Investigating the value of adding diverse ML models to the IMPROVER blend

Description

The Met Office needs to understand whether adding a broader and more diverse set of machine-learning weather models to the IMPROVER blend will improve forecast accuracy and provide sufficient operational value to justify the added complexity. Preliminary work indicates that adding AIFS can improve forecast accuracy, but further evidence is needed to determine whether additional ML models deliver incremental skill, where diminishing returns begin, and which model combinations offer the greatest complementary benefit. This evidence will support future decisions on model acquisition, integration, and operational implementation within probabilistic post-processing. Keywords: Closed Framework