Using Backpropagation to Predict Drought Factor in Keetch-Byram Drought Index

Abstract

Forest fires continue to rise during the dry season and they are difficult to stop. In this case, high temperatures in the dry season can cause an increase in drought index that could potentially burn the forest every time. Thus, the government should conduct surveillance throughout the dry season. Continuous surveillance without the focus on a particular time becomes ineffective and inefficient because of preventive measures carried out without the knowledge of potential fire risk. Based on the Keetch-Byram Drought Index (KBDI), formulation of Drought Factor is used just for calculating the drought today based on current weather conditions, and yesterday's drought index. However, to find out the factors of drought a day after, the data is needed about the weather. Therefore, we need an algorithm that can predict the dryness factor. So, the most significant fire potential can be predicted during the dry season. Moreover, daily prediction of the dry season is needed each day to conduct the best action then a qualified preventive measure can be carried out. The method used in this study is the backpropagation algorithm which has functions for calculating, testing and training the drought factors. By using empirical data, some data are trained and then tested until it can be concluded that 100% of the data already well recognized. Furthermore, some other data tested without training, then the result is 60% of the data match. In general, this algorithm shows promising results and can be applied more to complete several variables supporters.