Ground-based measurements for validation of L2A-products

Abstract

Multispectral Copernicus Sentinel-2 and hyperspectral DESIS and EnMAP missions serve for observation of land-cover change and deriving biophysical variables related to agriculture and forestry, monitor coastal and inland waters and are useful for risk and disaster mapping. Level-2A (L2A) data contain Bottom-of-Atmosphere (BOA) surface reflectance (SDR) products together with Aerosol Optical Thickness (AOT) and Integrated Water Vapour (WV) maps. A pre-condition for validation of SDR, AOT and WV data are in-situ measurements collocated spatially and in time with Satellite overpasses over the test site. AERONET provides such data for validation of AOT and WV retrievals for a large number of locations around the globe. Similarly, for the validation of SDR, permanently operating stations with standardized equipment are required. Such sites like the established RadCalNet sites have the advantage to provide data for each cloud screened satellite overpass. However, sites with permanent operating instruments measuring SDR are focusing on few homogeneous sites with typically low AOT, which are perfectly suitable for radiometric Cal/Val. They don’t offer the diversity of land surface types and aerosol conditions necessary for validation of L2A-products. Therefore, this validation data source has to be supplemented by in-situ measured SDR data from ad-hoc campaigns, as identified in projects such as Copernicus Cal/Val Solutions (CCVS). Ad-hoc campaigns are limited to quite short time range but provide reference data for substantially more surface types. DLR has a long-time experience in performing ground-based reference measurement campaigns for validation of multispectral and hyperspectral L2A products of, for example, Sentinel-2, Landsat, Rapid-Eye, DESIS and EnMAP data, as well as Cal/Val for airborne hyperspectral sensors. This covers all of SDR, AOT and WV retrieval. In this presentation we explain our measurement protocol applied for measuring SDR, AOT and WV on ground which was refined over years following own experience and recommendations from publications about best practices. It is necessary to know the instruments and sites used and to understand the measurement process and uncertainties involved. It is necessary to include into the measurement protocol the upscaling approach from the small footprint of instruments used near the ground to the size of satellite pixels. The completion of the validation procedure includes to convolve spectrally the ground reference measurements to the satellite-sensor-bands, to consider directional effects of surface reflection on the ground and careful geolocation of measurement positions in satellite image with respect to the selection of image pixels for spatial averaging. Application of the complete procedure is demonstrated on example of Sentinel-2 L2A-products, EnMAP data are used to demonstrate progress in accounting for BRDF-effects and open questions are discussed.