Pengembangan Model Prediksi Risiko Kebakaran Hutan dan Lahan di Kalimantan Barat menggunakan Data Satelit

Abstract

The frequency and intensity of forest and land fires in recent decades tend to increase, despite the prevention efforts have been made. In most cases, the prevention efforts were not effectice since the available early warning systems is on a daily basis or the longest is on weekly basis. The available time for the prevention is very short. This study aimed to develop seasonal fire early warning system, a system that can provide warning information with lead time of at least 1 to 3 month. With this system, it is expected that the preventive actions can be implemented more effectively. The method used was to build a hotspots prediction model based on information of sea surface temperature anomaly in the Pacific Ocean (NIÑO 3.4 region) and monthly and seasonally rainfall. Hotspots are used as an indicator to indicate the level of fire threat. Models that relate hotspot and sea surface temperature anomaly/seasonal rainfall were developed for seven land use/cover categories in two soil types, i.e. peat lands and minerals. The level of fire risk is determined based on the probability of having number of hotspot above threshold values under different condition of the sea surface temperature in the NIÑO 3.4. The threshold value is the same as the monthly average of hotspot number. The fire risk model was developed only for West Kalimantan province using hotspot data from 2001-2013 from Terra and Aqua satellites Modis. The results of the analysis suggested that the highest density of hotspots generally occured around August for non-forest land cover types of peat lands. Variability of hotspot density from July to October in this area was significantly correlated with the variability of SST anomalies in NIÑO 3.4 region with time lag of zero to two months. Based on risk matrix, it was found that the probability of having hotspot above the threshold value will be more than 50% if the the SST anomaly in NINO3.4 region was more than 0.5 oC. The probability to have hotspot number more than threshold value when the SST anomaly at lag 0, 1 and 2 month were 63%, 54% and 50% respectively. For the future research, it is suggested that the the development of hotspot forecasting model should use twophase regression techniques.