Prediction Model Of Fuel Moisture Content Based On Mixed Effects Model

In this paper,a real-time fuel moisture prediction model with the mixed effects is built based on the method of nonlinear mixed effect model(NLMEM).Among this model,the stand factors and spatial position are as the stochastic effect.The fuel moisture of this new model is related to the fuel moisture which were collected from three typical forests including Larixgmelinii,Quercus mongolica Fischer and mixture of Larix gmeliniiand and Betula platyphylla.And these data set are provided by the Nanwenghe Ecological Station located at Great Xing'an Mountains forest,Heilongjiang,China.This new model can solve the problems of heteroscedasticity by adding more weights to the residual variance.The results show the fitting effect of the real-Time NLME prediction model with the random effects and heteroscedasticity are better than it without considering the random effects.The random effects of mixed effect model is based on the type of stand factors,and the accuracy of the model(AIC=547.72,BIC=581.29,-2LL=527.72)using a constant plus power function as the heteroscedastic structure is the highest.It is obviously superior to the real-time NLME prediction model(AIC=961.65,BIC=988.50,-2LL=945.65)without adding more weights to the residual variance.The random effects of mixed effect model is based on spatial position,and the accuracy of the model using exponential function as the heteroscedastic structure is the highest.It is obviously superior to the real-time NLME prediction model(AIC=522.29,BIC=545.82,-2LL=508.29)without adding more weights to the residual variance.The results from the model test by using independent samples shows that the test precision of the NLME model with the random effects and heteroscedasticity structure has some improvements in comparison to the multivariate nonlinear regression model fitted by the least square method.The real-time fuel moisture content prediction model is based on the mixed effects method which can well describe the change laws in fuel moisture content for different types of forest and spatial position in the regional scale of interest.