F:\ALCES\Vol_38\PAGEMA~1\3806.PDF ALCES VOL. 38, 2002 KOITZSCH — MOOSE HABITAT SUITABILITY INDEX MODEL 8 9 APPLICATION OF A MOOSE HABITAT SUITABILITY INDEX MODEL TO VERMONT WILDLIFE MANAGEMENT UNITS Ky B. Koitzsch1 Wildlife and Fisheries Biology Program, University of Vermont, Burlington, VT 05405, USA ABSTRACT: Habitat Suitability Index (HSI) models translate existing knowledge of a species’ habitat requirements into quantitative measures of habitat quality. The HSI is a numerical index that represents the ability of a given habitat to provide life requisites for a species on a scale from 0 (unsuitable habitat) to 1 (optimal habitat). Habitat Suitability Index models are useful in natural resource planning for predicting the impacts of resource management practices on wildlife habitat. Many moose (Alces alces) HSI models require the labor-intensive collection of ground-level browse density data, which limits their applications for analyzing large landscapes required by moose. Some, however, have been developed utilizing remotely sensed data to analyze large study areas. I tested the usefulness of one of these models, created for the Lake Superior region, to 2 Wildlife Management Units (WMUs) in Vermont. Areas of study WMUs, “E1” and “I”, were 680 km2 and 729 km2, respectively. The model quantified 4 landscape-scale habitat variables representing annual cover types required by moose: percent area of regenerating forest, non-forested wetland, spruce/ fir forest, and deciduous/mixed forest. Model analyses were performed using a Geographic Information System (GIS). The model was useful in estimating relative habitat suitability of both WMUs, identifying within-WMU habitat variation, quantifying change in habitat suitability following a natural habitat-altering event, and predicting temporal change in moose habitat due to changes in forest management practices. The model revealed significant differences in habitat suitability of 0.64 for WMU E1 and 0.34 for WMU I. To determine within-WMU habitat variation, both WMUs were divided into 25-km2 evaluation units, which approximated the annual home range of moose in New England, and a HSI was calculated for each unit. Habitat suitability of 81 km2 of WMU I increased from 0.30 to 0.53 due to an increase in regenerating forest following heavy canopy damage from an ice storm in January 1998. A reduction in habitat suitability from 0.81 to 0.35 of Silvio O. Conte National Fish and Wildlife Refuge lands within WMU E1 was observed following a simulation in which all timber harvesting as a forest management practice was eliminated. Initial validation of this model for analyzing moose habitat at the WMU-scale is supported by correlation of HSI output to moose harvest data for WMU E1 25-km2 evaluation units and by comparison of HSI to estimated moose densities for both WMUs. ALCES VOL. 38: 89-107 (2002) Key words: Alces alces, Geographic Information System (GIS), Habitat Suitability Index (HSI) model, moose, Vermont, Wildlife Management Unit (WMU) In the last 40 years, Vermont has seen a considerable increase in the population size and distribution of eastern moose (Alces alces americana). The population, esti- mated at 20 animals in 1960, was thought to exceed 2,500 in 1998, and continues to grow at a predicted rate of 1.10 moose/year (Alexander 1993, Alexander et al. 1998). In the same period of time, moose distribu- tion has expanded from Vermont’s extreme northeast corner to the entire state. In 1993 the Vermont Department of Fish and Wild- 1Present address: P.O. Box 953, Waitsfield, VT 05673, USA MOOSE HABITAT SUITABILITY INDEX MODEL — KOITZSCH ALCES VOL. 38, 2002 9 0 life (VTDFW) initiated the first moose hunt in almost a century by issuing 30 harvest permits for one of Vermont’s 26 Wildlife Management Units (WMUs). By 1999, the moose hunt was expanded to 10 WMUs, representing approximately 51% of the state, and for the 2000 season 215 permits were issued. With an expanding moose popula- tion, public interest in non-consumptive uses of moose, such as viewing and photography, also have risen. In New England, moose habitat has been described in numerous studies (Cioffi 1981, Monthey 1984, Crossley 1985, Leptich and Gilbert 1989, Pruss and Pekins 1992, Thompson et al. 1995, Alexander et al. 1998, and K. Morris, Maine Department of Inland Fish and Wildlife 1999, unpublished data). Moose were found to require large habitats providing copious amounts of re- generating hardwood as their primary source of annual browse; young balsam fir (Abies balsamea) at higher elevations as a source of winter browse; mature spruce (Picea spp.) and balsam fir forests to escape the stressful effects of heat in summer and severe weather in winter (Renecker and Hudson 1986, 1990); and macrophyte-rich wetlands as a source of sodium in late spring and summer. Pletscher (1987), while studying nutrient budgets for white-tailed deer in north-central New Hampshire, dem- onstrated low sodium concentrations in ter- restrial vegetation and suggested that deer made up for this deficiency by utilizing natural salt licks, artificial salt licks along salted roads, and aquatic vegetation. As- suming sodium concentration in terrestrial vegetation throughout New England is low, as has been demonstrated for the Lake Superior region in work from Isle Royale (Jordan et al. 1973, Belovsky and Jordan 1981), moose probably make up for sodium deficiencies in the same manner. Wetland areas are also important for calving areas and as refugia from black bear (Ursus americana) predation, and insects. By occupying large home ranges, moose in New England are able to meet their sea- sonal life requisites, which are often spa- tially separated. Because of the presumed abundance and good quality of Vermont’s moose habi- tat, the VTDFW does not routinely inven- tory or monitor moose habitat. It does, however, collect physical measurements from legally harvested and incidentally killed moose as indicators of population health and habitat condition. While these meas- urements may indicate that a moose popu- lation is approaching carrying capacity or is in decline due to habitat deficiencies, they cannot specify which habitat component is deficient. The present physical condition of Vermont moose suggests that the herd is healthy and the habitat is productive (Alex- ander et al. 1998). However, with increas- ing demands on Vermont’s forests for rec- reation, forest products, conservation, and development, their ability to support moose could deteriorate. A tool that can inventory statewide moose habitat and predict habitat change due to change in forest management prac- tices and natural disturbances will aid in the conservation and management of valuable moose habitats. This tool should be com- patible with timber stand classification sys- tems developed for timber management since forestry practices have such a great impact on moose habitat throughout its east- ern range (Hurley 1986). The tool also should be simple, inexpensive to apply and capable of analyzing large habitats required by moose. An effective tool to meet all these demands is the moose Habitat Suit- ability Index (HSI) model (Allen et al. 1987). The concept of the HSI model began with the development of the U.S. Fish and Wildlife Service (USFWS) sponsored Habi- tat Evaluation Procedures (HEP) in 1976. The HEP quantified wildlife habitat based ALCES VOL. 38, 2002 KOITZSCH — MOOSE HABITAT SUITABILITY INDEX MODEL 9 1 on the Habitat Suitability Index (HSI) and total area of available habitat. They were created in response to the National Envi- ronmental Policy Act (NEPA) of 1969, which required that the environmental im- pacts on wildlife from any activity involving federal funding or a federal permit be de- scribed prior to implementation of the project. This act made it necessary for biologists to relate wildlife species to their habitat and to predict species response to habitat altera- tions (Thomas 1982). Between 1982 and 1989, the USFWS sponsored the development of over 160 HSI models for mammal, bird, reptile, amphib- ian, fish, and invertebrate species, and com- munities. These models translated existing knowledge of a species’ habitat require- ments into standard, quantitative measures of habitat quality on a scale from 0 (unsuit- able) to 1 (optimal). They are used to compare the ability of two or more study areas to provide habitat for a given species or to document habitat change over time within an individual study area. Habitat Suitability Index models also predict the consequences of proposed natural resource management on wildlife habitats and iden- tify suitable areas for development so nega- tive impacts on wildlife habitats can be minimized. Moose HSI Model II Two moose HSI models (Model I and Model II) were created by Allen et al. (1987) as part of the USFWS series and have served as the standard for more recent models (Allen et al. 1991, Courtois 1993, Palidwor et al. 1995, Hepinstall et al. 1996, Rempel et al. 1997, Romito et al. 1998, K. Morris, Maine Department of Inland Fish and Wildlife 1999, unpublished data). Model I and II were created for the evaluation of moose habitat in the Lake Superior region and were a product of a modeling workshop in Duluth, Minnesota in 1987. Model I was designed to evaluate the abundance and quality of growing season and dormant sea- son food and cover in study areas that approximate the size of annual habitats re- quired by moose (~600 ha). Intensive browse data collection is required for Model I. Model II was designed to rapidly evalu- ate and compare the ability of relatively large areas to provide annual habitat for moose using remotely sensed data. For this study, the usefulness of Model II for analyzing large tracts of habitat was tested by applying it to 2 of Vermont’s 26 WMUs, which is the geographic unit used by the VTDFW for moose management. Model II relates cover-type composition to moose habitat suitability and incorporates 4 cover-type variables that provide annual life requisites for moose. Model variables are: percent area of regenerating forests < 20 years old, used as a source of annual browse (variable 1); non-forested wetlands, used as a source of summer aquatic vegeta- tion (variable 2); spruce/fir forests > 20 years old, used as a source of summer and winter cover (suitable stands need > 50% spruce/fir canopy) (variable 3); and upland deciduous or mixed forests > 20 years old, used for both annual browse and cover (suitable stands must have > 25% canopy cover of trees, of which < 50% of the canopy must be spruce/fir) (variable 4) (Table 1). Model variables were based on research conducted by Peek et al. (1976) that described optimal moose habitat for northeast Minnesota. Twenty years was used as the cutoff age for regenerating forests because older trees are assumed to have little value as moose browse. Model II assumes that ideal year-round moose habitat requires the presence of all 4 habitat components and that model vari- ables are weighted equally. However, if any of the 4 habitat components is missing from the evaluation area, other than wetlands, suitability will equal zero regard- MOOSE HABITAT SUITABILITY INDEX MODEL — KOITZSCH ALCES VOL. 38, 2002 9 2 less of the percent area of the other cover types. It also assumes a positive correlation between species abundance and habitat quality (Allen et al. 1987). The degree of interspersion between food and cover is not addressed in Model II, however, in Ver- mont where logging operations are rela- tively small and scattered, and non-forested wetland and mature spruce/fir forests are distributed throughout the state, it is as- sumed that interspersion is adequate. Descriptions of preferred browse spe- cies and habitats of moose from studies in northern New Hampshire (Miller 1989, Pruss and Pekins 1992) and northern Maine (Cioffi 1981, Monthey 1984, Crossley 1985, Leptich and Gilbert 1989, Thompson et al. 1995) are similar to those described for the Lake Superior region (Allen et al. 1987) for which Model II was designed. Climate, which dictates annual habitat preference, is also similar between these two regions. A classification system developed by Wladimir Köppen shows that climate in the Great Lakes region and New England is similar based on vegetation types and annual monthly means of temperature and precipi- tation (Eichenlaub 1979). The U.S. Depart- ment of Commerce (1968) shows similari- ties between the two regions in annual minimum, maximum, and average daily tem- peratures, annual snowfall, mean annual numbers of days the minimum temperature was below 0oC, mean date for the last 0oC temperature in spring, and mean date for the first 0oC temperature in fall. Because moose habitat composition and climate in Vermont are similar to that of both northern Maine and northern New Hampshire, I con- sider it appropriate to assess moose habitat in Vermont using Model II. Specific objectives of this project were to: (1) generate GIS coverages converting vegetation data into 4 cover types upon which Model II is based; (2) apply Model II to 2 WMUs in Vermont to predict habitat suitability; (3) predict within-WMU habitat suitability variation; (4) demonstrate the Table 1. Description of model variables and the life requisites they provide, variable Suitability Indices (SI), and percent area of variables for optimum habitat suitability (Allen et al. 1987, Peek et al. 1976). Variable # Variable description Life requisites provided Variable % Area of (%area) by variable suitability variables for indices (SI) optimum habitat suitability 1 Regenerating Forest forage SI 1 40 – 50 < 20 years old 2 Non-forested aquatic forage, escape SI 2 5 – 10 Wetlands from insects and predation, thermoregulation 3 Spruce / Fir Forest winter and summer cover SI 3 5 – 15 > 20 years old 4 Upland Deciduous / forage and cover SI 4 35 – 55 Mixed Forest > 20 years old ALCES VOL. 38, 2002 KOITZSCH — MOOSE HABITAT SUITABILITY INDEX MODEL 9 3 – 1,000 m in elevation except to the south where the Nulhegan Basin lies. The basin is drained by the Nulhegan River, which flows eastward into the Connecticut River. The state’s most extensive bogs and softwood swamps are located in the basin, which averages 350 - 450 m in elevation. This WMU is characterized by a mosaic of young, intermediate, and mature stands of trees due to its logging history and diverse geography. Lowland areas are dominated by balsam fir, red spruce (Picea rubens), black spruce (P. mariana), poplar (Populus spp.), alder (Alnus spp.), and paper birch (Betula papyrifera). Intermediate eleva- tions contain primarily northern hardwood beech (Fagus grandifolia) / birch (Betula spp.) / maple (Acer spp.) forest. Sites above 800 m are predominately in red spruce and balsam fir stands. Deciduous, mixed, and coniferous forests cover approximately 54%, 25%, and 15% of the area, respec- tively (D. Williams, Spatial Analysis Labo- ratory, University of Vermont, unpublished data). This WMU has the greatest density of moose and the largest annual moose harvest of all WMUs in the state. Present estimate of moose density is 0.4 moose/km2 (C. Alexander, VTDFW, personal commu- nication). The second study area encompasses a 729 km2 portion of WMU I, and lies mostly within Addison County. This portion will be referred to as WMU I for the study. Wild- life Management Unit I is bordered by Route 17 to the north, Route 100 to the east, Route 73 to the south, and Route 116 to the west. It contains much of the northern half of the Green Mountain National Forest and strad- dles the 1,000 – 1,200 m Green Mountain spine. To the east, the Green Mountains drop steeply into the Mad and White River valleys, and to the west the mountains taper gradually into the Champlain Valley. Wild- life Management Unit I is dominated by mature northern hardwood forests at mid change in HSI of WMU I after a 1996 ice storm destroyed over half of the forest canopy at upper elevations; (5) predict change in HSI of WMU E1 after ownership passed from a commercial wood products company to a federal entity; and (6) support model validation for Vermont by correlating HSI values to population data from moose harvests. STUDY AREA Wildlife Management Unit E1 and a comparable-sized portion of WMU I were chosen as study areas (Fig. 1) because they both are very important moose habitats in Vermont, however, they vary greatly in vegetation composition, physiographic na- ture, and density of moose they support. Wildlife Management Unit E1 (680 km2) is located within Essex County in the north- east corner of Vermont. E1 is bordered by Canada to the north, the Connecticut River to the east, Route 105 to the south and Route 114 to the west. It is roughly circular in shape and surrounded by mountains 600 Fig. 1. Location of study areas WMU E1 and WMU I in Vermont. MOOSE HABITAT SUITABILITY INDEX MODEL — KOITZSCH ALCES VOL. 38, 2002 9 4 elevations, paper birch at high elevations, and pockets of spruce/fir on the highest peaks. Unit I has fewer non-forested wetlands than E1. Deciduous, mixed, and coniferous forests cover approximately 66%, 18%, and 10% of the area, respectively (D. Williams, Spatial Analysis Laboratory, Uni- versity of Vermont, unpublished data). Wildlife Management Unit I is one of 3 additional units opened to hunting in 1999 and has an estimated moose density of 0.1 moose/km2 (C. Alexander, VTDFW, per- sonal communication). Although WMU I is more densely populated by people than WMU E1, both areas represent 2 of the largest undeveloped tracts of land in the state. METHODS Moose HSI Model II and HSI Calcula- tion Model II: HSI = (SI 1 x SI 2 x SI 3 x SI 4 )1/4 In the model, HSI is the habitat suitabil- ity index for the study area or evaluation unit and SI 1 , SI 2 , SI 3 , and SI 4 are Suitability Index (SI) values for each of the 4 model variables. The HSI is the geometric mean of the 4 SI values. Percent areas of the 4 model variables are taken from variable cover type maps and plotted on Suitability Index graphs (Figs. 2 and 3) to determine SI values. Suitability index graphs were cre- ated for the model following a description of optimal habitat from Peek et al. (1976) (Allen et al. 1987). Percent area of vari- ables falling within optimal ranges will pro- duce a SI = 1.0 and percent areas less than or greater than optimal ranges will produce a SI < 1.0 (Figs. 2 and 3) (Allen et al. 1987). Data for each variable were compiled into an ArcView 3.1- based GIS (ESRI, Redlands, California, USA). Data were added to coverages by tablet digitizing with a CalComp Drawing Board II and WinTab digitizer software. Variable 1 — Percent area of regen- erating forest. The area of regenerating forest within each WMU was determined from numerous data sources. The Ver- mont Forest Resource Advisory Council (FRAC) quantified “heavy cuts” through- out Vermont between 1977-1996 and re- ported and mapped their findings (VDFPR 1996). These “heavy-cut” maps were used as base maps for variable 1. “Heavy cuts” were those visually determined from aerial flights or remotely sensed data to have been harvested below "C line". The C line represents the minimum amount of accept- able growing stock that makes a timber stand worth managing as defined by the Fig. 2. Suitability index curves showing relation- ship between percent area of regenerating forest and non-forested wetland variables and suitability index. Optimal coverage of regen- erating forest and non-forested wetland is 40- 50% and 5-10%, respectively (Allen et al. 1987). ALCES VOL. 38, 2002 KOITZSCH — MOOSE HABITAT SUITABILITY INDEX MODEL 9 5 U.S. Department of Agriculture silvicultural stocking guides (Long 1997). Stand stock- ing level is a function of basal area per acre (ft2) and the number of trees per acre. In New England, stands harvested below the C line can be expected to have large quan- tities of early successional regeneration from species such as aspen (Populus spp.), birch (Betula spp.), cherry (Prunus spp.), and maple (Acer spp.). Heavy cuts were easily discernible from maps because of their regu- lar shape and obvious contrast from adja- cent non-cut areas. Ancillary data used to complete the variable 1 map through 1999 f o r W M U E 1 i n c l u d e d 1 9 9 9 d i g i t a l orthophotography quadrangles (DOQ) and 1997-1999 Act 15 heavy-cut permits. For- est Service stand inventory data, which included all shelterwood, seed-tree, and clearcuts from 1977-March 1998, and 1997- 1999 Act 15 heavy-cut permits were used for WMU I. Color-infrared aerial photo- graphs, orthophotography, 1995 DOQs (WMU I), and ground-truthing were used to verify data. The final map displayed areas of regenerating forest as closed polygons. From these maps (Fig. 4), percent area of regenerating forests within the study areas was digitally queried. Percent area in “heavy cuts” < 23 years old (1977–1999) was used as an estimator of regenerating forests < 20 years old be- cause of data structure. I do not believe the addition of 2 years of data to variable 1 resulted in an overestimation of this vari- able, but rather made up for the small acre- ages of regenerating forests that were inad- vertently missed while analyzing data. Variable 2 — Percent area of non- forested wetlands. Non-forested wetlands included in this study followed suggested modifications to Model II by Adair et al. (1991) and wetland classifications from Cowardin et al. (1979). Adair et al. (1991) Fig. 3. Suitability index curves showing relation- ship between percent area of spruce/fir and deciduous/mixed forest variables and suit- ability index. Optimal coverage of spruce/fir forest and deciduous/mixed forest is 5-15% and 35-55%, respectively (Allen et al. 1987). Fig. 4. Variable 1 - regenerating forest. MOOSE HABITAT SUITABILITY INDEX MODEL — KOITZSCH ALCES VOL. 38, 2002 9 6 recommended that only wetlands with limnological conditions favoring macrophyte production should be included in the model. W e t l a n d s i n c l u d e d w e r e e m e r g e n t , unconsolidated bottom, rock bottom, and aquatic bed palustrine, scrub/shrub, dead forested, lower perennial riverine, littoral lacustrine, and beaver ponds. Mylar and digital National Wetland Inventory (NWI) maps were used to determine total area of these non-forested wetland types in each WMU. National Wetlands Inventory data were used because they are readily avail- able to the public, cover the entire United States, and are consistent with their wetland classifications. Suitable wetlands were first identified on NWI mylar maps and then labeled as such on digital maps. Where data were incomplete on digital maps, or had not yet been digitized, data were manually digi- tized from mylar maps. A coverage of suitable wetland polygons was then gener- ated and percent area digitally queried (Fig. 5). Variable 3 — Percent area of spruce/ fir forests. Variable area was taken from a vegetation grid map of New Hampshire and Vermont created for the USFWS Gap Analysis Project (D. Williams, Spatial Analy- sis Laboratory, University of Vermont, un- published data). Data used to create the map included 4 Landsat Thematic Mapper (TM) satellite scenes acquired for spring, summer, and fall of 1992 / 1993, a small portion of the May 1995 scene, and ancil- lary data. Resolution of the TM data was 30 m. The map was validated through inter- pretation of aerial videography linked to a Global Positioning System (GPS). The map accurately classified 85% of 907 GPS points examined (D. Williams, Spatial Analysis Laboratory, University of Vermont, unpub- lished data). Much of the error was asso- ciated with classification of mixed forest as either deciduous or coniferous forest. Of the 7 land-cover types classified in this mapping project, only deciduous, conifer- ous, and mixed-forest classifications were pertinent to this study. The others were omitted from analysis. Forests were classi- fied as coniferous or deciduous if either contributed > 65% of stand species, or mixed if neither contributed >65%. The “coniferous forest” classification from this map represented spruce/fir forest for my analysis. Since Model II requires that co- niferous or deciduous forests contribute > 50% of stand species (Allen et al. 1987), areas derived from this vegetation map may underestimate percent area of coniferous and hardwood forests and overestimate percent area of mixed forests. Using ArcView and ARC/INFO spatial analysis GIS software, the vegetation grid was clipped to the extent of the 2 study areas. Polygon coverages of regenerating forest (variable 1) and non-forested wetland (variable 2) were then erased from the vegetation coverage leaving coniferous, mixed, and hardwood forest cover-types (Fig. 6). Percent area in spruce/fir forest Fig. 5. Variable 2 - non-forested wetland. ALCES VOL. 38, 2002 KOITZSCH — MOOSE HABITAT SUITABILITY INDEX MODEL 9 7 was queried from the resulting coverage. Variable 4 — Percent area of de- ciduous/mixed forests. Percent area in upland deciduous/mixed forest was derived in the same manner as variable 3 (Fig. 7). Within-WMU HSI Variation To determine within-WMU HSI varia- tion, each WMU was divided into 25-km2 hexagonal evaluation units, and a HSI was determined for each unit as previously de- scribed. Twenty-five square kilometers was chosen as the evaluation unit size be- cause it approximates the annual home range of moose in New England (Crossley 1985, Miller 1989, Thompson et al. 1995, K. Mor- ris, Maine Department of Inland Fish and Wildlife 1999, unpublished data). Allen et al. (1987) and Schultz and Joyce (1992) recommended that size of the evaluation unit should approximate that of the animals’ home range for HSI analysis. A regular hexagonal pattern was chosen for evalua- tion unit shape because it is the best discon- tinuous sampling pattern for a spatial func- tion (Olea 1984). Evaluation units were assigned to 3 habitat categories based on HSI values: least suitable habitat (HSI = 0.0–0.31), suitable habitat (HSI = 0.32– 0.66), and most suitable habitat (HSI = 0.67–1.0) (Fig. 8). Effects of Habitat Alteration on HSI of Study Area The effect of a rapid increase in regen- erating forests on HSI was demonstrated by comparing HSI of 81 km2 of WMU I before and after heavy ice damaged much of the deciduous forest canopy above 1,000m in 1998 (Fig. 9). The Vermont Department of Forest Parks and Recreation (VDFPR) mapped statewide ice damage into 2 cat- egories labeled “heavy” and “moderate”. Forests were considered “heavily” dam- aged if > 50% of the forest canopy was damaged. This heavy damage to the forest canopy simulated the effect of harvest prac- tices, which stimulate regeneration in the Fig. 6. Variable 3 – spruce/fir forest. Fig. 7. Variable 4 – deciduous/mixed forest. ALCES VOL. 38, 2002 KOITZSCH — MOOSE HABITAT SUITABILITY INDEX MODEL 9 9 type classification from the vegetation grid map (D. Williams, Spatial Analysis Labora- tory, University of Vermont, unpublished data). Area of non-forested wetlands re- mained constant. Model Validation HSI model validation is a process that determines whether a model accurately pre- dicts habitat quality from the animal’s per- spective. Initial validation of Model II was accomplished by correlating October moose harvest density (moose/km2) to HSI for 25- km2 evaluation units within WMU E1. Wild- life Management Unit I was omitted from this analysis because it had been hunted for just 1 year and only 6 moose were har- vested. Moose harvest locations recorded for WMU E1 from 1993-1999 were used for analysis. Twenty-five kilometer square evaluation units with > 95% of their areas within WMU E1 were used in the analysis (n = 21). One hundred fifty-seven moose harvested were within the 21 evaluation units. The non-parametric Spearman rank correlation coefficient was used for analy- sis because distribution of evaluation unit HSI values was not normal. For this analy- sis, it was assumed that moose, during the fall hunting season, are occupying habitats with the greatest HSI values, and therefore moose harvest should be highest in units with the greatest HSI. A positive correla- tion between HSI and moose harvest would support validation of the model. Results Moose HSI Model II revealed large differences in habitat suitability between WMU E1 (HIS = 0.64) and WMU I (HIS = 0.34) based on differences in variable com- position (Table 2). Wildlife Management Unit E1 contained 17% regenerating forest (SI = 0.42), 2% non-forested wetland (SI = 0.48), 14% spruce/fir forest (SI = 1.0), and 63% deciduous/mixed forest (S = 0.82). Wildlife Management Unit I contained 4% regenerating forest (SI = 0.11), 1% non- forested wetland (SI = 0.32), 10% spruce/ fir forest (SI = 1.0), and 84% deciduous/ mixed forest (SI = 0.37). Wildlife Manage- ment Unit E1 contained greater amounts of regenerating forest and non-forested wetland, and lesser amounts of deciduous/ mixed forest than WMU I. Each WMU contained optimal amounts of spruce/fir forest, less than optimal amounts of regen- erating forest and non-forested wetland, and more than optimal amounts of decidu- ous forests. Within-WMU HSI variation was found in both study areas between 25-km2 evalu- ation units (Fig. 8). Evaluation units classi- fied as “most suitable” (HIS = 0.67-1.0) had the greatest area of regenerating forest, non-forested wetland, and optimal area in spruce/fir forest, while the “least suitable” (HIS = 0.0-0.31) units had a lesser abun- dance of regenerating forest and non- forested wetland, and an overabundance of deciduous/mixed forest. Wildlife Manage- ment Unit E1 had approximately 20% of its Table 2. Habitat Suitability Index values for WMU E1 and I (Variable 1 = regenerating forest, Variable 2 = non-forested wetland, Variable 3 = spruce/fir forest, Variable 4 = deciduous/mixed forest). Percent area of model variables / suitability index (SI) WMU Area (km2) Variable 1 Variable 2 Variable 3 Variable 4 WMU HSI E1 680 16.90 / 0.42 1.72 / 0.48 14.30 / 1.00 63.40 / 0.82 0.64 I 729 4.46 / 0.11 0.73 / 0.32 9.76 / 1.00 83.76 / 0.37 0.34 MOOSE HABITAT SUITABILITY INDEX MODEL — KOITZSCH ALCES VOL. 38, 2002 100 Within the 108 km2 of the Silvio O. Conte National Fish and Wildlife Refuge parcel, habitat suitability was shown to de- crease from 0.81 to 0.35 after 20 years of a simulated no-cut policy (Table 4). Over the 20-year simulation, a reduction of percent area of regenerating forest from 21% to 1% was offset by a corresponding 10% in- crease in both spruce/fir and deciduous/ mixed forests. Within the parcel, SI of regenerating forest decreased from 0.53 to 0.03 and SI of spruce/fir and deciduous/ mixed forest also declined. Within the entire WMU E1, HSI was predicted to decrease from 0.64 to 0.60. Correlation of HSI to moose harvest density for 25-km2 evaluation units within WMU E1 revealed a Spearman coefficient of r = 0.53 (P = 0.013) and an increasing trend in moose harvests with an increase in HSI of the evaluation unit (Fig. 10). A HSI of 0.64 for WMU E1 and 0.32 for WMU I compared proportionately to estimated moose density of 0.40 moose/km2 and 0.10 moose/km2, respectively (C. Alexander, VTDFW, personal communication). DISCUSSION Moose HSI Model II predicted differ- ences in habitat suitability between WMU Table 3. Change in HSI of 81 km2 of heavily damaged forest and WMU I following an ice storm in January 1998 (Variable 1 = regenerating forest, Variable 2 = non-forested wetland, Variable 3 = spruce/ fir forest, Variable 4 = deciduous/mixed forest). Percent area of model variables / suitability index (SI) Study Area Variable 1 Variable 2 Variable 3 Variable 4 HSI HID1 (B)2 81 2.95 / 0.07 0.00 / 0.20 26.60 / 0.86 70.10 / 0.67 0.30 HID1 (A)2 81 55.54 / 0.89 0.00 / 0.20 26.60 / 0.86 17.51 / 0.51 0.53 WMU I (B)2 729 4.46 / 0.11 0.73 / 0.32 9.48 / 1.00 78.92 / 0.47 0.36 WMU I (A)2 729 10.30 / 0.26 0.73 / 0.32 9.48 / 1.00 73.06 / 0.60 0.47 1HID = Heavy Ice Damage. 2B = before ice damage, A = after ice damage. area in “most suitable” habitat and 80% in “suitable” (HIS = 0.32-0.67) habitat. “Most suitable” habitats were located in the south- central portion of WMU E1 and approxi- mated the boundary of the Nulhegan Basin. One evaluation unit in the northeast corner also contained “most suitable” habitat. Wildlife Management Unit I lacked any units in the “most suitable” category but had approximately 50% of its area in “suitable” habitat. “Suitable” habitats were located in the western half of the study area and in the east-central portion. The model predicted that HSI increased from 0.30 to 0.53 in 81 km2 of WMU I, which was heavily damaged by an ice storm in 1998 (Fig. 9) (Table 3). An increase in percent area of regenerating forest from 3% to 56% caused an increase in SI from 0.07 to 0.9. A corresponding decrease in deciduous/mixed forest from 70% to 18% caused a decrease in SI from 0.67 to 0.51, but caused an increase in the deciduous/ mixed forest SI for the entire WMU. Due to ice storm damage, area in regenerating forest doubled in the entire WMU and caused an increase in SI from 0.11 to 0.26. Habitat Suitability Index of the entire WMU in- creased from 0.36 to 0.47 as a result of the storm. ALCES VOL. 38, 2002 KOITZSCH — MOOSE HABITAT SUITABILITY INDEX MODEL 101 E1 and WMU I that reflect differences in VTDFW moose densities of 0.4 moose/km2 and 0.1 moose/km2, respectively. Calcu- lated HSI values were 0.64 for WMU E1 and 0.34 for WMU I. This difference in HSI value was due to the greater percent- age of regenerating forest and non-forested wetland, and the lesser percentage of ma- ture deciduous/mixed forest in WMU E1 compared to WMU I. On a percentage basis, WMU E1 had 278% more regenerat- ing forest, 136% more non-forested wetland, and 24% less deciduous/mixed forest than WMU I. These data support the theory of Telfer (1978) and Collins and Helm (1997) who have indicated that the abundance of regenerating forest is often the most limit- ing factor to moose density. These data, which describe relative moose habitat suit- ability per WMU, will be useful to wildlife agencies for meeting their moose manage- ment objectives and for assisting members of the public in choosing their desired re- gional moose population levels in states where public opinion is considered for moose management. For instance, within Ver- mont’s “Moose Investigation Project State- ment”, the VTDFW identifies the need to determine relative habitat suitability per WMU using remotely sensed land-use databases, GIS, and HSI models. The VTDFW also solicits public opinion when making moose management decisions. Model II also identified variation in habi- tat suitability within WMUs. The abun- dance of “most suitable” habitat in WMU E1 was due to a concentration of regener- ating forest and non-forested wetland habi- tats in the Nulhegan Basin. “Suitable” habitat was found throughout the rest of the WMU where lesser amounts of these com- ponents exist, and “least suitable” habitat was found on the perimeter of the WMU associated with high concentrations of de- velopment and agriculture. WMU I con- tained no evaluation units with “most suit- able” habitat because the WMU as a whole was deficient in regenerating forest and non-forested wetlands. In WMU I “suit- able” evaluation units were found to the Table 4. Change in HSI of Silvio O. Conte National Fish and Wildlife Refuge lands and entire WMU E1 following a simulated 20 year no-cut policy (Variable 1 = regenerating forest, Variable 2 = non- forested wetland, Variable 3 = spruce/fir forest, Variable 4 = deciduous/mixed forest). Percent area of model variables / suitability index (SI) Study Area Area (km2) Variable 1 Variable 2 Variable 3 Variable 4 HSI Silvio Conte 108 21.20 / 0.53 4.09 / 0.85 19.00 / 0.95 51.63 / 1.00 0.81 Silvio Conte +20 108 1.00 / 0.03 4.09 / 0.85 29.60 / 0.82 62.24 / 0.84 0.35 WMU E1 680 16.90 / 0.42 1.72 / 0.48 11.32 / 1.00 63.40 / 0.82 0.64 WMU E1 +20 680 13.73 / 0.34 1.72 / 0.48 13.03 / 1.00 65.15 / 0.78 0.60 Fig. 10. Moose harvest density versus evalua- tion unit HSI for WMU E1. Spearman correla- tion: r = 0.53, P = 0.013, n = 21. MOOSE HABITAT SUITABILITY INDEX MODEL — KOITZSCH ALCES VOL. 38, 2002 102 west of the Green Mountain spine in asso- ciation with the highest concentrations of non-forested wetlands and regenerating forest (Figs. 4 and 5). Also, the east central portion of WMU I was heavily cut by the Forest Service in the 1980s and retains a “suitable” classification. Because of its steep mountainous terrain, the remainder of WMU I contains less standing water, is less accessible to logging, and has a “least suit- able” classification. Differences in habitat suitability of adjacent evaluation units in both WMUs can be attributed to differ- ences in topography, which determines the ability of the land to develop wetlands and dictates accessibility for logging. In WMU I, suitable habitats are concentrated in flat- ter areas to the west of the Green Mountain Ridge, and in WMU E1 habitat suitability of areas surrounding the Nulhegan Basin de- creases as the basin rises up to the sur- rounding mountains. The potential effects of ice damage on habitat suitability were illustrated through HSI analysis before and after an ice storm in January 1998 damaged the deciduous canopy of 81 km2 of WMU I (Fig. 9). Increased HSI from 0.30 to 0.53 (an in- crease of 77%) was due to a significant increase of regenerating forest and a corre- sponding decrease in deciduous/mixed for- est. This same increase in regenerating forest contributed to an increase in HSI of the entire WMU I from 0.36 to 0.47 (an increase of 31%). This effect of ice dam- age on HSI illustrates how natural events can rapidly affect habitat quality for moose. However, ice storm damage may not cause permanent canopy opening in the affected areas and the resulting increase in produc- tion of ground level browse may be short- lived. Studies are presently under way to determine the long-term effects of the 1998 storm on the forest canopy, and these should reveal how long damaged areas will con- tinue to provide regenerating browse. Other environmental factors such as heavy defo- liation by forest insects such as spruce budworm (Choristoneura fumiferana), which defoliated 56% of all spruce and balsam fir in Vermont in 1983 (R. Kelley, Vermont Department of Forests, Parks and Recreation, personal communication), tree disease, and wind-throw can have similar effects on moose habitat. The 108-km2 parcel purchased by the USFWS in 1999 from Champion Interna- tional Corporation contains much of the Nulhegan Basin. The Nulhegan Basin is arguably the most productive moose habitat in the state based on number of moose harvested, automobile-moose collisions, and sightings. Of 187 moose harvests located in WMU E1 from 1993-1997, 30% were within the approximate boundary of the Nulhegan Basin. Model II predicted that after 20 years of a no-cutting policy, HSI of the parcel would decrease from 0.81 to 0.35 (a reduction of 57%), due to a significant re- duction in regenerating forest and a corre- sponding increase in spruce/fir and decidu- ous/mixed forests. Following 20 years of maturation, this forest will still provide valu- able winter cover and non-forested wetland habitats, but would supply limited understory browse. If the suitability of the habitat is reduced by this amount, I project significant declines in the moose population and moose harvest in this area. Habitat Suitability Index output also was used to support validation of Model II. The ideal method to validate HSI models is to compare model output to known population numbers of target species within the study area. Since these data are usually unattain- able, indicators of species abundance are utilized (Clark and Lewis 1983, Laymon and Barrett 1986, Thomasma et al. 1991, Robel et al. 1993). Another common method used to validate models is to compare HSI output of known-use sites to random sites in order to test that the model can differentiate be- ALCES VOL. 38, 2002 KOITZSCH — MOOSE HABITAT SUITABILITY INDEX MODEL 103 tween the two (Allen et al. 1991, Brennan 1991, Apps and Kinley 1998). For this study, moose harvest density was used as an indicator of species abundance for initial validation. A Spearman correlation of HSI output to moose harvest density for 21- WMU E1 evaluation units (r = 0.53, P = 0.013) indicated the tendency for moose harvest density to increase with an increase in HSI (Fig. 10). A positive relationship between HSI and VTDFW estimated moose densities, for both WMUs, also supports validation of Model II. Additional research to further validate the use of Model II for Vermont includes locating heavy use-sites using GPS collars on moose and correlating these to HSI, and comparing HSI of heavy use-sites to that of random sites. A study to analyze the effects of road density on moose harvest, since hunter access to moose is most likely correlated to road access to hunting areas, also could be conducted. Results of HSI evaluation only should be used to predict the potential of habitat to support moose and not as a predictor of population density. Too many other factors exist, which can reduce moose abundance even when habitat is favorable, that are not addressed in the model. In Vermont, these include traffic and road density that influ- ence the number of moose/car collisions, deer density and infection rate of brain worm (Parelaphostrongylus tenuis), win- ter severity, illegal harvest, black bear pre- dation on moose calves, the number of hunting permits issued, and inter-specific competition for food. This study demonstrated that Model II is useful for analyzing large tracts of moose habitat. In view of the rate and scale at which humans alter the environment, it is important to look at habitat from a land- scape perspective. Too often we concern ourselves with ecological processes on a small scale, unaware of large changes oc- curring around us. With advances in higher resolution satellite imagery and the avail- ability of a greater selection of satellite data scenes, the analysis of large habitats will become simpler. Landsat 7 satellite data are currently available from the United States Geological Survey (USGS) at 30m resolu- tion. Landsat 7 records Enhanced The- matic Mapper Plus (ETM+) data in 7 spec- tral bands plus an eighth panchromatic band, combines synoptic coverage, high spatial resolution (15 m from the panchromatic band), a high spectral band range (450-2350 nm), increased spatial resolution of the ther- mal IR band (band 6), and 5% radiometric calibration (USGS 2000, unpublished data). Minimally processed data, known as Level 0-R data, are available at 475 U.S. dollars/ scene and Levels 1-R (radiometrically cor- rected) and 1-G (radiometrically and geo- metrically corrected) data are available at 600 U.S. dollars/scene. These costs are substantially lower than prices for current Landsat 5 data (USGS, http:edc.usgs.gov/ buspartners/satellite satellite-program.html). Because Landsat 7 records image data of the entire world every 16 days, users can choose data from up to 22 different dates of the year for a particular study area. Higher resolution imagery will enhance the ability to differentiate between regenerating and mature forest for HSI model applications for moose. Suggested Modifications to Model II Renecker and Hudson (1986, 1990) observed heat stress in moose, character- ized by an increase in metabolism and res- piration rate, when temperatures exceeded 14oC in summer and –5oC in winter. Such stress can result in depressed foraging ac- tivity and weight loss. Telfer (1984) ob- served that the southern limit of holarctic moose distribution corresponded closely to the 20oC July isotherm and that high tem- peratures that reduce reproductive perform- ance might restrict the southern expansion MOOSE HABITAT SUITABILITY INDEX MODEL — KOITZSCH ALCES VOL. 38, 2002 104 of moose range into areas with adequate habitat otherwise. To reduce effects of heat stress, moose seek shade in dense cover and wet areas to bed, thereby reduc- ing energy expenditure, respiration, and metabolism. From National Oceanic and Atmospheric Administration (NOAA) data (November 1998 – October 1999), I calcu- lated the number of days that temperatures exceeded 20oC between May and Septem- ber when moose are in summer pelage, and –5oC between October and April when moose are in winter pelage for both WMUs. Approximately 310 days in both Island Pond (WMU E1) and South Lincoln (WMU I) exceeded these limits. Temperatures at higher elevation beneath forest canopy where moose can escape heat would have been slightly lower, but these numbers still establish that moose are subjected to many days of heat stress. I believe there exists a threshold number of days above heat stress thresholds that moose simply cannot toler- ate, and that habitat selection in Vermont, and especially in southern New England, is temperature-dependent. The addition of a variable that quantifies the number of days, and the number of hours per day, tempera- tures exceed heat stress threshold likely would enhance accuracy of this model in New England and help predict the southern limit to moose range. Also, if trends in global warming continue, the management of heat sensitive species such as moose and caribou (Rangifer tarandus) will depend on determining the potential of traditional habitats to continue to provide for these animals. To validate this model with the ad- ditional variable for heat stress, moose habi- tat selection during times of heat stress should be monitored and correlated to HSI. The use of GPS collars would be essential in acquiring these data. A study of moose activity and metabolism during these times also would provide data on behavioral and physiological changes associated with heat stress. Collars fitted with an activity coun- ter and temperature sensor could gather these data. MANAGEMENT IMPLICATIONS The results of this study indicate that in both WMUs, percent area of deciduous/ mixed forest exceeds the amount for opti- mum habitat suitability, percent area of spruce/fir forest exists at optimal amounts, and the area of regenerating forests and non-forested wetlands exist in quantities well below that needed for optimum habitat suitability. Non-forested wetland and re- generating forest are therefore most limit- ing to moose habitat suitability. With the number of beavers in the state increasing due to a decline in trapping, and legislation protecting wetland habitats, it appears that the present quantity and quality of non- forested wetland habitats will improve. However, a decreasing trend in heavy cut- ting throughout much of the state since the mid-1980s could reduce the occurrence of regenerating forests. To achieve the goals of moose management in Vermont of main- taining moose populations at or above cur- rent densities, and to increase benefits as- sociated with moose such as viewing and hunting, the continued use of forestry prac- tices that create regenerating forests, and the continued protection of non-forested wetland habitat are desirable. Resource managers also should strive to maintain habitat quality within the “suitable” and “most suitable” habitats as identified by Model II. Since timber management has a great impact on moose habitat quality, (Courtois 1993, Palidwor et al. 1995, Rempel et al. 1997, Romito et al. 1998), forest and wildlife managers should strive to integrate the use of moose HSI models at the land- scape scale into timber management prac- tices. ALCES VOL. 38, 2002 KOITZSCH — MOOSE HABITAT SUITABILITY INDEX MODEL 105 ACKNOWLEDGEMENTS This study was funded by Remo Pizzagalli and the Pope & Young Conser- vation Fund. Thanks to Cedric Alexander for continued interest and guidance, and to the Vermont Department of Fish and Wild- life, Vermont Department of Forest, Parks & Recreation, National Wetland Inventory, and Vermont Mapping Program for provid- ing data. Thank you David Hirth, Jeffrey Hughes, and Ruth Mickey for guidance as committee members and to the entire Uni- versity of Vermont Spatial Analysis Lab staff for technical support. Special thanks to Lisa Osborn who offered her endless support throughout the project. REFERENCES ADAIR, W., P. A. JORDAN, and J. TILLMA. 1991. Aquatic forage ratings according to wetland type: modifications for the Lake Superior moose HSI. Alces 27:140- 149. ALEXANDER, C. E. 1993. The status and management of moose in Vermont. Alces 29:187-195. , P. FINK, L. E. GARLAND, and F. HAMMOND. 1998. Moose Management Plan for the State of Vermont 1998- 2007. Vermont Department of Fish and Wildlife, Waterbury, Vermont, USA. ALLEN, A. W., P. A. JORDAN, and J. W. TERRELL. 1987. Habitat suitability in- dex models: moose, Lake Superior re- gion. Biological Report 82(10.155), U.S. Fish and Wildlife Service, Fort Collins, Colorado, USA. , J. W. TERRELL, W. L. MANGUS, and E. L. LINDQUIST. 1991. Application and partial validation of a habitat model for moose in the Lake Superior region. Alces 27:50-64. APPS, C. D., and T. A. KINLEY. 1998. Development of a preliminary habitat assessment and planning tool for moun- tain caribou in southeast British Colum- bia. Rangifer 10:61-72. BELOVSKY, G.E., and P.A. JORDAN. 1981. Sodium dynamics and adaptations of a m o o s e p o p u l a t i o n . J o u r n a l o f Mammalogy 62:613-621. BRENNAN, L. A. 1991. Regional tests of a mountain quail habitat model. North- western Naturalist 72:100-108. CIOFFI, P. J. 1981. Winter cover and browse selection by moose in Maine. M.Sc. Thesis, University of Maine, Orono, USA. CLARK, J. D., and J. C. LEWIS. 1983. A validity test of a habitat suitability index model for clapper rail. Proceedings of the Annual Conference, Southeastern Association of Fish and Wildlife Agen- cies 37:95-102. COLLINS, W. B., and D. J. HELM. 1997. Moose, Alces alces, habitat relative to riparian succession in the boreal forest, Susitna River, Alaska. Canadian Field- Naturalist 111:567-574. , and C. C. SCHWARTZ. 1998. Log- ging in Alaska’s boreal forest: creation of grasslands or enhancement of moose habitat. Alces 34:355-374. COURTOIS, R. 1993. Description d’un indice de qualité d’habitat pour l’original (Alces alces) au Quebéc. Gouvernment du Quebéc, Ministére du Loisir, de la Chasse et de la Pêche, Direction générale de la resource faunique, Gestion intégrée des ressources, document tech- nique 93/1. COWARDIN, L. M., V. CARTER, F. C. GOLET, and E. T. LAROE. 1979. Classification of wetlands and deep water habitats of the United States. U.S. Fish and Wild- life Service FWS/OBS-79/31. CRÊTE, M. 1988. Forestry practices in Quebec and Ontario in relation to moose population dynamics. Forestry Chroni- cle 64:246-250. CROSSLEY, A. 1985. Summer pond use by moose in northern Maine. M.Sc. The- MOOSE HABITAT SUITABILITY INDEX MODEL — KOITZSCH ALCES VOL. 38, 2002 106 sis, University of Maine, Orono, USA. EICHENLAUB, V. L. 1979. Weather and climate of the Great Lakes region. University of Notre Dame Press, Notre Dame, Indiana, USA. HEPINSTALL, J. A., L. P. QUEEN, and P. A. JORDAN. 1996. Application of a modi- fied habitat suitability index model for moose. Photogrammetric Engineering and Remote Sensing 62:1281-1286. HURLEY, J. F. 1986. Summary: develop- ment, testing and application of wildlife- habitat models – the managers view- point. Pages 151-153 in J. Verner, M. Morrison, and C. J. Ralph, editors. Wild- life 2000: Modeling habitat relationships of terrestrial vertebrates. University of Wisconsin Press, Madison, Wisconsin, USA. JORDAN, P.A., D. B. BOTKIN, A. S. DOMINSKI, H. F. LOWENDORF, and G. E. BELOVSKY. 1973. Sodium as a critical nutrient for the moose of Isle Royale. Proceedings of the North American Moose Confer- ence and Workshop 9:13-42. LAYMON, A. L., and R. H. BARRETT. 1986. Developing and testing habitat-capabil- ity models: pitfalls and recommenda- tions. Pages 87-91 in J. Verner, M. Morrison, and C. J. Ralph, editors. Wild- life 2000: Modeling habitat relationships of terrestrial vertebrates. University of Wisconsin Press, Madison, Wisconsin, USA. LEPTICH, D. J., and J. R. GILBERT. 1989. Summer home range and habitat use by moose in northern Maine. Journal of Wildlife Management 53:880-885. LONG, S. 1997. Drawing the line in the stand-a guide to understanding the C line. Vermont Woodlands, Autumn Is- sue:23-26. MILLER, B. K. 1989. Seasonal movement patterns and habitat use of moose in northern New Hampshire. M.Sc. The- sis, University of New Hampshire, Dur- ham, New Hampshire, USA. MONTHEY, R. W. 1984. Effects of timber harvesting on ungulates in northern Maine. Journal of Wildlife Management 48:279-285. OLEA, R. A. 1984. Sample design optimiza- tion for spatial functions. Mathematical Geology 16:369-391. PALIDWOR, K. L., D. W. SCHINDLER, and B. R. HAGGLUND. 1995. Habitat suitability index models within the Manitoba Model Forest Region: Moose (Alces alces). Version 2. Manitoba Model Forest In- corporated, Pine Falls, Manitoba, Canada. PEEK, J. M., D. L. URICH, and R. J. MACKIE. 1976. Moose habitat selection and rela- tionships to forest management in north- eastern Minnesota. Wildlife Monographs 48. PLETSCHER, D. H. 1987. Nutrient budgets for white-tailed deer in New England with special reference to sodium. Jour- nal of Mammalogy 68:330-336. PRUSS, M. T., and P. J. PEKINS. 1992. Effects of moose foraging on browse availability in New Hampshire deer yards. Alces 28:123-136. REMPEL, R. S., P. C. ELKIE, A. R. RODGERS, and M. J. GLUCK. 1997. Timber-man- agement and natural-disturbance effects on moose habitat: landscape evaluation. Journal of Wildlife Management 61:517- 524. RENECKER, L. A., and R. J. HUDSON. 1986. Seasonal energy expenditure and ther- moregulatory response of moose. Ca- nadian Journal of Zoology 64:322-327. , and . 1990. Behavioral and thermoregulatory responses of moose to high ambient temperatures and insect harassment in aspen-domi- nated forests. Alces 26:66-72. ROBEL, R .J., L. B. FOX, and K. E. KEMP. 1993. Relationship between habitat suit- ability index values and ground counts ALCES VOL. 38, 2002 KOITZSCH — MOOSE HABITAT SUITABILITY INDEX MODEL 107 of beaver colonies in Kansas. Wildlife Society Bulletin 21:415-421. ROMITO, T., K. SMITH, B. BECK, M. TODD, R. BONAR, and R. QUINLAN. 1998. Moose winter habitat: Habitat suitability index model. Version 5. Weldwood of Canada, Hinton, Alberta, Canada. SCHULTZ, T. J., and L. A. JOYCE. 1992. A spatial application of a marten habitat model. Wildlife Society Bulletin 20:74- 83. TELFER, E. S. 1978. Cervid distribution, browse and snow cover in Alberta. Journal of Wildlife Management 42:352- 361. . 1984. Circumpolar distribution and habitat requirements of moose. Pages 145-182 in R. Olson, R. Hasting, and F. Geddes, editors. Northern Ecol- ogy and Resource Management. Uni- versity of Alberta Press, Edmonton, Alberta, Canada. THOMAS, J. W. 1982. Needs for and ap- proaches to wildlife habitat assessment. Transactions of the North American Wildlife and Natural Resources Con- ference 47:35-46. THOMASMA, L. E., T. D. DRUMMER, and R. O. PETERSON. 1991. Testing the habitat suitability index model for fisher. Wild- life Society Bulletin 19:291-297. THOMPSON, M. E., J. R. GILBERT, G. J. MATULA, and K. I. MORRIS. 1995. Sea- sonal habitat use by moose on managed forest lands in northern Maine. Alces 31:233-245. U.S. DEPARTMENT OF COMMERCE. 1968. Climatic Atlas of the United States. U.S. Government Printing Office, Washington, D.C., USA. (VDFPR) VERMONT DEPARTMENT OF FOR- ESTS, PARKS AND RECREATION. 1996. A report made by the Vermont Depart- ment of Forests, Parks and Recreation for the Assessment Working Group of the Forest Resource Advisory Council. Waterbury, Vermont, USA.