Empirical Probability Calculation Error for SEAS51 Data

[nishadhka]

The get_mean_ens_triggers function (

ibf-thresholds-triggers/utils.py

Line 333 in 39c32b5

fct_mod = (ens_data <= mod_thr).mean(dim="member")

) is currently miscalculating empirical probabilities when applied to SEAS51 forecast data. The issue stems from the differing number of ensemble members over time:

• 198101-201612: 26 members
• 201701-present: 51 members

The current implementation (fct_mod = (ens_data <= mod_thr).mean(dim="member")) incorrectly assumes 51 members for the entire period, leading to erroneous probability calculations.

Technical Details:

• The dim="member" argument in .mean() calculates the average across all members, regardless of whether they are present in the data for a given time period.
• This results in the average being artificially lowered for the earlier period (198101-201612) where there are fewer members.