Climate predicts wildland fire extent across China

Wildland fire extent varies seasonally and interannually in response to climatic and landscape-level drivers, yet predicting wildfires remains a challenge. Existing linear models that characterize climate and wildland fire relationships fail to account for non-stationary and non-linear associations, thus limiting prediction accuracy. To account for non-stationary and non-linear effects, we use time-series climate and wildfire extent data from across China with unit root methods, thus providing an approach for improved wildfire prediction. Results from this approach suggest that wildland area burned is sensitive to vapor pressure deficit (VPD) and maximum temperature changes over short and long-term scenarios. Moreover, repeated fires constrain system variability resulting in non-stationarity responses. We conclude that an autoregressive distributed lag (ARDL) approach to dynamic simulation models better elucidates interactions between climate and wildfire compared to more commonly used linear models. We suggest that this approach will provide insights into a better understanding of complex ecological relationships and represents a significant step toward the development of guidance for regional planners hoping to address climate-driven increases in wildfire incidence and impacts.

Automated summary

Wildfire extent shows seasonal and interannual variations due to climate and landscape-level factors, making prediction challenging. Linear models fail to capture the non-stationary and non-linear associations, reducing prediction accuracy. To address this, we utilize time-series climate and wildfire data from China, using unit root methods to improve wildfire prediction. Results highlight the sensitivity of wildland area burned to vapor pressure deficit (VPD) and maximum temperature changes in short and long-term scenarios. Furthermore, repeated fires limit system variability, leading to non-stationarity responses. We conclude that autoregressive distributed lag (ARDL) models better capture the interactions between climate and wildfire compared to linear models, providing insights into complex ecological relationships and guiding regional planners in addressing climate-driven increases in wildfire incidence and impacts.