Land evapotranspiration (ET) plays a crucial role in the Earth's water carbon cycle, and accurate estimation of global land evapotranspiration is essential for advancing our understanding of land atmosphere interactions. Although many evapotranspiration products have been developed in recent decades, the widely used products still have inherent uncertainties due to the use of different forced inputs and imperfect model parameters. In addition, due to the lack of sufficient global in-situ observation data, it is not realistic to directly evaluate evapotranspiration products, which hinders their utilization and assimilation. Therefore, it is crucial to establish a reliable global benchmark dataset and explore evaluation methods for evapotranspiration products.
The aim of this study is to address these challenges through the following methods: (1) proposing an alignment based approach that considers non-zero error cross-correlation during multi-source data merging; (2) Using this merging method, long-term global daily evapotranspiration products with resolutions of 0.1 ° (2000-2020) and 0.25 ° (1980-2022) were generated and included as inputs for ERA5L, FluxCom, PMLv2, GLDAS, and GLEAM. The resulting product is CAMELE, a multi-source collection of land evapotranspiration data for location analysis.
ERA5-Land： https://cds.climate.copernicus.eu/cdsapp# !/dataset/reanalysis-era5-land? tab=overview
GLDAS： https://disc.gsfc.nasa.gov/datasets?keywords=GLDAS
Global Land Evaporation Amsterdam Model 3.7 (GLEAM-3.7)： https://www.gleam.eu/
Penman–Monteith–Leuning version 2 global evaporation model (PMLv2)： https://developers.google.com/earth-engine/datasets/catalog/CAS_IGSNRR_PML_V2_v017
FluxCom： http://fluxcom.org/
Global in situ observation: FluxNet
The fusion of products in this study involves three steps:
(1) Using registration methods (IVD and EIVD) to calculate the random error variance of the selected input product, determine the optimal product in the region, and set the error threshold;
(2) Calculate the weights of different products on each grid with the goal of minimizing mean square error (MSE);
(3) Based on the weight fusion of products, obtain a long sequence evapotranspiration product. As IVD and EIVD were developed by combining instrumental variable regression and extended localization systems, they also include descriptions of TC and EC algorithms.
CAMELE has shown good performance in various types of vegetation cover and has been validated with field observation data. The Pearson correlation coefficients (R) obtained during the evaluation process were 0.63 and 0.65, respectively. In addition, the comparison results indicate that CAMELE can effectively describe the multi-year linear trend, average value, and extreme value of evapotranspiration. However, it tends to overestimate seasonality. In summary, we have proposed a reliable set of evapotranspiration data that can help understand changes in the water cycle and potentially serve as a benchmark for various applications.
This work is licensed under a
Creative
Commons Attribution 4.0 International License.
CSTR:11738.11.NCDC.ZENODO.DB6472.2024
10.5281/zenodo.8047038