This post discusses matchups between ACOLITE processed OLCI data and Bahia Blanca AERONET-OC data, and is a response to poor results published by Arena et al. 2024 (e.g. their Fig. 5 and discussion about ACOLITE/DSF).
AERONET-OC lev20 data (accessed 2024-02-13) were used, and good days for matchups were determined (i.e. those with at least 5 valid lev20 measurements) to obtain Sentinel-3 data from CDSE (accessed 2024-02-13). Satellite data between 2021-09-21 and 2022-12-28 were downloaded and processed with an ACOLITE GitHub clone dated 2024-02-14.
Three configurations of ACOLITE/DSF were used to test sensitivity to settings:
1) Crop, subset to a ~36 km ROI over the Bahia Blanca AERONET site,
2) CropFixed, using the same subset, but using a single aerosol optical depth estimate over the ROI,
3) Full, full scene processing with all the defaults.
Mean and standard deviation were computed for a 3x3 pixel box centred on the site location, taken as -39.148333N, -61.721667E. The closest same-day AERONET measurement was used as reference.
Example spectra:
Lwn scatter plots for bands 3 to 6:
Band Oa03 443 nm
MAPD <9%, RMSD <0.35, MAD <0.24
Band Oa04 490 nm
MAPD <5%, RMSD <0.24, MAD <0.10
Band Oa05 510 nm
MAPD <4%, RMSD <0.24, MAD <0.08
Band Oa06 560 nm
MAPD <4%, RMSD <0.24, MAD <0.06