Wishlist for disp-xr (and possibly opera-utils)

[scottstanie]

Let's track our possibly additions and improvements here. Some may be general enough to fall into opera-utils; those that seem further along the analysis pipeline should probably go here.

[bbuzzanga]

Here are a few notes/suggestions that occurred to me as I was estimating velocities on a frame

opera_utils.run_s1_frame.py

• there should be some metadata that indicates what corrections were/are applied (eg in the velocity tiff)

  • metadata for all layers with a description and units (including masking values)
  • there should be metadata showing reference point used (eg in the velocity tiff)
  • would be nice to match the metadata in the OPERA layers (not propagated to the tiffs)

• If correction layers are being applied they should be written to disk

• there should be a way to skip hitting the url in order to get needed products, in case the server is down or something

  • Perhaps give a flag to specify download or not and a path to existing disp ncs

• Average recommended Mask should be fixed to show an actual average, or percentage that represents the time the pixel is valid

  • Currently the max shows a 1 at all pixels and a NaN at pixels in the watermask
  • N.B. I specified watermask to be used

• Average_ps_mask seems wrong, has a range of values (e.g., 1-15) and includes 255

• Water mask does not get applied to:

  • average temporal coherence
  • average shp counts
  • average phase similarity
  • average timeseries residuals
  • (it does get applied to the velocity field)
  • Phase similarity Values aren't 0 but ..1e-9 (1e-10 for water pixels), expected?

• Include some sort of timestamp in displacement_* tifs

[scottstanie]

A couple initial questions (agreed on most of what you wrote):

Include some sort of timestamp in displacement_* tifs
Do you mean a generation time? or the reference/secondary as full datetimes, instead of YYYYmmdd?

would be nice to match the metadata in the OPERA layers (not propagated to the tiffs)
Do you have other metadata you're interested in? From the CEOS work, there are >50 items in the product.

The metadata item is also hard to fully propagate, as xarray doesn't seem to read all of it in. We should keep track of which essential items are missing though.

But I also thing that the problem could be partially avoided by switching to either the completely lazy Marin approach, or by using the Virtual Dataset feature of HDF5 to avoid needing the full reformat after copying.