Hi Patrick
Don't worry about asking questions!
I'm happy ACOLITE is being used, and even happier when users are trying to understand what is going on.
It would be helpful if I could learn a bit more about your project and study site. If you can share some imagery I could have a closer look at some of the potential issues and perhaps run some corrections under development here at the moment.
In response to your questions:
1) If removing the NIR from the DSF does not make a difference, then it was not used to determine the aerosol optical thickness. If you are using a fixed aot estimate, you can check in the L2R file (or output log) which bands were used, there is an ac_band1 variable in the global attributes. Since the DSF did not use the NIR band I expect that the lake is quite impacted by adjacency effects and/or glint.
2) The NIR results are not dissimilar from what I found (e.g.
https://doi.org/10.1364/OE.27.0A1372 and
https://doi.org/10.1364/OE.483418). There could be several reasons for this discrepancy, for example uncorrected glint and adjacency effects, other atmospheric correction issues like gas transmittance correction or underestimated aerosol reflectance, sensor calibration or TOA radiance conversion. Examples of glint and adjacency effects (in S2 imagery) are provided here:
viewtopic.php?f=5&t=310
3) For low signal waters, problems with the atmospheric correction (aerosol correction, gas transmittance correction) and the sensor (noise, calibration) become more problematic as their uncertainty will represent a large fraction of the actual signal. A change in aot, which is mainly what a change of ROI will lead to, will not drastically change the shape of the spectrum, but will largely just lead to a white shift up/down. By the construction of the DSF, it tends to bias the aerosol optical thickness low and the resulting water reflectance high, especially for noisy sensors. In low signal waters this bias can be of the magnitude of the water signal itself. We also can't forget that these sensors were not designed for water applications! Some algorithms (e.g. C2RCC, POLYMER) can get around some of these issues by including a water model, but I don't think they are currently available for Planet data.
4) Well, satellite remote sensing of waters involves many target types and many sensors, as well as many processing algorithms. For your study here, using SuperDove for a high altitude and low signal water body I would say these results are expected. See also the answers to 2 and 3.
For your follow up post, the band exclusion point should be answered by 1) above. The glint correction for PlanetScope needs some additional configuration. There will be no reference to a glint correction being performed in your log file with the current default settings.
I haven't set these as defaults as there are some caveats to using the NIR for glint correction, i.e. the water reflectance has to be negligible, which is not the case for very turbid/productive waters.
You will mainly have to switch bands to use:
Code: Select all
dsf_residual_glint_correction=True
glint_force_band=800
glint_mask_rhos_wave=800
glint_mask_rhos_threshold=0.05
You could also use the alternative glint correction that uses a slightly different spectral shape of the glint:
Code: Select all
dsf_residual_glint_correction=True
dsf_residual_glint_correction_method=alternative
dsf_residual_glint_wave_range=800,900
glint_mask_rhos_threshold=0.05
These are options "1" and "2" from the Appendix B in
https://doi.org/10.1364/OE.483418 Note that there was a small error with variable deletion that was corrected with a commit to the GitHub code this morning.
I hope this helps!
Quinten
