I have 99 LS8 scenes (mountainous terrain) and before processing them any further I would like to mask out cloudy and snow pixel since they might mess up the topographic correction I have to apply. My ultimate goal is a NDVI/EVI time series, so I can treat each scene for itself. I tried the USGS surface product since it contains already a cloud mask but it contains too many artefacts for my study area and also maps out landslides, which are my subject of study. I also tried FMask, providing really bad results.

Does someone have any recommendation how to handle this problem? I mean, in the end it might be quicker to build a separate mask for every scene instead of struggeling but an automatic approach would be ideal (like a DT).

  • 1
    Honestly, the FMask model is about the best game in town and will yield far better results than band thresholding. However, we have found that it is quite sensitive to the "p" parameter and that it is not necessarily a good idea to go with the models defaults. Given that caveat, you can get quite good results, even in areas with high water/moisture content (eg., wetlands). – Jeffrey Evans Mar 7 '18 at 17:57
  • That is true, and I tried the Fmask again with different p-parameters - stil giving me bad results. I ended with thresholding now. Pain in the ass, but the masks are sufficient. Thank you anyway. :) – Aldi Kasse 2 Mar 9 '18 at 14:59
  • Did you use BQA or pixel_qa? The easiest approach – aldo_tapia Mar 16 '18 at 19:53
  • @aldo_tapia Yes I ended up using the QA-band and got pretty good results. Thanks! – Aldi Kasse 2 Mar 19 '18 at 13:46

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.