期刊
SCIENCE OF THE TOTAL ENVIRONMENT
卷 409, 期 18, 页码 3472-3481出版社
ELSEVIER
DOI: 10.1016/j.scitotenv.2011.05.032
关键词
Global wildfire; Burned area; Anthropogenic change; Climate change; Regression trees; Random forest method
资金
- AXA Research Fund
Identifying and quantifying the statistical relationships between climate and anthropogenic drivers of fire is important for global biophysical modelling of wildfire and other Earth system processes. This study used regression tree and random forest analysis on global data for various climatic and human variables to establish their relative importance. The main interactions found at the global scale also apply regionally: greatest wildfire burned area is associated with high temperature (>28 degrees C), intermediate annual rainfall (350-1100 mm), and prolonged dry periods (which varies by region). However, the regions of highest fire incidence do not show clear and systematic behaviour. Thresholds seen in the regression tree split conditions vary, as do the interplay between climatic and anthropogenic variables, so challenges remain in developing robust predictive insight for the most wildfire-threatened regions. Anthropogenic activities alter the spatial extent of wildfires. Gross domestic product (GDP) density is the most important human predictor variable at the regional scale, and burned area is always greater when GDP density is minimised. South America is identified as a region of concern, as anthropogenic factors (notably land conversions) outweigh climatic drivers of wildfire burned area. (C) 2011 Elsevier B.V. All rights reserved.
作者
我是这篇论文的作者
点击您的名字以认领此论文并将其添加到您的个人资料中。
推荐
暂无数据