降雨条件下蒙古栎和樟子松林地表死可燃物含水率预测方法的修正与比较

[Objective]Analyze the impact of rainfall conditions on the moisture content of surface fuel moisture content(FMC),establish machine learning models and traditional models to predict the moisture content of surface fuel under rainfall conditions,and evaluate and compare the established prediction models.[Method]Selecting the broad-leaved forest Quercus mongolica and coniferous forest Pinus sylvestris var.mongolica in northeast China,surface dead fine fuel as the research object,by indoor simulated rainfall experiment to explore the influence of rainfall for the surface fuel moisture content,combined with the direct estimation method to modify the fine fuel moisture content prediction model,finally using field data will be revised directly estimation method,the direct estimation method and convolution neural network(CNN)model is used in the comparison.[Result]Compared with the unmodified direct estimation method,the modified direct estimation method significantly improves the prediction accuracy,and the goodness of fit(R2)increases from 0.85-0.94 to 0.94-0.96;MAE decreased from 9.18%-18.33%to 6.86%-10.74%,MRE decreased from 3.97%-17.18%to 3.53%-14.48%.Compared with the convolutional neural network model,the R2 is above 0.90,the MAE is below 8.11%,and the MRE is below 8.87%,indicating high prediction accuracy.[Conclusion]Under the condition of rainfall simulation results show that indoor fuel moisture content in the form of logarithmic increase trend.Rainfall and last fuel moisture content have a significant impact on the fuel moisture content prediction model,and both the relational model and nonlinear model perform well in predicting fuel moisture content under indoor rainfall conditions.Comparing the three models,it can be concluded that the modified direct estimation method has the best prediction effect.