Find Nearest non-nan Elements in a Dfsu file

[Barcelona2019]

Is your feature request related to a problem? Please describe.
I am trying to sample modelled flood surfaces at locations where we have surveyed flood marks to compare to, but at some locations (x,y), the .find_nearest_elements gave me nan values - the problem is the surveyed flood marks are almost always on the edge wet/dry boundary.

Describe the solution you'd like
I would like to have a similar function to x.find_nearest_elements, where it finds the nearest non-nan elements, possibly x.find_nearest_nonnan_elements ?

[jsmariegaard]

Hi @Barcelona2019, this is a good idea, the main challenge here is that dfs.find_nearest_elements() rely only on the file header, whereas the method you suggested would rely on data. I wonder if a solution could be to give dfs.find_nearest_elements() an "elements" argument where you specify a subset of elements that the nearest elements should be found within. You could then for each timestep do the following:

wet_elements = np.where(~np.isnan(data[0][time_step,:]))
nearest_wet_elements = dfs.find_nearest_element(..., elements=wet_elements)

Do you think that would work?

[Barcelona2019]

Hi Jesper,

Thank you for your work on this thread - I think that would be great!

I spent some time and found a clumsy solution with a while loop ----

jj=1
idx = dfs.find_nearest_elements(x, y, n_nearest=jj)
aa = np.nanmax(ds.data[0][:,idx])
while np.isnan(aa):
    jj=jj+1
    idx = dfs.find_nearest_elements(x, y, n_nearest=jj)
    aa = np.nanmax(ds.data[0][:,idx[:]])

The above code helps me locate the maximum flood level at (x,y) or its nearest points - but I still think your suggestion would lead to a more elegant solution.

[ecomodeller]

In order to test this we would need a Dfsu test file where some elements are dry.

• Find suitable Dfsu example file of flood/dry