MissingCRS error while opening MODIS data using rioxarray

Question

I have a MODIS data. It is a hdf file (not hdf5). I can open it using ArcGIS and QGIS with no crs. When I use the following script, I get MissingCRS: CRS not found. Please set the CRS with 'rio.write_crs()'. error. Since I cannot open the file, I cannot use rio.write_crs().

import rioxarray

modis_pre_path = "C:/Users/user/Desktop/modis.hdf"
modis_pre = rioxarray.open_rasterio(modis_pre_path)

How can I solve this issue?

The full traceback:

MissingCRS                                Traceback (most recent call last)
~\AppData\Local\Temp/ipykernel_1520/3420315969.py in <module>
----> 1 modis_pre = rxr.open_rasterio(modis_pre_path)
      2 type(modis_pre)

C:\OSGeo4W\apps\Python39\lib\site-packages\rioxarray\_io.py in open_rasterio(filename, parse_coordinates, chunks, cache, lock, masked, mask_and_scale, variable, group, default_name, decode_times, decode_timedelta, **open_kwargs)
    926     # open the subdatasets if they exist
    927     if riods.subdatasets:
--> 928         return _load_subdatasets(
    929             riods=riods,
    930             group=group,

C:\OSGeo4W\apps\Python39\lib\site-packages\rioxarray\_io.py in _load_subdatasets(riods, group, variable, parse_coordinates, chunks, cache, lock, masked, mask_and_scale, decode_times, decode_timedelta, **open_kwargs)
    680         with rasterio.open(subdataset) as rds:
    681             shape = rds.shape
--> 682         rioda: DataArray = open_rasterio(  # type: ignore
    683             subdataset,
    684             parse_coordinates=shape not in dim_groups and parse_coordinates,

C:\OSGeo4W\apps\Python39\lib\site-packages\rioxarray\_io.py in open_rasterio(filename, parse_coordinates, chunks, cache, lock, masked, mask_and_scale, variable, group, default_name, decode_times, decode_timedelta, **open_kwargs)
   1033         result.rio.write_crs(rio_crs, inplace=True)
   1034     if has_gcps:
-> 1035         result.rio.write_gcps(*riods.gcps, inplace=True)
   1036 
   1037     if chunks is not None:

C:\OSGeo4W\apps\Python39\lib\site-packages\rioxarray\rioxarray.py in write_gcps(self, gcps, gcp_crs, grid_mapping_name, inplace)
   1226         data_obj = self._get_obj(inplace=True)
   1227 
-> 1228         data_obj = data_obj.rio.write_crs(
   1229             gcp_crs, grid_mapping_name=grid_mapping_name, inplace=inplace
   1230         )

C:\OSGeo4W\apps\Python39\lib\site-packages\rioxarray\rioxarray.py in write_crs(self, input_crs, grid_mapping_name, inplace)
    485 
    486         if data_obj.rio.crs is None:
--> 487             raise MissingCRS(
    488                 "CRS not found. Please set the CRS with 'rio.write_crs()'."
    489             )

MissingCRS: CRS not found. Please set the CRS with 'rio.write_crs()'.

Answer 1

Maybe you can try opening it with a context manager

import xarray as xr
import rioxarray

modis_pre_path = "C:/Users/user/Desktop/modis.hdf"

# Open the data using a context manager
with xr.open_dataset(modis_pre_path) as file_hdf:
    # You can open the data using the code below, however if you use rio.write_crs
    # it will ensure that crs for your data persist throughout your analysis
    # modis_pre = file_hdf # not sure if this should or not be commented
    modis_pre = file_hdf.rio.write_crs(file_hdf.rio.crs)

# view the data
modis_pre

Another option to try is to extract the numpy.array data with a context manager...

with rio.open(modis_pre_path) as file_hdf:
    modis_pre = file_hdf.read(1)

... and then create a xarray object with it

Answer 2

I've been plagued with this problem for a few days, and realized that it's most likely an issue with some MODIS product. For example, the lat and lon stored in MYD021KM are for 5km (406 x 271) while the radiation data is for 1km (2030 x 1354), and this inconsistence would cause this 'MissingCRS error'.