Sentinel-2 Level-2A Prototype Processor: Algorithms, Architecture and First Results
Mueller-Wilm, Uwe1; Louis, Jerome2; Richter, Rudolf3; Gascon, Ferran4; Niezette, Marc1
1Telespazio VEGA, GERMANY; 2Telespazio, FRANCE; 3DLR, GERMANY; 4ESA / ESRIN, ITALY
A prototype processor implementation for the Sentinel-2 Level-2A product generation and formatting is presented. Level-2A processing includes a Scene Classification and an Atmospheric Correction applied to Top-Of-Atmosphere (TOA) Level-1C ortho-image products. Level-2A main output is an ortho-image Bottom-Of-Atmosphere (BOA) corrected reflectance product.
Additional outputs are an Aerosol Optical Thickness (AOT) map, a Water Vapor (WV) map and a Scene Classification map (SCM) together with Quality Indicators for cloud and snow probabilities at 60 m resolution. The input Level-1C TOA products include 13 JPEG-2000 images associated to the 13 Sentinel-2 spectral bands at three different spatial resolutions with a GSD (Ground Sampling Distance) of 10, 20, and 60m. Level-2A output image products will in contrast be resampled and generated with an equal spatial resolution for all bands, based on the requested resolution on user's choice (10m, 20m or 60m). A 10 m resolution product contains the spectral bands 2, 3, 4 and 8 and an AOT map resampled from 20m. A 20 m product contains band 2 - 7, the bands 8A, 11 and 12 and an AOT and WV map. A 60m product contains all components of the 20m product and additionally the 60m bands 1 and 9. The cirrus band 10 will be omitted, as it does not contain surface information.
The processor algorithm is a combination of state-of-the-art techniques for performing Atmospheric Corrections (including Cirrus clouds correction) [R1], which have been tailored to the Sentinel-2 environment together with a Scene Classification module described in [R2].
The Scene Classification algorithm allows to detect clouds, snow and cloud shadows and to generate a classification map, which consists of 4 different classes for clouds (including cirrus), together with six different classifications for shadows, cloud shadows, vegetation, soils / deserts, water and snow. The algorithm is based on a series of threshold tests that use as input top-of-atmosphere reflectance from the Sentinel-2 spectral bands. In addition, thresholds are applied on band ratios and indexes like NDVI and NDSI. For each of these thresholds tests, a level of confidence is associated; it produces at the end of the processing chain a probabilistic cloud mask quality indicator and a snow mask quality indicator. The algorithm uses the reflective properties of scene features to establish the presence or absence of clouds in a scene. Cloud screening is applied to the data in order to retrieve accurate atmospheric and surface parameters, either as input for the further processing steps below or for being valuable input for processing steps of higher levels.
The aerosol type and visibility or optical thickness of the atmosphere is derived using the DDV (Dense Dark Vegetation) algorithm [R3]. This algorithm requires that the scene contains reference areas of known reflectance behaviour, preferably DDV and water bodies. The algorithm starts with a user-defined visibility (default: 20 km). If the scene contains no dark vegetation or soil pixels, the surface reflectance threshold in the 2190 nm band is successively iterated in order to include medium brightness reference pixels. If the scene contains no reference and no water pixels the scene is processed with the start visibility instead.
The water vapour retrieval over land is performed with the APDA (Atmospheric Precorrected Differential Absorption) algorithm [R4] which is applied to the two Sentinel-2 bands (B8a, and B9). Band 8a is the reference channel in an atmospheric window region, band B9 is the measurement channel in the absorption region. The absorption depth is evaluated in the way that the radiance is calculated for an atmosphere with no water vapour assuming that the surface reflectance for the measurement channel is the same as for the reference channel. The absorption depth is then a measure of the water vapour column content.
The Atmospheric correction is performed using a set of Look-up tables generated via libRadtran (www.libradtran.org). Baseline processing is the rural/continental aerosol type; other Look-Up tables can also be used according to the scene geographic location and climatology.
In order to assess the Sentinel-2 Level-2A prototype processor performance, simulated Sentinel-2 Level-1C test datasets have been generated, derived from Hyperion hyperspectral instrument flown on board EO-1 and from simulated Level-1C furnished by ESA. The computational performance and output product format validity is assessed to demonstrate the capabilities of the processor. Some comparisons are drawn with external data (Aerosol Optical Thickness, Water Vapor, manual cloud masks, etc.) and alternative cloud detection and atmospheric correction techniques to evaluate and to compare the results of the Level-2A processing.
Figure 1 below shows (from left to right (1) a simulated Level-1C TOA input Image (2) the Cirrus and Atmospheric corrected level-2A BOA Image, the Cirrus Band Nr. 10 and (4) the Scene classification of the Level-1C input image. The size of the images does currently not represent the final Sentinel-2 image resolutions, whereas the band criteria are representative.
[R1]: Richter, R., Wang, X., Bachmann, M., and Schlaepfer, D., "Correction of cirrus effects in Sentinel-2 type of imagery", Int. J. Remote Sensing, Vol.32, 2931-2941 (2011).
[R2]: J. Louis, A. Charantonis & B. Berthelot, "Cloud Detection for Sentinel-2", Proceedings of ESA Living Planet Symposium (2010).
[R3]: Kaufman, Y., Sendra, C. Algorithm for automatic atmospheric corrections to visible and near-IR satellite imagery, International Journal of Remote Sensing, Volume 9, Issue 8, 1357-1381 (1988).
[R4]: Schläpfer, D. et al., "Atmospheric precorrected differential absorption technique to retrieve columnar water vapour", Remote Sens. Environ., Vol. 65, 353366 (1998).