Patterns of roost site use by Asian hornbills and implications for seed dispersal

Animals spend a significant amount of time roosting. Therefore, understanding roosting patterns and the processes that influence roosting behaviour and roost site choice is essential. Hornbills exhibit interesting roosting patterns with some species roosting communally in large flocks. They are important seed dispersers and patterns of roost site use can have a significant influence on seed dispersal distributions and thereby on plant recruitment. We documented roost site use by four Great Hornbills (Buceros bicornis) and one Wreathed Hornbill (Rhyticeros undulatus) at a site in north-east India using GPS telemetry. We examined the influence of riverine habitats, nests and foraging range on roost selection. We determined the proportion of seeds that hornbills disperse at roosts and the dispersal distances of seeds dispersed at roosts from the source trees. Through telemetry, we found that roosts of Great Hornbills were generally in forested habitats. Our telemetry data showed that Wreathed Hornbill roosts were close to the river. These results were corroborated by observational data from roost sites where we had regular detections of relatively large flocks of Wreathed Hornbills and occasionally Great Hornbills. The roost sites were not close to the nest sites and were generally within the 95% kernel density diurnal activity ranges. Hornbills dispersed a small proportion of seeds at roost sites. Seeds dispersed at roost sites had almost twice the dispersal distances compared to those dispersed at non-roost sites. This study highlights variation in roost site pattern across individual hornbills and its implications for seed dispersal.


Introduction 36
We carried out the study in Pakke Tiger Reserve (area: 861.9 km 2 ; 92°36′-93°09′E and 136 26°54-27°16′N), which is part of the Eastern Himalaya Biodiversity Hotspot, in Arunachal 137 Pradesh state, in north-east India. We tagged the hornbills over two years in the south-eastern 138 part of the reserve, an area dominated by tropical semi-evergreen forest ( . We only tagged adult males. Since hornbills are diurnal animals, tags were 153 programmed to take locations at 15-minute intervals throughout the day and turn off at night 154 to save tag battery power. Reliable roosting information was not available for one of the 155 Great Hornbills, whose tag was programmed to shut down at sunset and turn on at sunrise. 156 For all the other birds the tag was programmed to shut down at least 45 min after sunset (~ 157 19:00 hr IST) and turn on at least two hours before sunrise (~ 03:10 hr IST), which allowed 158 us to extract information on hornbill roosting reliably. Based on our field observations, 159 hornbills arrive at roosts latest by 17:00-18:00 hr (IST) in June when the days are longest.
6 Additional details on the methods and the study area can be found in (Naniwadekar et al. 161 2019a, b). The GPS data for this study can be accessed from Naniwadekar et al. (2019c). 162 We monitored one roost site for 45 days between April to June (breeding season) in 163 2015 and two roost sites (including the one monitored in 2015) for a total of 211 roost watch 164 days (190 unique days on 21 days both roosts were observed) across breeding and non-165 breeding season in 2016. One of the monitored roosts was next to the Pakke river. The other 166 roost was on a hill slope in Darlong village, which is on the banks of Pakke River. While the 167 first roost site was 20 m from the riverbank, the other was 370 m from the riverbank. Both the 168 roost sites were located outside the boundary of Pakke Tiger Reserve in the adjoining Papum 169 RF and were close to human habitation. Two-three observers counted hornbills at roost sites 170 between 16:00 to 18:00 hr (IST). We recorded the species, time of arrival and number of 171

individuals. 172
Analysis 173 To determine roosts of hornbills, we calculated mean displacement distances between 174 consecutive time points (for every 15-min interval) between 03:15-19:00 hr and found that 175 displacement distances were least for 19:00 hr for the five hornbills (< 32 m; range across 176 individuals: 17.9-31.2 m). Therefore, we used the location of hornbills at 19:00 hr as the 177 roost location for the day. Due to GPS tag errors, location data was not available for 19:00 hr 178 (and often for few periods before that) for all days for all birds. Mean displacement distances 179 were higher for time intervals before 18:30 hr. Days for which we had obtained data at 18:30 180 hr, we had received data at 19:00 hr also. Therefore, we used the data only for the days on 181 which we had location information at 19:00 hr. We used the roost data to calculate the mean 182 displacement distances between roosts on consecutive days to determine how far, on average, 183 were the roosts located on successive days for the different hornbill individuals. 184 We used hierarchical cluster analysis with complete-linkage method implemented 185 through the 'stats' package in R to identify the cluster of points that were within 200 m from 186 each other (R Core Team 2019). Our observations at communal roosts of hornbill indicate 187 that often hornbills roost on multiple trees at a single site and the distance between the trees 188 can be around 100 m from each other. After arriving at the roost sites, they also move 189 between individual trees. While individual roost locations may vary, we considered all roost winter season (GH2NBr), we used the locations between 06:00 and 16:00 hr since the sunrise 211 and sunset is later and earlier in winters respectively. Hornbills start arriving at the roost sites 212 up to half an hour before sunset (Datta 2001) and our long-term observations of hornbills at 213 select communal roosts indicate that the birds mostly leave the roost before sunrise. We 214 independently validated these timings with the mean displacement in every 15-min intervals 215 for our tagged birds to confirm that our selected timings coincided with the diurnal activity of 216 the different hornbills. We plotted the roost locations as identified using the hierarchical 217 cluster analysis on these diurnal activity ranges of the hornbills to determine if the roost 218 locations were within the diurnal activity range or outside it. 219 We followed the method outlined in Naniwadekar et al. (2019b) to estimate the 220 relative proportion of seeds that were deposited at roost and non-roost (other) sites and 221 determine the dispersal distances of seeds that were deposited at roost and non-roost sites. In 222 Naniwadekar et al. (2019b), we have outlined the method that we followed to estimate the 223 relative proportion of seeds that were dispersed at the nest and non-nest sites. A random 224 starting point was selected following the distribution of foraging sightings across the entire 225 day. We excluded roost and nest locations of birds from this starting point selection since 226 they were unlikely to be fruiting trees. We integrated the movement information with the gut 227 passage time data to determine the end location where the hornbill potentially dispersed the 228 seed. If the end location was within 50 m of the roost location for that particular day from 229 which the starting point was selected, then the seed was classified as dispersed at the roost 230 site. In this case, the roost location was the precise, daily roost location and not the 'roost 231 site' (which was centroid of all roost locations within 200 m from each other) that was 232 identified using the hierarchical cluster analysis. We used the 50 m buffer around the roost to 233 account for both the GPS error and the typical canopy extent of the large trees which 234 hornbills often use for roosting. We also determined the dispersal distances of the seeds from 235 their random start location. Additional information on the distribution of foraging sightings 236 over time, gut retention times and the analytical framework can be found in Shukla  We had a total of 214 days of roosting data for the five hornbills (Table 1). The number of 242 days of data available for a single individual varied between 19-72 days ( Table 1) Tiger Reserve across the Pakke River close to human settlements in the neighbouring state of 248 Assam (Fig. S1). The mean distance between roosts on successive nights for the different 249 Great Hornbills varied between 130-1051 m ( Table 2). For the Wreathed Hornbill, the mean 250 distance between roosts on successive nights was 1305 m (Table 2). There was no consistent 251 difference between breeding and non-breeding Great Hornbills (Table 2). However, the 252 maximum distance between roosts on successive nights was greater than 1.18 km for the two 253 non-breeding Great Hornbills but was less than 710 m for the two breeding Great Hornbills 254 (Table 2).  (Table 1). There was only 19 days of data available for GH4Br. 260 The maximum number of days for which an individual bird used a roost site over the entire 261 tracking period was 17 days for both GH3Br and GH4Br ( Fig. S1A and B). For the non-262 breeding Great Hornbills, the number of roost sites used during the entire tracking period 263 varied between 11-33 for GH3NBr and GH5NBr (for which data was available for 30-55 264 days) ( Table 1). The maximum number of days for which an individual bird used the same 265 roost site during the entire tracking period was 8 and 11 days for GH2NBr and GH5NBr, 266 respectively ( Fig. S1C and D). For the breeding Wreathed Hornbill, we identified ten roost 267 sites during the entire tracking period from 72 days of available data (Table 1)

. Wreathed 268
Hornbill used two of the ten roost sites for up to 18 days each (Fig. S1E). Both these roost 269 sites were close to the river (Fig. S1E). The mean number of successive nights the five birds 270 used the same roost site (indicating repeated use of the same roost) varied between 2.6-5.3 271 days (Table 2). 272 Roosts of Great Hornbills were generally away from the river bank, but those of 273 Wreathed Hornbill were close to the river. The mean (± SE) distance of the roost sites from 274 the river was not very close for the 1850.2 (± 326.2) m for GH2NBr, 3054.4 (± 80.3) m for 275 GH3Br, 941.5 (± 219.4) m for GH4Br, 1536.5 (± 176.4) m for GH5NBr and only 157.6 (± 276 65.2) m for WH1Br (Fig. 1). The hornbills did not roost near the nests. The mean (± SE) 277 distance of the roost sites from the nest site was 423.9 (± 86.8) m for GH3Br, 964.6 (± 231.4) 278 m for GH4Br and 1754.2 (± 473.5) m for WH1Br (Fig. 2). All the roost sites of the breeding 279 Great Hornbills (GH3Br and GH4Br) (except one for GH3Br) were outside the 50% kernel 280 density utilization distribution (Fig. 2). However, eight of the 11 roost locations of GH2NBr 281 and 16 of the 33 roost locations of the GH5NBr were within the 50% kernel density 282 utilization distribution (Fig. 2). For the Wreathed Hornbill, six of the 10 locations were 283 outside the 50% kernel density utilization distribution (Fig. 2). All hornbills appear to exhibit 284 relatively long bout of flying when they leave their roosts in the morning and when they 285 arrive at their roosts in the evening as was evident by examining the mean displacement at 286 every 15-min interval (Fig. 3).  (Fig. 4). 301 The relative percentage of seeds dispersed at the roost sites varied between 7 -17% 302 (Fig. 5). The breeding hornbills (GH3Br: 7%; GH4Br: 7%) dispersed fewer percentage of 303 seeds under the roost trees as compared to the non-breeding hornbills (GH2NBr: 17%; 304 GH5NBr: 10%) (breeding vs. non-breeding hornbills: χ 2 1 =146.6, P < 0.001) (Fig. 4). The 305 estimated percentage of seeds dispersed under the roost trees for the breeding Wreathed 306 Hornbill was 9% (Fig. 5). The mean dispersal distances of seeds is higher when they are 307 dispersed at roost sites as compared to non-roost sites (Fig. 5). 308

Discussion 310
This is the first study to examine the individual patterns of roost use by Asian Hornbills and 311 the influence of specific habitat, nest and diurnal foraging range on roost site selection. While 312 Wreathed Hornbills tend to roost near rivers, individual Great Hornbills mostly roost in 313 forested sites away from the river. Both Great and Wreathed Hornbills show some roost site 314 fidelity with individuals using some roosts more often than others. This study highlights that 315 despite exhibiting relatively long commutes to the roost, almost all the roost locations of 316 different hornbill individuals were within the diurnal activity ranges of the hornbills. This 317 study also highlights that hornbills dispersed relatively small proportion of seeds at roost 318 sites. Hornbills dispersed the bulk of the seeds at non-roost sites, which are likely to be more 319 suitable for germination of seeds. Given that individual hornbills use multiple roosts, not all 320 roosts are likely to be used frequently. The infrequently used roost sites might offer 321 favourable opportunities for seeds to establish. Interestingly, the seed dispersal distances at 322 roost sites were more than twice compared to sites where hornbills perch but are not roost 323 sites, facilitating very long-range seed dispersal events during roosting by hornbills. The roost 324 site monitoring data corroborates the finding from the telemetry study demonstrating that 325 Wreathed Hornbills prefer to roost close to the river often in relatively larger numbers as 326 compared to the Great Hornbill.  consecutive days when the birds used the same roost, and the number of days for which roost 641 data from consecutive days was available is also given. * -Number of days for which the 642 roost data was available for successive nights. This number is different from the number of 643 days for which the roost data is available (which is summarized in Table 1) since no roost 644 data was available for some nights during the tracking period. within the 95% (area enclosed within the black line) kernel density diurnal activity range for 653 the five hornbills but not necessarily within the 50% kernel density activity range (area 654 shown in grey). The locations used for the kernel density diurnal activity range estimation are 655 those between 05:00-17:00 hr for the five hornbills, thereby excluding the roost locations. 656 The black dots are the roost locations of the bird identified using hierarchical cluster analysis. 657 One roost location for GH4Br which was outside the 95% kernel density diurnal activity 658 range is not shown since it was used for only one night and it was far away from its activity 659 range. Coordinates on the map represent the north and east latitudes and longitudes 660 respectively. 661 Great Hornbills with the former tagged in the winter season (which is the non-breeding 664 season) and the latter in the early summer (which coincides with the breeding season). 665 WH1Br is the breeding Wreathed Hornbill. There is a spike in the displacement just after and 666 before the bird leaves the roost. It indicated that birds travel a long distance after they leave 667 the roost in the morning and before returning to the roost in the evening. 668  GH4Br