BMA-ET: A new global terrestrial evapotranspiration dataset from multi-datasets fusion based on Bayesian model averaging covering 1980-2020

1. Dataset information Dataset name: BMA-ET Summary: A long-term (1980-2020) global terrestrial evapotranspiration (ET) product with a spatial resolution of 0.5 degree and 1 degree. This is a merged product from thirty ET products using the Bayesian model averaging (BMA) method. These include four main types: remote sensing–based, machine learning–based, reanalysis-based, and land-surface-model-based. 2. Content of the data set The dataset includes 3 files organized as follows: (1) Metadata for BMA-ET.docx: This document provides detailed information about the dataset. (2) ET_BMA_1980_2020_05degree.nc: The file containing global terrestrial evapotranspiration dataset at 0.5 degree resolution in NetCDF format. (3) ET_BMA_1980_2020_1degree.nc: The file containing global terrestrial evapotranspiration dataset at 1 degree resolution in NetCDF format. 3. Brief calculation introduction Evapotranspiration fusion from thirty sets of ET datasets (Figure 1) was completed using a BMA method. The 30 sets of ET datasets are clustered and then fused. First, we calculated a Pearson correlation coefficient matrix using the residuals from the 30 sets of ET datasets with observations from FLUXNET2015 sites. Second, the 30 sets of ET datasets were clustered based on the residual correlation coefficient matrix. Third, for each vegetation type, the BMA fusion of the ET data within each cluster was performed first, and then the BMA fusion of the fused data for each cluster was performed. Figure 1. Years of coverage for each evapotranspiration dataset, with 1982–2011 being the common period of coverage for all evapotranspiration datasets. Dataset types are labelled as follows, by group: RS = remote sensing; ML = machine learning; RA = reanalysis; LSM = land surface model. 4. Details of the variables in the file Three dimensions are included in the file (e.g., ET_BMA_1980_2020_1degree.nc), and they are longitude (lon), latitude (lat), time (in month). The length of these dimensions was listed as follows: Dimensions: lon= 360 ; lat= 180 ; time= 492 ; Data variable: Size: 360*180*492 Dimensions: lon, lat, time Datatype: double Attributes: longname = ‘evapotranspiration’ units = ‘mm/mon’ _FillValue = NaN missing_value = NaN Using Matlab as an example, the method for reading data is as follows: nc = ncread(‘ET_BMA_1980_2020_1degree.nc’,’ET’); 5. Authors and contacts Authors: Yi Wu ([e-mail]) Chiyuan Miao ([e-mail])