论文部分内容阅读
目前,森林火险气象指数被广泛用于世界多个国家和地区。本研究目的为,基于火险气象指数,在中国不同气候区建立火险概率模型。本文在中国4个气候区,使用1998-2007年的气象及火灾数据,以位置变量、月份、海拔、加拿大、美国及澳大利亚的气象火险指数、植被指数为自变量,建立了半参数化Logistic回归模型,分析各自变量与着火概率及大火发生概率之间的非线性关系。在不同区域,模型所选自变量组合不同,这与各气候区不同气象及植被状况有关。通过模型模拟数据和实际观测数据散点图、火险概率图、大面积火灾数量预报曲线图,分析了模型的预测能力。研究结果表明,在4个气候区,海拔和NDVI 指数对着火概率影响显著。模拟可燃物含水量的气象火险指数由于反映出了植被的季节变化特征,在中国北部成为火险概率模型中的重要因子。模拟土壤有机层可燃物状况的火险气象指数在中国南部(东南、西南)成为火险概率模型的重要因子。在中国4个气候区,应用半参数化Logistic回归模型,可以有效模拟月时间尺度着火概率及大火发生概率,并为分析火险气象指数的预报能力提供了有效途径。本研究为进一步分析气候与火险之间的动态关系提供了理论基础。
At present, the forest fire danger meteorological index is widely used in many countries and regions in the world. The purpose of this study is to establish the probability model of fire danger in different climatic zones in China based on the fire danger meteorological index. In this paper, meteorological and fire data from 1998 to 2007 were used in four climatic zones in China. Semi-parametric Logistic regression was established based on the location variables, monthly, elevations, meteorological fires in Canada, the United States and Australia and the vegetation index. Model to analyze the nonlinear relationship between each variable and the probability of fire and the probability of fire. In different regions, the variables selected by the model are different, which is related to the different meteorological and vegetation status in each climatic zone. Through the model simulation data and the actual observation data scatter plot, the fire danger probability map and the large area fire forecast curve, the model predictive ability is analyzed. The results show that in four climatic zones, elevation and NDVI have a significant impact on the probability of fire. The meteorological fire risk index that simulates the moisture content of combustibles has become an important factor in the probability model of fire risk in northern China due to the seasonal variation of vegetation. The fire danger meteorological index simulating the combustibles of soil organic layers has become an important factor in the fire risk probability model in southern China (southeast and southwest). In the four climatic zones in China, the semi-parametric Logistic regression model can effectively simulate the probability of fire and the probability of a large fire on the monthly time scale, and provide an effective way to analyze the forecast ability of a fire-weather meteorological index. This study provides a theoretical basis for further analysis of the dynamic relationship between climate and fire hazards.