Fire Risk Assessment and Warning Based on Hierarchical Density-Based Spatial Clustering Algorithm and Grey Relational Analysis

Abstract

Understanding the spatial and temporal distribution patterns of fire is of ecological, social and economic importance. The purpose of this study is to examine the spatial distribution of high fire risk using machine learning algorithms and early warning weather in high-risk areas. Take the satellite monitored fire point data in Yunnan Province during 2015-2019 as an example. The spatial distribution law of high-density parts is found using hierarchical density-based spatial clustering of applications with noise (HDBSCAN) algorithm, and the correlation degree analysis of high-density fire areas based on clustering and annual mean meteorological factors is carried out using grey correlation analysis (GCA) method. Results illustrates that within five years, fires frequently occurred in the seven regions of Wenshan Zhuang and Miao Autonomous Prefecture, Honghe Hani and Yi Autonomous Prefecture, Lijiang City, Pu'er City, and Xishuangbanna Dai Autonomous Prefecture. Among them, the fire in Lijiang had the greatest relationship with precipitation, Pu'er and Xishuangbanna had the greatest correlation with temperature, and Honghe Hani and Wenshan Zhuang were most affected by wind speed. This article acclaims fire prevention in key periods, key areas, key weather and reinforces the protection of transmission lines in the risk area.