European eVALuation sites (EVAL): A European sampling for regional satellite product intercomparison

For the quality assessment of satellite-based products it is necessary to perform product intercomparisons at global and continental scales. In the context of the Copernicus Land Monitoring Service High-Resolution Vegetation and Productivity Parameters (HR-VPP) project, a network of European eVALuation (EVAL) sites was defined to perform product intercomparisons over a representative sampling in terms of land cover and geographical area. EVAL is thus a spatial sampling of 3800 evaluation sites over the 39 countries of the European Economic Area (EEA39) [1]. The main selection criterium is the homogeneity, in terms of land cover, over areas of 1 km x 1 km (more than 70% of the area corresponds to the same land cover class) to be useful for the evaluation of decametric and hectometric (300 – 500 m) resolution products. 

The selection was based on surface albedo validation sites (SAVS1.0) [2], and European long-term ecosystem research (eLTER) sites [3]. Those sites were first screened to be homogeneous in land cover type, and then visually inspected using Google Earth to select homogeneous sites. Initially, only 597 sites (158 SAVS1.0 and 439 eLTER) were selected. Then, the network was complemented with Land use and land cover survey (LUCAS) samples [4] to get a better representativeness of European land cover types and geographical distribution.  10 additional LUCAS sites were selected for each 1 x 1 degree according to the same homogeneity criteria.  Thereafter, random rejection of sites was conducted to keep the distribution of EVAL sites similar to that of the EEA-39 area (see distribution in Figure_1). In the last step, under sampled regions (e.g., Turkey or Iceland) were filled through visually identification of additional homogeneous sites.

A total of 3800 sites were finally selected. The CORINE Land Cover 2018 (CLC2018) [5] was used to assign the land cover class to each site to allow the analysis per cover type. An aggregation of the CLC land cover classes was performed to reduce the number of classes to 8 main classes: Broad-leaved forest (BLF), Needle-leaf forest (NLF), Mixed Forest (MF), Crops, Grass, Shrubs, Agroforestry and Sparse. Figure_2 shows the distribution of the 3800 EVAL sites per aggregated land cover type as compared to the actual distribution over the whole EEA-39. 

 The xlsx file provided in this dataset contains two spreadsheets with the following information:

‘EVAL_sites’: ‘Site label’, ‘Latitude’, ‘Longitude’, ‘CLC 2018’,’ Aggregated Land Cover’, and ‘Environmental zone’ for each location. 

 ‘Legend’: information about the legends used in the dataset (CLC2018, aggregated land cover and environmental zone) and the number of EVAL samples for each class.

References:

[1] https://sdi.eea.europa.eu/catalogue/srv/api/records/8526ff78-b000-42e1-8360-a2fb3a51e4ac

[2] Loew, A., Bennartz, R., Fell, F., Lattanzio, A., Doutriaux-Boucher, M., Schulz, J., 2016. A database of global reference sites to support validation of satellite surface albedo datasets (SAVS 1.0). Earth Syst. Sci. Data 8, 425–438. https://doi.org/10.5194/essd-8-425-2016

[3] https://www.lter-europe.net/

[4] https://ec.europa.eu/eurostat/statistics-explained/index.php?title=LUCAS_-_Land_use_and_land_cover_survey

[5] https://doi.org/10.2909/960998c1-1870-4e82-8051-6485205ebbac