Estimation of fuel load using remote sensing data in Hulunbuir Grassland

• Published in: Natural hazards research 2022-12, Vol.2 (4), p.375-383
• Main Authors: Bao, Wurigula, Yu, Fangbing, Chang, Yu, Guo, Meng, Zhou, Fenfen, Zhong, Chao
• Format: Article
• Language: English
• Subjects: Fuel load; Hulunbuir grassland; Optical image; Radar image
• ISSN: 2666-5921; 2666-5921
• DOI: 10.1016/j.nhres.2022.11.004

Grassland fires are one of the main disturbances in Hulunbuir Grassland and have a great effect on the regional ecological environment and the production and life of herdsmen. The characteristics of fuel load and spatial distribution of grassland in the fire season are the main indices in simulating grassland fire behavior. Based on a field survey of fuel load in Hulunbuir Grassland in different seasons, vegetation indices of optical images (Landsat-8 OLI) and the texture characteristic parameters of radar data (Sentinel-1) were selected as input parameters, and three models, multiple linear regression (MLR), stepwise regression (SR) and random forest (RF), were used to estimate fuel load in the study region. Considering the limitations of optical images in estimating vegetation changes in fire seasons, this study explored a method of estimating vegetation biomass in the growing season and then transforming it to fuel load in fire seasons based on the relationships between biomass and fuel load. The results show that a combination of optical and SAR images is advantageous for estimating grassland fuel load. Of the three estimation models, the accuracy of RF is better than that of SR and better than that of MLR. This study provides a new method for large-scale grassland fuel inversion, and the continuous data of large-scale grassland fuel can improve the simulation accuracy of grassland fire behavior and provide strong support for grassland fire prediction. Furthermore, the policy of national grassland contracts to herdsmen in China has made enormous contributions to decreasing grassland fires.