Environmental Factors Determinate Roadkill Levels of the Endemic Iberian Species, Iberian Hare (Lepus Granatensis)

16 Lepus granatensis is an Iberian Peninsula endemic species and one of the most 17 important small game species. We surveyed Iberian hare-vehicle accidents in roads 18 network in southern Spain, analysing the Mediterranean landscape, the main habitats 19 of this species. We recorded roadkill of roads during 6-month, compared hare roadkill 20 densities to hare hunting yields. We analyzed the spatial patterns and factors that 21 could be influencing the hare road kill. We detected blackspots of hare road kill in 22 areas with high landscape heterogeneity and included embankments, intersections 23 roads and high traffic intensity. The hare roadkill ranged between 6% and 41% of the 24 annual harvest of hares killed on neighbouring hunting estates. We therefore consider 25 it highly relevant to take into account the hare road kill, especially in hare hunting 26 areas, suggesting to gamekeepers and managers addressing the issue of road kill of 27 hares. It would be necessary that hunting quotas be adjusted in territories where the 28 additive effect of these non-natural hare mortalities converge. Results point to future 29 directions for applied research in road ecology, which would include demographic 30 compensation and roadkill mitigation. Our methodology could be of wide use to 31 identify lagomorphs’ road kill blackspots by analysing environmental spatial patterns.

Roads were surveyed weekly for a total period of six months, between March 1, 134 2003 andJuly 31, 2003. This is the period of maximum reproductive activity for hares 135 in the region (26). Three surveyors were present in each survey (always the same 136 during the study period to avoid inter-observer biases). Surveys were driven in a car 137 vehicle at 10 km/h. To survey by foot was not allowed by the police in these roads. 138 Before undertaking the first sampling, we cleaned and removed all carcasses on the 139 sampled road sections. Surveys were carried out at dawn. We recorded the UTM 140 coordinates of each collision point using a GPS eTrex Vista Cx (Garmin, USA). 141 When a carcass was detected, we removed it to avoid double counting during 142 subsequent sampling. All other wild species killed by vehicle collisions were also 143 recorded. Kill rates were standardized as the number recorded per 100 km (19). 144 We investigated whether hare roadkills were aggregated in certain road sections 145 i.e., blackspots. We considered two approaches for this estimating the possible 146 aggregation in 100-m and 500-m road sections. These distances have been proposed 147 to be as far as the road habitat disturbance effect reaches (4). In addition, we also 148 computed the density of roadkills in two 100m and 500m buffer radios, as the number 149 of hares killed per km 2 . 150 To compare hunting bags with hares killed on roads, we used the annual hunting 151 reports (AHRs) for the period 1993 to 2001 from 181 game estates. These were all the 152 fame estates that in the seven municipalities which were traversed by the sampled 153 roads. We analysed 1,282 AHRs from these game estates and estimated the hunting 154 yield (HY) as ∑ mean annual number of hares hunted per game estate / ∑ areas of the 155 game estates in km2 (40; 41; 42). Hunting data were taken from Farfán (32). 156 157 Roadkill modelling 158 The number of collisions for any target species depends on a number of factors 159 related to road features and traffic volume (43; 44), animal behaviour and phenology 160 (45; 46; 7). The surrounding habitat structure and landscape can also play an 161 important part (1; 14; 47). To consider the incidence of these possible factors in our 162 sample, we overlaid hare collision points on habitat maps derived from digital 163 orthophotographs (scale 0.5 m/pixel) using ArcGis 9.3 software (Esri, USA). All 164 roads containing the collision points were also digitized onto the habitat maps. 165 We measured variables related to road features, surrounding habitat and 166 landscape ( We also generated random points without hare-vehicle collisions on these roads 181 as controls in the statistical tests. We applied the same procedures as used with the 182 buffers and environmental variable measurements. 183 184

Data analysis 185
We tested if the spatial pattern of collisions in road sections fitted a pattern 186 expected at random through the Wald-Wolfowitz run test (54). If the random 187 hypothesis was rejected, we estimated a spatial index of dispersion as the 188 variance/mean ratio. If this ratio yielded values >1 hares roadkills were dispersed as 189 contagiously objects (55) in those road sections. 190 To detect potential multicollinearity between variables, we developed a 191 correlation matrix and obtained a Spearman's rank correlation coefficient (rho). Based 192 on this value, the coefficient of determination (R²) and the value of the Variance 193 Inflation Factor (VIF) were calculated to measure possible collinearity between 194 variables (VIF >5, 56), removing one of the variables involved in the cases. Only 195 those that captured the effects of any set of highly correlated variables were allowed 196 to continue. The VIF statistic was calculated as: 197 We generated predictive models for hare roadkills using a GLMM with a 199 binomial error distribution and a logit link function (57)

208
A total of 1,336.8 km of roads were sampled during the study period. The field 209 effort involved 171.9 observer-hours. Over a period of 6 months, we recorded a total 210 of 162 dead animals; 68.5% Iberian hares, 17.9% wild rabbits, 4.9% other mammals, 211 5.6% birds, and 3.1% reptiles (Table 3). Of the 111 hares found dead near the roads, 212 only 80 could be clearly attributed to a vehicle collision; these were taken into 213 account for further analysis. We estimated a standard kill rate for the area of 6.0 214 hares/100 km. 215 216 Hare roadkills were not randomly distributed neither in 100-m road sections 220 (test Wald-Wolfowitz; N = 552; Z = -5,782; p < 0,001) nor in 500-m sections (test 221 Wald-Wolfowitz; N = 113; Z = -4,024; p < 0,001) suggesting the possible existence 222 of black spots. However, the variance/mean ratio was only >1 in the 500-road sections 223 (0.71 ± 0.12 hares killed per section; s2 = 1,21) confirming the existence of black 224 spots at least in road sections from this size onwards. A total of 68.7% of the hare 225 accidents were concentrated in 18.8% (10.5 km) of the road network sampled (Fig. 2). 226 Density of hares killed on the roads was 4.6 ± 0.5 hares/km2 in 100-m buffers 234 and 0.9 ± 1.4 hares/km 2 in 500-m buffers. The hunting yield in neighbouring game 235 estates was 15.1 ± 14.8 hares/km 2 . Therefore, roadkills can account between 8% -236 40% of hares hunted in the area (Table 4). 237 238

Implications of the hare's roadkill on its conservation 258
Our results indicated that in the study area more than half of all reported 259 vertebrate roadkills were Iberian hares. This frequency is higher than that observed 260 numbers for ungulates or other medium and large-sized mammals (59; 1) and for 261 other hare species (19; 20). The standard kill rate is also almost five times higher than 262 for other studied hare species (18; 17) and noticeably striking in contrast with the 263 represent 31.1% of the regional landscape (64) as well as olive groves, vines and 280 cereals cover 26.6% of the soils (65). Therefore, the landscape conditions, soil use, 281 and road network in the study area are typical of almost a third of the region. 282 An important third limitation of the study arises from research evidences 283 published some years after our sampling period. Santos et al. (66) shown that small 284 mammal carcasses, like lagomorphs, do not persist on roads more than one to two 285 days, being lower specially in summer due to same factor as scavengers, weather 286 conditions or people removing them. This suggest that for a study like ours the 287 optimal monitoring frequency should have been daily and not weekly. Further, this 288 divergence in the weekly vs daily sampling could involve an important false negative 289 rate in the estimated hotspots, missing "true" hotspots (67). In the case of lagomorphs, 290 these authors suggest an underestimation of about 40% which implies that our 291 estimated roadkill mortality rate is much lower than real, and some hotspots were 292 missed. In any case, this divergence evidences the relevance of hare road mortality in 293 the area and enhances our results because an underestimation supposes a higher 294 source of additional mortality and bias to hunting planning. 295 An additional consideration about the representativeness of our results is related 296 to hare density. It has been argued that road casualty counts are not correlated to 297 population densities and that traffic flow is the most important explanatory variable 298 considering variance in road accidents for certain taxa (68; 69). However, although 299 traffic has a role to play, in the case of wild rabbits a density dependent relation 300 between roadkills and the population living in wider landscapes has been shown (70; 301 71). This direct relationship may suggest something similar is affecting the Iberian 302 hare. In fact, this relationship is true for hare hunting yields (72) since higher hare 303 densities and more yields in fertile plains in which dry wood crops and irrigated 304 herbaceous crops are intensively managed (23) such as in our study area. 305 Finally, although our study period was short it agrees with the species' (2), landscape heterogeneity is the main factor influencing Iberian hare fatalities. 355 Mixed patches of forested areas with pastureland or farmland create habitat mosaics 356 where an increase of resource availability for wildlife tends to increase the presence 357 of species and then the likelihood of crossing nearby roads (11; 1). The proximity of 358 forests to open areas is also a key factor in collisions (89). Road borders and verges 359 may act as feeding areas for some species (14) as well as they occur during regular 360 animal movement in their home range (90). 361 Forested areas in our study area are made up by olive groves, a woody crop that 362 allow water to become available to smaller species due to their widespread trickle 363 irrigation systems, as well as providing food resources (91). It is also worth 364 considering that weather conditions or seasonal variations affecting food availability 365 influence roadkill rates (92). This is a likely effect in the dry summers of the study 366 area that could push hares to cross roads looking for available water. Vineyards also 367 provide refuge, water and food. It should be noted that the Iberian hare follows a 368 heterogeneous habitat selection pattern and move frequently between habitat patches 369 (22). Therefore, hares may cross roads in points of high landscape heterogeneity in 370 search of food (i.e.: herbaceous crop shoots, weeds or early summer grapes); looking 371 for roadside vegetation or road verges (8); because of changes in food availability due 372 to harvesting (93), drove by farm works or machinery (94); or simply because it's the 373 rooting season and they are looking for mates in high diversity patches. 374 We identified the presence of embankments as the second factor favouring hare 375 accidents. Some authors have suggested that embankments may act as barriers that 376 prevent animals from crossing the road (95), and collisions occur when the road and 377 the adjacent landscape are at the same level (96). However, the difference in level in 378 the case of the sampled roads was always less than 1 m and in most cases was even 379 less than 0.5 m. We think that these differences do not prevent hares from accessing Iberian hares that include traffic and road mortality, before estimate hunting quotas. In 425 addition, given that hare hunting may accrue economic benefit for the local estates, a 426 proposed solution in an inter-disciplinary field like road ecology must consider also 427 whether reducing hunting quotas has any repercussions for estates' profit. 428 429

Conclusions 430
Finally, the results and discussion point to future directions for theoretical and 431 applied research in road ecology, which would include demographic compensation 432 and roadkill or the assessment of specific mitigation measures to protect lagomorphs. 433 Due to the Iberian hare distribution as well as the common design of two-lane roads, 434 we consider also that our methods and results could help to the management of roads 435 and associated landscape throughout the national territory. In summary, this work 436 proposes a method that allows detecting favorable spatial configuration observing a 437 priori the spatial structure of a territory as preventive tools, and a propose a 438 modification in the management of hunting quotes only in territories with that   Location of the study area in the northeast of Malaga province to the southern of Spain. Stretches of highways analysis for hare roadkills (55.7km). Note: The designations employed and the presentation of the material on this map do not imply the expression of any opinion whatsoever on the part of Research Square concerning the legal status of any country, territory, city or area or of its authorities, or concerning the delimitation of its frontiers or boundaries. This map has been provided by the authors.

Figure 2
Spatial study context. Grey and red circles indicate the points with hare roadkill events; the red circles, those hare roadkill points that we have added a photograph with the around 100m buffer habitat (showing heterogeneous habitats). Rectangles indicate some points without hare roadkill detected which we have added a photograph with the 100m buffer habitat (showing homogeneous habitats).