Gridding sparse data (fine to coarse)

[NicWayand]

hi @JiaweiZhuang, I am trying to regrid sparse (~1 km) Ice Bridge data (airborne flight transects) to a coarser 25 km grid, using the new conservative_normed method from your master branch. But I am getting only NaNs out.

Even if there is only one non-NaN fine cell within a coarse cell, it should assign that only fine cell's value to the coarse cell, correct?

A runable example can be found here.

[JiaweiZhuang]

In your case, getting all NaNs is the expected, default behavior for most algorithms, as explained in #22 (comment).

To get rid of NaNs, you need to apply masking like #22 (comment). conservative vs conservative_normed only affects the magnitude of the regridding result, not the occurrence of NaNs.

[JiaweiZhuang]

There should be a ds_fine['mask'] that masks all cells out of your data trajectory (i.e. all NaNs in the input data).

[NicWayand]

Ah thanks! I had the ds_target['mask'] but forgot the ds_fine['mask'].

Now getting non-nan values in the coarse grid as expected:

However, now all other coarse cells were set to 0, when I would have expected them to be NaN. From past convos I understand this is the expected current behavior.

Using the add offset, mask > offset, subtract offset, workaround works for now but feels a little hacky. Thanks for all the work on #22!