Comparing SEN2COR and FORCE BOA

[JonasardoLundUniv]

We compared reflectances (BOA) for Sentinel-2 (band 2,3,4,8) from SEN2COR (S2B_MSIL2A_20240905T102559_N0511_R108_T33UUB_20240905T163108.SAFE) and FORCE (S2B_MSIL1C_20240905T102559_N0511_R108_T33UUB_20240905T154834.SAFE) as shown below. Only pixles with highets FORCE quality flag (0) included.
A systematic bias for the visual bands and some spread in NIR, even if R2 are fairly high. In 2018, an Atmospheric Correction Inter-Comparison Exercise including both SEN2COR and FORCE was published https://www.mdpi.com/2072-4292/10/2/352 . I assume there been some changes to both SEN2COR and FORCE since then. When computing NDVI for example, one adds these biases in the denominator reducing comparability.

Anybody who looked into such differences in more detail?

Regards
/Jonas Ardö, Lund University

[davidfrantz]

Hi @JonasardoLundUniv,

as far as I know, the ESA L2 product adds an offset of 1000 for scaling purposes. I think you find this offset in their metadata.

Cheers,
David

[JonasardoLundUniv]

Thanks David, it is correct, I missed that (<RADIO_ADD_OFFSET band_id="1">-1000</RADIO_ADD_OFFSET> in MTD_MSIL1C.xml). Looks better now, but with some spread in NIR.
/Jonas Ardö, Lund University

[kelewinska]

Dear @JonasardoLundUniv,
In both, FORCE and SEN2CORE, one can run the TOA to BOA processing with different parameters, enable the atmospheric, topographic, and BRDF corrections. Furthermore, if i remember correctly, SEN2CORE offers (or used to offer) different models for the atmospheric correction.
From your description does not transpire how you parametrized either processor, hance it is impossible to tell what drives the differences you pointed out, nor it is possible to recreate your runs.

All the very best,

[JonasardoLundUniv]

Dear @kelewinska

You are right. Sorry for not providing full information. This was quite a quick & dirty test.
SEN2COR version: (S2B_MSIL2A_20240905T102559_N0511_R108_T33UUB_20240905T163108.SAFE) was downloaded from https://dataspace.copernicus.eu/ so their default SEN2COR settings were used.

Regards
/Jonas

FORCE PRM file below:

++PARAM_LEVEL2_START++

INPUT/OUTPUT DIRECTORIES

------------------------------------------------------------------------

FILE_QUEUE = /projects/eko/fs6/pers/JA/2025/FORCE/33UUB_list_260213.txt
DIR_LEVEL2 = /projects/eko/fs6/pers/JA/2025/FORCE/33UUB_LEVEL2/
DIR_LOG = /projects/eko/fs6/pers/JA/2025/FORCE/33UUB_LEVEL2/log
DIR_PROVENANCE = /projects/eko/fs6/pers/JA/2025/FORCE/33UUB_LEVEL2/provenance
DIR_TEMP = /projects/eko/fs6/pers/JA/2025/tmp

MASKING

------------------------------------------------------------------------

FILE_AOI = NULL

JA: DEM data from GET in 2 meter reolution have the missing value = NoData Value=-9999

FILE_DEM = /projects/eko/fs6/pers/JA/2025/FORCE/33UUB_DEM.vrt
DEM_RESAMPLING = BL
USE_DEM_DATABASE = FALSE
DEM_NODATA = -9999

DATA CUBES

------------------------------------------------------------------------

DO_REPROJ = FALSE # TRUE
DO_TILE = FALSE # TRUE
FILE_TILE = NULL # /projects/eko/fs6/pers/JA/2025/FORCE/Alnarp_tiles.txt
TILE_SIZE = 90000 90000
RESOLUTION_LANDSAT = 30
RESOLUTION_SENTINEL2 = 10
ORIGIN_LON = 7.9944820999523601
ORIGIN_LAT = 69.7058069302281122
PROJECTION = PROJCS["SWEREF99 TM",GEOGCS["SWEREF99",DATUM["SWEREF99",SPHEROID["GRS 1980",6378137,298.257222101,AUTHORITY["EPSG","7019"]],TOWGS84[0,0,0,0,0,0,0],AUTHORITY["EPSG","6619"]],PRIMEM["Greenwich",0, AUTHORITY["EPSG","8901"]],UNIT["degree",0.0174532925199433,AUTHORITY["EPSG","9122"]],AUTHORITY["EPSG","4619"]],PROJECTION["Transverse_Mercator"],PARAMETER["latitude_of_origin",0],PARAMETER["central_meridian",15],PARAMETER["scale_factor",0.9996],PARAMETER["false_easting",500000],PARAMETER["false_northing",0],UNIT["metre",1,AUTHORITY["EPSG","9001"]],AUTHORITY["EPSG","3006"]]
RESAMPLING = NN

RADIOMETRIC CORRECTION OPTIONS

------------------------------------------------------------------------

DO_ATMO = TRUE
DO_TOPO = TRUE
DO_BRDF = TRUE
ADJACENCY_EFFECT = TRUE
MULTI_SCATTERING = TRUE

WATER VAPOR CORRECTION OPTIONS

------------------------------------------------------------------------

DIR_WVPLUT = NULL
STRICT_WATER_VAPOR = FALSE
WATER_VAPOR = NULL

AEROSOL OPTICAL DEPTH OPTIONS

DO_AOD = TRUE
DIR_AOD = NULL

CLOUD DETECTION OPTIONS

------------------------------------------------------------------------

ERASE_CLOUDS = FALSE
MAX_CLOUD_COVER_FRAME = 99
MAX_CLOUD_COVER_TILE = 99
CLOUD_BUFFER = 300
CIRRUS_BUFFER = 0
SHADOW_BUFFER = 90
SNOW_BUFFER = 30
CLOUD_THRESHOLD = 0.225
SHADOW_THRESHOLD = 0.02

RESOLUTION MERGING

------------------------------------------------------------------------

Type: Character. Valid values: {IMPROPHE,REGRESSION,STARFM,NONE}

RES_MERGE = IMPROPHE

CO-REGISTRATION OPTIONS

------------------------------------------------------------------------

DIR_COREG_BASE = NULL
COREG_BASE_NODATA = -9999

MISCELLANEOUS OPTIONS

------------------------------------------------------------------------

IMPULSE_NOISE = TRUE
BUFFER_NODATA = FALSE

TIER LEVEL

------------------------------------------------------------------------

TIER = 1

PARALLEL PROCESSING

------------------------------------------------------------------------

NPROC = 30
NTHREAD = 1
PARALLEL_READS = TRUE
DELAY = 5
TIMEOUT_ZIP = 30

OUTPUT OPTIONS

------------------------------------------------------------------------

OUTPUT_FORMAT = COG
FILE_OUTPUT_OPTIONS = NULL
OUTPUT_DST = TRUE
OUTPUT_AOD = FALSE
OUTPUT_WVP = FALSE
OUTPUT_VZN = TRUE
OUTPUT_HOT = TRUE
OUTPUT_OVV = TRUE

++PARAM_LEVEL2_END++