SPATIAL ASSESSMENT OF PEAT-LAND FIRES UTILIZING BINARY LOGISTICS REGRESSION IN WEST KALIMANTAN

Abstract

This study contributes to the understanding of forest fire susceptibility by applying a binary logistic regression model combined with a Geographic Information Systems (GIS) to map hotspot vulnerability in West Kalimantan, Indonesia, an approach not extensively explored in previous research. Forest fire is one of the environmental problems. In West Kalimantan, land fires are a routine disaster that is experienced almost every year. In this paper, a binary logistic regression model was used to identify land fire in west Kalimantan. In addition, mapping of confidence of hotspot susceptibility was carried out in West Kalimantan. The data used were 72 hotspots spread across in seven districts of West Kalimantan in 2020. The independent variables used were land cover, slope, topography, distance of hotspots to rivers, distance of hotspots to roads and distance of hotspots to settlements. While the dependent variable was the point which was classified into hotspots and non-hotspots. Results showed that the method identified that the variables significantly influencing land fires include the distance of the points to the river and the distance of the points to the road. The Binary Logistic Regression model of the land fire in West Kalimantan has a classification accuracy rate is 84.03%. From the results of weighting and visualization using GIS shown that the area that has a very high level of vulnerability is the city of Pontianak (42.97%). Meanwhile, areas that have a moderate level of vulnerability include Kayong Utara, Kubu Raya, Mempawah, Sambas, Sanggau, Sekadau and Sintang districs. Kubu Raya and Kayong Utara districts in the medium vulnerability level have the largest forest fire districts (43.70% and 41.25%). Meanwhile, districts that are in the very low vulnerability level are Bengkayang, Singkawang, Landak and Melawi districts.