THE MAP OF THE PREDICTION OF THE PROBABILITY OF AT LEAST ONE HOTSPOT DUE TO FOREST AND LAND FIRES IN RIAU PROVINCE, INDONESIA

• Main Author: Tesalonika Riwie W., Maria
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/64957
• Institution: Institut Teknologi Bandung

because of the flammable characteristics of the forest vegetation and of the land. On the other hand, the forest and land fires could also occur caused by human factor, such as deliberate acts of arson or negligence. In addition to the possibility of human casualties and to respiratory problems due to the resulting haze, forest and land fires could also cause interruptions to businesses or to daily activities which could result in financial losses. Riau province in Indonesia often experiences forest and land fires. The impact of the resulting haze could cause respiratory problems which could result in deaths; business interruptions, such as flights cancellations; and interruptions to daily activities, such as temporarily stopping school and office activities, not only in the city of Pekanbaru and its surroundings but also in neighboring countries, such as Singapore and Malaysia. This thesis attempts to predict the probability of at least one fire hotspot in Riau province based on peatland types, types of land cover, forest area allocations, types of land-use, distance to rivers, and distance to roads, using the historical data of the number of hotspots caused by forest and land fires in Riau province from January 2007 to December 2019. The prediction of the probability is determined using a logistic regression model. Although the forest and land fires in Riau province often occur during the dry or hot season, which is around June to August, determining the prediction of the probability was carried out without considering the weather or climate factors nor the fraudulent activities by human; and without considering the time of the occurrence of the fire hotspots. The estimations of the parameters of the logistic regression model are determined using the Maximum Likelihood method and the selection of the best model is determined based on the smallest value of the Akaike Information Criterion (AIC). The model goodness-of-fit is based on the largest value of the area under the Receiver Operating Characteristics (ROC) curve or the Area Under Curve (AUC). After the best model was determined, the predictions of the probabilities of at least one hotspot caused by forest and land fires in Riau province are classified into five categories: very-low, low, medium, high, and very-high. Lastly, the maps of the predictions of the probabilities are constructed for twelve districts in Riau province.