Meteorological Deviations from Monthly Climatological Reference Values and Their Association with Forest Fire Burned Area and Duration, in Three Greek Regions

Earth may persist without humanity, but human survival and well-being depend on a healthy planet and natural systems. This asymmetry highlights a fundamental challenge: humans are part of nature and depend on its systems, yet human activities can also disrupt them. Our growing capacity to understand these relationships offers an opportunity and a responsibility to pursue solutions that support the flourishing of ecosystems alongside human health. Forest fires illustrate this interdependence through their potential consequences for ecosystems and communities, although these impacts were not directly assessed in the present study. Within this planetary health context, this study examines how deviations from climatological reference values are associated with forest fire burned-area classes, in three Greek regions. Specifically, we analyzed temperature, wind speed, and humidity at forest fire onset in Attica and Crete (2011–2025) and in Eastern Macedonia and Thrace (2022–2025), comparing observations with region-specific climatological reference periods. Duration patterns were examined descriptively, while six machine-learning methods were evaluated for classifying burned-area outcomes, with incidents of at least 5 ha assigned to the larger burned-area class. Adam, GenClass, and FC2 achieved similar mean classification errors across the regional datasets (12.12–12.18%), whereas L-BFGS and RBF performed slightly worse, and NEAT showed a substantially higher error. These results indicate comparable performance among the three leading methods for the available data; they do not demonstrate causal relationships or operational forecasting capability. The meteorological associations identified may contribute to understanding forest fire risk and could serve as one input to future integrated approaches to preparedness and environmental management. Further validation, together with ecological, health, and social data, is needed to assess their wider relevance to planetary health and community resilience.