ENVIRONMENT AND ENERGY

Project

Ocean Temperatures and European Rain

Global warming is fundamentally altering the water cycle, making the accurate projection of regional weather patterns a matter of importance for European infrastructure, agriculture, and natural hazard management. However, regional climate models (RCMs), which simulate atmospheric processes at high resolution over specific continents, heavily depend on the data they receive at their geographical boundaries. One of the most critical inputs is sea surface temperature (SST). Even minor discrepancies in how ocean temperatures are measured or defined can cascade through atmospheric simulations, leading to considerable uncertainties in rainfall predictions over land.

To quantify these uncertainties, a research team at the working group on Geodetic Earth System Science at the University of Bonn, launched the high-performance computing project "Modeling Oceanic Signals in Europe's Atmospheric Water Cycle" (Project ID: moistly). The team focused on the EURO-CORDEX domain, that is, the entire European continent and its adjacent marine regions. At the heart of the considerations was a subtle but consequential physical distinction: the difference between bulk sea surface temperatures (defined as average over a few meters below the surface) and skin temperatures associated with the actual micro-surface layer. Due to the day-night cycle and the action of atmospheric winds, the ocean's skin is typically ~0.2 degrees cooler than the underlying bulk water. Climate simulations have long overlooked this distinction, often driving atmospheric models with foundation or bulk temperatures, or even switching between the two variables. The project therefore set out to discover how this systematic temperature difference alters the moisture transport from the Northeast Atlantic Ocean, the Baltic Sea, and the Mediterranean Sea onto the European continent in a state-of-the-art RCM.

The Computational Challenge

Simulating atmospheric processes—such as the exchange of heat, moisture, and momentum—with the necessary detail over many years is a major computational task. To that end, the research team deployed the Icosahedral Nonhydrostatic modeling framework in climate limited-area mode (ICON-CLM), configured with 60 vertical layers and a spacing between horizontal grid points of about 12 km. See Animation 1 for an illustration of how water vapor in ICON-CLM is transported from the ocean to the continent for the time period January 2003. To isolate the impact of the marine boundary conditions, the team performed extensive parallel sensitivity experiments spanning 456 historical months (in repeated integrations of the period 2002–2011). The project was very data-intensive, involving the processing of roughly 12.2 terabytes of read inputs and generating 9.4 terabytes of written model outputs for every single simulated month. Managing these massive workflows while solving the model’s differential equations required the high-performance computing infrastructure of the JUWELS Cluster Module at the Jülich Supercomputing Centre. For each individual model integration, the team utilized 16 computational nodes (768 physical cores) simultaneously to achieve the necessary parallel efficiency.

Key Findings

The ICON-CLM simulations performed on JUWELS revealed that fractional changes in SST matter (Figure 1). Using foundation ocean temperatures instead of the slightly cooler skin temperatures increased the simulated annual rainfall over the European continent by 2.4% when averaged over all land points. This difference translates to an additional 12 millimeters of water per year, driven entirely by enhanced evaporation over the warming seas and subsequent moisture transfer to land. In several regions, such as over the Eastern European Plain and the coastlines of Italy or Spain, the increase in annual rainfall reached 6% or more, with larger contributions from summer than winter months. Crucially, the supercomputing data showed that this additional moisture does not manifest as a gentle, uniform drizzle. Instead, it systematically shifts statistics toward more intense precipitation events. Furthermore, the simulations highlighted the important role of internal model variability. During winter, European weather is largely dictated by strong, predictable Atlantic jet streams. In summer, however, when these large-scale atmospheric steering forces are weak, local thermodynamic processes can become very prominent and amplify the model response to small differences in the adopted SST fields.