Xarray and collections of forecasts

Science
6

I thought I had seen an issue along those lines, but I think I searched ‘coordinates’ and ‘exception’ (as technically I’m selecting a non-dimension coordinate in one case) so I didn’t find that one. dt.sel(time=dt, errors="ignore") would work great. I’ve got some more ideas that I’ll continue in the issue.

A method like .concat_child_datasets would also be helpful. Between those two it could simplify two of my methods from.

    def constant_offset(self, offset: Union[str, int, float, timedelta]) -> xr.Dataset:
        timedelta = pd.to_timedelta(offset)

        filtered_ds = []

        for child in self.datatree_obj["model_run"].children.values():
            try:
                selected = child.ds.sel(forecast_offset=timedelta)
                filtered_ds.append(selected)
            except KeyError:
                pass

        combined = xr.concat(filtered_ds, "time")
        combined = combined.sortby("time")
        return combined

to

    def constant_offset(self, offset: Union[str, int, float, timedelta]) -> xr.Dataset:
        return self.datatree_obj["model_run"].sel(forecast_offset=offset).concat_child_datasets("time").sortby("time")

I think it could allow other datatrees with similarly really closely related but non-alignable datasets to be quickly collapsed into aligned datasets with less boilerplate.

2 Likes