I I I I ANWR Progress Report FY87 I I I i I ACCURACY ASSESSMENT OF LANDSAT I I LAND COVER MAPS I 1-... _1 OF THE COASTAL PLAIN OF THE NATIONAL WILDLIFE REFUGE, ALASKA, 1987 I

N.A. Felix I M. McWhorter · D.L. Binney I L.M. Koestner II

I Key Words: Landsat land cover map, vegetation, classification, habitat types, airphoto analysis, , Alaska, Arctic National I Wildlife Refuge, Arctic-Beaufort I I I Arctic National Wildlife Refuge u.s. Fish and Wildlife Service I 101 12th Avenue I Fairbanks, Alaska 99701 FWLB 0254 I 1e Manager Approval~~~, Date:~ I I I. I

Accuracy assessment of Landsat-assisted land cover maps of the I coastal plain of the Arctic National Wildlife Refuge, Alaska, 1987 I Nancy A. Felix, Marta McWhorter, Daryl L. Binney, and Lorna M. Koestner. u.s. Fish and Wildlife Service, Arctic National Wildlife Refuge, Fairbanks, Alaska. II Abstract: Accuracy assessments of two versions of Landsat­ assisted land cover maps were conducted on the coastal plain of the Arctic National Wildlife Refuge. Ground determinations and R map classifications were compared for 126 map polygons. Agreement occurred at 53% of the polygons on the 1982 Landsat map and 48% on the 1985 map. The largest source of error (25% on the 1982 map and 27% on the 1985 map) was cutpoints between closely I related land cover classes. Classes are related to each other along gradients of moisture, cover, or total cover. Misclassification errors where the plant community found on the I ground clearly did not belong to the map class occurred at 17% of the 1982 map polygons and 22% of the 1985 polygons. Most commonly the map misclassified braided river floodplain I communities as wet or moist plant communities. Description errors indicating that the land cover class description needed additions or modifications to allow the user to accurately identify the class occurred on 6% of the 1982 polygons and 4% of I the 1985 polygons. Since cutpoint errors between closely related classes accounted for the majority of errors, the Landsat­ assisted maps are useful for showing general distributions of I land cover classes. I I

I ..;:t 1..0 ex:> N N I 1..0 0 0 0 1..0 I 1..0' "('I') I ('I') ARLIS . Alaska Resources · 2 Library & lf'lformauon SerVices I · · Anchw~~.:, .-t.:dska , · · ·. :•'. I I.

I ANWR Progress Report FY87 Assessment of Landsat land cover maps of the coastal plain of the I Arctic National Wildlife Refuge, Alaska, 1987 An accurate map of vegetation types on the coastal plain of the I Arctic National Wildlife Refuge (ANWR) is needed as a basis for wildlife habitat studies. Studies to determine the value of certain portions of the coastal plain to wildlife species are being initiated in order to predict the impacts of further oil I and gas exploration and development. These studies include: - availability of forage in areas used by the Porcupine I caribou herd during calving and post-calving periods - comparisons of habitat characteristics of high and low I muskox use areas - studies of snow geese staging grounds I If further oil and gas exploration and development is allowed on the coastal plain, a vegetation base map would be essential for aiding in site selection for facilities and related activities to I minimize environmental impacts. Three versions of a Landsat-derived land cover map were produced for the coastal plain between 1977 and 1985 (Nadler & LaFerriere I 1977, Walker et al. 1982, Markon 1986). An accuracy assessment of the most recent Landsat map was conducted over the entire refuge in 1985, and analyses of the data are in progress. I Preliminary analyses indicated that there was a large amount of disagreement between the Landsat map and the three botanists who determined vegetation types on the ground. There were also many I disagreements among the botanists indicating that better descriptions of the land cover classes are needed to allow consistent ground identifications. I Determining the accuracy of existing maps of the coastal plain will help determine the need for further habitat mapping. The objectives of this study were to evaluate the accuracy of the I 1982 and 1985 Landsat maps on the coastal plain and develop more detailed vegetation descriptions where possible to aid in I identification of land cover types on the ground. Methods I The study area is an irregularly shaped portion of the coastal plain of ANWR, lying between 142° W and 147° W and north of 69° 34'N, covering approximately 630,000 ha (Fig. 1). It is bordered I by the Brooks Range on the south, the Beaufort sea on the north,

3 ·.; . . _·: ' ·::-·,.' I .,.· '.' I ------ilil Iiiii IIIII 111111 !1!!!1! 1!111!!1 -- -

,. 145°W 143•w 10 5 0 10 20 MILES E'"3 I I KILOMETERS ,\. 5 0 10 20

Pigw:e 1. Map of study sites on the coastal plain of the Arctic National Wildlife Refuge, Alaska.· I.

the Aichilik River on the east, and the Canning River on the I west. The study area comprises tundra and is mostly continuously vegetated with low-growing plants, including sedges, grasses, mosses, , forbs, and dwarf . Taller shrubs are I generally restricted to drainages. Shallow soils are underlain with permafrost, and the ground surface remains frozen from about mid-September to mid-May. A detailed description of the study I area, including geology, climate, soils, vegetation, and wildlife can be found in Garner and Reynolds (1986). Eighteen sites were randomly selected on the coastal plain (Fig. I 1). Enlarged maps of each site were produced from digitized data of 1985 Landsat land cover classes. Five to ten unique polygons of uniform vegetation were delineated on the maps within a 3-mile H radius at each site, for a total of 126 polygons. Polygons were transferred from the maps to 1:63,000 color infrared photographs using a zoom transfer scope. These photos were used in the field I to locate the polygons. Two botanists were transported to each polygon by helicopter. The polygons were observed from the air to determine which I vegetation type covered the majority of the area. A brief data collection stop was made, and plant community descriptions, including moisture level, dominant species, and major life forms I were recorded. The polygon was then classified into the Landsat land cover types developed for the 1982 and 1985 maps {Walker et al. 1982, Markon 1986). Ground or aerial photographs were taken I at most polygons to provide a record of the vegetation. Map classification of polygons was determined on the enlarged Landsat 1985 maps. Landsat 1982 classifications were determined from a 1:250,000 scale map since the digitized data was unavailable and enlargements could not be produced. A zoom transfer scope was used to overlay polygons from the Landsat 1985 maps onto the 1982 map. The Landsat 1982 classification which covered the majority of the polygon was recorded. SAS Tabulate Procedure was used to.produce contingency tables ·I depicting agreements and disagreements between the following: - ground observations and Landsat 1982 map classes I - ground observations and Landsat 1985 map classes ground observations using Landsat 1982 and Landsat 1985 classes I - map classifications for Landsat 1982 and Landsat 1985 The major diagonal of each table shows the number of agreements that occurred, while the off-diagonal elements of the table show I disagreements. Column and row totals show commission and omission errors for each vegetation type. Commission errors occured when a site was interpreted as one type on the maps but I was found to be another type on the ground, while omission errors I 5 I I.

occurred when types on the ground were not recognized on the I maps. Detailed ground descriptions of plant communities were used to I determine why disagreements occurred. Map and botanist classifications and community descriptions were entered into dBase, data base management software (Ashton-Tate 1986). community descriptions often included more than one vegetation I type, and these were listed in order of importance. Percentages of.each community and surface form descriptions were included when available from field data. Data were ordered by map or I botanist classifications, and tables listing classifications and community descriptions were produced. B This data was used to classify disagreements in to the following types of errors: - cutpoint - plant community is intermediate between two I closely related land cover classes - misclassification - plant community clearly does not belong to map classification I - description - land cover class description needs modification to clearly include this plant community I Results and Discussion

I 1982 Landsat map Landcover types on the 1982 Landsat map had 53% agreement with I ground determinations (Table 1). Disagreements at 25% of the points w~re caused by difficulties in determining cutpoints between closely related land cover classes (Table 2). Seventeen percent of the points had been misclassified on the map, and I disagreements at 6% of the points were due to problems with the land cover class descriptions. I cutpoint errors frequently occurred between adjacent land cover classes on a moisture gradient: II - water (I) ' - aquatic tundra: pond/sedge tundra complex (II) - wet sedge tundra (III) II - moist/wet sedge complex (IVa) - moist sedge, prostrate shrub tundra (V) All sites with water alone were correctly classified by the map. Three disagreements occurred between.water and aquatic tundra on sites with small amounts of vegetation. The amount of vegetated area allowed in a site classified as water is not included· in the I land cover description. One disagreement occurred between I I ------

Table 1. Accuracy assessment table showing Landsat 1982 land cover types compared with ground determinations, coastal plain, Arctic NWR, 1987. 1982 Landsat-land cover types -=--~~~~M~a~p~c=la~ss~1~·f~i~c=a~t~i~o=n~-==-~~~Omission Ground determination I II III IV V VI VII VIII IX X XI errors

Water I 2 1 2 6/20 Aquatic tundra II 3/6 Wet sedge III 7/17 Moist/wet or Dry prostrate shrubs IV 2 1 • 8/17 Moist sedge, prostrate shrub2 v 1 • 13/32 Moist tussock, shrub VI • 0/2 Moist shrub, tussock VII • • 16/17 Shrub tundra VIII • 1 • 2/2 Partially vegetated IX 4/7 Barren gravel or rock X ~®: 0/5 Wet gravel or mud XI . . 0 0/1 Commission errors 2 2 ., 17 10 10 1 0 1 8 2 59 16 5 13 26 29 12 2 0 7 13 3 126

Agreement = 53% Errors = 47%

I Agreements between map clasification and ground determinations are circled. 2 Type V also includes moist graminoid/barren complex tUndra. ------liil iilil - - 1!1!!1 - - ',-. .

GroUnd descriptions for disagreements between botanist ground determinations and the 1982 Landsat land cover map, Arctic BWR, ;f ~,~~~;ic\~.. ~ •· Alaska 1987 • . '~~;:~-,:,·;;;;}".':.~ :":.-.-:...... ,--:-...,_":'-_.-,...,..----.A-:g::re=e:-----..,M;a::-p~------;:;--::----.,--::-.,.------..... ·· '"~ .' .. ;; :... ~'>- BOtariist c~ll- 'ment clase Ground description

,,, I •. ~ Water 14, Aquatic sedge Small ponds (80%) with wet sedge ';'< Partially vegetated Shallow ponds Water in river channel, moist low ahmb, and partially vegetated gravel bars Barren Water in river channel and partially vegetated gravel bare Wet gravel Water in river channel and partially vegetated gravel bare Tidal flats with shallow water and mud wa:ter Ponds with aquatic grass along edges Aquatic grass on edge of ponds and very vet sedge in low-centered polygons with standing water i · Wet sedge · Very wet sedge in low-centered polygons vith standing vater.and a' few ponds

10 Moist/vet Wet sedge and moist sedge, dwarf shrub strange or rima Wet sedge with some shrub cover

Moist sedge, shmb Wet sedge drainages with moist sedge, shrub

_Aqu.atic' sedge Dry dwarf shrub river terrace (over .25% barren) _-·Moist sedge, abm b Dry dwarf shrub, moss river terrace vitb barren mud areas Dry dwarf shrub, forb river terrace . Partially vegetated Dry dwarf shrub river terraces (over 4~ area) barren gravel rind river channel Barren gravel Dry dwarf shrub river terrace (over 2~ veg), barren gravel,· and channel

Wet. sedge Moist sedge, dwarf shrub Moist sedge, dwarf shrub with some wet sedge Moist sedge, dwarf shrub polygons vitb vet sedge troughs Moist sedge, dwarf shrub Moist graminoid, dwarf shmb, barren with Eriophorum angDstitolium~ bigelowii, integrifolia, and lichens . . '',- · ,· Moist sedge, dwarf shrub with patches of tussocks · · · ·~'··· Moist sedge, dwarf shrub with _Q!!:!!, aquatilis, Eriophorum :~ .. angustifolium and Salix planifolia · :-,;:.;:. tussock Moist sedge, dwarf shrub with Salix planifolia and moist dvart Shrub, sedge, barren with Dryas intesrrroria Moist dwarf shrub, sedge along stream and partially vegetated. gravel bars ------iiii - - -

~ble 2. continued

Agree­ Map Error type >.'· Botanist call ment class C M D Ground description

V:j:. Moist 2 tussock

VII. Moist shrub, 1 Moist/vat Moist dwarf shrub, sedge tussock high-centered polYgons vith vet tussock sedge troughs Moist sedge, prostrate Moist dwarf shrub, sedge water tracks and moist tussock, dwarf shrub shrub tundra 2 Moist dwarf shrub, sedge tussock tundra 1 Moist dwarf shrub high-centered polygOns vith vet sedge troughs Moist tussock, 4 Moist dwarf shrub, sedge water tracks and moist tussock, dwarf shrub shrub tundra 1 Moist dwarf shrub, sedge tussock tundra 2 Moist dwarf shrub high-centered polygons vith vet sedge troughs

VIII. Shrub 0 Moist/vet or 1 Moist low shrub, partially vegetated, and barren gravel bars along tundra Dry prostrate stream Barren gravel 1 Moist dwarf shrub, graminoid partially vegetated river bar (over 25% shrub cover)

IX. Partially ; Barren gravel 4 Moist low shrub or dry dwarf shrub, forb (5-20%), barren gravel vegetated bars and water in river channel

. x. Barren 5 gravel

XI. Wet gravel 1 or mud

Total 31 21 -7- ... ':.} Perc~nt of all 53% 25% 17% 6'/, points

I Error type: C - cut point problems between closely related types; M - misclassified by map; D - class descriptions need modification I, . ·: .

I aquatic tundra and wet sedge on a site with low-centered polygons with standing water. · I Cutpoint problems occurred between wet sedge and moistjwet complex at seven sites and between moist sedge, prostrate shrub and moistjwet complex at five sites (Table 2). All seven sites were classified as moistjwet complex on the map, but were called I wet sedge by botanists. Two types of outpoints occurred at these sites. Four sites were described as containing primarily wet sedge with less than 40% moist sedge, prostrate shrub strangs or I rims, and three sites were wet sedge tundra which contained small amounts of shrub cover. Three sites which were described by botanists as primarily moist sedge, shrub polygons with wet sedge I troughs were classified as moistjwet on the map. Conversely two sites classified by the map as moist sedge, shrub tundra were described by botanists as wet sedge drainages with moist sedge, dwarf shrub (moist/wet complex). Four sites which were clearly I described as moist sedge, prostrate shrub tundra by botanists were misclassified on the map, two as wet sedge and two as I moistjwet complex. Many disagreements occurred among moist tussock, shrub; moist shrub, tussock tundra and moist sedge, prostrate shrub; (Table 1). Botanists identified considerably fewer sites as tussock I tundra and more sites as moist shrub, tussock tundra than the map. This indicates that botanists used a stricter definition of tussock tundra and a more encompassing definition of moist shrub, I tussock tundra than the map. Agreement occurred at only one moist dwarf shrub, tussock site, which contained tussock tundra with dominant moist dwarf shrub, sedge water tracks. Seven sites I classified as tussock tundra by the map were determined to be moist shrub, tussock tundra by botanists. Five of these were outpoint errors with four sites described as having dominant moist dwarf shrub water tracks (Table 2). Two sites which I included moist dwarf shrub high-centered polygons with no tussocks were misclassified as tussock tundra. I Another three.water track sites; two moist dwarf shrub, tussock sites; and one moist dwarf shrub high-centered polygon site were misclassified as moist sedge, prostrate shrub tundra (Table 2). I Three sites with moist dwarf shrub, tussock high-centered polygons with wet sedge troughs were classified as moistjwet sedge on the map. The description of the moist;wet sedge land I cover type does not include this habitat type. Three sites identified as moist sedge, prostrate shrub tundra on II the ground were classified as tussock tundra on the map (Table 2). cut point disagreements occurreo at two sites described as containing moist sedge, prostrate shrub with patches of tussocks.· 'I The third site was misclassified as it contained no tussocks. I

I '· ~ .. •, ,· .~- ..

I. · .. . ''· ,. ' :. :'

On river floodplains, cutpoint problems occurred between classes I of increasing vegetative cover (Table 2): - barren gravel (X) II - partially vegetated (IX) - dry prostrate shrub (IVb) Amounts of vegetative cover found in each of these classes was I not specified in the class descriptions (Walker et al. 1982). Botanists therefore used the cutpoints which were included in the 1985 class descriptions: less than 5% vegetative cover for I barren gravel and 5-20% cover for partially vegetated areas. These cutpoints appeared to be lower than those used in the 1982 map classification. All polygons which botanists identified as Ia barren floodplain were classified as such on the map (Table 1). But four polygons which botanists called partially vegetated were classified as barren on the map, and two polygons which were called dry prostrate shrub tundra were classified as partially I vegetated. Three braided river beds (one with over 25% shrub tundra, one dryas terrace with over 20% dry prostrate shrub, and one dominated by moist sedge, prostrate shrub tundra) were I misclassified as barren gravel on the map (Table 2). Cutpoint disagreements also occurred between dry prostrate shrub I and moist sedge, prostrate shrub tundra (Table 2). These two land cover types are dominated by and similar forbs, lichens, and mosses. The amount of moss and sedge cover I separates these two classes. Three floodplain sites were clearly misclassified into unrelated land cover types (Table 2). One dry prostrate shrub river I terrace was classified as aquatic sedge. Two braided stream beds dominated by moist low shrub tundra were classified as moistjwet or dry prostrate tundra and barren gravel. I Three sites called water by botanists, which included primarily water in river channels with partially vegetated gravel bars, were classified as partially vegetated, barren gravel or wet I gravel (Table 2). The description of water should be modified to state the water in river channels may not include islands of gravel or vegetation. Two areas of tidal flats with shallow water I and mud were classified as wet gravel or mud on the map. One of these areas was called water by the botanists. The description I of wet gravel or mud should be modified to include this habitat. 1985 Landsat map I Overall, 48% agreement occurred between land cover types on the Landsat map and ground determinations (Table 3) •. The .majority of· disagreements, 27% of all points, were cutpoint problems between closely related land cover types .(Table 4) • Twenty-two percent I ,_ . - . ~ '. .

''-· .·. I '··: ., .. I ------

Table 3. Accuracy assessment table showing Landsat 1985 land cover types compared with ground determinations, coastal plain, Arctic NWR, 1987.

1985 Landsat land cover types Map classification Omission Ground determination CW OW SW VW WG MW MP TT ME ADS DP SF BF errors

Clear water (cw)@- • 1 2 3/17 Offshore water (OW) • 0 • • • 0/1 Shallow water (SW) • • 1 3/3 Very wet graminoid (VW) • 1/5 Wet graminoid (WG) • 6/17 Moist/wet complex (MW) • 5/8 Moist prostrate scrub (MP) 3 • 1 22/32 Moist graminoid tussock (TT) • 1/2 Mesic erect scrub (ME) 1 5 2 13/17 N 0) - Alluvial deciduous shrub (ADS) • 1 1 • 1 • 3/3 Dry prostrate scrub (DP) 1 2 ·@A· 7/9 Scarcely veg floodplain (SF) 1 -~A 2/6 Barren floodplain (BF) . . ~ 0/6 Commission errors 0 0 .i _2 17 _]_ 16 .2.. 0 5 _2 66 14 1 0 8 20 20 17 17 7 0 2 9 11 126

Agreement = 48% Errors = 52%

1 Agreements between map classification and ground determinations are circled. ------liiii == -

Table 4. Ground descriptions for disagreements between botanist ground determinations and the 1985 Landsat land cover map, Arctic NWR, Alaska 1987.

gree­ Map Error type Botanist call ment class C M D Ground description

Water 14 Very vet 1 Small ponds (80%) with wet sedge 1 Barren floodplain 2 Water in river channel ( over 50%) and gravel bars

Shallow water 1 Very vet 1 Shallow ponds with small areas of vegetation Vet graminoid 1 Water in river channel, moist low shrub, and partially vegetated gravel bar Barren floodplain 1 Tidal flats with shallow water and mud

Very wet 4 Vet graminoid 1 Very wet sedge in low-centered polygons with standing water and a few ponds

Wet graminoid 11 Very wet 1 Wet sedge, very wet sedge in low-centered polygons, and dried ponds Moist/vet '3 Wet sedge and moist sedge, dwarf shrub 2 Wet sedge with some shrub cover

Moist/wet Vet graminoid 3 Wet sedge and"moist sedge, dwarf shrub Moist prostrate 2 Moist sedge, dwarf shrub and wet sedge drainages

Moist prostrate 10 Wet graminoid 1 Moist sedge, dwarf shrub with Dryas integrifolia and Salix reticulate 1 Moist sedge, dwarf shrub; wet sedge; and tussock tundra Moist/vet '3 Moist sedge, dwarf shrub and wet sedge 1 Moist sedge, dwarf shrub along stream drainage 1 Moist graminoid, dwarf shrub, barren Moist tussock 4 Moist sedge, dwarf shrub with Eriophorum angustifolium, , and Salix planifolia 4 Areas with some Eriophorum vaginatum tussocks scattered or in patches '3 Hummocky sedge, dwarf shrub communities with !• angustifolium and ~- planifolia, and and~ integrifolia Mesic erect '3 Moist sedge, dwarf shrub with Salix planifolia ssp. pulchra and F.riopho~lm angustifolium. Includes areas of hummocky sedge with Arctagrostis latifolia, ~ bigelowii, and Dryas integrifolia Barren floodplain 1 Moist dwarf shrub, sedge along stream, partially vegetated gravel bars, and water

Moist tussock 1 Moist prostrate 1 Moist sedge tussock, dwarf shrub with a fev water tracks of moist sedge, dwarf shrub ------

Table 4. continued

Agree­ Map Error type Botanist call ment class C M D Ground description

Mesic erect Wet graminoid 1 Moist dwarf shrub, sedge polygons, and wet sedge low areas (40%) Moist/wet 2 Moist dwarf shrub, sedge tussock high-centered polygons with wet sedge troughs 2 Moist sedge tussock, dwarf shrub with moist dwarf shrub, sedge water tracks 1 Moist dwarf shrub, sedge tussock tundra Moist prostrate 1 Moist dwarf shrub, sedge tussock, high-centered polygons with wet sedge troughs 1 Moist sedge tussock, dwarf shrub with moist dwarf shrub, sedge water tracks

Moist tussock 2 Moist dwarf shrub, sedge water tracks and moist sedge tussock dwarf shrub tundra 2 Moist dwarf shrub, sedge tussock 1 Moist dwarf shrub high-centered polygons and wet sedge troughs

Alluvial shrub 0 Wet graminoid 1 Barren and partially vegetated gravel bars (50-60%); moist lov shrub, dwarf shrub, forb river terrace (30%); stream (15-20%)

Moist/wet 1 Moist low shrub and dwarf shrub (over 25%) with partially-vegetated and barren gravel bars along a stream

Scarcely vege~ted 1 Moist dwarf shrub, graminoid, forb partially vegetated river bar (over 25% shrub cover where vegetated)

Dry prostrate 2 Moist/vet 1 Dry dwarf shrub, moss ~iver terrace Moist prostrate 1 Dry dwarf shrub, moss river terrace with barren mud areas 1 Dry dwarf shrub, forb river terrace Scarcely vegetated 4 Partially vegetated river terraces with well over 20% vegetation

Scarcely vegetated 4 Very wet 1 Barren gravel bars, moist low shrub, and water in river channel. Barren floodplain 1 Barren gravel bars; dry dwarf shrub, forb (5-20%); and water in river channel Total ()() 34 27 -5-

Percent of all 48% 27% 22% 4% points

1Error type: C - outpoint problema between closely related types: M - misclassified by map; D - class descriptions need modification.

·. ;· I

of the disagreements were misclassifications on the map I and 4% were caused by incomplete descriptions of land cover types. Present land cover type descriptions with suggested additions of modifications are included in Appendix I. Cutpoint errors frequently occurred between adjacent land cover classes on a moisture gradient: I - water -·very wet graminoid - wet graminoid I - moist/wet complex - moist prostrate shrub I Two areas containing ponds with small areas of wet sedge were classified as very wet graminoid on the map, but as water or shallow water by botanists. This indicates that water classes are tightly defined, and do not contain even small inclusions of I vegetation. The distinction between very wet graminoid and wet graminoid tundra was unclear in two instances, where areas included a combination of wet sedge, low-centered polygons with standing water, and small ponds. A large amount of disagreement occurred between moistjwet complex and wet graminoid or moist prostrate scrub (Table 3). Moistjwet I complex is defined as being intermediate between these two classes, and often contains both classes intermixed. The botanists assumed that at least 40% of each class must occur in a I moistjwet complex site based on descriptions developed for the 1982 map (Walker et al. 1982). The map disagreed with botanist calls of moistjwet complex five out of eight times. Three of I these areas called wet graminoid on the map were described on the ground as primarily wet graminoid tundra with moist prostrate scrub as a secondary type, and the two areas called moist prostrate scrub on the map were described as having moist I prostrate scrub as the primary type and wet graminoid as the secondary type. This indicates that the map could distinguish between wet graminoid and moist prostrate shrub, even when I botanists determined that each class occurred over 40% of the area. I Of those polygons classified as moistjwet complex on the map, botanists identified five as wet graminoid and five as moist prostrate scrub (Table 4). Moist prostrate shrub strangs and rims were an important component of many of the wet sites, and I wet graminoid tundra occurred in four of the moist prostrate scrub sites, however botanists determined that the primary type in each polygon covered over 60% of the area. The high numbers I of both omission and commission errors in moistjwet complex indicate that the map classification is not consistent, and that ·accuracy could~ be improved by dropping this class.

15, <;_ •' I

Another five sites classified as moistjwet tundra were called I mesic erect dwarf scrub on the ground (Table 4). Three of these sites, which were dominated by dwarf shrubs and tussocks, were misclassified by the map. The other two sites included I high-centered polygons with wet sedge troughs. These disagreements were probably due to a misinterpretation of the class description for moistjwet complex. The present description does include high-centered polygons, but does not include the I species usually dominant on these sites (Appendix I). Much confusion occurred among moist prostrate scrub, moist I graminoid tussock tundra, and mesic erect scrub (Table 3). Seven moist prostrate scrub polygons were misclassified as tussock tundra on the map (Table 4). Four of these were typical moist I sedge, dwarf shrub tundra with Eriophorum angustifolium, Carex aguatilis, and Salix planifolia ssp. pulchra. Three sites were hummocky sedge, dwarf shrub tundra, which included the above plant community, and moist sedge, dwarf shrub tundra with frost I scars dominated by Carex bigelowii and Dryas integrifolia. Botanists called only two polygons tussock tundra, and therefore I must have been using a stricter definition of this class than that used for the map (Table 3). Eight polygons mapped as tussock tundra were described as having some tussocks, but were called moist prostrate or mesic erect scrub (Table 4). Botanists I assumed that up to 20% cover of cottongrass tussocks could be included in moist prostrate scrub based on descriptions for the 1982 map (Walker et al. 1982), but this percentage may be lower I for the 1985 map. The mesic erect dwarf scrub class is described as part of a continuum with tussock tundra, but is dominated by dwarf shrubs with over 25% cover. Two of the sites where I disagreements occurred were described as dwarf shrub, tussock tundra with over 25% cover, and the other two sites had dominant water tracks with Salix planifolia interspersed with tussock tundra. Another polygon which was misclassified as tussock I tundra on the map, included moist dwarf shrub high-centered polygons with no tussocks. I The distinction between moist prostrate scrub and mesic erect scrub was not clearly defined in the land cover class descriptions (Appendix I). The ten sites for moist prostrate I scrub where botanists and the map agreed represented three major community types: - moist sedge, dwarf shrub flat--centered polygons I - hummocky moist sedge, dwarf shrub - frost-scar tundra I With only three sites in agreement, the plant communities present in mesic erect scrub are less clear, but appear to include: I - moist dwarf shrub, tussock tundra 16. ------

I. ,·;;:_... ·..

- tussock tundra with well-defined water tracks I - moist dwarf shrub high-centered polygons However, the map did not consistently classify these plant I communities as mesic erect scrub (Table 4). One water track site and one high-centered polygon site were classified as moist prostrate scrub, and four sites with these communities were I classified as moistjwet complex, as discussed previously. Three sites were classified as mesic erect which botanists called moist prostrate (Table 4). Two of these sites were misclassified II by the map as they contained communities previously identified as moist prostrate scrub. The third site contained a mixture of I both moist prostrate and mesic erect shrub communities. Cutpoint determinations and misclassifications were the main sources of disagreement in floodplain areas (Table 4). Cut point :D errors occurred between the following land cover types which have increasing amounts of vegetative cover: - barren floodplain I - scarcely vegetated floodplain - dry prostrate shrub scrub or alluvial deciduous shrub I One barren floodplain area on the map was determined to be scarcely vegetated, and five scarcely vegetated polygons in the map were determined to be dry prostrate scrub or alluvial deciduous shrub with well over 20% vegetative cover on the I ground. Another barren floodplain site on the map was determined to have over 20% moist prostrate scrub on the ground. Barren floodplain and scarcely vegetated floodplain classes in the 1985 I map were described as having less than 5% and 5-20% vegetative cover, respectively. Ground descriptions indicated that these I percentages are low. Classification errors occurred between the floodplain classes and very wet graminoid, wet graminoid, moistjwet complex tundra (Table 4). Floodplain habitats included braided stream beds with I barren gravel bars, partially vegetated gravel bars, islands of dry prostrate scrub or alluvial deciduous shrub, and water·· in river channels. The spectral reflectances of these different I components are apparently averaged together in the Landsat classification, and appear similar to the graminoid classes. This error in misclassifying braided stream beds was also noted I in a number of areas visited during the accuracy assessment conducted in 1985 (P. Kuropat, pers. comm.). Another floodplain area classified as wet graminoid on the map I was called shallow water by botanists (Table 4). Shallow .water is defined to include riverine areas with shallow water, or combination of water and gravel bars~ Although braided I floodplains with shallow water and gravelbars are common·on all

...... ' - ; ·-··. ., , ' : _·, l ... c ---.: -. 1: I

rivers of the coastal plain, the map shows this class occurring 'I only on portions of the Canning River. Three polygons called water or shallow water by botanists were I classified as barren floodplain on the map (Table 4). one of these included tidal flats with shallow water and mud. Another tidal flat area was also classified as barren floodplain. The I I description of this class should be expanded to include tidal mud flats which may be covered by shallow water. Two river channels with over 50% water were called clear water by I botanists, but classified as barren floodplain on the map (Table 4). This shows that water is strictly defined and does not contain inclusions of any other land cover type, as was seen in R the earlier very wet graminoid examples. A change in the definition of water will correct this problem (Appendix I). n A comparison of the 1982 and 1985 Landsat land cover maps Botanists treated land cover classes on the 1982 and 1985 maps as being directly related so that each land cover class on the 1982 I map had an equivalent class on the 1985 map (Table 5). The descriptions developed for the 1982 map (Walker et al. 1982) were often used to supplement the 1985 descriptions. On the maps, 67% of the polygons were classified in these equivalent land cover classes (Table 6). Differences between the 1982 and 1985 map classifications are discussed below. I A number of disagreements between the 1982 and 1985 Landsat classifications occurred between adjacent classes on a moisture gradient (Table 6). Five polygons classified as moistjwet tundra I complex in 1982 were called wet graminoid in 1985. Botanist calls were in agreement with the 1982 map twice and with the 1985 map three times (Table 7). Another two polygons classified as I moist/wet in 1985 were called wet graminoid in 1982; botanists were in agreement with the 1982 map on one and the 1985 map on the other. These disagreements illustrate the difficulties in I distinguishing among these classes. Two polygons which included tussock tundra with dominant water tracks were classified as tussock tundra in 1982, and I misclassified as moist/wet tundra in 1985 (Table 7). Another polygon which included moist dwarf shrub, tussock tundra was also I misclassified as moistjwet complex in 1985. A large number of disagreements occurred in the tussock tundra class (Table 6) Eleven polygons classified as moist sedge, prostrate shrub in 1982 were called tussock tundra in 1985 (Table I 7). Botanist calls agreed with the 1982 map on nine of these. Six polygons included hummocky moist sedge, shrub tundra with frost scars or scattered tussocks which fall into the ·moist :~ ' - ·~ ~ :. ' sedge/barren complex .class. (V) on the 1982 map. The 1985 map

.·:, '·.:. ·-·_.'•;' I. . '· ..:-· .. .··:. ------

Table 5. Botanist determinations of land cover classes according to Landsat 1982 and Landsat 1985 classification systems, coastal plain, Arctic NWR, 1987.

Botanist determinations Landsat 19821 Landsat 1985 I II III IV v VI VII VIII IX X XI Total

Clear water 1 17 Offshore water 1 Shallow water . . 3 Very wet graminoid ~ ~- 5 Wet graminoid .@ . . 17 Moist/wet complex @. 8 Moist prostrate shrub .@ . . 33 Moist graminoid tussock ®· 2 Mesic erect shrub .@ . 16 -o Alluvial deciduous shrub . @ 1 3 Dry prostrate shrub ® . . 9 Scarcely veg. floodplains ® . . 6 Barren floodplain . ®0 6 Total 20 6 17 17 33 2 16 2 7 5 1 126

Equivalent classes = 98%

I I-Water; II-Aquatic sedge; III-Wet sedge; IV-Moist/wet complex or Dry prostrate shrub; V-Moist sedge, barren complex; VI-Moist tussock, shrub; VII-Moist shrub, tussock; VIII-Shrub tundra; IX-Partially vegetated; X-Barren gravel or rock; XI-Wet gravel or mud. 2 Equivalent classes are circled. ------

Table 6. Landsat 1982 classification compared to Landsat 1985 classification. coastal plain, Arctic NWR, 1987.

Land cover classifications Landes t 19821 . Landsat 1985 I II III IV v VI VII IX X XI Total Clear water 14 Offshore water . . 0 1 Very wet graminoid ~ @ 1 . 2 8 Wet graminoid . @ 5 1 2 20 Moist/wet complex 0 • @ 1 2 20 Moist prostrate shrub 2 0 1 . 17 Moist graminoid tussock 11 ® . 17 Mesic erect shrub . 2 3 ® • 7 Dry prostrate shrub ® 2 ..., Scarcely veg. floodplain 1 0) . 9 0 Barren floodplain 4 Q) 11 Total 16 5 13 26 29 12 2 7 13 3 126 Equivalent classes = 67% I I-Water; II-Aquatic sedge; III-Wet sedge; IV-Moist/wet complex or Dry prostrate shrub; V-Moist sedge, prostrate shrub or barren complex; VI-Moist tussock, shrub; VII-Moist shrub, tussock; IX-Partially vegetated; X-Barren gravel or rock; XI-Wet gravel or mud. 2 Equivalent classes are circled. ------·------·Table 7. Disagreements between the 1985 and 1982 Landsat maps, coastal plain, Arctic National Wildlife RefUge, 1987.

Map Botanist - Map Map classifications disagree­ Misclass­ Agreements 1985 1982 menta ification 1985 1982 Ground description

Very vet I. Water 1 Aquatic graminoid on edge of ponds and vet sedge low-centered III. Wet sedge polygons with standing water 1 1 Wet sedge, very wet sedge with standing vate! and dried ponds with bacterial benthic mat and algae IX. Partially veg. 1 85 Shallow ponds 1 85 Braided river bed with barren gravel bars and moist low shrub tundra

Wet graminoid IV. Moist/vet 2 2 Wet sedge and moist sedge, dwarf shrub complex 1 Wet sedge with moist sedge, dwarf shrub strange or rima 2 Wet sedge low-centered polygons v. Moist sedge, 1 Wet sedge and moist sedge, dwarf shrub complex prostrate shrub IX. Partially veg. 2 85 Water in river channel, moist low shrub and partially vegetated gravel bars

Moist/vet V. Moist sedge, 1 85 Moist dwarf shrub, sedge tussock tundra prostrate shrub VI. Moist tussock, 2 85 Moist tussock tundra with dominant water tracks of moist shrub dwarf shrub, sedge

Moist prostrate IV. Moist/vet 1_. 1 Moist sedge, dwarf shrub, wet sedge, and ponds (beaded stream) shrub 1 1 Moist sedge, dwarf shrub flat-centered polygons , N VI. Moist tussock, 1 1 Moist sedge tussock, dwarf shrub with few water tracks ...... shrub

Tuaaock tundra V. Moist sedge, 3 85 3 Moist sedge, dwarf shrub with Eriophorum angustifolium, Carex prostrate shrub aquatilis, and Salix planifolia ssp. pulchra or barren complex 3 3 Hummocky moist sedge, dwarf shrub with frost scars 3 3 Hummocky moist sedge, dwarf shrub with scattered tussocks 2 82 Moist dwarf shrub, sedge tussock and moist tussock, dwarf shrub tundra

Mesic erect v. Moist sedge, 1 82 1 Moist dwarf shrub, sedge water tracks and moist tussock, shrub shrub (ME) prostrate shrub tundra or barren complex 1 1 Hummocky dwarf shrub, sedge with Salix planifolia ssp. pulchra, .§_. reticulate, Eriophorum angustifolium, !· vaginatum and Carex bigelowii VI. Moist tussock, 1 82 1 Moist dwarf shrub high-centered polygons shrub 2 82,85 Moist sedge, dwarf shrub with~ planifolia and EriOphOrum angustifolium

Scarcely veg. II. Aquatic sedge 1 82 Dryas river terrace i"loodplain X. Barren gravel 3 82 3 Braided river floodplain with partially vegetated (20% cover) or rock bars

. Alluvial deciduous 1 82 Partially vegetated river bar with 25% cover of willows ·shrub· 1 82 Dryas river terrace (over 20% cover) barren gravel bars and river terrace. Totals I '.; .. -~ ·.·

appears to have redefined the tussock tundra class to include this habitat type. Modifications are needed in the 1985 class description to clearly indicate this change (Appendix I). Another three sites which included typical moist sedge, dwarf shrub tundra with Eriophorum angustifolium, Carex aguatilis, and Salix planifolia ssp. pulchra were clearly misclassified on the 1982 map. The remaining two sites included moist dwarf shrub, tussock tundra and were thus misclassified by the 1982 map. Only two agreements occurred in mesic erect scrub (Table 6). Two polygons called mesic erect scrub on the 1985 map were called I moist sedge, prostrate shrub on the 1982.version; botanists agreed with each version once (Table 7). Another three sites classified as mesic erect shrub were called tussock tundra on the I 1982 map. -Botanists agreed with the 1985 map at one site with moist dwarf shrub high-centered polygons, and disagreed with both maps at the other two sites. These sites were moist sedge, dwarf I shrub with Salix planifolia ssp. pulchra and Eriophorum angustifolium which clearly fit the description of moist prostrate shrub. I Three braided floodplains which were classified as partially vegetated floodplains on the 1982 map were misclassified as very wet or wet graminoid on the 1985 map (Table 7). In contrast, one I Dryas river terrace site which was classified scarcely vegetated floodplain in 1985 was misclassified as aquatic sedge in 1982. Five braided floodplains with partially vegetated river bars (two with over 20% plant cover) were misclassified as barren gravel in I 1982. The three scarcely vegetated sites were correctly classified by the 1985 map, while the two with over 20% cover I were misclassified as scarcely vegetated floodplain. I Conclusions and Recommendations Since the largest source of error was cutpoints between closely related types (25% in 1982 and 27% in 1985) and description errors (6% in 1982 and 4% in 1985) can be corrected, we conclude I that these maps show general distributions of land cover classes accurately 83% and 78% of the time. Therefore, Landsat MSS data provides an efficient, cost-effective method for mapping over I large areas. These maps provide basic information for wildlife habitat studies and are useful for regional conservation planning (Bailey 1984, Talbot and Marken 1986). Wildlife studies designed I to predict or minimize the impacts of oil and gas exploration and development will require more detailed site-specific vegetation maps. Thus further mapping efforts are needed on the coastal I plain of ANWR. In order to correct cutpoint problems which were the largest source of error, further mapping efforts must include detailed I vegetation sampling to allow quantification_of the cutpoints. At

'.:; :22 '>;.<' ",, .. , . :_. ' I­ ~;;_. '. ·. :;: .. ' 'i.:·: ·.\' >"· .. I. . . ' . . ,, ~. . ' ... L I

this point it is hard to know whether satellite technology can I consistently separate closely related types. However, the land cover class descriptions developed in the past often did not specify quantitative cutpoints for shrub cover, tussock cover, I total plant cover, moisture levels, and other attributes used to separate the classes. The ground data used to develop these maps may have been inadequate for defining the differences between I closely related types. A number of the misclassifed polygons were located on river floodplains where the map was unable to distinguish braided I floodplain communities with river channels from wet or moist plant communities. Higher resolution Thematic Mapper or SPOT data may help to resolve this problem. Description errors, although comprising only a small portion of all errors, illustrate the need to develop detailed ground descriptions of vegetation classifications to allow accurate interpretation of map information.

I Acknowledgements We wish to thank Dave Douglas of Alaska Fish/Wildlife Research I Center for producing maps of the study sites from the digitized Landsat data, Trans-Alaska helicopter pilot Ken Butters for aerial support, and Cathy Curby for assistance with data I analysis. I Literature Cited Bailey, T.N .. 1984. Terrestrial habitats and wildlife species. Technical Supplement, Kenai National Wildlife Refuge I Comprehensive Conservation. U.S. Fish and Wildlife Service, Anchorage, Alaska. Garner, G.W., and P.E. Reynolds. 1986. Final report baseline I study of the fish, wildlife, and their habitats. u.s. Fish and Wildlife Service, Anchorage, Alaska. I Marken, C.J. 1986. Arctic National Wildlife Refuge land cover mapping project users guide. u.s. Geological Survey, EROS I field office, Anchorage, Alaska. Nodler, F.A. and A.J. LaPerrier. 1977. Landsat mapping -north slope Arctic National Wildlife Refuge. Report in files of I the u.s. Fish and Wildlife Service, Fairbanks.

,··,.

I .. ·. ~ .,,.<:,_-_ : I ,. . J . : " -~. ' ·> ... · 23 ,. ''-

~: •, .·. ·' ., : ' ,' • ' 'r • •· ~.. ; ;

. ' . . .·.. ·.·•-·. I

Talbot, s.s. and C.J. Markon. 1986. Vegetation mapping of I Nowitna National Wildlife Refuge, Alaska using Landsat MSS digital data. Photogrammetric Engineering and Remote :I Sensing 52(6): 791-799. I Walker, D.A., K.R. Everett, w. Acevedo, L. Gaydos, J. Brown, and P.J. Webber. 1982. Landsat - assisted environmental mapping in the Arctic National Wildlife Refuge, Alaska. I u.s. Army Cold Regions Research and Engineering I Laboratories, Hanover, N.H. CRREL Report 82-27. 52pp.

I I I I I I :I I I­ 24 I ------

Appendix I. Land cover class descriptions, typical communities, and suggested additions or changes for the 1965 Landsat map, coastal plain, Arctic NWR, 1987.

Land cover class descriptions (Markon 1987) 'rypica 1 communities Additions or changes

CLEAR WATER including lakes, pond, and rivers. Clear lake'S and ponds Includes deep clear water lakes or river channels, Lakes and deep river channels with with clear or silty water. muddy water May have small inclusions of very wet tundra with Arctophila Lake with small inclusions of fUlva. No inclusions of wet aquatic Arctophila fulva sedge or river gravel bars graminoid tundra occur in this land .cover class.

SHALLOW WATER includes riverine areas in which the water is Occurs only on the Canning shallow or when the satellite sensor received spectral data River on the coastal plain. from both water and gravel bars and recorded,them as one class.

OFFSHORE WATER is the Beaufort Sea shoreline as was digitized Lagoon Includes ocean, lagoons, and · on the Flaxman Island, Barter Island, Demarcation Point, and deep mouths of drainages. Mount Michelson quads and applied to the land cover image. Mouth of drainage Those water areas north of the shoreline were labeled offshore water.

VERY WET GRAMINOID is a graminoid-dominated formation associated Aquatic Arctophila fulva with a~atic habitats surrounding large, open bodies of fresh graminoid tundra/la~ water; very wet habitats which contain numerous small bodies wet sedge of open water; and coastal habitats frequently inundated with salt water. Surface forms include low-centered polygons with Very wet Carex Aquatilis abundant standing water, thaw lake basins, the littoral zones sedge tundra/ponds of lakes and coastline. Arctophila ~ is the primary species (75% of area) in deeper water, up to 1m deep, with Carex aquatilis, , and Eriophorum angustifolium dominating areas where the water 1s less than 30 em deep.

WET GRAMINOID formations are associated with wet habitats. These Wet Eriopho~Jm angustifolium, habitats often receive water by surface and subsurface flow from Carex aquatilis sedge tundra surrounding terrain. The habitats generally have standing water with some standing water , :throughout the summer. Vegetative cover is continuous, as depth ,'of water is not a limiting factor to plant establishment and Wet Carex chordorrhiza, ·growth. The habitat has fev drained microsites associated with EriophO'ruin angu stifolium, polygon rims, strangmoor, hummocks, etc. Landforms where these Carex aquatilia, Potentilla habitats occur are river deltas, drained lake basins, and river ~tria/pond (less than 20% channels, where surface forms are low-centered polygons, and of area) strangmoor. Primary taxa include numerous Carex spp. and Eriophorum spp. Common species occurring in-this type include Wet Carex aquatilia, Eriophorum Carex aquatilis, .£· microglochin, .£. atrofusca, .£• amblyorhyncha, an~stifOlium, E. Scheuchzeri , £. scirpoidea, .£· rostrate, .£• bigelowii, .£. physocarpa, £_. (low-centered polyRQns)/moist misandra, Eriophorum vaginatum, !· angustifolium, !· russeolum, £_. aqu~tilis, !· angustifolium, Equisetum fluviatile, Scir;pus sci oideas, ~· caespitosua, Dryas integrifolia, Salix 'Pedicularis spp., Valeri gonum spp., ~lata, S. planifOTiB spp. · 'l'omenthypnum nitens, and s spp. Some shrub species pulchra (rima) include Arctostiijjiiy"!os !!!!!!:!.• Salix anata, and s. arctophila. ------

Appendix I. continued

Land cover class descriptions (Markon 1987) Typical communities Additions or changes

MOIST/WET TUNDRA COMPLEX is a type in which dwarf shrubs and Moist Eriophorum angustifolium, Botanists assumed that at least graminoids occur together in habitats intermediate in moisture Carex aquatilis, Salix planifolia 40% of both vet graminoid and regime between the wet graminoid and moist prostrate dwarf shrub sap. pulchra, ~· reticulate, Ruble moist prostrate dwarf shrub formations. Wet and moist microsites are often intermixed in a chamaemorua, Aulacomnium tu tundra must occur in moist/vet pattern in this habitat. High-centered and flat-centered A. palustre, Dicranum sp., complex baaed on Walker et al, polygons are common surface features in river delta and drained splendena sedge, dwarf ahru , moss (1982). However, our results lake basin landforms. Along river drainages, disjunct string tundra (flat-centered polygons, rims)/ indicated that the 1985 map bogs are the most common surface form. Common species on these wet Carex aquatilis, Eriophorum did not use consistent sites include Dryas integrifolia, Salix lanata, S. reticulate, angustifolium, E. rusaeolum sedge percentages in distinguishing Caasi tetragona; Vaccinium uligiii'Osilm-;-Eri"Ophorum triste, tundra (low-centered polygons, between these classes. The tum, Carex bigelowii, C. membranacea, Polygonum troughs) map would be more accurate bis-;:t:"':o~r~"',.;.;;;rellii:riillaeta, Senecio spp., Tomenthyp~um nitens, if areas called moist/vet · and Hylocomium spp. Moist Carex aquatilis, Salix planifolia tundra were reclassified into asp. pulchra, S.lanata sedge, dwarf either wet graminoid or moist shrub tundra/wet--c:-8quatilis sedge prostrate shrub tundra. tundra (flat-centered polygons with well developed thermokarat pits) Areas of high-centered polygons with moist Salix planifolia Wet Eriophorum angustifolium, !• ssp. :p.tlchra, Betula nsns, Scheuchzeri, Carex aquatilis sedge Vaccinium vitis-idaea dwarf tundra/moist Dryas integrifolia, shrub tundra and extensive vet Salix reticul~E.angustifolium sedge troughs were also included dwarf shrub, sedge-tundra (strangmoor) in this class, but vera not clearly included in the class description.

MOIST PROSTRATE DWARF SCRUB contains prostrate dwarf shrub Moist Salix planifolia asp. p1lchra, A list of typical communities (over 25% cover) and sedge formations occupying mesic habitats Eriophorum anguatifolium, Carex is needed in this class on gentle to moderately steep slopes. In the foothills, these aqu~tilis dwarf shrub, sedge tundra description to separate moist habitats are frequent on mid to lower slopes which receive prostrate dwarf scrub from BUbsurface drainage from adjacent terrain. Dryas integrifolia Moist Eriophorum anguatifolium, !• mesic erect dwarf scrub. is often the dominant species, Equisetum arvense and the moss, vaginatum, Salix reticulata, Dryas Tomenthypnum ~· are characteristic species of this formation. integrifolia, s. planifolia ssp. Botanists assumed that up to 20% Carex bigelowii gives the habitat a hummocky surface. Moist pulchra, ~· arotica, Tomenthypnum cover of tussocks could be habitats on slightly elevated microsites in the coastal plain, and nitena hummocky sedge, dwarf shrub included in this class based on alluvial terraces in the foothills and mountains are often drier, tundra Walker et al. (1982), but this ·.owing to greater exposure and lack of water from surrounding percentage appeared to be lover terrain. Lichens are more important than mosses in these drier Moist Carex bigelowii, Arctagrostis for the 1985 map. habitats. These habitats are very similar to the moist micrositea latifolia, Dryas integrifolia, Salix of the vet/moist dwarf shrub, graminoid land cover class. Other phlebophylla, Dicranum sp., •'". species important to this type include Salix arctics, S. lanata, Tomenthypnum nitens, Aulacomnium ,~· pulchra, ~ chamaemorus, Saxifrag"iliiirculus, ~· pi.m~ turgidum, Tha~ sp. crustose Petasites frigidus, Eriophorum vaginatum, and Carex aqu~tilis. , graminoid, dwarf shrub, barren frost-scar tundra ------

Appendi% I. continued

Land cover class descriptions (Markon 19B7) Typical communities Additions or changes

MOIST GRAMINOID TUSSOCK is a subclass related to part of the Moist Eriophorum vaginatum, Salix Botanist observations indicated scrub subclass mesic erect dwrf scrub. Species dominating planifolia ssp. pulchra, ~· reti'CUlata, that this class could include this class include the tussock producing Eriophorum vaginatum Dryas integrifolia, Vaccinium vitis­ less than 20% tussocks, and over and· Carex bigelowii. Also occurring are Betula nana, Salix idaea, Hylocomium splendens, Pti1JGffum 25% cover of shrubs. planifOITa ssp. pulchra, S. reticulate, Dryas in~if~ CIIIare, Tomenthypnum nitens sedge Vaccinium uliginosum, V. vitis-idaea, Pyrola spp., Polygonum tussock, dwarf shrub tundra/moist The tussock tundra class also bistorta, .!_. viviparum-;- Cetraria spp.,""""Toiii.eiithypnum ni tens, _!. angustifolium, ~· planifolia ssp. included hummocky moist sedge HYlocomium splendens, and Ptilidium ciliare. This is an pu lchra sedge, dwarf shrub tundra shrub tundra with frost scars herbaceous class which is dominated ~ graminoids and has less (water tracks) and scattered tussocks. The than 25% cover of shrubs. minimum percentage of tussocks allowed in this class needs to be defined.

MESIC ERECT DWARF SCRUB contains over 25% cover dwarf shrubs Moist Salix planifolia ssp. pulchra, This class included: primarily from the taxa Betula spp., Salix spp., Vaccinium Eriophorum angustifolium (water tracks)/ - moist dwarf shrub, tussock uliginosum, and Cassiope~gona. These shrubs are usually moist E. vaginatum, E. angustifolium, tundra O.l m to 0.5 m in height with interlocking branches. This type Carex bigelowii, Arctagrostis latifolia, - tussock tundra with well­ is common on low rolling hills. On hill sides at lower ~· planifolia ssp. pulchra, Cassiope defined water tracks ·elevations (below 900 m), graminoid tussocks often occur with tetragons, S. reticulate, Vaccinium - moist dwarf shrub hip­ the dwrf shrubs. Major tussock producing plants include vitis-idaea-;- Ledum palustre ssp. centered polygons Eriophorum vaginatum and ~ bigelowii. ·Major shrub species decumbens sedge tussock, dwarf shrub include Betula glandulosa, B. nana, Salix glauca, s. reticulate, tunilra However these plant communities ~· plani~ssp. pulchra,-Ledum deCiiiiiben'S;Iiii:Ccinium ~ were not consistently idaea, and Empetrum nigrum. '"Otii'er species present ma.v include Moist Salix planifolia ssp. pulchra, classified as mesic erect dwarf ~ lugens, Carex scirpoidea, Equisetum arvense, !· scirpoides, Betula~ ssp. exilis, Ledum palustre scrub HYlocomium splendens, Tomenthypnum nitens, and Sphagnum spp. ssp. deCii.iiibens, Erropil'Oruiii"Viiiinatum dwarf shrub, sedge tussock tundra/moist planifolia ssp. pulchra, Carex aqua tilis, Eriophorum angu stifolium dwarf shrub, sedge tundra (water track)

Moist Salix planifolia ssp. pulchra Betula nana ssp. exilis, Vaccinium Vit'IS-id"iiea, Spha~pp., Hylocomium splendens, Aulacomnium turgidum, A. palustre dwarf shrub tundra/wet sedge (well-developed high-centered polygons with troughs) ------

Appendix I. continued

Land cover class descriptions (Markon 1987) Typical communities Additions or changes ALLUVIAL DECIDUOUS SHRUB occurs on frequently flooded gravel Moist Salix brachycarpa ssp. niphoclada, This class was not mapped on the sites dominated by~ planifolia ssp. pulchra and~· S. hastata, S. lanata ssp. Richardsonii coastal plain. alaxensis. On some sites especially on the coastal plain, low shrub riverbar Betula spp. (dwarf birch) may occur with Salix in older allUvial terraces. The number of species~rring with Moist Salix glauca, !!· reticulate, Dryas the above' species as co-dominants or as understory are many integrifolia, Arctosta os rubra, and may include: Salix lanata, s. richardsonii, S. glauca, Oxytropis maydelliana, sarum sp. low ~· brachycarya, !!.lliiStata, !!• reticulate, Arctostap1iY"iO"i3 shrub, dwarf shrub, forb river bar rubra, as integrifolia, Eguisetum arvense, !· variegatum, E. sci es, Carex spp., Festuca spp., Juncus castaneus, 'P'etasi spp. ,Hedysarum spp., and HylocOiiiiii'iilspp. On the coastal plain, this class usually occurs in small patches which can not be distingnished on Landsat imagery, and is classified as scarcely vegetated floodplain.

DRY PROS'l'RATE DWARF SCRUB formation occurs on river terraces, Dry Dryas integrifolia, Cassiope slightly elevated microsites on the coastal plain, and upper tetragons, Carex Bigelowii, C. slope positions in the foothills. Bare soil is often an misandra, C~ia cucullata­ important component of the ground surface as a result of frost dwarf shrub, aed~e river action. Because of the harsh environment, plants do not achieve terrace. heights greater than 10 em. Some of the more commonly occurring , shrubs are Dryas integrifolia (usually dominant) and 1?_. octopetala Dry Dryas integrifolia, ,llith Arctostaphylos~. Salix reticulate, ~· rotundifolia and Sali'ier'Otundifolia, Cassiope tetragons. Nonwoody species include Saxifrage hirculus, ~opis nigrescens !!~ oppositifolia, Polygonum bistorta, Petasites arctics, dwarf shrub, forb river Polemonium spp., Flquisetum arvense, Carex spp., Festuca spp., terrace Hierochloe spp., Epilobium latifolium, ~ glaciale, and the lichen Cetraria spp.

SCARCELY Vl!XlETATED FLOODPLAIN is a subclass that is a result Moist Salix spp., Arctagrostis Area with over 20% cover were of the initial invasion of plants on recent river alluvium. latifolia, Festuca sp., included in this class. Plant cover averages 5-20 percent. Some of the more common Oxytropis app. low shrub, species include Epilobium latifolium, Calamagrostis canadensis, graminoid, forb partially , Bromus spp., and Salix spp. On the coastal plain (below the vegetated gravel bar (5-20% , 500 meter contour) this type includes alluvial deciduous scrub plant cover)/shallow stream communi ties. Barren gravel bars/dry Dryas integrifolia, Salix ova~ia, Artemisia arct~Oxytropis nigrescens dwarf shrub, forb river terrace (overall less than 20% plant cover)/river

, BARREN FLOODPLAIN is less vegetated than scarcely vegetated Barren gravel bars/river/ Areas with over 5% cover were floodplain. It consists of alluvium and includes silt, sand, few partially vegetated bars, included in this class. and rocks. Plant cover is less than 5 percent and includes with Artemisia arctics, the same species as scarcely vegetated floodplain if present. Epilobium latifolium, and Tidal mud flats should be Salix alaxensis added to the description.

Tida 1 flats (mu'd and shallow water)