9:15 AM - 9:30 AM
[AAS04-02] Comprehensive Wildfire Characteristics Analysis Based on the Bivariate Copulas in the United States
Keywords:Wildfire, Frequnency analysis, Return period, Copula, Probability distributions, Wildfire bivariate dependence
Wildfires constitute an integral ecological process in the natural Earth system associated with regional and global biogeochemical cycles, human activities, and vegetation structure. Nevertheless, current research has often only focused on univariate wildfire characteristics in risk analysis, which is inadequate to accurately describe the multivariate phenomenon probabilistic properties of wildfires. Moreover, it is difficult to evaluate in which year or month the wildfire risk is higher from a comprehensive perspective. Accordingly, one objective of this study is to perform a bivariate wildfire risk assessment considering the wildfire bivariate statistical characteristics.
Application of a traditional copula approach to wildfire characteristics remains inadequate, which may overlook single mega-wildfire events and super frequent wildfires. Single wildfire characteristics and direct application of copula theory are both insufficient to reveal the complicated wildfire bivariate statistical characteristics, and the wildfire risk cannot be accurately assessed. To overcome this issue, another objective of this study is to construct a probability distribution of the wildfire priority (WP) to comprehensively measure the wildfire risk through the weighted average of the univariate and joint probabilities of wildfire statistics.
Accordingly, in this study, we simultaneously focus on the univariate extreme probability distribution of wildfire characteristics and copula-based bivariate probability distribution considering the wildfire activity and burned area. First, we determine the univariate distribution of these two target wildfire characteristics. Then, the weighted average method is applied to balance the copula-based bivariate and univariate probability distributions to calculate the WP probability. Finally, through this WP analysis approach, the spatiotemporal variability and RP trends of the wildfire risk can be explored. Since the United States is experiencing increasingly extreme wildfires, this region is selected as the study area. And the results indicate that: (1) the regional difference of the optimal marginal distribution of the burned area is more significant than that of the wildfire activity; (2) compared with direct joint probability, the weighted joint probability is slightly more sensitive in capturing extreme wildfire events; and (3) wildfires risk has an increasing trend in California, Oklahoma and the junction of Kentucky and West Virginia, while most of the southeastern United States have shown decreasing trends in wildfire risk. Overall, the framework of wildfire frequency analysis proposed in this study will provide a reference for a better understanding of the spatiotemporal characteristics of wildfire statistics.
Application of a traditional copula approach to wildfire characteristics remains inadequate, which may overlook single mega-wildfire events and super frequent wildfires. Single wildfire characteristics and direct application of copula theory are both insufficient to reveal the complicated wildfire bivariate statistical characteristics, and the wildfire risk cannot be accurately assessed. To overcome this issue, another objective of this study is to construct a probability distribution of the wildfire priority (WP) to comprehensively measure the wildfire risk through the weighted average of the univariate and joint probabilities of wildfire statistics.
Accordingly, in this study, we simultaneously focus on the univariate extreme probability distribution of wildfire characteristics and copula-based bivariate probability distribution considering the wildfire activity and burned area. First, we determine the univariate distribution of these two target wildfire characteristics. Then, the weighted average method is applied to balance the copula-based bivariate and univariate probability distributions to calculate the WP probability. Finally, through this WP analysis approach, the spatiotemporal variability and RP trends of the wildfire risk can be explored. Since the United States is experiencing increasingly extreme wildfires, this region is selected as the study area. And the results indicate that: (1) the regional difference of the optimal marginal distribution of the burned area is more significant than that of the wildfire activity; (2) compared with direct joint probability, the weighted joint probability is slightly more sensitive in capturing extreme wildfire events; and (3) wildfires risk has an increasing trend in California, Oklahoma and the junction of Kentucky and West Virginia, while most of the southeastern United States have shown decreasing trends in wildfire risk. Overall, the framework of wildfire frequency analysis proposed in this study will provide a reference for a better understanding of the spatiotemporal characteristics of wildfire statistics.