Unleashing the Power of Diverse Climate Data for generating new knowledge and better Predictions
In the complex world of climate science, data are a cornerstone. Over the years, the European Union has invested heavily in producing vast amounts of climate-related data from both observations and model simulations. Despite the availability of these valuable datasets, many remain underutilised. Theme 1 of the EXPECT project aims to change that by combining these different data sources to generate new climate knowledge and improve predictive models.
Why Does it Matter?
Understanding climate processes and their impacts is essential for responding effectively to climate change. However, three major challenges hinder our progress. First, the data we have—especially as we go back in time towards the early 20th century—is incomplete, which limits our ability to fully grasp past climate extremes and then to apply this new knowledge from understanding past extremes for predicting the future. Second, current climate models often lack the accuracy and high resolution needed to capture fine-scale climate dynamics and processes. These models struggle to precisely simulate complex interactions within the Earth’s climate system, particularly at regional levels. Third, climate predictions are limited by incomplete knowledge of the initial states and forcing data (e.g. aerosol emissions). By addressing these issues—filling in historical data gaps, improving model precision and data for initializing and forcing the climate model predictions—we can place recent climate events, such as droughts, into a longer and more detailed historical context and increase the accuracy of future climate predictions. This approach provides scientists and policymakers with a better understanding of the drivers behind these extremes and helps predict future climate changes with greater accuracy.
Innovative Approaches to Data Analysis
EXPECT is pioneering a range of innovative methods to make the most of existing data. One key approach involves using machine learning techniques to fill in missing observational data, particularly focusing on hot and dry climate extremes. By training machine learning models with historical and new datasets, EXPECT aims to reconstruct past climate conditions more accurately. This will allow us to study the drivers of extreme weather events, providing insights that could lead to breakthroughs in climate science.
Another crucial aspect of Theme 1 is the use of data assimilation techniques to improve the accuracy of climate models. Data assimilation combines observational data with model simulations to produce more reliable predictions. EXPECT plans to apply novel data assimilation methods to use satellite-based soil moisture data to improve the representation of soil model parameters, which affect the simulation of hot and dry extremes. By improving the parameters used in climate models, EXPECT will help produce more accurate and useful predictions of future climate conditions.
EXPECT will further make use of AI to emulate high-resolution climate simulations, which traditionally demand massive computing power. Although these high-resolution models were designed to have an unprecedented process realism, they are so resource-intensive that the number of simulations that can be run is limited. By creating sophisticated AI-based replicas of these complex simulations, EXPECT aims to produce a multitude of virtual climate scenarios at a fraction of the computational cost and computing time.
This AI-driven method promises to significantly enhance our ability to quantify uncertainties in climate predictions, offering a more robust understanding of potential climate scenarios.