The issue of global warming and more specifically its health impact on populations is increasingly concerning. The aim of our study was to evaluate the impact of temperature on dairy cattle mortality in France during the warm season (April–August). We therefore devised and implemented a spatial partitioning method to divide France into areas in which weather conditions were homogeneous, combining a multiple factor analysis with a clustering method using both weather and spatial data. We then used time-series regressions (2001–2008) to model the relationship between temperature humidity index (an index representing the temperature corrected by the relative humidity) and dairy cattle mortality within these areas. We found a significant effect of heat on dairy cattle mortality, but also an effect of cooler temperatures (to a lesser extent in some areas), which leads to a U-shaped relationship in the studied areas. Our partitioning approach based on weather criteria, associated with classic clustering methods, may contribute to better estimating temperature effects, a critical issue for animal health and welfare. Beyond the interest of its use in animal health, this approach can also be of interest in several situations in the frame of human health.