Estimation Models Of Solar Radiation At Different Time Scales

In agricultural meteorology, accurate solar radiation data is crucial for crop simulation. However, compared to sunshine hours and temperature, solar radiation data is scare. So it is very necessary to find a way to estimate solar radiation with conventional meteorological elements.With the development of science and technology, solar radiation is not the most suitable. Sometimes, we need to be more accurate hourly solar radiation for other models. So it is important to find a simple method for estimating the hourly solar radiation. Solar radiation can provide data support for other models.The time serial of solar radiation and sunshine hours at six sites were analyzed in this study. The models based on sunshine duration were constructed for daily, monthly, seasonally and yearly solar radiation estimation. Moreover, this study focused on the interchangeability of the coefficients at different time scales. There are22sites, which have hourly solar radiation. Daily solar radiation, which can estimate hourly solar radiation with Collares-Pereira model, was estimated in this paper using the coefficients at four time scales. Results showed that:1) In addition to Geermu site and Harbin site, yearly mean daily solar radiation of the other four sites showed obvious downward trend from1961to2000. In four seasons, the volatility of solar radiation in summer was the largest, followed by spring. The trend of multi-year mean solar radiation for12months showed unimodal type. Solar radiation values are generally lowest in winter (January or December), the highest in summer. The change trend of annual average sunshine hours in each site was similar, presented a stronger or weaker declining trend. In various seasons, sunshine durations in summer showed the largest range, followed by winter. Volatility of sunshine hours in spring and autumn was weaker. The trend of multi-year mean solar radiation for12months showed larger difference. To find out the relationship between sunshine duration and solar radiation at different time scales, Pearson correlation coefficients were calculated and showed from high to low in turn:daily time scale>monthly time scale> seasonally time scale>yearly time scale.2) The Angstrom-Prescott model was used to estimate global solar radiation at different time scales., Mean R2of all stations showed from high to low in turn:daily time scale>monthly time scale> seasonally time scale>yearly time scale. Through the Pearson correlation analysis, the relationship between the calibrated coefficients and geographical information of stations was concluded. The results showed that the western highland area tended to have more clear days than the eastern lowland area. On the contrary, coefficient a and (a+b) were negatively correlated with longitude which meant a general increasing trend for a and (a+b) with the decreased longitude from east to west in the study area.3) The calibrated coefficients at four time scales were interchanged in this study. Nevertheless, further analysis demonstrated that the calibrated coefficients at monthly time scale could be applied to estimate global solar radiation at larger time scales with an acceptable accuracy. On the contrary, the calibrated coefficients at larger time scale were applied to estimate global solar radiation at daily time scale with a bad accuracy.4) The parameter a and b for each site, which were different with the change of sunset angle every day, were concluded according to the Collares-Pereira model in this paper. For parameter a at each site, it was increased at first and then gradually decline. Peak value of a in the southwest of China were smaller, but they were generally high in the northeast, north and northwest of China. In contrast with the parameter a, the parameter b were declined d at first and then increase.5) The paper used six stations (Harbin, Beijing, Wuhan, Chongqing, Hangzhou and Lanzhou) as example to analyze results of hourly solar radiation estimation. Better fit were found at yearly time scale between the observed and predicted hourly solar radiation, the determination coefficient (R2) in Harbin station, Chongqing station, Hangzhou station and Lanzhou station were0.8493,0.7873,0.8362and0.8645respectively. However, R2in Beijing and Wuhan station were0.8619and0.8183respectively, which were found at daily time scale.6) The Collares-Pereira model was used to estimate hourly solar radiation with global solar radiation, which could be estimated by the calibrated coefficients at four time scales. Better fit were found at daily and monthly time scales (RMSE=0.51MJ·m-2·h-1, MAE=0.26MJ·m-2·h-1) between the observed and predicted hourly solar radiation. Next, RMSE between the observed and predicted hourly solar radiation at seasonally time scale was0.52MJ·m-2·h-1. The worst fit were found at yearly time scale. Pearson correlation coefficients were calculated between root mean squared error and geographical information of stations. RMSE were negatively correlated with longitude which meant a general increasing trend for RMSE with the decreased longitude from east to west in the study area. However, RMSE were positively correlated with latitude and elevation.