscience and Remote SensingSym . '90. p. 2401.

Sun, G. and D. S. Simonett (1988). Simulation of L-band HH microwave backscattering from coniferousforest stands: a comparison with SIR-B data. Int. J. Rem. Sens., vol. 9, pp. 907-925.

Sun, G., D. S. Simonett, and A. H. Strahler (1991). A radar backscatter model for discontinuous forestcanopies. IEEE Trans. Geosci. Rem. Sens., vol. 29, pp. 639-650.

Sun, G. and K. J. Ranson (1995). A three-dimensional radar backscatter model for forest canopies. IEEETrans. Geosci. Rem. Sens. (in press).

Tans, P. P., I. Y. Fung, and T. Takahashi (1990). Observational constraints on the global atmospheric CO2budget. Science, vol. 247, pp. 1431-1438.

The Integrated Land and Water Information System, ILWIS 1.4, User's Manual. International Institute forAerospace Survey and Earth Sciences, The Netherlands.

Thompson, M. D., and R. V. Dams. 1990. Forest and Land Cover Mapping From SAR: A Summary ofRecent Tropical Studies. Proc. 23rd Int. Symp. on Remote Sensing of Environinent. pp. 509-516.

Thompson, M.D., B.C. Macdonald, and W.C. Jefferies. 1993. Progress in Operational ForestMapping Using Airbome Radar in Tropical Regions. Int. Symp. Operationalization ofRemote Sensing. pp 163-175.

Trevett, J.W., 1986. Imaging radar for resources surveys. London, New York (Chapman and Hall), 313p.


Ulaby, F. T., Moore, R. K. and Fung, A. K. (1981). Microwave Remote Sensing,Vol I and Vol. II. Addison-Wesley, Reading, Mass.

Ulaby, F. T., B. Brisco, and M. C. Dobson (1983). Improved spatial mapping of rainfall events withspaceborne SAR imagery. IEEE Trans. Rem. Sens., vol. 21, pp. 118-121.

Ulaby, F. T., K. Sarabandi, K. McDonald, M. Whitt, and M. C. Dobson (1990). Michigan microwave canopyscattering model. Int. J. Rem. Sens., vol. 11, pp. 1223-1253.

Ulaby, F. T., R. K. Moore, and A. K. Fung. 1981-1986 (86a). Microwave Remote Sensing: Active andPassive, Vols. 1, 2, & 3. Addison-Wesley, Reading,

Ulaby, F. T., F. Kouyate, B. Brisco, and T. H. L. Williams. 1986b. Textural Information in SAR Images.IEEE Transactions on Geoscience and Remote Sensing. 24(2):232-245.

Umali, Ricardo T. (1978). Landsat Assisted Forest Inventory of the Philippine Island. Proceeding of the TwelfthInternational Symposium on Remote Sensing of Environment, Vol. 2. Ann Arbor, Michigan. pp. 1401-1405.

Umali, Ricardo T. (1985). Remote Sensing for Natural Resources Management in the Philippines. Developmentand Applications of Remote Sensing for Planning, Management and Decision-Making. Seminar Proceedings.UNDP/ESCAP Regional Remote Sensing Programme, Beijing. pp. 59-70.

Van der Sanden, J. J. 1991. Microwave Remote Sensing For Forest Monitoring Purposes. Proc. Int.Geoscience and Remote Sensing Symp. '91. pp. 1507-1510.

Van der Sanden, JJ, 1997.’Radar remote to support tropical forest management. Wagenignen, theNetherlands (Wagenignen Agricultural University), Tropendos-Guyana Series 5, doctoral thesis, 330p.

van Zyl, J. J. (1989). Unsupervised classification of scattering behavior using radar polarimetry data. IEEETrans. Geosci. Rem. Sens., vol. 27, pp. 36-45.

van Zyl, J. J. and C. F. Burnette (1992). Baysian classification of polarimetric SAR images using adaptive apriori probabilities. Int. J. Rem. Sens., vol. 13, pp. 835-840.

van Zyl, J. J. 1993. The Effect of Topography on Radar Scattering from Vegetated Areas. IEEETransactions on Geoscience and Remote Sensing. 31(1):153-160.

Viksne, A., T. C. Liston, and C. D. Sapp. 1970. SLR Reconnaissance of Panama. PhotogrammetricEngineering. 36(3):253-259.

Viereck, L. A. (1983). The effects of fire in black spruce ecosystems of Alaska and northern Canada in TheRole of Fire in Northern Circumpolar Ecosystems (Scope 18) (R. W. Wein and D. A. MacLean, eds.), JohnWiley and Sons, Chichester, England, pp. 201-220.

Viskne, A., Liston, T. C. and Sapp, C. (1970). SLR Reconnaissance of Panama. Photogrammetric Engineering,36(3): 253-259.

Wang, Y., F. W. Davis, and J. M. Melack (1993a). Simulated and observed backscatter at P-, L-, and C-bands from ponderosa pine stands. IEEE Trans. Geosci. Rem. Sens., vol. 31, No. 4, pp. 871-879.

Wang, Y., J. L. Day, F. W. Davis, and J. M. Melack (1993b). Modeling L-band radar backscatter fromAlaskan boreal forest. IEEE Trans. Geosci. Rem. Sens., vol. 31,pp. 1146-1154.

Wang, Y., F. W. Davis, J. M. Melack, E. S. Kasischke, and N. L. Christensen, Jr. (1994a). The effects ofchanges in forest biomass on radar backscatter from tree canopies. Int. J. Rem. Sens. (in press).


Wang, Y., E. S. Kasischke, F. W. Davis, J. M. Melack, and N. L. Christensen, Jr. (1994b). The effects ofchanges in loblolly pine biomass and soil moisture variations on ERS-1 SAR backscatter--a comparison ofobservations with theory. Rem. Sens. Environ., vol. 49, No. 1, pp. 25-31.

Wang, Y., L. L. Hess, S. Filoso, and J. M. Melack (1995). Canopy penetration studies: modeled radarbackscatter from Amazon floodplain forests at C-, L-, and P-band. Rem. Sens. Environ. (in press).

Waring, R. H., J. B. Way, R. Hunt, Jr., L. Morrisey, K. J. Ranson, J. Weishampel, R. Oren, and S. E.Franklin. Remote sensing with synthetic aperture radar in ecosystem studies. Bioscience (in press).

Way, J. B., J. Paris, E. Kasischke, C. Slaughter, L. Viereck, N. Christensen, M. C. Dobson, F. T. Ulaby, J.Richards, A. Milne, A. Sieber, F. J. Ahern, D. Simonett, R. Hoffer, M. Imhoff, and J. Weber (1990). Theeffect of changing environmental conditions on microwave signatures of forest ecosystems: preliminaryresults of the March 1988 Alaskan aircraft SAR experiment. Int. J. Rem. Sens., vol. 11, pp. 1119-1144.

Way, J.B. and E.A. Smith (1991). The Evolution of Synthetic Aperture Radar Systems and their progressionto the Eos SAR. IEEE Trans. on Geosci. and Rem. Sens, vol. 29, pp. 962-985.

Way, J. B., E. J. M. Rignot, K. C. McDonald, R. Oren, R. Kwok, G. Bonan, M. C. Dobson, L. A. Viereck,and J. A. Roth (1994). Evaluating the type and state of Alaska taiga forests with imaging radar for use inecosystem models. IEEE Trans. Geosci. Rem. Sens., vol. 32, pp. 353-370.

Wedler, E., S. Pala, and A. Jano. 1980. SAR Image Response Over a Conifer Regeneration Site. 6thCanadian Symp. on Remote Sensing, Nova Scotia. pp.179-183.

Wedler, E., and R. Kessler. 1981. Interpretation of Vegetative Cover in Wetlands Using Four-Channel SARImagery. American Society of Photogrammetry 47th Annual Convention. pp. 111- 124.

Weishampel, J. F., G. Sun, J. K. Ranson, D. LeJune, and H. H. Shugart (1994). Forest textural propertiesfrom simulated microwave backscatter: the influence of spatial resolution. Rem. Sens. Environ., vol. 47, pp.120-131.

Werle, D. 1989. Radar Remote Sensing for Application in Forestry. Canada Centre for Remote SensingEnergy, Mines & Resources Canada. Ottawa, Ontario, Canada. 35 pp.

Werle, D. 1989b. Potential Application of Imaging Radar for Monitoring the Depletion of Tropical Forests.Proc. Int. Geosciences and Remote Sensing '89. pp. 1383-1386.

Yueh, S. H., J. A. Kong, J. K. Jao, R. T. Shin, and T. LeToan (1992). Branching model for vegetation. IEEETrans. Geosci. Rem. Sens., vol. 30, pp. 390-402.

Wu, S.T. 1984.’Analysis of synthetic aperture radar data acquired over variety of land cover. ’IEEETransaction Geoscience and Remote Sensing vol. GE-22, no.6, pp 550-557.

Wu, S. T. (1986). Preliminary Report on Measurements of Forest Canopies with C-Band Radar Scatterometer atNASA/NSTL. IEEE Transactions on Geoscience and Remote Sensing, 24(6):894-899.

Wu, S.T, 1987. “Potential application of Multipolarization SAR for the Pine plantation biomass estimation”.IEEE Transactions on Geoscience and Remote Sensing, vol. GE-25, No. Pp. 403-409.

Wu, S.T. 1989. Assessment of Tropical Forest Biophysical Characteristics With Multipolarization SARData Acquired Over a Mountainous Region in Costa Rica. IGARSS '87: 12th Canadian Symposium onRemote Sensing, Vol 3:1378-1382.


Wu, S.T., 1990. Assessment of tropical forest stand characteristics with multipolarization SAR data acquiredover a mountainous region in Costa Rica”. IEEE Transactions on Geoscience and Remote Sensing, vol.28,no.4, pp.752-755.

Wu, S. T. and Sader, S. A. (1987). Multipolarization SAR Data for Surface Feature Delineation and ForestVegetation Characterization. IEEE Transactions on Geoscience and Remote Sensing, 25(1):67-76.

Yelizavetin, I. V. 1993. Digital Terrain Modeling From Radar Image Stereo Pairs. Mapping Sciences andRemote Sensing. 30(2):151-161

3.4 The Applications of Videography in Forestry

1. Assessing the Spatial Impact of Hurricanes on Forest Ecosystems in Southern Louisiana

Hurricanes wreak havoc on societies and ecosystems alike. In 1992, Hurricane Andrew ripped through southFlorida and left a swath of economic and ecological disaster in its wake. Forest inventories were conductedbefore and after the hurricane as part of an interagency research project to assess and to monitor the itseffects on mangrove forests (McInthosh, 1993). Field surveys were slow following Andrew due to siteinaccessibility and the extensive forest damage that hindered ground movement. To overcome theselogistical problems, aerial videography was taken at low altitude by helicopter over mangrove forests alongthe southwest coast of Florida. A coastal and inland transect were flown within the forest boundary ofmangrove extent over Ten Thousand Island National Wildlife Refuge and Everglades National Park. Thetransects were perpendicular to the hurricane path. Continuous video footage was taken along these transectswith recorded voice transmissions of co-ordinate location, altitude, flight speed, bearing, and other pertinentobservations of ground damage on the tape. A separate global positioning system (GPS) unit was trackingexact helicopter movement along with the video. Video analysis involved both visual and image analysis andgeoreferencing. Percent bole damage was determined for each video frame and placed into one of four broadcategories. These interpretations were grouped into similar damage zones in a GIS for interpolation betweenvideo frame locations and flight lines. Five damage zones were established from the four damageclassifications: 4=severe, 3=moderate, 2=light, 1=scattered light damage, and 0=no damage. Theclassification result is shown in Figure 1. At the end of the study it was concluded that the use of aerialvideography techniques allowed a rapid assessment of damage to forest resources in southern Louisianacaused by Hurricane Andrew. Aerial videography reduced the need for ground analysis of the damaged area.In addition, GPS coupled with the aerial videography provided for quick orientation of the video data andallowed the video frame location and corresponding classification to be entered into a GIS.


Figure 1. Path of Hurricane Andrew in southern Louisiana and zones of forest damage (McInthosh, et a., 1993).

2. Distinguishing Species Composition and Vegetation Pattern in Riparian Forests

A study was conducted utilising aerial colour infrared (CIR) videography to assess the plant composition andzonation of riparian vegetation along the Lower Rio Grande in Cameron and Hidalgo Counties, Texas.Results showed that several major vegetation cover types could be delineated at these sites. Aerial CIRvideography appears to be an excellent method for mapping the distribution of trees in the three layer, butnot for the distribution of understory shrubs and herbaceous species.

3. Estimation of Leaf Area Index and Light Intercepted by Shrubs from Digital Videography

In 1991 research was done by using digital videography as a part of Oregon Transect Ecosystem Research(OTTER) Project, which was conducted by Beverly E. Law. Because the development of digital videographyhas generated interest in the use of these instruments to estimate vegetation variables such as leaf area index(LAI) and light intercepted (f IPAR ) by shrubs in shrublands and open forest canopies. The objective of thisstudy was to examine the relationships between the percentage of scene classified as foliage in colour digitalvideography acquired in the field, and LAI and f IPAR by canopies of two shrub species, bitterbrush (Purshiatridentata) and manzanita (Arctostaphylos patula) (Law et al., 1995). Researches collected still video andf IPAR data on 23 bitterbrush and 27 manzanita plots. The digital video camera integrated broad blue, green,and red bands, and recorded 50 digital images on a single diskette. Video images were transferred to acomputer with a Matrox graphic board, and analyzed with resource Imaging Graphics System software. The


images were processed to determine four components in a scene: the percentage of illuminated and shadedleaves, and illuminated and shaded background (soil, branches). Supervised classification was used, with thefour-pixel categories specified by training sites on known scene components. The minimum distance to meanclassifier was applied to classify pixels into the four categories.According to results there was a positive linear relationship between the combined percentage of sceneoccupied by sunlit leaves, shaded leaves, and shaded background as the dependent variable, and

f IPAR (R 2 =0.66 manzanita, and 0.61 bitterbrush) (Figure 2). A logarithmic relationship was observedbetween the combined percentage of scene occupied by sunlit leaves, shaded leaves and shaded background,and the independent variable LAI (R 2 =0.88 manzanita, and 0.65 bitterbrush) (Figure 3). A green/red ratiowas useful for classification of video scene components into illuminated and shaded leaves, and shadedbackground. This study demonstrates that colour digital videography may be suitable for field estimates off IPAR and low LAIs in open shrublands.

Figure 2. The relationship between percentage of scene occupied by illuminated leaves, shaded leaves, and shaded background as the dependent variable, and the fraction of incident radiation intercepted by vegetation (f IPAR ), as observed in bitterbrush (n=23) and manzanita (n=27).


Figure 3. The relationship between percentage of scene occupied by illuminated leaves, shaded leaves, and shaded background as the dependent variable, leaf area index (LAI), for bitterbrush (n=23) and manzanita (n=27).

4. Differentiating bottomland tree species with multispectral Videography

The Atchafalaya Basin was selected by the Forest Service’s Southern Forest Experiment Station as a pilotstudy site for evaluating the accuracy of multispectral aerial videography in the determination of species ofbottomland trees, and for developing volume estimates from photogrammetric and regression methodsapplied to the videography. Thomasson et al., (1995) focused in their identification of individual trees inlarge-scale videography on the first part of this study.In this study, three plots chosen and designated as 65,66,70. According to field measurements plot 66consists of (97 % percent) bald-cypress (Taxodium distichum (L.) Rich.), black willow (Salix nigra Marsh.),green ash (Fraxinus pennsylvanica Marsh.), and water tupelo (Nyssa aquatica L.). Plot 70 consists of (92 %)American sycamore (Platanus occidentalis L.), eastern cottonwood (Populus deltoides Batr. ex. Marsh.), andboxelder (Acer negundo L.). Within plot 65 the species composition was primarily willow, bald-cypress, andsycamore.

The timing of the missions was designed to exploit differences in vegetative reflectance between leaf-out andsenescence. The video camera used was the Xybion Electronic Systems MSCO2 which contained a 6.6-mmby 8.8-mm CCD image sensor and a rotating filter wheel that allowed a set of sequential images to beobtained in different spectral bands (550 nm visible green, 800nm NIR cw, and 1000 nm NIR cw). Duringflight, images were recorded on VHS cassettes. Digitization and sequential image were performed on anIBM computer. Supervised computer classification procedures were employed. Training statistics werecollected from unshaded areas of individual trees whose species were known. The classification routineschosen for use in this study were the minimum distance and the maximum-likelihood classifier. At the end ofthe study although both methods gave similar accuracy, the maximum-likelihood classifier requiredapproximately 12 times the calculation time of the minimum distance classifier. For the six speciesconsidered- namely, willow, cypress, green ash, sycamore, cottonwood, and boxelder- the averageclassification accuracy found was 70 percent. According to the this result concluded that bands 1,2, and 3were useful for differentiating tree species in the Atchafalya Basin and also classification of multispectralaerial videography shows promise for use in forest inventories.


Table 1: Contingency table for classification of plot 66 (Thomasson et al., 1995).

Known Number Percent Number of Trees Classed into Category

Tree of Correct

type trees (%) Cypress Willow Ash

Cypress 91 79(78) 72 13 6

Willow 59 73(71) 16 43 0

Ash 29 93(92) 0 2 27

Percent correct of trees classed

in this category 82(81) 74(72) 82(79)

*Numbers in parenthesis exclude trees used to generate training statistics. (Thomasson et al.,

1995)

5. Describing urban forest cover

In 1990, the USA Forest Service has started the project to better understand the effects of urban vegetationon the environment. As part of this project research has been done by Sacamano et al., (1995) to demonstratethe utility of two airborne videography systems compared to traditional aerial photography for interpretingurban land cover. Comparisons are made among the different types of imagery in terms of accuracy. Cost ofimagery and implications for integration with GIS are also considered. The results are intended to identifythe more useful and economical means for accurately describing urban forest cover.In order to test the application of airborne videography in urban areas, Oak Park, Illinois, was selected as acase study. This study compared SRSC multispectral digital colour video, FPM/ MAG super VHS formatanalog composite colour video, and B&W and colour infrared photography of Oak Park. Researchesemployed a consistent, uniform method for interpreting the various types of imagery, which has been usedfor aerial photographic interpretation and was adapted in this study for the video imagery. For interpretationmore than one interpreter was used in effort to diminish the effect of individual interpreter bias. Analysingthe accuracy of interpretations compared the analog and digital videography and the B&W and colourinfrared aerial photography. The results of interpretation are given in Table 2.

Table 2: Percent of points correctly interpreted on imagery (Sacamano et al., 1995)

B&W Colour IR Analog Digital

Cower Class Photo Photo Video Video

Tree/Shrub 46.4 (G) 69.8 (G,P) 62.6 (G) 64.5 (G,P)

Grass/Soil 56.9 (T) 45.6 (T) 37.5 (B) 49.3 (T)

Building 79.4 (P) 68.1 (T) 61.7 (P) 70.8 (P)

Paving 65.4 (B,G) 61.6 (G) 68.0 (B,G) 66.7(G,B,P)

Most frequent reason for misinterpretation is indicated within () where B= Building, G= Grass: P= Paving;

T= Tree

According to the result of this study it concluded that for overall estimates of percent cover, photography andvideography are both fairly accurate and useful. Colour infrared photography offers the highest degree ofaccuracy and, considering the limited availability of airborne videography systems and their costs, may offera better tool for urban forest mangers and planners interested in urban land cover, particularly trees.


6. Application of multispectral aerial video in forest and and cover classification

FiltersLenses

Black&

White CD

Cameras

Figure 4: Schematic of 4 CVS (King et al., 1990)

A 4-camera aerial multispectral video system (4CVS) was developed by King et al., (1983) and evaluated inforest and land cover classification. The system (Figure4) is a low-cost alternative to aerial photographic andother electro-optical scanning systems. It is capable of imaging in narrow spectral bands in the visible andnear infrared. Both black and white band sequential imagery and colour/ false colour imagery are generated.The feasibility of multispectral aerial video in land cover classification was evaluated by comparingclassification accuracies for a selected test site using several classifiers and several image processingmethods. A conservation area north-west of Toronto was selected as a test site because it contained a widevariety of cover types in discrete areas suitable for training and classification. In overall assessment of 4CVS-image quality, several image-processing procedures were tested to determine optimum radiometricenhancements or corrections for video based land cover classification. These are: noise reduction, reductionof spatial variations in image plane irradiance, reduction of spectral dimensionality, addition of textureinformation, contrast stretching, edge enhancement, and post- classification thematic map smoothing.Comparison of the accuracy of five spectral classification routines which were parallelpiped, Suits-Wagner,minimum distance to mean, supervised maximum likelihood, and unsupervised maximum likelihood, and theeffects of the image processing procedures on classification accuracy were evaluated using table analysis.Each classification accuracy were compared to a digital ground truth map generated from interpretation ofaerial photographs, extensive field survey, and analysis of Toronto Region Authority maps. As a resultaverage accuracy of classification found was 70 %. According to the result of this study it was concludedthat classification of multispectral video for detailed cover types was significantly improved (70 %) byreducing noise and spatial radiometric variations in the data, as well as by adding texture information.Additional improvements in imaging conditions, spectral band selection, and feature extraction combine togive low cost multispectral video high potential in land cover type analysis and thematic map updating.

Black and WhiteMultiplexer

Colour compositeEncoder 4 CVS

VCR VCRMonitor

Visual ImageInterpretation

ComputerProcessing


7. Aerial videography for verification of forest resources assessment

Aerial Videography combined with GPS coordinates, was flown to verify interpretations of five 1990Landsat Thematic Mapper (TM) images in Mexico. These images were paired with nominal 1980 Landsatimages to estimate forest cover changes for FAO Forest Resources Assessment 1990 Project. The results ofthis projects presented by McIntosh et al., (1993). In this project video transects of the scenes were flown at1 000 meters above ground level (AGL), with an 8.5 millimetre focal length video camera lens, producing animage swath approximately flying height (AGL). Video imagery of the four southern and northern (whichwas under cloud shadow or rainy conditions) FAO TM scenes was acquired and field verification wasaccomplished. A videocassette recorder and television were used for manual verification and correctioninterpreted TM print overlays. The FAO forest cover classification scheme included open forest, closedforest, fragmented or discontinuous forest, and shrubland. These were subdivided into broad crown coverdensities of 10 to 40 percent, 40 to 70 percent, or 70 to 100 percent crown cover. According to theconclusion this project the quality of the video imagery of the southern for FAO sample areas acceptable forinterpretation. Vegetation types according to FAO definitions were distinguishable under partly cloudyconditions and with videography high quality imagery possible under low conditions as well.

8. Multispectral videography for assessing the accuracy of wildlife habitat maps

Information on the extent and condition of forest lands throughout the U.S. is becoming increasinglyimportant for the preservation of habitat and the management of wildlife species. In northern Californiaresearch has been done by Biging et al., (1995) in which the use of high-resolution multispectral aerialvideography was evaluated in relation to ground survey for assessing the accuracy of wildlife habitatthematic maps. It was tested how well videography could identify tree species, crown size, hence crowncover percentage and tree sizes (diameter at breast height). These three parameters are used to classifywildlife habitat in California.In this project forest environment was selected from a fixed-wing aircraft using a suite of image sensors toprovide spectrally complimentary high spatial resolution video imagery. This imagery was acquired usingthe RECO 1 proprietary remote sensing data acquisition system that emulates the following LandsatThematic mapper satellite bands: blue (B), green (G), red(R), near-infrared (NIR), and thermal infrared(TIR). Three bands (B, G, R) were acquired on one high-resolution three-chip camera, and NIR wascollected on a monochrome video camera with extended sensitivity in the NIR. The thermal infraredimagery was acquired from a liquid nitrogen-cooled thermal scanner. Image blur was minimised through theuse of image motion-compensating camera mount. The interference filters used were band 1(R: 400-500nm), band 2 (G: 500-600nm), band 3(R: 600-700 nm) band 4(NIR: 780-1000 nm), and band 5 (TIR: 8-12µm). The three separate channels (RGB, NIR and TIR) were multiplexed (recorded in alternating frames)to a single record to facilitate image band registration. The analog data from the video frames were convertedto a digital format using frame grabber technology.The main goal of this research was to develop videographic methods for estimating the requisite forestparameters for characterising wildlife habitat so that this information can be used to perform an accuracyassessment of existing wildlife habitat maps. At the and of the study, results showed that digitizing crownsto determine crown cover percent and calculating the crown cover percent after classification are promisingtechniques for acquiring reference data for accuracy assessment. However, both techniques are substantiallyless accurate if there is significant shadowing in the image. Because of this reason it is highly desirable toacquire imagery in early to mid-summer, when shadowing is minimized. Based on data analysis, it appearsthat videography can be used, in certain conditions, to attain reasonable estimates of crown cover percent andpossibly quadratic mean diameter, and to confirm or deny the reasonableness of satellite classifications ofwildlife habitat. After investigation it was found that the videographic procedure, provided good distinctionbetween conifer and hardwood components, but not for identifying individual trees. Because it was notpossible to identify individual tree species at all resolution levels. However, several techniques weredeveloped for estimating individual tree crown sizes, which were used to calculate crown cover percentagewith acceptable accuracy. According to the researchers those methods need further development, but holdpromise in providing a readily acquired and cost-effective way of checking the accuracy of crown coverpercent estimates. Tree size, as measured by diameter at breast height, is difficult to infer, and their methodswere not accurate for estimating this parameter.


9. Conclusions

Aerial vidography is a new remote sensing technique. It has expanded very rapidly in the past few years. Itprovides immediate imagery with very low cost and high light sensitivity making aerial videographyattractive to the remote sensing users. Beside these advantages aerial videography has some limitations aswell especially low resolution in comparison with aerial photography.Aerial videography is a technique which use single black and white, single color, multispectral black andwhite, and multispectral color which can be sensitive visible or visible/NIR portion of electromagneticspectrum. This system can be mounted on regular aeroplane or helicopter which provides close rangingresearch.

Aerial videography has been used successfully in forestry applications as outlined below:• Monitoring the forest damage. Videography was very useful remote sensing tool in terms of time

consumption to assess the damage.• Mapping trees distribution with very good accuracy.• Color digital videography was found good in estimating of f IPAR and LAI in open shrublands.

• For differentiating tree species, aerial videography gave overall accuracy of 70 percent.• Aerial videography systems offered better tool for urban forest management.• Forest and land cover classification done with videography shows average accuracy of 70 percent. FAO

forest classification was verified by using aerial multispectral videography.

10. Literature Cited

Aerial Videography &Land-Cover Mapping, http://www.snr.uvm.edu/airvideo.html

Anonymous 1, Videography, http://globike.com/vidgrafyO.html

Anonymous 2, Using Videography to Evaluate Impacts to Ecological Resources,http://www.ead.anl.gov/~web/newead/prgprj/proj/video/video.html

Bakker, J.G.M., Braam,B.M., van Leeuven, H.J.C. and de Boer,R.J., 1993, AirborneCCD-Video Monitoring Performance and Application of A Colour Infrared Video System, NRSP-2,92-30.

Bakker, J.G.M., Huising E.J., Kroon, R.J.G.A. and Loedeman J.H., 1994,3D Object Geometry FromOblique Recorded Video Frames; An Approach to a Flexible, User Friendly, Fast and Cheap MonitoringTechnique, NRSP-2, 94-07.

Biging,G.S., Dobbertin, M. and Murphy, E.C., 1995, A Test of Airborne Multispectral Videography forAssessing the Accuracy of Wildlife Habitat Maps, Canadian Journal of Remote Sensing, Vol.21, No.3,pp.357-367.

Everitt, J.H., Lulla, Escobar, D.E. and Richardson, A.J., 1990,Aerospace Video Imaging Systems forRangeland Management, PE & RS, Vol.56, pp. 343-349.

Hutchinson, C.F., Schowengerdt, Baker, L.R., 1990,A Two-Channel Multiplex Video Remote SensingSystem, PE&RS, Vol.56, pp.1125-1128.

Jacobs, D.M., and McIntosh, S.E., 1993,Forest Damage Assessment of Hurricane Andrew in SouthernLouisiana Using Airborne Videography, Proceedings of the 14th Biennial Workshop on Color AerialPhotography and Videography for Resource Monitoring, Logan, Utah, pp.115-124.

Judd,F., Leonard,R., Everitt,J., Escobar, D., and Davis,M., Using Multispectral Videography to CompareThe Pattern of Zonation Between BrackishWater Marshes and Salt Water Marshes of the Rio Grande Delta,http://www.nal.usda.gov/ttic/tektran/data


King, D. and Vlcek, J., 1990, Development of a Multispectral Video System and Its Application In Forestry,Canadian Journal of Remote Sensing, Vol.16, No.1 pp.15-22.

Law, B. E., 1995Estimation of Leaf Area Index and Light Intercepted by Shrubs From Digital Videography,Remote Sensing of Environment, Vol. 51, pp.276-280.

Leonard,R., Judd, F., Everitt, J., Escobar, D., and Davis, M., Using Multispectral Videography inDistinguishing Species Composition and Vegetation Pattern in Riparian Forests of The Lower Rio Grande,http://www.nal.usda.gov/ttic/tektran/data

Lillesand, T.M. and Kiefer, R.W., 1987, Remote Sensing and Image Interpretation, pp.120-125.

Mausel, P.W., Everitt, J.H., Escobar, D.E., and King, D. J., 1992,Airborne Videography: Current Status andFuture Perspectives, PE&RS, Vol.58, No: 8, pp.1189-1195.

McIntosh S.E., Jacobs D.M., and Robles A. F., 1993, Acquisition of Airborne Videography of Mexico forVerification FAO Forest Assessment Landsat Interpretations, Proceedings of the 14th Biennial Workshop onColor Aerial Photography and Videography for Resource Monitoring, Logan, Utah, pp.135-143.

Meisner, D., E., 1986, Fundamentals of Airborne Video Remote Sensing, Remote Sensing of Environment,Vol.19, pp. 63-79.

Myre,R.J., 1993, Acolour Airborne Video System Developed within The U.S. Forest Service, Proceedings ofthe 14th Biennial Workshop on Color Aerial Photography and Videography for Resource Monitoring, Logan,Utah, pp.9-18.

Raper, J. and McCarthy T., 1996, Data on Demand, GIS Europe, pp.28-30.

Sacamano, P.L., McPherson, E.G., Myhre, R.J., Stankovich, M., and Weih, R.C., May 1995 Journal ofForestry, pp.43-48.

Saner, E., Eronat, A.H., Basoz, C., and Uslu O., 1998, An Economical New Approach to AirborneVideography, Case Study: Aegean Region of Turkey, May 1998, Symposium on Operational RemoteSensing for Sustainable Development, ITC, Enschede, The Netherlands (Unpublishd).

Thomasson, J.A., Bennett, C.W., Jackson, B.D., and Mailander, M.P., 994, Differentiating Bottomland Treespecies with Multispecral Videography, PE&RS, Vol. 60, No: 1, pp.55-59.

Vlecek, J., 1983, Videography: Some Remote Sensing Applications, Proc. 50th Meeting ASPRS, pp., 63-69.

3.5 Application of Laser (Lidar) in Forestry

The applications of two types of laser (Lidar) sensors will be discuss, which are the most potential andpromising in the forestry applications Bathymeter and Fluorescence. The interesting issue of the two types ofLidar system that each one of them used to a certain different application because of the their differentfunction that they accomplished.

3.5.1 The Applications of Bathymeter Laser System

1) Basically this system work as explained before send the pulse or the wave and receive it, therefore thetime which that pulse take to hit the target and return back to the system account and use to convert to adistance. A distance profile then establish for this purpose.

2) Another more accurate way that can be used to handle the result is to use the direct time curve.


Tree Height

The first method that explained above is used when the forest stand has breaks frequently (e.g. homogeneousstand height),so the distance profile will be exactly like the topographic profile. The profile is considered tobe the height of the object in this case, example of this method is shown in Figure 1 (Apr et al., 1982) and(Aldred and Bonnor 1985).

The second method which depend on the direct sequential profile curves that the laser system produced. Thismethod is mostly used by the researchers. laser profile data used to give two peak for each object. The firstpeak is the strong one that come from the target and the second weak one come from the next layer i.e. theground or it can be the under story vegetation (Nelson et al 1984) and (Aldred and Bonnor 1985). Thedifference between the two peak can be used as the tree height in this case, example of this data shown inFigure 2. This method can be relay on, in case where breaks happen very frequently. Figures 3 and 4 showtypical returns of two cover types deciduous and coniferous. This give higher accuracy than the first one. Theprecision in the first method according to Apr et al., (1982) + 3 meter while in the second method it can beup to a meter and that will still depend on the following factors:

1. Flight Height (500 meter or even less give much more accuracy than a data collected from 850 m asl).

2. Number of the events (e.g. times that pulses send to the target along the profile, the high number ofevents the accurate the results).

3. Laser beam divergence in mR. (e.g. laser beam width, the wider the beam the more accurate results).

4. The cover type (e.g. Deciduous or coniferous).

5. The season ( e.g. leaf off or leaf on specifically for the deciduous cover type).

According to Aldred and Bonnor (1985), the linear regression results show low residuals and highcorrelation between the predicted heights and the actual tree heights and when they plotthe two data sets mentioned above the got very closed results.

See Figures 5, 6, 7 and Table 1 that show the above results. Figure 8 shows the same results that Nelson et al(1988) obtained for the tree heights prediction. Moreover, the source of the error that happened in theprediction of the tree height most probable because the laser energy can be reflected from the point on thecrown of the tree (e.g. flanks of the tree) other than the top point (e.g. peak of the tree crown) and thatexplained in Figure 9. Also Nelson et al a 1988 show the same under estimate of the tree heights, Figure 10explain this differences.

Crown Cover Density

This method has been explained by Aldred and Bonnor (1985). They stated that the following variables usedto estimate the crown cover density of the stand:

1. Missing pulses: percentage of the return pulses.2. Peak 1 : average maximum amplitude of return from canopy.3. Peak 2 : average maximum amplitude of return from ground.4. Amplitude ratio: average ground to canopy amplitude.5. Number of peaks: average number of peaks per multiple return6. Total area: average total area under the pulse return.7. Ground area: average area under ground portion of multiple return.8. Canopy area: average area under canopy portion of multiple return.9. Height: average height of stand.


The derivation of some of these variables are shown in Figure 11. The control data for crown cover densitywas determined photogrammetrically. The results of this prediction then compared statistically with themeasurement done used the aerial photos and the residuals and the correlation shown in Table 2.

While Nelson et al. 1984 found that more than 60% of the variance in Photogrammetrically estimated canopydensities could be explained by two or more laser variables using multiple linear regression techniques.

Forest Cover Type

Aldred & Bonnor (1985) used the method of "along each f light line for each stand the strip corresponding tothe laser's footprint examined by photo- interpreter and categorized into one of four standard cover typeclasses". Then the statistical effectiveness of the result has been tested and they found that only thetwo-peaks return can have a correlation with the cover type. Moreover best classification accomplished whenthey used 10 mR laser beam data. However Nelson et al 1984 used the almost the same method for thispurpose and they had three different type of laser profile in which each cover type was related to one of them(e.g. undefoliated forest, shrubs, and defoliated forest.

Forest Tree Volume and Biomass

The procedure to predict these parameter was explained only by Nelson et al a 1988 and Nelson et al.,(1988). For both papers they basically used the same laser and ground variables and the same equation also.Briefly the variables that used for the laser prediction of volume and biomass are:

a) Laser Height Variables

1. LAVG3B average of the three largest laser canopy heights.

2. LHT mean plot height in which all pulses considered includes pulse direct to ground. this variable is directly proportional to the canopy profile area.

3. LHTNZ mean canopy height. Only pulses which intercept the canopy are considered here.

4. MCP2 modified canopy profile with a 2-M exclusion limit; that area between the top of the canopy and a line drawn 2M above ground trace.

5. MCP5 modified canopy profile with-a-5 M exclusion limit.

6. MCP10 modified canopy profile with a 10-M exclusion limit.

b) Laser Canopy Density Variables

7. INTER the interaction term of the two laser variables which infer canopy density.

The ground variables that used in this case are :

1. TTV a ground estimate of the total tree volume.2. TTGW a ground estimate of the total tree green weight.


Volume Estimation

Laser height variables and laser canopy density were tested to determine which will be the best estimator tothe forest volume. For this reason the following two equation used to this purpose:

In(TTV) = m (laser height metric)+n(laser density metric)+b

In(TTV) = m (laser height metric) +n{ In (laser height metric)}+b

Where TTV equal the total tree volume.

Biomass Estimation

Laser height variables and laser canopy density were tested to determine which will be the best estimator tothe forest biomass. For this reason the following two equations used to this purpose:

In(TTGW) = m (laser height metric)+n(laser density metric)+b

In(TTGW) = m (laser height metric) +n{In (laser height metric)}+b

Where MG equal the total green weight that used as a total biomass.

The results indicated after applying the above two methods that total tree volume and total tree biomass canbe estimated. "Based on two random samples test consisting of 38 test plots each, the best model predictedtotal tree volume within 2.6% of the true mean value, and biomass within 2.0%. Figure 12 shows therelationship between ground measurements of the volume and biomass and the laser predicted volume andbiomass.

3.5.2 Flunrencence laser_System

Basically this system as explained before send the laser energy that hit the target which will reemit theenergy after it reach the saturation point and the remitted energy will reflect the characteristics of thematerials of that object or target. It is usually called LIF Laser-Induced Fluorescence.

Species Identification

For this purpose laboratory experiment are the most methodology place used. In which they prepare theplants inside the lab and expose it to the laser energy and then record the results digitally and graphically.Chappelle et al., (1985) were able to get separate reflectance curves for seven different forest tree speciesthree deciduous and the other four were coniferous. Figure 13 show the signature or the spectral reflectanceof the four different forest tree species which are. The statistical analysis of the spectral reflectance valueshow significant difference between the coniferous and the deciduous forest tree spectra. Using airbornelaser fluorescence sensor system to detect and differentiate plants species, Hoge et al., (1983) found it isfeasible to detect four different type of vegetation (i.e. tree, shrub, grass, and herbaceous plants) using thelaser fluorescence sensor.


Plant stress

For this purpose as the previous one the researchers depend on the pigments and chlorophyll response to theradiation that the plants receive. Chappelle et al., (1984a) reported some species differentiation and plantstress of some herbaceous plants. They used laser fluorescence to identify the Corn healthy plants from theplants the suffer from the nutrient deficiencies. McFarlane et al., (1980) studied the laser fluorescencereflection from Citrus trees suffering from water stress. They found that an increase in fluorescence wascorrelated with the Citrus plant water stress as measured by stomata resistance and twig water potential.

Conclusion

Laser Bathymeter and Induced-Fluorescence types lidar sensor that used mostly in the field of remotesensing applications.

Laser remote sensing systems have been successfully used in several environmental and natural resourcesapplications.

Laser airborne sensor found very useful and can accomplished the following aspects in the forestry survey:

1. Tree height prediction.2. Forest cover types determination.3. Tree species differentiation.4. Crown cover or canopy density.5. Tree volume prediction.6. Tree woody biomass prediction.7. Tree water stress detection.8. Vegetation nutrient deficiency.

Since the reflected data profile show two kind of curve one is the forest and the other is the backgroundwhich usually weak, it is very possible that using the second curve to predict some information about theunderstory vegetation or at least about the forest ground.

Forest stand parameters should be correlate with the laser data from forest cover to see if there are moreinformation can be derived or modelled from the laser data.

Since the laser fluorescence was very beneficial in water stress and nutrition deficiency to the citrus tree andother plants, one can expect that this type of data may be very useful for these purposes and may more (e.g.forest disease and insect cases detection) to the forest tree.

There are some limitation that should be mentioned:

For Laser Bathymeter:a. Flight height must be less than 500 meter.b. The number of the events are mostly the important variable which determine the accuracy (e.g. Increase the number means more data collected and that reduces the experimental error) of the result and represent the observations in the field of the experimental design.c. Laser beam width insure more accurate date that will registered or scattered from the targets.


3.5.3 Literature Cited

Abreau,V.J.1980. "Lidar From Orbit" 19:489-493. Optical Engineering.

Aldred, A. H. and G. M. Bonnor.1985. "Application of Airborne Laser to Forest Surveys". InformationReport PI-X-51. Petawawa National Forestry Institute. Canadian Forest Survey. 62 pp.

Apr,H,Greisbach,J.C.,Tranrg, C.A. and J. P. Burns. 1982. "Mapping in Tropical Forest:A New ApproachUsing the Laser APR". Photogrammetric Engineering and Remote Sensing.Vol.XLVIII, No.1:91-100.

Bristow,M. P. 1979. "Fluorescence of Short Wavelength Cutoff Filter Applied Optics. 18:952-955.

Chappelle, M. P, McMurtrey, J. E., Wood, F. M. and W.W.Newcomb.1984."Laser-Induced Fluorescence ofGreen Plants. 2: LIF Caused by Nutrient Deficiencies in Corn". Applied Optics. 23:139-142.

Chappelle, E.W., Wood, F. M., McMurtrey,J.E.andW.W.Newcomb.1984."Laser Induced Fluorescence ofGreen Plants. 1: A technique for the Remote Detection of Plant Stress and Species Differentiation". AppliedOptics. 23:134~138.

Chappelle, E.W., Wood, F. M., Newcomb, W. W. and Mcmurtrey.J.E.1985. "Laser-Induced Fluorescence ofGreen Plants. 3: LIF Spectral Signature of Five Major Plants Types". Applied Optics. 24:74-80.

Cooney,J.A.1971. "Remote Measurement of Atmospheric Water Vapour Profiles Using Raman Componentof Laser Backscatter". Journal of Applied Meteorology.10:301-308.

El-Sayed,S.Z.1970."Phytoplankton Production of the south Pacific Sector of the Antarctic". ScientificExploration of the south pacific, National Academy of Sciences. From: R. M. Measures. 1984.Laser RemoteSensing. John Wiley & Sons. 510 pp.

Hoge, F. E., Swift, R. N. and J.K.Yungel.1983. " Feasibility of Airborne Detection of Laser-InducedFluorescence Emissions From Green Terrestrial Plants". Applied Optics.22:2991-3000.

Johnson,C.M.1970. "Laser Radar". From: Skolnik.M.I.1970. "Radar Handbook ".McGrow-Hill BookCompany.(37-1)-(37-69) pp.

Kennie,T.J.1985. "Remote Sensing Scanning System". From: Kennie & Matthews.1985. Remote Sensing inCivil Engineering. John Wiley & Sons. 82-85 pp.

Leonard,D.A.1967."Observation of Raman Scattering from Atmosphere Using Pulsed Nitrogen UltravioletLaser". From: R. M. Measures. 1984. Laser remote Sensing. John Wiley & Sons. 510 pp.

Lillesand, T. M. and R.W. Kiefer. 1987. Remote Sensing and Image Interpretation". 2nd Edition. John Wiley& Sons. 525-527 pp.

Link, L. E. and J.G.Collins.198l."Airborne Laser Systems use in Terrain Mapping". 15th InternationalSymposium on Remote Sensing of Environment. Ann Arbor MI. 11-15 May 1981.1:95-110.

Lowe,D.S.1980."Acquisition of Remotely Sensed Data From: B. S. Siegal, Remote Sensing in Geology.John Wiley & Sons. 47-90 pp.

Lowry, R. T. and C.J.Brochu.1979."An Interactive Correction and Analysis System for Air-Borne LaserProfiles of Sea Ice". Canadian Journal of Remote Sensing. Vol.4 No. 2. August 1979.


McFarlane.J.C.,Watson,R.D.,Theisen,A.F.,Jackson,R.D.,Ehrler,W.L-Pinter,P.J.,Idso, S.B.andR.J.Reginato.1980. "Plant Stress Detection by Remote Sensing Measurement of Fluorescence". AppliedOptics. 19:3287-3289.

Measures,R.M.1984."Laser Remote Sensing: Fundamentals and Applications". John Wiley & Sons. 510 pp.

Melfi,S.H.1972."Remote Measurements of the Atmosphere Using Raman Scattering". Applied Optics.11:1605-1610.

Mulders,M.A.1987. "Remote Sensing In Soil Sciences". Elsevier Publishing Company. 314-315 pp.

Nelson, R, Krabill, W. and G.Maclean.1984."Determining Forest Canopy Characteristics Using AirborneLaser Data". Remote Sensing of Environment. 15:201-212.

Krabill, W. and J.Onelli.1988. 'Tstimating Forest Biomass and Volume Using Airborne Laser Data". RemoteSensing of Environment. 24:247-267.

Swift, R. and W.Karbill.1988. "Using Airborne Laser to Estimate Forest Canopy and Stand Characteristics".Journal of Forestry. October 1988. 31-38 pp.

Northam, D.B.,Guerra, M. A., Mack, M. E., Itzkan, and I. Deradourian.1981. "High Repetition RateFrequency-Doubled Nd: YAG Laser for Airborne Bathymeter". Applied Optics. 20:968-971.

Pal, S. R. and A.I.Craswell.1973. "Polarization Properties of Lidar Backscattering from Cloud". AppliedOptics. 12:1530-1535.

Sandford,M.C.W.1967."Laser Scatter Measurements in the Meosphere and above". From: R. M. Mea Bures.1984. Laser remote Sensing. John Wiley & Sons. 510 pp.

Suits, G. 1983. "The Nature of Electromagnetic Radiation". from: N. R. Colwell. 1983. Manual of RemoteSensing. ASPRS. 37-60 pp.

Williamson,S.J.1973."Fundamental of Air Pollution". from: R.M.Measures.1984.Laser Remote Sensing.John Wiley & Sons. 510 pp.

3.6 Combined use of optical and radar satellite images for forestry applications

Multisensor data have been combined for a number of years to help obtain the synergistic effect thatimproves classification results. Combining Landsat MSS or TM with Spot XS or panchromatic data is anearly example using optical data (Ehlers, 1987). Merged multiband radar data such as L, P, C, and K havealso been used for a number of years (Rosenthal, 1981). Optical and microwave data were combined evenearlier, including work in analog format (Daily et al., 1979) to identify landform characteristics. Digitalformats were used later (Welch, 1984; Evans, 1986).

Image transformations are arithmetic operations that allow us to generate a new composite image from twoor more images (e.g. multispectral, multitemporal, multifrequency, multipolarization, multiple incidenceangle). The main reasons for image transformations are to extract information that was not clear or directlyavailable in the original images and to reduce data dimensionality, which will reduce space, time, effort andcost of image analysis. Correlation of different bands of images (multispectral or multitemporal) implies thatthere is redundancy in the data. Image fusion is the most recent technique. It combines or mergesmultisensor, multitemporal, multispectral, etc. images, and also transforms them to reduce the size of thedata set and enhance the data so we can extract most of the information available for a particular purpose(Welch and Ehlers, 1987; Welch and Elhers, 1988; Ehlers, 1988).


Leckie, (1990) have used airborne radar (X-VV, X-VH, X-HH, X-HV, C-VV, C-VH, C-HH, C-HV) andairborne MSS (five visible, three near-infrared, one mid-infrared band) in a study to differentiate forestspecies (both hardwood and softwood) in Ontario, Canada. Natural Stands. Best discrimination of specieswas on accuracy 74%..

Nezry et al., (1993), used SIR-B data and SPOT data of central Sumatra for land cover classification: water-burnt/unclassified-bare soil-rubber-crops-regrowth-deforested. Overall classification accuracy was far betterthan 75 %. Only the distinction between bare soil and burnt/unclassified encounters some problems.

Hussin and Shaker, (1996) using Landsat TM and Shuttle Imaging Radar-B images of Northern Florida,USA 14 land cover classes (including different types of natural forests and different ages of plantationforest) were found with overall accuracy with three SIR-B bands 88%, overall accuracy with seven TMbands 91.6 % and overall accuracy for different combinations of TM + SIR-B: 92.5 %.

Hussin and Shaker, (1996) used Landsat-TM (1990) in combination with ERS-1 (1993), JERS-1 (1993,1994) for an area of tropical forest in Central Sumatra, Indonesia and reported that with a combination ofTM and ERS more land cover classes could be distinguished with visual interpretation than on either one ofthe images alone.

Paris and Kwong, (1988) combined Thematic Mapper (TM) and Shuttle Imaging Radar (SIR-B) image dataof an area in Fresno County, California, USA.Land cover classes including crops and different types oforchards were classified. They concluded that the sensors provided quantitative information on the amountsof herbaceous and woody vegetation.

Ehlers, (1991) investigated the multi-image fusion techniques of Landsat TM and SIR-B data.He concluded that more cartographic information could be obtained from fused TM & SIR-B images thanfrom each of the images alone.

Literature Cited

Daily, M. T., Farr, T., Elachi, C. and Schaber G. 1979. Geologic interpretation from composite radar andlandsat imagery". Photo. Eng. & Remote Sensing, 45:1109-1116.

Ehlers, M. 1987. "Integrative Auswering von digitalen Biddaten aus der Satellitenphotogrammetrie und-Fernkundung in Rahmen von Geographischen Informationssystemen". Universitat Hannover Pub. 149, 139pp.

Ehlers, M. 1988. "Multisensor image fusion techniques in remote sensing". Proc. of the XVII Congress ofISPRS, Kyoto, Japan, Vol B7 pp 152-162.

Ehlers, M., 1991: Multisensor image fusion techniques in remote sensing. ISPRS Journal of Photogrammetryand Remote Sensing, 46, pp 19-30.

Evans, D. L. 1986. "Geological applications of multipolarized SAR data". Proc. 2nd Spaceborne ImagingRadar Symposium. JPL. Pasadena, CA. JPL Pub. 86-26, pp. 36-41.

Hussin, Y.A., S.R. Shaker, 1996: Optical and radar satellite image fusion techniques and their applications inmonitoring natural resources and land use changes. AEÜ International Journal of Electronics andCommunications Vol. 50, No. 2, pp 169-176.

Leckie, D.G., 1990: Synergism of synthetic aperture radar and visible/infrared data for forest typediscrimination. Photogrammetric Engineering and Remote Sensing, Vol. 56, no. 9, September 1990, pp.1237-1246.


Lozano-Garcia, F.L., 1988. "Analysis of synergistic effects of a combined spaceborne multispectral scannerand L-band multiple incidence angle radar data set. Ph.D. Dissertation. Dept. of Forestry and NaturalResources, Purdue University. 241 pp.

Rosenthal, W. D. 1981. "Development of visible/infrared/microwave agriculture classification and biomassalgorithms". Ph.D. Thesis, Kansas State University. 213 pp.

Nezry, E., E. Mougin, A. Lopes, J.P. Gastellu-Etchegorry, Y. Lamonier, 1993: Tropical vegetation mappingwith combined visible and SAR spaceborne data. Int. J. remote Sensing, Vol. 14, No. 11, pp.2165-2184.

Paris, J.F., H.H. Kwong, 1988: Characterization of vegetation with combined Thematic Mapper (TM) andShuttle Imaging Radar (SIR-B) image data. Photogrammetric Engineering and remote sensing, Vol. 54, No8, August 1988, pp. 1187-1193.

Welch, R. 1984. "Merging Landsat and SIR-A images data in digital formats". Imaging Technology andResearch and Development, July, pp. 11-12.

Welch, R. and Ehlers M. 1987. "Merging multiresolution SPOT HRV and Landsat TM data". Photo. Eng.and Remote Sensing, (53): 301-303.

Welch, R. and Ehlers M. 1988. "Cartographic feature extraction with integrated SIR-B and Landsat TMimages". Int J. Remote Sensing. Vol 9 No. 5, pp 873-889.


4. CURRENT AND FUTURE SATELLITS

4.1 Current Satellites

Tables 1 through 5 show some important characteristics of current optical and radar satellite systems.

4.2 Satellites of the Next Decade

The next decade promises an explosion in the quantity and quality of global land data available fromsatellites. If all the government and commercial land viewing satellite systems orbit as currently scheduled,in the year 2000 a minimum of 19 satellites or maximum of 31 satellites will be in polar orbit providing landdata at resolutions from 1 to 30 meters in panchromatic, multispectral, and radar formats. Table-6 belowshows some of the land data satellites currently planned for the next decade.

4.3 The Future of Satellite Remote Sensing

Short (1997) prepared a study on the future of remote sensing in the next decade. The study includes theplanned satellites for next decade. See the study of Short (1997) in Appendix 1.

TABLE-1 PLATFORM AND ORBIT CHARACTERISTICS OF LANDSATDescription Landsa

t1Landsat -2 Landsat -3 Landsat -4 Landsat -5 Landsat -6 Landsat -7 Landsat -

8Satellite operator NOAA NOAA NOAA EOSAT,

USAEOSAT,USA

EOSAT Co,USA

Lunch Date 23/06/72

05/03/78 16/07/82 01/03/84 5/10/93 N/A

CCRS Reception 7/08/72-26/10/77

l 9/04/75-7/02/82

17/05/78 -07/02/83

17/08/82-01/09/87

TM 1986,MSS 1984

N/A N/A

Mission Date 23/06/72 --02/01/78

22/01/75 -22/01/80

05/03/78to 7/01/83

16/07/82Satelliteturned offin Jan’ 86& keptstandbyStoppedtrack- ing 1/9/87

01/03/84 –to future

5/10/93*Satellitefailed toorbit,contactlost duringlaunch

January1999(proposed)

2003(proposed)

GroundResolution

80 m 80 m 40m(RBV)80m(MSS)120 m forThermalband

75 m(MSS)30 m TM120 mband6

75 m forMSS &30 m forTM, butfor Band6-120m

ETM :PB 13 x15 mPMB30mTB 120m

ETM30 m (MB)120m(SWTB)60m(MTB)15 m (PB)ALS10 m(VNIR)20m(SWIR)

ETMsame asLandsat-7

Swath Width 185km

185 km 185 km 185 km 185 km 185 km(MB)41 km


(ALS)Altitude 918

km918 km 918 km 705 km 705 km

Orbit type sun-synchrono-us

sun-synchronous

sun-synchronous

sun-synchronous

Nearpolar –sun-synchronous

Sun-synchronous

Sun-synchronous(proposed)

Orbit time - - - - 99 min. - -Orbits per day 14 14 14 14.5 14.5 - -Number of orbits 251 251 251 233 233 - -Repeat cycle 18

days18 days 18 days 16 days 16 days 16 days -

Image sidelap atequator

14% 14% 14% 7.6% 7.6% -

Crosses 40 N°latitude(at local suntime)

09:30am

09:30 am 09:30 am 10:30 am 10:30 am -

On board datastorage

Yes Yes Yes No No -


Imaging Systems :Status Turne

d off31/12/72

Removedfrom serviceon 25/02/82

Removed in 01/28/81 &fix againin13/04/81on Limitedbasis

off Band 4turnedoff.permanently rest 3bandsnominal

- N/AMSS

Bands

Therm-alBands

0.5-0.6µµm0.6-0.70µµm0.7-0.8µµm0.8-1.1µµm-

0.5-0. 6µµm0.6-0.70µµm0.7-0. 8µµm0.8-1.1µµm-

0.5-0.6µµm0.6-0.70µµm0.7-0.8µµm0.8-1.1µµm10.4-12.6µµmfailed3/3/79

4) 0.5-0.6µµm5) 0.6-0.70µµm6) 0.7-0.8µµm7) 0.8-1.1µµm

4) 0.5-0.6µµm5) 0.6-0.70µµm6) 0.7-0.8µµm7) 0.8-1.1µµm

- *MTB(Multispectral ThermalBands)1) 8.2-8.75µµm2) 8.75-9.3µµm3) 10.2-

11.0µµm4) 11.0-

11.8µµm

Status - - - off Nominal N/ATM/ETM

Bands - - - 1).45-. 52µµm2).52-.60µµm3).63-.69µµm4).76-.90µµm5)1.55-1.75µµm6)10.4-12.5µµm7)2.08-2.35µµm

1).45-. 52µµm2).52-.60µµm3).63-.69µµm4).76-.90µµm5)1.55-1.75µµm6)10.4-12.5µµm7)2.08-2.35µµm

PB 0.50-0.90µµmPMB1) 0.45-0.52µµm2) 0..52-0.60µµm3) 0.63-0.690µµm4) 0.76-0.90µµm5) 1.55-1.75µµm6) 2.08-2.35µµmTB 10.4-12.5µµm

MB1).45-. 52 µµm2).52-. 60µµm3).63-. 69µµm4).76-. 90µµm5)1.55-1.75µµm6)2.08-2.35µµmSWTB 3.53-3.93µµmPB 0.5-0.9µµmALS: 32 bands (16 VNIR,16SWIR)

StatusTurned off31/12/72

Removedfrom serviceon 25/02/82

No - - - *ALS(AdvancedLandsatSensor)

RBV

Bands .475-.575µµm0.58-0.68µµm0.58-0.68µµm

.475-.575µµm0.58-0.68µµm0.58-0.68µµm

0.5-0.75µµm

- - - -

Source: Sabins (1978); Australian Centre for Remote Sensing on the Web http://ww.auslig.au/acres/facts.htm#top;http://www.ccrs.nrcan.gc.ca/ccrs/tekrd/satsens/sats/landsate.html


TABLE -2 PLATFORM AND ORBIT CHARACTERISTICS OF SPOT (SYSTE′ME POUR I ′OBSERVATION DE LATERRE) HRV (HAUTE RESOLUTION VISIBLE) HIGH-RESOLUTION VISIBLE SYSTEM

Description SPOT-1 SPOT-2 SPOT-3 SPOT-4 SPOT-5Satellite operator CINES/IGN

FranceCNES France

CNES France

CNES France

CNES France

Launch date 22/02/1986 21/01/ 1990 26 /09/ 1993 March 1998 -

Mission date 22/02/1986 21/01/ 90- till now 26 /09/ 1993 March 1998

CCRS Reception 17/05/86-10/06/90,08/04/93-02/08/93stopped trackingon 02/08/93

January 1991 28/03/94-14/11/96(satellite declaredlost as of thisdate)

N/A

Altitude 830 km 830 km 830 km -

1999(proposed)Plannedcapabilities include5mresolutionpanchromatic &10 mresolutionmultispectral data.

Orbit type near polar, sunsynchronous

near polar, sunsynchronous



sun-synchronous(proposed)

Orbit period 101 minutes 101 minutes -

Repeat cycle 26 days 26 days 26 days -

Repeat cover 4-5 days whensensorsprogrammedmax. off nadir

4-5 days whensensorsprogrammedmax. off nadir

4-5 days whensensorsprogrammedmax. off nadir

-


Sensor type &Status

HRV1 & HRV2(N/A)

HRV1 & HRV2(Nominal)

HRV1 &HRV2POAM2 &DORIS(N/A)

HRV1 & HRV2

Imaging Systems Multi-spectralMode

Panchromatic Mode

Multi-spectralMode

PanChromaticMode

Multi-spectralMode

Panchromatic Mode

Multi-spectralMode

Panchromatic Mode

VMIVegetation.Monitoring.Instrument.

Ground resolution 20m 10m 20m 10m 20m 10m 20m 10m 1km

Swath Width 60 km 60km 60 km 60 km 60 km 60 km 60 km 60 km 2000km

Spectralbands(microns)

1)0.50-0.592)0.61-0.683)0.79-0.89

1)0.51-0.73

1)0.50-0.592)0.61-0.683)0.79-0.89

1)0.51-0.73

1)0.50-0.592)0.61-0.683)0.79-0.89

1)0.51-0.73

1)0.50-0.592)0.61-0.683)0.79-0.894)1.58-1.75

1)0.61-0.68

1)0.43-0.472)0.50-0.593)0.61-0.684)0.79-0.895)1.58-1.75

*PASTEC**PASTEL***DORIS*PassagerTechnolo-gique**PassegerSPOTdeTelecomm-unicationlaser***DopplerOrbitograp-hy andRadioposit-ioningIntregratedbySatellite

Source: Sabins (1978); Borel (1985); http://www.ccrs.nrcan.gc.ca/ccrs/tekrd/satsens/sats/landsate.html AustralianCentre for Remote Sensing on the Web http://ww.auslig.au/acres/facts.htm#top


TABLE-3 PLATFORM AND ORBIT CHARACTERISTICS OF RADARSAT

DESCRIPTION RADARSAT-1 RADARSAT-2 RADARSAT-3

SatelliteOperator

RADARSAT INTERNATIONALCANADA

RADARSATINTERNATIONALCANADA

RADARSATINTERNATIONALCANADA

Mission date 04/11/1995 2001 (proposed) 2004 (proposed)

Date testreceptioncommenced

March 1998 N/A N/A

Altitude 798 km(average) - -Orbit type Circular/sun-synchronous

(descending/dawn, ascending/dusk)Earthobservation/near-polar (proposed)

Earthobservation/near-polar (proposed)

Orbit period 100.7 minutes - -

Repeat cycle 24 days - -

Current crossingtime

descending 0600 hrs, ascending 1800hrs

- -

RadarsatPayload

SAR(Synthetic Aperture Radar)• Status: Nominal• Frequency-5.3 Ghz (C-Band)• SAR will have 6 different

operating mode

Same asRADARSAT-1

*The satellite wouldbe designed for a 10year lifetime.

Incidenceangle (°°)

Resolution(m)

Swath(km)

- -

Standard 20-40 28××25 100 - -

Wide Swath 20-39 28××35 150 - -

Fine Resolution 37-48 10××9 45 - -

ExentedCoverage

49-59 20××28 75 - -

ScanSAR(narrow)

20-3931-46

50××5050××50

300 - -

ScanSAR (wide) 20-49 100××100 500 - -

Source : Sabins (1978); Borel 1985; http://www.ccrs.nrcan.gc.ca/ccrs/tekrd/satsens/sats/landsate.htmlAustralian Centre for Remote Sensing on the Web http://ww.auslig.au/acres/facts.htm#top


TABLE-4 PLATFORM AND ORBIT CHARACTERISTICS OF ERS(EUROPEAN RMOTE SENSING SATELLITE)

DESCRIPTION ERS-1 ERS-2Satellite Operator ESA ESA

Mission date 17 July 1991 20 April 1995

Date receptioncommenced

17 July 1991 785 km

Altitude 785 km 785 kmOrbit type near polar, sun synchronous near polar, sun synchronousOrbit period 100 minutes 100 minutes

Repeat cycle 3 day cycle(17/7/91-1/4/92)35 day cycle (2/4/92-22/12/93)3 day cycle 23/12/93-9/4/94)168 day cycle (10/4/94-20/3/95)35 day cycle (21/3/95-3/6/96)

35 dayspositioned one day behindERS-1 on the same track

Current crossingtime

1000 - 1030 hrs local standard time 1000 - 1030 hrs local standardtime

ERS Payload AMI (Active Microwave Instrument) ATSR-M Same as ERS-1 with additionof GOME & AATSR

Sensor type/Altimeter

SAR-ImageMode

SAR-WaveMode

ScatterometerMode

RadarAltimeter

InfraredRadiometer

MicrowaveSounder

GOME(Global OzoneMonitoringExperiment)

AATSR(AdvanceAlongTrackScanningRadiometer)

Resolution <30 m <30 m 50 m 10 cm(accuracy)

1km ××1km

22km 40 km ××2 km40 km××320 km

0.5km

Swath (km) 80-100 5 500 N/A 500 500 960 500

Frequency 5.3GHz 5.3GHz 5.3GHz 13.5GHz

- - - -

Bands C-Band C-Band) C-Band KU-Band

1.63.71112

23.536.5

1) 0.24-0.295µµm (512bands)2) 0.29-0.405µµm (1024bands)3) 0.40-0.605µµm (1024bands)4) 0.59-0.79µµm (512bands)

0.650.851.271.63.711.012

Source: Sabins (1978); Borel (1985); http://www.ccrs.nrcan.gc.ca/ccrs/tekrd/satsens/sats/landsate.html AustralianCentre for Remote Sensing on the Web http://ww.auslig.au/acres/facts.htm#top


TABLE -5 PLATFORM AND ORBIT CHARACTERISTICS OF JERS-1(JAPAN EARTH RESOURCES SATELLITE)

DESCRIPTION JERS-1Satellite Operator NASDAMission date 11 Feb 1992First SAR Image Received 21 April, 1992Altitude 568 kmOrbit type near polar, sun synchronousOrbit period 96 minutesRepeat cycle 44 daysCurrent crossing time 1000 - 1030 hrs local standard timeJERS-1 PayloadSensor type SAR(Synthetic

Aperture Radar)OPS(Optical Sensor)

Resolution 18m××18m 18m××18m Swath (km) 75 75Frequency 1275 MHz -Bands L-band

• *band 4 is for forward viewing (15.33°)• band 3 and 4 make a stereo-pair• bands (5,6,7,8) not available

1)0.52-0.60 µµm2)0.63-0.69 µµm3)0.76-0.86 µµm4)0.76-0.86*µµm5)1.60-1.71 µµm6)2.01-2.12 µµm7)2.13-2.25 µµm8)2.27-2.40 µµm

Source : Sabins (1978); Borel 1985; http://www.ccrs.nrcan.gc.ca/ccrs/tekrd/satsens/sats/landsate.htmlAustralian Centre for Remote Sensing on the Web http://ww.auslig.au/acres/facts.htm#top


TABLE –6 LAND DATA SATELLITES CURRENTLY PLANNED FOR THE NEXT DECADEResolution in MetersCountry Owner/

OBJ (1)Program Schedule

d dateInstrument Type(2)

P M R# ColorsBands

StereoType (3)

France G/O Spot 5B ‘04 P & M 5 10 - 4 F/A

U.S G/O EOSAM-2/ L-8

‘04 P & M 10 30 - 7 -

U.S G/O Landsat-8

‘03 P & M 15 30 - 7 -

France G/O Spot 5A ‘99 P & M 5 10 - 4 F/A

India G/O IRS-1 D ‘99 P & M 10 20 - 4 C/T

U.S C/O SpaceImaging

‘98 P & M 1 4 - 4 F/A

Korea G/O KOMSAT

‘98 P & M 10 10 - 3 F/A

U.S/Japan

G/O EOSAM-1

‘98 M 15 15 - 14 F/A

U.S G/O Landsat-7

‘98 P & M 15 30 - 7 -

ESA G/O ENVISAT

‘99 R - - 30 0 -

U.S C/O SpaceImaging

‘97 P & M 1 4 - 4 F/A

U.S C/O Eyeglass ‘97 P 1 - - - F/A

France G/O Spot-4 ‘98 P & M 10 20 - 4 C/T

U.S C/O EarthWatch

‘98 P & M 1 4 - 4 F/A

U.S C/O EarthWatch

‘97 P & M 3 15 - 3 F/A

Notes: (1) Owner / OBJ: G= Government Funded, C=Commercially Funded, O=Operational,E=Experimental(2) Inst. Type: M=Multispectral only, P & M=Panchromatic & Multispectral, P=Panchromatic,R=Radar(3)Stereo-Type: F/A=Fore & Aft Stereo, C/T=Cross Track Stereo. Source: American Society forPhotogrammetry and Remote Sensing, 1995http://www.ccrs.nrcan.gc.ca/ccrs/tekrd/satsens/sats/landsate.html


5. OPTIMAL RADAR AIRBORNE AND SATELLITE SYSTEMS FOR HIGH-RESOLUTIONFOREST OBSERVATION

5.1 Systematics approach to determine optimal radar satellite configurations

Many trade-offs have to be made to determine on optimal configuration for a radar monitoring satellite. Inthe first place a certain balance has to be found between the role of satellite and aircraft remote sensing. Inprinciple radar satellites can give a lot of information. However, higher spatial and thematic detail may comewith a disproportionate increase of costs. Since detailed observations are usually required at specific areas ofinterest the role of satellite monitoring may be limited (role 1:) to collect information that can be retrievedrelatively accurate and easy (i.e. affordable) and (role 2:) to detect timely those areas of specific interestwhere more detailed observations are required. It is understood that the latter point is not easily evaluatedsince accessibility to and affordability of airborne remote sensing systems varies strongly, bothgeographically and from user to user.

Timely detection of areas of interest (role 2) requires a sufficient observation frequency. Consequently, thisconstraints spatial and radiometric resolution. Observations at the tree level are unlikely (unaffordable) to bemade with satellites. At the best we can hope for detection of small clear-cut areas or expansion of timberroad networks (van der Sanden, 1997). Timely observations at individual tree level are likely to be madesufficiently well with airborne radar (see section 2). Satellites are likely to be useful for monitoringdeforestation, forest and land degradation, land cover change and secondary re-growth (Hoekman andQuiñones, 1997, 1998).

The design of such a monitoring mission is not straightforward since many trade-offs have to be made. Atthe one hand the system has to be simple (to keep it affordable) at the other hand a certain accuracy orrobustness is required. To study this it is proposed to use the AIRSAR C-, L- and P-band fully polarimetricdata of the Colombian Amazon and evaluate certain scenario’s. Within these scenario’s four major landcover types play an important role. These are: primary forest, secondary forest, recently cut forest andpastures.

Results are summarised in a series of tables and figures. Table 1 shows the data set used for the evaluation.Since the incidence angle dependence was not found to be very high, the data in the 45-50 degrees incidenceangle range were used for the assessment of appropriate configurations, while the remaining data were usedfor validation. To evaluate the effect of spatial and radiometric resolution, which are interchangeable to alarge extent, the number of looks per unit area was taken as the main parameter (Table 2). Three cases havebeen studied in particular: i.e. the case where analysis can take place at the level of spatial units (imagesegments) with a sufficient number of ‘looks’ allowing the effect of speckle to be ignored (the ‘0 dB’ level),and for the situation that analysis takes place at the pixel level with 5- and 20-look data (the ‘2 dB’ and ‘1dB’ speckle level, respectively). KHAT statistics are used to indicate significance of classification results,allowing, for example, comparison of different results with the best and worst result.

The simulated classification results for the four main land cover types using a single channel configurationare shown in Table 3. At the ‘1 dB’ level only L-band HV-polarisation (Lhv) and P-band HV-polarisation(Phv) show reasonable results (80%). Table 4 shows that at the ‘1 dB’ level many combinations of 2channels reach results over 90%. There are many combinations of Lhv with C- or P-band and combinationsof C- and P-band which are successful. Single frequency combinations are inferior to many of the bestdouble frequency combinations. The effect of speckle can be studied in many ways. Tables 3 and 4 alsoshow results at the ‘0 dB’ and ‘2 dB’ levels. Figure 1 shows results for some combinations as a function ofspeckle level. It clearly illustrates that for some of the better combinations results do not improveconsiderably anymore after 20 looks. Consequently, on average, an area related to 20 looks may beconsidered as the smallest area detectable with any system that is using these combinations. Using three ormore channels doesn’t improve results considerably.


Though results improve when polarimetry is used this improvement is not significant in L—or P-band. Forexample, in table 6, it is shown that a multi-polarization L-band yields an overall result of 85.0%, while afull-polarimetric L-band system only improves to 85.8%. For C-band the overall results are poor, however inthis case the improvement is significant. Full-polarimetric classification is introduced by Hoekman andQuiñones (1998).

Existing or planned systems can now be compared to ‘dedicated’ forest monitoring systems. An overview isgiven in table 6. To be able to evaluate different applications it is necessary to study the possibilities fordiscriminating individual pairs of land cover. These are summarised in Table 7. It clearly shows that P-bandis basic to discriminate secondary from primary forest. To detect recently cut areas in primary forest C-bandalone would suffice. More inferences can be made. Besides classification however, it is important to get anindication of biomass levels. These can be related to land degradation in areas classified as pasture, or forestdegradation in areas classified as primary forest, or can be used to assess secondary re-growth. Table 5 andFigure 2 show some results, clearly indicating the utility of P-band and Lhv.

Validation results for classification are not included here. Focus of the above-mentioned results is on thesystematics rather than the assessment of optimal configurations. It is very useful to expand the proposedsystematics with more data from other sites before drawing firm conclusions.

5.2 New airborne radar systems for high-resolution forest observation

Parameters such as tree position, tree crown dimensions, canopy cover and terrain slope angle, and thelocation and condition of skid trails and logging roads, are of particular interest. In principle suchinformation can be retrieved on a routine basis over large areas from aerial photographs. Repetitiveobservation would allow assessment of logging intensity, erosion and fire susceptibility, verification ofreforestation obligations, etc. However, cloud cover too often prevents timely observation.

Radar does not have this limitation. Moreover, images of short-wave high-resolution radar, in principle, maygive sufficient information. Since other physical mechanisms underlie radar imaging radar images can not betreated in the same way as aerial photographs. Notably effects of ‘radar shadow’ and ‘layover’ should behandled with care. Layover, for example, occurs where two tree crowns with different heights are located atthe same range distance. In non-interferometric radar images these two tree crowns will be imaged on top ofeach other, without being able to detect such a situation. In interferometric images this situation can bedetected through the measurement of phase coherence.

Currently several interferometric airborne SAR systems exist. Of particular interest are the X- and C-bandDornier SAR (Faller and Meier, 1995), because of its high resolution (1.5 m) and recent deployment duringthe INDREX campaign in Indonesia, and the TOPSAR of NASA (Zebker et al, 1992), which has thecapability of interferometric observation in C- and L-band.

The capability of interferometry to measure terrain height is well known. Vegetation cover is oftenconsidered as a cause of error, with a magnitude less or equal to the vegetation height. Alternatively,interferometry is considered as a new technique to measure vegetation height. For repeat-pass measurementsof the ERS satellite relations between interferometric phase coherence and vertical distribution of scatterersin forest plantations have been found (Hagberg et al, 1995). Recent results from the Indrex campaign withthe Dornier SAR indicate that coherence is useful to describe vertical distributions at the tree level andsuggest a way to correct tree height and displacements errors due to overlay (Hoekman and Varekamp,1998). In tropical forests height differences between individual trees can be substantial and are common,hence overlay is common. Emergent trees in primary and logged-over forests can reach more than 10 mabove other upper canopy trees. The same is true for secondary forest which often comprises remnants of theformer primary forest. It seems likely 3-D tree (upper canopy) maps can be reconstructed with sufficientaccuracy to meet current information needs for sustainable forest management in Indonesia.


It seems even likely that forest height can be measured in addition to true terrain height when the radarwavelength is sufficiently long (Treuhaft et al, 1996). A robust system may need two frequency bands. Anairborne X-band interferometer to map vegetation upper canopy height level in combination with a P-bandinterferometer to map vegetation height, using the phase coherence, has been proposed by NASA (Paylor,1997).

5.3 References

Faller, N.P. and E.H. Meier, 1995, First results with the airborne single-pass DO-SAR interferometer, IEEETransactions on Geoscience and Remote Sensing, Vol.33, pp.1230-1237.

Hagberg, J.O., L.M.H. Ulander and J. Askne, 1995, Repeat-pass SAR interferometry over forested terrain,IEEE Transactions on Geoscience and Remote Sensing, Vol.33, pp.31-340.

Hoekman, D.H., 1991, Speckle ensemble statistics of logarithmically scaled data, IEEE Transactions onGeoscience and Remote Sensing, Vol.29, pp.180-182.

Hoekman, D.H. and C. Varekamp, 1998, High resolution single-pass interferometric radar observation oftropical rain forest trees, Proceedings Second International Workshop on Retrieval of Bio- and GeophysicalParameters from SAR data for Land Applications, 21-23 October 1998, ESTEC, Noordwijk, TheNetherlands.

Hoekman, D.H. and M.J. Quiñones, 1997, Land cover type and forest biomass assessment in the ColombianAmazon, Proc. IGARSS'97, 3-8 August 1997, Singapore, pp.1728-1730.

Hoekman, D.H. and M.J. Quinoñes, 1998, Land cover type and biomass classification using AIRSAR datafor evaluation of monitoring scenarios in the Colombian Amazon, Second International Workshop onRetrieval of Bio- and Geo-physical Parameters from SAR data for Land Applications, 21-23 October 1998,ESA-ESTEC, Noordwijk, The Netherlands

Paylor, E., 1997, personal communication

Treuhaft, R.N., S.N. Madson, M. Moghaddam and J. van Zijl, 1996, Vegetation characteristics andunderlying topography from interferometric radar, Radio Science, Vol.31, pp.1449-1485.

van der Sanden, J.J., 1997, Radar remote sensing to support tropical forest management, Ph.D. ThesisWageningen Agricultural University.

Zebker, H.A., S.N. Madson, J. Martin, K.B. Wheeler, T. Miller, Y. Lou, G. Alberti, S. Vetrella and A. Cucci,1992, The TOPSAR interferometric radar topographic mapping instrument, IEEE Transactions onGeoscience and Remote Sensing, Vol.30, pp.933-940.


4. TABLES and FIGURES

Tables 1, 2, 3, 4a-c, 5, 6, 7 and Figures 1, 2a-d

Table 1. Number of plots per vegetation cover class and selected incidence angle ranges.

25°°-60°° 30°°-35°° 45°°-50°°Primary Forest 233 38 34Secondary Forest 227 38 25Recently cut 93 4 24Pastures 225 31 22Total 778 111 105

Table 2. Speckle levels used (See Hoekman, 1991)

Averaging Speckle levelPolygons > 800-looks < 0.154 dB ‘0 dB level’Simulated 20-look data 0.983 dB ‘1 dB level’Simulated 16-look data 1.103 dBSimulated 5-look data 2.043 dB ‘2 dB level’Simulated 4-look data 2.314 dB

Table 3. Overall Maximum Likelihood (MLH) classification results (expressed in percentages) at the 95%level of confidence for the 45°-50° incidence angle range using a single channel, for speckle levels of 0, 1and 2 dB, for all AIRSAR channels studied and for 4 land (vegetation) cover types. The enlarged boldnumbers indicate the best result plus the results that are not significantly different from the best result at the95% level of confidence. The numbers in the shaded boxes indicate the worst result plus the results that arenot significantly different from the worst result at the 95% level of confidence.

Channel‘0 dB’ ‘1 dB’ ‘2 dB’

Chh 59.0 45.4 36.0Chv 71.4 56.7 42.3Cvv 67.6 53.0 40.6Lhh 69.5 60.1 52.8Lhv 91.4 79.4 65.4Lvv 82.9 66.2 53.0Phh 81.9 73.0 60.4Phv 88.6 80.3 67.5Pvv 79.0 70.4 57.4Prr 81.0 74.0 62.5Pll 78.1 72.2 62.3


Table 4a-c. Overall Maximum Likelihood (MLH) classification results (expressed in percentages) at the95% level of confidence for the 45°-50° incidence angle range using 55 combinations of two channels, forspeckle levels of ‘0, 1 and 2 dB’ (in table a, b and c, respectively) and for 4 land (vegetation) cover types.For each table the enlarged bold numbers indicate the best result plus the results that are not significantlydifferent from the best result at the 95% level of confidence. The numbers in the shaded boxes indicate theworst result plus the results that are not significantly different from the worst result at the 95% level ofconfidence.

Chv Cvv Lhh Lhv Lvv Phh Phv Pvv Prr PllChh 77.1 70.5 95.2 97.1 97.1 96.2 98.1 97.1 98.1 97.1Chv 73.3 95.2 97.1 93.3 96.2 98.1 95.2 100 99.0Cvv 91.4 95.2 96.2 98.1 98.1 97.1 98.1 99.0Lhh 95.2 94.3 84.8 87.6 83.8 82.9 79.0Lhv 91.4 97.1 96.2 94.3 96.2 94.3Lvv 96.2 92.4 91.4 96.2 93.3Phh 95.2 94.3 96.2 96.2Phv 91.4 91.4 92.4Pvv 88.6 87.6Prr 83.8

Chv Cvv Lhh Lhv Lvv Phh Phv Pvv Prr PllChh 58.5 54.0 75.7 83.7 74.8 85.3 88.7 83.6 86.5 86.8Chv 61.0 81.7 86.9 80.6 90.3 92.9 89.1 92.2 92.0Cvv 81.1 86.1 78.7 90.3 93.1 89.1 92.3 91.9Lhh 83.1 74.6 72.7 78.8 71.8 73.3 72.5Lhv 80.4 91.4 91.2 87.5 91.0 91.2Lvv 86.1 85.3 80.8 84.3 84.7Phh 84.3 80.1 79.9 78.0Phv 80.0 80.8 81.3Pvv 76.3 77.1Prr 76.4

Chv Cvv Lhh Lhv Lvv Phh Phv Pvv Prr PllChh 46.6 43.3 59.8 68.1 57.3 69.5 74.7 64.7 71.0 69.9Chv 49.7 67.4 72.9 63.4 76.8 81.1 71.9 77.8 77.3Cvv 65.8 71.2 61.4 75.6 79.5 70.4 76.8 76.2Lhh 68.8 61.3 63.3 68.5 60.9 64.4 64.0Lhv 66.5 78.5 78.5 72.8 76.9 76.9Lvv 71.6 72.9 65.4 71.2 70.5Phh 73.5 67.4 68.7 67.9Phv 69.7 71.4 71.2Pvv 66.5 65.9Prr 67.7


Figure 1. Overall Maximum Likelihood (MLH) classification results (expressed in percentages) at the 95%level of confidence for the 45°-50° incidence angle range for several combinations as function of the specklelevel expressed in number of looks.


Table 5. Relationship between backscatter, expressed as γ[dB], and biomass expressed as log10 of theabove ground fresh biomass in tons/ha, for several frequency and polarisation combinations. The regressionresults shown apply to the lowest Standard Error of Estimate (SEE) and the associated correlation coefficientr2. Further, the total range of γ of the experimental data and the ratio of range and SEE are shown.

r2 SEE [dB] range [dB] range/SEE [dB]

C-HH 0.31 0.44 3.25 7.3C-HV 0.61 0.34 3.195 9.3C-VV 0.66 0.52 4.82 9.3L-HH 0.81 1.16 9.312 8.0L-HV 0.93 1.05 11.59 10.9L-VV 0.77 0.95 7.75 8.2P-HH 0.90 1.43 11.12 7.8P-HV 0.94 1.71 16.12 9.4P-VV 0.91 0.87 9.58 11.0P-RR 0.93 1.24 13.15 10.6L-HV+P-RR 0.94 0.94 12.27 13.1

Figure 2a. L-band with HV-polarisation backscatter as function of biomass. The biomass is the fresh weightabove ground biomass (in tons/ha) at the logarithmic scale (i.e. 1.0 is 10 ton/ha, 1.5 is 31.6 ton/ha, etc.).Experimental data for primary forest (�), secondary forest (�) and pasture (*) are fitted to a curve of theform γ[dB]= a + b(1-exp(-cx)), where x is the logarithm of the biomass.


Figure 2b-c. Idem for P-band with HV and VV polarisation


Figure 2d. Idem for P-band with RR polarisation


Table 6. Overall Maximum Likelihood (MLH) classification results (expressed in percentages) at the 95%level of confidence for the 45°-50° incidence angle range for a selection of combinations at the ‘1 dB specklelevel’. The enlarged bold numbers indicate the best result plus the results that are not significantly differentfrom the best result at the 95% level of confidence. The numbers in the shaded boxes indicate the worstresult plus the results that are not significantly different from the worst result at the 95% level of confidence.

Channel remark result kappa VarkappaCpol, Lpol, Ppol 94.7 0.931016 0.000768‘All’ ‘11-channel’ 94.6 0.929309 0.000777Chv, Phv, Prr ‘best 3-channel’ 94.0 0.921145 0.000885Cpol, Prr 93.1 0.909599 0.001018Phv, Cvv ‘best 2-channel’ 93.1 0.908934 0.001032Cvv, Chh, Chv,Prr

ASAR+Prr 92.6 0.902984 0.001077

Lpol, Prr LightSAR+Prr 91.9 0.893512 0.001174Lvv, Lhh, Lhv,

Prr

LightSAR+Prr 91.3 0.886642 0.001225

Lhv, Prr 91.2 0.883844 0.001275Cpol, Lpol 90.4 0.873539 0.001365Lpol, Cvv LightSAR+ERS 89.5 0.862657 0.001479Lvv, Lhh, Lhv,

Cvv

LightSAR+ERS 89.0 0.856137 0.001521

Ppol 87.1 0.829398 0.001803Pvv, Phh, Phv 86.1 0.816590 0.001904Lpol LightSAR 85.8 0.813185 0.001926Lvv, Lhh, Lhv 85.0 0.803859 0.001996Lhh, Cvv JERS+ERS 81.1 0.751464 0.002492Lhh, Prr JERS+Prr 73.3 0.647533 0.003142Cpol Radarsat-2 ? 68.4 0.581294 0.003612Cvv, Chh, Chv ASAR 61.9 0.497471 0.003859Lhh JERS 60.1 0.465214 0.003928Cvv, Chh ERS+Radarsat 54.0 0.393662 0.004008Cvv ERS 53.0 0.373692 0.004144Chh Radarsat 45.4 0.275277 0.004154


Table 7. Percentage of confusion between land cover type pairs for the 45°-50° incidence angle range at the 1dB speckle level for the cases given in table 6. Each entry gives the result as the percentage of the sum of classa samples classified as b and class b samples classified as a from the sum of class a and class b samples, in theabsence of other classes (i.e. absence of class c, etc., and the class ‘unknown’). The expected value formaximum confusion therefore is 50%.

Monitoring Scenarios

Bands and polarisation Chh Cvv CvvChh

Lhh CvvChhChv

LhhPrr

LhhCvv

LvvLhhLhv

PvvPhhPhv

CvvChhChvPrr

PhvCvv

ChvPhvPrr

‘All’

Loss offorests

Primary –Pastures 1-4

20.1 10.6 9.1 0.5 7.7 0.0 0.4 0.0 0.0 0.0 0.0 0.0 0.0

Secondary-Pastures 2-4

18.6 10.3 8.0 1.9 7.5 0.8 1.5 0.1 0.8 0.7 1.0 1.0 0.0

Newlydeforestedareas

Primary-Recently Cut1-3

12.6 3.3 1.8 40.9 0.4 29.1 3.0 2.6 10.9 0.4 1.7 1.2 0.0

Secondary-Recently cut 2-3

11.3 3.2 1.6 30.8 0.3 17.3 2.2 9.5 9.6 0.3 1.9 1.2 0.0

Pastures –Recently Cut3-4

38.6 32.3 32.2 1.1 16.8 0.0 0.8 1.0 0.0 0.0 0.1 0.0 0.0

Foreststages

Primary-Secondary 1-2

46.0 46.9 45.3 23.8 43.1 5.3 24.4 11.7 2.1 5.8 4.6 2.0 0.6

Bands and polarisation PrrLhv

Cpol Lpol LpolCvv

LpolPrr

CpolPrr

Ppol CpolPpol

CpolLpolPpol

1-4 0.0 4.0 0.0 0.0 0.0 0.0 0.0 0.0 0.02-4 0.3 4.3 0.1 0.0 0.0 0.5 0.6 0.1 0.01-3 2.9 0.5 2.2 0.5 2.2 0.4 10.2 0.1 0.12-3 4.6 0.6 9.0 1.4 3.6 0.2 9.3 0.1 0.13-4 0.1 12.3 0.9 0.7 0.1 0.1 0.0 0.0 0.01-2 4.3 40.0 11.1 10.8 4.0 5.8 2.0 10.1 1.2

User Requirements Study: Workpackage 6 – Remote Sensing Applications for Forest Management A2

6. ASSESSMENT OF THE ROLE OF NEW DEVELOPMENTS FOR HIGH -RESOLUTIONSPACEBORNE SENSOR SYSTEMS.

6.1 Introduction

This document assesses the role of new technological developments for the performance of a dedicatedforestry mission. Two technological developments will be studied in particular:

- Local ground receiving stations- Internet technology

6.2 Local ground receiving stations.In this document local ground receiving stations are defined as ground receiving stations that can be movedquite easily from one place to the other (transportable) and cover only a limited, but local, area.

6.2.1 Ground receiving stations in general

Ground receiving stations are available on the market in various sizes, with associated various costs. Ingeneral for reception of high resolution satellite data it is required to have a large antenna (order 8-13 mdiameter), which can acquire from a large area (see the coverage of the SPOT ground receiving stations) andto meet the high data rates of up to 105 Mbits/sec.

These ground receiving stations cover a large area, which introduces all kinds of side effects such as priorityproblems between users and countries, long lead times, exchange of foreign currencies, political problemsetc. (see for reference RESPAS User Consultation, BCRS Report 93-15).

The above figure shows that almost every piece on earth is covered by these ground receiving station, butstill it is observed that many users do not get their data, because of the above, mostly political, problems.The lion share of the market for these large ground-receiving stations is taken by the Canadian companyMDA. Reported cost figures, depending on the size of the antenna, are between 4.5 MUS$ and 12 MUS$.

In addition to these high resolution ground receiving stations there are a number of low resolution groundreceiving stations (for reception of 1km resolution NOAA-AVHRR data), which require small antenna’s(order 0.5-1m diameter). These ground stations are available in almost every country and are mainly used inmeteorological departments for weather forecasting. There is a tendency to work with the 1km NOAA-datafor other applications as well.

There are various companies who supply these ground-receiving stations, even to the private market. Aprofessional ground receiving station will cost approximately 75.000 US$. Consumer ground receivingstations are offered for 10.000 US$.

6.2.2 Local ground receiving stations

There is a mismatch between the unsuccessful, in terms of revenues, operation of these large groundreceiving stations and the effectiveness of the low resolution ground receiving stations. A trend is observedthat some organizations have started to investigate the possibility of combining the local low resolutionreception effectiveness with the high resolution reception capabilities. The result is a local high resolutionreception capability at moderate costs. Currently, the only system available on the market is the RAPIDSsystem, developed by Natural Resources Institute, BURS Ltd. (both UK) and the National AerospaceLaboratory NLR (the Netherlands).


The Real-time Acquisition and Processing - Integrated Data System (RAPIDS) is a PC-based X-bandreceiver designed to provide local users with rapid access to earth observation data at least cost.The design philosophy of the RAPIDS PC-based transportable ground station is to meet national and/or localneeds for timely environmental data. In most countries, a large number of resource managers, planners anddecision makers would benefit from timely information on their environment, if it wereAvailable promptly (on demand), reliably and as inexpensively as possible. For these local areas/users thedemands are less than for horizon-to-horizon systems. Thus the resulting ground segment can beinexpensive, easy to transport, install and maintain.

The principal design requirement is for a system to handle capture for local areas when the satellite is within± 45° of the overhead position. The system has to maximize control during these passes where the rate ofchange in satellite position is highest. The ± 45° cone of acquisition enables capture of small unit volumes ofdata of local interest. The system also has to be robust enough to minimize the effect of wind forces duringtracking, and to be simple to maintain and operate. Standard PCs were selected as the platforms formanagement, tracking, capturing and processing of data. This is because of their increasingperformance/cost advantage and their widespread availability and use (compared to UNIX workstations) indeveloping countries. This makes for easier local maintenance and cost-effective integration with existingcapacity.

The current set-up for receiving consists of a dish antenna of 2.7 metres that can be tilted over a range of+60° to -30° in two perpendicular directions (figure 1). This range is enough to capture data within a circulararea of approximately 1000 km diameter, depending on the site. A set of four path aerials each with its ownlow noise amplifier for the 2.2 GHz tracking system is mounted at the dish centre. The 8 GHz data receptionLNA is positioned in the same focal plan as the patch aerials at the centre of the dish.

Figure 1. RAPIDS transportable ground station deployed with 2.7m dish aerial at NLR facility.


The antenna dish is moved by a hydraulic configuration. A hydraulic power unit with oil reservoir pumps,motors, and valves controlling the drive rams is used to move the aerial. Position monitors are linked to thepower supply for safety cut out, “safe park” mode and alarms. The receiver system is thus capable oftracking on S-band beacon signals and capturing X-band data signals from satellites over 90 arc-degreesusing a sub 4 metre antenna. Capability currently includes ERS, SPOT and JERS. Potentially, other satellites(e.g. IRS, Landsat, EOS et al) could be added to the capability.

The coverage of the RAPIDS facility depends on the configuration at installation. The NLR facility has a 90degrees coverage from 60 degrees to the North to 30 degrees to the South. This covers an ellipsoidal area ofapproximately 1000 km diameter (see figure below).

RAPIDS is put on the market for 0.5 MUS$.

6.2.3 The advantages of local ground receiving stations for a dedicated forestry mission.

Local ground receiving stations are essential in providing a direct link between the data and the use of thatdata. The advantages are:

1. Investment costs are less compared to large ground receiving stations.2. Running costs are less compared to large ground receiving stations.3. Data needs can be tuned to the actual needs of the user of which it is assumed that the user will be

the operator of the ground receiving station.4. Local area reception reduces the long lead times before data actually gets in the hands of the users.5. The data access is controlled by the user, the user remains autonomous6. There is no exchange of foreign currency across borders.

6.3 Internet technology

Internet is more and more used as a medium for the transportation of small amounts of data, but will, mostlikely, be used for delivery of large amounts of data in the future. The Internet infrastructure is wellstructured and well placed in the industrialized countries. In tropical countries the infrastructure is still notwell developed and thus unreliable. In these countries people still have to rely on telephone connections inthe absence of high-speed connections. However, there is a big development on-going to provide a globalbroadband “Internet-in-the sky”. The US-based company Teledesic (shareholders: Microsoft’s Bill Gates,Boeing, Motorola and Matra Marconi Space) will create the world’s first network to provide affordable,worldwide access to telecommunications services such as broadband Internet access and other digital dataneeds. The service that will use some 288 small telecommunications satellites is intended to start in 2003.The Teledesic system will offer user equipment to access the network. Most users will have access to thenetwork using a 64 Mb per seconds connection. The user equipment consists of small low-power terminalsand antenna’s. The laptop-size terminals will mount flat on a rooftop and connect inside to a computernetwork or single PC.Compared to standard land line connections, currently running at 33.6 Kb per second, an increase in capacityof approx. 2000 times is expected.


For example: A digital forest/non-forest map of 100 MB will take approximately 12.5 seconds using theTeledesic system and about 25000 seconds.

6.4 Conclusions

The key issue in all remote sensing missions is how to get the data/information to the users. There are severalways of doing that. Main aspect to be considered here is the autonomy of users. Users want to beindependent, they do not want to be controlled, they simply want control over what to get. The only way ofachieving that is to give the local user the tools to receive data and the tools to extract the informationlocally. A local, low cost ground station seems the only possibility at this particular moment. The Teledesicdevelopment as sketched above is simply too far away to investigate its opportunities.

technical document 6 (without appendices) · ITC In cooperation with FAO IKC N NIVR ... FPM-MAG US...

Documents

Transcript of technical document 6 (without appendices) · ITC In cooperation with FAO IKC N NIVR ... FPM-MAG US...