G E O M A T I C A

HISTORICAL MAP COLLECTIONS ONGEOSPATIAL WEB

Sabarish Senthilnathan Muthu1, Eleni Gkadolou2, Emmanuel Stefanakis1,2

1Department of Geodesy and Geomatics Engineering, University of New NrunswickFrederiction, New Brunswick

2Department of Geography, Harokopio University of Athens, Kallithea, Athens, Greece

The Provincial Archives of New Brunswick (PANB) and the University of New Brunswick Archives(UNBA) maintain a rich collection of historical maps and plans of the City of Fredericton dating back to the17th Century. This content is in the form of microfiches and paper maps, which are difficult to be accessedand used by researchers and educators. Recently, a series of microfiches and maps have been scanned andare already available in digital form. Going a step further, the advancements in geospatial web technologyhave been applied to make this digital content a valuable resource of historical information for the end-users. A prototype system has been developed and tested for a set of representative historical maps andplans. The system provides the following functionality: (a) geo-referenced scanned maps can be archivedinto a geospatial catalog; (b) appropriate metadata can be attached to them; (c) the maps can be superim-posed and visualized on top of recent base maps (e.g., Google Maps, Bing Maps, Open Street Map); and (d)the catalog content can be searched and disseminated on the web using geospatial standards. End-usershave been exposed to the prototype and evaluated its functionality. Their comments highlight the potentialof the prototype in the study of cultural heritage and historical documentation.

1. Introduction

Historical map collections form one of themost significant resources of a community’s cultur-al heritage. Old maps usually have great aestheticand scientific value. While they reveal older meth-ods for representing space, they also display theartistic style of the time of their creation.Furthermore, historical maps are a valuable sourcefor retrieving historical data and documenting thespatio-temporal evolution of geographic entitiesdepicted therein. One can discover, for example,where a—now dried up—lake used to be or theurban sprawl of a town. Occasionally, the informa-tion that historical maps carry is not documentedelsewhere and can be valuable in a plethora ofapplications of historical, cultural and educational

interest. Historical maps may be a vital documentof a community’s history as they were produced toestablish it, e.g. maps of new borders, plans of mil-itary missions, etc.

In order to promote collections of historicalmaps to a wider public, international initiatives havealready emerged. For instance, the InternationalCartographic Association (ICA) has formed aWorking Group on Digital Technologies inCartographic Heritage. The scope of this group is toinvestigate new methods for advanced analysis andreuse of maps and their content through the Internet.

Typically, historical maps can be found on theweb either as objects of cultural heritage manage-

GEOMATICA Vol. 67, No. 3, 2013 pp. 279 to 290

Eleni Gkadolouelenigadolou@

gmail.com

EmmanuelStefanakis

estef@unb.ca

SabarishSenthilnathan

Muthusabarish.sm@

gmail.com

G E O M A T I C A

ment systems [e.g. Europeana 2013] or as base mapsfor visualizing historical and other information (e.g.in projects like Philaplace [2013], where audiovisu-al material and narrations have been correlated withgeographic—historic or not—regions on historicalmaps). The publication of maps on the Internet ispossible by developing either digital catalogs, usual-ly according to bibliographic metadata standards(e.g. MARC21) [Montaner 2009] or portals (e.g.DIGMAP [2013]) by reusing the metadata that col-lections’ providers adopt [Borbinha et al. 2009;Simon et al. 2010].

A remarkable initiative in publishing historicalmap collections on the web is the David RumseyMap Collection [2013]. It contains more than150,000 maps mainly of North and South America(from the 18th and 19th century), but also globes,atlases etc. From these, 42,000 items have beendigitized and georeferenced and can be viewedusing a free browser (called Luna). Users can alsovisualize these items as layers in Google Earth andGoogle Maps. Searching and retrieving maps isconducted either in Luna browser (using criteriafrom the metadata catalogue, e.g., scale, authoretc.) or in MapRank [2013]. MapRank is a sophis-ticated tool that enables geographic searching bymap location and coverage in Google Maps.

What is missing from the aforementionedpractices is an explicit documentation of the mapsbased on their content. The geographic informationa map carries, if appropriately modeled; can beused to support a content-based searching of oldmaps. In some cases, this is a prerequisite, e.g.,when an old place name is the only information auser has to conduct a search for a map. Catalogingmaps and their content using spatial metadata itemsis a new practice yet to be studied. Recent advanceson the Semantic Web introduce new needs in themanagement and documentation of historical maps.

Having said that, the semantic description of thecontext of old maps becomes a necessity andallows for new trends in the accessibility andusability of historical map collections.

This study is aligned with the new trends in thedissemination of historical map collections. Byapplying advanced methods and tools available ongeospatial web, a prototype web-based system hasbeen implemented to disseminate a series of histor-ical maps of the city of Fredericton and make themaccessible to the end-users in an innovative way.Firstly, users can search for appropriate historicalmaps based on metadata items. Secondly, they areable to superimpose and visualize them on top ofrecent base maps, provided by the earth browsers(such as Google Maps or Open Street maps).Educators and historians can easily access historicalmaps that meet specific criteria, discover the wealthhidden in their content, and highlight evolution pat-terns between historical periods and differencesamong the past and the present.

This paper presents the components and func-tionality of the prototype system. Specifically,Section 2 gives an overview of the system architec-ture by introducing its components and their interac-tion. Section 3 provides an insight of the methodol-ogy applied to implement the system and highlightsthe system functionality. Section 4 summarizes theend-user comments and findings during the systemevaluation phase. Finally, Section 5 concludes thediscussion by introducing the short and long-termobjectives of the research team.

2. System ArchitectureThe proposed system is an open-source Web-

based system as shown in Figure 1. The systemconsists of the following four components and fol-lows client-server architecture:

1. Client: A typical Web browser (such as Chrome,Mozilla or IE) plays the role of the client in thisclient-server architecture. It facilitates display-ing the results fetched from the server to end-users. It is not anticipated that the clientmachine has any special software installation(apart from the Java Runtime).

2. Web server: The Web server facilitates servingthe users’ requests through HTTP protocol andproduces the response according to the request.The Web server houses all Web pages of theapplication and interacts with the GIS Server.Apache [Fielding and Kaiser 1997], a publicdomain, open-source Web server is used in thisapplication.280

Figure 1: System architecture.

G E O M A T I C A

3. GIS server: The GIS server allows serving thegeospatial content on the Web. GeoServer[OpenPlans 2013], an open-source server builton Java, is used as the GIS server for the Web-based system. It interacts with the Web clientand facilitates publishing the geospatial data.

4. Geospatial catalog server: The geospatial cata-log server is a single repository for storing,searching and retrieving the metadata elementsof all datasets. We use GeoNetwork [OpenSource Geospatial Foundation 2012], an open-source catalog application implemented in Javais used to manage geographic referenced data.

5. Web mapping framework: The Web mappingframework used is a JavaScript API;OpenLayers [OpenLayers 2012] for buildingrich Web based mapping applications withoutserver-side dependencies.

3. Methodology

The project was conceived in two phases asshown in Figure 2. Namely,

Phase 1: Creation of appropriate metadatafor the maps

In this phase, the main objective was the docu-mentation of the historical maps by enriching themaps with metadata; the maps were georeferencedand appropriate metadata were attached to each ofthe historical maps using the ISO 19115:2003 stan-dard. Additionally, tools for searching the metadataelements were also integrated.

Phase 2: Developing the Web application tovisualize the historical maps

In this phase, the Web application was devel-oped using HTML, CSS and JavaScript, along withOpenLayers for visualizing the historical maps ontop of modern Web mapping applications such asGoogle Maps, Bing Maps, OpenStreet Maps, etc.

3.1 Georeferencing and MosaicingThe first step in the first phase of the project

was the georeferencing of the scanned maps. Thesoftware application which was used for georefer-encing is Quantum GIS [QGIS Development Team2012], an open-source geographic information sys-tem. Quantum GIS is a user friendly Open SourceGeographic Information System (GIS) licensedunder the GNU General Public License havingcapabilities for storing, managing, editing and per-

forming analysis of geographic data. By virtue ofgeoreferencing, the image coordinates are convertedto real world coordinates.

The scanned maps had no projection system andthese maps were georeferenced to WGS84 referencesystem. Clearly identifiable point features such asroad intersections were identified as ground controlpoints and these points were distributed homoge-neously across the map. The coordinates for theseground control points were obtained from GoogleEarth, which is also in WGS84 reference system.The scanned maps were georeferenced using secondorder polynomial transformation and bilinear inter-polation resampling. The residual errors obtainedwere reasonable enough for the maps to be displayedon top of modern web mapping applications such asGoogle maps, Bing maps, etc. It may be noted that itis almost impossible to perfectly align an old map tomodern coordinate systems because mapping meth-ods before the age of aerial photography often onlyvery imprecisely represented scale, angle, distance,and direction [Knowles and Hillier 2008].

Once the georeferencing is completed, mosaic-ing was done as each map was acquired from sev-eral microfiches, thereby creating a single compos-ite. Quantum GIS was again used in mosaicing.The mosaiced images consisted of a section ofblack regions on all four sides, denoting that thearea is occupied by ‘null’ values not containing anynumerical values, which is relevant to the scannedimages. These no data values were eliminated usingGeospatial Data Abstraction Library, which is atranslator library for raster geospatial data formats[GDAL 2009]. By command line utilities, it facili-tates in transformation and processing of rasterdata. GDAL translate tool was used to accomplishthis task. Figure 3 shows the result obtained by thisprocess. 281

Figure 2: Flow diagram.

G E O M A T I C A

3.2 Creating the Geo-DatabaseOnce the scanned maps were georeferenced,

the second step is to add and publish these maps aslayers in GeoServer. GeoServer is an open sourcesoftware server written in Java that allows users toshare and edit geospatial data. Designed for inter-operability, it publishes data from any major spatial

data source using open standards [OpenPlans2013]. GeoServer allows to implement OpenGeospatial Consortium (OGC) defined Web MapService (WMS) standard, which facilitates inrequesting georeferenced maps served in geospatialdatabases. The response got back is an image inJPEG, PNG, etc. file formats, which can be visual-ized on the browsers.

The maps added in GeoServer are in WGS84reference (EPSG:4326) and these need to be trans-formed to Spherical Mercator projection(EPSG:900913). In order to properly overlay dataon top of the maps provided by the commercial APIproviders, it is necessary to use this projection (seePhase 2). This applies primarily to displaying rastertiles over the commercial API layers—such asTMS, WMS, or other similar tiles [OpenLayers2008]. Figure 4 shows the images which wereadded as layers in GeoServer.

3.3 Creating Metadata ElementsThe final step in the first phase of the project

was to create the geographic catalog application. Tofacilitate access to all the historical maps online, todocument and describe the maps and to searchcapabilities on the georeferenced datasets, a geo-graphic catalog server is used. The application usedto create the geospatial catalog is GeoNetwork, acatalog application to manage spatially referencedresources. It provides powerful metadata editingand search functions as well as an embedded inter-active web map viewer [Open Source GeospatialFoundation 2012].

This catalog server will be able to disseminatethe map collections and store the metadata elementsfor each of the historical maps. The metadata ele-ments were stored in GeoNetwork according to thespecifications of ISO 19115:2003 standard (Figure5). Only a small subset of the metadata elementsdescribed by ISO was used. Table 1 lists the meta-data elements which were used for each dataset.

Figure 6 shows the metadata in ISO19115:2003 standard of one particular dataset, thetown plots of Fredericton surveyed in 1788.GeoNetwork also provides tools to search the cata-log based on title, abstract, keywords, the geo-graphic extent and temporal extent. Users withoutaccess to the server would only browse and searchthe catalog. Users with access permissions wouldbe able to add and edit the metadata of the datasets.

3.4 Web applicationThe second phase of the project was the devel-

opment of the Web application using282

Figure 4: The Screenshot of the geo-database created.

Figure 3: Result obtained by removing the black regions on the mosaicedimages.

Figure 5: The home page of the catalog server. Available at:http://gaia.gge.unb.ca:8080/geonetwork/srv/en/main.home

G E O M A T I C A

HTML/JavaScript. OpenLayers [OpenLayers2012], a JavaScript framework for creating richWeb mapping applications is used to call the WMSlayers from the client-side without any server-sidedependencies. From this request an image is gener-ated from GeoServer, which is shown over contem-porary mapping applications such as Google Maps,Bing Maps, etc.

jQuery [The jQuery Foundation 2013] is usedto develop the slider function, which facilitates theuser to control the transparency of the WMS layers.jQuery is a JavaScript allowing users to simplifythe development of client-side scripting. It offersseveral other features such as event handling,effects and animations, User Interface elements,etc. By controlling transparency, the users couldidentify the temporal changes in the landscape.

The end-users would be able to access this visu-alization tool through the online resource metadataelement in GeoNetwork. Although GeoNetworkprovides an interactive map visualizer, it is not capa-ble of visualizing the datasets on modern mappingapplications such as Google and Bing Maps and alsooffers no capability to control the transparency of thelayers. These features are included in the Web appli-cation developed. Figure 7 presents the screenshot ofthe Web prototype.

4. System Evaluation

The prototype system has been developed inclose collaboration with the end-users so that it isfunctional and valuable to them. The research teamhas got advice from historians and local educatorsin various phases of the project.

Firstly, a historian along with the map archivistin the Provincial Archives of New Brunswick(PANB) have delved into the archive collections todiscover and scan a representative set of historicalmaps for the city of Fredericton (seeAcknowledgments). These maps were available inpaper form and microfiches. These media were digi-tized and further processed, as of Figure 2. Table 2summarizes the historical maps that were accommo-dated in the prototype. These maps date from 1788 to1906 and constitute a valuable resource for the evo-lution of the city of Fredericton in the 19th century.

Subsequently, the end-users helped in therecognition and population of the metadata items(Table 1) for each map in the geospatial catalog.These metadata items are very useful while search-ing the catalog based on various criteria.

The catalog and the visualizer were then intro-duced to two university professors and two

researchers with expertise in either history or edu-cation of social sciences (see Acknowledgments).In the sequel, a class of 30 students, pursuing aBachelor in Education at Queens University, werealso exposed to the system. Although, the sample issmall, the comments collected from them werevery encouraging.

The possibility of (a) switching on and off thehistorical maps, (b) having them overlaid either ontop of modern base maps (Google maps and Bing 283

Figure 6: Metadata elements for a historical dataset in ISO19115:2003 standard.

G E O M A T I C A

maps) or between them, and (c) changing theiropacity to make them transparent, opened newways to both their research and educational prac-tices. End-users have spent many hours in compar-ing each individual historical map against the mod-ern base maps, street by street, building block bybuilding block, house by house. They noticedchanges that they were not aware of, changes thatwere not documented elsewhere.

Figure 8 summarizes some representativechanges that can be readily highlighted through thevisualizer. Figure 9 shows the integrated interface

to both the geospatial catalog and the visualizer inthe most recent version of the system.

5. Conclusion

Future research of the research team will befocusing on the historical maps’ semantic documen-tation. This includes the semantic definition of: (a)the attributes (both geometric and thematic) of thehistorical map (e.g. creator, technique, scale), and (b)the geographic entities depicted in a historical mapas well as their spatio-temporal interrelations.

This is necessary as it enhances the mechanismsfor retrieving and reusing historical maps and theircontent [Gkadolou et al. 2012]. Specifically,

• the semantic search of geospatial data on theweb using spatial and conceptual criteria (forexample based on toponymy or geometry) ispossible

• the correlation between an entity’s spatiotem-poral changes and their causes is explicitlyattributed

• the development of interoperability mecha-nisms in several web—spatial or not—applica-

284

Figure 7: A screenshot of the visualizer. Map of 1832 on top ofBing maps.

Metadata element DescriptionTitle Name of the resourceDate Date at which the resource was createdPresentation form Map represented in raster or vector formAbstract A summary of the datasetPurpose The reason for which the dataset was createdStatus Whether the update of the dataset is ongoing or completedPoint of contact Describes the name, organization name, position and role

individual responsible for the creation of the datasetKeyword The theme and place keyword used to describe the resourceMaintenance and update frequency The frequency with which the dataset is updatedDenominator The scale of the resourceExtent The temporal extent and geographic bounds of the resourceDistribution information The format and version used to distribute the datasetReference system information The spatial and temporal reference system used in the datasetOnline resource Information about the online resources where dataset could be

obtained and visualizedMetadata File identifier Unique identifier for the metadata fileMetadata language Language used to describe the metadataMetadata character set The character coding standing usedMetadata date stamp Data at which the metadata was createdMetadata standard name The standard used is ISO 19115:2003/19139Metadata standard version The version used is 1.0Metadata contact details Individual name, position, role and e-mail address of the

person responsible for creating the metadata

Table 1: The metadata elements with which the dataset is described.

G E O M A T I C A

285

Title Description

Town Plots – 1832 Town plots drawn on February 22nd 1832 by T. Kelcher

Town Plots - Redrawn 1876 Map of the town plots of Fredericton redrawn on 1876.The map was surveyed on April 12th 1832 by T. Kelcher

Map of Fredericton - 1882 Map of the city of Fredericton, NB drawn byAlexander M. Hubly on 1882

Town plots flat, town plat - 1832 Town plots flat, town plat of T. Keleher, surveyed in 1832

Route sketch - 1799 Sketch of the route Fort Cumberland to Fredericton fromD. Campbell Mil, 1799

Public landings - 1864 Public landings of the city of Fredericton, surveyed in 1864

Town plat of Fredericton - 1848 Skeleton plan of the town plat of the city of Fredericton,surveyed in 1848 by William Percival

Town plots – 1788 Town plots of Fredericton, surveyed in 1788

River bank - Fredericton Map showing both banks of the River at Fredericton

Town plan – 1906 Town plan drawn by Odell, 1906

Table 2: The historical map collection available in the prototype.

(a) The riverbank in 1832 and today (b) The cemetery used to be a set of parcelscrossed by a street.

(c) Street name change: from Smyth (1786) toSmythe (today)

(d) Comparing two historical maps. The river-bank in 1832 and 1882.

Figure 8: Changes over time as discovered by superimposing the historical maps on top of modernbase maps.

G E O M A T I C A

tions that use maps (for example toponymydigital catalogs, historical archives, etc.) isfacilitated.

In order to achieve the above scope, cartograph-ic ontologies are being examined closely. Eventhough there are ontologies for defining geographicfeatures or for describing cultural heritage objects,there is not any specially designed for historicalmaps. As a result, a formal vocabulary for historicalmaps will be created, which can be used for theextension of the spatial metadata of ISO 19115:2003standard (used in Geonetwork catalog). The newmetadata items will be accommodated in the catalogto support an enhanced semantic-based searching ofgeospatial content stored in historical maps.

Finally, the extension of the system spatialcoverage to host historical map collections of otherareas in the Province of New Brunswick (GrandLake Meadows, Kouchibouguac Park, Mactaquac,etc.) based on the resources of PANB and UNBA as

well as the enhancement of the system functionalityare currently in progress.

Acknowledgements

Thanks to (a) UNB Work Study program 2012-13 for partially funding this research; (b) SusanMcAdam, historian, for delving into the ProvincialArchives of New Brunswick (PANB) and scanningthe historical maps and plans of the City ofFredericton; (c) Mary-Ellen Badeau, map archivistat PANB, for her valuable contribution in searchingthe archives; (d) Dr. Ted Christou, AssistantProfessor at Queens University, Dr. BonnieHuskins, Assistant Professor at Saint ThomasUniversity, Mrs. Kelly Jean Harris, BA, BEd, andMrs. Koral June Lavorgna, BA, MA, for the dili-gent evaluation of the system; and (e) the Editorand the anonymous reviewers of GeomaticaJournal for their valuable comments.

References

Borbinha J., G. Pedrosa, J. Luzio, H. Manguinhas, and B.Martins B. 2009. The DIGMAP Virtual DigitalLibrary. e-Perimetron, 4(1), 1-8

David Rumsey Map Collection, http://www.davidrum-sey.com.

DIGMAP (Discovering our Past World with DigitizedMaps), http://portal.digmap.eu.

EuropeanaConnect, http://www.europeanaconnect.eu.Fielding, R. T., and G. Kaiser. 1997. The Apache HTTP

Server Project. Internet Computing, IEEE, 1(4), pp.88-90.

GDAL. 2009. Geospatial Data Abstraction Library.[Online], www.gdal.org.

Gkadolou E., E. Tomai, E. Stefanakis, and G. Kritikos.2012. Ontological standardization for historical mapcollections: Studying the greek borderlines of 1881.In Proceedings of the XXII ISPRS Congress,Commission II, ICWG II/IV, August 25-September 1,Melbourne, Australia.

Knowles, A. K., and A. Hillier. 2008. Placing History:How Maps, Spatial Data, and GIS are ChangingHistorical Scholarship. Esri Pr.

MapRank Search, http://www.klokantech.com/mapranksearch.Montaner, C. 2009. Disseminating Digital Cartographic

Heritage: Standards and infrastructures. e-Perimetron,4(1), 53-54.

Open Source Geospatial Foundation. 2012. GeoNetworkUser Manual. [Online]. Available at http://geonetwork-opensource.org/manuals/2.8.0/eng/users/GeoNetworkUserManual.pdf.

OpenLayers. 2008. The OpenLayers LibraryDocumentation. [Online]. Available at

286

Figure 9: The interface to the system. Available at:http://gaia.gge.unb.ca/wsp/maps/index.html

G E O M A T I C A

http://docs.openlayers.org/library/spherical_merca-tor.html.

OpenLayers. 2012. The OpenLayers JavaScript Library.[Online], http://www.openlayers.org.

OpenPlans. 2013. GeoServer User Manual. [Online],http://docs.geoserver.org/stable/en/user/.

PhilaPlace. http://www.philaplace.orgQGIS Development Team, 2012. QGIS User Guide.

[Online], http://docs.qgis.org/1.8/pdf/QGIS-1.8-UserGuide-en.pdf.

Simon R., C. Sadilek, J. Korb, M. Baldauf, and B.Haslhofer. 2010. Tag Clouds and Old Maps:Annotations as Linked Spatiotemporal Data in theCultural Heritage Domain. In: Proceedings of theLinked Spatiotemporal Data Workshop (LSTD 2010),12-23

The jQuery Foundation. 2013. jQuery: The write less, domore, JavaScript library, http://jquery.com.

MS Rec’d 13/06/19Revised MS rec’d 13/08/09

AuthorsSabarish Senthilnathan Muthu is a Master of

Science in Engineering (MScE) candidate in theDepartment of Geodesy and Geomatics

Engineering at University of New Brunswick sinceFall 2012. He completed his Bacherlor’s degree inGeoinformatics Engineering at the College ofEngineering, Guindy, Anna University, India. Hisresearch focuses on Web-GIS, Spatial Databasesand Knowledge Discovery.

Eleni Gkadolou holds a Dipl.Eng. in Surveyingfrom the National Technical University of Athens;and a M.Sc. in Applied Geography from HarokopioUniversity of Athens. Currently, she is a PhD stu-dent in the Department of Geography at HarokopioUniversity of Athens. Her research work lies in thethematic field of Web Cartography with focus onthe usability of historical maps for educational andcultural purposes.

Emmanuel Stefanakis, Ph.D., PEng., is anAssociate Professor in the Department of Geodesyand Geomatics Engineering at University of NewBrunswick. His teaching and research is currentlyfocused on Geospatial Web and KnowledgeDiscovery from Geographic Data. He has over 60articles published in international journals and con-ferences in Geomatics.

287