TCRM 3.2.0

TCRM long-term hazard showed unrealistically high winds along the south coast of Indonesia and too-low winds over southeast Western Australia.

The issue was traced mainly to changes in the synthetic track movement model statistics. In v3.1.17, the movement model had shifted away from the v2.1 behavior in three important ways:

1. Speed and bearing tendencies had been using rate-based statistics rather than direct moving speed and bearing statistics.
2. The statistical persistence calculation had changed from the v2 raw np.correlate() autocorrelation formula to statsmodels.tsa.acf(), which strongly reduced AR(1) persistence and increased the random innovation term.
3. For angular variables such as bearing, v3 used SciPy circular standard deviation, which produced larger bearing spread than the legacy TCRM circular statistics.

These changes made synthetic TC motion too noisy and less able to maintain recurving tracks toward southeast WA.

Changes Made in this version:

• Restored use of direct moving speed and bearing statistics, rather than speed-rate and bearing-rate statistics.
• Restored the v2 autocorrelation formula in StatInterface/generateStats.py using raw np.correlate().
• Restored TCRM’s legacy circular mean/std functions for angular statistics, especially bearing.
• Kept the v2-style speed update behavior in the track generator.
• Some other small changes for package updating