search

GeoStat-Framework / PyKrige

Kriging Toolkit for Python
https://pykrige.readthedocs.io
BSD 3-Clause "New" or "Revised" License
759 stars 188 forks source link

Bug: Drift function applied to isotrope coordinates #174

Open MuellerSeb opened 3 years ago

MuellerSeb commented 3 years ago

This is a bug in my opinion: https://github.com/GeoStat-Framework/PyKrige/blob/48bc43362218b283bcba51b2c0990b72f7e29fbb/pykrige/uk3d.py#L166

In GSTools, drift is applied to the input coordinates and I think this should be done here as well.

@rth @bsmurphy @mjziebarth opinions?

bsmurphy commented 3 years ago

Sorry for the slow response on this, @MuellerSeb. I wouldn't necessarily call it a bug (since I think the kriging system is happy solving for the drift terms in any arbitrary coordinate system), but I do agree that maybe this implementation is suboptimal. (For example, if you wanted to extract the drift values and actually use them for something, having them calculated in the adjusted coordinate system would probably add another layer of difficulty in getting them back into a meaningful and usable form.) So, I agree a change here might be useful.

OK, now that I looked at the exact line you're pointing to, I think I see more exactly why you're calling it a bug... So you mean specifically with the arbitrary drift function capability? If so, then yes I agree this is more of a bug (or at least a non-ideal way of exposing this functionality), since really the function should probably operate in the same space as the original non-adjusted data, for physical (and intuitive) consistency. I probably just implemented it this way originally out of simplicity, but I agree changing this would be worthwhile.