Improving weather and climate forecasting – Why collaboration between domain scientists and High-Performance Computing experts is key
Weather and climate forecasts are vital tools to help save lives and to protect property and the natural environment. Indeed, extreme weather events such as storms, floods and extreme temperatures have a huge impact on communities all over the world and result in significant financial costs, too – the European Environment Agency reported annual losses of around 13 billion euros on average between 2010 and 2017 in EEA member countries. Close cooperation between earth systems scientists and High Performance Computing (HPC) experts is the key to achieve more accurate and reliable prediction capabilities which can help prevent these disasters and inform decision-making to better prepare us for an uncertain climate future.
Why HPC is critical for weather and climate numerical simulation
Since knowledge of the weather has always been crucial for human affairs, its forecasting has a very long history – from the observation of cloud patterns to current numerical weather prediction and climate models. These models require an enormous number of computations and produce massive volumes of data, especially when simulating the dynamics of the Earth System globally – including atmosphere, ocean, sea ice, land ice and vegetation, along with their interactions. In addition, to enhance the accuracy and the reliability of predictions, the resolution of simulations must be increased up to the kilometer scale, with the effect that models become more complex as more physical phenomena must be considered. This makes climate modelling one of the most demanding supercomputing applications in terms of computing power and data handling requirements.
Challenges
Successful weather and climate prediction poses a perpetual computing challenge: it requires the co-designing of Earth System models and dedicated HPC systems. This means profiling, analyzing and optimizing the implementation of current models, but also anticipating the future development of new mathematical methods and numerical algorithms, together with their deployment in future supercomputers. This is why collaboration between domain scientists and performance-focused HPC experts is critical.
Europe is moving forward
The weather and climate community is well organized and structured to meet this challenge. In the EU, there are interesting collaborative projects that are paving the way to weather and climate at exascale.
The Center of Excellence in Simulation of Weather and Climate in Europe (ESiWACE), funded by the EU's Horizon 2020 program, brings together climate modeling groups, HPC centers and partners from the HPC industry to improve all aspects of the weather and climate modeling workflow.
For example, it provides modeling teams with HPC development services and fosters collaboration between technical experts.
Collaboration fosters the exchange of expertise, enabling researchers and industry to be more productive and paving the way to scientific excellence.
ESCAPE projects (standing for Energy-Efficient Scalable Algorithms for Weather and Climate Prediction at Exascale) aim to develop world-class, extreme-scale computing capabilities for European operational numerical weather and climate prediction systems.
As a further step, Destination Earth (DestinE) will enable the monitoring and simulation of natural environments and human activity by modelling Earth “digital twins”, including the development of test scenarios.
Collaboration is key
Collaboration fosters the exchange of expertise, enabling researchers and industry to be more productive and paving the way to scientific excellence.
Results from collaborative projects are promising. Good examples include work published by ESCAPE, the ESiWACE consortium and, more recently, the ESiWACE HPC services initiative. Atos’ experts have been cooperating actively with organizations including DKRZ and Météo-France, both of which have redoubled their focus in this area by acquiring new systems from Atos: the successor of Mistral at DKRZ and the Taranis/Belenos successors of Beaufix/Prolix at Météo-France, boosting the supercomputing capabilities of these two centers by five. Another example comes from the Center of Excellence in Weather & Climate Modelling recently established by Atos and ECMWF, which is supporting researchers with HPC, AI and quantum capabilities.
These examples demonstrate that when experts work together, outcomes can be uniquely valuable – enabling researchers and industry to be more productive and ultimately bolstering scientific excellence.