Intercomparison of a High-Resolution Regional Climate Model Ensemble for Catchment-Scale Water Cycle Processes under Human Influence

TL;DR

This study presents a multi-model, high-resolution regional ensemble (ICON and TSMP1) for Europe within the DETECT framework, evaluating catchment-scale hydroclimate variables (T2m, P, ET) against E-OBS, GPCC, and GLEAM for 1990–2020 across four basins. It analyzes ten EURO-CORDEX-domain simulations, exploring convection-permitting and parameterized schemes, SST boundary conditions, irrigation, and sparse-input reanalysis to quantify structural differences and anthropogenic influences on the land–atmosphere–subsurface system. Key findings show persistent cold biases in T2m, large and seasonally variable precipitation biases, and divergent ET behaviors across models, with irrigation and SST forcing notably improving Po and Rhine performance and highlighting basin-dependent skill. The work provides a baseline for DETECT’s multimodel ensemble, informs future model development and scenario analyses, and establishes a publicly available dataset for basin-scale hydroclimate research and decision-support.

Abstract

Understanding regional hydroclimatic variability and its drivers is essential for anticipating the impacts of climate change on water resources and sustainability. Yet, considerable uncertainty remains in the simulation of the coupled land atmosphere water and energy cycles, largely due to structural model limitations, simplified process representations, and insufficient spatial resolution. Within the framework of the Collaborative Research Center 1502 DETECT, this study presents a coordinated intercomparison of regional climate model simulations designed for water cycle process analysis over Europe. We analyze the performance of simulations using the ICON and TSMP1 model systems and covering the period from 1990 to 2020, comparing against reference datasets (E-OBS, GPCC, and GLEAM). We focus on 2 m air temperature, precipitation and evapotranspiration over four representative basins, the Ebro, Po, Rhine, and Tisa, within the EURO CORDEX domain. Our analysis reveals systematic cold biases across all basins and seasons, with ICON generally outperforming TSMP1. Precipitation biases exhibit substantial spread, particularly in summer, reflecting the persistent challenge of accurately simulating precipitation. ICON tends to underestimate evapotranspiration, while TSMP1 performs better some seasons. Sensitivity experiments further indicate that the inclusion of irrigation improves simulation performance in the Po basin, which is intensively irrigated, and that higher-resolution sea surface temperature forcing data improves the overall precipitation representation. This baseline evaluation provides a first assessment of the DETECT multimodel ensemble and highlights key structural differences influencing model skill across hydroclimatic regimes.

Figures (22)

• Figure 1: Overview map of our study area in the EURO-CORDEX domain. Highlighted are the four selected river catchments used in this study for area-averaged analysis: Rhine River (Rees station), Rio Ebro (Tortosa station), Fiume Po (Pontelagoscuro station), and Tisa River (Senta station, tributary to Danube River). The named stations are GRDC-registered locations of river gauges and thus define the upstream portion of the drainage basins, respectively GRDC_hydrosheds.
• Figure 2: TSMP1 CTRL (TSMP1_12_ctrl) 1990-2020 mean biases of T2m, P and ET with respect to E-OBS, GPCC and GLEAM respectively.
• Figure 3: Boxplots of seasonal interannual T2m biases for different watersheds for the period 2003--2010.
• Figure 4: Boxplots of seasonal interannual precipitation biases for different watersheds for the period 2003--2010.
• Figure 5: Boxplots of seasonal interannual ET biases for different watersheds for the period 2003--2010.