int.j. remote sensing, 2002, vol. 23, no. 21,4551– 4570 TheCircumpolar Arctic Vegetation Map: AVHRR-derived base maps, environmentalcontrols, and integrated mapping procedures D.A.W ALKER,W. A.GOULD, H.A.MAIERand M.K.RAYNOLDS Instituteof Arctic Biology, University of Alaska Fairbanks, F airbanks, AK99775–7000, USA; e-mail: [email protected] Abstract. Anewfalse-colour-infrared image derived from biweekly 1993 and 1995Advanced V eryHigh Resolution Radiometer (A VHRR) dataprovides a snow-freeand cloud-free base image for the interpretation of vegetation as part ofa 1:7.5M-scale Circumpolar Arctic V egetationMap (CA VM).A maximum- NDVI( NormalizedDi VerenceV egetationIndex) image prepared from the same dataprovides a circumpolarview of vegetation green-biomass density across the Arctic.This paper describes the remote sensing products, the environmental factorsthat control the principal vegetation patterns at this small scale, and the integratedgeographic information-system (GIS) methods used in making the CAVM. 1.Introduction Remote sensing products from Earth-orbiting satellites provide the imagebase to make the rst detailed vegetationmap of anentire globalbiome, the arctic tundra. Anewvegetation map of the Arctic is needed for numerous international e Vorts, including global-changeand conservation studies, land-use planning, large-scale resource development, and education. The Circumpolar Arctic VegetationMap (CAVM)willbe based on current knowledge of arctic plant communities and their environmental controls. During the past sixyears, the CAVMparticipants have dened the project organization and methods in aseries of workshops (Walker 1995, D. A. Walker et al. 1995,W alker and Markon 1996,W alker and Lillie1997, Markon and Walker 1999,Walker 2000,Raynolds and Markon 2002).Sixgroups of collabor- ators arenow working on regionalmaps of Alaska,Canada, Greenland, Iceland, Svalbard,and Russia.This paper presents two newAVHRR-derived images of the circumpolar Arctic, the keyenvironmental variablesused for interpreting the vegetation,and the GIS framework for the mapping methods. 2.Remote-sensing products 2.1. AVHRR-derivedbase map Ahigh-resolution falsecolour-infrared (CIR)image of the circumpolar region representing the vegetationat maximum greenness is anessential element of the Thispaper was presented at the6th Circumpolar Symposium on Remote Sensing of Polar Environmentsheld in Y ellowknife,Northwest T erritories,Canada, from 12– 14 June 2000. InternationalJournal of Remote Sensing ISSN0143-1161 print /ISSN1366-5901 online ©2002Taylor & FrancisLtd http://www.tandf.co.uk /journals DOI:10.1080 /01431160110113854 4552 D. A. Walker et al. mapping method (gure 1). This imageis derived from AdvancedVery-High Resolution Radiometer (AVHRR)data that wereobtained from the USGeological Survey (USGS) AlaskaGeographical Science O Yce. It is composed of 1km ×1 km picture elements (pixels).The imageis acomposite of pixels,where each pixelwas selected byusing the date with highest Normalized Di Verence VegetationIndex (NDVI)for that pixelamong biweeklyimages taken during the periods from 1April to 31October in 1993and 1995.These periods cover the vegetationgreen-up-to- senescence period during two relativelywarm years when summer-snow cover was atminimum in the Arctic. This permitted the construction of animage with minimum snow and cloud cover. The white areasare glaciers that weremasked out using information from the DigitalChart of the World. The ocean icewas masked out using the coastlines from the DigitalChart of the World. The southern border of the CIRimage is clipped to tree line, which wasderived from the best available Figure1. A VHRR-derivedfalse colour-infrared image of thecircumpolar Arctic. The image isderived from the pixels with highest NDVI among biweekly images from 1993 and 1995.The southern border of the CIR imageis clipped to treeline, which was derived fromthe best available vegetation maps. Sixth CircumpolarSymposium on Remote Sensing of Polar Environments 4553 vegetationmaps and the most recent information (Raynoldsand Markon 2002). The imageis used asabase for drawingvegetation map polygons. Most boundaries on the vegetation map correspond to features that canbe seen on the AVHRRimage (see §4). 2.2. Maximum-NDVIimage Animage portraying maximum NDVI (gure 2)is used to delineate areasof high and lowvegetation biomass. This imagewas created from the same dataas the falseCIR image ( gure 1). NDVI=(NIR± IR)/(NIR+IR), where IR is the spectral reectance in the AVHRRnear-infrared channel (0.725–1.1 mm), where light- reectance from the plant canopy is dominant, and R is the reectance in the red channel (0.5–0.68 mm), the portion of the spectrum where chlorophyll absorbs maxim- ally.The NDVI valueswere grouped into eight classes that meaningfullyseparate the vegetationaccording to biomass. Red and orange areasin gure 2areareas of shrubby vegetationwith high biomass, and blue and purple areasare areas with low biomass. Figure2. Maximum NDVI for the circumpolar Arctic derived from the same data as gure5. 4554 D. A. Walker et al. 3.Environmental controls of vegetation at 1:7.5 M scale The approach used for makingthe CAVMis based on manual‘ photo- interpretation’of the AVHRRsatellite image.A map based purely on automated remote sensing procedures could not portray the details of plant communities for the entire circumpolar region because there is largevariability of tundra vegetation with similarspectral properties. Due to the smallscale of the map and lackof previous vegetationmaps for much of the Arctic, vegetationinformation must be inferred from expert knowledge of the plant communities in relation to principal landforms and other terrain features that arevisible on small-scalesatellite imagery. In the Arctic, vegetationof agivenlandscape canbe predicted on the basis of summer temperature regime (bioclimatic subzones), availableplants in the regional ora (oristic sectors), soil chemistry, and prevailingdrainage conditions (Walker 2000). 3.1. Bioclimaticzonation Afundamental problem for the CAVMis how to characterize the transitions in vegetationthat occur across the roughly 10 °Cmean Julytemperature gradient from the tree line to the coldest parts of the Arctic. The important role of summer temperature has been noted with respect to awidevariety of ecologicalphenomena, including phenology (Sorensen 1941),species diversity (Young 1971,Rannie 1986 ), plant community composition (Matveyeva1998 ),biomass (Blissand Matveyeva 1992,Bazilevich et al. 1997),and invertebrate and vertebrate diversity (Chernov and Matveyeva1997 ).Various authors, working with di Verent geobotanical traditions, havedivided the Arctic into bioclimatic regions using avarietyof terminologies (table 1).The origins of these di Verent terms and approaches havebeen reviewed for the PanarcticFlora ( PAF)initiative( Elvebakk1999 ).The PAFand CAVMhave accepted ave-subzone version of the Russian zonal approach (gure 3).The subzone boundaries aresomewhat modied from the phytogeographic subzones of Yurtsev (1994)based on recent information from avarietyof sources (Raynoldsand Markon 2002).Abrief description of each subzone follows: 3.1.1. SubzoneA: Herb subzone Subzone Aincludes mostly fog-shrouded islands within the permanent arctic ice pack where Julymean temperatures areless than about 2–3 °C,such asEllefRingnes, Amund Ringnes,King Christian, northern Prince Patrick and nearby islands in the northwest corner of the CanadianArchipelago. It also includes the coastalfringe of northernmost Greenland and northern Ellesmere and northern AxelHeiberg islands, the northeastern portion of Svalbard,Franz Josef Land,Severnaya Zemlya, the northern tips of the TaimyrPeninsula, and northern tip of NovayaZemlya. The summer temperatures inthese areasare near freezing allsummer due to acombina- tion of nearlycontinuous cloud and fogcover, which limits solar radiation, and the close proximity to the ice-covered ocean (Bay1997, Razzhivin 1999 ).More contin- ental inland areasof the largerislands areoften considerably warmer.Permanent ice covers largeareas of the land. Major parts of the nonglacialland surfaces are largelybarren, often with <5%cover of vascularplants, however, meadow-like plant communities arenot uncommon on mesic ne-grained soils, where there is suYcient moisture provided bythe cold humid oceanic climate. Woody plants areabsent on zonal sites. Zonalsites are ator gentlysloping, moderately drained sites with ne-grained soils that arenot inuenced byextremes Sixth CircumpolarSymposium on Remote Sensing of Polar Environments 4555 of soil moisture, snow, soil chemistry, or disturbance and which fullyexpress the inuence of the prevailingregional climate. Lichens, bryophytes, cyanobacteria,and scattered forbs (e.g. Papaver, Draba, Saxifraga and Stellaria)arethe dominant plants. Manyof the forbs, lichens and mosses havea compact cushion growth form. In midsummer, the arctic poppy, Papaverradicatum ,isthe most conspicuous plant over largeportions of this subzone. Other important low-growingcushion-forb genera include Minuartia and Cerastium.Soil lichens, mosses, and liverworts cancover a high percentage of the surface, particularlyin more maritime areassuch asNovaya Zemlya(Alexandrova 1980 ).Rushes ( L uzula and Juncus)and grasses( Alopecurus, Puccinellia , Phippsia, and Dupontia)arethe main graminoid groups. Sedges (Cyperaceae)are rare, and wetlands lackorganic peat layers.There is little contrast in the composition of vegetationon mesic sites, streamside sites, and snowbeds. The