Livestock predation by large cats represents a threat both to livestock production and to large cat conservation when retaliation occurs. Therefore, understanding the factors that influence the occurrence of attacks and their spatial distribution has become an important task for conservation managers. However, the importance of spatial and temporal scales has often been overlooked. In this study, we investigated the risk of attack on livestock by large cats while selecting the appropriate temporal and spatial scales at which these attacks occur. We collected geospatial data on attack and non-attack sites (2011-2015) in the region of Calakmul, which hosts the largest population of jaguars in Mexico, and obtained additional information relative to livestock management through 165 interviews with ranchers. We derived ecological and anthropogenic variables from two land-use maps (2000 and 2015) at four scales relevant for large cats (0,5 km, 2,5 km, 5 km and 10 km-radius). We built two sets of models, one expressing the potential effects of landscape characteristics (structural and functional), and the other expressing the potential effects of human pressure (human population, land use, and livestock production). Following an information-theoretic approach that integrates spatial correlation, we selected the appropriate temporal and spatial scales and then applied a hierarchical selection of our models. Five variables were best explained at a specific spatial scale, while the amount of forest in 2000 appeared more important than the amount of forest in 2015. The species of livestock raised was by far the main determinant of attack in the region, with sheep particularly at risk, which prompted us to integrate this variable into every models. Attack occurrence in the region was best explained by the functional characteristics of the landscape that included the fragmentation process; the effect of human pressure was of lower importance. This study, based on a robust approach using on state of the art procedures in landscape ecology, shows the importance of incorporating multiple spatial and temporal scales. It also highlights the benefit of using a two-dimensional approach to support conservation management measures at the appropriate organizational level. For instance, livestock management might be better addressed at a community level, whereas landscape fragmentation will be better tackled at a regional or state level.