A hybrid deep learning framework for regional reference crop evapotranspiration estimation in the Hetao Irrigation District using limited meteorological data
Study region: Hetao Irrigation District, China Study focus: Reference crop evapotranspiration (ETo) estimation is crucial for further calculating crop evapotranspiration and supporting water allocation decisions. The Penman Monteith 56 (P-M 56) formula is considered as the standard method for estima...
Saved in:
| Main Authors: | , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Elsevier
2025-10-01
|
| Series: | Journal of Hydrology: Regional Studies |
| Subjects: | |
| Online Access: | http://www.sciencedirect.com/science/article/pii/S2214581825005385 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Study region: Hetao Irrigation District, China Study focus: Reference crop evapotranspiration (ETo) estimation is crucial for further calculating crop evapotranspiration and supporting water allocation decisions. The Penman Monteith 56 (P-M 56) formula is considered as the standard method for estimating ETo but requires extensive meteorological data, limiting its use in data-scarce regions. In response to this concern, this study proposed two integrated deep learning models, i.e., CNN-Transformer and CNN-Informer, to estimate ETo based on three meteorological factor input combinations (temperature-based, radiation-based, and mass transfer-based). New hydrological insights for the region: The results indicated that the mass transfer-based CNN-Informer3 model achieved the best estimation performance, with its average Determination Coefficient (R2), Nash-Sutcliffe efficiency coefficient (NSE), Root Mean Square Error (RMSE), Relative Root Mean Square Error (RRMSE), Mean Absolute Error (MAE) and Percentage Bias (PBIAS) of 0.9712, 0.9665, 0.3764 mm/d, 0.1192, 0.2591 mm/d and -3.8143 % for accuracy evaluation, significantly outperforming other models. In conclusion, the key contribution of this study is the proposal of a hybrid deep learning framework that expands the applications of emerging deep learning algorithms and ensemble learning in the field of evapotranspiration estimation, which enables a more precise matching between irrigation water extraction and actual crop needs. Regional hydrological processes can thus better adapt to natural fluctuations, reducing the risks of droughts (due to insufficient irrigation) or waterlogging (due to excessive irrigation). |
|---|---|
| ISSN: | 2214-5818 |