LAND VALidation (LANDVAL) V2: Representative global sampling for satellite product intercomparison and/or calibration The LAND VALidation (LANDVAL) V2 network of sites was generated with a semi-automatic algorithm developed in Python open source code to provide a global sampling representative of global conditions in term of land cover, continental area and ecoregion. The source code needs the following inputs provided by the user: number of samples of the final dataset and the global maps used to classify the global conditions in terms of land cover and ecoregion. LANDVAL V2 was defined with a number of samples of 2000 sites and is based on Copernicus Global Land Cover based on PROBA-V (Buchhorn et al., 2020) and World Wild LiFe Terrestrial Ecoregion of the World classification (Olson et al., 2001). The code also includes the following quality criteria to select the coordinates along the globe: i) same land cover and ecoregion (>90% homogeneity) over an area of 3 km x 3 km; ii) distance between samples at least of 2 degrees on the hypotenuse; and iii) slope within the 3 km x 3 km area lower than 100 m. The atached figues show the percentage of samples per continental region, land cover and ecoregion of the whole globe (red), LANDVAL V2 (blue) and LANDVAL V1 (yellow). LANDVAL V2 provides almost identical distribution than the global case, improving the global representative of LANDVAL V1 used in several validation studies for vegetation LAI/FAPAR or surface albedo products (Fuster et al., 2020; Sánchez-Zapero et al., 2023, 2020). The csv file provided in this dataset contains the 2000 coordinates of LANDVAL V2 along with the following auxiliary information for each pair of latitude and longitude values: continental area, main Land Cover (LC), ecoregion (WWF), homogeneity provided as the percentage of the majority land cover within the 3 km x 3 km area and the slope within the 3 km x 3 km area (i.e., difference in maximum and minimum elevation).