A Statistical Analysis of Wildland and Prescribed Fires in the Changing California Climate
Author: Sequoia Andrade
Faculty Supervisor: Alexandra Piryatinska
Department: Mathematics
With climate change-driven large, destructive wildfires becoming more frequent, using data to identify effective or ineffective prevention strategies, such as prescribed burns, can inform future prevention strategies. In this work, a suite of statistical methods, including descriptive, inferential, and predictive methods are applied to a new dataset to analyze if prescribed burns are effective in mitigating large wildfires in recent history while accounting for climate change. Descriptive and inferential methods are used to summarize and identify significant trends. A time series analysis with cross correlation is used to identify which variables are related to wildfire acres burned at different time lag values. Additionally, a regression analysis is performed to identify significant predictors of wildfire acres burned and to estimate expected acres burned when climate variables, such as average temperature and precipitation, change in accordance with climate models. When grouped across the entire state with a monthly time scale, results indicate prescribed burns do not impact the number of acres burned due to wildfire. However, when considering smaller regions, the impact of prescribed burns is measurable. At a state-wide level, increased temperature, increased drought, and lower precipitation are related to increased wildfire acres burned. Overall climate variables have a stronger impact on wildfire size, though prescribed burns are associated with slightly smaller wildfires.