S3 LSR - Sentinel-3 Optical pre-calculated Land Surface Reflectivity Auxiliary Product

About

Earth’s land surface reflectance significantly contributes to top-of-atmosphere measurements in the visible and infrared spectral ranges. Decoupling the atmospheric and surface signals plays a crucial role in successful retrieval of properties of atmospheric constituents such as clouds, aerosols, and trace gases. Therefore, the quality of atmospheric products strongly depends on the accuracy, temporal, spatial, and spectral completeness of the surface reflectance data.

The optical measurements from the instruments onboard Sentinel-3, specifically OLCI and SLSTR, provide crucial information for deriving atmospheric and Earth surface (land and water) parameters in the VIS and NIR spectral ranges. Currently, most OLCI and SLSTR atmospheric processors require a priori knowledge of surface reflectance. Such an auxiliary data product does not yet exist within the dedicated Sentinel-3 operational atmosphere ground segment and associated processing chains.

Objectives

This project is dedicated to the generation of a pre-calculated Sentinel-3 optical Earth Land Surface Reflectivity (LSR) auxiliary product, tailored to the needs of Level 2 algorithms for Sentinel-3 atmospheric processing. The main requirements for the generated LSR product are as follows:

1. Possibility of integration into Level 2 atmospheric composition processors as static auxiliary data files with low-frequency updates.
2. LSR product should be based on statistically significant average values of surface reflectance and include information on the variability around these average values.
3. As much as possible, the product should provide global, gapless spatial coverage.
4. Stability under various atmospheric conditions, including high aerosol loading and cloudy cases.
5. The LSR product should contain different surface characteristics, such as full BRDF, LER, and various albedos characteristics, to be used by different atmospheric processors.
6. The accuracy of the LSR product should meet the requirements of atmospheric processors.
7. Spectral coverage and the number of spectral channels in the LSR product should be sufficient for applications such as aerosol, clouds, total column water vapour, and Aerosol Layer Height (ALH) retrieval.

Overview

Correctly accounting for Earth’s land surface reflectance is crucial for various atmospheric composition retrievals from space-borne measurements. Simultaneously, the accuracy of surface retrieval from space-borne instruments is strongly affected by accurate cloud masking and accounting for atmospheric aerosol properties and gases. Numerous studies have demonstrated that the quality of both atmospheric and surface characterisation can be essentially improved by performing joint simultaneous retrieval of aerosol and surface properties. This methodology has been implemented in the GRASP algorithm (Generalised Retrieval of Atmosphere and Surface Properties).

In this project, a hybrid GRASP retrieval approach is used to derive global LSR products from S3/OLCI and SLSTR. In this approach, aerosol properties are not retrieved but rather re-used from existing aerosol daily datasets retrieved from POLDER/PARASOL and VIIRS measurements over several years (Fig.1 and 2). Additionally, the PARASOL/GRASP surface product is used as a priori information for GRASP full BRDF retrieval from S3/OLCI (and optionally from S3/SLSTR) at 1-2 km spatial resolution (Fig.3). This hybrid approach essentially reduces the number of parameters to be retrieved by re-using available aerosol and surface datasets. At the same time, newly derived products at higher resolution will fully benefit from the completeness and high accuracy of the results obtained by full retrieval at lower resolution. It also allows for starting surface retrieval with initial information close to the sought solution. 

The proposed hybrid full BRDF retrieval approach is expected to be much faster than traditional GRASP simultaneous aerosol/surface retrieval without losing accuracy. It completely satisfies the computational requirements for generating a global LSR auxiliary product with 1-2 km spatial resolution based on a few years of S3 processing. The general processing scheme of the approach is presented in Fig. 4.