This thesis took the three essential factors of forest fire (fire source, environment, and litter) as the point of departure,and has approached the forecast method of forest fire in tropical area of southwest China in a macroscopic scale.

This study supported by mathematical method was from a new angle. For example,in the aspect of forest fire forecast, the main cause of forest fire was artificial fire, which had lots of randomness. So Markov random processes could be employed in the study. In the aspect of the analysis of environment that contributes to forest fire,correlation would reveal the relation among cosmos earth and life,for instance,the relation between the El Nino (or La Nina) and the forest fire. Especially in the aspect of estimating the amount of litter,a new estimate method based on space remote sensing data was raised. It provided an effective and quantitative way with higher reliability for forest fire forecast on region scale.