Avoid 1e-5 added to precipitation for all timesteps

[LINAMARIAOSORIO]

Checklist

• Have you followed the guidelines in our Contributing document?
• Have you checked to ensure there aren't other open Pull Requests for the same update/change?
• Added tests for changed code.
• Updated documentation for changed code.

Description

Interception.get_interception guards the interception-rate equation I_R = 1 - exp(-I_D * d_p / P_m) against a zero monthly precipitation by replacing P_m with 1e-5 in the denominator. The guard was written as prec = P * (P != 0) + (P * (P == 0) + 0.00001), so the + 0.00001 applied to every value and positive precipitation reached the denominator as P + 1e-5 mm. The guard is now pcr.ifthenelse(precipitation != 0, precipitation, pcr.scalar(1e-5)): positive precipitation is used unchanged and the epsilon only applies where P == 0.

Follow-up commits on top of the fix:

• ruff format of the changed module (the lint job runs ruff format --check).
• Regression tests in tests/unit/hydrological_processes/test_module_interception.py: positive precipitation reproduces the equation evaluated in float64; a small positive precipitation (P = 1e-4 mm) that the old guard lowered by about 6 % (this test fails on main); zero precipitation stays finite and zero.
• Golden fixtures regenerated (28 of the 45 files moved), SHA256SUMS rewritten and a new entry at the top of tests/fixtures/AUDIT.md with the provenance, the impact table and the replaced checksums.
• Changelog entry under Unreleased › Fixed.
• After review: two sentences of the audit notes corrected (the sampled-series justification and the rejected cell, which is also in srn0000001.tif), and origin/main merged so the branch is up to date for the protected merge.

Related Issue

• Resolve 1e-5 is added to precipitation for all timesteps in interception calculation #319

Motivation and context

The 1e-5 mm perturbation of the denominator diverges from the documented interception rate (doc/source/overview.rst, equation interception-r; supplementary document, equation SE3). The hydrological effect is numerical noise (at most 3.05e-5 mm on the regression rasters, 2.4e-4 mm on smc), but the semantic golden comparison of the CI matrix (rtol = 1e-5) rejected the previous goldens on one srn cell, so the fixtures had to be regenerated.

How has this been tested

• Local suite (pytest --ignore=tests/integration/doc): 997 passed, 1 skipped.
• test_interceptionCalc_small_positive_precipitation_keeps_denominator fails with the module from main (4.90e-5 instead of 5.23e-5) and passes with this branch.
• Goldens regenerated with tests/fixtures/regenerate_golden.py in a Docker container of ubuntu:24.04 (glibc 2.39-0ubuntu8.8) with the environment from ci/golden-env.lock and ci/golden-pip.lock. Before promoting them, the container reproduced the previous goldens byte-for-byte on main; the 28 regenerated digests are identical to the ones the exact job reported for this branch before the regeneration (run 33662562795) and the 17 untouched files keep their digests.
• The lint, docs and exact CI jobs were also run locally with act (ubuntu-24.04 runner image): green.
• CI on this pull request: ci-success green.

Screenshots

• N/A

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 92.47%. Comparing base (9deebb4) to head (1613d18).

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #323      +/-   ##
==========================================
- Coverage   92.48%   92.47%   -0.01%     
==========================================
  Files          60       60              
  Lines        3910     3908       -2     
  Branches      487      487              
==========================================
- Hits         3616     3614       -2     
  Misses        238      238              
  Partials       56       56              

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.