weightmap (pixel_overlaps) warnings and errors

[kerriegeil]

Hi, thanks for the helpful package.

On a Windows machine, I'm using the package successfully on ERA5 reanalysis data although I do get a user warning when calling pixel_overlaps. It occurs after the output "calculating overlaps between pixels and output polygons...". The warning is:

"/home/kgeil/miniconda3/envs/xagg/lib/python3.9/site-packages/xagg/core.py:308: UserWarning: keep_geom_type=True in overlay resulted in 1 dropped geometries of different geometry types than df1 has. Set keep_geom_type=False to retain all geometries overlaps = gpd.overlay(gdf_in.to_crs(epsg_set),'

When I try generating a weight map with the exact same shapefile but on AVHRR NDVI data instead I get a full error at the same-ish location:

"ValueError: GeoDataFrame does not support setting the geometry column where the column name is shared by multiple columns."

It looks like something is going wrong in get_pixel_overlaps around line 323 overlap_info...

I've tried rewriting the NDVI netcdf to be as identical as possible as the ERA5 file (same coord and dim names, etc) and both files are epsg:4326.

Any ideas how to get past this error?

[ks905383]

Hi Kerrie, thanks for the post.

First of all, I'm glad to hear that you were able to install xagg on a windows machine - that's been a problem for some users (#23 and #22 e.g.) - was there anything special you did to make the installation work?

Second of all, that's odd to hear. The first warning I've run into a few times, and though worrying, I don't think it's affected the final calculation (though it would be good if I would double check). The error is new to me, however. This may be a silly question, but does the AVHRR NDVI data happen to have a variable called 'geometry' or 'polygon' or something like that?

Could you link to the datafile you were using (or attach a subset version that still reproduces the error)?

Thanks!

[kerriegeil]

So really I'm using Ubuntu WSL2 on my window's 10 machine (I enabled WSL2 subsystem linux with Ubuntu- there are various instructions on the web). I installed xagg using conda from the conda-forge channel. So nothing different from how I would install any other packages.

The AVHRR NDVI data does not have any variables named geometry or polygon. It does have a variable named crs, but I tested out rewriting the data file without the crs variable and it didn't solve the problem. The data file (~60MB) I've been troubleshooting with can be downloaded here: https://www.ncei.noaa.gov/data/avhrr-land-normalized-difference-vegetation-index/access/1981/AVHRR-Land_v005_AVH13C1_NOAA-07_19810625_c20170610042839.nc. The shp file I'm using is us_county_mex_munis.zip.

Thanks for the help!

[ks905383]

Ah, fair!

Just to double-check, you said the error was also happening in xa.pixel_overlaps()? If so, I wasn't able to reproduce it with the files you linked. Could you copy in the full error message when it happens to you?

Also, this is a good reminder too (for me) to work through a few pending performance upgrades - that takes a while with the high resolution data you're working with 😅

[ks905383]

This should be now fixed through #39 , and the new update (only a few years late,...)

[ks905383]

Closed by #39 , upcoming in v0.3.1