The Study Of Forest Fuel Moisture Content Modeling

Fuel moisture content (FMC) is a very important component in fire danger rating, which has a direct influence on the difficulty level of fire and an indirect influence on fire intensity as well as fire spread rate, so its accuracy becomes the crucial determination of fire behavior and fire weather forecasting. At present, for predicting FMC there are four important methods which are method based on equilibrium moisture content (EMC), method based on regression of meteorological elements, method based on telemetry and method based on processing modeling. However, mathematical modeling is important component of all these methods above. So, models describing the moisture content of forest fuels are an integral component of most fire behavior prediction systems.In this paper, by analyzing the observed data of forestry station of Hailin Shihezi and Hailin Xingnong, the varying prediction model for forest fuel moisture by differential equations, fuel moisture model based on EMC method and model based on BP neural net are built. The results indicate as below:Firstly, the varying prediction model for forest fuel moisture by differential equations can predict fuel moisture content one hour later according to before time temperate, relative humidity and wind speed. The model has an average precision of 98%, which means this model mirrors the relation between the fuel moisture content with before time temperate, relative humidity and wind speed.Secondly, in the initial analysis of predicting fuel moisture content by method based on EMC, by analyzing the observed data of forestry station, one hour is taken for the fuel's time lag and the EMC model is built which the fuel moisture content can be predicted by. The fuel moisture contents predicted conform to the real truth, although the model's correlation coefficient is not as good as the varying prediction model.At the last of this paper, the method for predict fuel moisture content based on BP neural net is given. Air temperate, relative humidity and wind speed are taken as the basic input characteristic quantity, fuel moisture content taken as the output characteristic to built BP neural networks models. By simulating, the relative errors are less than six percentage points, which means predicting fuel moisture content based on BP neural net can be a way for predicting fuel moisture content and be popularized in the future.