Application of artificial neural network for forest fire risk mapping based on physiographic, human and climate factors in Sarvabad, Kurdistan province

The protective role of the Zagros forests in preventing soil and water erosion is very important. Therefore forest fires account as major environmental hazards for Zagros forests. The aim of this research was to study the influence of variables on fire occurrence and producing fire susceptibility map. Factors affecting fire incidences include altitude، slope، aspect، distance to residential areas، distance to roads and farmlands، temperature and rain. These factors were examined to determine the effect of each variable in occurrence of fire. Sampling procedures was performed in forest areas with previous fire records and in non- affected forest areas and MLP method was used to determine the importance of each factor in the occurrence of fires. Sensitivity maps of different areas to fire were produced using artificial neural network method. Results showed that the slope (100%) and distance to roads (95%) variables are most important factors influencing fire occurrence while aspect with 55% effect has the least correlation with fire occurrence. The model was further validated with Kappa coefficients. Results indicate that the sensitivity map is produced with 90% accuracy، and that 98% of the grid cells are correctly classified.