Climate predicts wildland fire extent across China.

Area of Science:

• Ecology
• Climate Science
• Environmental Modeling

Background:

• Wildland fire extent is influenced by seasonal and interannual climate and landscape factors.
• Current linear models struggle to predict wildfires accurately due to non-stationary and non-linear climate-fire relationships.

Purpose of the Study:

• To develop an improved approach for wildfire prediction by accounting for non-stationary and non-linear climate-fire dynamics.
• To investigate the sensitivity of wildland fire extent to climate variables like vapor pressure deficit and temperature.

Main Methods:

• Utilized time-series climate and wildfire extent data from China.
• Employed unit root methods to address non-stationarity and non-linearity.
• Applied an autoregressive distributed lag (ARDL) approach for dynamic simulation modeling.

Main Results:

• Wildland area burned demonstrates sensitivity to changes in vapor pressure deficit (VPD) and maximum temperature across different time scales.
• Repeated fire events were observed to influence system variability, leading to non-stationary responses.
• The ARDL model provided a more nuanced understanding of climate-wildfire interactions compared to traditional linear models.

Conclusions:

• The autoregressive distributed lag (ARDL) approach offers a superior method for modeling dynamic climate-wildfire interactions.
• Findings provide crucial insights for understanding complex ecological relationships and informing regional planning for climate-driven wildfire risks.