Data for "A climate similarity based global 0.1° daily evapotranspiration product from 1980–2025" (1980-1999)

You can access the remaining part of the dataset via Qian et al. (2026) using the following reference:Qian, L., Wu, L., & Yu, X. (2026). Data for "A climate similarity based global 0.1° daily evapotranspiration product from 1980–2025" (2000-2025) (Version 1) [Data set]. Zenodo. https://doi.org/10.5281/zenodo.18891803This dataset provides a long-term global daily evapotranspiration (ET)product generated using the Climate-Similarity-constrained deep learningframework (CLSIM). The dataset covers the period from 1980 to 2025 witha spatial resolution of 0.1° (~10 km) and daily temporal resolution. Evapotranspiration (ET) is a fundamental component of the global watercycle, linking land surface energy balance, ecosystem carbon exchange,and hydrological processes. However, existing global ET datasets derivedfrom reanalysis, land surface models, and remote sensing often showlarge discrepancies, particularly across heterogeneous climate regimesand under extreme climatic conditions. To address these limitations, the CLSIM framework integrates aclimate-similarity-based spatial upscaling strategy with deep learningmodels trained using global eddy covariance observations. Each globalland grid cell is assigned to its most representative flux tower sitebased on similarity in long-term climate statistics, and ET is thenpredicted using a site-specific CNN–LSTM model trained at that site. In addition, an Extreme-Climate-Aware Weighted Loss (ECA-WL) strategyis introduced during model training to improve ET estimation underextreme climatic conditions such as heatwaves, high vapor pressuredeficit (VPD), and drought. The resulting CLSIM dataset provides a spatially and temporallyconsistent global ET record that is suitable for a wide range ofapplications, including climate change studies, ecohydrology,water resource assessment, drought monitoring, and hydrologicalmodel benchmarking.