Winter models.

Of the variables used in the winter analysis (both sexes combined), growing season length had the highest mean increase in AUC (ΔAUC = 0.18), followed by winter temperature range (ΔAUC = 0.11). All SDM forms suggested that potential habitat suitability improved as growing season length increased beyond 200 days, with all model forms except Maxent predicting maximum habitat suitability with growing season length of 250–275 days. All model forms suggested that habitat suitability improved in areas with lower winter temperature ranges, where the difference between the maximum and minimum seasonal temperature is <20°C. The winter ensemble map suggests that male and female hoary bats might find favorable winter habitat along both coastlines and across large areas of the southcentral and southeastern United States, as well as much of Mexico (Fig 3).

Spring models.

In the spring analysis of females, growing season length had the highest mean increase in AUC (ΔAUC = 0.18), followed by spring temperature (ΔAUC = 0.11). As with the winter analysis, all SDM forms suggested that habitat suitability improved as growing season length increased beyond 200 days, with all model forms except Maxent predicting maximum habitat suitability with a growing season length of 250–275 days. All model forms suggested that habitat suitability improved in areas with mean spring temperatures of 10–20°C. In the spring analysis of males, growing season length had the highest mean increase in AUC (ΔAUC = 0.14), followed by precipitation seasonality (ΔAUC = 0.12), and winter temperature (ΔAUC = 0.10). All SDM forms suggested that habitat suitability improved as growing season length increased beyond 200 days; habitat suitability also improved in areas with higher precipitation seasonality, with most models predicting that habitat suitability increased with increasing variation in seasonal precipitation. Spring ensemble maps suggest that female hoary bats might encounter broader regions of favorable habitat farther north in the continental interior than males during that season (Fig 3).

Summer models.

In the summer analysis of females, summer temperature had the highest mean increase in AUC (ΔAUC = 0.22; this was the highest ΔAUC in all analyses), followed by spring temperature (ΔAUC = 0.17), and season length (ΔAUC = 0.12). All model forms suggested that habitat suitability improved in areas with mean summer temperatures > 15°C, and BRT and RF suggested that maximum habitat suitability occurred with mean summer temperatures between 15–25°C. All model forms suggested that habitat suitability for females during summer improved in areas with mean spring temperatures < 20°C, and BRT and RF suggested that maximum habitat suitability occurred with mean summer temperatures between 0–20°C. All SDM forms suggested that summer habitat suitability for females improved as growing season length increased beyond 150 days. In the summer analysis of males, spring temperature had the highest mean increase in AUC (ΔAUC = 0.14), followed by season length (ΔAUC = 0.13). All SDM forms suggested that maximum habitat suitability for males occurred in areas with spring temperatures between 0–20°C, and with a growing season length of > 150 days. Summer ensemble maps suggest that favorable habitat for female hoary bats might extend farther north and be more restricted to the interior of the continent than males; ensemble maps suggest proportionally less suitable habitat for males in the interior of the continent during summer, but more habitat than females in western areas of the U.S. and mountains of Mexico (Fig 3).

Autumn models.

In the autumn analysis of females, growing season length had the highest mean increase in AUC (ΔAUC = 0.18), followed by autumn temperature (ΔAUC = 0.12), and winter temperature (ΔAUC = 0.10). All SDM forms suggested that habitat suitability improved for females in autumn as season length increased beyond 200 days, and in areas with mean autumn temperatures of 5–25°C. All model forms except Maxent suggested that maximum habitat suitability occurs in areas with mean autumn temperatures of about 15°C. In the autumn analysis of males, growing season length had the highest mean increase in AUC (ΔAUC = 0.16). All SDM forms suggested that habitat suitability improved as growing season length increased beyond 200 days, with all model forms predicting maximum habitat suitability in areas with a growing season length of 250–275 days. Autumn ensemble maps suggest a general similarity between sexes in the distribution of potential suitable habitat, with modeled areas of suitable habitat for both sexes covering greater expanses than during other seasons (Fig 3). The possible exception to the general similarity in potential suitable habitat between sexes in autumn is the comparative sparsity of potential suitable habitat for males in the northern Rocky Mountains and adjacent plains in Canada (Fig 3).

Wind turbine data.

Maps of each seasonal ensemble map and the locations of 47,687 onshore industrial-scale wind turbines for the U.S. (and contained within our template) are shown in Fig 4. Mean seasonal habitat suitability for all wind turbines in our study area are: winter = 0.247 (range 0.005–0.817); spring female = 0.386 (range 0.003–0.805); spring male = 0.340 (range 0.002–0.832); summer female = 0.299 (range 0.013–0.851); summer male = 0.379 (range 0.013–0.884); autumn female = 0.407 (range 0.049–0.878); and autumn male = 0.439 (range 0.035–0.772). Histograms showing the empirical distribution of mean ensemble habitat suitability for each season and sex at all wind turbine locations contained within the template are also shown in Fig 4. In general, the greatest overlap in potential suitable habitat for hoary bats and the location of existing wind turbines (negative [left] skewed histograms of Fig 4) occurs during autumn.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. Ensemble maps for each season and sex for hoary bats (Lasiurus cinereus) in the United States using ensemble binary prediction from 5 species distribution model forms.

Locations of > 47,000 industrial wind turbines in the contiguous United States are indicated by black stars. The color ramp indicates the proportion of the 5 SDM model runs for each season that are considered suitable habitat. Darkest blue indicates that all of the models predict the area to be suitable habitat, while lightest blue indicates that all of the models predict the area to be unsuitable habitat. The histogram associated with each map shows the empirical distribution for mean habitat suitability for all wind turbines in the United States, estimated using the 5 SDM approaches. Negative (left) skewness in the histograms, or higher frequencies of values toward the right side (= 1) of the horizontal axis, indicate overlap between modeled suitable habitat and turbine locations.

https://doi.org/10.1371/journal.pone.0132599.g004

Winter.

Convention suggests that hoary bats migrate south during autumn to escape cold northern winters [12,36–41]. Subsequent observations led to the conclusion that at least some hoary bats might hibernate in northern areas [26,42–44]. Cryan [10] concluded that hoary bats occurred most often during winter in California, Mexico, and scattered locations in coastal areas of the eastern U.S. In our analysis, potential habitat suitability increased in areas with long growing seasons (>200 days) and temperature ranges of < 20°C. The winter ensemble map suggests great expanses of potentially suitable winter habitat throughout Mexico, along the Pacific Coast of the U.S., and across most of the south-central and eastern U.S. Our map of hypothesized winter distribution is consistent with earlier [10,12,38–41,45–48] and more recent [13] claims that warmer parts of North America provide suitable winter habitats for hoary bats. However, the general lack of occurrence records for hoary bats during winter from the south-central and eastern parts of the U.S. [10,12] raise the question of whether the broad expanses of potential habitat extending as far north as the Great Lakes that we modeled and mapped are actually occupied by hoary bats in winter. Given the conservation and management value of determining the true wintering grounds and other important seasonal habitats of hoary bats, areas where models predict habitat suitability but where evidence of species presence is lacking could be targeted for future presence-absence surveys to assess the validity of this seasonal hypothesis and further refine distributional understanding. The winter roosting behaviors and thermoregulatory strategies of hoary bats remain to be determined. Hoary bats can use torpor to save energy, at least for short periods of time, in air temperatures down to 0°C [11]. Although we do not know whether hoary bats remain active in winter, hibernate, or use an intermediate thermoregulatory strategy, our findings suggest they may select winter habitats that promote a combination of torpor and daily activity during warmer periods. Due to the relatively small number of winter records in our analysis, we pooled winter occurrence records of both sexes; however, we speculate that the habitat needs of male and female hoary bats may be similar during winter, and that males and females both require roosting conditions that are not excessively cold and that promote survival until spring. Testing for differences in the hypothesized winter habitat for males and females will require additional occurrence information from wintering bats.

Spring.

Hoary bats have been captured in large numbers in western and southwestern parts of the U.S. during spring, presumably on migration [10,12,43,49,50]. Findley and Jones [12] presented evidence that female hoary bats began spring migration earlier than males and that females might move from wintering grounds in Mexico to eastern and northern parts of the continent during spring and early summer, whereas males seemed to migrate later and concentrate more in distribution in western areas of the continent during spring and early summer. Cryan [10] concluded that in addition to bats originating from wintering grounds in Mexico, some hoary bats observed in the southwestern U.S. during spring, and particularly females, were likely migrating eastward from California; analysis of occurrence records also provided little evidence that female hoary bats migrate northward out of California up the Pacific Coast during spring. Our maps of hypothesized spring distribution of males and females suggest that hoary bats tend to move through areas with relatively long (> 200 days) growing seasons and mean seasonal temperatures of 10–20°C during spring, but that they might not migrate through the northern Rocky Mountains and northern Great Plains in that season. It is unclear why spring temperature was an important variable in spring models for females only, or why potentially suitable habitat in the ensemble maps is more extensive for females than males during spring. Adult male hoary bats are more likely than females to use torpor in some areas during spring [11] and are assumed to migrate later than females, so differences in our models and maps could have been due to bias in detectability, sampling, or may reflect actual differential habitat selection between the sexes during spring. In general, our maps of hypothesized spring distributions of females and males indicate similar patterns, and suggest that hoary bats are transitioning during spring from wintering regions in maritime climates near the edges of the continent toward summering grounds near the interior of the continent, with greater expanses of contiguous potential spring habitat predicted for the western half of the continent. Our maps are consistent with prior hypotheses of spring movement by hoary bats out of wintering grounds in Mexico, California, and eastern coastal regions of the U.S. Despite evidence of movement toward the continental interior during spring, suitable habitat conditions may also persist in western wintering grounds (e.g., Mexico and California) into the spring. Our maps do not indicate that hoary bats are strongly associated with any mountain ranges during spring, although association with mountains seems to occur during summer and autumn.

Summer.

Hoary bats have been observed throughout the U.S. and the southern half of Canada during summer [10,12,26], and several researchers have noted that they are not regularly documented in the southeastern U.S., east of the Mississippi River and south of the Ohio River [10,12,51]. Earlier studies presented evidence that adult males tend to be concentrated in distribution in mountainous areas of the western continent (including Mexico) during early summer, whereas females mostly occur east of the Rocky Mountains. Our maps of hypothesized distribution for both sexes predict potential summer habitat in areas with spring temperatures averaging < 20°C and with growing seasons of more than 150 days. Models for females included summer temperatures averaging > 15°C as having an important influence, but summer temperature was not an important predictor of male occurrence. Sex-specific differences in distribution, roost use, and thermoregulatory strategies are common in bats during summer [14,52], and it has been proposed that female hoary bats may be limited to areas of the continent with conditions favorable for gestation, lactation, and pup rearing, whereas males may not be as constrained by habitat needs during summer [11]. For example, our maps of hypothesized summer distribution indicate that California, with its seasonally dry Mediterranean climate, may not be suitable habitat for females during the summer months, but may be suitable for males. Our ensemble maps for summer females and males generally concur with prior understanding, although considerable areas of the western continent show potentially suitable habitat for female hoary bats, which contrasts the lack of previous evidence that female hoary bats regularly occur in western mountainous regions of the U.S. during summer. This is another area where our hypothesis maps differ from known patterns of distribution and could reflect incomplete characterization of the factors influencing the distribution of female hoary bats or the undetected presence of female hoary bats during summer across large swaths of the continent. Although these results suggest that females may occasionally travel farther between winter and summer grounds, they also suggest that there is considerable overlap in the summer distributions of male and female hoary bats.

Autumn.

Earlier scientists concluded that hoary bats likely migrated south in the autumn to avoid colder northern winters and lack of available prey [12,38–40,53]. As more details on distribution emerged, others proposed that hoary bats may move toward coastal areas in autumn where winter temperatures are warmer than in the interior of the continent [10,12,13,45–48,54]. In our analysis, areas with growing season lengths of > 200 days were predicted to be suitable habitat for hoary bats during autumn, but average temperature was again only an important predictor in the models for females, suggesting females may be more likely to occur in areas with autumn temperatures that average between 5–25°C. The maps of hypothesized distribution during autumn suggest that contiguous expanses of potential habitat occur across much of the species’ range and over more of the continent than during any other season. Analysis of occurrence records previously indicated outward range expansion and coastward movements of hoary bats during autumn [10]. Although confirmatory studies are lacking, hoary bats are considered habitat and roosting generalists [26]. Our maps are consistent with the previously proposed hypothesis that hoary bats occur in more regions and habitat types during autumn than during any other season. With the possible exception of western Canada, where mostly female habitat is predicted, our maps also indicate fewer qualitative differences between the sexes in potential habitat during autumn compared to the other seasons for which sex-specific models were created. Hoary bats begin mating during late summer and autumn [55] and it is suspected that sex differences in distribution and habitat use are less prevalent in autumn as the species returns to wintering grounds. Similar to the presumably transitional habitats inferred from spring models, autumn maps of hypothesized distribution do not indicate that hoary bats would be unlikely to encounter abundant suitable habitat in southeastern coastal regions while moving from their summering areas toward wintering grounds during autumn.

Impacts of wind energy development.

Migratory tree bats represent a substantial proportion (> 75%) of bat fatalities at wind energy facilities, and hoary bats represent the largest fraction (approximately 40–50%) of all documented bat fatalities in North America [15,17,20,23]. Our results offer new insight and testable hypotheses about where important seasonal habitats of hoary bats may occur and the areas through which they may travel among seasons. Our model results and maps are not suitable for predicting the actual presence of hoary bats, but represent hypotheses of where the species may occur at any given time of year. We emphatically state that confirmatory field studies are necessary to assess the models and hypothesized distributions described herein, preferably by following the detailed movements of many individuals with tracking devices (e.g., satellite tags) or in a probabilistic sampling framework using presence-absence sampling methods. Most bat fatalities at wind energy facilities in the temperate zones of North America occur during late-summer and autumn [18,20] and our modeling results may help better determine where hoary bats are likely to occur during that period of susceptibility. Unfortunately, fatalities at turbines occur most during the season in which our models suggest that potentially suitable habitats for hoary bats span across broader areas of the continent than during any other time of year. If these hypothesized distributions are confirmed, then hoary bats may be more seasonally general in their habitat use (or probability of being found out of their usual habitat) in autumn than during any other time of year, making it more difficult to predict the specific places they are likely to occur during the period of greatest susceptibility. However, alternative approaches to determining autumnal habitats of hoary bats may be possible.

Developing a better understanding of where hoary bats winter may be a key element of trying to predict and avoid fatalities at wind turbines. Our maps of hypothesized distribution indicate that hoary bats winter in locations with relatively long growing seasons and where winter temperatures are moderated by proximity to oceans. California and Mexico might be used consistently during winter, but the extent of wintering habitat in coastal regions of the eastern and southeastern U.S. remains to be determined due to disagreement between our winter model results and the lack of prior occurrence records for more interior regions of the southeastern U.S. during winter [10]. Future studies are needed to assess these new hypotheses of seasonal distribution, which may indicate that risk of wind energy to bats may be highest in habitats between interior summering areas and coastal wintering grounds (e.g., near passes in coastal mountain ranges). We predict that wind turbines built between the summering and wintering grounds of hoary bats will cause more fatalities during the period of susceptibility than those built out of transitional areas. Ironically, strategically sampling wind turbines at a continental scale for hoary bat fatalities and then comparing fatality numbers and sex composition to SDM-predicted, sex-specific habitat suitability may be a feasible method of evaluating our seasonal distribution and migration hypotheses. Only by such validation might the use of SDMs to predict fatality and avoid high-risk sites for wind energy development move from the realm of theoretical to practical.