Citation

Zhang Z, Feng Z, Zhang H, Zhao J, Yu S, Du W. Int. J. Wildland Fire 2017; 26(3): 209-218.

Copyright

(Copyright © 2017, International Association of Wildland Fire, Fire Research Institute, Publisher CSIRO Publishing)

DOI

10.1071/WF16026

PMID

unavailable

Abstract

Grassland fires are major disturbances to ecosystems and economies around the world. Therefore, research on the spatial patterns of grassland fires is important for understanding the dynamics of fire occurrence and providing evidence for fire prevention and management. One of the problems in grassland fire risk analysis is that historically observed fire data are generally in the point format, with imprecise positions, whereas other influencing factors are often expressed in continuous areal units. To minimise the influences of inaccurate locations and grid size, density estimates can be produced using kernel density estimation (KDE) - a nonparametric statistical method for estimating probability densities. This method has been widely used to convert historical fire data into continuous surfaces. In this study, KDE was applied to grassland fire events in the eastern Inner Mongolia of China, based on Moderate Resolution Imaging Spectroradiometer (MODIS) Terra and Aqua daily active fire data from 2001 to 2014. The bandwidth choice was based on the mean random distance method. Annual and seasonal kernel density maps were produced, showing that the spatial patterns of grassland fire events remained temporally consistent. These results were used to create grassland fire risk zones on the basis of the mean density values in the study area. Grassland fire prevention and planning may focus on high-risk areas identified using this method.

Language: en