Modeling Habitat Suitability for Moose (Alces alces) in Maine
Lauren Hendricks (11)
Environmental Studies Program
Colby College, Waterville, Maine
The moose is very important to Maine for many reasons, including economic and aesthetic value. It can also be very dangerous to vehicles. To effectively manage moose populations, it is important to know where these animals might be located. Using a GIS and information about moose habitat preferences, I created a model of suitable habitat for moose in Maine. This model, based primarily on land cover, is supported by data on the actual distribution of moose in Maine.
Habitat suitability index (HSI) models rely on field data and expert knowledge of the species in question to determine its requirements, which are ranked by relative importance. Data for variables such as land cover can then be classified according to rank and ultimately combined to create a final suitability model (Store and Jokimaki 2003).
Maine is well known for its moose, Alces alces. Moose have economic value, through hunting and moose spotting tours. Knowing where suitable moose habitat is located is crucial for population management. The limited geographic range of moose reflects the narrow habitat tolerance of this species. Variables such as snow depth, distance from roads and railroads, predation, elevation, slope and distance from human settlement have been identified as important (Messier 1991; Dettki et al. 2003; Maier et al. 2005). Two primary factors have been identified: cover and food availability (Allen et al. 1987; Hepinstall et al. 1996; Dussault et al. 2006). Cover protects moose from heat stress in the summer and snow in the winter (Allen et al. 1987). Moose have specific forage requirements, feeding primarily on the leafy parts of broadleaf trees in the summer and twigs and conifers in the winter (Allen et al. 1987; Dussault et al. 2006; IUCN Redlist 2008). Moose can often be found in recently disturbed areas in the process of regeneration (Allen et al. 1987; Dussault et al. 2006). Wetlands are also an important habitat feature, providing both relief from hot summer temperatures and important nutrients (Allen et al. 1987).
All analyses were performed with ESRIs ArcGIS 9.3. Land cover data for Maine were obtained from the Maine Office of GIS. Data were re-sampled to a cell size of 100m by 100m.
Six intermediate layers were created and combined with a weighted sum on a cell by cell basis (see Figure 1 for model; see Table 1 for weights). Cover, water and wetlands, and development were extracted from the original land cover layer. Layers representing distance to each feature were created using the Euclidean distance tool. Raw distances were reclassified to represent the suitability for moose. For distance to development, larger raw distances were more suitable. For distance to cover, water, and roads, smaller raw distances were more suitable. Each cover classification was evaluated for its cover suitability and feeding preference (Table 2). Food suitability values represent a combination of food preference ranking and distance from cover, as moose will not utilize a food source if it is too far from cover.
The output of the weighted sum was a raw HSI. Results were normalized to a scale from 0 to 1, with 0 representing habitat least suitable for moose and 1 representing the most suitable habitat.
Zonal statistics were calculated for the final HSI to assess the valididty of the model. Biophysical region data from the Maine Office of GIS were used to create analysis zones. Data from Philip Nyhus and Caitlin Dufraine (Colby College Environmental Studies Program) showing moose harvest zones were also used.
Figure 1. Graphic representation of model used to create final Habitat Suitability Index.
Figure 2. Raw input land cover layer.
Figure 3. Sample intermediate layers used to generate final HSI. Layers shown, from left to right, are Food Suitability, Distance to Cover and Distance to Development.
Results and Discussion
Figure 4. Habitat suitability index output. Areas of high HSI values are most suitable for moose, while areas of low HSI values are least suitable.
The model predicts that 74% of Maine has a moose HSI rating of 0.5 or greater, but only 5.7% of the state has a HSI rating of 0.75 or greater. However, when analyzed by zone, there is a considerable amount of variation in predicted HSI values.
When mean HSI values for each of the biophysical regions of Maine are considered, there is little variation. Mean HSI values range from 0.51, for the South Coastal Region and the Aroostook Lowlands, to 0.62 for the Eastern Interior. However, when the amount of land in each region that is suitable is considered, there is more variation. The Eastern Lowlands and Eastern Interior have the highest amounts of land with a HSI rating of 0.75 or higher, at 10.2% and 9.7%, respectively. The Central Mountains region has the smallest amount of highly suitable land (HSI at or above 0.75), at 0.2%. The biophysical regions of Maine were developed using environmental variables, such as temperature and vegetation. It is logical to expect that these differences will be reflected in the HSI predictions. This is reflected in the model.
Figure 5. Map showing mean HSI values calculated for the 15 biophysical regions of Maine.
Figure 6. Chart showing percent of area in each biophysical region with predicted HSI values equal to or greater than 0.5 and 0.75. Over half of each region has a predicted HSI of 0.5 or higher. However, substantially less of each region has a HSI value of 0.75 or higher.
When analyzed by hunting zone, mean HSI values range from 0.52 to 0.65. The habitat suitability predicted by this model roughly corresponds with the observed densities of moose in Maine. Generally, the southern part of the state is less suitable habitat for moose. The central and northern areas are more suitable. However, it is difficult to compare the HSI output with the observed density, as they measure very different things. The HSI output is only a suitability ranking for each cell, not a predicted density. To convert these suitability rankings to predicted density, it would be necessary to determine the maximum moose density. In Minnesota, this has been observed to be 2 moose/km2, but results from that study cannot be directly applied to another geographic area (Allen et al. 1987). Additionally, the observed densities are based on moose harvest and may not be an accurate representation of moose density. An example of this is Zone 0, which covers Baxter State Park. As hunting is not allowed in the park, measures based on moose harvest would suggest that this area is not suitable for moose. However, both the HSI predictions and anecdotal evidence would suggest otherwise.
Figure 7. Comparison between mean predicted HSI values and observed moose density, by hunting zone (numbered). Moose density, calculated from harvest records, have been normalized for easier comparison.
Finally, it is important to remember that the minimum stand size for a habitat type to be truly suitable is not known. Allen et al. (1987) suggested a that at least 8 stands of 2ha or more every 600ha are necessary to support moose populations in Minnesota. This model uses cell sizes of 1ha. Further research is needed to determine the actual habitat requirements of moose in Maine.
This project was created for ES212: Introduction to GIS and Remote Sensing. I would like to thank Philip Nyhus and Manny Gimond for their expert guidance and help in creating this model.
Allen, A.W., Jordan, P.A. and J.W. Terrell. 1987. Habitat Suitability Index Models: Moose, Lake Superior Region. U.S. Fish Wildl. Serv. Biol. Rep. 82 (10.155). 47 pp.
Dettki, H., Lofstrand, R. and L. Edenius. 2003. Modeling Habitat Suitability for Moose in Coastal Northern Sweden: Empirical vs. Process-Oriented Approaches. Ambio 32 (8): 549 556.
Dussault, C., Courtois, R. and J.P. Ouellet. 2006. A Habitat Suitability Index Model to Assess Moose Habitat Selection at Multiple Spatial Scales. Canadian Journal of Forest Research 35: 1097 1107.
Geist, V., Ferguson, M. and J. Rachlow. 2008. Alces americanus. In: IUCN 2010. IUCN Red List of Threatened Species. Version 2010.4. <www.iucnredlist.org>. Downloaded on 03 April 2011.
Hepinstall, J.A., Queen, L.P. and P.A. Jordan. 1996. Application of a Modified Habitat Suitability Index Model for Moose. Photogrammetric Engineering & Remote Sensing. 62 (11): 1281 1286.
Maier, J.A.K., Ver Hoef, J.M., McGuire, A.D., Bowyer, R.T., Saperstein, L. and H.A. Maier. 2005. Distribution and Density of Moose in Relation to Landscape Characteristics: Effects of Scale. Canadian Journal of Forest Research 35: 2233 2243.
Messier, F. 1991. The Significance of Limiting and Regulating Factors on the Demography of Moose and White-Tailed Deer. Journal of Animal Ecology 60 (2): 377 393.
Store, R. and J. Jokimaki. 2003. A GIS-based Multi-Scale Approach to Habitat Suitability Modeling. Ecological Modeling 169: 1 15.