SiB2 Simulation And Random Forest Method Correction Of Turbulent Heat Flux And CO₂ Flux In Wheat Fields In East Chin

The exchange of energy,momentum,water vapor,and carbon dioxide（CO₂）between the ground and the atmosphere has a significant impact on weather and climate.It is essential for both single-point and regional atmospheric simulations.However,there are still uncertainties in the simulation results of land surface processes in the wheat field ecosystem in East China.This study analyzed the characteristics of radiation,heat flux,and CO₂ flux in a wheat field in Dongtai City,East China from January to May 2015.The Second Simple Biosphere Model（Si B2）was used to simulate these characteristics,and the performance of the Si B2 model on this specific land surface was studied based on Eddy Covariance（EC）observational data.Subsequently,the Si B2 model’s simulation results were corrected using the Random Forest（RF）machine learning algorithm.The results are shown as follows:（1）Meteorological elements,radiation,heat flux,and CO₂ flux not only have daily variations,but also change continuously with the growth of wheat.Latent heat flux plays a dominant role in surface energy allocation,followed by sensible heat flux.There is an energy imbalance in each growth stage,and the energy imbalance at night is higher.（2）The soil,atmospheric,and vegetation parameters needed for Si B2 were accurately determined,and Si B2 was used to dynamically simulate net radiation,sensible heat flux,latent heat flux,soil heat flux,and CO2 flux.The simulated results were consistent with the observed daily variation characteristics,but there were some deviations in size.Net radiation,sensible heat flux,and latent heat flux were overestimated by 13%,25%,and 36%,respectively;while soil heat flux and CO₂ flux were underestimated by 37%and 40%,respectively.The determination coefficients（R²）between the simulated net radiation,sensible heat flux,latent heat flux,soil heat flux,and CO₂ flux and the measured values were 1.00,0.59,0.75,0.60,and0.62,respectively.（3）A Random Forest correction model based on Si B2 was successfully constructed,and the corrected daily variation results were more consistent with the on-site observed values.The determination coefficient between the simulated net radiation and the actual observation was1.00,so net radiation was not corrected.After correction,sensible heat flux and latent heat flux were overestimated by 15%and 6%,respectively;while soil heat flux and CO₂ flux were underestimated by 20%and 17%,respectively.The determination coefficients between the corrected sensible heat flux,latent heat flux,soil heat flux,and CO₂ flux and the measured values were 0.99,0.85,0.78,and 0.71,respectively.（4）For each growth period,the mean value of R² for four fluxes simulated by the Si B2model had the best performance in the reproductive stage except for the net radiation,the mean value of R² for four fluxes revised by the RF model had the best performance in the ripening stage.For the whole growth period,R² of latent heat fluxes simulated by the Si B2 model was the highest except for the net radiation,R² of sensible heat fluxes revised by the RF model was the highest.