Spatial Distribution Of PM_2.5Concentration From Surface Fuel Combustion Of Pinus Koraiensis Plantations Based On Indoor Simulation

With the global warming and frequent forest fires,a large number of smoke from forest burning is discharged into the atmosphere,which seriously threatens the safety of ecological environment and people's lives.PM_2.5 is one of the main products of forest fires.It is of great significance to study on the influential factors and spatial distribution of PM_2.5 emissions for the future PM_2.5 occupational exposure standards,forest fire smoke management and forest fire management in China.In this study,Pinus koraiensis plantation,a typical coniferous tree in the Mao'ershan area of the Northeast China,was selected as the research object.Surface fuels in the forest were used as experimental materials.Fuel beds with different moisture content and load were constructed.144 times burning experiments were carried out under different wind speeds by using wind tunnel device.PM_2.5 particles at different spatial points were collected and their concentrations were calculated.According to the results,the spatial distribution model of PM_2.5 concentration was established based on BP neural network,and the model was analyzed.The results are as follows:?1?It is found that PM_2.5 concentration is positively correlated with wind speed within designed wind speed;PM_2.5 concentration is positively correlated with fuel moisture content under the condition of single variable,but it is easily affected by the comprehensive effect of wind speed and fuel loading;PM_2.5 concentration is positively correlated with fuel load.?2?In the laboratory scale of spatial points,from PM_2.5 concentration which collected at each point in the horizontal and vertical directions,the overall trend is that PM_2.5 concentration increases along the horizontal downwind direction and the vertical upward direction.?3?Based on BP neural network,the spatial distribution model of PM_2.5 concentration with single hidden layer was established.The prediction accuracy of modeling samples and validation samples is balanced when hidden layer nodes is 5.The RMSE of modeling samples is 750.3?g·m-3,MRE is 53.5%,and MAE is 582.2?g·m-3,the RMSE of the validation samples is 985.6?g·m-3,MRE is 53.6%,and MAE is 582.2?g·m-3.The R2 of the measured and predicted values of modeling samples and validation samples are 0.742 and 0.569,respectively.The model has good prediction accuracy,fitting accuracy and high reliability.