Response of Bat Activity to Land-Cover and Land-Use Change in Savannas is Scale-, Season-, and Guild-Specific

Tropical savannas are biomes of global importance that are under severe pressure from anthropogenic change, including land-cover and land-use change. Bats, the second-most diverse group of mammals, are critical to ecosystem functioning, but may be vulnerable to such anthropogenic stresses. However, there is little information on the response of savanna bats to land-cover and land-use change, especially in Africa. This limits our ability to develop conservation strategies for bats and maintain the ecosystem functions and services they provide in this biome. Using acoustic monitoring, we measured how guild-specific (aerial, edge, and clutter forager) bat activity responded to both fine-scale metrics of vegetation structure and landscape-scale metrics of land-cover composition and configuration across the wet and dry seasons in a savanna in southern Africa undergoing rapid land-cover and land-use change. We found that all three guilds responded more strongly to landscape metrics than fine-scale vegetation structure, although the specific metrics varied between guilds. Aerial and edge bats responded most strongly to the percent savanna cover and savanna fragmentation in both seasons while clutter bats responded to percent rural cover in the wet season and percent water cover in the dry. All three guilds responded more strongly to the landscape in the dry season than the wet season. Our results show it is possible to conserve bats, and the ecosystem services they can provide, in savannas undergoing anthropogenic land-use and land-cover change but strategies to do so must consider foraging guild, large spatial scales, and seasonal variation in bat activity. Highlights Bats in savannas respond to land-cover and land-use change on large spatial scales Landscape had a greater influence on bat activity in the dry season than the wet Aerial and edge forager activity responded to savanna cover and fragmentation Clutter forager activity was best explained by rural and water cover Minimizing fragmentation and maintaining water promotes bat activity in modified savannas


Introduction 51
Tropical  This study was conducted across an area of approximately 2,300 km 2 in the eastern low-lying 136 region of Eswatini referred to as the "Lowveld" which is bordered by the Drakensberg 137 Mountains in the west and the Lubombo Mountains in the east (Figure 1) We resampled this population count raster to the resolution of the classified raster using the 167 nearest-neighbor algorithm. We overlaid the population raster on the classified raster and 168 reclassified any cells with population count >1 as rural (Figure 1). 169 170

Acoustic Sampling 171
To capture variation in landscape cover across our study site we created a grid of 3 km 2 (~1.73 172 km × ~1.73 km) blocks (hereafter "block"). We then overlaid this grid on the classified raster. 173 We randomly selected 30 blocks (out of a possible 780) for acoustic surveys. These blocks were 174 stratified between the three land-cover categories, with ten blocks for each type (10 rural, 10 175 savanna, 10 sugarcane). Within each block, we deployed five Anabat Express detectors (Titley,  176 Inc., Ballina, Australia) at randomly placed points (hereafter "points") from November 2015 -177 July 2016 (Figure 1). Each detector was attached to a tree trunk or electric pole at 1.5 m above 178 the ground. Anabat detectors were set to record starting half an hour before sunset and continued 179 recording for six hours. We standardized the number of calls per sono-species by counting each species a maximum of 209 once per minute (Miller, 2001 We quantified the environment at two spatial scales: a fine scale around each sampling point and 221 the landscape scale within each sampling block. At the fine scale, we measured vegetation cover 222 and structure. In order to do so, we established a 30 m transect in each of the cardinal directions 223 from the sampling point. We evaluated canopy and ground cover at the sampling point where the 224 Anabat detector was placed and at points at 10 m intervals along each 30 m transect (total of 225 thirteen measurements) while shrub cover was measured along the length of each 10 m interval 226 within each transect (total of twelve measures). We measured the canopy cover using a spherical 227 densiometer (Forestry Suppliers, Inc., Jackson MS) (Lemmon, 1956). We visually estimated 228 ground cover in 1 × 1 m quadrats. We classified ground cover as: sugarcane, crop (all crops other 229 than sugarcane), grass, bare ground, and water. We measured shrub cover, woody vegetation <2 We calculated a variety of land-cover composition and configuration metrics within each 239 sampling block (Gustafson, 1998). To account for land-cover composition, we measured the 240 percent cover of savanna, rural, sugarcane, and water. For configuration metrics, we used

248
We calculated pairwise correlations between all fine-scale metrics and all landscape-scale 249 metrics using the function "rcorr" in the package Hmisc (Harrell, 2006 We evaluated a priori suites of models to explain bat activity at both the fine and landscape 264 scales. Each fine scale model included one of the fine-scale measures of vegetation structure: 265 canopy cover, shrub cover, sugarcane cover, bare ground cover, water cover, and distance to 266 water. We also included a null model ( Within each scale and for each season, we compared models using Akaike Information Criterion 286 corrected for small sample size (AICc) using the function "model.sel" in the package MuMIn 287 (Barton, 2017). We considered models within 2 AICc units to be competing models. We then 288 compared the point response models to each other and the best block response models to each 289 other, using AICc. We evaluated the parameters of the top models by examining their 95% 290 Confidence Intervals (CIs) and considered those that did not cross 0 to be relevant. We then 291 graphed relevant parameters to understand how activity changes across variables of interest.

293
Finally, we compared the fit of the overall best fine-scale models to the overall best landscape-294 scale models using Pseudo R 2 (McFadden, 1974). Pseudo R 2 measures the deviance explained by 295 a given model compared to the null model. We used Pseudo R 2 because the local and landscape 296 models had different responses (e.g. activity at Anabat points vs. activity summed across all 297 Anabat points within a block, respectively) and are therefore not directly comparable. 298 299 3. Results 300 We recorded acoustic data for a total of 3,408 hours during 120 sampling nights across the 30 301 sampling blocks. During this period, we identified a total of 69,897 bat calls. These calls were 302 predominantly from aerial bats (n=48,466), followed by edge bats (n=21,361), and finally clutter 303 bats (n=70). In general, we found that all three guilds responded more to the landscape scale than 304 the fine scale and this response was stronger in the dry season than the wet season, but each guild 305 responded differently to the landscape (Table 3). 306 307 3.1 Aerial foraging guild 308 At the fine scale, the best model to explain activity of aerial foragers during both seasons was 309 water cover. Activity increased with increasing water cover during both the wet season (β = 0.09, 310 [95% confidence interval: 0.08, 0.10]) and dry season (β = 0.14 [0.13, 0.16]). There were no 311 other competing models (Table 3, Table S1, Fig. 2). The Pseudo R 2 for top models in both 312 seasons was relatively low, though higher in the dry (0.07 vs. 0.04) ( Table 3).

314
At the landscape scale, the best model to explain activity in both seasons was a model with 315 interactive effects of savanna cover and savanna splitting ( 2.85, -2.08]). During the wet season, activity increased more quickly with increasing savanna 320 splitting where there was greater savanna cover. In the contrast, in the dry season, activity 321 decreased with increasing savanna splitting, with a more rapid decline when savanna cover was 322 higher (Fig. 2). There were no other competing models (Table 3, Table S1). Pseudo R 2 was over 323 twice as high for dry season models as the wet (0.28 vs. 0.12) (  (Table 3, Table S1). Similar to aerial bats, Pseudo R 2 was 332 twice as high for dry season models as the wet (0.08 vs. 0.04) ( Table 3, Table S2, Figure 3).

334
At the landscape scale, the best model to explain the activity of edge bats was a model with the 335 interaction between savanna cover and splitting (  (Fig. 3). There were no competing models ( We found that these landscape characteristics explained more of the bat activity response in the While we predicted that bats would respond more strongly to landscape composition than 393 configuration, we found that both composition and configuration, particularly fragmentation of 394 savanna land cover, were important for aerial and edge foraging bats. that sugarcane had a significant, positive effect on clutter bats at the fine scale in the dry season. 416 During this season, sugarcane plantations may offer resources, such as water from dams or 417 irrigation canals and insects that are scarce in savannas or rural areas. In addition, sugarcane is 418 densely planted and may reach two meters in height and therefore may provide suitable habitat 419 for clutter foragers. The resemblance of vegetation structure to native vegetation in areas of 420 agricultural land use may be more important for bats than the production intensity.

422
We found that water was important for all three foraging guilds in the dry season, although there 423 was variation in the spatial scale at which water drove activity for each guild. Water availability 424 is important for bats in general, providing both water for drinking and insect foraging (Adams, 2004; Pinto and Keitt, 2008), which we were unable to take into account.

445
Increasing levels of anthropogenic land-cover change around the world are cause for concern for 446 many wildlife species and biodiversity as a whole (