mdtanker
commented
1 year ago It looks like the rounding error is occurring somewhere in the call to hm.filters._fft.ifft(), which is the function that appears to call xrft, so it might be an issue with xrft.
I think a simple fix is to add the following lines to the end of the apply_filter() function to copy the original coordinate values onto the new filtered grid:
def apply_filter(grid, fft_filter, **kwargs):
...
rest of function
...
# use original coordinates on the filtered grid (rounding errors likely occurred)
filtered_grid = filtered_grid.assign_coords(
{
dims[1]:grid[dims[1]].values,
dims[0]:grid[dims[0]].values
})
# Check that coordinates exatctly match the original coordinates
np.testing.assert_equal(filtered_grid.easting.values, grid.easting.values)
np.testing.assert_equal(filtered_grid.northing.values, grid.northing.values)
return filtered_gridHappy to open this in a PR if people think it's the best fix for now!
leouieda
LL-Geo
santisoler
roxyboy
Description of the problem:
When using the new grid transformations, I noticed that several (at least
derivative_upwardandgaussian_lowpass) functions are slightly changing the easting coordinates of the dataarrays. I was attempting to add the resulting grids as a new dataset variable with:Plotting the new grid shows:
Compared to the normal method of creating a new array shows an issue with the high and low easting values.
Numpy testing confirms that the easting values of the produced grid don't match the original past the 10th decimal place.
The northing coordinates are exactly equal for this dataset, but northing and easting both fail the test on a dataset of min.
I have a workaround, which is to round the coordinate values to the same number of decimal places before merging, which gives the desired results.
I need to add the resulting data back to the original dataset since my grid has NaN's, which I've filled with -999 and need to mask the results with the original data extent.
Any ideas what's causing this issue?
Otherwise, these new functions are great!
System information
conda listbelow:Output of conda list