Using life history trade-offs to understand core-transient structuring of a 1 small mammal community 2

23 An emerging conceptual framework suggests that communities are comprised of two 24 main groups of species: core species that are temporally persistent, and transient species that are 25 temporally intermittent. Core and transient species have been shown to differ in spatiotemporal 26 turnover, diversity patterns, and importantly, survival strategies targeted at local vs. regional 27 habitat use. While the core-transient framework has typically been a site-specific designation for 28 species, we suggest that if core and transient species have local vs. regional survival strategies 29 across sites, and consistently differ in population-level spatial structure and gene flow, they may 30 also exhibit different life-history strategies. Specifically, core species should display relatively 31 low dispersal rates, low reproductive effort, high ecological specialization and high survival rates 32 compared to transient species, which may display a wider range of traits given that transience 33 may result from source-sink dynamics or from the ability to emigrate readily. We present results 34 from 21 years of capture-mark-recapture data in a diverse rodent community, evaluating the 35 linkages between temporal persistence, local abundance, and trade-offs among life-history traits. 36 Core species at our site conservatively supported our hypotheses, differing in ecological 37 specialization, survival and dispersal probabilities, and reproductive effort from transient species. 38 Transient species exhibited a wider range of characteristics, which likely stems from the multiple 39 processes generating source-sink dynamics and nomadic transience in local communities. We 40 suggest that trait associations among core-transient species may be similar in other systems and 41 warrants further study. 42 43

We use 21-years of data from a diverse desert rodent community containing information 136 on movement, mark-recapture rates, and reproduction to test the hypothesis that core species have 137 7 fundamentally different life history strategies than transient species as expected from differences 138 in local vs. regional habitat use. We predict that core species will generally be associated with 139 relatively low movement rates, high survival rates, and low fecundity. We predict that non-core 140 species will display a mixture of traits, depending on whether they are source-sink or nomadic 141 transients, but generally have more incidence of high movement rates, low survival rates, high 142 fecundity, and resource or habitat generalism (Figure 1 behaviorally dominant kangaroo rats (Dipodomys spp.) have enlarged auditory bullae that make 154 it possible to selectively exclude them from plots that have a smaller gate size (n=8). Total 155 rodent removal plots have no gates (n=6), while control plots (n=10) have relatively large gates 156 that allow all species access (Brown 1998 being differently suited to local and regional habitat use strategies. We analyzed data for 170 individuals where there was no discrepancy in recorded species across captures. During 1989-171 1999, individuals were marked using ear and toe tags, and during 2000-2009, individuals were 172 mainly marked with PIT tags. We conducted extensive data cleaning and error checking to 173 ensure that potential problems in the data (e.g., duplicate tags, uncertainty in sex or species) were 174 resolved. In cases where the data with identical tags could be clearly partitioned into unique 175 individuals, we assigned new unique tag numbers to each individual. In cases where data could 176 not be clearly partitioned into individuals, or where species identity was questionable, the data 177 were excluded from analysis. 178 Core and transient species status was assigned based on temporal persistence, as defined 179 by the proportion of years that each species was present

184
Using individual-level recapture data, we assessed movement characteristics for each 185 species. Locations of the permanently marked trap stakes were recorded in 2010 using ProMark3 186 GPS Units with error < 2 cm. We recorded the distance in meters between trap stakes among 187 chronologically ordered capture histories for each individual. For each species, we binned the 188 individual movement data by 6-meter increments that roughly represent the distance between 189 stakes (with bin 1 representing distance 0-3 meters, or recapture at the same stake), and plotted 190 the data in histograms. For each species, we calculated the modal distance and the mean plus one 191 standard deviation of the log (Y+1) transformed data to determine a species-level benchmark at 192 which each movement distribution transitions into long-distance movements. We chose this 193 transformation to meet the assumptions of normality and because there are many 0 m movements 194 (Sokal and Rohlf 2012). For each species, these histograms provide insight into the frequency at 195 which individuals move short versus long distances. Using the combined individual movement 196 distances of the core species within each feeding guild (granivore, folivore, and carnivore), we 197 set the mean plus one standard deviation of log(Y+1) transformed data as our guild-level 198 benchmark defining a short versus a long distance movement to compare across all species. We 199 used this method because home range size likely differs based on trophic group (Mace and 200 Harvey 1983). probabilities (ψ), we used a two-state model where all species were first recorded in state 1, and 208 were switched to state 2 (or from state 2 back to state 1) conditional on the distance between 209 trapping stakes upon recapture being greater than the guild-level benchmark defining short 210 distance movements. In this two-state CMR model, transitioning between states indicates long 211 distance movement, and staying in the same state indicates short distance movement, conditional 212 on apparent survival and recapture probabilities. We defined apparent survival probability as the 213 probability that an individual alive in trapping period i survived and did not emigrate from the 214 entire study area by trapping period i+1. We defined recapture probability at trapping period i+1 215 as the probability that a live individual on the study area was recaptured in a trap. All 216 probabilities were measured over a time scale of approximately one month, the time between 217 trapping events. To address inconsistencies in the data, we controlled for omitted trap periods 218 (when trapping did not occur or the site was only partially trapped) by fixing recapture 219 probability to zero for those occasions. It should be noted that we could not differentiate between 220 permanent emigration and death, which may affect interpretation of our survival estimates. Thus, 221 low apparent survival probabilities may indicate low actual survival, high permanent emigration 222 from the study area, or both. We evaluated each species separately in RMark to estimate 223 apparent survival, recapture, and transition probabilities (White and Cooch 2012), except for 224 transient granivores, which we grouped together because there were not enough captures to 225 analyze species separately. Pooling data for all species into a single dataset, and designating 226 species or strategies with factors, led to an extremely large CMR dataset that prohibited 227 11 computational analysis using MARK and RMark. We thus used post hoc analyses to compare the 228 estimates for core versus intermediate and transient species. For further details on our RMark 229 analysis, please refer to our code, which is maintained online in a public GitHub repository along 230 with the data (https://github.com/weecology/portal-rodent-dispersal) and is available in the 231 online supplement. 232 To assess reproductive effort for each species, we tracked the reproductive history for 233 captured individual females within each calendar year. We considered females with enlarged 234 and/or red nipples or who were pregnant (researcher could feel embryos) to be actively 235 reproducing. If a female was marked in reproductive condition during consecutive trapping 236 periods, we assumed it to be a single reproductive event. Reproductive condition recorded across 237 non-consecutive trapping periods was considered as multiple reproductive events. We used data 238 from females because males display reproductive signals for a larger portion of the year, and 239 male reproductive status does not necessarily indicate recent copulation or reproductive success. 2013) and using the assumption of linear decrease in trait covariance (bm gls; Brownian motion 249 model). We also compared species using a linear regression and PCA biplots to determine the 250 12 relative influences of temporal status, feeding guild, and phylogenetic relatedness on movement 251 patterns and life history traits. 252

254
Core-transient species designation 255 During the 21-year study period, we captured 12,651 individuals from the 21 species 256 included in the analysis (Table 1). Based on the proportion of years that each species was 257 present, we placed species into three temporal persistence categories ( Figure 2) Figure 3; Figure S1). Transient species that had few long-distance movements 275 may be attributed to high mortality, low detectability on the site, low recapture due to rapid 276 movement off the site, or a combination of these. Core species tended to move shorter distances 277 Reproductive results from the field data were best explained by phylogeny. For all 294 species in Heteromyidae (5 core and 3 non-core), the majority of captured females were never 295 14 recorded in reproductive condition (Table 1; Figure S2). However, despite generally much lower 296 abundance, species in Cricetidae were observed in reproductive condition more often. For 297 example, nearly 50% of Peromyscus eremicus (core) and P. maniculatus (non-core) were 298 recorded in reproductive condition at least once per year (Table 1) and N. albigula (core) females 299 were often found in reproductive condition. However, Sigmodon (non-core) females were almost 300 never recorded as reproductive ( Table 1). The lack of observed reproduction may suggest that 301 Sigmodon rarely reproduce at the site or that sampling error associated with the small number of 302 captures affected our results. Onychomys (core) females were rarely captured when reproductive, 303 but other data suggest that O. torridus may reproduce multiple times per year (Table 1, Table 2). 304

305
PCA results suggested that species can be grouped in multivariate space by their traits 306 and core-transient status (Fig 4), and that traits appear to be strongly conserved within family 307 ( Fig A3). Phylogeny (family) was a significant predictor of ψ (linear model; lm (ψ~family), 308 p=0.016, r 2 =0.28), Φ (lm (Φ~family), p=0.006, r 2 =0.35), and mean abundance (lm 309 (abundance~family), p=0.004, r 2 =0.37), but not for the proportion of years a species was present. 310 Mean abundance was positively related to the proportion of years a species was present in the 311 study area when phylogeny was controlled for (bm gls, p=0.005). Body size was not a significant 312 predictor for the proportion of years present, mean abundance, survival, reproduction, or 313 movement (bm gls, p >> 0.05). We did not detect strong movement-survival trade-offs or 314 movement-reproduction trade-offs in this community using linear regression or phylogenetic 315 methods (Figs A4, A5). There was a relationship between observed modal movement distance 316 and temporal persistence (lm(distance mode ~ proportion years present), p=0.03, r 2 =0.23; bm 317 gls, p=0.05) where species that persisted longer moved shorter distances. 318

320
Our study provides the first test, to our knowledge, of whether life history traits are 321 associated with the temporal persistence patterns of species in a community. Life-history traits 322 were generally conserved within evolutionary lineages, which in our system are also related to 323 higher or lower degrees of adaptation to desert environments. Our results provide some Unfortunately, it is not possible to definitively distinguish between these groups using our data, 352 but using a combination of our results and the well-studied natural history of these species, we 353 can make some informed predictions. Species that exhibited trait correlations more similar to 354 core species could be source-sink transients (Lenormand 2002 would be predicted to exhibit relatively small change in response to the same scenario. In the 406 drive to better understand the response of biodiversity to perturbations, a temporal perspective of 407 species demographics and persistence represents a critical link in identifying the linkages 408 between local and regional richness patterns and predicting community response to change.  for each bin, but that there is a large difference in total number of individuals captured in each 571 group, that is not represented in the histograms (but see Table 1)