tastatham / coss Goto Github PK

Not all areal interpolation methods use unique source and target identifiers.
_check_uid calculates a unique identifier if one is not present, regardless of the interpolation method.
areal_checks should specify an additional parameter, to determine whether a unique identifier should be checked/calculated.

One enhancement that could speed up both areal weighting / dasymetric mapping is to use Dask GeoPandas instead of GeoPandas for the underlying geometric calculations.

However, the GeoPandas overlay function used by areal weighting is not yet available in Dask GeoPandas.
To move this issue forwards, the overlay function must be merged into Dask GeoPandas.

This would also involve merging DataFrames based on the index, which mean #5 would be resolved.

Methods that interpolate values based on areal weighting (as well as pycno) produce erroneous results when there are are duplicate geometries.
Geometries may be duplicated when handing certain types of datasets e.g. survey data.
In this scenario, I was using survey data and I simply forgot to perform a groupby per unique identifier, prior to running the interpolation.
This returned interpolated values for all targets but the results were erroneous.

My initial thought was to update all areal_checks to add an additional check for duplicate geometries/unique identifiers;

My idea is to add a check, to see whether duplicate values per geometry/unique identifier are present.

Currently the coss package merges pd.DataFrames based on a column.
However, merging on a set index is more performant.
Once unique identifier bugs/enhancements are complete, merges will be based on the index.

An Error is thrown in the _check_uid function , when GeoDataFrame.name does not exist.
.name is used to create a unique identifer when one does not exist or is not specified.
A solution is to use the uuid Python module.

Add WIP example notebook for all areal interpolation methods, using both intensive and extensive variables.