1. Estimation of shortwave solar radiation using the artificial neural network from Himawari-8 satellite imagery over China.
- Author
-
Peng, Zhong, Letu, Husi, Wang, Tianxing, Shi, Chong, Zhao, Chuanfeng, Tana, Gegen, Zhao, Naizhuo, Dai, Tie, Tang, Ronglin, Shang, Huazhe, Shi, Jiancheng, and Chen, Liangfu
- Subjects
- *
PHOTOSYNTHETICALLY active radiation (PAR) , *SOLAR radiation , *ARTIFICIAL neural networks , *REMOTE-sensing images , *STANDARD deviations , *GEOSTATIONARY satellites , *ORTHOGONAL functions - Abstract
• The ANN algorithm model can quickly and accurately calculate the hourly SSR by directly connect the satellite data and in-situ SSR measurements in a simple way rather than the complex scheme based on the lookup table estimation approach. • The ANN model was applied to Himawari-8 geostationary satellite data and the estimated SSR show a good agreement with in-situ observation. • We analyze the spatio-temporal characteristics of SSR in China on the hourly, daily and monthly scale in 2016. The previous studies have less to analyze the characteristics of SSR spatio-temporal at the daily scale, especially at the hourly scale. In this study, we use an artificial neural network (ANN) method to estimate the downward surface shortwave radiation (DSSR) over China from Himawari-8 geostationary satellite data. As the training data of the DSSR estimation algorithm, ground-observed DSSR (GOS) data is compiled to complete the ANN method. GOS data from 89 stations over mainland China in 2016 are divided into training, testing and validation samples with a proportion of 3:1:1, in order to perform the DSSR estimation and accuracy validation. As a result, estimated DSSR from Himawari-8 data in 2016 shows good consistency with validation samples of ground observed DSSR, holding the determination coefficient and root mean square error of 0.90 and 88.86 W m−2 for the hourly mean DSSR, and 0.96 and 24.46 W m−2 for the daily mean DSSR, respectively. To investigate the spatio-temporal variation of daily DSSR in China, we performed the first empirical orthogonal function (EOF) analysis based on the ANN-derived DSSR estimates. The spatial pattern of the first EOF mode reveals a larger DSSR variation in northeast and northwest China, in contrast to the smallest variation appeared in the southwest. The hourly ANN-derived DSSR analysis shows a peak point at noon (local time) in each region, particularly in the Tibetan Plateau. In terms of the monthly and annual DSSR spatial distribution, the regions with higher elevations, such as Northwest, Tibetan Plateau and Inner Mongolia Plateau in China, have larger DSSR than other regions. In contrast, relatively low DSSR appears in the southwest and the northeast of China through the year. [ABSTRACT FROM AUTHOR]
- Published
- 2020
- Full Text
- View/download PDF