Study On High Resolution Solar Radiation Simulation Based On Multi-source Remote Sensing Data

Solar radiation is the main source of energy on the earth,and it is also an important input parameter for models of hydrology,ecology,and climate change.Therefore,Research on solar radiation is extremely important.Remote sensing provides continuous spatial data and it is an important source of data for solar radiation simulation.Although there are already many free radiation products released at present,the general spatial resolution is low,and it is difficult to meet the analysis and application of research solar radiation in regions or small and medium scales;the resolution of the solar radiation is low by using a relatively low resolution meteorological satellite,and the model of regional scale or basin scale needs input points.Therefore,based on the shortcomings of the above research,this paper aims to simulate the solar radiation data with high spatial resolution.The purpose of this study is to simulate the high spatial resolution of the solar radiation data,but the spatial resolution of the TRMM precipitation data is low(0.25degrees),so it is necessary to downscale the TRMM3B43 monthly precipitation data as the input parameter of the neural network.Because of the greater correlation between the enhanced vegetation index(EVI),DEM,slope,aspect and the precipitation,this paper uses the method of geographic weighted regression(GWR)to downscale the TRMM precipitation data,and improve its spatial resolution to 1km.The results of TRMM monthly precipitation data reduction are verified.The results are that the average correlation coefficient R is 0.92,MBE is-1.07 mm,and RMSE is49.78 mm.Before and after the reduction,the correlation with the measured value has almost unchanged,MBE becomes smaller,RMSE is also reduced,and the average deviation is still stable at about ± 5mm after reducing the scale.As a matter of fact,the accuracy of precipitation increases after the downscaling,and the spatial variation is more detailed.The artificial neural network model can well express the complex nonlinear coupling relationship between the solar radiation intensity and the influence factor.Compared with other empirical models,the artificial neural network model is simple,and it is widely used in meteorological geography.The simulation of the solarradiation is also of high precision by using neural network method.So this paper chose the neural network model to simulate the solar radiation.The average day and night temperature(Tm),day and night temperature difference(?T),precipitation(P),enhanced vegetation index(EVI)and air pressure ratio(?)are selected as input parameters of the simulated solar radiation,and the monthly mean of the total solar radiation(S)is used as the output simulation of the neural network to obtain spatial continuous spatial distribution of the solar radiation.The monthly data of 17 radiation stations were selected to verify the simulation results.The results showed that the average R was 0.85 and MBE% was 14%.The simulation results of each site are compared and analyzed.The results show that the accuracy of most sites is very good.In addition to the Ankang site,the R of the rest of the sites is more than 0.8,of which,the simulation results in Gangcha,fruit and Guyuan sites are very good,but the simulation results in Hongyuan,Taiyuan and Ankang sites are not good.Through the study of this paper we have obtained the following main conclusions:1.The precipitation accuracy of TRMM precipitation based on the GWR downscaling method is high,indicating that the method can be applied to the downscaling of precipitation in the Loess Plateau.2.Using neural network and multi-source remote sensing products,we can simulate the continuous spatial distribution of solar radiation data.3.When using neural network to simulate solar radiation,high resolution solar radiation data can be obtained by improving the resolution of input parameters.This provides a good method for the preparation of high resolution solar radiation data sets.4.The solar radiation of the Loess Plateau in 2015 is the first trend to increase and then decrease,the most in July,and the least in January,the distribution of solar radiation is higher in the northwest and southwestern regions,but the distribution of solar radiation in the southeastern regions is less.