Publications

Zhao, L; Liu, Y; Xing, XG; Ma, XY; Cui, NB (2020). ARTIFICIAL NEURAL NETWORK APPROACH FOR EVAPOTRANSPIRATION DOWNSCALING BASED ON REMOTE SENSING. FRESENIUS ENVIRONMENTAL BULLETIN, 29(4), 2041-2054.

Abstract
Timely and accurate determination of evapotranspiration (ET) is crucial for water resource management. This study presented a method based on an artificial neural network (ANN) approach that can conduct ET downscaling with or without using thermal infrared data. However, obtaining ET data with both high temporal and spatial resolutions is a challenge because of the constraints of satellite technology. Therefore, this study adopted an ANN approach for ET downscaling to resolve the aforementioned problem. This approach provided ET data with high temporal and spatial resolutions. The path analysis results indicated that in the study areas (Wugong and Fufeng, China), the land surface temperature (LST) was the dominant factor that affected daily ET. Obtaining LST information from thermal infrared data is difficult because of the sensor's spatial and temporal resolutions, and an ANN approach that does not require thermal infrared data for ET downscaling is necessary. Results revealed a strong correlation between the normalized difference vegetation index (NDVI) and LST during summer. Based on this finding, the NDVI can be considered a predictor variable to obtain LST data by using moderate-resolution imaging spectroradiometer (MODIS) data. Finally, we applied the established ANN model to Huanjing- 1B data, and the results demonstrated that the ANN model can suitably estimate summer ET data by using a combination of downscaled LST data and a digital elevation model. Thus, the ANN model, which does not require thermal infrared data, can be accurately used for ET downscaling during summer.

DOI:

ISSN:
1018-4619