Human subsidies facilitate hyperpredation of Mediterranean island wildlife by outdoor cats

1 1 Human subsidies facilitate hyperpredation of Mediterranean 2 island wildlife by outdoor cats 3 4 Jeffrey R. Ferrer1*, John Vandermeer2&, Collin J. Richter1 & Erin R. Baldwin1&, Johannes 5 Foufopoulos1& 6 7 1School for Environment and Sustainability, University of Michigan, Ann Arbor, Michigan, 8 United States of America 9 2Department of Ecology and Evolutionary Biology, University of Michigan, Ann Arbor, 10 Michigan, United States of America 11 12 *Corresponding Author 13 [email protected] (JRF) 14 15 & Collin J. Richter, John Vandermeer, and Erin R. Baldwin contributed equally to this work 16 17 18 19 20 .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted July 31, 2025. ; https://doi.org/10.1101/2025.07.27.667076doi: bioRxiv preprint 2 21 Abstract 22 Domestic cats (Felis catus) are avid wildlife predators and one of the most harmful invasive 23 species. The lethal impacts of an introduced predator such as cats on wildlife can be further 24 exacerbated by the introduction of an additional abundant non-native prey species capable of 25 supporting an exceptionally dense predator population, a phenomenon known as hyperpredation. 26 A special case of hyperpredation involves human food subsidies, when an invasive predator 27 impacting native wildlife is supported not by another invasive taxon, but by human-derived food 28 sources. To test whether access to anthropogenic food subsidies by cats is causing hyperpredation, 29 mark-recapture methodology was used to measure twelve cat populations experiencing a gradient 30 of human subsidies on the Mediterranean island of Naxos, Greece. Line-transect surveys were 31 conducted at each site to measure reptile population abundance across this gradient, and other 32 factors including human density and distance of reptiles from villages were considered as well. 33 Strong evidence was found that the population size of cats is a direct result of the density of 34 anthropogenic food available to them, and a corresponding decline in the size of reptile populations 35 as cat density increased was observed. It was also shown that as cat populations exceed the 36 available supply of human food, the negative effects of cat density on reptile populations are 37 exacerbated. These results demonstrate that access to anthropogenic subsidies has allowed cat 38 populations to expand to exceptional levels, driving hyperpredation on island wildlife. 39 Introduction 40 Invasive species are one of the greatest threats to biodiversity worldwide [1-3]. Invasive 41 species have been implicated in over 50% of all modern extinctions whose causes are known, and 42 in about 20% of total cases, invasive species have been identified as the sole driver of extinction .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted July 31, 2025. ; https://doi.org/10.1101/2025.07.27.667076doi: bioRxiv preprint 3 43 [4-5]. One of the most harmful groups of invasive species are mammals, who currently threaten 44 more IUCN listed critically endangered terrestrial vertebrate species than any other group, an 45 impact driven primarily by black rats (Rattus rattus) and domestic cats (Felis catus) [6]. 46 Cats have been linked to ≥ 63 vertebrate extinctions [7] and are recognized as one of the 47 most harmful invasive species due to their global distribution and devastating impacts on wildlife 48 [8]. In the United States alone, it has been estimated that cats kill billions of birds, mammals, and 49 reptiles each year, with mortality driven primarily by unowned free-roaming cats, rather than pets 50 [9]. In addition to direct predation, cats can have many other negative effects on wildlife such as 51 influencing prey behavior by increasing prey wariness and decreasing foraging time [10-11], 52 transmission of diseases such as rabies, toxoplasmosis, cutaneous larval migrans, tularemia, and 53 plague [12], and loss of genetic integrity in native species such as the Eurasian wildcat (Felis 54 silvestris silvestris) through hybridization [13]. 55 The damaging effects of cats are particularly prevalent on islands [14-15]. Despite making 56 up only 5.3% of the Earth’s landmass, 61% of all known extinctions have occurred on islands, and 57 invasive species have been cited as the most frequent cause of insular species extinctions [16]. 58 Island communities are disproportionately affected by invasive mammalian predators, and 59 especially cats, where a general lack of native predators typically results in down-regulated 60 antipredator responses, rendering island wildlife relatively tame and easy to capture [17-19]. Cats 61 have emerged as the causal factor in at least 33 extinctions of insular vertebrate species [9], and 62 because they have already been introduced to most suitable islands worldwide [14], they represent 63 a threat to numerous additional island endemics [10]. 64 The impacts of an introduced predator on native prey can be exacerbated by the 65 introduction of an additional non-native prey species, a phenomenon known as hyperpredation .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted July 31, 2025. ; https://doi.org/10.1101/2025.07.27.667076doi: bioRxiv preprint 4 66 [20-23]. When this occurs, a rapidly reproducing non-native prey species supports an unusually 67 dense population of predators. This can have severe consequences for the native prey species, 68 which may be decimated by the resulting increase in predators. For example, the extinction of the 69 endemic Macquarie Island Parakeet (Cyanoramphus novaezelandiae erythrotis) is thought to have 70 occurred following hyperpredation by cats. Although cats were introduced onto Macquarie Island 71 within ten years of its discovery in 1810, the native parakeet withstood predation pressure and 72 remained numerous there until about 1880. In 1879, European rabbits (Oryctolagus cuniculus) 73 were released onto the island and reproduced rapidly. Within a few years, cat populations 74 persisting primarily on rabbits expanded rapidly and opportunistically consumed parakeets. The 75 parakeet was last definitively seen in 1890 [18, 20, 24]. Hyperpredation has since been recorded 76 in several systems [25-28], and understanding how to predict and manage it will only increase in 77 importance as the spread of invasive species is expected to increase with the expansion of global 78 trade networks [29]. 79 A related situation can occur when an invasive predator, impacting native wildlife, is 80 supported not by another invasive taxon, but by human-derived food sources [30] (Fig 1). In many 81 urban environments, densities of predators increase simply through ease of access to anthropogenic 82 resources like trash and hand-outs [31-33]. For example, while cats are typically solitary and 83 territorial hunters, the presence of large quantities of a stable food resource can alter their behavior 84 by reducing their home-range size and territoriality, increasing tolerance of home-range overlap, 85 and leading to the formation of cat colonies [34-35]. While densities of cats can vary greatly, 86 Liberg et al. (2000) [31] found that densities above 100 cats/km 2 are only found in urban areas 87 with a constant source of anthropogenic food supplementation. These unnaturally high densities 88 of cats can potentially result in a unique form of hyperpredation where the introduced ‘prey’ that .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted July 31, 2025. ; https://doi.org/10.1101/2025.07.27.667076doi: bioRxiv preprint 5 89 is supplementing the predator population is simply anthropogenically derived food. As cats have 90 been shown to have high fidelity to a single feeding site [34], it is expected that the number of 91 feeding sites available to cats should be correlated with the size of the cat population. However, 92 this has yet to be demonstrated, and the extent to which increased food supplementation affects 93 greater impacts on wildlife remains unclear as for example, more subsidies may simply result in a 94 switch away from wildlife to anthropogenic sources. 95 Fig 1. Conceptual diagram showing how the addition of anthropogenic food supplementation 96 into a predatory-prey relationship can result in hyperpredation on wildlife. 97 Cats were first introduced to the Mediterranean islands thousands of years ago [36-38] and 98 are playing an important predator role in both urban and rural Mediterranean ecosystems [39-40] 99 where they have been implicated in significant wildlife mortality [11,41,42]. The goals of this 100 study were to establish which factors support cat populations in a representative Mediterranean 101 ecosystem (the island of Naxos in the Aegean Sea, Greece), and to determine how cats are affecting 102 resident wildlife populations while considering other factors that may shape a possible relationship 103 between cats and wildlife. 104 Methods used to answer these questions included, 1) obtaining population and density 105 estimates of twelve discrete cat populations experiencing a range of food supplementation, 2) 106 quantifying the amount of food supplementation available to each cat population, 3) surveying 107 reptile communities in the natural habitat in the immediate vicinity of each of these locations, and 108 4) providing visual support through the development of heat maps showing the spatial distribution 109 of cats. 110 By investigating these population level relationships along a gradient of food 111 supplementation within a representative island system, this study is taking a novel and fine-grained .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted July 31, 2025. ; https://doi.org/10.1101/2025.07.27.667076doi: bioRxiv preprint 6 112 approach that will allow a quantitative demonstration of the connection between human food 113 subsidies and decreases in wildlife in Mediterranean wildlife. 114 Methods 115 Study Site 116 Field work was conducted during the months of May and June 2024 on the Aegean Island 117 of Naxos, Greece. Naxos (430km2) is the largest island in the Cyclades archipelago with a typical 118 Mediterranean climate consisting of warm dry summers, and mild moist winters. The landscape 119 of Naxos is characterized by xeric phrygana, a type of thorny summer-deciduous Mediterranean 120 cushion scrub community dominated by Sarcopoterium spinosum and Genista acanthoclada, as 121 well as high evergreen maquis dominated by Quercus coccifera and Juniperus turbinata. The wide 122 distribution of these communities is consistent with a long history of disturbance and degradation 123 of the original oak forest habitats through grazing and agricultural activities on the island [43]. 124 Areas near villages consist of a mosaic of these communities mixed with irrigated gardens, 125 grainfields, as well as olive groves. 126 Cycladic wildlife is dominated by reptiles which, by virtue of their life history, are 127 particularly well suited for the Mediterranean climate [44]. Naxos is home to 13 reptile species, 128 including two introduced species (Hemidactylus turcicus and Chalcides ocellatus). While reptiles 129 are overall common, communities on the island are dominated by the Aegean wall lizard (Podarcis 130 erhardii), hereafter “wall lizard”, which can easily be observed. Both wall lizards and other reptile 131 taxa occur in particularly high densities along the drystone walls and terraces which act as focal 132 points of reptile activity, and are a ubiquitous feature of the Mediterranean landscape [45-46]. 133 Reptile populations in the region are likely shaped by a combination of both bottom-up, as well as 134 top-down effects. Among lizard predators, snakes are probably the most important [47]. The only .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted July 31, 2025. ; https://doi.org/10.1101/2025.07.27.667076doi: bioRxiv preprint 7 135 other mammalian predator besides cats are stone martens (Martes foina), although because of their 136 strictly nocturnal activity pattern, they do not likely encounter the broadly heliothermic reptiles. 137 Because of this high aggregate biomass, reptiles are a particularly important component of 138 Cycladic ecosystems and constitute an ideal system to investigate the impacts of cats on local 139 species communities. 140 Naxos contains over 30 towns and villages that range in population size from being 141 essentially abandoned, to the capital of Chora (with ca. 6000 inhabitants) [48]. Naxian settlements 142 have traditionally been built in a very compact manner, facilitating defense against pirate raids 143 [49]. They typically encompass a dense network of stone-lined footpaths that allow access to every 144 part of the village. Household refuse is collected in select sites located at the edge of a settlement 145 where garbage truck access is possible, creating distinct hot spots of waste accumulation. At the 146 same time, there is a clear border and very sharp transition from the last outer row of houses to the 147 surrounding agricultural matrix where resident wildlife occurs. 148 Cats can be found in large numbers in nearly every town on Naxos. At the same time, 149 because of Naxos’ arid landscape, cats are present but rare in the remote countryside, especially 150 more than 1-2km away from human habitations [11]. They appear to be highly dependent on the 151 shelter of human settlements where they readily consume nutritionally dense cat food provided by 152 residents as well as garbage from open dumpsters (S1 Appendix). These conditions are 153 representative of many coastal areas around the Mediterranean Basin [40]. Within villages, cats 154 exist across a spectrum of associations with humans, ranging from fully domesticated and fed to 155 completely ignored and even persecuted (personal observation). Even when affiliated with a 156 household, cats on Aegean islands are traditionally kept as outdoor pets and are allowed to roam 157 freely. Because they are typically fed in a yard or in the street, and because of the open nature and .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted July 31, 2025. ; https://doi.org/10.1101/2025.07.27.667076doi: bioRxiv preprint 8 158 dense path network of island villages, it is possible to obtain a good estimate of the extent of food 159 supplementation in each village. 160 This study focuses on 12 representative villages selected randomly as study sites (Fig 2, 161 S1 Table) while ensuring that the following criteria were met: 1) presence of a discrete village 162 edge and a sufficiently dense network of village paths allowing easy access and high visibility to 163 the entire settlement, and 2) the presence of dry-stone walls extending radially away from the 164 village into the surrounding habitats to facilitate reptile surveys. Two of these rock walls ( ≥ 200m 165 in length) were randomly selected for surveys for each village, except for one smaller village that 166 contained a single wall. 167 Fig 2. Map of the study area. Survey locations are indicated by black stars. Naxos, Greece and 168 surrounding islets shown in main map. Entire country outlined in the inset, with a square around 169 Naxos for reference) Map created in ArcGIS Pro v3.4 (2024). Produced with custom ESRI 170 National Geographic Basemap (2025) [50-51]. 171 Reptile surveys 172 Reptile population abundance for each village were quantified through standardized 173 surveys. Each survey was conducted along 200-meter transects parallel to drystone walls, which 174 have been shown to act as foci for reptile activity in the region [11,52]. All surveys were completed 175 during the months of May and June 2024 during peak reptile activity time (8am – 12pm), and 176 under favorable weather conditions [53]. Weather metrics (temperature, wind speed, relative 177 humidity, cloud cover) were collected at the beginning and end of each survey (Kestrel 5000 178 Environmental Meter, 2024, Nielsen-Kellerman Company, Boothwyn, Pennsylvania). Surveys 179 began at the edge of a village and were conducted by slowly walking along the length of the 180 drystone wall. Species name and distance (m) from the origin were recorded (Garmin, 2024, .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted July 31, 2025. ; https://doi.org/10.1101/2025.07.27.667076doi: bioRxiv preprint 9 181 GPSMAP® 65s Handheld GPS device, Olathe, Kansas) for all reptiles observed on the wall or on 182 the ground within 1 meter from the wall (using ‘unknown’ if no species ID was possible). Each 183 wall received four surveys spread over the duration of the field season. 184 Cat surveys 185 Cat surveys were conducted for three consecutive nights within each village. Surveys were 186 performed in the evening during the time of peak cat activity (5pm – 8pm) and under appropriate 187 weather conditions [54]. Surveys were conducted by walking along every path in a village once, 188 and taking a picture and recording the GPS location of every cat observed. Survey tracks were 189 recorded and uploaded into ArcGis Pro v3.4 [50] to confirm comprehensive coverage of each 190 village. Great individual variability in terms of cat color, size, condition, and unique markings 191 allowed cats to be identified to the level of individual [55]. Each cat was given a unique identifier 192 code, and on the second and third survey nights, cats were designated as new individuals or 193 recaptures from previous nights. Cat population size in each village was calculated using the 194 Schnabel Index, a method used to estimate population size in mark-recapture studies when 195 multiple sampling events are conducted [56-57]. Kittens were also recorded in the surveys but 196 were excluded from the population size calculation due to their secretive nature making 197 comprehensive identification difficult. 198 Additional field data 199 During the first cat survey conducted at each village, the locations of all food available to 200 cats in the form of either pet food dishes or communal garbage dumpsters were recorded. The 201 aggregate number of food dishes and dumpsters were combined into a number of ‘food stations.’ 202 Human population data for every community on the island were provided by the municipality of 203 Naxos and confirmed through a review of the Greek National Statistical Agency. The area of each .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted July 31, 2025. ; https://doi.org/10.1101/2025.07.27.667076doi: bioRxiv preprint 10 204 village (ha) was quantified in ArcGIS Pro v.3.4 by manually digitizing the perimeter of each 205 village. Larger villages have more cats, as well as humans and feeding sites, necessitating the need 206 to establish a correction for community size. To correct for area effects, cat population, number of 207 food stations, and human population were divided by the village surface area. 208 Statistical Analyses 209 Statistical analyses were performed in R version 4.4.2 [58] with the following the lme4 210 [59] and AICcmodavg [60] packages with the significance threshold set at 0.05. 211 Cat density 212 Ordinary least squares regression (OLS) was used to first explore univariate associations 213 between cat density, human density, and food station density and then to develop more complex 214 models that included combinations of these predictors and their interactions. Corrected Akaike 215 information criterion (AICc) [61] was used to determine the best fit model. 216 Reptile populations 217 A cat ‘satiation index’ was calculated for each village as the residuals of the regression of 218 cat density against food density. Positive residuals were interpreted as cats being in a food deficit, 219 and therefore hungrier, while negative residuals were interpreted as cats being in a food surplus 220 and more satiated. The residual values were rescaled so that the most satiated (index = 0) had a 221 value of 0, and higher positive values indicated lower satiation, or increased hunger. This index 222 was used as a model parameter to investigate whether level of cat satiation influenced reptile 223 populations. 224 To analyze the overall effect of distance from a village on reptile populations, all reptile 225 observations were combined, and the total number of reptiles observed at each 1-meter interval on 226 the transects were calculated. These data met the assumptions of a linear regression, with a linear .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted July 31, 2025. ; https://doi.org/10.1101/2025.07.27.667076doi: bioRxiv preprint 11 227 relationship between count and distance, normally distributed residuals, and homoscedasticity. 228 Therefore, OLS was performed to analyze the association between distance and reptile counts for 229 the combined dataset. Surveys for the village that only had one rock wall were given twice as much 230 weight to prevent reptile counts for this village to be smaller due to reduced sampling effort. To 231 further test whether cat density influenced the effect of distance on reptile counts, the number of 232 reptiles observed at each 1-meter interval were summed within each village, and a Poisson 233 regression was run for each village with reptile counts regressed against distance. The coefficient 234 for distance was exponentiated to obtain a slope, and the slopes were regressed against cat density 235 to investigate the association between cat density and the change in reptile counts with distance. 236 The association between reptile counts and cat density, cat satiation, and distance were 237 analyzed in two separate analyses. For analysis one, reptile counts in each village were grouped 238 into eight 25-meter distance bins. Then, through Poisson generalized linear mixed effect models 239 (GLMM), with village ID included as a random effect, combinations of model parameters were 240 explored to determine the best fit model, with AICc criteria used to determine model fit. An 241 interaction term between cat density and satiation index was included in model exploration as well. 242 For analysis two, reptile counts were analyzed at the survey level, through Poisson generalized 243 linear mixed effect models (GLMM) with village ID included as a random effect. These models 244 were used to analyze the association between the number of reptiles observed in each survey and 245 cat density and cat satiation index across all villages without distance. An interaction term between 246 cat density and satiation index was explored in this analysis as well. 247 Results 248 Cat density analysis .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted July 31, 2025. ; https://doi.org/10.1101/2025.07.27.667076doi: bioRxiv preprint 12 249 Cat densities in all villages were higher than what is normal for the species under rural or 250 natural conditions (Fig 3, S2 Appendix) [31], except for one village with no cats that also had no 251 additional food supplementation or permanent human residents. 252 Cat density was significantly positively associated with both predictor variables when analyzed 253 independently- food density (adjusted R2 = 0.66, p = 0.009, n = 12, OLS, Fig 3), and human density 254 (adjusted R 2 = 0.45, p = 0.001, n = 12, OLS, Fig 4). When multiple predictor variables were 255 included in model selection, two models showed nearly equal fit for predicting cat density. The 256 first model (Model A) included only food density as a predictor (AICc = 65.54, AICcWt = 0.67, 257 Log Likelihood = -28.27, Table 1) and the second model (Model B) included both food density 258 and human density (AICc = 67.47, AICcWt = 0.26, Log Likelihood = -26.88, Table 1). Food 259 density was significantly positively associated with cat density in both models (adjusted R 2 = 260 0.657, p = 0.0001 for Model A and adjusted R 2 = 0.698, p = 0.014 for Model B, Multiple 261 Regression, Table 2), while human density was not significant in the model that included it. When 262 the type of feeding supplementation (garbage bins and food dishes) was separated, cat density was 263 significantly positively associated with both density of garbage bins (adjusted R2 = 0.28, p = 0.044, 264 OLS) and density of feeding dishes (adjusted R2 = 0.73, p = 0.0002, OLS). 265 Fig 3. Ordinary least squares regression showing association between cat density and feeding 266 density (p = 0.0008, adjusted R2 = 0.66). Shaded region represents 95% confidence interval. 267 Fig 4. Ordinary least squares regression showing association between cat density and human 268 density (p = 0.01, adjusted R2 = 0.45). Shaded region represents 95% confidence interval. 269 270 271 .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted July 31, 2025. ; https://doi.org/10.1101/2025.07.27.667076doi: bioRxiv preprint 13 Predictors AICc ΔAICc AICcWt. Log Likelihood Cat Density ~food density 65.54 0.00 0.67 -28.27 ~human density + food density 67.47 1.93 0.26 -26.88 ~human density 71.2 5.66 0.04 -31.10 ~ human density + food density + interaction 72.16 6.62 0.02 -26.08 null model 75.87 10.33 0.00 -35.27 272 273 Table 1: Model selection for factors predicting density of cats, Models are ranked in ascending 274 order by corrected Akaike corrected information criteria (AICc). ΔAICc, AICc weight, and Log 275 Likelihood are provided for each model. Bold selections show models with near equal fit. Model 276 variables include food density, human density, and an interaction between food. Statistic Estimate Std. Error t-value p-value Model A. Intercept 0.443 1.534 0.289 0.779 Food Density 3.624 0.771 4.702 0.0001* Model B. Intercept -0.675 1.6`4 -0.418 0.779 Human Density 0.068 0.045 1.532 0.160 Food Density 2.769 0.914 3.031 0.014* 277 278 Table 2: Results for cat density analysis using multiple linear regression showing two models with 279 similar fit for predicting cat density. Asterisks indicate significant predictors (p < 0.05). Estimates, 280 standard errors, and z-values are also shown. Model A adjusted R 2 = 0.657. Model B adjusted R2 281 = 0.698. 282 Reptile population analysis .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted July 31, 2025. ; https://doi.org/10.1101/2025.07.27.667076doi: bioRxiv preprint 14 283 A significant positive correlation was found between reptile counts and distance from the 284 village edge, indicating that reptile numbers declined closer to human habitations and cats 285 (Pearson’s R = .2176, p = 0.002, OLS, n = 200, Fig 5). 286 Fig 5. Ordinary least squares regression showing association between the number of reptiles 287 observed at each 1-meter interval distance from village for the entire study (p = 0.002, adjusted R2 288 = 0.43). Shaded region represents 95% confidence interval. 289 For analysis one, when variables with the binned dataset were examined individually, 290 Poisson GLMM’s showed that both cat density (marginal R 2 = 0.33, conditional R 2 = 0.82, p = 291 0.007, n = 96, Poisson GLMM), and satiation index (marginal R2 = 0.38, conditional R2 = 0.82, p 292 = 0.002, n = 96, Poisson GLMM) were significantly negatively associated with reptile counts. 293 When models that included different combinations of predictor variables and their interactions 294 were compared, the model that best fit the data included all of the predictor variables; distance (p 295 = 0.0003), cat density (p = 0.053), satiation index (p = 0.212), as well as the interaction between 296 cat density and satiation index (p = 0.001) (marginal R 2 = 0.67, conditional R 2 = 0.83, AICc = 297 569.11 and AICcWt = 0.87, Tables 3 & 4, compared to next best model AICc = 574.91, and 298 AICcWt = 0.05,). Hence, distance and the interaction term were found to be significant, while cat 299 density and satiation index were not. 300 For analysis two, when variables were examined individually, cat density (marginal R 2 = 301 0.33, conditional R2 = 0.83, p = 0.008, n = 96, Poisson GLMM, Fig 6) and cat satiation (marginal 302 R2 = 0.39, conditional R 2 = 0.82, p = 0.002, n = 96, Poisson GLMM, Fig 7) were significantly 303 negatively associated with the number of reptiles observed in a survey . The full model that best 304 fit the data included cat density (p = 0.047) cat satiation (p = 0.218), and the interaction between 305 cat density and cat satiation (p = 0.001) (marginal R2 = 0.67, conditional R2 = 0.83, AICc = 686.08 .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted July 31, 2025. ; https://doi.org/10.1101/2025.07.27.667076doi: bioRxiv preprint 15 306 and AICcWt = 0.89, Tables 3 & 5, compared to the next best model AICc = 692.09 and AICWt = 307 0.04, Table 3). In this model, cat satiation was not significant, cat density was significant, and there 308 was a significant negative interaction between cat density and cat satiation. 309 Fig 6. Poisson generalized linear model showing association between reptile counts per survey 310 and cat density. Shaded region represents 95% confidence interval for fixed effects, p = 2e-16, 311 null deviance = 544.02 on 95 degrees of freedom, residual deviance = 462.78 on 94 degrees of 312 freedom. 313 Fig 7. Poisson generalized linear model showing association between reptile counts per survey 314 and satiation index. Shaded region represents 95% confidence interval for fixed effects, p = 2e-16, 315 null deviance = 544.02 on 95 degrees of freedom, residual deviance = 440.90 on 94 degrees of 316 freedom. 317 318 319 320 321 322 323 324 325 .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted July 31, 2025. ; https://doi.org/10.1101/2025.07.27.667076doi: bioRxiv preprint 16 Reptile Count ~ AICc ΔAICc AICcWt. Log Likelihood A Reptile Analysis 1 ~ distance + cat density + cat satiation + interaction 569.11 0.00 0.87 -278.09 ~ distance + cat satiation 574.91 5.79 0.05 -283.23 ~ distance + cat density + cat satiation 575.04 5.93 0.04 -282.19 ~ cat density + distance 576.16 7.05 0.03 -283.86 ~ cat density + cat satiation + interaction 579.63 10.51 0.00 -284.48 ~ distance 579.83 10.71 0.00 -286.78 ~ cat satiation 585.52 16.4 0.00 -289.63 ~ cat density + cat satiation 585.61 16.49 0.00 -288.58 ~ cat density 586.77 17.66 0.00 -290.26 null model 590.48 21.37 0.00 -293.18 B Reptile Analysis 2 ~ cat density + cat satiation + interaction 686.08 0.00 0.89 -337.71 ~ cat satiation 692.09 6.01 0.04 -342.91 ~ cat density + cat satiation 692.24 6.16 0.04 -341.90 ~ cat density 693.59 7.51 0.02 -343.66 null model 697.27 11.2 0.00 -346.57 326 327 Table 3: Model selection for factors predicting number of reptile observations from two different 328 analyses. Model A shows results from reptile analysis 1 and Model B shows results from reptile 329 analysis 2. Models are ranked in ascending order by corrected Akaike corrected information 330 criteria (AICc). ΔAICc, AICc weight, and Log Likelihood are provided for each model. Model 331 variables include distance (model A only), cat density, cat satiation, and an interaction between 332 cat distance and cat satiation. 333 Heat maps .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted July 31, 2025. ; https://doi.org/10.1101/2025.07.27.667076doi: bioRxiv preprint 17 334 Heat maps showing the spatial distribution of cats, dumpsters, and food dishes were created 335 for each village in ArcGIS Pro v3.4 to provide visual supplementation to these results. Heat maps 336 include every cat observation across all three surveys in each village. These heat maps clearly 337 show that cats form high density aggregations near human-derived food resources. Due to large 338 cat population size, these results were best visually observed in the village of Vivlos (Fig 8), where 339 the highest cat density regions (red regions) are all associated with a dumpster or a food dish, while 340 none of the lowest cat density areas (blue regions) had a detectable source of human-derived food 341 within them. Results from heat maps from other villages with smaller populations were consistent 342 with these findings. 343 Fig 8. Heat map showing the spatial distribution of cats (n = 171 cat observations) in relationship 344 to human subsidies in a typical Naxian village (Vivlos). Black dots show all cat observations made 345 during three nights of surveys. Cats typically stayed in close proximity to sources of food, whether 346 deeding dishes or dumpsters Map created in ArcGIS Pro v3.4. Produced with ESRIi World Image 347 WGS1983 World Image Base Map (2025) [62]. 348 Discussion 349 Cats pose an increasingly pressing ecological problem across most island ecosystems 350 globally [14-15], yet managers and local decisionmakers still have a poor understanding of the 351 magnitude of the impacts, as well as the causative factors. By surveying outdoor cats along a 352 gradient of increasing food supplementation at twelve discrete sites within a single representative 353 Mediterranean ecosystem, this study demonstrated that cat population size is the direct product of 354 the quantity of human-derived food resources available to them. It was found that food resources 355 come primarily in the form of food scraps or dry cat food placed by local residents in their yards 356 or village paths. Equally important are the communal dumpster bins that are usually open and .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted July 31, 2025. ; https://doi.org/10.1101/2025.07.27.667076doi: bioRxiv preprint 18 357 sought out by groups of stray cats. These results demonstrate that cat densities in villages are 358 correlated with the density of both these resources. Beyond density of food resources, a separate 359 and distinct effect of human population on cat density was detected, potentially reflecting 360 additional benefits in the form of shelter and medical care some cats derive from a subset of each 361 community’s inhabitants. While previous studies have shown that cats will congregate and form 362 high density colonies around easily accessible food resources [34,63,64], these results further 363 demonstrate that cats are not just simply reorganizing themselves spatially around these resources 364 (Figs 4 & 10), but rather that population level cat density readily rises to match bottom-up human 365 supplementation effects. 366 These analyses assume that the cat population within each village constitute a discrete unit. 367 The existing literature indicates that while the home range size of cats varies widely [64-65], home 368 range decreases with increasing productivity of the landscape [66]. The high density of food dishes 369 and garbage sites within each village in this study (Fig 8) indicates that sufficient resources are 370 available to collapse home ranges and congregate cats within the individual village. The cat 371 surveys, conducted over three consecutive nights demonstrate high level of philopatry of 372 individual cats with most animals not moving more than a few dozen meters from day to day. The 373 assumption of closed village populations was further supported by a conspicuous absence of cats 374 away from human settlements (personal observation). These observations agree with past research 375 [11] and indicate that cats are not roaming the landscape of Naxos or travelling between villages 376 in significant numbers, but rather remain close to their within-village food sources. Consequently, 377 increases in cat density within a village are the result of local legacies rather than the result of 378 extensive inter-village movement. .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted July 31, 2025. ; https://doi.org/10.1101/2025.07.27.667076doi: bioRxiv preprint 19 379 Is it possible that the association between cats and food dishes is simply the result of 380 increased provisioning by compassionate locals in response to rising cat populations, rather than 381 the other way around? While some of this cannot be completely ruled out, field observations 382 suggest instead that feeding intended for a local house pet ends up inadvertently supporting other 383 stray cats and hence promotes population increases. Furthermore, cat populations closely reflect 384 the density of garbage dumpsters, the number of which certainly is not determined by cat density. 385 In summary, the existing evidence suggests that intentional or unintentional supplementation of 386 cats drives population numbers rather than vice versa. 387 We found that cat population density is negatively correlated with resident reptile 388 population abundance, indicating substantial levels of predation on lizards and snakes. This 389 corroborates regular but ad hoc observations during the study of cats consuming wildlife (S1 390 Appendix). While data collection in this study focused on reptiles, parallel trends are known to 391 occur within birds as well [40]. 392 This study also reveals that while cat impacts on local wildlife are both pervasive and 393 density-dependent, they are also modulated by the degree of cat satiation. Because cat populations 394 do diverge somewhat from what would be predicted based on available resources, the residuals of 395 this relationship were used as a metric of the relative resources available per cat capita, essentially 396 obtaining an index of over/under provisioning (termed cat satiation). The results indicate that not 397 only is the extent to which cat satiation is proportional to decreases in wildlife populations 398 independent from cat density (Fig 7), but also that the interaction between cat density and satiation 399 is particularly important. Hence, cat numbers are more negatively associated with wildlife 400 populations when satiation indices are high, i.e. cats are hungry. This is consistent with a scenario 401 where cats preferentially consume human-derived resources and switch to preying local wildlife .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted July 31, 2025. ; https://doi.org/10.1101/2025.07.27.667076doi: bioRxiv preprint 20 402 when these are not sufficient. Despite the strong evidence of direct cat impact on reptiles through 403 predation, a different explanatory process cannot be completely excluded. In this scenario, higher 404 cat density and reduced provisioning results in reptiles that are more wary, more cryptic, and less 405 active, therefore reducing detectability during surveys. 406 A clear linear rise in reptile abundance with increasing distance from a village edge was 407 found (Fig 5). Reptile counts increased on average by ~40% from the immediate edge of the village 408 to 200 meters away from a village, suggesting that proximity to villages was detrimental to reptiles. 409 While this pattern is consistent with increased cat-induced mortality close to villages, there was 410 no relationship between cat density and how rapidly reptile numbers rose with increasing distance 411 from a village. Hence, the steepness of this decline in reptile abundance when approaching a 412 village was not significantly related to cat density or cat satiation, perhaps because of insufficient 413 sample sizes or non-linear relationships. While the possibility that cats do not influence the effect 414 of distance is not being ruled out, alternative causes may be that 1) as lizards are predated by cats, 415 the high density of highly active lizards means that the new spaces created by cat predation are 416 readily filled by a lizard from further away and 2) the increase in reptiles away from villages could 417 be attributed to other factors such as habitat suitability increasing with greater distance from human 418 settlements. 419 Because anthropogenic food subsidies are critical for cat population growth, and increasing 420 cat population size has proportional negative impacts on resident wildlife population sizes, this 421 study strongly indicates that human subsidies are facilitating hyperpredation of wildlife on Naxos. 422 As a result, in many Mediterranean regions, human supplementation drives the impacts of an 423 artificially elevated cat population on resident wildlife. 424 Management recommendations .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted July 31, 2025. ; https://doi.org/10.1101/2025.07.27.667076doi: bioRxiv preprint 21 425 Public opinion regarding free-ranging cats on the Aegean islands varies greatly, with some 426 residents considering them pests while others taking care of them through regular feeding (personal 427 observation). On Naxos and other islands, cats are increasingly becoming a source of amusement 428 for tourists, and cat-themed merchandise can be found on sale in many retail stores. Local 429 businesses have also capitalized on the tourist’s affection for cats by selling bags of cat food to 430 allow for a petting zoo-like experience with the local strays. Over the last several years, there has 431 been a pronounced shift in public opinion regarding cat welfare away from the past laissez-faire 432 attitude toward a more compassionate stance. As a matter of fact, it is the attitude of support and 433 consequent feeding that has resulted in the present-day ballooning of cat populations on Aegean 434 islands. Concurrently, there is a growing realization that cat overpopulation is a problem, primarily 435 due to animal welfare rather than wildlife conservation reasons, and cat population control 436 programs are increasingly discussed. 437 This study has shown that increasing cat densities translate directly into higher mortality 438 for reptiles, and that the negative impacts of cats is strongest when human food becomes limiting, 439 and cats are forced to rely more on wildlife for nutrition. These results may be interpreted as 440 providing conflicting management suggestions. While in the short term, overprovisioning of cats 441 reduces predation of wildlife, in the long term this will result in accelerated reproduction and 442 expanding cat populations. This conflict between short-term benefits versus long-term detriments 443 can best be navigated through a carefully balanced approach. If cat populations are to be managed 444 as to reduce wildlife mortality, several steps need to be taken including expansion of existing trap- 445 neuter-release programs, as well as reduction in supplemental feeding of unsterilized cats to reduce 446 reproductive potential. More specifically, proposed recommendations include 1) improving waste 447 management policies to decrease accessibility of this resource to cats, 2) increasing trap-neuter .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted July 31, 2025. ; https://doi.org/10.1101/2025.07.27.667076doi: bioRxiv preprint 22 448 release efforts, and 3) providing local education about the deleterious impacts of cats on wildlife 449 and encouraging both residents and tourists to only feed sterilized cats. This last point is 450 significant, as keeping only sterilized cats fed will decrease their dependence on wildlife while 451 also not contributing to cat population growth. The reduced access to human-derived food by 452 unsterilized cats through recommendations 1 and 3 may cause them to consume wildlife more in 453 the short term, but the loss of easily available human food resources should also reduce their 454 reproductive output. While controversial, it is worth mentioning that humane cat culling is an 455 option as well, as it has been shown to be highly effective, especially on islands [67]. Additionally, 456 the implementation of cat management needs to be followed by close environmental monitoring 457 to assess unintended consequences such as mesopredator release of rats following cat population 458 declines [22]. 459 Conclusion and future research directions 460 As global biodiversity continues to be lost at an alarming rate [68], the importance of 461 understanding the mechanisms driving species extinction is increasing. The conclusions of this 462 study should be applicable to numerous regions in the Mediterranean Basin, at least those where 463 cats impact small- and medium- bodied vertebrates. Further research efforts need to center on 464 investigating how the extent of human subsidies influences that distance cats will roam from a 465 village, and stable-isotope analysis of scat can help elucidate how the ratio of anthropogenic and 466 natural food sources vary along this gradient. Lastly, this study focused on the high tourism 467 summer period when large numbers of visitors come to the island. Extending the study to the 468 winter months when less food supplementation is available can help to understand the true impact 469 on resident wildlife. .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted July 31, 2025. ; https://doi.org/10.1101/2025.07.27.667076doi: bioRxiv preprint 23 470 Acknowledgements 471 We would like to thank Father George Palamaris for providing us with housing in the 472 Venetian castle of Naxos while conducting this research, and to Mayor Dimitris Lianos for being 473 supportive of the research the Foufopoulos lab conducts in Greece. We would also like to thank 474 Andressa Viol for her assistance in developing conceptual figure 1. 475 References 476 1. Wilcove DS, Rothstein D, Dubow J, Phillips A, Losos E. Quantifying Threats to 477 Imperiled Species in the United States: Assessing the relative importance of habitat 478 destruction, alien species, pollution, overexploitation, and disease. BioScience. 1998; 479 48(8): 607-615. 480 2. Tilman D, Clark M, Williams DR, Kimmel K, Polasky S, Packer C. Future threats to 481 biodiversity and pathways to their prevention. Nature. 2017; 546: 73–81. 482 3. Hald-Mortensen C. The Main Drivers of Biodiversity Loss: A Brief Overview. Journal 483 of Ecology & Natural Resources. 2023; 7. 484 4. Clavero M, García-Berthou E. Invasive species are a leading cause of animal extinctions. 485 Trends in Ecology & Evolution 2005; 20(3): 110. 486 5. Bellard C, Cassey P, Blackburn T. Alien species as a driver of recent extinctions. Biology 487 Letters. 2016; 12(2). 488 6. Dueñas MA, Hemming DJ, Roberts A, Diaz-Soltero H. The threat of invasive species to 489 IUCN-listed critically endangered species: A systematic review. Global Ecology and 490 Conservation. 2021; 26; e01476. .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted July 31, 2025. ; https://doi.org/10.1101/2025.07.27.667076doi: bioRxiv preprint 24 491 7. Doherty TS, Glen AS, Nimmo DG, Ritchie EG, Dickman CR. Invasive predators and 492 global biodiversity loss. Proceedings of the National Academy of Sciences. 2016; 493 113(40): 11261–11265. 494 8. Loss SR, Boughton B, Cady SM, Londe DW, McKinney C, O’Connell TJ, et al. Review 495 and synthesis of the global literature on domestic cat impacts on wildlife. Journal of 496 Animal Ecology. 2022; 91(7): 1361–1372. 497 9. Loss SR, Will T, Marra PP. The impact of free-ranging domestic cats on wildlife of the 498 United States. Nature Communications. 2013; 4(1396). 499 10. Stone PA, Snell HL, Snell HM. Behavioral Diversity as Biological Diversity: Introduced 500 Cats and Lava Lizard Wariness. Conservation Biology. 1994; 8(2): 569-573. 501 11. Li B, Belasen A, Pafilis P, Bednekoff P, Foufopoulos J. Effects of feral cats on the 502 evolution of anti-predator behaviours in island reptiles: Insights from an ancient 503 introduction. Proceedings of the Royal Society B: Biological Sciences. 2014; 281(1788): 504 20140339. 505 12. Gerhold RW, Jessup DA. Zoonotic Diseases Associated with Free-Roaming Cats. 506 Zoonoses and Public Health. 2013; 60(3): 189–195. 507 13. Beutel T, Reineking B, Tiesmeyer A, Nowak C, Heurich M. Spatial patterns of co- 508 occurrence of the European wildcat Felis silvestris silvestris and domestic cats Felis 509 silvestris catus in the Bavarian Forest National Park. Wildlife Biology. 2017; 2017(1): 1- 510 8. 511 14. Medina FM, Bonnaud E, Vidal E, Tershy BR, Zavaleta ES, Donlan CJ, et al. A global 512 review of the impacts of invasive cats on island endangered vertebrates. Global Change 513 Biology. 2011; 17(11): 3503–3510. .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted July 31, 2025. ; https://doi.org/10.1101/2025.07.27.667076doi: bioRxiv preprint 25 514 15. Medina FM, Bonnaud E, Vidal E, Nogales M. Underlying impacts of invasive cats on 515 islands: Not only a question of predation. Biodiversity and Conservation. 2014; 23(2): 516 327–342. 517 16. Tershy BR, Shen KW, Newton KM, Holmes ND, Croll DA. The Importance of Islands 518 for the Protection of Biological and Linguistic Diversity. BioScience. 2015; 65(6): 592– 519 597. 520 17. Moors PJ, Atkinson IAE, Sherley GH. Reducing the rat threat to island birds. Bird 521 Conservation International. 1992; 2(2): 93-114. 522 18. Courchamp F, Langlais M, Sugihara G. Control of rabbits to protect island birds from cat 523 predation. Biological Conservation. 1999; 89(2): 219-225. 524 19. Blumstein DT, Daniel JC. The loss of anti-predator behaviour following isolation on 525 islands. Proceedings of the Royal Society B: Biological Sciences. 2005; 272(1573): 526 1663-8. 527 20. Smith AP, Quin DG. Patterns and causes of extinction and decline in Australian 528 conilurine rodents. Biological Conservation. 1996; 77(2): 243–267. 529 21. Bonsall MB, Hassell MP. Apparent competition structures ecological assemblages. 530 Nature. 1997; 388: 371-373 531 22. Courchamp F, Langlais M, Sugihara G. Cats protecting birds: Modelling the 532 mesopredator release effect. Journal of Animal Ecology. 1999; 68(2): 282–292. 533 23. Courchamp F, Langlais M, Sugihara G. Rabbits killing birds: Modelling the 534 hyperpredation process. Journal of Animal Ecology. 2000; 69(1): 154–164. 535 24. Taylor RH. How the Macquarie Island parakeet became extinct. New Zealand Journal of 536 Ecology. 1979; 2: 42-45. .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted July 31, 2025. ; https://doi.org/10.1101/2025.07.27.667076doi: bioRxiv preprint 26 537 25. Imber M, Harrison M, Harrison J. Interactions between petrels, rats and rabbits on Whale 538 Island, and effects of rat and rabbit eradication. New Zealand Journal of Ecology. 2000; 539 24(2): 153–160. 540 26. Roemer GW, Coonan TJ, Garcelon DK, Bascompte J, Laughrin L. Feral pigs facilitate 541 hyperpredation by golden eagles and indirectly cause the decline of the island fox. 542 Animal Conservation Forum. 2001; 4(4): 307–318. 543 27. Pope KL, Garwood JM, Welsh HH, Lawler SP. Evidence of indirect impacts of 544 introduced trout on native amphibians via facilitation of a shared predator. Biological 545 Conservation. 2008; 141(5): 1321–1331. 546 28. Carpio AJ, Tortosa FS, Barrio IC. Rabbit Abundance Influences Predation on Bird Nests 547 in Mediterranean Olive Orchards. Acta Ornithologica. 2015; 50(2): 171-179. 548 29. Levine JM, D’Antonio CM. Forecasting Biological Invasions with Increasing 549 International Trade. Conservation Biology. 2003; 17(1): 322–326. 550 30. Maeda T, Nakashita R, Shionosaki K, Yamada F, Watari Y. Predation on endangered 551 species by human-subsidized domestic cats on Tokunoshima Island. Scientific Reports. 552 2019; 9(1): 16200. 553 31. Liberg O, Sandell M, Pontier D, Natoli E. Density, spatial organization, and reproductive 554 tactics in the domestic cat and other felids. In: Turner DC, Bateson P, editors. The 555 Domestic Cat: The Biology of its Behavior. 2nd ed. Cambridge: Cambridge University 556 Press; 2000. p. 119-48. 557 32. Fedriani JM, Fuller TK, Sauvajot RM. Does availability of anthropogenic food enhance 558 densities of omnivorous mammals? An example with coyotes in southern California. 559 Ecography. 2001; 24(3): 325–331. .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted July 31, 2025. ; https://doi.org/10.1101/2025.07.27.667076doi: bioRxiv preprint 27 560 33. Newsome TM, Dellinger JA, Pavey CR, Ripple WJ, Shores CR, Wirsing AJ, et al. The 561 ecological effects of providing resource subsidies to predators. Global Ecology and 562 Biogeography. 2015; 24(1): 1-11. 563 34. Izawa M, Doi T, Ono Y. Grouping patterns of feral cats living on a small island in Japan. 564 Japan Journal of Ecology. 1982; 32: 373-82. 565 35. Macdonald DW. The ecology of carnivore social behaviour. Nature. 1983; 301: 379–384. 566 36. Serpell JA. Domestication and history of the cat. In: Turner DC, Bateson P, editors. The 567 Domestic Cat: The Biology of its Behaviour. 3rd ed. Cambridge (UK): Cambridge 568 University Press; 2014. p. 179-92. 569 37. Vigne JD, Guilaine J, Debue K, Haye L, Gérard, P. Early Taming of the Cat in Cyprus. 570 Science. 2004; 304: 259-259. 571 38. Driscoll CA, Clutton-Brock J, Kitchener AC, O’Brien SJ. The Taming of the Cat. 572 Scientific American. 2009; 300(6): 68–75. 573 39. Mori E, Menchetti M, Camporesi A, Cavigioli L, Tabarelli de Fatis, K., Girardello M. 574 License to Kill? Domestic Cats Affect a Wide Range of Native Fauna in a Highly 575 Biodiverse Mediterranean Country. Frontiers in Ecology and Evolution. 2019; 7. 576 40. Krawczyk E, Hedman H, Pafilis P, Bergen K, Foufopoulos J. Effects of touristic 577 development on Mediterranean island wildlife. Landscape Ecology. 2019; 34. 578 41. Bonnaud E, Bourgeois K, Vidal E, Kayser Y, Tranchant Y, Legrand J. Feeding Ecology 579 of a Feral Cat Population on a Small Mediterranean Island. Journal of Mammalogy. 580 2007; 88(4): 1074–1081. .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted July 31, 2025. ; https://doi.org/10.1101/2025.07.27.667076doi: bioRxiv preprint 28 581 42. Bonnaud E, Berger G, Bourgeois K, Legrand J, Vidal E. Predation by cats could lead to 582 the extinction of the Mediterranean endemic Yelkouan Shearwater Puffinus yelkouan at a 583 major breeding site. Ibis. 2012; 154(3): 566-577. 584 43. Sarika M, Bazos I, Zervou S, Christopoulou A. Flora and vegetation of the European- 585 network “Natura 2000” habitats of Naxos island (GR4220014) and of nearby islets 586 Mikres Kyklades (GR4220013), Central Aegean (Greece). Plant Sociology. 2015; 52(2): 587 3–56. 588 44. Valakos ED, Pafilis P, Sotiropoulos K, Lymberakis P, Maragou, P, Foufopoulos J. The 589 amphibians and reptiles of Greece. Ed. Chimaira: Frankfurt Am Main; 2008 590 45. Solomou A, Proutsos N, Karetsos G, Tsagari K, Chatzipavlis N. Micrometeorology of the 591 agricultural terraces and stone walls and impacts on biodiversity in the Mediterranean 592 landscape of Greece. International Journal of Agricultural Resources, Governance and 593 Ecology. 2022; 18(1-2): 3-21. 594 46. Nossan H. Biodiversity Implications of Agricultural Terrace Abandonment in a 595 Mediterranean Landscape [master’s thesis]. Ann Arbor: University of Michigan; 2023 596 47. Pafilis P, Foufopoulos J, Poulakakis N, Lymberakis P, Valakos ED. Tail Shedding in 597 Island Lizards [Lacertidae, Reptilia]: Decline of Antipredator Defenses In Relaxed 598 Predation Environments. Evolution. 2009; 63(5): 1262-1278. 599 48. Hellenic Statistical Authority. [homepage on the Internet]. Available from: 600 https://www.statistics.gr/en/home/ Accessed [6,1,2024]. 601 49. Semple EC. Pirate Coasts of the Mediterranean Sea. Geographical Review. 1916; 2(2): 602 134–151. 603 50. ESRI. ArcGISPro v3.4. Redlands. California: ESRI, Inc.; 2024. .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted July 31, 2025. ; https://doi.org/10.1101/2025.07.27.667076doi: bioRxiv preprint 29 604 51. ESRI. National Geographic World Map. Redlands. California: ESRI, Inc.; 2025. 605 52. Donihue CM. Drivers of Functional Trait Variability in Podarcis erhardii, the Aegean 606 Wall Lizard [dissertation]. New Haven: Yale University; 2016. 607 53. Catsadorakis G. Aspects of Ecology and Behaviour in the lizard Podarcis 608 erhardii. Biologia Gallo-Hellenica. 1984; 11: 99-110. 609 54. Konecny MJ. Home Range and Activity Patterns of Feral House Cats in the Galápagos 610 Islands. Oikos. 1987; 50(1): 17–23. 611 55. Elizondo EC, Loss SR. Using trail cameras to estimate free-ranging domestic cat 612 abundance in urban areas. Wildlife Biology. 2016; 22(5): 246–252. 613 56. Schnabel ZE. The Estimation of the Total Fish Population of a Lake. The American 614 Mathematical Monthly. 1938; 45(6): 348–352. 615 57. de Azevedo Carvalho D, Collins PA, De Bonis CJ. The mark–recapture method applied 616 to population estimates of a freshwater crab on an alluvial plain. Marine and Freshwater 617 Research. 2013; 64(4): 317-323. 618 58. R Core Team. R: A Language and Environment for Statistical Computing. Vienna 619 (Austria): R foundation for Statistical Computing; 2024 620 59. Bates D, Maechler M, Bolker B, Walker S. Fitting Linear Mixed-Effects Models Using 621 lme4. Journal of Statistical Software. 2015; 67(1): 1-48. 622 60. Mazerolle MJ. AICcmodavg: Model selection and multimodel inference based on 623 (Q)AIC(c) [R package]. Version 2.3.3. 2023. 624 61. McQuarrie A, Shumway R, Tsai CL. The model selection criterion AICu. Statistics & 625 probability letters. 1997; 34(3): 285-292. .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted July 31, 2025. ; https://doi.org/10.1101/2025.07.27.667076doi: bioRxiv preprint 30 626 62. ESRI. World Image WGS1983 World Image Base Map. Redlands. California: ESRI, 627 INC.; 2025. 628 63. Liberg O. Spacing Patterns in a Population of Rural Free Roaming Domestic Cats. Oikos. 629 1980; 35(3): 336–349. 630 64. Mirmovitch V. Spatial Organisation of Urban Feral Cats (Felis Catus) in Jerusalem. 631 Wildlife Research. 1995; 22(3): 299–310. 632 65. Jones E, Coman BJ. Ecology of the Feral Cat, Felis catus (L.), in South-Eastern Australia 633 III.* Home Ranges and Population Ecology in Semiarid North-West Victoria. Wildlife 634 Research. 1982; 9(3): 409–420. 635 66. Bengsen AJ, Algar D, Ballard G, Buckmaster T, Comer S, Fleming PJS, et al. Feral cat 636 home-range size varies predictably with landscape productivity and population density. 637 Journal of Zoology. 2016; 298(2): 112-120. 638 67. Nogales M, Martín A, Tershy BR, Donlan CJ, Veitch D, Puerta N, et al. A Review of 639 Feral Cat Eradication on Islands. Conservation Biology. 2004; 18(2): 310-319. 640 68. Wilson EO. Threats to Biodiversity. Scientific American. 1989; 261(3): 108-117. 641 S1 Table. Summary table showing data used in this study. Area and density columns are in 642 hectares, total food density column is equal to the summation of dish density and dumpster 643 density, the reptile count column shows the average and standard deviation of total reptile 644 counts observed at the corresponding village across all surveys. 645 S1 Appendix. A. Outdoor cats line a set of dumpsters in Chalkio. B. Outdoor cats feed out of 646 a feeding dish in Vivlos. C. Cat predates a Eurasian Collared Dove in Chora. D. Cat predates 647 a wall lizard in Moni. .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted July 31, 2025. ; https://doi.org/10.1101/2025.07.27.667076doi: bioRxiv preprint 31 648 S2 Appendix. Photo taken in Glynado demonstrating the exceptional densities cats can 649 achieve in urban environments. 650 .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted July 31, 2025. ; https://doi.org/10.1101/2025.07.27.667076doi: bioRxiv preprint .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted July 31, 2025. ; https://doi.org/10.1101/2025.07.27.667076doi: bioRxiv preprint .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted July 31, 2025. ; https://doi.org/10.1101/2025.07.27.667076doi: bioRxiv preprint .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted July 31, 2025. ; https://doi.org/10.1101/2025.07.27.667076doi: bioRxiv preprint .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted July 31, 2025. ; https://doi.org/10.1101/2025.07.27.667076doi: bioRxiv preprint .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted July 31, 2025. ; https://doi.org/10.1101/2025.07.27.667076doi: bioRxiv preprint .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted July 31, 2025. ; https://doi.org/10.1101/2025.07.27.667076doi: bioRxiv preprint .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted July 31, 2025. ; https://doi.org/10.1101/2025.07.27.667076doi: bioRxiv preprint .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted July 31, 2025. ; https://doi.org/10.1101/2025.07.27.667076doi: bioRxiv preprint