Raster.to_xarray() not masking NaN nodata values

[geograFie]

It seems like NaNs are not masked when using Raster.to_xarray(). The Raster object contains the nodata attribute which is set via Raster.to_rio_dataset(), so not sure what the issue is and if that if that is a common issue, or only happens in my case.

I solved this by manually masking NaN values slope_xr.where(slope_xr != slope_xr.rio.nodata).

%matplotlib inline
import pdemtools
import xdem
from geoutils import Raster
from pyproj import CRS
import matplotlib.pyplot as plt

# download some ArcticDEM data 
bounds = (-2789094.0, 708249.0, -2789044.0, 708299.0)
dem = pdemtools.load.mosaic(
    dataset='arcticdem',
    resolution=2,
    bounds=bounds,
    version='v4.1'
)

# Xarray to Raster
crs = CRS.from_wkt(dem['spatial_ref'].attrs['crs_wkt'])
epsg = crs.to_epsg()
affine = dem.rio.transform()

raster = Raster.from_array(
    data=dem.data,
    crs=epsg,
    transform=affine,
    nodata=-9999.           # set nodata value
)

# Raster to xDEM 
x_dem = xdem.DEM(raster)
print(type(x_dem))
print(x_dem.nodata)

# Compute terrain parameter
slope = xdem.terrain.slope(x_dem)

# Convert back to Xarray
slope_xr = slope.to_xarray()
print(slope_xr.data[0,:3,:3])

This code outputs:

[rhugonnet]

Thanks @geograFie, I also noticed this behaviour yesterday when adding the from_xarray() function to solve #519. Both are now fixed in #521 with more tests, awaiting peer-review before merging and making a new release. 😉