Estimation Of Solar Radiation Using NASA/POWER Data

As the primary energy source for everything on the earth, solar radiation is the most important factor of the earth’s climate formation, and has crucial significance on surface radiation balance, energy exchange and climate formation. Accurate solar total radiation data is not only very important to the radiation evaluation of climate system and agricultural growth simulation and yield prediction, but also essential demand of the effective design and operation, and energy load estimation of solar system. Because of the lack of solar radiation observation, accurate estimation of solar radiation become research focus.In this paper, site models, five zoning models and unified model of mainland China were established and validated based on solar radiation data from ground observations and NASA Prediction Of Worldwide Energy Resources (NASA/POWER) project data on a 1 ° x 1 ° geographic coordinate grid of 88 sites. The main results of research are as follows:(1) It was found that NASA/POWER data showed the best agreement with ground observation data by comparison of the acquired data, with many correlations of 0.9. Correlation coefficients of 95% of sites were more than 0.8, and only values of five sites were less than 0.80. The overall average of NASA/POWER data was 14.88 MJ m-2 d-1, higher than ground observation of solar radiation. The difference of each site varied between 2.69 and 1.64 MJ m-2 d-1.(2) Site models based on NASA/POWER solar radiation data were developed using unitary linearity regression. In the site models, the MBE (mean bias error) and CRM (coefficient of residual mass) from model of Hailar station were the lowest, was 0.01 MJ m-2 d-1 and 0, respectively. There are 37 stations (most of them are in north China, including Tibet, Qinghai, Gansu, Ningxia, Inner Mongolia and northeast three provinces) had good performance(10%<rRMSE(%)(relative root mean square error)<20%), and 46 stations (mainly distributed in north China, southwest region) had fair performance (20%< rRMSE(%)<30%), and only five sites (Mianyang, Chengdu, Mount Emei, Changning, Ganzhou) had poor performances (30%<rRMSE (%)<40%).(3) Five solar radiation zones were determined by self-organizing neural network (SOM) clustering analysis method. Results showed that zone 5 (mainly including northwest China) had most abundant solar radiation resources with annual average daily solar radiation of 16.57 MJ m-2 d-1, followed by zone 4 (including Yunnan-Guizhou Plateau, western Sichuan, eastern Qinghai, southeastern Tibet) which had annual average daily solar radiation of 15.78 MJ m-2 d-1. Zone 1 (mainly including southwest China) had lowest solar radiation resources with annual average daily solar radiation of less than 9.5 MJ m-2 d-1. For different zones and the whole mainland China, models based on the geographic information of actual site and grid cell were developed by establishing the relationship between coefficient of each site model and geographic information of actual site (longitude, latitude, altitude) or geographic information of grid cell (longitude, latitude, altitude of grid cell). The average difference of models based on the geographic information of actual site and grid cell was very small, and the error values of zone 3 (In northeast and north China) was minimum. Compared with site model, the accuracy of models has little difference. The imitative effect of total radiation model presented in this thesis is satisfactory and could be used to estimate solar radiation in mainland China.(4) In addition, the results of this study also showed that NASA/POWER data is relatively reliable. They represent a valuable source of solar radiation data for research concerned with regional to global geographic scales.