all-zeros result from imod.evaluate.streamfunction_line

In GitLab by @claumedimont on Jun 1, 2023, 18:13

The imod.evaluate.streamfunction_line takes two xarray.DataArrays as parameters: fff and frf. From a cross-section model, I only obtained the frf and flf arrays with dimensions (time: 278, layer: 39, y: 1, x: 258). I needed to create the fff array, which at first I did with fff = xr.zeros_like(flf).

The issue came when I tried to obtain the streamfunction using the obtained frf and the created fff. In the code that calculates the streamfunction, the flux on the edges is interpolated to the cell center (lines 66-67 in https://gitlab.com/deltares/imod/imod-python/-/blob/master/imod/evaluate/streamfunction.py). As the created fff only has 1 cell in the y-direction, this shift introduces nan along the entire cross section, resulting in all-zeros in the result.

The solution was to convert the fff into an array with 2 cells in the y-direction.

Start in imod/evaluate/streamfunction.py at the interpolation noted around lines 66-67, and reproduce the case with fff having one y cell and frf shaped (time, layer, y, x). Check how the shifted edge flux produces NaN values and an all-zero result. Done means the reported cross-section input no longer silently returns all zeros, with a regression test for the one-cell y dimension.