CAMELE: Collocation-Analyzed Multi-source Ensembled Land Evapotranspiration Data

The Collocation-Analyzed Multi-source Ensembled Land Evapotranspiration Data (CAMELE) is a long-term daily global evapotranspiration (ET) product. It integrates inputs from ERA5L, FluxCom, and PMLv2.v017, GLDAS-2, and GLEAMv3.7 datasets. Optimal weights for the inputs are based on error information derived from the Double Instrumental Variable Technique (IVD) and Extended Instrumental Variable Technique (EIVD), minimizing the variance of random errors in the least squares sense.

CAMELE provides long-term daily global ET estimates at two resolutions: 0.1° for 2000-2020 and 0.25° for 1980-2022. Rigorous validation against in-situ observations has demonstrated the excellent performance of CAMELE across various vegetation coverage types. The validation process yielded favorable results, including Pearson correlation coefficients (R) averaging 0.63 and 0.65, root-mean-square-errors (RMSE) of 0.81 and 0.73 mm/d, unbiased root-mean-square-errors (ubRMSE) of 1.20 and 1.04 mm/d, mean absolute errors (MAE) of 0.81 and 0.73 mm/d, and Kling-Gupta efficiency (KGE) averaging 0.60 and 0.65 at resolutions of 0.1° and 0.25°, respectively.

In addition to the data, we provide example MATLAB codes to read and plot CAMELE data and employ IVD and EIVD methods to merge the inputs. Compared to the previous version, this new version.202306, we have made the following substantial improvements: (1) all inputs have been updated to their latest and most suitable versions; (2) a more rigorous approach to calculate optimal weights, considering non-zero error cross-correlation and proper rescaling, is used; (3) a much more comprehensive site-scale evaluation has been conducted to verify the reliability of CAMELE; (4) the data has been compressed for easier usage.

If you have any questions or identify any shortcomings of CAMELE, please do not hesitate to contact us.