Forest fire threatens forest ecosystem stability,causing incalculable loss to forestry and forestry economy.As the basis material of the forest fire,forest fuel moisture content determines the difficulty of a fire and burn.For fire prevention and the control of forest fire spreading,a study on the change law of fuel moisture content has a vital significance and also provides a fundamental theory for forest fire forecast.First,based on varying-coefficient regression model,meteorological factors and current fuel moisture are chosen as explanatory variables and observed variables respectively.By using these variables a function and thrice function of fuel moisture content varying-coefficient regression models are built.As result the model accuracy are 92% and 94%,respectively.As a comparison,the fuel moisture content constant coefficient regression model based on the theory of linear regression and balance moisture content model exhibits a lower accuracy of 85%.Therefore the feasible and practical value of fuel moisture content varying-coefficient regression prediction model are determined.Furthermore,anther fuel moisture content prediction model based on semiparametic varying-coefficient regression model are established.This model shows an accuracy of 97% which indicts the feasibility of fuel moisture content semiparametic varying-coefficient regression prediction model.Finally,this paper discusses the different among linear regression model,varyingcoefficient regression model and semiparametic varying-coefficient regression model.Linear regression model only reflect the average impact response of influence factors.And varyingcoefficient regression model and semiparametic varying-coefficient regression model reflect variations on impacts when influence factors are in different states.A conclusion is drown that the semiparametic varying-coefficient regression model takes a lead on performance on simulation prediction precision,but varying-coefficient regression model is advantage in terms of convenience. |