[karimbahgat]

OCS core now ships several data sources that this repo also had, and core's ERA5-Land refactor (open-climate-service #505) broke an import here.

Removed, now in core

• CLMS NDVI: core has clms_ndvi_v3_dekadal (#507). The tests of the shared _clms.py moved to the productivity tests.
• Copernicus DEM 90 m: core defines copdem90_elevation_static, the same id (#503).
• ERA5-Land daily minimum and maximum temperature, and dewpoint: core has era5land_temperature_daily_min, era5land_temperature_daily_max and era5land_dewpoint_daily (#506). Note that the min/max temperature in ocs-plugin was local timezone aware, while in core they run in UTC+0 (local timezone support will be added later).

Fixed

• ERA5-Land soil moisture and wind speed imported helpers from core's era5_land.py, which no longer exists. The five helpers they use are now copied into rasters/_era5_land.py rather than imported from core's private code.

Checks

• Ruff and the full test suite pass. The lockfile tracks current core main.
• No dataset id here collides with one in core.

Not changed: credentials work as before, and should be standalone/separate from builtin credentials.

Note: tests/test_era5_land_daily_stats_plugin.py was already empty, so the two remaining ERA5-Land plugins have no tests.

[turban]

Thanks, this is a good cleanup, and copying the helpers is better than reaching into core's private modules. Approving. A few notes, none blocking:

1. The ERA5-Land replacements count days differently. The removed min/max/dewpoint templates asked CDS for days in the instance's own time zone. Core's era5land_temperature_daily_min/_max/era5land_dewpoint_daily work from the hourly ARCO store and their days run in UTC (their descriptions say so). For an instance like Nepal (+5:45), a "day" is shifted by almost six hours. That's probably fine, but worth a line in the PR body so nobody expects the same numbers. The ids also changed (era5land_temperature_min_daily → era5land_temperature_daily_min, clms_ndvi_dekadal → clms_ndvi_v3_dekadal), so instances that used the old ids need to re-ingest under the new ones. One leftover: the usage example in nepal-climate-service/scripts/animate.py still says clms_ndvi_dekadal.

2. Possible follow-up: soil moisture could become a YAML-only template using core's ERA5LandCDSDailyFromHourlyPlugin (variable: swvl1, temporal_aggregation: mean), since core's ARCO store mapping already includes swvl1. That would also give UTC days. Wind speed still needs its own plugin because of the hypot, so the helpers would stay for now. Not for this PR.

3. Tests: _era5_land.py is now maintained here, and the remaining ERA5-Land plugins have no tests. A couple of small unit tests for cds_fallback_area (the clamping at the poles and at ±180) and collapse_expver would be cheap and would catch drift.

[karimbahgat]

Thanks @turban . 1) True, added a brief note about it in the PR description. 2) Yes, and planning to include these in core as well, so waiting until then to make any improvements. 3) This was also raised while implementing this PR, but I don't think it's worth improving in this quick proof-of-concept repo, and will anyway disappear once soil moisture and wind speed is added to core.