Estimation method of daily global radiation under different sunshine conditions: A case study of Jiangsu Province

Global radiation is a key factor affecting carbon exchange and the surface energy budget of agroecosystems. To accurately estimate the daily global radiation (GR) under different sunshine conditions and to improve the research carried out on agroecosystems, this study used daily meteorological and radiation data collected between 2005 and 2020 at three radiation observation stations in Jiangsu Province, namely Huai’an, Lüsi, and Nanjing, to divide the research samples into two categories, namely with and without sunshine, according to whether the number of hours of sunshine per day was zero. In total, 24 observable meteorological factors and 3 geographical factors were identified, with the main factors influencing GR under different sunshine conditions being determined using correlation analysis. Daily data from the three stations collected during odd-numbered years between 2005 and 2016 were selected as the modeling dataset, and the least-squares stepwise regression method was adopted to establish the GR estimation models for conditions with and without sunshine, with GR and the daily atmospheric transparency coefficient (ratio of GR to sky radiation SR, GR/SR) representing the dependent variables. Daily data samples from the three stations collected during even-numbered years between 2005 and 2016 were selected as the between-group verification set, while daily data samples collected from 2017 to 2020 were selected as outside-group verification sets. The optimal GR estimation model for Jiangsu Province was determined by comparing the model fits and the estimation effects of the original models with the between-group and the outside-group verification sets. The results showed that first, GR was significantly correlated with most of the meteorological factors (P<0.01) regardless of the presence of sunshine. GR under sunshine conditions had the strongest correlation with sunshine factors, while GR under without sunshine condition had the strongest correlation with the daily maximum ground temperature (T_GMax). Furthermore, the correlation coefficient between GR and T_GMax was higher than the correlation between GR and other temperature factors. Second, the estimation model with GR as the dependent variable and T_GMax and daily dew point temperature as the independent variables was selected when the daily sunshine duration was zero; the coefficient of determination (R²) of this model was 0.650, and the estimation accuracy of GR was close to 75%. The estimation model with GR/SR as the dependent variable and daily percentage of sunshine and sunshine duration as the independent variables was selected when the daily sunshine duration was greater than zero; the R² of this model reached 0.769 and the average estimation accuracy of GR was 87.60%. On the basis of subsection of estimation models, the average accuracy of GR under different sunshine conditions in Jiangsu reached 84.71%, and the proportion of outliers in the total sample was 2.04%. The introduction of accurate GR estimation is greatly beneficial to carry out research on crop growth and yield simulation and soil moisture estimation, and ultimately provide a basis for related research on agroecosystems.