PyAPS ndarray is not C-contiguous error with scipy.interpolate

[bbuzz31]

I'm getting the following error when running MintPy with ERA5 pyAPS corrections:

Traceback (most recent call last):
  File "/Users/buzzanga/Miniconda3/envs/MintPy/bin/smallbaselineApp.py", line 8, in <module>
    sys.exit(main())
             ^^^^^^
  File "/Users/buzzanga/Software_InSAR/MintPy_git/src/mintpy/cli/smallbaselineApp.py", line 208, in main
    run_smallbaselineApp(inps)
  File "/Users/buzzanga/Software_InSAR/MintPy_git/src/mintpy/smallbaselineApp.py", line 1117, in run_smallbaselineApp
    app.run(steps=inps.runSteps)
  File "/Users/buzzanga/Software_InSAR/MintPy_git/src/mintpy/smallbaselineApp.py", line 898, in run
    self.run_tropospheric_delay_correction(sname)
  File "/Users/buzzanga/Software_InSAR/MintPy_git/src/mintpy/smallbaselineApp.py", line 648, in run_tropospheric_delay_correction
    mintpy.cli.tropo_pyaps3.main(iargs)
  File "/Users/buzzanga/Software_InSAR/MintPy_git/src/mintpy/cli/tropo_pyaps3.py", line 166, in main
    run_tropo_pyaps3(inps)
  File "/Users/buzzanga/Software_InSAR/MintPy_git/src/mintpy/tropo_pyaps3.py", line 716, in run_tropo_pyaps3
    calc_delay_timeseries(inps)
  File "/Users/buzzanga/Software_InSAR/MintPy_git/src/mintpy/tropo_pyaps3.py", line 610, in calc_delay_timeseries
    tropo_data = get_delay(
                 ^^^^^^^^^^
  File "/Users/buzzanga/Software_InSAR/MintPy_git/src/mintpy/tropo_pyaps3.py", line 488, in get_delay
    aps_obj.getdelay(pha)
  File "/Users/buzzanga/Miniconda3/envs/MintPy/lib/python3.11/site-packages/pyaps3/objects.py", line 310, in getdelay
    val = linearint(np.vstack((lati, loni, D)).T)*np.pi*4.0/(cinc*wvl)
          ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/buzzanga/Miniconda3/envs/MintPy/lib/python3.11/site-packages/scipy/interpolate/_rgi.py", line 331, in __call__
    indices, norm_distances = self._find_indices(xi.T)
                              ^^^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/buzzanga/Miniconda3/envs/MintPy/lib/python3.11/site-packages/scipy/interpolate/_rgi.py", line 499, in _find_indices
    return find_indices(self.grid, xi)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "_rgi_cython.pyx", line 95, in scipy.interpolate._rgi_cython.find_indices
  File "stringsource", line 660, in View.MemoryView.memoryview_cwrapper
  File "stringsource", line 350, in View.MemoryView.memoryview.__cinit__
ValueError: ndarray is not C-contiguous

If i run MintPy again, with the same config file, the error is not raised, and timeseries_ERA5.h5 is written
However, each date slice in the inputs/ERA5.h5 file is all 0s, and the resulting timeseries_ERA5.h5 is - as expected - identical to timeseries.h5. This happens with the the latest conda release and the last comment on main.

Please use this PR for testing with ARIA tools

ariaDownload.py -t 71 --bbox '32.9798 34.9440 -119.2723 -116.2523' -s 20200112 -e 20200130 -w ./products
ariaTSsetup.py -f './products/S1*.nc' -l 'unwrappedPhase' --bbox '32.9798 34.9440 -119.2723 -116.2523' -d ./DEM/SRTM_3arcsec.dem -w ./TEST5 -nt 1

mkdir TEST5/MP
cd TEST5/MP
smallbaselineApp.py TEST.cfg

and here's the config file (TEST.cfg), make sure you put it in TEST5/MP

mintpy.load.processor      = aria
mintpy.load.updateMode     = no
mintpy.load.unwFile        = ../stack/unwrapStack.vrt
mintpy.load.corFile        = ../stack/cohStack.vrt  
mintpy.load.connCompFile   = ../stack/connCompStack.vrt 

mintpy.load.demFile        = ../DEM/SRTM_3arcsec.dem 
mintpy.load.incAngleFile   = ../incidenceAngle/*.vrt 
mintpy.load.azAngleFile    = ../azimuthAngle/*.vrt  

mintpy.networkInversion.weightFunc = no
mintpy.network.coherenceBased  = no

mintpy.solidEarthTides = no #[yes / no], auto for no
mintpy.troposphericDelay.method       = pyaps #[pyaps / gacos / no]
mintpy.troposphericDelay.weatherModel = ERA5 #[ERA5 / MERRA / NARR], auto for ERA5a
mintpy.troposphericDelay.weatherDir   = auto

mintpy.topographicResidual                   = no

mintpy.reference.date = no # first date


########## 10. velocity
mintpy.timeFunc.polynomial = 1   #[int >= 0], auto for 1, degree of the polynomial function
mintpy.timeFunc.periodic   = auto   #[1,0.5 / list_of_float / no], auto for no, periods in decimal years

mintpy.timeFunc.startDate   = auto   #[20070101 / no], auto for no
mintpy.timeFunc.endDate     = auto   #[20101230 / no], auto for no
mintpy.timeFunc.excludeDate = no     # auto for exclude_date.tx

mintpy.save.kmz             = no
mintpy.plot                 = no   

[yunjunz]

Hi @bbuzz31, this is a recent bug introduced in scipy 1.10 in its interpolate sub-module. Since pyaps and mintpy use this sub-module, the circle CI testings have failed too, with the same error you have. This bug has been fixed and merged in scipy/scipy#17717, we will have it automatically after the next patch release of scipy. So let's just wait.

In the meanwhile, put a scipy<1.10 version constraint while installing the dependencies of your conda environment.

Let's keep this issue open so that other users are aware of this as well.

[bbuzz31]

aha thanks @yunjunz ! Sorry I missed that in their issues.

Naturally, we just forced scipy >= 1.10 in AT and RAiDER so that we can give descending values to the scipy interpolator... but luckily I think this latter functionality exists in 1.9.0 also