Handle multiple time periods of 1 model setup

[Hendrik1987]

Dear modelskill team.

I have one point observation and one model setup, run for two seperate periods in time, say a and b. I have scattered measurements, and don't want to run a long model during calibration phase.

Here is what I tried:

obs1 = ms.PointObservation(obs1.dfs0, item=0)
mr1a = ms.model_result(mdl1_part1.dfs0, item=0)
mr1b = ms.model_result(mdl1_part2.dfs0, item=0)

I can of course do seperate comparers and plotting:

cmp1 = ms.match(obs1, mr1a)
cmp2 = ms.match(obs1, mr1b)
cmp1.plot.scatter()
cmp2.plot.scatter()

But I actually want to concat mr1a and mr1b somehow. This does not work:

cmp = ms.match(obs1,[mr1a, mr1b])
cmp.plot.scatter()

I get ValueError: No data found in selection.

I did not expect this to work, as I suppose the logic is reserved for having two different model setups rather than periods.

Should I concat the model data as preprocessing step or is there a smarter way?

I guess, this also relates to the "max_model_gap" discussion that we had back then. But I can't figure out how this is handled in the latest modelskill version.

Thanks for help and hints!

[ecomodeller]

Have you tried to "add" two comparers?

If the model have the same name I think it should work as expected.

cmp = cmp1 + cmp2

https://github.com/DHI/modelskill/blob/main/notebooks/Combine_comparers.ipynb

[Hendrik1987]

Yes, that works, thanks!

Only the cmp.plot_timeseries() behaves (maybe) little unexpecte, as it zooms into the first model period only.

Inspecting cmp.data and the cmp.plot(), however, confirms that all data is correctly available in cmp.

[ecomodeller]

Thanks for pointing that out, seems like something someone😉 should look into.