ACOLITE forum

Hello Quinten,
We are trying to find a way to solve the problem of atmospheric correction in both clear water and turbid water.
As we know, the DSF mode in ACOLITE has achieved good results in turbid water. So we want to know whether the EXP mode can achieve the same good effect in clean water (because near-infrared extrapolation is Implemented)?
Looking forward to your reply.

The DSF tends to give a positive bias in the surface level reflectance outputs; as a result of an underestimation of the aerosol optical depth by selecting the darkest pixels/targets in a subscene. For turbid waters this is usually fine, but for clear waters this bias can be of the same magnitude as the water signal. I have been trying some methods to improve this, but at rather low priority since there are other processors that perform quite well over clear waters (e.g. https://doi.org/10.1016/j.rse.2021.112366), and the main intended use of ACOLITE is turbid waters. In the Copernicus Marine Service the High-Resolution Ocean Colour (HR-OC) products are produced by blending C2RCC for clear waters and ACOLITE/DSF for more turbid waters (see e.g. https://doi.org/10.48670/moi-00119).

The EXP method is rather limited, and by default for Landsat/Sentinel-2 it uses the SWIR bands for the atmospheric correction (https://doi.org/10.1016/j.rse.2015.02.007). It has some issues with glint, thin clouds, adjacency effects as it assumes the full SWIR signal is from aerosols. The exponential extrapolation also gives poor results at the blue end of the spectrum (https://doi.org/10.1016/j.rse.2019.03.010). In fact in that paper you can see for rather clear waters (from AERONET-OC) the DSF tends to outperform the EXP. Perhaps you are better off using the NIR-SWIR option as implemented in SeaDAS/l2gen or use a modeling approach such as POLYMER/C2RCC.

I hope this helps!

Quinten