We need to download cloudless Sentinel-2 data for a rather large number of regions of interest (ROIs), which are provided in the form of geojson polygons. Unfortunately, relying on the cloud cover metainfo tag will not help us, as the clouds can appear in a corner of the granule, which does not intersect with our ROI - we might up rejecting images that would be good for our ROI or keeping images that are good outside, but bad inside the ROI.

Thus, I thought, it could be a good idea to use Google Earth Engine (GEE) to search for granules, which are not affected in our ROI, exploiting the cloud band. Then, I would export a list of image IDs for subsequent downloading outside GEE.

In the end, the data is subject to a land cover classification task, and we want to have one (mosaicked, if necessary) image per season per ROI.

Any idea how I could get this list of image IDs for a bunch of ROIs? I am not looking for a full-fledged processing chain but rather for hints towards how to formulate the data query. The challenge is the required ROI-based point of view in contrast to the many answers with a pixel-based or granule/image-based point of view...

| improve this question | | | | |
  • You could use .reduceRegions to get the mean cloud cover for each Image and ROI as a table using the quality band of Sentinel-2. The table could then be exported for sorting by your cloud cover thresholdand get the download IDs. – Kersten Feb 1 '18 at 8:38

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.