Estimation Of Land Surface Shortwave Downward Radiation With Remotely-sensed Data And Its Temporal Upscaling Methods

The earth’s surface downward shortwave radiation(SWDR)is generally defined as the sum of the incoming solar energy in the shortwave spectrum(0.3–3.0μm),after being attenuated by the atmosphere,received by the land surface.As a key parameter of surface radiation budget,SWDR is an essential driving parameter for many earth system models.Obtaining accurate information about its temporal and spatial changes is of great importance for the study of surface radiation balance,water cycle,ecosphere productivity,agricultural resource management,and renewable energy planning.With the successful launch of more satellites and the improvement of the observation capabilities of on-board sensors in the past few decades,the advantages of satellites being able to timely and accurately capture the spatial distribution and dynamic changes of surface radiation in a large area become more and more obvious.As a result,remote sensing is playing an increasingly vital role in the study of SWDR.Although academic circles have carried out large-scale research and encouraging results have been achieved,there are still some problems,including:it is still necessary to conduct more comprehensive assessment of the existing SWDR products;different components of SWDR have been given insufficient attention by the majority of surface radiation products or remote sensing algorithms now available;there is still a lack of comprehensive and systematic research and comparison of different temporal upscaling methods based on polar-orbiting satellite observations.To address these problems,this dissertation mainly focuses on:the validation of different SWDR products,the estimation of all-sky SWDR(different components),and the comparison of temporal upscaling methods based on instantaneous SWDR.The results of this dissertation are as follows:(1)Himawari-8,GLASS,CERES-SYN and ERA5 SWDR products,which were still lack of horizontal comparison in previous studies,are validated and evaluated with measured data from 89 CMA sites in 2016,and the factors affecting the accuracy of products are analyzed and discussed.Results show that when at a spatial resolution of 1°,in terms of the overall results of daily SWDR,among these products,the CERES-SYN product has the highest accuracy(with an RMSE of 31.0W/m~2 and a bias of 2.6 W/m~2),and the ERA5 reanalysis product has the worst accuracy(with an RMSE of 41.5 W/m~2 and a bias of 11.0 W/m~2),and the algorithm adopted by Himawari-8 cannot effectively distinguish cloud from snow,which makes SWDR underestimated significantly during the snow cover period(with an RMSE of 37.1 W/m~2 and a bias of 6.7 W/m~2).In terms of spatial distribution of the validation results,these products have the highest accuracy in northern China while the lowest in the Qinghai-Tibet Plateau.Among them,Himawari-8 and GLASS overestimate the daily SWDR at most sites in central-eastern China but underestimate the daily SWDR in the Qinghai-Tibet region(the bias values of this region are-13.9 W/m~2 and-22.4 W/m~2,respectively),and ERA5 has a positive bias at most sites of the country.It can also be found that the validation results of each product vary with months and seasons,and the robustness of CERES-SYN and GLASS is the best out of the four SWDR products.Analyses show that the relative root mean square errors of validation results are positively associated with cloud amount(cloud optical thickness),while a negative relationship can be seen between coefficient of determination and cloud amount(cloud optical thickness).(2)All-sky total and direct SWDR are estimated with a narrow-band LUT method.Applications of the LUT method to both Moderate Resolution Imaging Spectrometer(MODIS)and Himawari-8 measurements are presented,and validation versus ground measurements from 21 sites shows that both the total and direct SWDR can be accurately derived from the above-mentioned method.An additional comparison with the official Himawari-8 SWDR product reveals that the estimated values of the LUT method(with an RMSE of 91.77 W/m~2 and a bias of-3.91 W/m~2)performs better than the official Himawari-8 product(with an RMSE of 97.30 W/m~2and a bias of 24.22 W/m~2),which further demonstrates that the LUT-based SWDR values are believable.Compared with existing methods,the LUT approach only need solar and observing angles,blue band radiance,elevation,and surface climatological albedo as its inputs.The operation of this method is simple and convenient,while ensuring the accuracy of the estimated values and avoiding the problems caused by atmospheric products as the inputs of parameterization methods.(3)On the basis of previous research results,the existing temporal upscaling methods based on a priori knowledge of daily curves suitable for low-middle latitude area or high latitude area are summarized,improved,compared and analyzed.Then the effects of observation frequency,cloud and instantaneous SWDR with different biases on the temporal upscaling results are analyzed by taking the middle and low latitudes at low-middle latitude area.The results show that:temporal upscaling method with the best performance should be selected for the middle and low latitudes and high latitudes respectively;different from the net surface radiation,the RMSE of the temporal upscaling results under cloudy sky conditions significantly larger than that under clear sky conditions,indicating that clouds can markedly affect the accuracy of the temporal upscaling results for SWDR;within a certain limit,the temporal upscaling method has good robustness and the time upscaling results do not change significantly with different instantaneous SWDR.