Optimizing climate model selection in regional studies using an adaptive weather type based framework: a case study for extreme heat in Belgium

Abstract

Abstract Selecting climate model projections is a common practice for regional and local studies. This process often relies on local rather than synoptic variables. Even when synoptic weather types are considered, these are not related to the variable or climate impact driver of interest. Therefore, most selection procedures may not sufficiently account for atmospheric dynamics and climate change impact uncertainties. This study outlines a selection methodology that addresses both these shortcomings. Our methodology first optimizes the Lamb Weather Type classification for the variable and region of interest. In the next step, the representation of the historical synoptic dynamics in Global Climate Models (GCMs) is evaluated and accordingly, underperforming models are excluded. In the last step, metrics are introduced that quantify the climate change signals related to the impact of interest. Using these metrics, a scoring method results in assessing the suitability of GCMs. To illustrate the applicability of the methodology, a case study of extreme heat in Belgium was carried out. The developed method offers a framework for selecting periods within climate model datasets while considering the changes in the large-scale circulation patterns and the changes in the climate signal, each step optimized for a specific climate impact driver. This framework provides a comprehensive method for selecting periods from large ensemble GCM simulations based on weather types, ensuring relevant climate projections for subsequent research which can be applied in model ensemble-based research for different climate variables and climate impact drivers.