ms.match() raises an unclear TypeError when observation and model result timezones differ

[jpalm3r]

Summary

ms.match() fails when the observation is timezone-aware and the model result is timezone-naive (or vice versa). The failure surfaces as a plain pandas TypeError from deep inside the matching code, with no indication of what actually went wrong.

SCADA exports are often timezone-aware UTC, so users can hit this without realizing their model result and observation disagree on timezone handling.

Root cause

Nothing in the codebase validates or normalizes timezone-awareness. A search of src/modelskill for tz_localize, tz_convert, tzinfo, and timezone returns no matches. The only related time-coordinate handling, _normalize_time_to_ns() (src/modelskill/timeseries/_timeseries.py:42-58), normalizes datetime64 resolution only, not timezone.

The first place a mismatch actually breaks something is observation.trim(period.start, period.end, ...) inside _match_space_time (src/modelskill/matching.py:405), which calls:

data = self.data.sel(time=slice(start_time, end_time))  # _timeseries.py:404

If one side is timezone-aware and the other isn't, this raises:

TypeError: Cannot compare tz-naive and tz-aware datetime-like objects

with no ModelSkill-specific context. The same class of failure can also occur further down the pipeline in align_data() (src/modelskill/timeseries/_align.py:51-85, the interp call on line 80) and TrackModelResult.subset_to() (src/modelskill/model/track.py:74-93, the join call on line 82), if trim doesn't fail first.

Suggested fix

Either:

• Normalize timezone-aware inputs to UTC-naive internally, or
• Validate the mismatch early and raise a clear, ModelSkill-specific error message, with documentation covering this requirement.

Found by a test user evaluating the network functionality ahead of a 1.4.0 release.

[jpalm3r]

Fixed on branch beta_test_found_bugs:

• 48b10134 — test reproducing the bug (a tz-aware/tz-naive pair now hits the same bare pandas TypeError the issue describes)
• 39c70f7d — _match_space_time now checks tz-awareness compatibility between the observation and each model result before trimming, and raises a clear ValueError naming both sides on mismatch

tests/test_match.py passes.