Kd_tree for model collocation

[vic1309]

Hey,

I am new to Pyresample, and I am currently struggling with collocating model and altimeter tracks. I have the following case: a geophysical model with a given variable (background map), and one X-track track (dotted line).

I want to collocate the modeled fields onto the Jason-1 lat,lon points. I am using the kd_tree.get_neighbour_info function for that:

ds_geo = pyresample.geometry.GridDefinition(lons=df.lon.values, lats=df.lat.values) # Model lat and lon values 
pos_geo = pyresample.geometry.SwathDefinition(lons=dx.lon.values, lats=dx.lat.values) # X-track lat and lon values

_, valid_output_index, index_array, distance_array = \
                            pyresample.kd_tree.get_neighbour_info(
                                source_geo_def=ds_geo,
                                target_geo_def=pos_geo,
                                radius_of_influence=5000,
                                neighbours=1,
                                reduce_data=False)

index_array_2d = np.unravel_index(index_array, ds_geo.shape)

However, I am obtaining the following error:

------> index_array_2d = np.unravel_index(index_array, ds_geo.shape)
ValueError: index 2347004 is out of bounds for array with size 2347004

Is it because part of the observations fall outside the model domain? Any tips?

Thanks in advance.

[djhoese]

Do you want to get the model data for those points? There is the get_sample_from_neighbour_info function for doing this. Is there a reason you can't use this?

I think the main issue you're running into which is currently handled in this function besides applying the valid_output_index and valid_input_index arrays, is that the underlying KDTree object is setting "out of bounds" index values to len(input_array). However, I'm surprised you're getting this index as the pyresample function you called is supposed to be reassigning this "out of bounds" index to -1.

As a first step, you could check if valid_output_index.all() is true and also valid_input_index.all() is true. If not then you'll need to make sure to take those into account when doing the indexing yourself.

...or use the get_sample_from_neighbour_info function.