|
Resource selection habitat model for Northern flying
squirrels in the Black Hills, South Dakota.
|
|
|
|
|
| Abstract: | Northern flying squirrels (Glaucomys sabrinus) of the Black Hills National Forest (BHNF) of South Dakota represent a unique and isolated population, but little is known about the distribution and habitat use of this population. A resource selection function (RSF) habitat model was created for northern flying squirrels (G. sabrinus) throughout BHNF. Using methods from Manly et al. (2002) and Johnson et al. (2006), logistic regression was used to compare habitat variables at used habitat locations (radio-tracking and trapping locations) to a random sample of available habitat locations throughout the study area. Logistic regression coefficient estimates of significant variables were incorporated into a GIS raster layer to produce a map with RSF values for BHNF. The RSF values were transformed to a relative probability of habitat use ranging from 0 to 1. Independent validation data were used to determine model fit based on predictive performance of the RSF. Data used in the model determined that northern flying squirrels in BHNF used habitats with higher precipitation, closer distance to a stream, aspen (Populus tremuloides), northwest aspect, higher basal area of snags and a higher density of live trees and snags than randomly available habitats. The RSF map identifies possible high use areas of habitat by northern flying squirrels throughout BHNF and is useful for management purposes, as well as a baseline for future research and monitoring for an isolated population at the southern edge of their range. |
|
|
|
| Article Type: | Report |
| Subject: |
Flying squirrels
(Research) Habitat selection (Research) Home range (Research) |
| Authors: |
Hough, Melissa J. Dieter, Charles D. |
| Pub Date: | 10/01/2009 |
| Publication: | Name: The American Midland Naturalist Publisher: University of Notre Dame, Department of Biological Sciences Audience: Academic Format: Magazine/Journal Subject: Biological sciences; Earth sciences Copyright: COPYRIGHT 2009 University of Notre Dame, Department of Biological Sciences ISSN: 0003-0031 |
| Issue: | Date: Oct, 2009 Source Volume: 162 Source Issue: 2 |
| Topic: | Event Code: 310 Science & research |
| Geographic: | Geographic Scope: United States Geographic Code: 1USA United States |
| Accession Number: | 211067840 |
| Full Text: |
INTRODUCTION Data on habitat use, obtained through radiotelemetry, can be integrated into geographic information systems (GIS) to create species-specific resource selection function (RSF) habitat models (Gibson et al., 2004; Posillico et al., 2004). Manly et al. (2002) defined RSFs as any function that is proportional to the probability of use by an organism. Mapping the potential distribution based on model predictions identifies possible high use areas that aid resource managers in determining important habitats for management (Fielding and Bell, 1997). The ecology of northern flying squirrels is as varied as the forest communities in which it occurs (Smith, 2007). Across their range northern flying squirrels use habitats with features typical of mature forest, including large-diameter trees (Holloway, 2006; Lehmkuhl et al., 2006; Smith et al., 2004), large snags (Carey, 1995; Carey et al., 1999; Holloway and Malcolm, 2006; Smith et al., 2004), coarse woody debris (Carey et al., 1999; Smith et al., 2004) and truffle abundance (Lehmkuhl et al., 2006; Waters and Zabel, 1995). Northern flying squirrels typically inhabit mixed-coniferous-hardwood forests or dominant coniferous forests across their range (Weigl, 1978; Wells-Gosling and Heaney, 1984; Carey et al., 1999; Holloway, 2006) Movements, home range size and use of space appear to be influenced by availability of food resources (Menzel et al., 2006a), primarily truffles (Gomez, 2005), as well as by the availability of suitable den sites (Carey, 1995; Carey et al., 1997). Northern flying squirrel diet primarily consists of mycorrhizal fungus fruiting bodies, truffles, (North et al., 1997; Carey et al., 1999; Ransome and Sullivan, 1997, 9004; Weigl, 2007), but they also consume lichens, buds, berries, staminate cones and animal material (Weigl, 2007). For denning, northern flying squirrels use cavities in live hardwoods (Carey et al., 1997; Cotton and Parker, 2000; Hackett and Pagels, 2003; Holloway and Malcolm, 2007; Hough and Dieter, 2009b), cavities in snags (Maser et al., 1981; Bakker and Hastings, 2002; Meyer et al., 2005; Hough and Dieter, 2009b) and conifers for external dray nests (Mowrey and Zasada; 1984; Carey et al., 1997; Cotton and Parker, 2000; Hough and Dieter, 2009b). In the Black Hills National Forest (BHNF), there is an isolated population of northern flying squirrel, Glaucomys sabrinus bangsi (Rhoads) (King, 1951; Wells-Gosling and Heaney, 1984) that is restricted to this region of western South Dakota. Due to its rarity, isolation and population risk because of restricted range, northern flying squirrels within the BHNF are considered a Forest Service Species of Local Concern (USDA Forest Service, 2005) and a Species of Special Concern (S2) by the South Dakota Natural Heritage Program (South Dakota Department of Game Fish and Parks, 2006). The objectives of this study were to determine probability of habitat selection throughout the Black Hills and potential distribution. Logistic regression was used to compare habitat variables at radio-tracking and trapping locations (used habitats) to characteristics within the Black Hills study area (available habitats). Logistic regression is commonly used in wildlife studies to predict habitat use throughout a study area (Fielding and Haworth, 1995; Mladenoff et al., 1999). The GIS map layer produced from the model will help wildlife managers determine important areas of northern flying squirrel habitats, as well as provide a baseline for future research. MATERIALS AND METHODS Study area.--This study was conducted in the BHNF, located in western South Dakota (UTM 13N 574719-641489 E, 4809979-4932866 N) (Fig. 1). The BHNF is a unique ecosystem that consists of forest surrounded by Great Plains grasslands (Froiland, 1990). The ponderosa pine dominated hills extend 900-1200 m above the surrounding Great Plains prairie. The Black Hills were formed by mountain uplift, extend 200 km north to south and 100 km east to west and encompass 486,000 ha. Topography varies from steep ridges, rock outcrops, canyonlands and gulches to upland prairie, rolling hills and tablelands. Their elevation range is approximately 1200 to 2207 m, with the forested region extending to 2102 m (Froiland, 1990). The southern Black Hills, has a warmer (9.3 C) and drier (45-51 cm/y) annual climate than the northern portion of the range (7.2 C and 61-66 cm/y; Shepperd and Battaglia, 2002). Ponderosa pine is found throughout the BHNF and is the most abundant tree species, dominating 83% of the landscape (USDA Forest Service, 2005). In the central to southern hills quaking aspen (Populus tremuloides) and paper birch (Betula papyrifera) are interspersed with pine in the bottomlands and along water sources, spruce occurs sporadically and is commonly found along streams, and there is little to no understory. In the northern hills white spruce is more abundant, comprising 2% of the vegetation (USDA Forest Service, 2005). The northern hills has an understory component, primarily bur oak (Quercus macrocarpa), but may also include American elm (Ulmus americana), green ash (Fraxinus pennsylvanica), box elder (Acer negundo) and eastern hop-horn-beam (Ostrya virginiana) (Hoffman and Alexander, 1987). Quaking aspen and paper birch dominate moister environments, particularly in disturbed areas, of the central to northern BHNF (Hoffman and Alexander, 1987). [FIGURE 1 OMITTED] Habitat model.--To create a RSF habitat model for northern flying squirrels, we downloaded digital databases from the Black Hills National Forest Service website (USDA Forest Service, 2006). The databases included a vegetation polygon layer, digital line graph of streams and digital elevation model (10 m resolution). Precipitation databases for South Dakota and Wyoming were downloaded from the Water and Climate Center of the Natural Resources Conservation Service website (USDA Natural Resources Conservation Service, 2007). The vegetation layer provided vegetation composition and structure variables along with landscape features for a site. Variables taken from the vegetation layer were habitat types (grass-shrub, aspen-birch, bur oak, pine and spruce), structural stage (1-2, 3A, 3B, 3C, 4A, 4B, 4C), aspect, # live trees/ha, basal area ([m.sup.2]) of live trees [greater than or equal to] 12.7 cm dbh, basal area ([m.sup.2]) of dead trees [greater than or equal to] 12.7 cm dbh, # dead and down trees [greater than or equal to] 12.7 cm dbh and # snags [greater than or equal to] 12.7 cm dbh. Descriptions of variables used in the model are provided in Table 1. Structural stage 1-2 is grass-shrub habitat, which was not used by northern flying squirrels; therefore, this variable was not included in the model. Spatial analyst extension (Environmental Systems Research Institute, 2004) was used in ArcGIS 9.1 to convert the vector polygon layer to a raster dataset with 30 m resolution. Spatial analyst was also used to create a 30-m raster dataset depicting distance from streams. All layers were projected to UTM World Geodetic System 1984 coordinate system. The model was produced for the contiguous national forest land within the Black Hills of South Dakota and Wyoming. JMPIN 4.0 (SAS Institute Inc., Cary, North Carolina) was used to create the model and the model was integrated into ArcGIS 9.1 to produce the RSF habitat suitability map. Where not already stated rejection limit was set to [alpha] = 0.05. Where applicable, Shapiro-Wilk's test was used to test variables for normality. Nonparametric tests were performed because assumptions of normality and heterogeneity were not always met. There were 1503 trapping (n = 223), tracking (n = 1255) and observational (n = 25) locations for northern flying squirrels used in producing and validating the RSF habitat model. The locations were collected from 2005-2007 and data were pooled across all study years. Three hundred (20%) random points were removed from the sample of used habitat locations for use in validating the model. A random sample of 1203 points, with replacement, was taken from the study area to describe the available habitat. Spatial analyst was used to reclassify categorical variables (aspect, tree species and structural stage class) into indicator variables and a separate layer was created for each indicator variable. If the variable was present then the cell was coded '1' and if absent the cell was coded '0'. For example, in the north aspect layer all cells with a north aspect will have a '1' and all other cells will have a '0'. This process was continued for the other seven aspects, as well as dominant tree species and structural stage class. The used habitat locations and available habitat locations point layers for building the model were intersected with variables considered for inclusion in the RSF model (Table 1) using the intersect point tool in Hawth's Analysis Tools (Beyer, 2007). The resulting attribute tables contained values of each variable at each point. The two attribute tables (used habitat locations and available habitat locations for model building) were exported to JMPIN and combined. In the table with combined data a column was added and all used habitat locations were coded '1', while the available habitat locations were coded '0'. Within the Black Hills, spruce only dominates approximately 2% of the landscape and northern flying squirrel locations from 2005-2007 did not fall within areas dominated by spruce. Therefore, spruce was not considered as a variable in the logistic regression model. However, spruce has been found to be significant in other northern flying squirrel studies across their range (Connor, 1960; Jackson, 1961; Musser, 1961; Weigl and Osgood, 1974; Mowrey and Zasada, 1984; Urban, 1988; Payne et al., 1989; Loeb et al., 2000; Odom et al., 2001; Hackett and Pagels, 2003; Smith and Nichols, 2003; Ford et al., 2004; Menzel et al., 2006b), therefore, a value of one was added to all cells containing spruce in the RSF model to give weight to spruce habitat. The model building data were subjected to a series of tests to determine which variables would be incorporated into the final model. Fisher's exact test was used to compare proportions of used habitat locations and available habitat locations categorical (indicator) variables. Wilcoxon rank sums test was used to compare means of continuous variables between used habitat locations (trapping and tracking) and available habitat locations (random points throughout the study area). Any categorical or continuous variables found to be significant at the 10% level were retained for further statistical analysis. A correlation matrix was established with retained variables. When two variables were highly correlated (r [greater than or equal to] 0.70), the one deemed to be less significant biologically was removed (Manly et al., 2002). Logistic regression was used to determine coefficients of remaining variables with the dependent variables being used habitats (1) or available habitats (0). Logistic regression has become the standard method of analysis for discrete response outcome variables (Hosmer and Lemeshow, 2000), including small mammal studies (Carey et al., 1999; Menzel et al., 2006a). In addition, the assumptions for the independent variables are more lenient and continuous and categorical variables can both be included in the model (Hosmer and Lemeshow, 2000). The final RSF model was determined by stepwise backward elimination. In the logistic regression model, likelihood ratio chi square test was used to determine which variables were significant in determining habitat use by northern flying squirrels at the 10% level (Manly et al., 2002). The variable with the highest likelihood ratio chi square p-value was removed and logistic regression was run on the remaining variables. This process continued until all variables in the model had a regression coefficient that was significant at the 10% level. The regression coefficients were included in the logistic regression equation w * (x)=exp ([[beta].sub.0] + [[beta].sub.1] [x.sub.1] + ... + [[beta].sub.p] [x.sub.p]) where [[beta].sub.0] is the estimated constant intercept coefficient, [[beta].sub.1 ... p] are the significant regression estimated coefficients and [x.sub.1...p] are the corresponding variables (Manly et al., 2002). RSFs in use-availability studies often take on an exponential form of function (Manly et al., 2002; Johnson et al., 2006). Contamination can occur where used habitat locations are included in the sample of available habitat locations. For use-availability studies, the sampling probabilities of used habitat locations (Pu) and available habitat locations ([P.sub.a]) are incorporated into the logistic regression model to modify the intercept coefficient ([[beta].sub.0]) ([[beta].sub.0] + [log.sub.e][(1 - [P.sub.a])[P.sub.u]/[P.sub.a]]) (Manly et al., 2002). If the sampling probabilities of used habitat locations and available habitat locations are not known, then the probabilities along with the constant intercept coefficient are removed from the logistic regression equation, which we did in this study. A RSF is estimated using the estimated coefficients and significant variables from the logistic regression model with the following formula w(x) = exp ([[beta].sub.1][x.sub.1] + ... + [[beta].sub.p] [x.sub.p]) where [[beta].sub.1...p] are significant regression coefficients and [x.sub.1...p] are the corresponding significant variables (Manly et al., 2002) The accepted model was then incorporated into ArcGIS 9.1 to produce a RSF value for each 30 m resolution cell in the study area. The resulting RSF values were used to predict relative probability of selection throughout the study area. RSF values are not an absolute probability, therefore, a high probability of use does not define optimal habitat (Manly et al., 2002), because these values are based on used habitat locations and available habitat locations, where sampling probabilities are not known. The raw RSF scores are transformed to scale the predicted values (w) between 0 and 1 using the following formula [??] = w(x)/1 + w(x) where habitat selection probability increases as the transformed RSF value ([??]) approaches 1. The RSF probabilities from the resulting map were then classified into 10 quantiles (ordinal bins) which represent categories of increasing habitat selection (Johnson et al., 2006). Where habitats with RSF values falling within bin 1 have the lowest habitat selection or lowest probability of use and habitats with RSF values falling within bin 10 have the highest habitat selection or highest probability of use. Statistical methods normally used to test validity of a logistic regression model (ROC, Kappa and confusion matrix) are not appropriate for use-availability research designs due to contamination (Boyce et al., 2002). The independent validation data were used to validate the model according to methods suggested by Johnson et al. (2006). The utilization U([x.sub.i]) value of each of the bins was determined using the formula U([x.sub.i]) = w([x.sub.i])A([x.sub.i]) / [summation over (j) w([x.sub.j])A([x.sub.j]) where w([x.sub.i]) was the midpoint value of the RSF bin (i) and A([x.sub.i]) was the area of bin i (Boyce and McDonald, 1999). The midpoint value w([x.sub.i]) for each bin was the midpoint of the RSF probability range for that bin and the area A([x.sub.i]) was total number of pixels for each bin (Johnson et al., 2006). The number of validation observation points within each bin were counted. The expected number of validation observations within each bin ([N.sub.i]) was estimated using [N.sub.i] = N * U([x.sub.i]) where N was the total number validation points and U([x.sub.i]) was the utilization function from above (Johnson et al., 2006). The proportion of expected observations was compared to the proportion of observed validation points using linear regression. The regression line was compared to a slope of zero (use = availability). [R.sup.2] was also calculated from the linear regression model comparing proportions of expected and observed points, with a high value indicating an acceptable fit of the model. Spearman rank correlation was conducted between observed and expected frequencies to determine degree of correlation. Chi-square goodness-of-fit was calculated and a nonsignificant p-value (P [greater than or equal to] 0.05) indicates an acceptable model. Lastly, Pearson chi-square test was used to determine which bins have an expected value different from observed value. RESULTS Habitats used by northern flying squirrels were at higher elevations, higher precipitation, closer to streams, had a higher basal area of live trees > 12.7 cm dbh and a higher basal area of dead trees > 12.7 cm dbh, contained more live trees per hectare, more dead down trees per hectare and more snags per hectare than randomly available habitats (Table 2). Used habitats were comprised of less grass-shrub habitat, more aspen-birch, less structural classes 3A and 3B and more structural class 4A and 4B than randomly available habitats (Table 3). Structural class 4C was marginal (P = 0.1008), therefore, 4C was considered for the logistic model (Table 3). Used habitats were also more likely to be on east and northwest slopes, but less likely to be on southeast and west slopes than randomly available habitats (Table 3). Two variables found to be significant were removed because they were highly correlated (r [greater than or equal to] 0.70) with another significant variable. Elevation and precipitation were correlated (r = 0.75). Elevation was removed because moisture levels can influence fungus abundance and distribution and therefore, northern flying squirrel abundance and distribution. The variable estimating the number of dead and down trees (DeadDown) was correlated with dead basal area (DeadBA) (r = 0.81). The variable estimating dead and down trees was removed because the dead basal area variable includes down mortality and down sound snags. The final estimated RSF was [??] = exp {0.0983(Precip) - 0.0009 (DistStrm) - 0.5634(GRA) + 1.3875(TAA) - 2.0695 (3A) - 1.8418(3B) - 0.9112 (WE) + 0.3129(NW) + 0.0005 (LiveHa) + 0.0387 (DeadBA) + 0.0372 (TotSnags) + 1 (TWS)} The final RSF indicated that habitats used by northern flying squirrels had higher precipitation and were closer to streams than randomly available habitats. Northern flying squirrels used aspen-birch habitat, but not grass-shrub habitat (Table 4). Structural classes 3A and 3B were used less than expected based on availability (Table 4). Northwest aspect was used more than expected, while west aspect was used less than expected based on availability (Table 4). Northern flying squirrels used habitats with a higher density of live trees and snags, as well as a higher basal area of dead trees > 12.7 cm dbh (Table 4) than randomly available habitat. The linear regression model comparing the proportion of expected to observed frequencies for each bin indicated a good model fit ([R.sup.2] = 0.85) (Fig. 1). The regression line was significantly different from zero (use [not equal to] availability) (P = 0.0002). Spearman rank correlation value was 0.9423 (P < 0.001). The goodness of fit chi-square test indicated expected and observed values were significantly different ([chi square] = 48.85, df = 9, P < 0.0001). Individual bin Pearson chi-square tests found that five of the ten bins had observed frequencies similar to expected frequencies (Table 5). Even though the goodness of fit chi-square test was significant, all other tests indicate a good model performance between observed and expected frequencies. Therefore, the model was accepted based on validation data. Our decision to accept the current model is based on two anomalies (Jasper Burn and Black Elk Wilderness Area) within the BHNF vegetation layer, which may have affected the validation data and are furthered explained in the discussion. A 30 m resolution map of transformed RSF values ranging from 0-1 probability of habitat selection is depicted in Figure 2. The RSF habitat model was compared to a GIS habitat layer for the Black Hills and areas of low, medium and high probability of use were delineated (Fig. 3). DISCUSSION Habitat suitability models are an important management tool for nocturnal and elusive species that are difficult to study. The RSF habitat model for northern flying squirrels in BHNF does not give an indication of presence or absence based on habitat characteristics at a site, but instead provides a management tool to identify areas of habitat suitability based on known locations. Trapping results suggest northern flying squirrels are distributed throughout the Black Hills at varying densities across the landscape (Hough, 2008). These densities are presumably related to habitat suitability, which is depicted in the habitat model. One caution in using RSF habitat models is that maps that predict a high probability of use for certain areas do not define optimal habitat (Manly et al., 2002). However, Rosenberg and Anthony (1992) suggested that because of the distribution of northern flying squirrels throughout their range, this species is highly adaptable and is a habitat generalist. Across their range, there is no single variable or group of variables that explained variation in abundance (Smith et al., 2004). Abundance has been found to be correlated with a range of variables, including density of large trees and snags, shrub and canopy cover, prevalence of old-forest features (e.g., coarse woody debris) and abundance of hypogeous mycorrhizal fungi (Smith, 2007). [FIGURE 2 OMITTED] We believe the population of northern flying squirrels in the Black Hills is stable, because northern flying squirrels were caught throughout the Black Hills and were found in less than optimal habitats (edge of harvested and burned areas) (Hough, 2008). Also, northern flying squirrels are breeding successfully and rearing young. During all three study years juveniles were captured while trapping and observed while radio-tracking. Northern flying squirrels used aspen and birch habitat, however, grass-shrub habitat does not provide essential resources. Aspen-birch (2% of the landscape) was found to be important denning habitat for northern flying squirrels in BHNF (Hough, 2008). Aspen and birch trees are important for den sites, because cavities are readily created by primary cavity users in live trees and snags (Carey and Gill, 1983; Shepperd and Battaglia, 2002; Menzel et al., 2004; Holloway, 2006). Aspen and birch do not provide optimal foraging habitats because hypogeous fungi are associated with conifers (Weigl, 1978; Maser et al., 1985; U.S. Fish and Wildlife Service, 1990, Loeb et al., 2000). Menzel et al. (2006b) found that several of the squirrels they tracked denned during the day in hardwood patches and foraged during the evening in conifers patches, in mixed conifer-hardwood patches. The same pattern was found in BHNF, with squirrels tracked to bottomlands of aspen and birch during the day and ponderosa pine dominated slopes and ridges during the night (Hough, 2008). Grass-shrub was not selected due to the species' dependence on trees. Grass-shrub habitat would not provide optimal foraging habitat for northern flying squirrels, because this habitat does not provide hosts for fungal growth or other favored food sources, in addition there is a lack of nest sites and protection from predators. Studies in the Appalachians have found spruce to be the primary predictor in northern flying squirrel presence (Weigl and Osgood, 1974; Weigl, 1978; Urban, 1988; Payne et al., 1989; Loeb et al., 2000; Odom et al., 2001; Hackett and Pagels, 2003; Ford et al., 2004; Menzel et al., 2006a, 2006b). Northern flying squirrels are also found in spruce forests across other parts of their range including, southeastern New York (Connor, 1960), Great Lakes region (Jackson, 1961), Alaska (Mowrey and Zasada, 1984; Smith and Nichols, 2003) and Utah (Musser, 1961). Across northern flying squirrel distribution range, spruce is an important component of habitat use because they are used for dray nests (Menzel et al., 2004; Mowrey and Zasada, 1984; Hough, 2008). Spruce seeds also provide a potential food source (Smith, 2007). Also, hypogeous fungi are associated with spruce (Luoma et al., 1991; Loeb et al., 2000; Pyare and Longland, 2001). Future research should determine northern flying squirrel abundance in spruce dominated areas of the Black Hills which were inadvertently neglected in this study, due to the small extent of spruce dominated habitat across the landscape. [FIGURE 3 OMITTED] Northern flying squirrel abundance increases with forest complexity (Weigl et al., 1992; Carey et al., 1999; Carey, 2001). The model results suggest that not only was density of live trees important as a predictor of northern flying squirrel presence, but so was the basal area of dead trees and total number of snags. Live trees provide food sources, such as mycorrhizal fungi and seeds, and canopy cover for protection from predators. In the Pacific Northwest, habitat features that consistently explain significant variation in northern flying squirrel abundance are forest characteristics associated with snags and down, dead wood or decay (Carey et al., 1999). Cavities in snags provide den sites and fungal diversity has been found to be positively correlated with the abundance of coarse woody debris (Carey et al., 1999). In Oregon, the highest density of northern flying squirrels occurred in stands with the highest coarse woody debris (Carey et al., 1999) and southeastern Alaska (Smith et al., 2004). Northern flying squirrels in BHNF also used habitats with larger trees, but not canopy cover (structural stages 4A, 4B and 4C). Large trees serve as den sites and an indication of food availability. Northern flying squirrels select larger trees for dray nests (Hough, 2008) and larger trees are presumed to indicate a mature stand, which has an established mycorrhizal community and more seed production (Fisher and Wilkinson, 2005). Also, mycorrhizal fungus is often clumped and short-lived, therefore relatively dense canopies, large tall trees, and open midstories are important for squirrels to move through their home range efficiently and safely (Vernes, 2001). Northern flying squirrels used habitats with higher precipitation and closer distances to streams compared to available habitat locations throughout the study area. Hypogeous fungi fruiting bodies require moist, nutrient-rich and cool environments to grow (Pyare and Longland, 2001). Northern flying squirrel population distribution may be heterogeneous across landscapes because of the heterogeneous spatial distribution of truffles (Pyare and Longland, 2002). In drier sites in the west, squirrels appear to be associated with riparian areas with cooler and wetter conditions (Meyer et al., 2005, 2007), because fungus composition and abundance may be related to moisture levels (Rillig et al., 2002). Southern BHNF is classified as Dry Coniferous Forest (Marriott et al., 1999); therefore, squirrels in the southern hills may be limited to riparian areas where fungus grows. The northern BHNF, is a Mesic Coniferous Forest (Marriott et al., 1999), which typically support higher population densities of northern flying squirrels than xeric forests (Lehmkuhl et al., 2006; Hough, 2008). Microhabitat use in mesic forests is not restricted to riparian areas (Carey et al., 1999). An increase in moisture results in an increase in fungal community diversity (Carey and Johnson, 1995) and abundance (Luoma et al., 1991). This in turn, results in an increase in northern flying squirrel abundance (North et al., 1997; Ransome and Sullivan, 1997; Carey et al., 1999; Pyare and Longland, 2001). One problem with the model is the degree of habitat selection in the Black Elk Wilderness Area [located just north of Custer State Park (Fig. 2 and 3)], which is reserved as a wildlife refuge where timber harvest is restricted. The Black Elk Wilderness area covers about 4080 ha and is a category 1 management area. The area is managed to protect and perpetuate natural ecological processes with little human interference (USDA Forest Service, 1996). Late successional vegetation occurs in the area. The area consists of a diversity of species composition and structure. We predicted the Black Elk Wilderness area to have a high probability of use because in the Pacific Northwest northern flying squirrel abundance was higher in old-growth and complex young forests than managed coniferous stands (Carey et al., 1992; Witt, 1992; Carey, 1995; Waters and Zabel, 1995). The model suggests that this area is not as highly selected as we predicted. We believe three of the habitat variables (number of live trees, dead basal area and number of snags) found to be significant in the model are not accurate for the Black Elk Wilderness area in the GIS database. All three variables had low values; therefore, indicating this area was not suitable as northern flying squirrel habitat. However, during 2005, which had a low trapping success overall, we trapped on the edge of the wilderness area. We captured five adult flying squirrels and four juvenile flying squirrels, indicating that not only are the squirrels in the area, but they are successfully reproducing. Another problem with the model is the high probability of selection indicated in the northern portion of the Jasper burn [located in the southwest contiguous portion of BHNF (Fig. 2 and 3)]. The burn area has a high number of snags and, therefore, a high dead basal area. These high values resulted in a high probability of use for this area. In other studies, northern flying squirrels did not use areas with stand replacing fires, such as the Jasper burn, and edges of the burn area were sink habitat (Fisher and Wilkinson, 2005). This study contributes to the knowledge of northern flying squirrels for an isolated population at the southern edge of their range, where information is lacking for this species (Smith, 2007; Weigl, 2007). The RSF habitat model could be used by land managers for determining areas and habitat characteristics important to northern flying squirrels. Understanding resources requirements is important for isolated populations in intensively managed forests, because across their range northern flying squirrels play an important role in improving forest health through dispersal of mycorrhizal fungus and management decisions can greatly impact the population. The map shows the predicted spatial distribution of northern flying squirrel habitat (and presumably population) throughout the Black Hills. Future research and monitoring will also benefit from the model and map, because northern flying squirrels are secretive and elusive and trapping success is low, therefore, the RSF habitat map can be used to determine where to focus research efforts. Acknowledgments--The South Dakota Department of Game, Fish and Parks provided funding for this project through the State Wildlife Grants Program, Federal Assistance Study Number 2414. Additional funding was provided by South Dakota State University. We would like to thank project collaborator A. Kiesow; BHNF Forest Service employees C. Staab and J. Rydalch; our technicians M. Greer, K. Cudmore, J. Booth, B. Seiler and N. Hough and volunteers P. McCarthy, M. Schickel, S. Gunsaulus, S. Leroux and V. Shamblen for their enthusiasm and dedication to the project. We thank P. McCarthy, M. Schickel, H. Britten and C. Johnson for reviewing the manuscript. SUBMITTED 4 SEPTEMBER 2008 ACCEPTED 21 JANUARY 2009 LITERATURE CITED BEYER, H. L. 2007. Hawth's Analysis Tools for ArcGIS. http://www.spatialecology.com/htools/index. php. Accessed 1 October 2007. BOYCE, M. S. AND L. L. MCDONALD. 1999. Relating populations to habitats using resource selection functions. Trend Ecol. Evol., 14:268-272. --, P. R. VENIER, S. E. NIELSEN AND F. K. A. SCHMIEGELOW. 2002. Evaluating resource selection functions. Ecol. Modell., 157:281-300. CAREY, A. B. AND J. D. GILL. 1983. Direct habitat improvements-some recent advances, p. 80-87. In: USDA Forest Service Rocky Mountain Forest and Range Experimental Station. Snag habitat management: proceedings of the symposium. Fort Collins, Colorado. RM-99. --, S. P. HORTON AND B. L. BISWELL. 1992. Northern spotted owls: influence of prey base and landscape character. Ecol. Mono., 62:223-250. --. 1995. Sciurids in Pacific Northwest managed and oldgrowth forests. Ecol. Appl., 5:648-661. --, T. M. WILSON, C. C. MAGUIRE AND B. L. BISWELL. 1997. Dens of northern flying squirrels in the Pacific Northwest. J. Wildl. Manage., 61:684-699. --, J. KERSHNER, B. BISWELL AND L. DOMINGUEZ DE TOLEDO. 1999. Ecological scale and forest development: squirrels, dietary fungi, and vascular plants in managed and unmanaged forests. Wildl. Mono., 142:1-71. --. 2001. Experimental manipulation of spatial heterogeneity in Douglas-fir forests: effects on squirrels. For. Ecol. Manage., 152:13-30. CAREY, A. P. AND M. L. JOHNSON. 1995. Small mammals in managed, naturally young, and old-growth forests. Ecol. Appl., 5:336-352. CONNOR, P. F. 1960. The small mammals of Ostego and Schoharie counties, New York. New York State Museum Science Serve Bulletin, 382:1-84. DUCKWITZ, J. J. 2001. Small mammal survey of Wind Cave National Park. M.S. Thesis. South Dakota State University. Brookings, South Dakota. 113 p. FIELDING, A. H. AND P. F. HAWORTH. 1995. Testing the generality of bird-habitat models. Conserv. Biol., 9:1466-1481. --AND J. F. BELL. 1997. A review of methods for the assessment of prediction errors in conservation presence/absence models. Environ. Conserv., 24:38-49. FISHER, J. T. AND L. WILKINSON. 2005. The response of mammals to forest fire and timber harvest in the North American boreal forest. Mammal Rev., 35:51-81. FORD, W. M., S. L. STEPHENSON,J. M. MENZEL, D. R. BLACK AND J. W. EDWARDS. 2004. Habitat characteristics of the endangered Virginia northern flying squirrel (Glaucomys sabrinus fuscus) in the Central Appalachians. Am. Midl. Nat., 152:430-438. FROILAND, S. G. 1990. Natural History of the Black Hills and Badlands. The Center for Western Studies, Sioux Falls, South Dakota. GABEL, A., E. KRUEGER, C. ACKERMAN, M. GABEL AND S. WEANS. 2006. Occurrence and diet of flying squirrels (Glaucomys sabrinus) in the northern black hills of South Dakota. Report to South Dakota Game, Fish and Parks and U.S. Fish and Wildlife Service. Federal Assistance Project T-14-R, Study 2414. GIBSON, L. A., B. A. WILSON AND J. G. ABERTON. 2004. Landscape characteristics associated with species richness and occurrence of small native mammals inhabiting a coastal heathland: a spatial modeling approach. Biol. Conserv., 120:75-89. GOMEZ, D. M., R. G. ANTHONY AND J. P. HAYES. 2005. Influence of thinning of Douglas-fir forests on population parameters and diet of northern flying squirrels. J. Mammal., 69:1670-1682. HACKETT, H. M. AND J. F. PAGELS. 2003. Nest site characteristics of the endangered northern flying squirrel (Glaucomys sabrinus coloratus) in southwest Virginia. Am. Midl. Nat., 150:321-331. HOFFMAN, G. R. AND R. R. ALEXANDER. 1987. Forest vegetation in the Black Hills National Forest of South Dakota and Wyoming: a habitat type classification. USDA Forest Service, Rocky Mountain Forest and Range Experiment Station, Research Paper RM-276, Fort Collins, Colorado. HOLLOWAY, G. 2006. Flying squirrel (Glaucomys sabrinus and G. volans) habitat use and ecology in landscapes managed with partial harvesting silviculture in central Ontario. Ph.D. dissertation, University of Toronto, Toronto, Ontario, Canada. --AND J. R. MALCOLM. 2007. Nest-tree use by northern and southern flying squirrels in central Ontario. J. Mammal., 88:226-233. HOSMER, D. E. AND S. LEMESHOW. 2000. Applied logistic regression. Wiley-Interscience, New York. 373 p. HOUGH, M. J. 2008. Research techniques, habitat use, and ecology of northern flying squirrels, and research techniques and distribution of red squirrels in the Black Hills National Forest and northeast South Dakota. M.S. Thesis. South Dakota State University, Brookings, South Dakota. 143 p. JACKSON, H. H. T. 1961. Mammals of Wisconsin. University of Wisconsin Press, Madison. 504 p. JOHNSON, C. J., S. E. NIELSEN, E. H. MERRILL, T. L. McDONALD AND M. S. BOYCE. 2006. Resource selection functions based on use-availability data: theoretical motivation and evaluation methods. J. Wildl. Manage., 70:347-357. KING, J. A. 1951. Subspecific identity of the Black Hills flying squirrels (Glaucomys sabrinus). J. Mammal., 32:469-470. KOPROWSKI, J. L. 2005. Pine Squirrel (Tamiasciurus hudsonicus): a technical conservation assessment. USDA Forest Service, Rocky Mountain Region. Available: http://www.fs.fed.us/r2/projects/ scp/assessments/pinesquirrel.pdf [23 April 2006]. KRUEGER, E. M. 2004. Distribution and habitat associations of northern flying squirrels in the northern black hills of South Dakota. In: South Dakota wildlife diversity small grants results and reports for 2002. South Dakota Department of Game, Fish and Parks, Pierre, South Dakota. LEHMKUHL, J. F., K. D. KISTLER, J. s. BEGLEY AND J. BOULANGER. 2006. Demography of northern flying squirrels informs ecosystem management of western interior forests. Ecol. Appl., 16:584-600. LOEB, S. C., F. H. TAINTER AND E. GAZARES. 2000. Habitat associations of hypogeous fungi in the southern Appalachians: implications for the endangered northern flying squirrel (Glaucomys sabrinus coloratus). Am. Midl. Nat., 144:286-296. LUOMA, D., FRENKEL, R. AND TRAPPE, J. M. 1991. Fruiting of hypogeous fungi in Oregon Douglas-fir forests: seasonal and habitat variation. Mycologia, 83:335-353. MANLY, B. F. J., L. L. MCDONALD, D. L. THOMAS, T. L. McDoNALD AND W. P. ERICSON. 2002. Resource selection by animals: statistical analysis and design for field studies. Second Edition. Kluwer, Boston, Massachusetts. MARRIOTT, H., D. FABER-LANGENDOEN, A. McADAMS, D. STUTZMAN AND B. BURKHART. 1999. Black Hills Community Inventory: final report. The Nature Conservancy, Minneapolis, Minnesota. 111 p. MASER, C. E., J. M. TRAPPE AND R. A. NUSSBAUM. 1978. Fungal-small mammal interrelationships with emphasis on Oregon coniferous forests. Ecol., 59:799-809. MENZEL, J. M., W. M. FORD, J. W. EDWARDS AND M. A. MENZEL. 2004. Nest tree use by the endangered Virginia northern flying squirrel in the central Appalachian Mountains. Am. Midl. Nat., 151:355-368. --, --, --AND L. J. CEPERLEY. 2006a. A habitat model for the Virginia northern flying squirrel (Glaucomys sabrinus fuscus) in the central Appalachian Mountains. USDA Forest Service Northeastern Research Station, Newton Square, Pennsylvania. NE-729. --, --, --AND T. M. TERRY. 2006b. Home range and habitat use of the vulnerable Virginia northern flying squirrel Glaucomys sabrinusfuscus in the Central Appalachian Mountains, USA. Oryx, 40:204-210. MEYER, M. D. AND M. P. NORTH. 2005. Truffle abundance in riparian and upland mixed-conifer forest of California's southern Sierra Nevada. Can. J. Bot., 83:1015-1020. --, D. A. KELT AND M. P. NORTH. 2005. Nest trees of northern flying squirrels in the Sierra Nevada. J. Mammal., 86:275-280. MLADENOFF, D. J., T. A. SICKLEV AND A. P. WYDEVEN. 1999. Predicting gray wolf landscape recolonization: logistic regression models vs new field data. Ecol. Appl., 9:37-44. MOWREY, R. A. AND J. C. ZASADA. 1984. Den tree use and movements of northern flying squirrels in interior Alaska and implications for forest management, p. 351-356. In: W. R. Meehan, T. R. Merrell Jr. and T. A. Hanley (eds.). Fish and wildlife relationships in old-growth forests: proceedings of a symposium, Juneau, Alaska, April 1982. American Institute of Fishery Research Biologists, Moorehead City, North Carolina. MUSSER, G. G. 1961. A new subspecies of flying squirrel (Glaucomys sabrinus) from southwestern Utah. Proceedings from Biological Society Washington, 74:119-126. NORTH, M., J. M. TRAPPE AND J. FRANKLIN. 1997. Standing crop and animal consumption of fungal sporocarps in Pacific Northwest forests. Ecol., 78:1543-1554. ODOM, R. H., W. M. FORD, J. W. EDWARDS, C. W. STIHLER AND J. M. MENZEL. 2001. Developing a habitat model for the endangered Virginia northern flying squirrel (Glaucomys sabrinus fuscus) in the Alleheny Mountains of West Virginia. Biol. Conserv., 99:245-252. PAYNE, J. L., D. R. YOUNG AND J. F. PAGELS. 1989. Plant community characteristics associated with the endangered northern flying squirrel, Glaucomys sabrinus, in the southern Appalachians. Am. Midl. Nat., 121:285-292. POSILLICO, M., A. MERIGGI, E. PAGNIN, S. LOVARI AND L. RUSSO. 2004. A habitat model for brown bear conservation and land use planning in the central Apennines. Biol. Conserv., 118:141-150. PYARE, S. AND W. S. LONGLAND. 2001. Mechanisms of truffle detection by northern flying squirrels. Can. J. Zool., 79:1007-1015. --AND--. 2002. Interrelationships among northern flying squirrels, truffles, and microhabitat structure in Sierra Nevada old-growth habitat. Can. J. For. Res., 32:1016-1024. RAYSOME, D. B. AND T. P. SULLIVAN. 1997. Food limitation and habitat preference of Glaucomys sabrinus and Tamiasciurus hudsonicus. J. Mammal., 78:538-549. --AND--. 2004. Effects of food and den-site supplementation on populations of Glaucomys sabrinus and Tamiasciurus douglasii. J. Mammal., 85:206-215. RILLIG, M., K. K. TRESEDER AND M. F. ALLEN. 2002. Global Change and Mycorrhizal Fungi, p. 135-162. In: M. G. A. van der Heijden and I. R. Sanders (eds.). Mycorrhizal Ecology. Springer, New York, New York. 469 p. ROSENBERG, D. K. AND R. G. ANTHONY. 1992. Characteristics of northern flying squirrel populations in second- and old-growth forests in western Oregon. Can. J. Zool., 70:161-166. SHEPPERD, W. D. AND M. A. BATTAGLIA. 2002. Ecology, silviculture, and management of Black Hills ponderosa pine. USDA Forest Service, Rocky Mountain Forest and Range Experiment Station, Fort Collins, Colorado. SMITH, W. P. AND J. V. NICHOLS. 2003. Demography of Prince of Wales flying squirrels, an endemic of southeastern Alaska temperate rainforest. J. Mammal., 84:1044-1058. --, S. M. GENDE AND J. V. NICHOLS. 2004. Ecological correlates of flying squirrel microhabitat use and density in temperate rainforests of southeastern Alaska. J. Mammal., 85:663-674. --. 2007. Ecology of Glaucomys sabrinus: habitat, demography, and community relations. J. Mammal., 88:862-881. SOUTH DAKOTA DEPARTMENT OF GAME, FISH AND PARKS. 2006. Rare, threatened or endangered animals tracked by the South Dakota Natural Heritage Program. http://www.sdgfp.info/wildlife/ diversity/RareAnimal.htm. Accessed 18 March 2007. URBAN, V. 1988. Home range, habitat utilization and activity of the endangered northern flying squirrel. M.S. Thesis. West Virginia University. Morgantown, West Virginia., 54 p. USDA FOREST SERVICE. 1996. Black Hills National Forest land and resource management plan: phase II amendment. USDA Forest Service, Black Hills National Forest, Custer, South Dakota. --. 2005. Black Hills National Forest land and resource management plan: phase II amendment USDA Forest Service, Black Hills National Forest, Custer, South Dakota. --. 2006. Black Hills National Forest Geographic Information Systems--Available Coverages. http://www.fs.fed.us/r2/blackhills/projects/gis/index.shtml. Accessed 30 April 2007. USDA NATURAL RESOURCE CONSERVATION SERVICE. 2007. National Water and Climate Center. http://www. wcc.nrcs.usda.gov/. Accessed 15 Oct. 2007. VERNES, K. 2001. Gliding performance of the northern flying squirrel (Glaucomys sabrinus) in mature mixed forests of eastern Canada. J. Mammal., 82:1026-1033. --, S. BLOIS AND F. BARLOCHER. 2004. Seasonal and yearly changes in consumption of hypogeous fungi by northern flying squirrels and red squirrels in old-growth forest, New Brunswick. Can. J. Zool., 82:110-117. WATERS, J. R. AND C. J. ZABEL. 1995. Northern flying squirrel densities in fir forests of northeastern California. J. Wildl. Manage., 59:858-866. WEIGL, P. D. AND D. W. OSGOOD. 1974. Study of the northern flying squirrel, Glaucomys sabrinus, by temperature telemetry. Am. Midl. Nat., 92:482-486. --. 1978. Resource overlap, interspecific interactions and the distribution of the flying squirrels, Glaucomys volans and G. sabrinus. Am. Midl. Nat., 100:83-96. --, T. W. KNOWLES AND A. C. BOYNTON. 1992. The distribution of the northern flying squirrels, Glaucomys sabrinus coloratus, in the southern Appalachians, North Carolina Wildlife Resources commission, Raleigh, North Carolina. --. 2007. The Northern Flying Squirrel (Glaucomys sabrinus): A Conservation Challenge. J. Mammal., 88:897-907. WELLS-GOSLING, N. AND L. H. HEANEY. 1984. Glaucomys sabrinus. Mammal. Spec., 229:1-8. WITT, J. W. 1992. Home range and density estimates for the northern flying squirrel, Glaucomys sabrinus, in western Oregon. J. Mammal., 73:921-929. MELISSA J. HOUGH (1) AND CHARLES D. DIETER Department of Biology/Microbiology, South Dakota State University, SAG 304, Box 2207B, Brookings 57007 (1) Corresponding author: Telephone: 315-663-7521; e-mail: melhough18@hotmail.com Table 1.--Description of variables considered for the resource
selection function (RSF) habitat model for northern flying
squirrels studied in the Black Hills National Forest (May-Aug.,
2005-2007)
Variable Description
Continuous
Elev Elevation in meters.
Precip Average annual precipitation in centimeters.
DistStrm Distance to nearest stream in meters.
LiveHa Total number of live trees per hectare.
LiveBA Sum of the basal area ([m.sup.2]/ha) of live trees
[greater than or equal to] 12.7 cm dbh.
DeadBA Sum of the basal area ([m.sup.2]/ha) of dead trees
[greater than or equal to] 12.7 cm dbh. This
includes all salvageable and unsalvageable
standing mortality, clown mortality, standing
sound snags, down sound snags, or standing
unsound snags.
DeadDown The number of dead and down trees with a dbh
[greater than or equal to] 12.7 cm. This
includes salvageable curl unsalvageable down
mortality, down sound snags, or down unsound
snags.
TotalSnags Total number of standing mortality trees, hard
snags and soft snags with a dbh >12.7 cm per
hectare.
Categorical
GRA Crass-shrub habitat
TAA Aspen-birch habitat
TBO Bur oak habitat
TPP Ponderosa pine habitat
3A Habitat structural stage with trees 2-23 cm dbh
and 11-40% canopy cover.
3B Habitat structural stage with trees 2-23 cm dbh
and 41-70% canopy cover.
3C Habitat structural stage with trees 2-23 cm dbh
and <71% canopy cover.
4A Habitat structural stage with trees 23-40 cm dbh
and 11-40% canopy cover.
4B Habitat structural stage with trees 23-40 cn dbh
and 41-70% canopy cover.
4C Habitat structural stage with trees 23-40 cm dbh
and <71% canopy cover.
NO North aspect
NE Northeast aspect
EA East aspect
SE Southeast aspect
SO South aspect
SW Southwest aspect
WE West aspect
NW Northwest aspect
Table 2.--Chi-square tests comparing means of trapping and
tracking locations (used habitat) (n = 1203) and random
locations throughout the study area (available habitat) (n =
1203) to determine continuous variables retained for resource
selection function (RSF) habitat model of northern flying
squirrels studied in the Black Hills National Forest (May-Aug.,
2005-2007). See table 1 for description of variables
Used Available
Variable mean [+ or -] SE mean [+ or -] SE
Elev (b) 1755.11 [+ or -] 5.89 1669.74 [+ or -] 7.07
Precip 25.14 [+ or -] 0.09 23.58 [+ or -] 0.74
DistStrm 269.14 [+ or -] 7.03 292.34 [+ or -] 7.30
LiveHa 1048.89 [+ or -] 33.70 471.50 [+ or -] 20.49
LiveBA 28.57 [+ or -] 0.47 24.64 [+ or -] 0.58
DeadBA 7.22 [+ or -] 0.22 4.07 [+ or -] 0.20
DeadDown (b) 9.97 [+ or -] 0.41 5.93 [+ or -] 0.42
TotalSnags 7.46 [+ or -] 0.26 3.36 [+ or -] 0.19
Variable Z p-value (a)
Elev (b) 8.14 0.0000
Precip 10.36 0.0000
DistStrm -3.36 0.008
LiveHa 15.87 0.0000
LiveBA 5.81 -0.0001
DeadBA 16.21 0.0000
DeadDown (b) 8.94 0.0000
TotalSnags 15.99 0.0000
(a) Means tested with Wilcoxon rank sums (P [less than
or equal to] 0.10)
(b) Removed due to correlation (r [greater than or equal to]
0.701
TABLE 3.--Fisher's exact test comparing proportion of trapping
and tracking locations (used habitat) (n = 1203) and random
locations throughout the study area (available habitat) (n =
1203) to determine categorical variables retained for resource
selection function (RSF) habitat model of northern flying
squirrels studied in the Black Hills National Forest (May-Aug.,
2005-2007). See Table 1 for description of variables
Variable Used (%) Available (%) p-value (a)
GRA 4.82 22.36 -0.0001
TAA 12.14 2.58 -0.0001
TBO 0.17 0.25 1.0000
TPP 82.88 72.74 -0.0001
3A 1.16 4.82 -0.0001
3B 1.16 6.98 -0.0001
3C 4.16 3.41 0.3927
4A 36.41 28.01 -0.0001
4B 30.67 23.69 0.0001
4C 12.97 10.72 0.1008
NO 10.81 9.39 0.2790
NE 18.12 16.63 0.3603
EA 17.87 14.63 0.0356
SE 8.06 10.31 0.0663
SO 6.65 8.40 0.1219
SW 16.87 19.29 0.1379
WE 3.24 12.64 -0.0001
NW 18.37 8.73 <0.0001
(a) Means tested with Fisher's exact test (P < 0.10)
TABLE 4.--Estimated coefficients for the resource selection
function (RSF) habitat model for northern flying squirrels
studied in the Black Hills National Forest (May-Aug.,
2005-2007). See Table 1 for description of variables
[chi p-value Odds
Coefficient Estimate SE square] (a) ratio
Precip 0.0983 0.0121 59.8 -0.0001 1.10
DistStrm -0.0009 0.0002 17.13 -0.0001 1.00
GRA -0.5634 0.0838 50.27 -0.0001 0.32
TAA 1.3875 0.1642 118.03 -0.0001 16.04
3A -2.0695 0.2657 88.28 -0.0001 0.02
3B -1.8418 0.2421 97.7 -0.0001 0.03
NW 0.3129 0.0741 18.48 -0.0001 1.87
WE -0.9112 0.1118 81.55 -0.0001 0.16
LiveHa 0.0005 0.0001 93.72 -0.0001 1.00
DeadBA 0.0387 0.0097 16.77 -0.0001 1.04
TotSnags 0.0372 0.0086 20.98 -0.0001 1.04
(a) Coefficients tested with likelihood ratio chi square
(P [less than or equal to] 0.10)
TABLE 5.--Comparison of observed and expected frequencies for
the 10 quantiles, which represent ordinal resource selection
function (RSF) bins (categories) with increasing habitat
selection to determine validity of RSF model. RSF values
(va(x)) are habitat selection probabilities for northern flying
squirrels in the Black Hills \ational Forest (May-Aug.,
2005-2007). RSF bin midpoint w([x.sub.i]) is the midpoint RSF
value for the range of the RSF bin (i) and A([x.sub.i]) is the
number of 30-m pixels within the study area for bin (i). The
obsewed frequency is the number of independent validation data
points within each bin range. The expected frequency of each
bin ([n.sub.ii]) was estimated using [MATHEMATICAL EXPRESSION
NOT REPRODUCIBLE IN ASCII] where N is the total number of
validation points (N = 300)
RSF Bin Observed
midpoint # pixels
RSF Bin (i) RSF Bin range w([x.sub.i] A([x.sub.i] n
1 0.044-0.690 0.367 512895 0
2 0.690-0.795 0.743 510101 5
3 0.795-0.866 0.831 525449 11
4 0.866-0.903 0.885 507894 4
5 0.903-0.933 0.918 541661 12
6 0.933-0.955 0.944 505975 14
7 0.955-0.974 0.965 544236 45
8 0.974-0.985 0.980 467932 33
9 0.985-0.993 0.989 401043 63
10 0.993-1.000 0.99i 651218 113
Observed Expected
RSF Bin (i) % n % p-value (a)
1 0.00 3 0.99 0.08
2 1.67 9 2.95 0.29
3 3.67 15 5.06 0.43
4 1.33 21 6.84 0.00
5 4.00 28 9.38 0.01
6 4.67 32 10.71 0.01
7 15.00 41 13.62 0.67
8 11.00 41 13.51 0.35
9 21.00 39 13.12 0.02
10 37.67 71 23.82 0.00
(a) Observed versus expected frequencies tested with Pearson
chi square (P [less than or equal to] 0.05) |
| Gale Copyright: | Copyright 2009 Gale, Cengage Learning. All rights reserved. |