Forest Fire Risk Zone Modeling Using Logistic Regression and GIS: An Iranian Case Study

Forest fires are an important environmental concern worldwide, affecting the soil, forests and human lives. During the process of burning, soil nutrients are depleted and the soil is subsequently more vulnerable to erosion. Nowadays it is necessary to identify the factors influencing the occurrence of fire and fire hazard areas, in order to minimize the frequency of fire and avert damage. Logistic regression was used to study the forest fire risk and identify the most influential factors in the occurrence of forest fires. Climatic variables (temperature and annual precipitation), human factors (distance from streams and farmland) and physiography (land slope and elevation) were considered and their correlation with the occurrence of fires investigated. Results of model validation and sensitivity of various areas to fire were examined with the ROC coefficient and Hosmer–Lemeshow test. The estimated coefficients for the independent variables indicated that the probability of occurrence of fire is negatively related to land slope, site elevation and distance from farmlands, but is positively related to amount of annual precipitation.