filtering at the aggregation stage???
WARNING: Do not start thinking about this yet, this is an edge case issue that Rich requested. In my opinion we should ignore this issue for now but I I just wanted to write it down.
So Rich wanted the ability to have L3 filtering such that we only use 1 granule per grid cell. So, VIIRS has a lot of overlap especially around the poles. Ideally he would like to only have a single granule contribute to a grid cell. The ideal selection algorithm in his mind would involve determine which granule to pull from by picking the one which has the scan angle closest to nadir.
This would be kind of tricky as it would require filtering at the aggregation stage. There are 3 possible ways I could see implementing this:
1). For each grid cell we could store the mean scan angle and orbit number such that when we aggregate to L3 that we can filter based on the lowest scan angle. The reason we would need orbit number is for adjacent granules which see the same 1 degree grid box. This is do-able but the one downside is it adds filtering to the aggregation stage and it makes the aggregation stage trickier
2). A second option would be to somehow use orbnav to decide which grid cells should be used from which granules. This is probably the best solution
- The final option which would require no action on our part would be to push this onto the ST member. Basically they could create a variable to filter on which uses Latitude and Scan Angle to determine which pixels to include. For instance at Lat=0 we might go out to 52 degrees and at Lat=50 go out to 45 degrees and Lat=70 go out to 20 degrees. This wouldn't be perfect but has the advantage that we push this onto the St member.
I think solution 2 might be the best but we should table this while we finish off release 1 of the gridding