Studying the factors affecting the risk of forest fire occurrence and applying neural networks for prediction

Abstract

Lebanon is considered the gate to east and west, with an area of 10452 km2. Its location has made of it a main destination for tourists from all over the world. Besides, it is rich in caves and its mountains are covered with different kinds of trees. Cedars and Pine trees are known in Lebanon since ancient times. Yet Lebanon has been facing a critical threat of losing its green mountains and fields. Forest fire has caused the loss of many green acres in the past years. This paper studies the effects of climatological data: temperature, relative humidity, wind speed and daily precipitation on the risk of forests fire occurrence in Lebanon. These impacts impose the adaptation of certain techniques that could help to predict fires and thus avoid their happening. Artificial Neural Networks have been utilized for the purpose. The weather data of the year 2012 collected from North Lebanon, Kfarchakhna station are taken for study. We have studied the effects of both the number of neurons in the hidden layer and the training technique on the network's performance and the mean squared error. This is an indication of the good performance of such network in adopting its predicting decision.