Direction of integration for cumint?

[chrisb13]

For aesthetic reasons I'd like to plot a barotropic streamfunction using an integral starting at the top rather than the bottom. (For an idealised southern ocean channel model.) Looking at the method-object-docs, I couldn't see a way to do this with cumint..? I briefly played with reindexing via strides but that feels un-xgcm like (and I didn't see how to make it work!)

Thanks for a handy package!

[jbusecke]

Hi @chrisb13, thanks for raising this issue.

I think this is related to #337 and earlier #184.

I agree this is a needed feature, but currently you would have to flip your values (not the coordinates see #605) apply the cumsum and flip them back (which is not ideal).

Would you be interested in working on this feature at all?

[dhruvbalwada]

Here is an easy and temporary solution using grid_ufunc feature:

ds_dens = xr.Dataset(
    coords={
        "zl": (["zl"], np.arange(26,28, 0.2)),
        "zi": (["zi"], np.arange(25.9, 28, 0.2))
    }
)

grid = Grid(ds_dens, 
            coords={"Z": {"center": 'zl', 'outer': 'zi'}},
           periodic=False)

ds_dens['vh'] = 0.*ds_dens.zl + 0.5


@as_grid_ufunc("(Z:center)->(Z:outer)")
def upward_sum(a):

    temp_zeros = np.zeros_like(a[...,0])[..., np.newaxis]

    temp = np.append(np.flip(np.cumsum(np.flip(a, axis=-1), axis=-1), axis=-1), temp_zeros, axis=-1)
    return temp

reg_cumsum = grid.cumsum(ds_dens.vh, 'Z', boundary='fill', fill_value=0.)
upward_cumsum = upward_sum(grid, ds_dens.vh, axis=['Z'])

Creates the following plot