Predicting fire risk for Scotland’s peatlands using statistical models based on weather conditions

Peatlands are globally important carbon-rich ecosystems but are increasingly vulnerable to fire risk due to climate change and human activity. Predictive modeling of peatland fire risk is essential for effective management and mitigation, particularly in regions like Scotland, where extensive peatlands face unique climatic and ecological pressures. This study aims to develop a weather-driven predictive framework for peatland fire risk, focusing on the weather data (e.g., temperature, precipitation, relative humidity) with drought and climate indices (e.g., SPEI, NAO) to enhance prediction accuracy for Scotland’s peatlands. Statistical models including machine learning (ML) techniques are utilized to capture seasonality, spatial variability and fine-scale hydrological dynamics in the fire risk. The study also evaluates the predictive skill of linear Log-Reg and ML-based models, proposing the best model to use to predict peatland fire risk probability. We highlight the gaps in peatland-specific fire modeling, and suggest future research priorities to effectively address and to improve fire risk predictions and inform peatland management strategies in Scotland and similar ecosystems.