obs_info and .res join/merge

[pkpatkel]

Broadly, on what fields does the obsinfo file join to the res.df?

The index in res.obsinfo matches the 'Group' column in res.df, but res.phi does not contain X,Y,etc. info...what am I missing?

[echristi]

It's been awhile since I looked at this. The obsinfo stuff is very much a work in progress. The index of res.obsinfo is suppose to be observation names. So, if the index is wrong then X,Y won't get joined in res.phi. Something must be getting messed up reading the obsinfo file

Andy might have more insight.

[pkpatkel]

So, it looks like the obsinfo file needs to have a record for every unique entry in the .res.df file...wouldn't it make some sense to have the index of the obsinfo file join on the 'Group' column of .res.df (assuming that your 'Group's in that file are individual boreholes)? X,Y won't be changing for each record - it might make the obs_info file a little easier to compile and work with if it's based upon boreholes instead of individual records in a time series.

[echristi]

Specifics for time series is not really implemented yet. I agree that it's a little clunky and not intuitive right now if dealing with time series data. We intend to get to it and will use the obsinfo to help with that but we aren't exactly sure how yet. 'Group' in obsinfo is intended to be more generic. For example you could have several time series belong to a single group. Plotting for time series is something we would like to get into PESTools soon. Suggestions or help are welcome.

[aleaf]

I haven't looked at it in awhile either, but seems like we need to generalize how obsinfo is joined- have it join on observation name by default (since every observation has to have a unique name), but as an alternative, allow the user to specify another column. In which case it would have to do a one-to-many join (distribute information about a group, such as x, y coordinates, to multiple obsevations in res.df)