asc_desc2horz_vert.py is too strict about reference point matching about mintpy HOT 8 CLOSED

The code does keep all the metadata during the loading process, for the sake of completeness, although it uses UPPER_CASE only. "mintpy.*" metadata is saved to check the processing configurations.

from mintpy.

Hi @EJFielding, the reference_point.py bug is fixed now.

from mintpy.

Hi @EJFielding, 0.0001 deg ~= 10 m on the equator, so I think relaxing it to 3 decimal digits should still be fine.

You could try reference_point.py --write-data to change the REF_X/Y/LAT/LON metadata and update the matrix value. MintPy uses UPPER_CASE style for all metadata and is case sensitive.

from mintpy.

Three digits sounds good. Thanks for also mentioning that the metadata is case sensitive and key parameters are UPPER_CASE. It is a bit confusing because info.py prints out a bunch of parameters that are lower case also, including mintpy.reference.lalo and ref_lat. I understand now that the one that matters is the REF_LAT.

I see that you already changed the comparison to 3 decimal digits or ~100 m, so that seems reasonable.

from mintpy.

By the way, I tried to run the reference_point.py --write-data command on my time-series step function file and it did not work.

reference_point.py --write-data -l 18.2 -L -67.15 A135_geo_timeseriesStepModel_msk.h5
Using default MintPy Path: /home/fielding/tools/MintPy
--------------------------------------------------
input reference point in lat/lon: (18.2, -67.15)
input reference point in y/x: (330, 179)
--------------------------------------------------
calculating the temporal average of timeseries file: A135_geo_timeseriesStepModel_msk.h5
reading timeseries data from file: A135_geo_timeseriesStepModel_msk.h5 ...
/home/fielding/tools/MintPy/mintpy/objects/stack.py:464: RuntimeWarning: Mean of empty slice
  dmean = np.nanmean(data, axis=0)
Traceback (most recent call last):
  File "/home/fielding/tools/MintPy/mintpy/reference_point.py", line 447, in <module>
    main(sys.argv[1:])
  File "/home/fielding/tools/MintPy/mintpy/reference_point.py", line 440, in main
    reference_file(inps)
  File "/home/fielding/tools/MintPy/mintpy/reference_point.py", line 186, in reference_file
    if mask[inps.ref_y, inps.ref_x] == 0.:
IndexError: too many indices for array

Maybe I should open a new issue?

Updating the REF_LAT and REF_LON with add_attribute.py works. I understand that it won't update the values of the array for the new location, but I am only changing the reference point by less than a pixel in this case.

from mintpy.

Hello, I am encountering a similar issue.
I am not able assign a single reference point to my data sets in a scripted fashion, as the rounding error is different for the ascending and descending data sets.

When I use refLat=37.07010 and refLon=89.20000 I get the following error message:

Traceback (most recent call last):
  File "/u/sarh0/rzinke/tools/MintPy/mintpy/asc_desc2horz_vert.py", line 278, in <module>
    main(sys.argv[1:])
  File "/u/sarh0/rzinke/tools/MintPy/mintpy/asc_desc2horz_vert.py", line 258, in main
    inps = cmd_line_parse(iargs)
  File "/u/sarh0/rzinke/tools/MintPy/mintpy/asc_desc2horz_vert.py", line 101, in cmd_line_parse
    raise ValueError('input files do not have the same reference point from REF_LAT/LON values!\n{}'.format(msg))
ValueError: input files do not have the same reference point from REF_LAT/LON values!
	file1: 41velocityMaskedRefd.h5, REF_LAT/LON: ['37.071', '89.200'] degrees
	file2: 121velocityMaskedRefd.h5, REF_LAT/LON: ['37.070', '89.200'] degrees

I have tried "jiggling" the numbers in various directions, but always I get a similar same error message, with the ascending reference point different from the descending.

I will try removing this check for now, but might need more attention in the future.

from mintpy.

Thanks for bringing it up @rzinke. The current way of checking is not very friendly to datasets with low resolution, we should change it to check if the two reference points are close in terms of pixels, e.g. <=3 pixels. Let's re-open for a reminder.

from mintpy.

@rzinke It seems that there are often some round-off errors that cause the reference points to be slightly different. Updating the REF_LAT and REF_LON with add_attribute.py works on the files to make them exactly the same. This worked for me.

from mintpy.