Cost-Loss Ratios
The Relative Economic Value (REV) of Ensemble Logistic Regression (ELR) fine-tuning of IFS and cGAN to predict probabilities of exceedances across different rainfall thresholds (in mm/day) and lead times (in days).
REV is plotted against the cost-loss ratio equivalent to the forecasted probability of threshold exceedance. REV scores are calculated using test years of 2021, 2023 and 2024.
ℹ️ A perfect REV score would be 1, whereas 0 means no value is added on top of a default always/never forecasted event.
For further information on the REV refer to Wilks (2001).
