Professional Documents
Culture Documents
net/publication/241324746
CITATIONS READS
40 1,468
2 authors, including:
Fiona Danks
UNEP World Conservation Monitoring Centre
21 PUBLICATIONS 394 CITATIONS
SEE PROFILE
Some of the authors of this publication are also working on these related projects:
All content following this page was uploaded by Fiona Danks on 17 November 2017.
To cite this article: F. S. Danks & D. R. Klein (2002) Using GIS to predict potential wildlife habitat:
A case study of muskoxen in northern Alaska, International Journal of Remote Sensing, 23:21,
4611-4632, DOI: 10.1080/01431160110113890
F. S. DANKS*
Department of Biology and Wildlife, University of Alaska Fairbanks,
Fairbanks, AK 99775-6100 , USA
and D. R. KLEIN
Institute of Arctic Biology, University of Alaska Fairbanks, Fairbanks, AK
Downloaded by [University of Cambridge] at 13:23 17 November 2017
99775, USA
1. Introduction
Geographical Information Systems (GIS) provide a set of powerful tools that
allow spatial data to be collected, stored, maintained, transformed and displayed for
a speci c purpose (Berry 1993, Johnson 1993, McLaren and Braun 1993, Markon
1994), thereby allowing an analysis of complex spatial problems. GIS is not limited
to spatial data but can also establish links with and accommodate attribute data.
GIS is able to assimilate and analyse a wide variety of data types in a database, and
has enabled complicated problems and issues to be addressed in a more
comprehensive way than previously. Its powers and uses are still being tested.
The work described here was a study aimed at determining the feasibility of
using GIS as a practical tool in wildlife habitat assessment and prediction in the
Arctic and its application in wildlife and related land management. In Greenland, a
GIS mapping project was used as a basis for setting regulations protecting muskox
and geese populations from human activity in critical habitat areas (Bay 1994). The
potential for GIS and remote sensing to be used in habitat studies is being realized.
Habitat of species such as black bears, deer, rabbits, small mammals and birds has
been examined (Clark et al. 1993, Homer et al. 1993, Lauver and Whistler 1993,
Coker and Capen 1995, Heaton 1996, Knick and Dyer 1997, Segura Lopez 1998).
Planning for resource development is obviously a key issue in the management
of wildlife, especially in the Arctic. Information from this project should be of
assistance in assessing how muskoxen and their habitat may be aVected by diVerent
Downloaded by [University of Cambridge] at 13:23 17 November 2017
by development activities because they are relatively sedentary, with high delity for
local habitats, and suitable muskox habitat is patchy in distribution (Klein 1988).
The diet of muskoxen is diverse and includes sedges, forbs, grasses and woody
shrubs (Smith 1994). Muskoxen vary their diet in relation to their seasonal locational
changes. Previous habitat studies in Greenland and Alaska have shown that in
summer muskoxen prefer low lying, often tussocky tundra, providing it is not too
wet, and riparian areas, while in winter they move to drier ridge tops and stream
bluVs (Wilson 1992, Klein et al. 1993, Ihl 1999). Variation in muskox diet can also
occur regionally and altitudinally, according to available vegetation types (Parker
and Ross 1976, Parker 1978, Rapota 1984, Thing et al. 1987, O’Brien 1988, Klein
et al. 1993, Department of Renewable Resources 1996, Nellemann 1997, Nellemann
1998, Ihl 1999). The muskox diet is heavily in uenced by terrain and associated
Downloaded by [University of Cambridge] at 13:23 17 November 2017
characteristics, particularly in winter months. Unlike caribou, muskoxen are not well
adapted for digging through snow to obtain forage, and in winter they seek areas
with low snow accumulations (Smith 1994, Klein 1996), i.e. exposed ridges and
bluVs. Snow depth and hardness are factors associated with terrain ruggedness
(Nellemann and Thomsen 1994, Nellemann and Reynolds 1997), which re ects the
interaction of slope and aspect with wind direction.
Figure 1. Maps showing areas covered by the Kuparuk River Drainage Basin and the
NPR-A, location of established muskox groups in the Itkillik Hills (Kuparuk River
Drainage Basin) and location of muskox sightings and evidence of presence along the
Kikiakroruk River (NPR-A). Boundary between the NPR-A and the Kuparuk River
Drainage Basin lies roughly along the Colville River.
A key assumption in this study is that muskoxen in Alaska possess the same or
similar habitat requirements as muskoxen in other similar Arctic regions, dependent
upon availability and abundance of common resources and extent of environmental
similarity. This allows previous muskox habitat studies to strengthen the basis for
assessing the models of predicted habitat. A second assumption in developing the
maps is that muskoxen in the future will select for similar habitat criteria as they
have done in the past. This means that the models’ predictions of suitable habitat
will be valid in the future if and when muskoxen do permanently expand their
habitat into the NPR-A.
2. Methods
2.1. Overview
Data layers were created representing all available aspects of muskox habitat.
These data were analysed and the ensuing GIS habitat model and subsequent habitat
maps created. The development of the GIS habitat model was based on an area of
known muskox presence in the Kuparuk River drainage basin ( gure 1). After
Sixth Circumpolar Symposium on Remote Sensing of Polar Environment 4615
evaluating the quality of the model for the Kuparuk region, the model was extrapol-
ated to an area of potential value as habitat for muskoxen, the NPR-A, which is
located west of the Colville River as indicated in gure 1. Because of its close
proximity and similarity in physical characteristics such as vegetation and topo-
graphy to the NPR-A, the Kuparuk region was chosen as the development site of
the model over other regions in northern Alaska, also with known muskox presence.
Muskox seasonal location data, from the Kuparuk River drainage basin region,
were required to create the initial habitat model and to establish a basis of assessment
of habitat suitability in the NPR-A. Vegetation maps were required to incorporate
seasonal forage needs. Elevation data, slope characteristics, aspect characteristics
and an index of the interaction between the latter two (terrain ruggedness) were
necessary to take into account terrain variability and its eVects on habitat selection.
Downloaded by [University of Cambridge] at 13:23 17 November 2017
All data were projected using Albers equal area projection, a common projection in
Alaska as it yields a lower distortion over a large area than many of the other
projections (Snyder 1987). All data layers ultimately had a spatial resolution of 50 m.
ARC/INFO was the GIS program used to produce, edit and manipulate the
coverages, images and grids, and the resulting maps and analyses (ESRI, Redlands,
CA). Components of the main ARC program, including GRID, ARCEDIT and
TABLES were also used. ArcView, an independent program that functions in con-
junction with ARC/INFO, was also used (ESRI, Redlands, CA). Arc Macro
Language (AML) programs were written and run through ARC/INFO to accomplish
more eYciently the functions and analyses.
et al. 1987, Wilson 1992, Ihl 1999, D. Klein personal communication, 1999). The
buVer sizes were then added into the GIS attribute tables and buVer coverages
created in ARC/INFO.
Since buVer coverages were to be overlain with the vegetation, elevation and
terrain layers, which were all in grid format, buVer coverages were also converted
to grids. All buVers were combined into two grids, one of buVers from winter
locations (April 1997, 1998 and 1999) (ADF&G data) for the winter habitat map
and the other of buVers from summer locations (June 1999 ADF&G data and ABR’s
Colville River data) for the summer habitat map.
Muskoxen were not resident during this study in the NPR-A so no locations
were available; however, presence of winter faeces and a weathered skull found in
the Kikiakroruk River area ( gure 1) and aerial survey sightings in summer indicate
Downloaded by [University of Cambridge] at 13:23 17 November 2017
Table 1. Comparison of classi cation schemes for Kuparuk River drainage basin and NPR-A
satellite-based vegetation maps.
Class Class
division Description division Description
the buVered muskox locations in the Kuparuk region was determined separately for
summer and winter buVered locations. Ranked schemes were then created according
to relative presence of each vegetation class in the buVers. Vegetation classes whose
representation was distinctly diVerent were assigned their own rank, while classes
that had very similar values were given the same rank. Classes with lowest representa-
tion typically were grouped together. This resulted in a six order ranking for summer
and a ve order ranking for winter (table 2). The nal summer and winter Kuparuk
vegetation suitability maps, re ecting muskox presence in or use of vegetation, were
then created by reclassing original vegetation class values to re ect the ranked orders.
The same procedure was applied to create summer and winter vegetation suitabil-
ity maps in the NPR-A. Since the Kuparuk vegetation classi cation scheme had
Table 2. Kuparuk River drainage basin vegetation map classes and rank (and per cent) of
each class within summer and winter muskox location buVers.
Summer Winter
been applied to the NPR-A to link the two areas, the same rank order scheme for
summer and winter vegetation used in the Kuparuk was applied to the NPR-A. It
is these summer and winter vegetation suitability maps that were used in the initial
model (Kuparuk maps) and its subsequent application to the NPR-A (NPR-A maps).
2.2.4. T errain ruggedness data
A number of methods for determining terrain ruggedness have been used in
habitat studies, ranging from simple contour counting by hand to more complicated
procedures based on computer algorithms, and from values that represent a single
point to values at a point that incorporate values of surrounding points (Nellemann
and Thomsen 1994, Gallant et al. 1995, Nellemann and Fry 1995, Nellemann and
Cameron 1996, 1998, Nellemann and Reynolds 1997, Nicholson et al. 1997 ). More
general geomorphometric works examine similar issues (Moore et al. 1991, Nogami
Downloaded by [University of Cambridge] at 13:23 17 November 2017
1995, Rhoads and Thorn 1996, Schmidt and Dikau 1999). A GRID ARC/INFO
based approach that calculates a variety of terrain characteristics including aspect,
slope, aspect ruggedness and slope ruggedness from DEM lattices was employed in
this study. This methodology was based on the one initially devised by Nicholson
et al. (1997) and relies on circular statistics (Zar 1996). This methodology incorpor-
ates the characteristics of surrounding areas in devising the resulting ruggedness
value for each grid cell and is therefore arguably better than a system that counts
contours. In addition, the terrain ruggedness index of Nellemann and Thomsen 1994,
Nellemann and Fry 1995, Nellemann and Cameron 1996, 1998, Nellemann and
Reynolds 1997, was used to combine the derived slope and aspect ruggedness
characteristics.
Aspect and slope values were derived from a base elevation map (DEM) and
intermediate aspect and slope layers created. In cells of the intermediate aspect grid
with a zero slope value, an aspect value of 1 resulted. This was reclassed in GRID
to a value of zero. Aspect is expressed as the sin or cos of the aspect in radians;
consequently two aspect coverages were created, an X component and a Y compon-
ent. Final aspect and slope coverages were converted to grids with integer values
rather than the raw oating point values.
Aspect ruggedness and slope ruggedness values were then derived from the aspect
and slope grids, with the ruggedness values calculated based on the values of the
contiguous surrounding cells. The number of cells included in the ruggedness calcula-
tions was chosen according to habitat selection characteristics of muskoxen, i.e. they
typically select habitat on a localized scale.
The slope ruggedness grid was created from the slope oating point grid. A grid
of slope standard deviation was created, which when converted into an integer grid,
represented slope ruggedness.
The aspect ruggedness grid was created through a process of intermediate grids.
The oating point aspect grids, the X and Y component grids, were the base layer.
A mean vector grid was then created (the R grid). Next, angular dispersion was
calculated using the X component grid, Y component grid and R (mean vector) grid.
Lastly, the angular grid was created using the intermediate X, Y and R grids. The
angular grid values were converted from oating point values into integer values
yielding the aspect ruggedness grid.
Finally, the two oating point ruggedness grids were combined into an overall
terrain ruggedness grid using Nellemann’s index as follows:
aspect ruggedness×slope ruggedness
Terrain ruggedness= (1)
aspect ruggedness+slope ruggedness
Sixth Circumpolar Symposium on Remote Sensing of Polar Environment 4619
Terrain ruggedness takes into account the combination of aspect and slope in
determining the terrain characteristics and therefore more accurately re ects the
in uence of climatic eVects such as wind and snow cover.
Since all of the terrain characteristic data layers were derived from DEMs, the
spatial resolution was established at 50 m.
Suitable data on snow cover and depth were not available in time for use as
initially hoped, but terrain ruggedness measurements were considered to be represent-
ative of potential habitat use by muskoxen through the relationship of snow accumu-
lation and its characteristics to terrain variability (Fleischman 1990, Nellemann and
Reynolds 1997).
The following data layers were included in the winter and summer habitat
suitability models for the inhabited Kuparuk River drainage basin: seasonal muskox
buVered locations, winter or summer vegetation according to ranking of types used,
DEM, slope, X component of aspect, Y component of aspect, slope ruggedness,
aspect ruggedness and terrain ruggedness (a total of nine data layers per map). In
the NPR-A, the same data were used with the exception of the muskox location data.
To diVerentiate the habitat selected by muskoxen in the occupied Kuparuk region
from the remaining areas according to each variable (the above data layers), a
multivariate maximum likelihood classi cation (Tabachnick and Fidell 1989) was
done within ARC/INFO. For this analysis, mask grids of muskox presence versus
absence were created with the buVered location grids (both winter and summer). A
value of one was assigned to the cells within the buV ers and a value of zero was
assigned to all the cells within the Kuparuk boundary but not included within the
buVered region. A total of 32 buVered locations were used in the winter analysis and
79 buVered locations were used in the summer analysis.
To run the maximum likelihood analysis a gridstack (a le containing all predictor
variables with the exception of the buVered muskox locations) was created for the
Kuparuk region, one for the summer habitat model and one for the winter habitat
model. A signature le that produced means and covariance matrices for both
seasons was created using the appropriate gridstack and buVered muskox location
grid. Next a maximum likelihood classi cation grid was created for summer and
winter habitats using signature le statistics from the buVered location grid and
appropriate gridstack. In the analysis, all cells in the stack are assigned to one of
the classes represented in the signature le, muskox presence or absence; the
maximum likelihood classi cation determines to which class each cell belongs.
Finally, prediction grids of probability of muskox presence or absence were
produced for both winter and summer data, again using the appropriate signature
le and gridstack. These grids were used to create the nal maps for the Kuparuk
model prediction of muskox habitat suitability. The models of habitat suitability
that resulted show relative suitability along a scale of 0 to 100, that is 0 represents
the lowest likelihood for suitable muskox habitat and 100 represents the highest.
The nal Kuparuk models were visually examined for anomalies in the prediction
of muskox presence or absence for winter and summer habitat use. None were found
and the resulting maps seemed consistent with reports in the literature on muskox
winter and summer habitat choices and characteristics of the areas within the
Kuparuk region (Robus 1981, Wilson 1992), for example, preference for drier more
elevated areas in winter versus wetter, riparian areas in summer.
4620 F. S. Danks and D. R. Klein
The model was then applied and extrapolated to the NPR-A region. To produce
the habitat models for this region, the same maximum likelihood analysis was
completed on the NPR-A. In addition to the gridstack of variables from the Kuparuk
region, gridstacks for the winter and summer variables in the NPR-A were created.
Exactly the same variables were incorporated, just covering a diVerent area. The
signature les that were produced for the Kuparuk region using the buVer locations
and the Kuparuk grid stacks were used again for creation of the maximum likelihood
prediction models, but with the NPR-A gridstacks instead in these last two steps.
There are two resulting maps for the NPR-A, both prediction models of relative
habitat suitability (values of 0 to 100), one for winter and one for summer. These
maps are equivalent to those produced in the initial Kuparuk region model and
serve as the nal results showing potential muskox habitat in the NPR-A. Again, as
Downloaded by [University of Cambridge] at 13:23 17 November 2017
with the base layer data, the nal models have a spatial resolution of 50 m.
3. Results
3.1. GIS analysis
3.1.1. Muskox location data
Muskox signs, winter faeces and a weathered skull were found in the Kikiakroruk
River area. ADF&G muskox location data from June 1999 included the location of
a herd of eight muskoxen in the NPR-A, quite close to the Kikiakroruk River
( gure 1). These observations suggest that muskoxen may potentially use certain
areas of the NPR-A and parts of the NPR-A may in fact provide suitable habitat.
Based on the seasonal division of muskox location data, winter and summer
habitat choices seemed to diV er markedly. In winter, the Itkillik hills was a more
favoured area, with group presence concentrated in a smaller area, while in summer,
animals spread out further, venturing away from the hills and often south towards
the coast. In summer, the Colville Delta and surrounding area seemed to be highly
favoured ( gure 2).
3.1.2. Vegetation maps
The Kuparuk winter and summer maps of vegetation type preference in ranked
order appeared accurate given knowledge of muskox seasonal habitat preferences.
The maps showed clear distinctions between winter and summer occupation of
vegetation types, indeed suggesting diVerent habitats are important seasonally (as in
table 2). The adjusted NPR-A vegetation map with classes of 1 to 8 based on the
Kuparuk vegetation classi cation appeared to be accurate. The original and adjusted
NPR-A vegetation maps were compared and all vegetation trends appeared similar
with the only di Verence being that the new spatial resolution was coarser and there
were fewer vegetation classes (8 versus 17 classes). This newly classi ed vegetation
map was divided into ranked summer and winter vegetation use maps using the
same schemes employed in the Kuparuk region. There is some degree of error in the
vegetation data, due to the natural patchiness of vegetation (and of some types in
particular such as heath or riparian vegetation) , the inability to ground truth entire
mapped regions, and the reclassi cation of vegetation types into a vegetation suitabil-
ity index. However, with the scale and the broad vegetation type divisions used, the
data are suitable and the quality acceptable.
3.1.3. Maximum likelihood analysis
Seasonal vegetation use (winter and summer), elevation, slope, X component of
aspect, Y component of aspect, slope ruggedness, aspect ruggedness and terrain
Sixth Circumpolar Symposium on Remote Sensing of Polar Environment 4621
Downloaded by [University of Cambridge] at 13:23 17 November 2017
Figure 2. Muskox winter and summer locations from aerial survey data overlaid with the
Kuparuk River drainage basin vegetation map.
4622 F. S. Danks and D. R. Klein
ruggedness were all incorporated into the resulting models of muskox habitat
suitability.
Elevation in the Kuparuk showed far greater variation towards the south and
the Brooks Range while variation in the northern parts was less extreme. However,
even small changes are important in muskox habitat selection. Slope in the Kuparuk
shows trends similar to elevation with greater variation found to the south. However,
some variation was present in the central and northern areas, evidence of additional
topographic relief that may not be evident in the elevation map scale presented here.
The Kuparuk River drainage basin maps of aspect result in a rugged display of
directional relief that is present. While distinct topographi c variation is evident for
the whole region, these maps clearly re ect terrain variability present in the less
extreme areas, something that the other maps failed to distinguish in such detail and
Downloaded by [University of Cambridge] at 13:23 17 November 2017
something that is important in muskox habitat selection ( gure 3). The nal three
maps for the Kuparuk, slope ruggedness, aspect ruggedness and terrain ruggedness,
variables that have been built on more basic ones, also distinguish more clearly
terrain variability throughout the region. Each of these maps represents a unique
component in muskox habitat determination. Similar trends in terrain distinction
and latitudinal variation were seen for the equivalent NPR-A maps.
The Kuparuk winter habitat prediction map ( gure 4) shows that higher, drier
regions have a higher relative suitability, with the Itkillik Hills region and other
areas of greater topographic variation and elevation further to the east and south
yielding higher probabilities of habitat selection. Vegetation in these more suitable
areas re ects winter diet. The lower wetter areas have less potential for use in winter.
Predictions of suitable summer habitat in the Kuparuk are markedly diVerent
( gure 5). Areas likely to be occupied by muskoxen in summer are drainages in
riparian areas, in the Colville River Delta area and in other lower lying areas of the
Kuparuk River drainage basin. Elevated areas and bluV regions with more sparse
vegetation are not likely to provide favourable habitat in summer, as at this time
the animals are able to take advantage of the richer vegetation in the lower wetter
areas that were inaccessible in winter due to snow. South of the Itkillik Hills, the
region was not shown to be suitable. Though potential winter habitat area extends
a bit further south than potential summer habitat, a large region before and including
the foothills and mountains of the Brooks Range is not suitable habitat in either
season. Elevation and topographic extremes may be too great for suitable habitat:
muskoxen do not require extreme topographic variation. For both the summer and
winter habitat maps in the Kuparuk region, areas larger than currently occupied
regions were suggestive of potential habitat. Since there are currently only around
100 muskoxen in the area, it is likely that the area is capable of supporting larger
numbers. Consequently maps of the Kuparuk region can be used in predicting future
muskox habitat occupancy as their numbers increase.
Unlike in the Kuparuk region, probability of habitat suitability in the NPR-A
cannot be supported with location data. However, since the model appears justi ed
in the Kuparuk region according to established muskox habitat knowledge, and
prior use of the area, and it is directly applied to the very similar NPR-A region, it
is assumed that its predictions will be supported. It should be noted however that
reliability of the extrapolated predictions is limited for a few reasons. (1) No ground
data for vegetation in the NPR-A was collected to test accuracy of the adjusted
vegetation classi cation. (2) The NPR-A is a vast area with varying terrain and
topographic characteristics, with diVerent wind patterns, which may aVect habitat
Sixth Circumpolar Symposium on Remote Sensing of Polar Environment 4623
Downloaded by [University of Cambridge] at 13:23 17 November 2017
Figure 3. Map of aspect (Y component) values for Kuparuk River drainage basin [(sin of
aspect value in radians)×1000 ].
selection di Verently than in the Kuparuk region. Snow patterns, however, in the
Kuparuk region appear to be a good analogue for the rest of the North Slope based
on both modelling and eldwork (M. Sturm, personal communication, 2000).
The most suitable winter habitat in the NPR-A is found in the middle section of
the NPR-A, stretching horizontally across most of the region, and in the more
topographically varied areas ( gure 6). Additional potential winter habitat is located
in the upper east corner, west-south-west of the Colville Delta, and includes the area
where eldwork was conducted and where signs of winter presence were found.
Wetter, lower areas do not show high potential for winter habitat, as expected, due
to the low accessibility of both these areas and the vegetation caused by heavy snow
4624 F. S. Danks and D. R. Klein
Downloaded by [University of Cambridge] at 13:23 17 November 2017
Figure 4. Predicted muskox winter habitat suitability in the Kuparuk River drainage basin.
Suitability ranking from 0 to 99 ( low to high).
Figure 5. Predicted muskox summer habitat suitability in the Kuparuk River drainage basin.
Suitability ranking from 0 to 100 ( low to high).
inadequate testing and a re ection that the model was not built for habitat
characteristics of more extreme values such as those found here.
Suitable summer habitat appears to be more plentiful and widespread than
suitable winter habitat in the NPR-A ( gure 7).The region of the NPR-A with highest
potential summer habitat value is the north-east corner, an area of less topographic
4626 F. S. Danks and D. R. Klein
Downloaded by [University of Cambridge] at 13:23 17 November 2017
Figure 6. Predicted muskox winter habitat suitability in the NPR-A. Suitability ranking from
0 to 99 ( low to high).
variability than the higher, hillier areas. The Kikiakroruk River area is included in
this region. In addition the same middle horizontal band is predicted to be favoured
summer habitat, though more value is placed on the south side of this band where
terrain variation is lower. Drainages found in the northern NPR-A where the land
is wetter and lower are shown to provide some potential summer habitat as well.
Vegetation in these areas re ects muskox summer preference. Again, just as with the
winter model, the southern portion below the middle band and including the foothills
of the Brooks Range, is not well suited for summer occupancy.
4. Discussion
Relationships between factors in uencing habitat selection are important and
play a role in determination of muskox habitat. A vegetation type suitable for
muskoxen may be located in two topographicall y di Verent areas, only one readily
accessible by or appropriate for muskoxen. Consequently it is not suYcient to
determine habitat simply based on one or even a few criteria. All potential factors
that in uence habitat selection should be considered. This is why an analysis of
potential habitat should incorporate a substantial set of variables, ideally all of those
that are important in habitat selection. This project has shown how GIS can be
applied and spatial data used to develop not only existing but predicted habitat
models, a tool that has great potential for addressing important habitat issues.
Seasonal variation in habitat use was evident in the Kuparuk region where
Sixth Circumpolar Symposium on Remote Sensing of Polar Environment 4627
Downloaded by [University of Cambridge] at 13:23 17 November 2017
Figure 7. Predicted muskox summer habitat suitability in the NPR-A. Suitability ranking
from 0 to 100 ( low to high).
known diVerences in muskox location over the seasons exist, but the GIS models
also distinguished between winter and summer habitats in both the Kuparuk and
NPR-A. Seasonal winter to summer movements in the Kuparuk region are north,
towards coastal areas along with a general dispersal from winter concentration areas.
Summer and winter habitat prediction maps for the NPR-A show similar patterns
of seasonally available habitat.
The method of determining ruggedness variables (aspect, slope and terrain) for
each cell based on all surrounding cell values is appropriate . As you move from cell
to cell surrounding terrain characteristics change and each cell’s value re ects this.
This is arguably how animals selectively use habitat, through the continuum in
change that develops. Terrain ruggedness measurements are representative of snow
accumulation through the direct relationship of snow accumulation to terrain variab-
ility (Fleischman 1990, Nellemann and Reynolds 1997). Areas chosen by muskoxen
in winter will presumably re ect their choice of habitats inclusive of in uences of
snow characteristics.
Inherent problems exist in a spatial analysis such as this one, where existing data
from a variety of sources, and collected for diVerent purposes, are relied upon.
However, these problems can be limited or overcome to yield valuable results. Some
degree of error, e.g. due to scale or GPS locations without diVerential correction, is
unavoidable in a project of this nature. Conclusions must be tempered accordingly.
4628 F. S. Danks and D. R. Klein
The 50 m spatial resolution data used in this project worked well, however, there
are issues of error and scale. The DEM cell values are taken from elevation contours
so the chosen value may not be truly representative of that particular region. Other
terrain characteristics developed are based on these elevation values. Changes in
terrain or vegetation that are not detectable at the scale used here, occurring within
a 10 m area for example, may aVect habitat suitability and choice for muskoxen.
Satellite maps at ner spatial resolution are becoming available for some areas but
at high cost, limiting the practicality of their acquisition, and they may not necessarily
be available for the desired region. In addition, unclassi ed data, if available, would
require involved methodology beyond the scope of this project, to create appropriate
maps. Furthermore, if all other data layers are not available at the same spatial
resolution, details in the layers that do have a ne spatial resolution will be lost.
Downloaded by [University of Cambridge] at 13:23 17 November 2017
The overall spatial resolution is only as good as the coarsest layer in the GIS.
Additional coverages derived from Advanced Very High Resolution Radiometer
(AVHRR) data, which shows time of green-up, were not added because of the coarse
1 km spatial resolution. Given the sedentary nature of muskoxen and the need for
more micro-scale habitat assessment, this did not seem appropriate. A coverage
representing rivers and streams was considered, however with its coarse scale, it did
not adequately detail smaller or more seasonal drainages which are important as
well in muskox habitat selection, thereby potentially biasing the results.
In addition to the error within the map layers, there is error in the GPS muskox
group locations. Because the GPS data were not diVerentially corrected, the margin
of error surrounding locational data could be up to approximately 60 m.
Maximum likelihood classi cation proved to be an eYcient analysis to use in
determination of predicted habitat based on present habitat. An alternate analysis,
logistic regression was tried, but the data set using all grid cells was too large for
the ARC/INFO system to handle, and limiting the extent was too great a comprom-
ise. Maximum likelihood classi cation seemed most appropriate given the objectives
and data formats, as compared to other analyses.
This project followed a unique approach in interpretation and assessment of
muskox habitat by basing analyses in part on spatial muskox location data from
aerial surveys. These data clearly distinguished distinct seasonal variation in habitat
use and selection, providing insight into the dynamic nature of muskox habitat. In
addition, the GIS analysis allowed the important, potentially in uential eVect of
group size in the determination of habitat suitability to be addressed.
The model of predicted muskox habitat was not extrapolated into other regions
because the focus was on the NPR-A, where suitable GIS data were available.
Application of the model was restricted to areas of similarity, the Kuparuk and
NPR-A, rather than introducing greater variability.
The approach used was to create a model of predicted muskox habitat in the
NPR-A based on extrapolation of muskox habitat criteria between areas currently
occupied and those not yet occupied. More speci cally, the GIS models of predicted
muskox habitat suitability should provide a sounder basis for consideration of the
value of muskox habitat in land management decisions that will need to be made
within the NPR-A and Kuparuk areas.
5. Conclusions
This study employed a unique application of GIS to generate a predictive habitat
model. Its results should have potential utility in land planning and resource manage-
Sixth Circumpolar Symposium on Remote Sensing of Polar Environment 4629
ment decision making. Results from this study suggest that the potential for continued
expansion of muskox numbers and distribution in Alaska’s Arctic exists.
There is great potential to build upon the foundations of this project. Re nements
in methodology could improve this study or result in bene t to the design of future
projects. Even the currently unavoidable errors in a project such as this may be
decreased should improved data become available. This project was designed so that
additional data or variables may be added as they become available, permitting the
analyses to be run again. In addition to the eVective terrain characteristic layers
employed here, data on snow cover and depth would also be of value. If more
location data were available, it would lend even greater support to the predictions
and might also enable a test of the model’s predictions to be run, providing a stronger
foundation for validity of the predictions. The in uence of global climate change on
Downloaded by [University of Cambridge] at 13:23 17 November 2017
Acknowledgments
We would like to thank the following for their support: Alaska Cooperative Fish
and Wildlife Research Unit; University of Alaska Graduate School; Bureau of Land
Management and David Yokel (Alaska); Alaska Department of Fish and Game;
Alaska Biological Research (ABR); Department of Biology and Wildlife, University
of Alaska Fairbanks; Associated Students of the University of Alaska Fairbanks; Dr
D. Verbyla, Dr A. D. McGuire, Dr R. Barry, and Dr J. Maier; and C. Ihl and fellow
students and scientists.
References
Bay, C., 1994, Review of vegetation mapping in Greenland. Circumpolar Arctic Vegetation
Mapping Workshop, L akta, Russia, 21–25 March 1994, pp. 38–43.
Berry, J., 1993, The application of GIS to mammalogy: basic concepts. In GIS Applications
of Mammalogy, edited by S. B. McLaren and J. K. Braun. Special Publication of the
Oklahoma Museum of Natural History in collaboration with the Carnegie Museum
of Natural History (Norman: University of Oklahoma), pp. 4–10.
Clark, J. D., Dunn, J. E., and Smith, K. G., 1993, A multivariate model of female black bear
habitat use for a geographic information system. Journal of W ildlife Management,
57, 519– 526.
Coker, D. B., and Capen, D. E., 1995, Landscape-level habitat use by brown-headed cowbirds
in Vermont. Journal of W ildlife Management, 59, 631–637.
Department of Renewable Resources, 1996, Muskox abundance, habitat use, and diet on
the Yukon Coastal Plain. Yukon Territorial Government TR-96-03, Whitehorse,
Yukon.
Ducks Unlimited, Inc., 1998, National Petroleum Reserve-Alaska earth cover classi cation
nal report. Submitted to US Department of the Interior, Bureau of Land
Management, Anchorage, AK. Contact John Payne, BLM, Anchorage, AK, USA.
4630 F. S. Danks and D. R. Klein
Fleishman, S. J., 1990, Lichen availability on the range of an expanding caribou (Rangifer
tarandus) population in Alaska. MS thesis, University of Alaska Fairbanks, Fairbanks,
AK, 81 pp.
Gallant, A. L., Binnian, E. F., Omernik, J. M., and Shasby, M. B., 1995, Ecoregions of
Alaska. USGS Professional Paper 1567. US Government printing oYce, Washington.
Heaton, J. S., 1996, Use of geographic information systems (GIS) and remotely sensed data
to characterize the habitat and predict the distribution of Peromyscus truie comanche.
MS thesis, University of North Texas.
Homer, C. G., Edwards, T. C., Ramsey, R. D., and Price, K. P., 1993, Use of remote sensing
methods in modeling sage grouse winter habitat. Journal of W ildlife Management,
57, 78–84.
Ihl, C., 1999, Comparative habitat and diet selection of muskoxen and reindeer on the Seward
Peninsula, western Alaska. MS thesis, University of Alaska Fairbanks, Fairbanks,
AK, 81 pp.
Jingfors, K. T., 1980, Habitat relationships and activity patterns of a reintroduced muskox
Downloaded by [University of Cambridge] at 13:23 17 November 2017
Nellemann, C., 1997, Grazing strategies of muskoxen (Ovibos moschatus) during winter in
Angujaartor up Nunaa in western Greenland. Canadian Journal of Zoology, 75,
1129–1134.
Nellemann, C., 1998, Habitat use by muskoxen (Ovibos moschatus) in winter in an alpine
environment. Canadian Journal of Zoology, 76, 110– 116.
Nellemann, C., and Cameron, R. D., 1996, EVects of petroleum development on terrain
preferences of calving caribou. Arctic, 49, 23–28.
Nellemann, C., and Cameron, R. D., 1998, Cumulative impacts of an evolving oil- eld
complex on the distribution of calving caribou. Canadian Journal of Zoology, 76,
1425–1430.
Nellemann, C., and Fry, G., 1995, Quantitative analysis of terrain ruggedness in reindeer
winter grounds. Arctic, 48, 172–176.
Nellemann, C., and Reynolds, P. E., 1997, Predicting late winter distribution of muskoxen
using an index of terrain ruggedness. Arctic and Alpine Research, 29, 334–338.
Nellemann, C., and Thomsen, M. G., 1994, Terrain ruggedness and caribou forage availability
Downloaded by [University of Cambridge] at 13:23 17 November 2017
during snowmelt on the Arctic coastal plain, Alaska. Arctic, 47, 361– 367.
Nicholson, M. C., Bowyer, R. T., and Kie, J. G., 1997, Habitat selection and survival of
mule deer: tradeoVs associated with migration. Journal of Mammalogy, 78, 483–504.
Nogami, M., 1995, Geomorphometric measures for digital elevation models. Zeitschrift fur
Geomorphologie, N.F. Supplement-Bd, 101, 53–67.
O’Brien, C. M., 1988, Characterization of muskox habitat in north-eastern Alaska. MS thesis,
University of Alaska Fairbanks, Fairbanks, AK, 114 pp.
Parker, G. R., 1978, The diets of muskoxen and Peary caribou on some islands in the
Canadian high Arctic. Canadian Wildlife Service, Ottawa, Ontario, Occasional
Paper No. 35.
Parker, G. R., and Ross, R. K., 1976, Summer habitat use by muskoxen (Ovibos Moschatus)
and Peary caribou (Rangifer tarandus pearyi) in the Canadian high Arctic.
Polarforschung , 46, 12–25.
Rapota, V., 1984, Feeding ecology of the Taimyr muskoxen. Biological Paper of the University
of Alaska, Special Report, No. 4, Fairbanks, AK, pp. 75–80.
Reynolds, P. E., 1984, Distribution, movements and herd dynamics of radio-collared mus-
koxen in the Arctic National Wildlife Refuge. Proceedings of the First International
Muskox Symposium. Biological Papers of the University of Alaska, Special Report.
No. 4, Fairbanks, Alaska, 63 pp.
Rhoads, B. L., and Thorn, C. E. (editors), 1996, The scienti c nature of geomorphology.
Proceedings of the 27th Binghamton Symposium on Geomorphology, 27–29 September
1996, University of Illinois at Urbana-Champaign , Illinois (Chichester: John Wiley and
Sons), 481 pp.
Robus, M., 1981, Muskox habitat and use patterns in north-eastern Alaska. MS thesis,
University of Alaska Fairbanks, Fairbanks, AK, 116 pp.
Schmidt, J. and Dikau, R., 1999, Extracting geomorphometric attributes and objects from
digital elevation models—semantics, methods, future needs. In GIS for Earth Surface
Systems —Analysis and Modeling of the Natural Environment, edited by R. Dikau and
H. Saurer (Stuttgart: Gebr Bornträger), pp. 153–173.
Segura Lopez, W., 1998, Application of the HEP methodology and use of GIS to identify
priority sites for the management of white-tailed deer. In GIS Methodologies for
Developing Conservation Strategies: T ropical forest recovery and wildlife management
in Costa Rica, edited by B. G. Savitsky and T. E. Lacher (New York: Columbia
University Press), 242 pp.
Smith , T., 1994, Muskox. ADF&G W ildlife Notebook Series. Revised by J. Coady and
R. Kacyon. http://www.state.ak.us/adfg/notebook/biggame/ muskoxen.htm.
Snyder, J. P., 1987, Map projections—a working manual. Geological Survey Professional
Paper 1395 (Washington: USGPO), 383 pp.
Tabachnick, B. G., and Fidell, L. S., 1989, Using Multivariate Statistics, 2nd edn (New York:
Harper Collins Publishers).
Thing, H., Klein, D. R., Jingfors, K., and Holt, S., 1987, Ecology of muskoxen in Jameson
Land, north-east Greenland. Holarctic Ecology, 10, 95–103.
4632 Sixth Circumpolar Symposium on Remote Sensing of Polar Environment
Wilson, K. J., 1992, Spatial scales of muskox resource selection in late winter. MS thesis,
University of Alaska Fairbanks, Fairbanks, AK, 90 pp.
Zar, J. H., 1996, Circular distributions: descriptive statistics. In Biostatistical Analysis, 3rd
edn (Upper Saddle River, NJ: Prentice Hall ), pp. 591–612.
Downloaded by [University of Cambridge] at 13:23 17 November 2017