Natural disasters have been a constant threat for humankind – think of earth quakes, volcano eruptions or tsunamis. In the age of rapid climate change, some kinds of hazards are getting both more frequent and more violent: Forest fires, heavy rains, storms and flooding, poisoning air or river pollution, or heat waves and droughts.

We can’t directly influence or stop any of those event when they happen – but it turns out that simulations backed by high performance computers can be quite helpful in devising emergency action, predict events and issue timely warnings, inform those in charge on how to best prepare to minimize negative impact, or – in some cases – how to reduce the risk of events occurring altogether: Action on these risks is one of the EU’s priorities.

Fig 1: In Autumn 2021, the eruption of Cumbre Vieja on La Palma caused lots of volcanic ashes being emitted. In an urgent computing effort, CoE ChEESE helped the local authorities by forecasting the distribution of ashes and thus supporting decisions on necessary measures to protect the population. Image (C) ESA under CC BY-SA 3.0 IGO

One example we all know and use every day – maybe without even realizing that HPC is involved –  are weather forecasts: Here meteorology is used  since decades and does good service in predicting not only normal weather but also warns about storms, heavy rainfall or extreme heat. But there’s more in the toolbox than is commonly known and used.

One striking example in the last few years is the devastating flooding in the Ahr valley in Germany in Summer 2021: Heavy rains in a narrow valley caused an enormous flooding. While the rain was predicted, the actual amount and effect of the flooding wasn’t, resulting in about 150 casualties – though in principle all tools were available to predict this event and evacuate people.

Fig 2: Damages in the Ahr valley, Germany, summer 2021: Over 150 people died during a nightly catastrophic flooding.  While the heavy rains causing the flooding were forecast correctly, the flooding itself was not – although capable simulation tools existed. Thus, local authorities failed to issue a warning and start evacuation.  Image (C) Phillip Reuther

And there’s much more in the toolbox:

• CoE ChEESE-2P develops applications to predict Tsunami waves, lava flow, ash dispersion of volcano eruption, and other geophysical modeling tools
• Applications of CoE HiDALGO2 can predict urban air pollution, heat and local wind effects, spread of wildfires and smoke, and pollution transport in rivers
• The ESiWACE-3 CoE has a range of high-resolution climate models in its portfolio that can help to predict regional likelihood of extreme weather in the long term and thus inform on adaptation strategies (in addition to globally limiting climate change to a minimum through appropriate measures).

Many actual  use cases underline the usefulness of those applications:  Analysing air pollution in Stockholm (a project initiated with Nation Competence Center (NCC) Sweden (ENCCS) and taken a step further with CoE HiDALGO2), or advising local authorities on La Palma during the eruption of the volcano Cumbre Vieja (CoE ChEESE). Further examples can be found in the compilations of use cases and success stories from the NCCs, for instance on using AI & satellite data for improved local weather forecasts, or building a database of climate-induced severe weather events for insurance companies. A recent workshop on natural hazard simulations  organized by ChEESE-2P highlighted many more examples.

Yet, there is no systematic support for using these tools routinely and generally, as it is standard for weather prediction.  Doing so would require more than just having the software codes sitting somewhere – as most of these events arise without much prior notice, we need to be able to instantly fire up a simulation programs (“urgent computing”) , with direct access to the necessary data to set up the case-specific models for the simulation in an automated way.

CoEs and NCCs have teamed up to build on the resources already available in the network and move things forward. Firstly, to raise awareness of the potentials of these approaches, and to reach out to potential users, materials will be compiled like collections of use cases and success stories. A systematic overview of the resources available will be created, centred around the simulation codes and related mature application services, including up-to-date references to related training opportunities and documentation. For the larger picture, a white paper is planned describing the state, the opportunities but also identifying barriers for uptake and work needed for removing them.