Estimation Of Downward Radiation Based On Multi-source Data Fusion

As one of the important components of surface radiation and energy budget,shortwave downward radiation is a key factor of surface process models such as hydrology,ecology and biogeochemistry.Shortwave downward radiation plays a vital role in estimating the surface energy budget,quantifying the changes in surface energy and the change of climate.Meanwhile,shortwave downward radiation is also an important driving factor for the earth circulation and climate change.Therefore,it is fundamentally important to estimate the accurate of shortwave downward radiation.As the vital sources for shortwave downward radiation,remotely sensed and reanalysis data have their own advantages and disadvantages.Remotely sensed observations have a higher accuracy while the spatiotemporal continuity is not sufficient.Conversely,the reanalysis data have a good spatiotemporal continuity,while the accuracy is rather limited.Thus,the attempt of this study is to develop a suitable method to fuse the remotely sensed and reanalysis data.This study explores two major aspects to solving the shortcomings in the study of shortwave downward radiation estimation.Firstly,there is a lack of systematic evaluation for shortwave downward radiation on specific application types in the evaluation of the accuracy and applicability of shortwave downward radiation products.This study evaluates the accuracy of remotely sensed CERES-SYN and reanalysis ERA-5 datasets based on the different four land use data including city,vegetation,water and snow or ice.Secondly,in the view that the spatial resolution and accuracy of existing products need to be further improved,the paper fuse the remotely sensed CERES-SYN data with higher accuracy and the reanalysis ERA-5 data with spatial-temporal continuity and higher temporary resolution to estimate shortwave downward radiation with high accuracy and spatial-temporal continuity.The main conclusions are as follows:(1)The spatial suitability of remotely sensed CERES-SYN dataset and reanalysis ERA-5 product were evaluated according to different land use cover types.The results show that CERES-SYN dataset performs better for city,vegetation and water types,while the ERA-5 product has higher accuracy for snow or ice type.Datasets in urban and water types overestimate the shortwave downward radiation,and snow or ice type underestimated the shortwave downward radiation.SYN data in vegetation types are underestimated in the second and third quarters,while other quarters are overestimated,which may be related to the growth pattern of vegetation.But the bias of the third and fourth quarters of the vegetation type is smaller than the first and second quarters.The radiation value of the water type has little change in the four quarters.The bias of the city type of the third and fourth quarters are smaller than that of the first and second quarters.All data has large bias in the four quarters of snow or ice types,and there are a large number of outliers in the second and third quarters,which have great impact on the assessment of data quality.(2)Based on the above analysis,a data fusion framework is developed to generate high-precision and high-resolution shortwave downward radiation results.The evaluation results show that the accuracy of the fused data in the city type has been slightly improved,in which the annual average bias is decreased by 30.52%,while the annual average RMSE is decreased by 2.14%.The shortwave downward radiations before and after the fusion of the city type are all overestimated.The accuracy of the data in vegetation types has been greatly improved.The annual average bias is decreased by 83.42% and the RMSE is decreased by 8.36%.The ERA-5 data of vegetation were underestimated in July,and were overestimated in the remaining months.While the fusion product was underestimated from June to September,and the rest of the month was overestimated.The accuracy of the water type was also improved,of which the annual average bias is decreased by 48.73% and the RMSE is decreased by 6.85%.The radiation values before the fusion were overestimated,while the data after the fusion were underestimated in October,and were overestimated in the rest months.In summary,a fusion scheme is determined through an analysis on of different landcover types.And the developed fusion framework is relatively effective,so that the accuracy of shortwave downward radiation on different landcover types has been improved to varying degrees.