Evaluation And Optimization Of Cropland Light Use Efficiency Models Under Increasing Diffuse Radiation Fraction

In recent years,the impacts of global climate change on the carbon cycle of terrestrial ecosystems have become increasingly difficult to predict.As a critical parameter for remote sensing estimation of Gross Primary Production（GPP）and an essential expression of plant photosynthesis,Light Use Efficiency（LUE）plays a crucial role in understanding cropland ecosystem function and supporting accurate remote sensing monitoring of global or regional crop productivity.Accurate estimation of crop LUE is dependent on a range of factors,including Diffuse Photosynthetically Active Radiation(PAR_diff),which directly affects the photosynthetic rate of vegetation.Therefore,an accurate assessment of the impact of PAR_diff on crop LUE is beneficial for understanding the impact of global climate change on the carbon cycle of terrestrial ecosystems.This paper aims to assess the extent to which the LUE estimation models consider diffuse radiation conditions in their modeling.Additionally,we optimized the LUE estimation model based on the Diffuse Photosynthetically Active Radiation Fraction(FDIFF_PAR)parameter,resulting in the development of the Diffuse Photosynthetically Active Radiation Light Use Efficiency Model（DP-LUE）.Finally,we utilized the DP-LUE model to explore the spatial and temporal characteristics of the LUE_max parameters of regional crops in China from 2000 to 2021.The main findings of this study are presented below.（1）This paper evaluated six remote sensing models（CASA,MOD17,EC-LUEa,VPM,EC-LUEb,and TL-LUE）and two crop production models（DSSAT and APSIM）to determine their ability to accurately estimate crop LUE.Results showed that these models were not able to accurately estimate crop LUE,with R² values less than or equal to 0.250.The remote sensing two-leaf models were found to be less applicable in predicting crop GPP,mainly due to their inability to distinguish between the effects of diffuse and direct radiation on crop canopy light interception in agroecosystems.In contrast,the crop production models performed better than the remote sensing models in predicting crop LUE.This was mainly attributed to their full consideration of soil water stress,whereas only the EC-LUEa model(R²=0.250,RMSE=0.868 g C^-1MJ^-1,Bias=-0.005 g C^-1MJ^-1)performed better among the eight representative models.This was mainly attributed to the use of a more accurate environmental stress calculation scheme in this model.（2）The results of this study indicate that PAR_diff positively influences crop LUE by promoting crop LUE_max and,consequently,LUE.Crop LUE was found to be actively correlated with FDIFF_PAR.Among the eight LUE estimation models evaluated,only the MOD17 and TL-LUE models were able to simulate the variation of crop LUE with FDIFF_PAR to some extent.The MOD17 model categorizes crop growth into cereal and broadleaf types,while the TL-LUE model considers the variation of light interception with the proportion of diffuse radiation when simulating LUE.Diffuse radiation can enhance photosynthesis both directly and indirectly by improving environmental conditions and alleviating photoinhibition.The study found that crop LUE_max increased with increasing FDIFF_PAR,a phenomenon that was not captured by any of the eight models evaluated.The crop LUE_max simulated by the remote sensing two-leaf models exhibited fluctuating characteristics with FDIFF_PAR,mainly due to their separation of the vegetation canopy into sunlit and shaded leaves.In contrast,the LUE_max parameters of both the remote sensing big-leaf models and crop production models remained constant.（3）Fully considering the effect of FDIFF_PAR on crop LUE_max when modeling can improve the accuracy of LUE prediction.In this study,we re-optimized the EC-LUEa model by combining FDIFF_PARparameters,while fully considering the effect of diffuse fertilization on the photosynthetic rate of crops.This resulted in the development of the DP-LUE remote sensing big-leaf model,which significantly considered the effects of PAR_diff and direct PAR on the LUE_max of shaded and sunlit leaves,respectively.The DP-LUE model demonstrated higher applicability in simulating crop GPP,while also significantly improving the prediction accuracy of crop LUE and simulating the effect of FDIFF_PAR on crop LUE.Compared to the EC-LUEa model,the DP-LUE model showed an 18.80%improvement in R²,while the RSME and Bias decreased by 19.82%and 100%,respectively.The DP-LUE model retained the advantages of the EC-LUEa remote sensing model while fully considering the effect of FDIFF_PAR on vegetation LUE_max.（4）Crop LUE_max is a dynamic parameter with temporal and spatial heterogeneity.This study explores the spatial and temporal variation characteristics of regional crop LUE_max in China from 2000 to2021 using the DP-LUE model combined with the LUE model inversion method.The annual mean crop LUE_max from 2000-2021 showed an overall fluctuating decreasing trend,while the seasonal means exhibited the following values:summer(0.1064 g C MJ^-1)>winter(1.052 g C MJ^-1)>spring(1.050 g C MJ^-1)>autumn(1.048 g C MJ^-1).The monthly variation of crop LUE_max in the Chinese region showed an overall fluctuating trend,peaking in June-August.The high value areas of interannual,seasonal,and monthly variations of crop LUE_max were mainly concentrated in central and eastern China,as well as in southern North China,southeastern Northwest China,and northeastern Southwest China.In contrast,the low value areas were mainly distributed in northeastern and southern China,as well as in northern North China,northwestern Northwest China,and southeastern Southwest China.Overall,these findings reveal the dynamic and heterogeneous nature of crop LUE_max in China and highlight the importance of considering both temporal and spatial factors in crop LUE modeling.