Nearest point 3D optimized (taking into account the height difference)

[heu3becteh]

Hello!

So as I understand right now topo_mask is used by Point-Stat only to filter the obs/fcs data not satisfying thresholds use_obs_thresh (discarding the stations with height difference above the threshold) and interp_fcst_thresh (discarding the forecast grid points with height difference above the threshold).

Geography Match uses these thresholds and land_mask to filter land/water difference between observation and forecast.

Is there any other way to use height difference between observation site and forecast grid point?
I would like to use the height difference to find Nearest Neighbor: min(horizontal_distance + vertical_distance*100). Is it possible?
May it be possible to use height difference as weights or some other way, apart from discarding the data not satisfying the threshold?

Best regards,
Alexander

[j-opatz]

Hi Alexander,

And thank you for your question. Unfortunately, there's no way currently in MET to use the height difference in the manner you're describing. As you noted, topo_mask allows filtering of observations and forecasts compared to an intermediate height difference of the topo map. This step is the closest you can get to comparing your forecast to observation height differences, as topo_mask will assume the topo map definition is within the provided forecast file if the file_name entry isn't used.

But from your description, it sounds like you'd like to extract the actual height distances to be used for nearest neighbor and weighting. Nearest Neighbor does exist as a regridding method in regrid, but those values cannot be accessed or saved in any method.

[heu3becteh]

Thank you for the reply!

But from your description, it sounds like you'd like to extract the actual height distances to be used for nearest neighbor and weighting. Nearest Neighbor does exist as a regridding method in regrid, but those values cannot be accessed or saved in any method.

Yes, I want to use height difference together with horizontal distance when looking for the nearest neighbour.
I will try looking into possibilities to make it work like that, but probably will just use topo_mask filtering in the end.