Patch release for the 7.3.x line. It corrects the threshold definitions of five DCSC indices (definitions from the DRIAS portal, Météo-France) and updates their documentation strings.
Fixed
- TXHWD, TNCWD and TNND: the comparison and offset were wrong, so the indices counted the wrong days. For example, an 8-day heat spell gave 357 days instead of 8. They now follow the DRIAS definition: TXHWD counts Tx >= normal + 5 °C, TNCWD and TNND count Tn <= normal - 5 °C, and a spell needs at least 6 consecutive days. (#616, #618, #619)
- TXND and TNHT: days exactly at normal + 5 °C are now counted (>= instead of >), as in the DRIAS definition. (#620, #621, #622)
- The generated API docstrings now carry the corrected definitions. (#624)
Notes
- These fixes change results for the five indices above compared with all earlier releases that contain the DCSC indices.
- icclim does not set the reference normal (DRIAS uses 1976-2005); you provide it with the
normalargument. - The exact percentile interpolation fix for SciPy 1.18.1 (issue #603) was released in 7.3.5.
Full changelog: https://github.com/cerfacs-globc/icclim/blob/v7.3.6/doc/source/references/release_notes.rst