Kilometre-Scale Climate Modeling: 5 Breakthroughs in IFS + FESOM/NEMO Simulations

Km-scale models, such as ECMWF’s Integrated Forecasting System (IFS), play a vital role in the understanding and prediction of weather patterns and long-term climate trends. They can resolve smaller features, such as mesoscale eddies in the ocean, urban areas over land, and convective cloud structures in the atmosphere. These smaller processes are crucial for simulating a realistic uptake of heat by the ocean, or for extreme weather events and regional climate dynamics. However, while high-resolution models offer greater spatial accuracy, they come with the trade-off of significantly higher computational demands, making them more resource-intensive, and often require longer run times for simulations than low-resolution models. Although low-resolution models are computationally less expensive, they may overlook important small-scale processes, leading to less precise predictions in certain regions and different transient behaviour. Additionally, they may misrepresent extreme weather conditions. Balancing these trade-offs and developing improvements for high-resolution simulations are key for advancing climate models and are the focus of the nextGEMS project.

In the paper „Multi-year simulations at kilometre scale with the Integrated Forecasting System coupled to FESOM2.5 and NEMOv3.4“ by Rackow et al., a group of scientists, from several modeling centers, academia, and the wider nextGEMS community, present and evaluate the first multi-year-long km-scale model simulations, developed within the nextGEMS project, that connect the IFS to the ocean models NEMO and FESOM. The goal of the study was to present the different nextGEMS simulations and the scientific and technical progress that was made over the first three model development cycles. The authors present the data that emerged from the multi-year simulations, explain how to adjust the code of the ocean, atmosphere, and land models, and detail how to set up km-scale simulations that more accurately represent real-world processes. A particular focus was set on sea ice leads, the urban diurnal temperature cycle, and other phenomena, including variability patterns such as the Madden-Julian Oscillation (MJO) and the Quasi-Biennial Oscillation (QBO).

The IFS-FESOM model that emerged as a result of the nextGEMS research better incorporates connections between the atmosphere, land areas, urban areas (e.g., cities), the oceans, and sea ice in the polar regions. These components are important parts of the climate system as their interactions regulate the Earth’s climate, drive weather patterns, and impact the environmental conditions on our planet.

This blog post discusses the five main technical changes that led to the improvement of the simulations:

1. Tackling the issue of water non-conservation

Processes like evaporation, condensation, and precipitation enable the distribution of water across the globe and between the land surface, the atmosphere, and oceans. They also change the phase in which water is present on Earth. However, the total amount of water remains nearly constant in the Earth system and should also be conserved in climate models, without artificial sources or sinks of water. This is especially important as the amount of water present in the system influences the representation of other key processes, such as extreme weather events, temperature rise, or the conditions for vegetation activity in different regions.

In their 2025 study, Rackow et al. found that in the IFS, the total amount of water does not remain constant over time. Instead, the model accumulates an increasing amount of water as simulations progress. The researchers traced back this issue of water non-conservation to the use of Semi-Lagrangian advection schemes. These schemes are implemented in the IFS to reduce computational costs when simulating atmospheric transport processes. However, they only approximate how air—and with it, water—is moved in the real world. As a result, the amount of water in the model after advection tends to be overestimated. This overestimation becomes more pronounced at higher spatial resolutions and when deep convective processes are not parameterized—that is, when they are resolved explicitly rather than represented through simplified equations based on numerous assumptions.

Rackow et al. (2025) addressed this problem of water non-conservation in high-resolution climate models by introducing a tool called global mass fixer. This tool corrects errors in the distribution of water and ensures that the total amount of water in the model is more realistically maintained at each time step. However, this approach also increases computational costs, and although it drastically improves water conservation on a global scale, some other imbalances still remain. To fully resolve these issues, the authors suggest the addition of a total energy fixer to address remaining inconsistencies in the model’s energy balance.

A more detailed explanation of the water non-conservation challenge is given by Tobias Becker and other members of the ECMWF in their 2022 newsletter article, available here.

2. Representing the top-of-the-atmosphere (TOA) radiation balance more realistically

The radiation balance refers to the difference between the energy the Earth receives from the sun and the energy it radiates back into space. This balance is crucial in climate modeling because it determines whether the Earth’s overall temperature is stable, warming, or cooling. In previous configurations, this balance was not accurately represented, leading to model drift and impacting the estimates of global warming from increased greenhouse gas concentrations in multi-year simulations.

To address this issue, the researchers made several adjustments focused on how clouds are represented and formed in the models:

1. Alteration of cloud edge erosion rates to slow down the clouds’ process of breaking down and to make them persist in the model over a longer period
2. Reduction of subgrid-scale cloud heterogeneity to slow down the process of accretion and therefore slightly increase cloud cover. 
3. Adjustment of processes connected to mid-level convection to detrain in the liquid form and not in the ice form, which reflects the real-world behavior more accurately. 
4. Simplification of how the model calculates the movement of moisture (semi-Lagrangian advection scheme) by switching from the complex method of cubic interpolation to a simpler one, namely linear interpolation. This change was applied to all forms of water except for water vapor, i.e., rain, snow, and ice. The linear interpolation is faster and uses less computing power while still being accurate enough for these calculations. Additionally, linear interpolation reduces the issue of water non-conservation before the global mass fixers are applied and increases the diffusion around updrafts.
5. Enhancement of high cloud representation by reducing the size of ice particles to better align it with observational data. High clouds are clouds that form at high altitudes in the atmosphere, typically above 6,000 meters (20,000 feet) and are made up of tiny ice crystals rather than water droplets. These types of clouds are likely to occur more often in storm-prone regions, which underscores the importance of this model change for future predictions.

3. Simulation of intense precipitation

Deep convection is a process by which clouds can form in the atmosphere and build up high enough to reach an altitude where temperatures are below 0°C. Previous km-scale model simulations at 4 and 9 km without parametrizing this process  (Deep OFF) led to unrealistically intense and disorganized convective systems. To fix these issues, the researchers of this study refrained from turning off the parametrized deep convection processes completely and instead reduced the amount of mass transported at the base of clouds by a factor of six compared to the default configuration at a 9 km resolution. This approach provides a more realistic transition between parametrized and explicitly resolved convection at the km-scale. Following this, the updated model can more accurately predict extreme weather events and how they change in a warming climate.

4. Enhancing the eddy-resolving features in the mid- and high-latitude oceans

Eddies are small, swirling currents in the ocean that can significantly impact weather and climate. To capture these features more accurately, the model was refined with a resolution finer than 5 km in large parts of the global ocean, which allowed it to represent mesoscale eddies and sea ice leads. In combination with the high atmospheric resolution, this higher resolution of the ocean processes led to a better understanding of the complex air-ice-ocean interactions, revealing new aspects of how these three components work together. One important difference was to have the model explicitly represent the shape and make-up of sea ice leads and to represent the resulting heating of the atmosphere. By including these interactions, the model now simulates novel energy exchange and climate processes, with potentially significant implications for understanding extreme weather events in polar regions and the effects of climate change on sea ice and ocean circulation.

5. Improving the representation of cities and urban areas

The urban scheme, developed at ECMWF, provides a simplified method to improve the simulation of weather conditions in cities. Tested over multiple years, the scheme has demonstrated its ability to enhance the simulation of near-ground temperatures, particularly in urban areas. When combined with higher-resolution land static information and integrated into the IFS model, it significantly increased the model’s accuracy in representing land surface temperatures over urban areas, capturing both temporal variability (changes over time) and spatial variability (differences between urban areas and their rural surroundings). The IFS-based nextGEMS simulations are the first coupled climate simulations including explicitly the effect of urban areas. However, some limitations persist, such as discrepancies in nighttime temperatures, which may result, among other causes, from an inaccurate representation of heat emissions resulting from human activities, such as the heating of buildings or the operation of vehicles at night.

In addition to these five technical improvements, the study has also shown that km-scale models can improve the representation of atmospheric circulation and extreme precipitation events, and revealed previously unknown interactions between the climate components, which will be the focus of future research projects.

Source: Rackow, T., Pedruzo-Bagazgoitia, X., Becker, T., Milinski, S., Sandu, I., Aguridan, R., Bechtold, P., Beyer, S., Bidlot, J., Boussetta, S., Deconinck, W., Diamantakis, M., Dueben, P., Dutra, E., Forbes, R., Ghosh, R., Goessling, H. F., Hadade, I., Hegewald, J., Jung, T., Keeley, S., Kluft, L., Koldunov, N., Koldunov, A., Kölling, T., Kousal, J., Kühnlein, C., Maciel, P., Mogensen, K., Quintino, T., Polichtchouk, I., Reuter, B., Sármány, D., Scholz, P., Sidorenko, D., Streffing, J., Sützl, B., Takasuka, D., Tietsche, S., Valentini, M., Vannière, B., Wedi, N., Zampieri, L., and Ziemen, F. (2025).Multi-year simulations at kilometre scale with the Integrated Forecasting System coupled to FESOM2.5 and NEMOv3.4, Geosci. Model Dev., 18, 33–69, DOI: 10.5194/gmd-18-33-2025.