Masking out cloud values with rioxarray

[nkorinek]

Hello!

I am currently trying to figure out the best way to mask out cloud values in a rioxarray object. The Earth Lab team and I have been updating our lessons to lean more on rioxarray then rasterio, and I got a little stuck on this lesson here about masking cloud values: https://www.earthdatascience.org/courses/use-data-open-source-python/multispectral-remote-sensing/landsat-in-Python/remove-clouds-from-landsat-data/.

As you can see in that lesson, currently we do it by just getting the numpy arrays out of the rioxarray objects with .values and then applying a mask to it that with numpy. I'll post the relevant code below, where landsat_pre is the landsat image with the clouds, and landsat_qa is the layer we used to produce our mask.

landsat_pre_cl_free = em.mask_pixels(
    landsat_pre.values, landsat_qa.values, vals=all_masked_values)

em.mask_pixels turns the landsat_qa array into a mask based on the values in all_masked_values and applies it to landsat_pre. Sorry if this is confusing! Just want to make sure all the relevant information is here.

The question boils down to this: is there a way to apply a mask like this to a rioxarray object that doesn't involve turning it into a numpy array? Any insight here would be greatly appreciated, thanks for your time!

[nkorinek]

Just to maybe clarify a bit more, here's what em.mask_pixels is essentially doing behind the scenes:

from numpy import ma
all_masked_values = [96, 112, 160, 176, 224]
mask = landsat_qa.isin(all_masked_values)
return ma.masked_array(landsat_pre.values, mask)

[snowman2]

Looks like xarray also has the isin method.

Side note: xarray doesn't support masked arrays, it just converts to float and uses nan ref.

So, the equivalent xarray form would be:

all_masked_values = [96, 112, 160, 176, 224]
masked = landsat_pre.where(~landsat_qa.isin(all_masked_values))

[nkorinek]

Thanks for the feedback, that looks great!