A benchmark dataset for half-hourly evapotranspiration estimation in China from 2000 to 2024

(1) Half-hourly LE data (2000–2024).This dataset provides seamless half-hourly latent heat flux (LE) data for flux tower sites across China, generated using the AutoML-H2O framework. The half-hourly LE data represent the core product of this dataset and cover the period from 1 January 2000, 00:00 to 31 December 2024, 23:30.Each record is associated with a quality flag indicating the data source: T denotes half-hourly LE derived from original flux tower observations; F represents gap-filled LE within the observation period; and P indicates LE values generated through temporal prolongation beyond the observation period.All timestamps are recorded in local site time and follow the format YYYYMMDD–HHMM (e.g., 20120101–1530). The data provide complete, temporally continuous half-hourly LE time series for all sites. (2) Daily, monthly, and annual aggregated LE data.Daily, monthly, and annual LE products were directly aggregated from the gap-filled and temporally prolonged half-hourly LE data, ensuring full temporal consistency across all time scales. These aggregated datasets cover the period 2000–2024.For each temporal scale (day, month, and year), quality control information is provided in the form of the proportion of valid half-hourly observations contributing to the aggregated LE value within the corresponding period. This unified definition of data availability allows users to consistently assess data reliability across temporal scales. No distinction is made between gap-filled and prolonged data at aggregated scales, with the aim of providing stable and continuous datasets for long-term hydrological and climate change studies. (3) Auxiliary driving data.To support reproducibility and extended analyses, auxiliary driving variables used for LE gap-filling and temporal prolongation are also provided. These include hourly meteorological and energy balance variables from ERA5-Land, as well as NDVI and LAI derived from MODIS products. All auxiliary datasets have been spatially and temporally matched to the LE data at the site scale. NDVI and LAI data are available from 18 February 2000, 08:00 (local time) onward; for earlier periods, vegetation indices are not provided. In addition, the code used for LE gap-filling and temporal prolongation is publicly available and provided alongside the dataset, enabling full reproducibility of the data processing workflow and facilitating reuse for similar applications. All data files are released in CSV format, with file names clearly indicating site information, temporal scale, and variable type.