Forest fire risk prediction usinggeospatial data and machine learning techniques, a case study in the western region of Nghe An province

Abstract

Nghe An is the province with the largest area of forests and forestry land in the country with more than 1 million hectares of forest, coverage rate reaching 58,33%. Due to the influence of climate change and human activities, forest cover in Nghe An has profound fluctuations, of which forest fires are one of the main causes. This article presents the results of developing a forest fire risk prediction model in the western region of Nghe An province from geospatialdata andmachine learning algorithms. From the analysis of natural and social conditions in the study area, 9input data layers include: (1) elevation, (2) slope, (3) aspect, (4) vegetation cover density, (5) population density, (6) land surface temperature, (7)evapotranspiration, (8) wind speedand(9) average monthly rainfall is used to build a forest fire risk prediction model.In the study, we tested with 02machine learning algorithms, including Random Forest (RF)and Gradient Tree Boosting (GTB), then selected the appropriate algorithm byevaluating accuracy using the fire point data set as well as model performance.The obtained results showed that the AUC(Area Under the Curve)value of the GTB(350) algorithm reached 0,948, higher than the RF(100) (0,947). From this result, the study used the GTB algorithm with 350 trees to create a forest fire risk prediction map in the western region of Nghe An province