A meteorologist from the Karlsruhe Institute of Technology (KIT) has received a starting grant from the European Research Council (ERC) to pioneer a method for more accurate and energy-efficient weather forecasts.
The ASPIRE project, led by Dr Julian Quinting from KIT’s Institute of Meteorology and Climate Research, aims to optimise weather forecasts and reduce their computational efforts to save costs and energy. The project will employ recurring signals in the tropical Pacific that have an essential influence on atmospheric circulation in Europe, developing machine learning models to replicate the effects of a high resolution.
How can accurate weather forecasts save energy?
Due to the impacts of climate change and the ongoing energy crisis, reliable weather forecasts for a time frame between two weeks and two months are becoming increasingly essential. This is because predicting what the temperatures will be in the coming weeks can be vital information for estimating a building’s heating requirements and filling gas storage.
Moreover, accurately predicting extreme weather events, such as heat waves, droughts, or floods, can help the public and authorities prepare accordingly, providing enough time to mitigate potential damage.
What will the ASPIE project involve?
The Advancing Subseasonal PredIctions at Reduced computational Effort (ASPIRE) project is advancing weather forecasts in this timeframe, called subseasonal forecasts. Now backed by the fiscal power of the ERC starting grant – which provides up to €1.5m annually for up to five years – Dr Quinting will have the resources to develop precision forecasts, decreasing computing effort, which will reduce costs, energy, and greenhouse gas emissions.
Dr Quinting commented: “My underlying idea is to make more use of sources in the atmospheric system with high intrinsic predictability. These sources are, for instance, recurring patterns in the atmosphere that vary on the time scale of two weeks to two months.”
Quinting believes that recurring signals in the tropical Pacific hold the potential for predicting atmospheric circulation in Europe. However, these tropical signals are underrepresented in current numerical weather prediction models, which prevents the full utilisation of intrinsic predictability.
ASPIRE aims to enhance the representation of tropical signals using a high spatial resolution in the tropics. A high resolution usually requires more computing power. To combat this, Quinting will develop machine learning models that imitate the effects of a high resolution to reduce computing effort.
If the method is successful, it could be used for climate research on other aspects of the atmospheric system that have a high intrinsic predictability and are incorrectly represented in weather prediction models.
Quinting concluded: “With ASPIRE, we want to showcase the potential of simulations with locally high spatial resolution. Ideally, weather services will be able to utilise existing computing power even better.”