download MODIS (crs: SR-ORG:6974) data appears obscured

[cug-xyx]

Dear Developers:

Thank you so much for being able to develop the awesome xee.

I was downloading the MODIS series product (MODIS/061/MOD16A2GF) and noticed that the downloaded data was shaded.

I think it's a problem with the crs.

Here are the images and code:

ic_name = 'MODIS/061/MOD16A2GF'

ic = ee.ImageCollection(ic_name) \
  .select("ET") \
  .filterDate('2013-06-02', '2013-06-03')

ds = xarray.open_dataset(
    ic, engine='ee',
    crs = 'EPSG:4326', 
    scale=0.25
)
ds

ds.to_netcdf('test.nc')

r = xr.open_dataarray('test.nc')
r

r \
    .assign_attrs(crs='EPSG:4326') \
    .transpose() \
    .plot(
    vmin=0, vmax=300,
    levels=[i for i in range(0, 270, 30)],
    colors=['#ffffff', '#fcd163', '#99b718', '#66a000',
            '#3e8601', '#207401', '#056201', '#004c00', '#011301']
)

[simonff]

What specifically do you mean by "shaded"?

[cug-xyx]

I'm sorry. Maybe I didn't describe it properly.

This image is ok on GEE.

var dataset = ee.ImageCollection('MODIS/NTSG/MOD16A2/105');
var ET = dataset.select('ET').filterDate('2013-06-02', '2013-06-03');
var ETVis = {
  min: 0,
  max: 300,
  palette:
      ['#ffffff', '#fcd163', '#99b718', '#66a000',
            '#3e8601', '#207401', '#056201', '#004c00', '#011301'],
};
Map.addLayer(ET, ETVis, 'Evapotranspiration');

But exporting locally using xee results in something like the following:

[simonff]

Ah, you mean the weird yellow area over Americas and Antarctica. I agree this is wrong.

[cug-xyx]

Indeed. But xee is so good that I just can't live without it, so I'm looking for a solution here

[simonff]

Let's see what the actual developers say - I'm just another xee user

[scheip]

Have you tried this with different times or different bands and does it yield the same result? I am also having a problem with bringing MODIS data across via xee, though a bit different to what you are showing.

[cug-xyx]

Have you tried this with different times or different bands and does it yield the same result? I am also having a problem with bringing MODIS data across via xee, though a bit different to what you are showing.

@scheip, I've tried different periods and it's the same problem :(

[noahgolmant]

What I notice about this dataset is that each band has a different scale, which is the same case as #122. There is a chunk of code in get_info where xee infers a number of collection-wide attributes like the primary coordinate. Just spitballing, but maybe that has something to do with it? Like that data differs between bands with different scales?

[tylere]

I'm attempting to create a reproducible example of this issue, but I am encountering an error when converting to NetCDF

https://colab.research.google.com/gist/tylere/3faef2df6558a6458b896c1dd58024b1/xee_issue_119.ipynb

---------------------------------------------------------------------------
ValueError                                Traceback (most recent call last)
<ipython-input-9-5c431fe9edba> in <cell line: 1>()
----> 1 ds.to_netcdf('test.nc')

8 frames
/usr/local/lib/python3.10/dist-packages/scipy/io/_netcdf.py in _write_var_metadata(self, name)
    483         self._pack_int(len(var.dimensions))
    484         for dimname in var.dimensions:
--> 485             dimid = self._dims.index(dimname)
    486             self._pack_int(dimid)
    487 

ValueError: 86400 is not in list

@cug-xyx are you still able to reproduce the error you originally reported? If so, could you fully describe your environment, including the versions of Python, Xee, and any installed packages?

[cug-xyx]

Hi @tylere,

I don't know what happened, but I think the bug has been fixed, here is the same example i reproduced.

https://colab.research.google.com/drive/1QL8-_NL_Y4vupzhdAvfpvlWD0IYf7OZX?usp=sharing