rileyhales / geomatics Goto Github PK

Currently you can only do series on 2 dimensional data, or 3 if the 3rd is time. we want to be able to do up to 4 if the 4th is time. Case in point soil moisture data and wind speed data often stored with a depth/height 3rd dimension

Probably necessary changes:

• generalize the x_var, y_var to be dim1, dim2
  • can we still have default dims for lat/lon/time??
  • maybe we can try to guess them?
• add a new dim3 parameter
• modify slicing_info utility to intelligently determine indices, dimension order, shape, etc
• can we still call it point_series for 1 dimensional data?? Do we need to rename the 3 kinds of series??

option 2:

• have a series_1d, series_2d, series_3d function. duplicate code and more to maintain and merge later but probably quicker to generate. pushes more responsibility onto the user. less ability to be generally applied.

• single file to multiband geotiff

xarray.open_rasterio looks like a good option

use geopandas so you can accept a geojson?? (can we reliably turn a geojson into a shapefile)
we should change the conversion tools between geojson and shapefile to use this. We can go back and forth easier.

There is also the option of using the rasterstats mask option in addition to zonal_stats which may perform better. Could open possibility of accepting more geometry inputs/require fewer different functions.

pygrib is on conda-forge so we should be able to use it as a dependency.

• need to use it as an engine -> swap var for band_number.
• will need to modify the _check_var_in_dataset option to check that the band number is in the length of files available.
• speed test to understand the potential benefits.
• does this impact any other dependencies?
• will implementing this disrupt the other engine workflows?