Assessing the Extent and Severity of Erosion on the Upland ...€¦ · Final Report 6 . Assessing...

Assessing the Extent and Severity of Erosion on the

Upland Organic Soils of Scotland using Earth Observation

A GIFTSS Implementation Test

FINAL Report

October 2009

The views presented in this report are those of the contractor and do not necessarily reflect

those of the Scottish Government or Scottish Ministers

Assessing the Extent and Severity of Erosion on the Upland Organic Soils of Scotland using Earth Observation

Table of contents Glossary ....................................................................................................................... 3 Executive Summary...................................................................................................... 5 1. Introduction............................................................................................................... 6

1.1 Aims.................................................................................................................... 6 1.2 Soil and peatlands............................................................................................... 7 1.3 Earth observation ................................................................................................ 9

2. Study Site ............................................................................................................... 11 2.1 Monadhliath mountains overview ...................................................................... 11

3. Data........................................................................................................................ 15 3.1 EO data............................................................................................................. 15 3.2 Ancillary data..................................................................................................... 16

4. Image Processing and Data Preparation................................................................. 17 4.1 Image correction................................................................................................ 18 4.2 Generation of derived layers.............................................................................. 21 4.3 Development of classification classes ............................................................... 21 4.4 Selection of testing and training sites ................................................................ 22 4.5 Image segmentation.......................................................................................... 25

5. Analysis – Integration with Current Methods ........................................................... 27 5.1 Mapping of peat erosion features ...................................................................... 27 5.2 Investigation of ASTER and peat composition................................................... 31 5.3 Peat erosion and impacts upon soils ................................................................. 34

6. Maps of Classification Results ................................................................................ 34 7. Precision and Accuracy .......................................................................................... 39

7.1 Mapping of peat erosion features ...................................................................... 39 7.2 Evaluation of suitability for purpose ................................................................... 49

8. Implementation Plan ............................................................................................... 56 9. Conclusions ............................................................................................................ 59 10. Acknowledgments................................................................................................. 60 11. References ........................................................................................................... 61 Annex 1 Satellite and Airborne Sensor Specifications................................................. 65 Annex 2 Image Processing Technical Specifications .................................................. 67 Annex 3 Classification Class Descriptions .................................................................. 71 Annex 4 EO Classification Using Traditional Methods................................................. 82 Annex 5 Classification Areas ...................................................................................... 84 Annex 6 Project Deliverables ...................................................................................... 85

Report prepared by

Steve Keyworth, Mark Jarman, Dr. Katie Medcalf

Environment Systems Limited

11 Cefn Llan Science Park

Aberystwyth

Ceredigion

SY23 3AH

Tel: +44 (0)1970 626688

http://www.envsys.co.uk


Glossary

ArcGIS Geographic Information System (GIS) softwareA

ASTER Advanced Spaceborne Thermal Emission and Reflection Radiometer

ATCOR Atmospheric/Topographic Correction software

Atmospheric The correction made to remotely sensed radiance to account for correction effects related to the intervening atmosphere between the

earth’s surface and the satellite.

BNSC British National Space Centre

Boolean A system of logic/algebraic processes (e.g., AND, OR)

CASI Compact Airborne Spectrographic Imager

Confusion matrices A matrix that displays statistics for assessing image classification accuracy by showing the degree of misclassification among classes.

Definiens AG A German-based company producing Definiens Developer software

Definiens Developer Graphical user interface for developing Definiens processes

(eCognition)

DEM Digital Elevation Model

DTM Digital Terrain Model

Endmember Spectra associated with the pure samples of surface materials (e.g., vegetation).

EO Earth Observation

Fuzzy membership A mathematical function that defines the degree of an elements function membership in a fuzzy set

GIFTSS Government Information From The Space Sector

GIS Geographic Information System

HRG High Resolution Geometric

IRS Indian Remote Sensing Satellite

Kappa coefficient A statistical measure of the agreement, beyond chance, between two maps (e.g. map of classification and ground truthed map)

Linear spectral The process of extracting the proportions of surface materials unmixing from spectrally mixed image pixels based on knowledge of pure

spectra (endmembers)

NDVI Normalized Difference Vegetation Index

NIR Near infrared

Orthorectification A process of geometric referencing of an image to a map coordinate system that considers variations in the topography of

3


the earth surface and the tilt of the satellite sensor.

Pre-processing Ortho, atmospheric, topographic and other corrections to prepare imagery for classification.

Radiance Conversion of satellite image digital numbers to radiance. calibration

Segmentation Grouping of pixels based on similar values – a type of automated vectorisation (digitising).

Server engines The components of Definiens software accessible over a network.

Shapefile A set of files used by ESRI Arcmap the contains points, arcs or polygons holding tabular data and spatial information

SPOT A French satellite supporting the HRG sensor

SWIR Short Wave Infrared

Topographic Shadowing of a surface by the surrounding topographic relief shadowing and as a function of solar angle.

4


Executive Summary Environment Systems were commissioned by the Scottish Government, working with the British National Space Centre (BNSC) under the GIFTSS programme (Government Information from the Space Sector), to evaluate satellite earth observation as a cost-effective method of assessing the extent and severity of erosion in the upland organic soils of Scotland.

Soil is a non-renewable resource which performs many vital functions as part of the systems supporting food and other biomass production, storage, filtration and transformation of many substances including water, carbon and nitrogen. Peat soils are formed as a result of organic material that has accumulated over centuries. This build-up of organic deposits can be several metres in depth. The spatial extent of exposed peat, intact peat, vegetated bog surfaces and pools is a highly important component in regards to the potential modelling of the carbon budget of peatland and global climate change. As long as a peat bog accumulates more organic matter than it loses, it functions as a carbon sink. Previous attempts to map the distribution of upland peatlands have been reliant on aerial photography or satellite sensors (jointly known as ‘earth observation’) and have only attempted to answer first level ecological questions such as the extent of exposed peat. Advances in techniques and understanding of earth observation now allow for more detailed mapping and monitoring the surface of the Earth. The GIFTSS study was set up to deliver an implementation test of mapping peat erosion using earth observation. The study was based on the Monadhliath Mountains which are located within the westernmost range of the Grampian Mountains in the highlands of Scotland. For the study area SPOT5, IRS P6 and ASTER satellite imagery were prepared; including full geometric and atmospheric correction. Critical to the success of the project was the availability and full integration (into the automated processing chain) of current digital aerial photography. This image data was then complemented by GIS datasets that, whilst often historical and at differing scales and nomenclatures, provide a set of core topographic and thematic information. The Definiens Developer object orientated rule-based classification software was successfully used to classify the imagery to produce a set of ‘core level’ data, which in turn was used to produce ‘application level’ data; in this case peat erosion maps. The advanced image segmentation (automated digitising) within Definiens provided the spatial framework for the classification, with a bespoke set of spectral and spatial rules driving the classification of the image segments. Overall map accuracy was calculated as over 84%, with clear visual coincidence between the classified map and both the in-situ field data and aerial imagery.

Indicative costs of implementing this method for Scotland have been provided to enable the costs and benefits of different approaches to soil monitoring to be evaluated.

Final Report 5


1. Introduction In 2009 the Scottish Government launched its Soil Framework for Scotland which

highlighted the need for a monitoring scheme to identify trends in soil condition. A central theme for this framework is concerned with the link to climate change and the large soil carbon reserves that exist in the soils of Scotland. Environment Systems were commissioned by the Scottish Government, working with the British National Space Centre (BNSC) under the GIFTSS programme (Government Information from the Space Sector); to evaluate satellite earth observation as a cost-effective method of assessing the extent and severity of erosion in the upland organic soils of Scotland. This approach has already been evaluated through a BNSC funded scoping study and a joint BNSC and Scottish Government Workshop held in 2007. Participants at the workshop included leading scientists engaged in research on upland organic soils in England, Wales and Scotland. One of the conclusions was that it was likely to be “feasible” that satellite remote sensing could contribute to assessing the state of upland organic soils in Scotland (Birnie and Puri 2007).

1.1 Aims

The aim of this project is to provide the Scottish Government (as the user) with the necessary information to assess the practical scope for the routine use of EO information to support the delivery of data on national trends on the severity and extent of erosion in upland organic soils.

In particular the following information is required, in Phase 1;

• A demonstration of the effectiveness of high and very high resolution satellite data for identifying and quantifying peat erosion on organic soils in Scotland.

• Processing and use of remotely sensed data for a demonstration area.

• Assessment of the suitability of satellite earth observation, at various resolutions,

• Comparison of the effectiveness of aerial and satellite based remote sensing in achieving national level monitoring information.

• Consideration of appropriate remote sensing methods to identify erosion extent and severity.

In Phase 2;

• Analysis and comparison of achievable precision and associated methods.

• Assessment of the suitability of this data to provide a national overview of erosion in organic upland soils in Scotland.

• Assessment of the potential for use of this data to provide broad national trends for soil erosion extent and severity in the upland organic soils of Scotland.

Final Report 6


The following tasks have been considered during both Phase 1 and Phase 2:

• Providing expert scientific advice to the Scottish Government,

• Providing key factual information on the extent and severity of erosion in Scotland’s upland organic soils,

• Evaluating the feasibility of monitoring change in the extent and severity of upland organic soil erosion in Scotland using Earth Observation,

• Evaluating the best compromise between data resolution, acquisition timescale and quality (accuracy, usefulness) of output results,

• Evaluating the cost effectiveness of EO compared to other available techniques to provide information on the extent and severity of erosion in Scotland’s upland organic soils.

1.2 Soil and peatlands

Soil is defined as the top layer of the earth’s crust, consisting of weathered mineral and biological matter. Soil is a non-renewable resource which performs many vital functions as part of the systems supporting food and other biomass production, storage, filtration and transformation of many substances including water, carbon and nitrogen. Soil has a role as a habitat and gene pool, serves as a platform for human activities, landscape and heritage and acts as a provider of raw materials.

Soil degradation (e.g. erosion, compaction, loss of organic material, contamination) is accelerating, with negative effects on human health, natural ecosystems and climate change, as well as on our economy.

Soil protection does exist in European legislation in e.g. CAP, Landfill and Nitrates Directives), but at EU level there is no instrument specifically addressing the protection of soil. The 6th Environment Action Programme commits the EU to treat soil the same as water and air as an environmental media and as a non renewable resource to be conserved. These commitments include COM (2006) 231 on the Thematic Strategy for Soil Protection, COM (2006) 232 establishing a framework for the protection of soil and amending Directive 2004/35/EC and SEC (2006) 620 of the Thematic Strategy for Soil Protection.

Peat can be defined as ‘Organic material which accumulates in wet places where plant growth exceeds the rate of residue decomposition’ (Brady & Weil 2002). These residues accumulate over the centuries from wetland plants and suffer from very low levels of decomposition due to the lack of oxygen present (and in some cases low temperatures). This leads to a build-up of organic deposits which can be several metres in depth which are collectively known as peat.

Blanket peat covers almost 9% of the land surface of the United Kingdom and 10.45% of Scotland’s (Taylor 1983), which is a significant proportion of the global peat resource. Most of these peatlands are bogs, and the majority of these (over 70%) are found in Scotland (Lindsey and Immirzi 1996).

Peat can be classified into four defined groups as specified by;

The remains of mosses such as Sphagnum.

The remains of herbaceous plans such as sedges and reeds

7

1. Moss Peat

2. Herbaceous Peat

Final Report


Forms the remains of woody plants, such as trees and shrubs

Remains of aquatic plants (e.g. algae) and of faecal material of aquatic animals

Typically, organic deposits contain two or more of these kinds of peat, often forming in an alternating layered fashion. Within the peat layer itself, woody materials tend to dominate the surface layers due to the succession of plants.

Structurally peat is made up of two distinct entities, an upper layer known as the acrotelm, and a lower layer known as the catotelm. The acrotelm is very thin in form (0.2-0.8 m), located at the surface, and the layer within which the top of the water table can always be found. The catotelm is much thicker (1 – 10 m), has a water content of around 95-97%, and consists of dead or decaying plant material.

Peatland and Carbon

Peatlands have been described as being one of the most important ecosystems in terms of their feedback to global climate change (Bridgham et al., 2008). They play a key role in the global carbon cycle through the sequestration of atmospheric carbon into peat, and the release of carbon gases in the form of CO2 and methane.

Peatlands contain between 329 and 525 x 1015 g C, or roughly 19% of the global soil carbon pool (Batjes 1996). It has also been suggested that the global peatland soil- carbon pool is 75 times larger than annual anthropogenic carbon dioxide emissions (Bridgham et al., 2008).

Erosion Features

Peat erosion ranges in scale from local events to the extensive degradation of plant cover and associated exposure of bare peat. This exposure of the bare peat leads to the surface layers within the peat mass becoming less structurally cohesive through the action of frost and desiccation. Rain may then penetrate down these desiccation cracks leading to the development of gully systems. Other erosion features prevalent in peatland environments included rills and sheet features induced by low magnitude, high frequency events, and high magnitude low frequency events such as peat slips and peat bursts.

Existing ground measurements of contemporary erosion rates on bare peat surfaces suggest that changes are in the order of 1-5 cm per annum. A conservative figure of 350,000 ha has been suggested as to the total area of blanket peat in the British Isles currently suffering from peat erosion processes (Tallis 1998).

Key Peat Erosion Studies

• An extensive study of peat erosion in Scotland has been undertaken by Ian Grieve of Stirling University. Around 20% of the total upland area of Scotland was surveyed by generating a 5 x 5 km grid and surveying 144 sites selected using a stratified random sampling approach.

Eroded features were grouped into three classes:

• Narrow well defined gullies • Broad gullies • Remnant peat blocks

Final Report 8

3. Woody Peat

4. Sedimentary Peat


These eroded features were identified through the examination of stereo of aerial photography and features mapped with a quantification of these areas produced. Total area of eroded peat was calculated to be ~219 km2 or 6% of the total area sampled. There was significant variation in levels of erosion throughout Scotland with peat erosion levels in the Monadhliath found to be over 20%.

• A report was undertaken by Scottish Natural Heritage (SNH) in 2006 to

assess the impacts of herbivore species upon blanket bog in a Special Area of Conservation (SAC) within the Monadhliath. An association was found between heavy trampling due to red deer and sheep, and areas of peat erosion through the exacerbating of peat hags and the prevention of the re-vegetation of bare peat (Dayton 2006).

Areas of blanket bog were located within the SAC and a weighting of the extent of trampling for each site was given based as specified by MacDonald (2008). It was noted that upland plateau areas were the most subjective trampling and within little or no re-colonisation of eroded bare peat. It was however noted that these areas are subject to natural process such as peat hagging which is associated with the creation of areas of bare peat. Therefore all erosion could not be put down to trampling.

Mapping Peat

The spatial extent of exposed peat, intact peat, vegetated bog surfaces and pools is a highly important component in regards to the potential modelling of the carbon budget of peatland and global climate change. As long as a peat bog accumulates more organic matter than it loses by aerobic decay in the surface layers (the acrotelm) and by anaerobic decay in the permeability saturated lower layers (the catotelm), it functions as a carbon sink.

A definition employed for the purposes of mapping Scottish peat land is an area with an organic horizon of over 50 cm with the state and condition of a peat land ecosystem related to vegetation, soil and water. It is worth noting that organic soils occur at all altitudes in Scotland, and are often present at sea level especially in the Western and Northern Isles.

1.3 Earth observation

Earth Observation (EO), including satellite and airborne systems, allow for mapping and monitoring the surface of the Earth. EO is the ‘earth facing’ discipline of remote sensing.

EO technologies are most commonly used through the acquisition and use of aerial photography, with satellite-based EO starting up in 1972 with the launch of the first Landsat satellite. Since then, there have been progressive improvements in spatial, temporal and spectral resolution, across a range of mapping scales for a variety of mapping requirements. The operational adoption of EO systems is often held back in the UK by our weather patterns, where the build up of a time series of satellite images can be challenging. With this in mind hybrid solutions such as satellite imagery in combination with airborne imagery can offer the flexibility of focussing operations in clear weather breaks. Operating costs associated with this can be high, but systematic capture of aerial photography is now more routinely planned allowing many applications to benefit from the ‘collect once, use many times’ ethos.

Final Report 9


For the purposes of this study and based on the accepted definitions (including used by the EU Global Monitoring for Environment and Security), satellite systems are categorised into four resolution classes:

• Very High Resolution (VHR) ≤ 5 m

• High Resolution (HR) ≤ 30 m

• Medium Resolution (MR) ≤ 300 m

• Low Resolution (LR) ≤ 5 km

All aerial photography should be considered ‘Very High Resolution’. Higher spatial resolution typically means a smaller geographic footprint. Wider area coverage can be achieved by mosaicing several scenes together.

EO in the uplands

EO can be used to estimate the variety, type and extent of land-cover throughout a study area, meeting an elementary need that is common to many ecological applications. With fine spatial and/or spectral resolution remote sensing sensors becoming increasingly available within both airborne and spaceborne platforms, the possibility for understanding these factors and overcoming them is affording the potential for improved understanding and monitoring of land-cover and biodiversity.

Previous attempts to map the distribution of upland peatlands have been reliant on aerial photography and broad EO sensors such as Landsat (McMorrow et al., 2004). These attempts have only sought to answer first level ecological questions such as the extent of exposed peat.

Increases in spectral and spatial capabilities within EO sensors (e.g. using airborne hyperspectral), combined with the necessary expertise and processing techniques, offer the possibility to answer second and third level ecological questions such as the composition of exposed peat, and physical and biochemical properties. Other relevant EO/peat activities

• A report was commissioned by SNH in 2000 to provide a large scale assessment of the blanket bogs which are found across Grampian and adjoining areas of north-east Scotland through utilising a remote sensing approach. It was undertaken by coupling knowledge of remote sensing with peatland ecology and allowed for a broad-scale ecological assessment of this remote and extensive habitat (Johnson and Morris 2000).

A ground survey of the area was undertaken and peatland classes indentified. These areas were then located upon photographs and the Landsat data and manually digitised on screen. An unsupervised image classification process was undertaken within areas known to contain peat as identified using SSSI data. The concluded that it was not possible to establish with high levels of accuracy a detailed blanket bog classification due to the coarseness of the Landsat (HR-MR) imagery.

• LUCAS (Land Use / Cover Area Frame Statistical Survey) is a major pan-

European project, launched in 2000 by DG Agriculture and Eurostat, and based on the application of area frame techniques for collecting land cover and land use. The current LUCAS survey is taking place during 2009, with the aim of repeating this every 3-5 years. The current survey

Final Report 10


has been extended (with cooperation of the European Commission Joint Research Centre, Ispra) to incorporate a soil component in the survey in order to improve the quality of soil modelling and monitoring in Europe. The results will be used for a number of purposes, including updating European soil maps, validating soil models, and measuring the quantity of organic carbon in soil. There are 3,417 LUCAS sample points in Scotland, 834 of which will include a soil sample.

2. Study Site

2.1 Monadhliath mountains overview

The Monadhliath Mountains (Figure 1) are located within the westernmost range of the Grampian Mountains in the highlands of Scotland. They lie to the west of the Cairngorms and traverse in a northeast to southwest direction located to the south of Loch Ness. The mountains rise to over 900 m with high ridges and plateaus (Dayton 2006). Carn Dearg is the highest peak located within the area with an altitude of 3,100 feet (945 metres).

The area has many small steams and rivers flowing away from the mountain range. The largest of these is the River Findhorn which rises on the north western slopes of Carn Mairg and flows northward to the North Sea. Areas within the Monadhliath have been designated as ‘Special Areas of Conservation’ (SAC) and as an SSSI under the Wildlife and Countryside Act (1981). The area has been considered of such importance due to the presence of breeding birds, blanket bog, upland assemblage, vascular plant assemblage and the moth Psodos coracina (Dayton 2006).

The Monadhliath are located on bedrock of a predominantly metamorphic quartz- feldspar-granite makeup, with some mica-schist located to the east of Carn Ban. Other geological units noted include intrusive granite, syenite, granophore and allied types which are predominantly located along the hilltops that run along the boundary between Coignafearn and Glen Banchor (Dayton 2006). The climate of the Monadhliath is one of the wettest in Britain with the area receiving on average over 1691 mm of rainfall a year. The area also receives large amounts of snow with snow present for up to 106 days a year. Average annual temperatures for lowland areas are around 8oC with temperatures barely above freezing for parts of the uplands (Met Office 2009).

Much of Monadhliath is inaccessible by road but is increasingly frequented by hill walkers since the introduction of the ‘Landform Reform Act’ (2003) legislation with Scotland. The area is also populated by numerous grazing species including sheep and deer. These have been associated with heavy trampling of blanket bog areas within the upland plateau (Dayton 2006).

Vegetation

The area in and around the Monadhliath comprises a very complex vegetation structure which includes areas of plantation evergreen woodland, heaths, bogs and mires as well as extensive grassland areas. As the topography of the area is highly undulating with altitudes up to 950 m, it has enabled a broad range of habitats and species to thrive and also created different environments within these areas. Final Report 11


Species found in abundance there include:

• Heath – Erica tetralix (Ling heather), Vaccinium myrtillus (Bilberry);

• Blanket bog - Sphagnum spp. (Bog moss) Eriophorum spp. (Cotton grass);

• Montane heath – Rhacomitrium spp. (Wolly fringe moss).

• Woodland – Larix decidua (Larch), Picea abies (Norway Spruce), Quercus spp.(Oak);

• Grassland – Nardus stricta (Matt grass), Molinia Caerulea (Purple moor grass), Festuca spp. (Fescues);

• Other – Pteridium acquilinum (Bracken), Ulex spp. (Gorse), Juncus Effusus (Rush).

The area is known to contain large areas of blanket bog. These are located within areas of gently sloping ground and are concentrated within the valley bottoms and upland plateaus. Smaller areas of blanket bog are found throughout the area forming in depressions and in mosaic with areas of wet heath and acid grassland (Dayton 2006).

Scottish Borders – White Coomb and Moorfoot Hills overiew

Primarily due to the availability of airborne near infrared, but also as a comparison area, two small sites in the Scottish Borders (figure 2) have been selected for inclusion in this study. The Scottish Borders Council (licence holders of the near infrared imagery) has a draft Phase 1 Habitat vegetation classification which could be subsequently used in association the NIR data. The Council sees merit for its use within the targeted upland areas and have declared interest in providing the project with the data for the area, if we can share the results with them.

The Scottish Borders cover a large area of land on the border between Scotland and England, extending from the uplands of the Lammermuir Hills in the north, the Southern Uplands in the west and the Cheviots Hills in the south, through the valleys of the Tweed, Teviot and Till, to the town of Berwick in the east. The area has a rich diversity of habitats and species and it is an area of exceptional natural beauty. Much of the landscapes and wildlife within these areas are recognised as highly important and some are protected through specific nature conservation designations. Some of the most important habitats in the area include:

• Heather, blanket bogs and grassy moorlands on the uplands

• An outstanding variety of wetland habitats, which range from the internationally recognised mosses and lowland raised bogs to large lochs and reservoirs.

• Large areas of forestry plantations

• More intimate areas with small fields of permanent pasture and woodlands.

• Improved farming areas in the more fertile valleys, intermingled with areas of wetland and marsh.

Final Report 12


Figure 1: Map of the Monadhliath Mountains study site

Final Report 13


Figure 2: Map of the Scottish Borders supplementary study sites

Final Report 14


3. Data The main focus of the data activities in this study is on the opportunities presented by EO derived sources. They come from a combination of satellite and airborne sensors and can provide both topographic and thematic information. These EO sources are then complemented by GIS datasets that, whilst often historical and at differing scales and nomenclatures, provide a set of core topographic and thematic information, to which the EO information can be added.

3.1 EO data The primary EO data investigated here are from optical (or passive) sensors i.e. where the satellite requires daylight to acquire an image. Radar (or active) sensors are considered more generally, but have not been processed in this project.

Optical multispectral sensors record a reflected signal from the surface of the earth, and are further categorised by their whereabouts in the electromagnetic spectrum that they operate as shown by the wavelengths in Figure 3.

Figure 3: Electromagnetic spectrum EO provides the opportunity for a consistent, objective mapping, from a selection of sensors over a range of mapping scales. EO-based applications may be deployed for contiguous wide-area coverage, to;

• baseline landscape features.

• record changes in the landscape brought about by seasonality.

• as a tool for building a time series of land cover change and phenology.

• targeted site-specific mapping, e,g. for catastrophic peat erosion event. EO data is not a universal panacea. Not everything can be mapped, all of the time. However, imagery at different working scales and timings can provide field scale to

Final Report 15


the wider area perspective, as well as tracking cause, effect and change not directly possible with field work (including areas difficult to access on foot).

Indices

Utilising the EO data, there are over 150 vegetation indices (indicators of the presence of different vegetation) that can be generated, some of which have been published within scientific literature, but only a small subset have a substantial biophysical basis, or have been systematically tested.

The performance and suitability of a particular index can generally be determined by the sensitivity of the index to a characteristic of interest. `The Normalised Difference Vegetation Index’ (NDVI), the most widely used index separates green vegetation from other surfaces because the chlorophyll of green vegetation absorbs the red light for photosynthesis, and reflects in the near infrared wavelengths due to scattering which is induced by the internal leaf structure. A high NDVI value will therefore indicate high leaf biomass, canopy closure, or leaf area. NDVI cannot itself be used as a direct indicator of land cover type. Image acquisition

The EO data being investigated during this study has been acquired over the period 2006–2007, when off-the-shelf imagery was available (without any tasking of satellites required). Cloud-free or relatively cloud free images have been acquired by the French Système Pour l'Observation de la Terre (SPOT-5) High Resolution Visible Infrared (HRVIR), the US Terra-1 Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) and the Indian Remote Sensing Satellite (IRS) Linear Imaging Self-Scanning (LISS) sensor. Imagery covered both spring and summer so a time series was available for analysis. In addition to the spaceborne images, very high resolution aerial photography (red, green, blue) was purchased from Getmapping, with additional aerial photography (red, green, blue, near infrared) accessed from the Scottish Borders Council for the Tweed Catchment.

A technical description for each of the EO data sources is set out in Appendix 1.

3.2 Ancillary data

The data layers outlined in Table 1 were used to orientate the study team, assist in the development of the classification process and to plan future field working. Specifically, Ordnance Survey MasterMap data provided information on the spatial distribution of roads and some urban areas (e.g., buildings). The NextMap Digital Terrain Model (DTM) data provided information on landscape position. Various vegetation layers were also available through discussions with Scottish Natural Heritage.

Table 1: Used GIS data layers

Data Origin Application

MasterMap Ordnance Survey Topographic structure

1:25,000 maps Ordnance Survey Situation planning

SIACS Scottish Government Topographic structure for agricultural areas

Soils (1:250,000 general Macaulay Institute Contributes to the coverage, 1:25,000 for definition of specific areas) biogeographical zones

Final Report 16


with small supporting role in the development of rules.

Landscape Character Scottish Natural Heritage Defining the area of the Assessment (LCA) Monadhliath

NEXTMap Britain Intermap Elevation, slope, aspect and orthorectified radar image data

National Inventory of Forestry Commission Woodland masking Woodland and Trees (NIWT)

Upland Survey Scottish Natural Heritage Detailed vegetation surveys for familiarisation, training and testing.

National Vegetation Scottish Natural Heritage Detailed vegetation Classification (NVC) surveys for familiarisation,

training and testing.

Aerial photo prints Scottish Natural Heritage Site familiarisation and (Grieve 2005) time series comparison

Table 2: Unused GIS data layers

Data Origin Reason for not using

Land Cover Map 2000 Scottish Natural Heritage Habitat classification too (Originated Centre for broad for use within this Ecology & Hydrology) study.

National Nature Reserves Scottish Natural Heritage Not located within study (NNR) area

RAMSAR sites Scottish Natural Heritage Not located within study area

Sites of Special Scientific Scottish Natural Heritage Not within areas of Interest (SSSI) interest.

Special Areas of Scottish Natural Heritage Not within areas of Conservation (SAC) interest.

4. Image Processing and Data Preparation The successful classification and understanding of the opportunities presented by earth observation require careful initial preparation of the satellite imagery. This comprises three main technical elements; geometric, atmospheric and topographic corrections.

These processes allow us to align the satellite images with other datasets (e.g. in a GIS), then it allows for a consistent application of classification techniques by removing the earth atmosphere effects from the imagery. Errors in any of these corrections compromise the extraction of trajectories of reflectance and associated measures and the successful application of the decision rules, particularly where these require temporal imagery.

Technical details of the image processing that have been applied are presented in Annex 2.

Final Report 17


4.1 Image correction Geometric Correction

The key priority is to register images acquired by sensors with different orbital and observing configurations (e.g., spatial resolutions, viewing angles). All imagery supplied were geo-referenced to the British National Grid.

To do this, we generated Erdas Imagine ground control point (GCP) files. Using independent procedures available in Erdas Imagine we orthorectified and compared the corrected imagery by evaluating against known features (e.g., cairns, wall boundaries, field boundaries, buildings, road junctions), as identified using OS MasterMap data, aerial photography and NEXTMap ORI data.

In the registration process, we sought to minimise resampling by using image to map registration for all scenes (and referring to the GCP database). All images were resampled to 5 m using near neighbour techniques. We consider that images resampled using cubic convolution might be preferable for segmentation as edges are better preserved but nearest neighbour resampling is still best for classification, even though pixels are often removed or duplicated. Once completed the geometrically processed imagery is correctly positioned (see Figure 4) and is by definition, co-registered with all other imagery and spatially referenced datasets (Figure 5).

Figure 4. Registration of SPOT5

(a) Unregistered SPOT5 (b) Registered SPOT5

Final Report 18


Figure 5. Co-registration of SPOT 5 and ASTER satellite imagery

Registered ASTER

Registered SPOT5

Atmospheric Correction

All imagery supplied was calibrated to radiance (W m2 sr-1 µ-1) and subsequently to surface reflectance (%). According to the plan, it was vital that these stages were delivered, and delivered accurately, to ensure the successful development of the rule base. Our previous work has demonstrated that atmospheric correction of optical (e.g., SPOT and IRS sensor) data to surface reflectance is a critical stage in developing consistent decision rules and classifications. Using ENVI FLAASH, the following tasks were completed:

• A sensitivity analysis of key input parameters (e.g., visibility) required by the FLAASH algorithm.

• Reference to Meteorological Office data on the state of the atmosphere at the time of image acquisitions.

• Inclusion of aerosol loadings (from 2000) reported on a daily basis based on observations from coarser resolution optical sensors, namely MODIS and MISR.

• Identification of temporally invariant dark and bright reflectance that can be used to evaluate the success of the atmospheric correction.

The result of this processing is a set of imagery that has comparable reflectance data (see Figure 6), ready for classification, allowing;

• Comparison of reflectance data acquired by different sensors observing on the same date.

• The spectral reflectance curves associated with vegetation to be better interpreted in terms of the photosynthetic pigment amounts (in the visible

Final Report 19


channels), internal leaf structure (near infrared) and moisture content (Figure 7).

• Vegetation indices and endmember fractions can be calculated with greater confidence and inconsistencies associated with the atmosphere were removed.

•

Figure 6. SPOT5 and ASTER reflectance curves for the Monadhliath Mountains

Figure 7. Typical spectral reflectance curve for vegetation.

Final Report 20


Topographic Correction

Topographic correction (presented as units of reflectance, %) help compensate for differences in illumination as a function of time of year, time of day and slope and aspect. This can prove particularly useful for mountainous areas. As assessment was made of the study areas at the start of the project as, whilst the Monadhliath is clearly upland, there were not significant areas of steep slope that were going to affect the classification process. As a result, this correction has not been applied to the imagery, but it is likely to be an important part of the image processing stages for a wider roll out of any EO solution.

4.2 Generation of derived layers

For the classification of the landscape within the Monadhliath, over 30 EO-derived features were utilised, in the form of either a derived layer, or as a ‘customised feature’ created within Definiens Developer software. Derived data layers to be used within the classification included vegetation indices (namely the Normalised Difference Vegetation Index; NDVI), endmember fractions (of bare soil, water/shade, photosynthetic vegetation and non-photosynthetic vegetation; NPV) and topographic features (slope, aspect and convexity). The former were generated through using the fully processed EO data and applying linear spectral un-mixing or band math procedures to each image. The topographic features were obtained from the NEXTMap DTM data using a specialist topographic feature extraction tool, within ENVI.

Other indices were generated within the Definiens Developer software itself by generating an ‘arithmetic function’ using both the equations of known vegetation indices, and through the use of indices created in previous landscape classifications by Environment Systems. With vegetation indices being a combination of surface reflectance at two or more wavelengths that are designed to highlight a particular property of vegetation, each of the vegetation indices generated will accentuate a particular vegetation property and therefore aid in the separation of the various classification classes.

The topographic layers will be used to locate and distinguish between habitats located on steeper and shallower slopes and those inhabiting ‘sink’ areas in the landscape where water commonly accumulates. Particularly this will enable discrimination to vegetation groups such as bogs and mires from grassland areas.

4.3 Development of classification classes

In order to produce a classification of peat erosion it was necessary to split the vegetation up into classes which were of relevance to peat stability and the surrounding area. Several standard systems exist for describing detailed vegetation classes on areas such as the Monadhilaith Mountains e.g. National Vegetation Classification (Rodwell 1992) or Phase 1 habitats (JNCC). However these classification systems are not suitable as they stand, for use with satellite imagery. Phase 1 mapping includes land use as well as land cover classes, which are problematic to map from space.

Experience on work in Wales (Lucas et al., (in press)) has also shown that the Phase 1 classes are too broad, and classes need to be established that can be broken down into areas of similar species coverage’s, or similar mixes of species that are identifiable spectrally from the satellite data. The NVC classification on the other

Final Report 21


hand are too detailed and many classes depend on the presence of plants which are very small in size and infrequent in occurrence throughout the sward, and again these cannot be identified from space. Instead a pragmatic approach was taken to dealing with the vegetation classes as this project was about looking at soil erosion classes rather than habitats. We therefore chose the vegetation classes based on the amalgamation of land cover types that could be separated into the fewest groups by the spectral data from the remote sensing imagery and the air photography. The broad habitats that resulted from this were named with descriptive titles which outline the sort of plant communities that could be expected to fall within these areas.

A full description of all classes, including definition, photo and erosion risk category is presented in Annex 3.

4.4 Selection of testing and training sites

Field work planning

A systematic route was planned around the study area for identifying the erosion features and associated land covers. Analysis was carried out on the number of field samples that needed collecting for each class to ensure that enough replicates were collected to allow for a statistically valid accuracy assessment to be completed at the end of the project. It was established (Congalton and Green) that 50 replicates were needed for each class, with the opportunity to over sample the key classes (e.g. eroding blanket bog) and to under sample classes that we are mapping simply due to their location in relation to peat erosion; but which we are less interested in e.g. acid grassland.

Working with knowledge of the Grieve study, we targeted areas that had been previously visited and surveyed for peat; these areas are NH3500, NN4090, NH5005, NH5505 and NN5095 (see Figure 8 for a subset and Figure 9 for the complete area). Replicates were established for each of the classes (as mapped in draft stage during Phase 1 of the project) and a full set of field maps (based on topographic, satellite, derived imagery and air photos) and recording sheets were produced. Access permissions from the Estates and Glendoe Hydro Scheme were obtained and basic daily routes established for five full days in the field. A full health and safety risk assessment was undertaken.

Field working

The total number of points visited (267) is shown in Figure 8. Each field work day was designed to collect the best coverage of data across the whole study area, whilst collecting enough data for each replicate. The pre-defined sample points were used to structure each day, with supplementary points then collected during transects between each (Figure 8).

A Trimble Nomad GPS (~5 m accuracy) and Garmin eTrex Summit (~15 m) were used to record the location of in-situ data. The Nomad was used to record smaller

features, with the Garmin used to record larger, more homogeneous areas. At each point a record sheet was filled including; unique ID, grid reference, class, description, picture (with photo ID). In addition, areas of relevant peat/associated habitat were mapped on the field sheets.

A key element of the field work was ecologists working with remote sensors. It was only by combining expert knowledge of the manifestation of different surfaces (e.g.,

Final Report 22


vegetation and soils) within remotely sensed data and derived products (e.g., shade fractions, the NDVI) that a full understanding and therefore accurate recording could take place.

Post field work

After the field work, all the data collected was brought together and analysed for completeness. All in-situ points were georeferenced and associated photos checked. The data associated with these objects were then divided into a “training” dataset that can be used to refine and/or develop decision rules and a reserved “testing” dataset that was used to independently evaluate the success of the decision rules in classification.

Figure 8. Comparison of in-situ data points

(a) Survey points, with topographic map (b) Survey points, with SPOT 5 image

Final Report 23


Figure 9. Location of in-situ data collection

Final Report 24


4.5 Image segmentation Segmentation techniques are not a new domain within computation, they are however a relatively new way of classifying remote sensing data. Historically, a process of manual digitisation of areas of interest following an extensive field campaign would be undertaken.

However with the increased necessity for high-spatial resolution imagery analysis, and the availability of commercial or non-commercial software packages, segmentation processes have become increasing useful as these enable a rapid separation of image pixels into a representative form of the real work component they depict (Blaschke et al., 2004).

In high spatial resolution imagery, pixels when grouped together can better represent the characteristics of land-cover groups far better than single pixels can. Therefore groups of adjacent pixels will be organized into objects which will in turn be treated as a minimum classification unit (Yu et al., 2006). These objects are defined as basic entities which are located within an image, where each pixel group is composed of similar digital values, and possesses an intrinsic size, shape, and geographic relationship with the real-world scene component it models. Therefore, the objects can be said to be spectrally more homogeneous within individual regions than between them and their neighbours. The Definiens Developer multi-resolution segmentation algorithm creates image or grid data segments which are primarily based upon three criteria: scale, colour and shape (smoothness and compactness). Scale is the heterogeneity acceptance within a segment whilst colour, smoothness and compactness are all variables that accentuate the segments spectral homogeneity and spatial complexity. A comparison of the segmentation process within Definiens Developer has been noted to have performed more to reality than many others have (Neubert and Meinal 2003). By emphasising the scale rather than compactness, it can allow for polygons to follow natural features more naturally. When shape information was strongly emphasized instead of colour, the resulting polygons were unstructured and did not closely follow feature boundaries. As with previous work, segmentation has been undertaken within Definiens software. Most peat areas will occur in the uplands outside of any SIACS boundaries. For areas outside of the SIACS boundaries, objects of only a few pixels have been generated using the spectral information contained within the imagery. Within discrete areas, consideration will be given to the use of finer (< 5 m) spatial resolution e.g. aerial photography data in complex environments. Where available (given the open, upland nature of the area), we used the SIACS (field boundary) data to align the segments with the boundaries of farm units (e.g., fields) and OS MasterMap layers relating to buildings, roads and rivers will also be used to guide the segmentation.

The segmentation procedure involves the creation of three levels which will be used to classify based upon the level of detail required. Within the upper level (Integration Level), all external dataset (i.e. OS MasterMap) were synced within a chessboard segmentation procedure and classified out. The multi-resolution segmentation algorithm was then used to segment the remaining areas of the image to aid in the removal of all other areas not associated with peat erosion features (i.e. Woodland, Non Peat vegetation) and to separate the areas of blanket bog.

Final Report 25


This segmentation was then carried down to create an EO level and a finer multi- resolution segmentation within the SPOT data applied to areas of blanket bog to aid in the classification of smaller peat features within the EO data.

Finally within the bottom level (Air Photo Level), a very fine multi-resolution segmentation using the aerial photography was applied to pick out the peat erosion features present within the imagery. This segmentation was applied within the areas noted to be peat erosion classes within the EO layer.

Figure 10. Levels of segmentation employed.

Level 1 – external data and chessboard segmentation

Level 2 – satellite EO segmentation

Level 3 – airborne EO segmentation

Final Report 26


5. Analysis – Integration with Current Methods There are a number of ways of classifying EO imagery and associated data. These are typically presented as;

• simple spectral or pixel-based classifiers (supervised and unsupervised classifiers), that are the traditional approach to classifying satellite imagery.

• object-based classifiers (currently the only commercial software available for this is Definiens Developer as used on this project), that include image segmentation and a hierarchical technique for rule based classification that allows for a fuzzy mapping (useful for mapping vegetation continuums).

• Parcel-based classifiers (used on the latest version of the UK Land Cover Map) and these use additional datasets e.g. OS MasterMap to help structure the landscape in the classification. It is worth noting that the Definiens Developer software also allows for parcel-based classification.

• Visual interpretation, which is still a valid and very important technique in the wider process of mapping from EO data. The specialist staff that train the expert systems (e.g. Definiens Developer) require good to excellent satellite and airborne imagery interpretation skills to be able to develop classification rules.

During the course of Phase 1 we have focussed primarily on the object-based classification opportunities presented by Definiens Developer, and have undertaken some simple supervised classifications of the study area to enable comparison of results.

5.1 Mapping of peat erosion features Rule based Classification Approach

A rule based classification was undertaken within the Definiens Developer Software using the classes described in section 4.3. The rule based approach was chosen as it permits knowledge of both the ecology and the content within the EO data to be incorporated into the classification process. Numerically derived rules can then be established based on noted differences and changes within the imagery, and thus used progressively to produce a classification (Lucas et al., 2007).

The Definiens Developer software was chosen due to its ability to take into account both the spatial and spectral information in high-resolution remote sensing imagery, its relative ease in realising the processing of a large remote sensing dataset, its ability to include ancillary information in the segmentation process, and its fast execution. It is also robust, does not have any parameters that require detailed tuning, and is relatively easy to apply the output results in subsequent. To undertake the rule based classification, thresholds were developed for each class using various image layers and from the derived indices. For consistency, the decision rules were based on the opinion of a single operator utilising knowledge acquired during previous projects of a similar nature. Rules for each class were acquired by obtaining upper and lower boundaries for each band through using the ‘feature view’ function within Definiens Developer. The bands that best enabled the discrimination of each class were selected, and the upper and lower boundaries used as selected threshold values.

Final Report 27


A rule base (Figure 10) was devised and implemented using a three tiered hierarchical approach. The first level (Integration Level) was used to remove any unwanted features such as the forestry and urban areas and to distinguish areas where peat erosion features will most likely be present (Blanket Bog) through the incorporation of EO and aerial photography within the segmentation processes. The purpose of the second level (EO Level) was to establish the level of detail with which peat erosion features could be established from the EO data. This was achieved by bringing the classification down from ‘Integration Level’ and re-segmenting within classes that potentially may have peat erosion features present. Through analysis of the EO data, thresholds were then developed for the peat erosion classes developed for this level. The SWIR bands within the SPOT and ASTER were of great importance here. An example of the use of the ASTER SWIR band can be seen in Figure 13 along with a range of other derived indices.

A third level was then introduced to enable the classification of peat features from the high spatial resolution aerial photography. This process again involved the classification from the level above (EO Level) being brought down and a much finer segmentation utilising the aerial photography applied. Peat gullies and areas of bare peat could then be distinguished by generating thresholds using indices and bands taken from the aerial photography. Specialist shape features within Definiens were used within the delineation of gullies as these features were noted to be long and thin within the segmentation produced. The integration, EO and Air Photo levels all combine to form the ‘Core Level’ process tree (as shown in Figure 11). This provides the basic data building blocks for the production of a range of maps. In this example, this then forms the input into the ‘Application Level’ process tree (as shown in Figure 12), illustrating how it is possible to construct maps of peatlands, peat erosion risk, or even other application level maps.

Final Report 28


Figure 11. The Monadhliath Mountains draft rule base, showing the ‘Core Levels’.

Input Data

MasterMap

Water

NIWT

Woodland

MasterMap

Urban

EO Level

Acid

Grassland

Blanket Bog

Montane

Grassland

Bog Myrtle

Flush

Air Photo Level

Bare Peat

Gully

Red Box – Acquired using Air Photo Blue Box – Acquired using SPOT Orange Box – Acquired using SPOT/ Air Photo

Final Report

Bog

Calluna Heath

Wet Heath

Rocky Heath

Bracken

Peat

Erosion

Feature

Peat

Key

Green Box – Acquired using SPOT and ASTER Purple Box – Acquired using ASTER Black Box – Acquired using external datasets

29

Blanket

Bog -

Residual

Ev

Tussocks

Other Blanket

Bog Feature

Blanket Bog -

Eroding

Bare Peat

Bare Peat

Area of

Interest

Rock

Shadow Blanket Bog -

Stable

Peat

Gully

Cloud

Shadow

Vegetation

Non – Blanket

Integration Level

Non - Vegetation

Known Features

SIAC’s Field

Cloud/


Figure 12. The Monadhliath Mountains draft rule base, showing the ‘Application Levels’.

Core Levels (Integration, EO & Air Photo)

Peatland

Areas

• Bare Peat

• Peat Gully

• Blanket Bog –

Eroding

• Blanket Bog – Stable

• Peat Substrate

• Blanket Bog –

Residual

• Ev Tussock

• Bog Myrtle

Final Report

Very High

• Bare Peat •

• Peat Gully

• Ev Tussock •

• Bare Peat

Substrate

•

30

Bracken

• Rock

• Snow

• Water

Man-made

• Cloud

• NIWT

Woodland

Shadow

•

•

User Desired

Non Peatland

Area

Output

Moderate Low

Blanket Bog • Montane

– Stable Grassland

• Flush • Rocky Heath

• Acid • Calluna

Grassland Heath

Wet Heath

Bog Myrtle

Erosion

Risk

High

Blanket Bog •

– Eroding

Blanket Bog

- Residual

•

•


Alternative classification tests carried out

Prior to selecting the Definiens approach, consideration was given to other classification techniques. These techniques, described below, are now considered older and less subtle. A short note on each is provided, with example outputs presented in Annex 4. Supervised and Un-supervised classification approaches

In order to assess the merit for using an object based approach for classifying peat erosion features fro EO data, a supervised and an unsupervised classification approach was applied to both the SPOT data and to a subset of the aerial photography. The Maximum Likelihood approach was chosen as the supervised classification method, and the K-means selected as an unsupervised approach.

Maximum Likelihood Classification

An assumption is made that the statistics for each class in each band are normally distributed. This assumption of ‘normality’ is generally reasonable for common spectral response distributions whereby the response pattern can be completely described by the ‘mean vector’ and the ‘co-variance matrix’ (Lillesand and Kiefer 2004). Given these values, a calculation of the probability that a given pixel belongs to a specific class can be computed. An advantage of using the maximum-likelihood classifier (MLC) is that it assesses both variance and covariance when classifying an unknown pixel (Jollineau and Howarth 2008). Regions of interest representing the same classes at level two for the SPOT classification and at level three for the Air Photo classification were delineated by selecting areas within the corresponding imagery. These were then used as input into the maximum likelihood classifier and a classification produced from these. K-Means Classification

An unsupervised classification does not utilise training data, but instead uses an algorithm that examines unknown pixels in an image and aggregates them into a number of classes based upon the natural groups or clusters present in the image values (Lillesand and Kiefer 2004). Hence, the resulting classes are spectral classes.

The K-means approach enables the analyst to set the number of classes to be present and locates the number of clusters in the multidimensional measurement space. Each pixel in the image is then assigned to the cluster whose attributes mean vector is closest (Lillesand and Kiefer 2004). Class numbers were selected on the

basis of the image imputed and the amount of classes used within the rule base at that level.

5.2 Investigation of ASTER and peat composition

ASTER did facilitate the separation of peat and non-peat areas. It was not though possible to separate the peat up into different compositions, namely classifying the Acrotelm and Catotelm. Figure 13 shows an example of the use of the ASTER SWIR band for the purpose of highlighting peat areas. The dark areas represent peat and the very dark areas bare peat. Final Report 31


Figure 13. ASTER satellite imagery in view, with a Definiens report on the coincident datasets available (red dot) for analysing

Final Report 32


Figure 14. Soil map of the Macaulay (organic soils presented in shades of purple)

Final Report 33


5.3 Peat erosion and impacts upon soils Scotland has a diverse soil coverage driven by the range of different geology, climate and landform. Many scales of landform processes operate on the soils of Scotland; there are some basic trends at the broadest scale of mountain ranges, driven by their underlying geology and climatology. But, also at a micro level the soil types can be very different from one slope to another, dependant on the steepness of slope whether it is shedding or receiving water, its aspect and altitude.

Across Scotland the Macaulay Institute have mapped the country in terms of broad soil types at 1: 250,000 scale (Soil Survey of Scotland). The lowlands of Scotland have been mapped in more detail at 1: 63,360 or 1:25,000 scales. Both of these scales of mapping provide a very good picture of the soil types of Scotland and the broad macro level. The more detailed maps also give a good indication of soil types and location on a more local setting. The data sets are therefore of great use and importance in designing bio-geographical zonings of the landscape and looking at what sort of soil types can be expected and where these will be found. However the peat erosion work carried out as part of this project, must take place at the micro level as we are attempting to find features with are a metre or less in size (with the mapping of peat gullies aiming to pass the ‘jump test’). If a broader scale map is introduced into the rule base as a thematic layer to identify land which can then be considered in more detail, it masks many of the affects of erosion because of the generalisation of the mapping. We have therefore found it expedient to use the soil maps as an excellent information resource, but not as a thematic layer within the classification.

Their importance would increase during any future roll out of an EO-based system for mapping peat erosion, helping to define and target biogeographical zones across the country. Figure 14, shows the extent of different soils in the Monadhliath area from the Macaulay Institute soils data.

6. Maps of Classification Results Maps developed from each of the Core and Application processing levels, as previously set out in Section 5.1, Figure 10 and 11, have been produced and are presented below.

Final Report 34


Figure 15. EO level classification

Final Report 35


Figure 16. Air photo level classification

Final Report 36


Figure 17. Peatland level classification

Final Report 37


Figure 18. Erosion risk level classification

Final Report 38


7. Precision and Accuracy

7.1 Mapping of peat erosion features

Both a qualitative (visual) and quantitative (statistical) accuracy assessment have been carried out. The results of this are presented below.

Qualitative Accuracy

To present the qualitative accuracy of the classifications, a series of comparison maps are presented. At the scales presented, they demonstrate similarity of pattern and colour. Looking at the full classification in a GIS system will allow for a more detailed qualitative accuracy assessment. The following maps are presented;

• Figure 19. Erosion risk and aerial photography classifications in comparison with

the 25 cm aerial photography • Figure 20. Earth Observation classification in comparison with the ASTER data • Figure 21. SPOT and aerial photography data with related classifications.

• Figure 22. Gully field sampling and comparison with gully classification at the

Earth Observation Level

Final Report 39


Erosion risk and aerial photography classifications in comparison with the 25 cm aerial photography

Aerial Photography Level Classes

Bare Peat

Peat Gully

Acid Grassland

Blanket Bog - Eroding

Peat Substrate

Blanket Bog - Stable

Blanket Bog - Residual

EV Tussock

Water

Non Peatland Area

Figure 19.

Final Report

Erosion Risk

Very High

High

Moderate

Low

Non Peatland Area

40


Figure 20. Earth Observation classification in comparison with the ASTER data

Final Report 41


Figure 21. SPOT and aerial photography data with related classifications

Final Report 42


Final Report 43


Figure 22. Gully field sampling and comparison with gully classification at the Earth Observation Level

Final Report 44


Quantitative accuracy

In Phase 2, following the successful completion of the field work and production of the Definiens classification, it was possible to establish statistical measures of accuracy.

Background to quantifying accuracy

Within accuracy assessment for a land cover maps there are several layers of uncertainty. The primary layer of uncertainty is found in the field due to ecotones between one cover type and another e.g., where a degraded bog becomes a peat erosion feature is not necessarily a ‘hard’ line on the ground. In many cases the slump at the end of a peat hag has resulted in clumps of bog vegetation within the bare peat area. Where these are frequent the area is eroding bog, where they are infrequent the area would be considered ‘bare peat’, but the land in-between has a degree of uncertainty. This uncertainty is mirrored within the fuzzy mapping of the vegetation and peat classes. In areas where the vegetation meets the whole criteria of one class

we are very certain about it group, where the criteria are less definite we are less certain of it. A map could therefore be drawn which moves the line on the ground from ‘most

likely area of bare peat’ to ‘any likely chance of bare peat’. This uncertainty is also reflected in the various type of bog vegetation especially when the frequency of a cover plant makes the difference between one vegetation type and another. For example intact bog will have a moderate amount of Eriophorum vaginatum with it, but eroding bog is likely to have more. Where the imagery is identifying Eriophorum vaginatum both types of bogs are possible, one being just slightly more likely than another.

A further level of uncertainly is added by the process of capturing information in the field. We will have accurately identified a peat erosion feature, but in collecting the data we may have recorded it up to 5 metres away from its ‘actual location’ (due to GPS accuracies). Five metres is good when working in an environment like the Monadhliath, but as the imagery is very well georeferenced, this inaccuracy in recording the field data leads to a further source of uncertainly and error. In order to build an error matrix that takes these issues into account it is necessary to move away from straight boolean classes, as they do not exist on the ground. Instead we need to introduce the concept of plausibility. For the example explained above; on the edge of the eroding bog, a field data collection point that recorded bare peat, eroding bog, or tussocks would have an equal chance of being a ‘correct’

Final Report 45


interpretation of the situation on the ground. Added to this is the scale of mapping issue; for example our minimum mapping unit within the air photo level is a metre; that means any feature of less than a metre is going to be consumed within another class. Within the satellite data, the minimum mapping unit is 5 m; that is anything of less than five square metres will be consumed into a surrounding class. These two scale issues mean it is incorrect to use point data for accuracy assessment as points in themselves would not exist as a class in their own right. Instead features of the size of the minimum mapping unit must be created, so that classes can be accurately described. Within our field data collection we also have to deal with the accuracy of the field collection devices, therefore any features within a five metre radius of our mapped point is also a ‘plausible‘ class.

When we come to examine the accuracy therefore, we could have several data tables or matrices. One is for where the field class is exactly the same as the map class or is a plausible explanation of the map class. Another table would show where the map class seems to be correct and the field class incorrect or visa versa, or there is a possibility that neither of the data are correct. In order to resolve these disagreements a further data set would be needed. This could be an air photo interpretation exercise.

Once each of the matrices has been resolved, matrix algebra could be used to report the outcome of the error investigation. Within this report we have just reported the basic error with those that agree or are plausible against those that disagree. We have not tried to delineate whether the disagreement arose from the field or map data being incorrect. As the accuracy assessments of these types of maps is very much still a research topic we are aware that this is a simplistic explanation of a more complex truth. The result (and current view of on going research) is that any accuracy tables shown are likely to understate the potential of the map to describe the situation on the ground in a useful way; that is real accuracy is likely to be higher than the ones reported here. Accuracy calculation An error matrix allows for the comparison on a category by category basis of the relationship between known reference data (‘ground truth’), and the corresponding results from the classification process (Lillesand and Kiefer 2004). Multiple descriptive measures can be derived from these error matrixes which include overall accuracies, user and producer accuracies and the kappa coefficient. Overall accuracies are computed by dividing the number of correctly classified pixels or objects by the total number of reference pixels or objects. An overall accuracy of 75% when using a direct pass/ fail accuracy approach suggests that the accuracy of the classification was good, but not perfect within confusion existing between bare peat and eroding blanket bog for example. However when bringing a plausible fail logic into the assessment, the accuracy increases to 84%. As each object is composed of representative pixels, there is the implicit assumption that the image is composed of pure pixels. Remotely sensed data is often dominated by mixed pixels that contain more than one class and therefore this is a major problem when obtaining accuracy. This explains the confusion that exists between bare peat and blanket bog - eroding classes.

User accuracy is obtained by dividing the number of correctly classified pixels/ objects in each category by the total number of pixels that were classified in that category. Producer accuracies however are obtained by dividing the number of correctly classified pixels per category by the number of training samples per

Final Report 46


category (Lillesand and Kiefer 2004). Good producer accuracy indicates that the reference data was correctly mapped, whereas high levels of user accuracy indicate that the land cover map has a good correlation with the reference data.

The classes of rocky heath and blanket bog – eroding, both had high producer and user accuracies which indicates that there is a very good correlation between the reference data, the classification produced and the land cover within the classification area. The class of blanket bog - stable however had much higher producer accuracy in relation to producer accuracy which suggests that although the reference data was correctly mapped, the relation between the reference data and the land cover map was not as good.

The kappa coefficient, describes when the value is closer to one then there is a true agreement regarding the extent to which the percentage correct values of an error matrix are compared to. When it’s closer to 0 then it’s seen as being ‘chance’. As the kappa coefficients achieved in this classification are 0.68 and 0.80, it can be said there is a true agreement and therefore the classification is between 68 and 80% better than one resulting from chance. Accuracy statistics were calculated for the map from in-situ field data collected by the project team (see Table 3).

Final Report 47


Users

% % Plausibly

Correct Correct

55 61.82 92.73

77 89.61 90.91

79 68.35 68.35

35 71.43 88.57

21 90.48 95.24

Overall

75.28

% Plausibly Correct

Kappa coefficient

% Correct

% Plausibly Correct

Class key

Class A – Bare peat

Class B – Blanket bog - eroding

Class C – Blanket bog - stable

Class D – Rocky heath

Class E – Acid grassland

Final Report

Table 3. Classification accuracy

Classes

A

B

C

D

E

Total

Ground Truth Circles

C D E

0 1 3

6 1 0

54 0 22

0 25 6

1 1 19

61 28 50

88.52 89.29 38.00

72.13 92.86 58.00

Pass

Plausible

Fail

48

Total

267

Overall

A B

34 17

1 69

0 3

0 4

0 0

35 93

97.14 74.19

40.00 88.17

0.68

0.80

Overall

84.64

Producers % Correct

Overall


7.2 Evaluation of suitability for purpose Evaluation of the effectiveness of satellite and aerial data

The purpose of this study is to assess the erosion of organic soils in Scotland using EO. As set out earlier in this report, EO is any system that remotely collects information about the surface of the earth; whether it is space or airborne. To begin with the focus of this work was largely on the spaceborne EO component due to the greater complexities of the pre-processing and interpreting. In practice, equal focus was been placed on both the air and space-derived information. The scale of working that is required for mapping the erosion features (e.g. the peat gully ‘jump test’) means that VHR and HR satellite imagery can contribute to the mapping process (in particular by providing a time series of thematic information) but to get the best possible delineation of peat erosion features currently airborne provides the optimum scale. Historically this required manual digitisation and interpretation, which is both time consuming and costly. The opportunities presented by applying satellite-derived methods of processing (e.g. image segmentation and feature extraction) mean that aerial photography can now be processed in an efficient and controllable manor to provide the topographic structures for classification in areas that are often devoid of any features mapped by the Ordnance Survey. Evaluation of satellite data for classifying peat erosion features

Specific small scale peat erosion features could not be established through the use of EO data such as SPOT and ASTER. These sensors do not possess the spatial resolution to permit the identification of peat gullies. It was however possible to establish the larger areas of bare peat which are present. An example of the classification results achievable through the use of EO data is shown in Figure 23. Potential areas of peat are clearly distinguishable (purple) with the areas of bare peat shown in black.

Evaluation of aerial photography for classifying peat erosion features

The use of aerial photography enabled the smaller scale peat erosion features to be classified (Figure 24). Features such as peat gullies could be classified and to a high level of detail. The use of aerial photography though did not enable the discrimination between areas of bare peat, and dark vegetated areas containing species such as Calluna vulgaris, therefore it was necessary to use the satellite –based EO to initially target areas, within which the finer detailed imagery could be applied.

Final Report 49


Figure 23. Peat classification (left) of ASTER (right)

Figure 24. Peat classification (left) of aerial photography (right)

Final Report 50


The use of the CIR (colour infrared) aerial photography though did enable this separation between bare peat and heath species such as Calluna Vulgaris. CIR data was tested for an area within the Scottish Borders (White Coomb) within an area containing evidence of peat erosion. A clear separation can been see in Figure 25 whereby peat is classified in yellow, cleared areas in orange, and Calluna Vulgaris in purple. Figure 25. Definiens airborne infrared classification in the Scottish Borders. (a) original infrared aerial image

(b) Classified image

Legend

Cleared Ground

Bare Peat

Calluna Heath

Other Heath

Final Report 51


Evaluation of Object vs Per Pixel Classification

It was apparent that the classifications produced using an object orientated rule based approach were far more representative of the Monadhliath landscape. They allowed for the use of many external datasets and derived layers/ indices within the classification process which removed the many classification confusions that existed through using the pixel based classification approaches. This was noted by Smith and Fuller (2001) who attributed this confusion to noise in the data, atmospheric effects and natural variations within the land-cover type. These factors will therefore adversely impact upon the spectral information present within each pixel, and ultimately the accuracy of the classification produced. By incorporating this information, knowledge on the context of the landscape used within the classification process will permit accuracies attainable to be considerably higher. The object based classification produced within Definiens Developer made use of features present within the landscape such as a DTM and MasterMap stream vectors in the classification of specific land-cover classes.

Gaps and opportunities

Phase 1 of the project has been based on the use of multispectral satellite and airborne imagery, combined with GIS data. It is considered that this is a robust basis for operational mapping of peat erosion; however consideration has also been given to the opportunities presented by incorporating;

Figure 26. TerraSAR-X image

• Synthetic Aperture Radar (SAR) imagery (e.g. TerraSAR-X, Figure 26) to help characterise and extract information on the structural attributes of the terrain and its land- cover characteristics.

• Radar expertise to help with the processing and development of algorithms for integrating SAR into a classification production chain. SAR processing is inherently different to that of

multispectral imagery.

Final Report 52


• Unmanned Aerial Vehicles (UAVs). A UAV is an aircraft designed or adapted to operate with no human pilot on-board which, when combined with some form of payload, avionics, and appropriate (ground-based) infrastructure offer a flexible and relatively low cost EO data source. UAV technologies and wider operational infrastructure are developing rapidly e.g. the miniaturisation of sensors and data storage. Currently the main limitation for routine deployment is the regulation of civil airspace, requiring UAVs to be flown within direct visual range of the ‘pilot’. It maybe a few years yet, but in the medium term this is likely to be an important tool in the box.

• Airborne hyperspectral has not been explicitly studied in this project, but existing knowledge of the offering makes it clear that the technology provides opportunities as a source of a large amount of spectral data, but over specific targeted areas e.g. strips of 500 m. The additional information provided by a hyperspectral sensor would be valuable over sites at very high risk for detailed monitoring.

• A systematic acquisition programme of all available SPOT5, ASTER and IRS imagery for Scotland. This acquisition programme (an equivalent is currently underway in Wales) provides a library of information about the surface of the earth in Scotland. Allowing for both a multi-faceted and ongoing and retrospective viewing and analysis.

• All image pre-processing (geometric, ortho, radiometric and topographic corrections) should be undertaken to a high specification. Off-the-shelf services are available from the image suppliers, but consideration should be made of the input data used (e.g. topographic maps) in these processes. This is of particular importance when subsequently undertaking automated processing (e.g. multi-stack image segmentation and classification).

• Topographic correction should be performed when undertaking a systematic, operational programme of EO-based mapping in areas of varying terrain.

• Cloud masking should be considered to enable the use of partially cloudy images in the mapping chain.

Biogeographical zones

As the reflectance of vegetation (from EO imagery) varies as a function of, for example, season and the physical environment, the division of the landscape into distinct biogeographical regions that best captured this variation is desirable. It is proposed that a series of biogeographical zones are established, based on knowledge of habitat distributions and environmental influence across Scotland. Between and even within these regions, several habitats could vary in terms of timing of leafing, flowering and senescence but also their overall structure and species composition. The biogeographical zones, unless they exist already for Scotland, can be produced primarily from national data on soils, elevation and rainfall. They would only need to be broad (estimated at around 50-60 to cover Scotland) to allow for the logical handling of how habitats present in EO and also in the logical sub-division of Definiens image processing tasks.

Final Report 53


Monitoring

Assessing peat erosion is inherently time-based i.e. a time series of information is required to be able to assess the change in habitat cover over time. Currently the best available dataset specific to this is the Grieve study. As part of the assessment on this project, the original stereo air photo prints (see Figure 27) were examined and their possible role in monitoring was discussed. The study, originating from 1995, has an information rich dataset in the stereo pairs of black and white aerial photo prints. These were interpreted using a stereoscope and consideration was given to how the information could be used;

1. Mapping of peat gullies and bare peat from manual stereo air photo interpretation.

2. Scanning, georeferencing and the digital mapping of peat gullies and bare peat from air photo interpretation.

3. Targeting areas of change for identification of a time series of air photos.

Final Report 54


Figure 27. Comparison of (a & d) air photo prints, digital air photos (b & e), classification (c & f)

(a) 15/6/1989 Stereo Photograph (b) 2006 Aerial Photography (c) Bare Peat and Peat Gully classification

(d) 15/6/1989 Stereo Photograph (e) 2006 Aerial Photography (f) Bare Peat and Peat Gully classification

Final Report 55


8. Implementation Plan The starting point for establishing an implementation plan was a discussion with the Scottish Government on key preferences and options for the roll out of any future system. Information was collected on current data and software licensing, hardware and staff skills.

The Scottish Government currently has;

• A wide range of GIS topological and thematic data together with air photo coverage. The only additional data needed would be the satellite imagery (and additional air photo coverage for future change detection mapping).

• Wide access to ESRI GIS software for processing vector and some raster data, together with limited licences of Erdas Imagine and ER Mapper to cover most of the image pre-processing requirements.

• Good quality PC hardware, but consideration would need to be given to dedicated hardware and storage.

• Some skills in image processing, but not for routine and large scale work. There are currently no known skills in advanced classification (e.g. Definiens).

The likely implementation plan is to use external expertise in the short term; followed by capacity building, then the full take-up of EO-based peat erosion mapping by the Scottish Government if required (Table 4). Table 4. Implementation handover

Delivery 2010 20xx

External External/internal

Internal

Based on this framework, work was carried out, calculating the broad cost of initially delivering the work externally (by consultants) based on the method, results and lessons learnt from this project. This is presented in Table 5. The costs have been calculated for mapping a similar size area to the Monadhliath. The Monadhliath is approximately 2,500 sq km, which equates to roughly 1/30th of the area of Scotland (78,772 sq km).

Based on the soils map of Scotland, and the identification of land characterised as having organic soils, approximately 65% (ECOSSE, 2007) of the country will need mapping to identify possible peat erosion.

Clearly economies of scale will be achievable, market conditions at the time of testing, and the final specification for the mapping will all contribute to the final budget required. Final Report 56


The outputs from this EO-based system will include;

• Production of the ‘core level’ data

• Production of peat erosion ‘application level’ data

• Opportunities for additional application level data, based on the core level data e.g. habitat maps, carbon budgeting (with knowledge of peat depth), ecosystem goods and services, planning (e.g. wind farm applications).

Following agreement with the Scottish Government, a direct comparison has not been made on the financial cost of implementing a field-based survey of peat erosion across Scotland.

Notes to outline costs

The following items have not been included below (Table 5), but should form part of the financial element of any wider case or decision;

1. Project management (can account for 10% if mapping small-medium areas, likely to extend to 15%+ if run as a larger single programme).

2. Current digital air photo coverage of the areas of Scotland with organic soils; ideally with a concurrent near infrared band (the collection of which is now standard by all UK air photo acquisition companies).

3. Specialist software not currently licenced by Scottish Government e.g. Definiens Developer. At time of writing embracing Definiens Developer would include;

• Software licence ~ £25,000 (includes 1 year maintenance)

• Training (initial familiarisation, 1 week intensive formal training, immediate and on going software use) ~ £5,000

4. PC and/or server hardware for storage and processing of large volumes of EO and GIS data (~50Gb of permanent space, with 2-3 times that for temporary processing of the Monadhliath data).

5. Scanning and digitising the Grieve black and white stereo aerial prints.

6. SPOT5 (or equivalent) satellite tasking to take the form of ‘work horse’ imagery for mapping. All satellite costs presented above are based on archive imagery.

7. VAT.

Final Report 57


Table 5: Outline costs for implementing the EO-based system (Monadhliath size area)

Assumptions

All data (inc. soils) is available from Scottish Government under licence and at no additional cost.

A ‘Monadhliath’ size area to be completed with 3 scenes. Based on 1 day per scene for geometric, atmospheric and topographic correction, plus 2 days for generating end-members and final checks.

(estimated Scotland coverage: ~45 scenes for SPOT5, 5 scenes for IRS, 40 scenes for ASTER).

A knowledgeable field team (2 people for 3 days), built around an ecologist with good understanding of EO information content. It is crucial for the ecology team to be closely integrated with the remote sensing team.

Utilising the Monadhliath rule-base, with biogeographic- related fine tuning.

The Monadhliath area takes 6 hours to process (counted here as essentially zero cost). This would be planned to run over night, with a short period of checking the following morning.

Calculation of accuracies, map production etc.

Days

1. Data collation

2. Image processing

3. In-situ data collection

4. Rule development

5. Classification processing

6. Post

processing

TOTAL

Final Report

Average day Total time Total fixed rate (£) cost cost

2 350 700 -

5 450 2,250 SPOT5

4,000

IRS

1,600

ASTER

55

6 400 3,000 600

5 550 2,750 -

1 550 550 -

1 450 450 -

£9,100 £6,255

58


9. Conclusions The aim of this study was to build on existing knowledge from research projects ongoing in Scotland and the UK on upland peat erosion. A considerable wealth of knowledge already existed into the types and causes of upland peat erosion. There are also studies which have evaluated erosion using different methods ranging from field measurements at set points through to detailed air photo interpretation of rill and gully size and pattern.

Of the main types of peat erosion in the uplands, large single episodic events such as bog bursts and peat slides are the easiest to sense and trace remotely. However these events are infrequent and although of great extent in themselves, have been shown to be less significant than continual local erosion through gulling and rills on blanket bogs at high altitude and shallow slopes. The focus of this work has been on the identification of these gullies and rills through the application of EO.

EO, and in particular the combination of airborne and spaceborne imagery, provides the opportunity for a consistent, objective mapping, from a selection of sensors over a range of mapping scales.

Following assessment, an approach utilising an object orientated rule based classification was chosen as it permits knowledge of both the ecology and the content within the EO data to be incorporated into the classification process. The results from both Phase 1 and 2 show that using a combination of very high resolution aerial photography, with high and very high resolution multispectral satellite imagery it is possible to integrate these with knowledge of the landscape processes to extract erosion features at the appropriate fine scale.

Success of the project came from not simply considering this as an ecological project, or as a remote sensing project. Rather the success came from coupling ecological knowledge, with remote sensing expertise and EO information content. The combination of this skill set cannot be underestimated.

The traditional presentation of EO accuracy information has been questioned and proposals have been made that there is a need to move away from the concept of crisp Boolean mapping to one based on upper and lower bounds of belief; to try to manage and represent the variation and uncertainty on the ground (e.g., mixed pixels, mosaics, etc). Particularly given promotion of deeper non-expert access to spatial data and environmental decision making through a range of initiatives.

During the Phase 2 work, consideration has been given to a possible future classification (in particular the rule-based approach) across Scotland and the successful transfer of rules established in one area to another.

It is proposed that the Scottish Government should adopt the use of EO for mapping and monitoring peat erosion across Scotland. Initially through a short project to confirm the transferability of the approach to a second area of Scotland. This could take the form of a ‘rapid implementation’ to assess and confirm the biogeographical effect on the process. A project of this nature would also provide an additional short term output for the Scottish Government, including further examination of additional ‘application levels’ that could usefully be generated from the ‘’core levels’ in support of policy delivery, whilst full funding is being considered.

Final Report 59


10. Acknowledgments The Environment Systems team (Steve Keyworth, Dr. Katie Medcalf, Mark Jarman, Johanna Breyer, Stephen Ray and advisors for Phase 1, Dr. John Scullion, Dr. Pete Bunting, Professor Richard Lucas, Dan Clewley (Aberystwyth University) and Phase 2 advisor Dr. Ian Grieve (Stirling University)) are grateful for the knowledge, advice and support of the project steering group, including Geeta Puri (Scottish Government), Matt O’Donnell (BNSC) and Richard Turner (GIFTSS).

The project team are very grateful to Andrew Coupar for the invaluable discussions whilst joining us on the first day of field work on the Monadhliaths; Dr. John Gordon for the loan of the air photo prints from the Ian Grieve study; Dr. Andy Tharme (Scottish Borders Council) for his support and use of the airborne colour infrared imagery over the Scottish Borders study sites. Also to the discussions and data provided by Dave Horsfield and Jenny Bryce (Scottish Natural Heritage) and to Alan Brown (Countryside Council for Wales).

Final Report 60


11. References

Ahlqvist, O., Keukelaar J. and Oukbur, K. (2000). Rough classification and accuracy assessment. Int. J. Geographical Information Science, 14(5): 475-496.

Batjes, N. H. (1996), Total carbon and nitrogen in the soils of the world. European Journal of Soil Science, 47, 151–163.

Birks, J. and Ratcliffe, D.A. (1980) Classification of upland vegetation types in Britain. Nature Conservancy Council, Peterborough. Birnie, R.V, Puri, G (2007) Scottish Government and British National Space Centre Workshop, Assessing the extent and severity of erosion on the Upland organic soils of Scotland by remote sensing, Summary Report.

Blaschke, T., Burnett, C. and Pekkarinen, A. (2004). Image segmentation methods for object-based analysis and classification, Remote Sensing Image Analysis: Including the Spatial Domain (S.M.D) [In] Yu, Q., Gong, P., Clinton, N., Biging, G., Kelly, M. and Schirokalle, D. (2006). Object-based Detailed Vegetation Classification with airborne high spatial resolution remote sensing imagery, Photogrammetric Engineering and Remote Sensing, 72, 799-811.

Brady, N.C. and Weil, R.R. (2002), The Nature and Properties of Soils – Thirteenth Edition, Pearson Education Ltd, New Jersey. Bridgham, S.D. Pastor, J. Dewey, B. Weltzin, J.F. and Updegraff, K. (2008), Rapid Carbon Response of Peatlands to Climate Change, Ecology, 89, 3041–3048. Burrough, P. A., McDonnell, R. A. (1998) Principles of Geographical Information Systems: Spatial Information Systems and Geostatistics. Oxford University Press, Oxford.

Comber, A.J., Law, A.N.R. and Lishman, J.R. (2004) Application of knowledge for automated land cover change monitoring. International Journal of Remote Sensing, 25, 3177-3192.

Congalton, R.G. and Green, K. (1993) A practical look at the sources of confusion in error matrix generation. Photogrammetric Engineering and Remote Sensing, 59, 641- 644.

Congalton, R.G. and Green, K (1999)., Assessing the Accuracy of Remotely Sensed Data: Principles and Practices, Lewis Publishers.

Curran, P.J. (1985) Principles of remote sensing. Longman, London.

Cutler, M.E.J. McMorrow, J. and Evans, M. (2002), Remote Sensing of upland peat erosion in the southern Pennines, North West Geography, 2, 1476-1580. Dayton, N. (2006), An assessment and evaluation of herbivore impacts on blanket bog habitat in the Monadhliath Special Areas of Conservation. Scottish Natural Heritage Commissioned Report No.165.

Definiens (2008). http://www.definiens.com/

Final Report 61


ECOSSE: Estimating Carbon in Organic Soils - Sequestration and Emissions: Final Report, 2007, http://www.scotland.gov.uk/Publications/2007/03/16170508/0 (Accessed 25-09-09).

Encyclopaedia Britannica (Online) 2008, http://www.britannica.com/EBchecked/topic/388785/Monadhliath-Mountains, Accessed (4-4-09).

European Commission (COM) (0nline) 2006, Soils, http://ec.europa.eu/environment/soil/index_en.html, (Accessed on 15-4-09) Fuller, R.M., Groom, G.B. and Jones, A.R. (1994). The land-cover map of Great Britain: an automated classification of Landsat Thematic Mapper data. Photogrammetric Engineering and Remote Sensing, 60, 553-562.

Fuller, R.M., Smith, G.M., Sanderson, J.M., Hill, R.A., Thomson, A.G. (2002). The UK Land Cover Map 2000: Construction of a parcel-based vector map from satellite images. Cartographic Journal, 39, 15-25. Grieve, I.C., Davidson, D.A. and Gordon, J.E. (1995) Nature, extent and severity of soil erosion in upland Scotland. Land Degradation and Rehabilitation, 6, 41-55.

Grime, J.P., Hodgson, J.G., Hunt, R (1998) Comparative Plant Ecology: A functional approach to common British Species. Unwin Hyman, London. Harris, P. M., and Ventura, S. J. (1995). The integration of geographic data with remotely sensed imagery to improve classification in an urban area. Photogrammetric Engineering and Remote Sensing, 61, 993–998.

Johnson, G. M., and Morris, J.M. (2000), Scottish Blanket Bog Inventory: Grampian Region and adjacent areas – Characterisation of blanket bogs using Landsat Thematic Mapper. Scottish Natural Heritage Research Survey and Monitoring Services. No. Perth.

Jollineau, M. Y., and Howarth, P. J. (2008). Mapping an inland wetland complex using hyperspectral imagery, International Journal of Remote Sensing, 29 (12), 3609- 3631.

Lillesand, T.M. and Kiefer, R.W. (2004). Remote Sensing and Image Interpretation, John Wiley and Sons INC.

Lindsay, R. and Immirzi, P. (1996), An Inventory of Lowland Raised Bogs in Great Britain. [In] The sensitivity of peat-covered upland landscapes, Catena, 42, 345–360. Lucas, R.M., Rowlands, A., Keyworth, S., Brown, A. and Bunting, P. (2006f). Mapping habitats and agricultural land covers in Wales using multi-temporal remote sensing data. International Journal of Photogrammetry and Remote Sensing, 62 (3), 165-185.

Lucas, R., Rowlands, A., Brown A., Keyworth S. and Bunting, P. (2007) A Rule- based classification of multi-temporal satellite imagery for habitat and agricultural land-cover mapping, ISPRS Journal of Photogrammetry and Remote Sensing, 62, 165–185.

Final Report 62


Lucas. R., Brown, A., Bunting. P., Medcalf. K., Clewley. D., Breyer. J. and Keyworth. S., (In Press) Updating the Phase 1 Habitat Map of Wales using Satellite Sensor data.

Macdonald. A., Stevens. P., Armstrong. H., Immirizi, P. and Reynolds. P. (1998) A guide to upland habitats surveying land management impacts [In] Dayton, N. (2006), An assessment and evaluation of herbivore impacts on blanket bog habitat in the Monadhliath Special Areas of Conservation. Scottish Natural Heritage Commissioned Report No.165.

McMorrow, J.M., Cutler, M.E., Evans, M. and Al-Roichdi, A. (2004), Hyperspectral indices for characterising upland peat composition. International Journal of Remote Sensing, 25 (2): 313-325.

Nature Conservancy Council (England Field Unit) (1990) Handbook for Phase 1 habitat survey: A technique for environmental audit. Field manual Dept PHS Nature Conservancy Council, Peterborough. Neubert, M. and Meinel, G. (2003) Evaluation of segmentation programs for high resolution remote sensing applications. Proceedings of the Joint ISPRS/EARSeL Workshop: High Resolution Mapping from Space 2003. (ed. by M. Schroeder, K. Jacobsen and C. Heipke), Hanover, Germany, October 6-8, 2003. Rodwell, J.S. (ed.) (1991). British Plant Communities. Volume 1. Woodlands and scrub. Cambridge University Press.

Rodwell, J.S. (ed.) (1991). British Plant Communities. Volume 2. Mires and heath. Cambridge University Press.

Rodwell, J. S. (ed.) (1992). British Plant Communities. Volume 3. Grassland and montane communities. Cambridge University Press. Rodwell, J.S. (ed.) (1995). British Plant Communities. Volume 4. Aquatic communities, swamps and tall-herb fens. Cambridge University Press.

Rodwell, J.S. (ed.) (2000). British plant communities. Volume 5. Maritime communities and vegetation of open habitats. Cambridge University Press. Smith, G. M. and Fuller, R. M. (2001). An integrated approach to land-cover classification: an example in the Island of Jersey, International Journal of Remote Sensing, 22 (16), 3123–3142.

Tallis, J.H. (1998), Growth and degradation of British and Irish blanket mires [In] The sensitivity of peat-covered upland landscapes, Catena, 42, 345–360. Taylor, J.A., (1983), The peatlands of Great Britain and Ireland [In] The sensitivity of peat-covered upland landscapes, Catena, 42, 345–360.

The Met Office (Online) 2008, www.metoffice.gov.uk/climate/uk/location/scotland (Accessed 1-4-09).

Wang, L., Sousa, W. P. and Gong, P. (2004). Integration of object-based and pixel- based classification for mapping mangroves with IKONOS imagery, International Journal of Remote Sensing, 25 (24), 5655-5668.

Final Report 63


Yu, Q., Gong, P., Clinton, N., Biging, G., Kelly, M. and Schirokalle, D. (2006). Object- based Detailed Vegetation Classification with airborne high spatial resolution remote sensing imagery, Photogrammetric Engineering and Remote Sensing, 72, 799-811.

European Commission (COM) (0nline) (2006), Soils, http://ec.europa.eu/environment/soil/index_en.htm , (Accessed on 15-4-09).

Final Report 64


Annex 1 Satellite and Airborne Sensor Specifications SPOT5

SPOT has four colour bands and an optional panchromatic band (the latter has not been used here). The four colour bands are green, red, near-infrared and mid- infrared.

Satellite SPOT5

Scene Size ¼ scene (30*30km)

Image acquisition data 08/05/06

Angle of incidence 27.27o

Resolution 10 m

Spectral Coverage Green-SWIR

GRS Scene Reference 17-233

Acquisition time 12-05pm

Sun Azimuth 177o

Price £1,239 (£930 1/8 scene)

ASTER – (Advanced Spaceborne Thermal Emission and Reflection Radiometer) The ASTER imagery is acquired within 14 spectral bands with 15-90 m resolution depending upon the bands used. (VNIR - 4band – 15 m, SWIR - 6bands – 30 m, TIR - 5bands – 90 m)

Satellite ASTER

Resolution 15 m

Scene Size Full Scene

Image Acquisition Data 22/08/2007

Scene ID SC:AST_L1A 003:2046459185

Cloud Cover 0%

Sun Azimuth 163.15o

Pointing Angle -8.5o

Price £55

IRS – (Resoursesat-1) The IRS-P6 satellite operates within the green to SWIR and carries three different resolution cameras. The only imagery available for the study area is through the medium resolution camera (LISS-III) which operates in three spectral bands in VNIR, and one in Short Wave Infrared (SWIR) band with 23.5 metre spatial resolution.

Satellite IRS-P6 LISS-III

Scene Size ¼ scene (70*70 km)

Resolution 20 m

Spectral Bands 4 (vis.,NIR,SWIR)

Image Acquisition Data June 2006 Scene ID

Cloud Cover

Sun Azimuth

Price £1,622

Aerial Photography The aerial photography prices have been sourced through Getmapping and are priced based upon the study area size. The aerial photography purchased was 25 cm

Final Report 65


and was captured in 2005. This finer spatial resolution (versus the 50 cm also available) will allow for features such as peat hags and smaller gullies to be detected, thus increasing the level at which we can detect erosional features. NIR Aerial Photography At the start of Phase 1 we proposed to use a secondary study site to evaluate the use of NIR (Near-infrared) aerial photography for mapping peat erosion features. Using NIR aerial photography will allow for the use of the NDVI (Normalized Differential Vegetation Index) within the classification procedure at a much finer spatial level due to the 25 cm photographic resolution. As aerial photographic capture today (in this case the Vexcel camera) and in the future will also include a NIR camera, there exists the potential to be a national NIR data set, something which will be of much use in projects like this. Quickbird and IKONOS After consulting the available coverage for both IKONOS and Quickbird, it was deemed that there was not any suitable imagery available for the study area, and so we do not propose to use any of these as an image source. Imagery that potentially was useful contained cloud or had snow cover present which makes it unsuitable for use within this study.

With consideration of possible Scotland-wide mapping, the image footprint for Quickbird and Ikonos lend themselves to targeting specific areas, but not part of a wide ranging operational process.

Final Report 66


Annex 2 Image Processing Technical Specifications

SPOT

Geometric Correction

• Undertaken within Erdas Imagine 9.3 using the .dimap file. • Correction achieved using the ‘SPOT5 – orbital pushbroom’ geometric model

coupled with the NEXTMap Britain DTM data. • SPOT data then outputted in .tif format and resized to 5 m pixels.

Radiometric Correction

• Undertaken with ENVI 4.5 using the geometrically corrected .tif file exported

from Erdas Imagine. • Gain values for each band acquired from the SPOT header file. • Equation inputted into the ‘band maths’ function for each band separately to

calibrate pixels from digital number to radiance.

L = D/G

L = Radiance

D= Digital number of pixel

G = Gain

• Bands re-stacked using the ‘Layer Stacking’ tool. • Re-stacked image then converted to a .bil format for use with the FLAASH

tool used within the Atmospheric Correction process. • FLAASH requires all pixels to have no ‘no zero pixels’, so using ‘band maths’,

a value of 0.000001 should be added to all pixel across all bands. (Note - 0.000001 is smallest value possible).


• Undertaken using the FLAASH tool within ENVI 4.5. • All input coefficients except ‘average scene elevation’ and ‘zenith angle’ are

located within the header. • Zenith Angle = 180 – Viewing Angle • Average scene elevation calculated by generating a shape file covering the

scene and using the ‘zonal statistics’ tool in conjunction with the NEXTMap DTM within Erdas Imagine, a mean elevation value is calculated.

• Once all required coefficients have been inputted, the FLAASH tool is applied to the SPOT image.

• A value of 0.000001 is subtracted from all exterior pixels to return them to 0 by creating a mask of all pixels less than 0.000002 and applying it to the atmospherically corrected image.

• Image then outputted in a .tif format. ASTER


• Undertaken within ENVI 4.5 using the .hdf file.

Final Report 67


• VNIR 3B, SWIR bands 5-9 and all TIR bands are removed as they were not no seen to contain relevant information.

• RPC projection emulation activated within the ‘edit ENVI header function’ and set to ‘United Kingdom’.

• Orthorectify all VNIR bands within one process and then SWIR band4 within a singular process.

• 1 GCP (Ground control point) used to othorectify the VNIR bands. • Point located on the image and on a registered image (NEXTMap ORI data

used within this instance), with an elevation value obtained from the

NEXTMap dtm. • Values inputted into the ‘Orthorectification with Ground Control’ tool within

ENVI, setting the projection to United Kingdom. • Image orthorectified using ‘nearest neighbour’ sampling technique and

outputted in ENVI format with a pixel size of 5 m. • Process repeated for SWIR band 4 but using 2 GCP’s due to the coarseness

of the SWIR resolution in comparison to the VNIR (15 m ~ 30 m) • VNIR bands and SWIR band 4 then stacked using the ‘Layer stacking’

function within ENVI to create a new ASTER file. • No requirement to calculate radiance as ASTER data is already converted to

Radiance. • File converted to .bil format for use within FLAASH and a value of 0.000001

added to all pixels across all bands using the ‘band math’ tool. Atmospheric Correction

• Undertaken using the FLAASH tool within ENVI 4.5. • All input coefficients except ‘average scene elevation’ are located within the

header. • Average scene elevation calculated by generating a shape file covering the

scene and using the ‘zonal statistics’ tool in conjunction with the NEXTMap DTM within Erdas Imagine, a mean elevation value is calculated.

• Once all required coefficients have been inputted the FLAASH tool is applied to the ASTER image.

• A value of 0.000001 is subtracted from all exterior pixels to return them to 0 by creating a mask of all pixels less than 0.000002 and applying it to the atmospherically corrected image.

• Outputted image converted from ENVI format to .tif. IRS

Image generation

• Image contained within separate GeoTiff bands. • Bands imported into ENVI and stacked together using the ‘layer stack’ tool. • Converted .bil format for use within FLAASH and a value of 0.000001 added

to all pixels across all bands.


• Data already calibrated to radiance due to quarter scene purchase. (Whole

scenes would require calibrating to radiance) • Undertaken using FLAASH within ENVI 4.5. • All input coefficients except ‘average scene elevation’ are located within the

header.

Final Report 68


• Average scene elevation calculated by generating a shape file covering the scene and using the ‘zonal statistics’ tool in conjunction with the NEXTMap DTM within Erdas Imagine, a mean elevation value is calculated.

• Once all required coefficients have been inputted the FLAASH tool is applied to the IRS image.


• Undertaken within Erdas Imagine 9.3 using the atmospherically corrected file

outputted from ENVI. • Correction achieved using a 3rd order ‘polynomial’ geometric model. • GCP’s collected on IRS image and from the corresponding point on a

registered image. (NEXTMap ORI and the register SPOT imagery used from this purpose)

• GCPS values collected from the registered imagery inputted as ‘reference data’ within the Erdas Imagine geometric model.

• Projection set to ‘British Grid’. • IRS data then outputted into a .tif format and resized to 5 m pixels.

Creation of Endmember Fractions

• Photosynthetic, Non-Photosynthetic and Shade endmember fraction images

generated for SPOT, ASTER and IRS images. • Undertaken within ENVI 4.5. • Chosen image is opened and a 2D scatter plot created using the NIR and

Red bands (Fig 1)

Figure 1 - 2D Scatter Plot

• The furthest pixels are selected from each corner of triangle and a ‘region of

interest’ (ROI) generated. • Using the NIR and Red bands following corners represent the following

fractions.

Bottom Right – Photosynthetic

Final Report 69


Bottom Left – Shade Top – Non-Photosynthetic

• Using the ‘Linear Spectral Unmapping’ tool, the ROI’s are imported and a

scale of between 0-1 set for the output image. • Process is then applied to the imagery and exported as a .tif file.

Creation of Vegetation Indices

Various indices generated within Definiens Developer and as separate

images for use within the segmentation stage.

Vegetation Indices image generation

• Created using the ‘band maths’ function within ENVI 4.5. • To create a Red NDVI image layer from the SPOT data, the following

equation was inputted using the appropriate bands within ‘band maths’ and a new layer generated.

(SPOT NIR – SPOT Red) / (SPOT NIR + SPOT RED)

• Other indices layers such as a Green NDVI were also create using the same

process as above, but using the correct equations. • Indices created in Definiens Developer are developed and applied to the

segmented images.

Final Report 70


Annex 3 Classification Class Descriptions

Classification Class Description Erosion Risk • Areas no smaller in size than 5 m2 within the EO

Bare Peat data and of no smaller than 3 m2 within the aerial Very High photography.

• Areas of bare peat of less than 25% vegetated. • Eroding peat features • Obtained through the use of the EO data. present which are exhibiting

active erosion processes.

• Clearly defined channels within the areas of Peat Gully blanket bog with a width of between 0.5 – 4 m and

a max depth of 3 m. • Commonly found within a network structure with

the density of gullies dictated by the slope angle and the duration of the gully.

• Obtained through the use of the aerial photography.

• Vegetation within the gully must not exceed 30% with bare peat clearly visible within the bottom and sides if viewable.

• Where gullies are vegetated, bare peat must be present on the side.

Final Report 71


• Bare substrate material found within heavily Peat Substrate eroding blanket bog and bear peat areas.

• Consists of rocks and gravel materials. • Areas delineated through the use of the aerial

photography and obtained through using thresholds within Definiens Developer.

• Tussocks of Eriophorum vaginatum / Trichophorum Ev Tussocks cespitosum within areas of heavy eroding blanket

bog and bare peat. • No larger than 2 m2 in size.

Final Report 72


• Areas of blanket bog which are seen to be in a Blanket Bog – Eroding state of erosion due to the presence of over 20% of High

the land cover having evidence of erosion. • Large scale evidence of peat gullying and areas of • Areas whereby actively

bare peat. eroding peatland features • Comprised of sphagnum-rich vegetation with a are present, but which

hummock and hollow structure. Other species contain vegetated areas present in abundance includes Calluna vulgaris, continuing species such as Trichophorum cespitosum, Eriophorum vaginatum Calluna vulgaris, and Erica tetralix. Trichophorum cespitosum,

Eriophorum vaginatum and Erica tetralix.

• Little change needed to instigate more erosion.

• Areas of blanket bog not delineated through the

Blanket Bog – Residual use of the EO data but present within the aerial photography.

• Larger than 2 m2 in size. • Species present include Eriophorum angustifolium

and Molinia caerulea.

Final Report 73


Areas of blanket bog which are seen to be in a state of stability due to presence of less than 20% of the land cover having evidence of erosion. Little evidence of peat gullying or areas of bare peat. Gullies or bare peat areas that are present are in a process of being re-vegetated. Within the study area stable bog comprised Eriophorum vaginatum / Calluna vulgaris bog, Myrica gala bog, and Eriophorum angustifolium / Molinia caerulea bog. The Eriophorum angustifolium and Molinia caerulea species were commonly seen on areas that were re-vegetating or accumulating peat.

Flush areas of Eriophorum vaginatum and Molinia caerulea.

Form on gently sloping areas ground and are often linear or triangular in form and may include watercourses. Typically include Sphagnum and/ or other bryophytes along with other sedges and Juncus species.

74

•

Blanket Bog – Stable

•

•

•

• Flush

•

•

Final Report

Moderate

• Areas which exhibit fewer

areas of active erosion, but which exist on peat soils.

• Acid grassland areas located near to rivers on areas where peat depths to the substrate are low and which are subjected to grazing.

• Wet heath areas are known to be a delicate community which exist on shallower

peat areas which maybe prone to erosion.


• Heath areas on steep slopes which contain various Wet Heath other heathland species such as Calluna vulgaris,

Molinia caerulea and Eriophorum vaginatum that exist within wet environments.

• Delineated through the use of thresholds through the use of the EO data.

• Bog Myrtle dominated environments within gently Bog Myrtle sloping areas up to an altitude of 450 m amsl.

Final Report 75


• Areas of acid grassland (U4 Festuca Agrostis Acid Grassland grassland, around stream edges) which are larger

than 10 m2 in size. • Delineated using of thresholds developed through

the use of the EO data.

• Grassland areas found above 775 m in elevation. Montane Grassland • Delineated through the use of thresholds through Low

the use of the EO data. • Areas where vegetation

cover is far more dense and

which exhibit very few areas of active erosion.

• Calluna Vulgaris areas are densely covered with vegetation, but should management practices change, these areas may begin to grade towards becoming moderately at

Final Report 76


• Areas of heath intersected with areas of exposed risk. Rocky Heath bedrock.

• Species include Rhacomitrium spp, Vaccinium myrtillus and Calluna vulgaris in a dwarf form.

• Predominantly found on the steeper slopes and on ridges above blanket bog areas.

• Areas of heath on steep slopes which contain Calluna Heath areas of dense stands of older Calluna vulgaris

bushes. • Delineated through the use of thresholds through

the use of the EO data.

Final Report 77


• Bracken (Pteridium) species delineated through the Bracken use of thresholds using the EO data. Non Peatland Area

• Areas which are not located

on peatland soils or which are not classed as being a peatland feature.

• Water Features Present within Ordinance Survey Water MasterMap data. This includes all rivers, streams,

ponds and lakes.

Final Report 78


• Urban Features Present within Ordnance Survey Man-Made Features MasterMap data. This includes all roads and

tracks. • Areas related to the Glendoe Hydro-scheme

development.

• Woodland present within the National Inventory of Woodland Woodland and Trees (NIWT) for Scotland.

Final Report 79


• All snow present with the EO and aerial Snow photography derived through the use of thresholds

within Definiens Developer.

• Areas of bare rock present within the EO data Rock delineated through the use of thresholds within

Definiens Developer.

Final Report 80


Areas of shadow present within the EO data where in-sufficient spectral data was available to obtain accurate classification derived through the use of thresholds within Definiens Developer.

Clouds present within the EO data derived through the use of thresholds within Definiens Developer.

81

•

Shadow

•

Cloud/ Cloud Shadow Final Report


Annex 4 EO Classification Using Traditional Methods (a) Aerial Photography

(b) Aerial Photography – Maximum Likelihood Classification

Legend

Water

Rock

Modified Bog

Heath

Other Vegetation

Peat

Final Report 82


(c) SPOT Image

(d) SPOT – Maximum

Likelihood supervised classification

Legend

Manmade

Water

Cloud Shadow

Blanket Bog

SIAC's Fields

Other Vegetation

Snow

Modified Bog

Rock

Final Report 83


Annex 5 Classification Areas

Erosion Risk Level Total Area (sq km) EO Level Total Area (sq km) AP Level Total Area (sq km)

Very High 25.35 Bare Peat 2.09 Bare Peat 6.95

High 37.37 Man-made 1.43 Man-made 1.43

Moderate 166.21 Wet Heath 18.20 Wet Heath 18.20

Low 64.97 NIWT Woodland 7.06 NIWT Woodland 7.06

Non Peatland Class 25.30 Bracken 0.86 Bracken 0.86

(Total) 319.21 SIAC Field 0.12 SIAC Field 0.12

Calluna Heath 10.36 Calluna Heath 10.36

Rocky Heath 46.19 Rocky Heath 46.19

Montane Grassland 8.42 Montane Grassland 8.42

Acid Grassland 84.02 Peat Gully 14.25

Shadow 3.17 Acid Grassland 84.02

Blanket Bog - Eroding 60.63 Shadow 3.17

Bog Myrtle 6.03 Blanket Bog - Eroding 35.71

Rock 1.60 Bog Myrtle 6.03

Blanket Bog - Stable 50.54 Peat Substrate 3.35

Snow 1.98 Rock 1.60

Cloud/ Cloud Shadow 4.86 Blanket Bog - Stable 50.54

Water 4.22 Snow 1.98

Flush 7.42 Blanket Bog - Residual 1.72

(Total) 319.21 Ev Tussocks 0.76

Cloud/ Cloud Shadow 4.86

Water 4.22

Flush 7.42

(Total) 319.21

Final Report 84


Annex 6 Project Deliverables The project deliverables, in addition to the project report, are a full set of data;

• SPOT5, IRS and Aster satellite imagery (orthorectified and corrected to reflectance) – TIF format.

• Endmember images from each of the satellite imagery – TIF format.

• DTM, slope and aspect models for the study area – TIF format.

• In-situ data collected during the field work, including points and photos – ESRI Shp and Jpeg.

• Classification maps for EO, Air Photo, Peatland and Peat Erosion Risk levels – TIF and ESRI Shp.

• Accuracy tables – Microsoft Xls.

Final Report 85

Assessing the Extent and Severity of Erosion on the Upland ...€¦ · Final Report 6 . Assessing...

Documents

Transcript of Assessing the Extent and Severity of Erosion on the Upland ...€¦ · Final Report 6 . Assessing...