Sensitivity Analysis of ICOsahedral Nonhydrostatic (ICON) Simulations at Convection Permitting Resolutions over Western Türkiye

The main objective of this study is to analyze the performance and sensitivity of the ICOsahedral Nonhydrostatic (ICON) regional climate model at convection-permitting resolutions in reproducing temperature and precipitation patterns over Western Türkiye. Because the computational cost of ICON scales strongly with horizontal resolution, domain size, and simulation length, conducting multi-decadal integrations for each parameterization choice is typically prohibitive. To facilitate efficient sensitivity testing, this study prioritizes single-year simulations, which allow rapid iteration of physical parameterizations relative to multi-decadal integrations.  To balance computational efficiency with climatic relevance, we selected a representative baseline year that best represents the 1960–2024 climatology, based on observational metric. This selection was determined by evaluating 64 years of meteorological data (1960–2024) to find the year that most closely aligned with long-term regional averages of 212 stations. Based on this comparison, 2015 was selected as the representative year with the long-term mean temperature of 13.3 °C compared to 13.9 °C in 2015, and the long-term mean precipitation of 615.464 mm compared to 614.58 mm in 2015. With 2015’s temperature and precipitation values nearly mirroring the multi-decadal mean, it provides a robust platform for executing physical-parameter testing more efficiently than full long-term integrations. Consequently, the ICON simulations in this study use 2015 as the baseline year.

To provide realistic large-scale atmospheric forcing, initial and boundary conditions are prescribed from the ERA5 reanalysis data which offers high-resolution, temporally consistent fields and has been widely demonstrated to represent synoptic variability relevant to Western Türkiye.  Because regional model simulations are strongly influenced by unresolved (sub-grid) processes, we implement a structured sensitivity assessment within the ICON modeling system. Parameterization schemes governing deep convection, cloud microphysics, and radiative transfer are varied in a controlled experimental design to quantify their influence on simulated hydroclimate. The objective is to identify the physical configuration that minimizes systematic errors in temperature and precipitation and yields a robust representation of coastal-inland and topography-driven gradients characteristic of Western Türkiye. Therefore, the simulation outputs of each model configuration are evaluated against observations from 221 meteorological stations obtained from the General Directorate of Meteorology. Model and observation comparisons are performed at monthly and seasonal scales, and skill is summarized using RMSE and mean bias, alongside correlation and index of agreement to assess the consistency of simulated spatial patterns and temporal evolution across seasons.

Keywords: ICOsahedral Nonhydrostatic (ICON) Model;  Türkiye; Regional Climate Modelling