Moving towards the end of 2021, more and more of EUCP’s projects are coming to fruition. It’s great to start seeing the impact of our work on climate science and how useful it can be for our users. As EUCP starts drawing to a close, we will continue refining these products, working alongside our key partners and producing truly useful, actionable climate information. In this newsletter, we’ll take a look at some of the impact that EUCP has had in the climate science community, and at some of our latest products.

August 2021 saw the release of the Sixth Assessment Report from the Intergovernmental Panel on Climate Change. This landmark report represents the state of the art in our understanding of the climate system and how it is changing.

A great deal of EUCP research has contributed to the AR6 report, which also played a major role in the run-up to the COP26 UN climate conference in November. Here, we’re going to take a look at a snapshot of some of this key research that made its way into AR6. Additional details on these papers and more can be found in the Publications section of our website.

EUCP has made important advancements in the use of high-resolution climate models that can simulate atmospheric convection. Nikolina Ban and colleagues used the first multi-model ensemble of these high-resolution climate simulations to show they provide more realistic, finer spatial details and better represent hourly precipitation and the associated daily cycle. They also better simulate summer and autumn heavy rainfall. A paper led by Emanuela Pichelli then used this ensemble to study future climate, showing that high resolution models project even fewer, and even more intense, Alpine rainfall events by 2100 compared to lower resolution models.

High-resolution models were also used by Steven Chan and colleagues, who found they project greater changes in European extreme rainfall than their lower-resolution counterparts.

On the more technical side, the AR6 report featured research led by Daniel Befort on how to combine climate information from decadal predictions and uninitialised projections to improve prediction skill beyond 10 years, part of EUCP Work Package 5. Members of the team also described promising results for improving our climate projections by calibrating them using observations.

Studies led by Lukas Brunner looked at different ways of constraining climate model projections, a key part of our work in EUCP, and at how weighting climate models by their performance and similarity to each other gives more skilful climate change predictions.

EUCP research also covers how the climate may change in the future and how we can use our models best. Andrew Schurer and colleagues showed that human influence on the climate could strengthen the contrast between wet and dry regions in the tropics, while a paper led by Panos Athanasiadis described factors that can influence North Atlantic weather patterns.

Leo Borchert and colleagues found that our latest climate models have become much better at predicting future North Atlantic temperatures. This will help predict how extreme weather events may change in the future.

Finally, two papers led by Doug Smith are among those cited in AR6. Doug and colleagues found that climate a decade in advance is more predictable than previously thought, and that the climate of the North Atlantic can be made far more predictable while still using our current climate models.

This is only a small selection of the research from EUCP that has helped make the Sixth Assessment Report happen. This is also, of course, not to forget our colleagues who helped write AR6 directly, working with dozens of other authors around the world. Their work will help inform climate policies and negotiations for years to come.

In EUCP three different climate modeling systems are used to project future climate change: global climate models (GCMs), regional climate models (RCMs), and convection permitting models (CPMs). These systems differ mainly in terms of resolution, ranging from a grid spacing of ~100 km in the latest generation GCMs, ~10 km in RCMs, and ~2 km in CPMs. While GCMs and RCMs basically share the same physics and dynamics – usually optimized for different resolutions – CPMs are fundamentally different in terms of their small-scale dynamics, which allows them to directly simulate atmospheric convection without needing to be supplied with strongly simplified convection schemes.

One may question the benefits of the high-resolution climate modeling systems compared to the low-resolution systems. Added value is expected from better resolving small-scale features related to geography – topography, small lakes, cities, surface differences, land-sea borders – and also small-scale atmospheric phenomena, like convective showers and wind gusts. While it is relatively straightforward to show “added value” for the present-day climate, it is not so obvious to “prove” that high resolution models are better at future changes. Various efforts in EUCP Task 5.4 aim to quantify the “added value” of the high-resolution systems. ICTP, for instance, compares present-day climate with high-resolution observation datasets for high- and low-resolution modeling systems, establishing whether high-resolution results have better (extreme) statistics compared to low resolution model results. They can further evaluate whether future changes are systematically different between the modeling systems. Other teams focus on statistics more directly related to physical processes, the simulation of the diurnal cycle of convective rain (ETH) , temperature/humidity dependencies of extreme precipitation (KNMI), and specific geographically related features such as snowfall on mountains (SMHI).

RCMs are embedded in the output of GCMs, and typically can only be run for a (often rather small) sub-selection of the available GCMs (see Fig. 1 for a hypothetical example). This implies that not all possible large-scale future climate states (as simulated by the GCMs) are covered with the current set of RCMs. This could introduce a bias in the climate response in the RCMs with respect to the GCMs as well as “unexplored futures” that may be important from a certain user risk perspective. With international coordinated actions – like the EURO-CORDEX initiative – a large set of RCMs has been created in recent years, mitigating these problems to a considerable extent.

This is far more problematic for CPMs, which are computationally very demanding and therefore can only simulate rather short time slices of typically 10-30 years. Despite large efforts in Work Package 3 of EUCP to produce these simulations, they are still a so-called ensemble of opportunity, and no effort has been made to systematically explore the GCM uncertainty range. This limits the straightforward applicability of CPM results, and careful analysis has to be made to extract the useful information. For instance, though changes in convective rain statistics on a specific location are dominated by large uncertainty due to random climate noise, robust changes can be obtained by spatially pooling larger areas.

The other main aim of Task 5.4 is to investigate how to deal with the fact that we have only a relatively small set of downscaling experiments, given the wealth of GCM simulations available in CMIP5 and CMIP6. Methods are being developed that aim to statistically fill in, or emulate, these gaps in the matrix (red blocks in Fig. 1). This work is still very experimental: uncertainties in the appropriate large-scale predictors and how to combine them using statistical methods are still abundant. Methods range from very simple interpolation methods such as pattern scaling (DMI/NBI) to rather advanced methods using neural network machine learning techniques based on various large-scale predictors (CNRM).

Given the diversity of these approaches – in methodology, input, and what they can emulate – a simple common framework to test and compare them has been designed. In this framework a mid-century climate is emulated using information from the present-day and an end-century climate scenario (RCP8.5), using three models: CNRM-CM5 r1i1p1@150km (GCM), CNRM-ALADIN63@12km (RCM) and CNRM-AROME41t1@2.5km (CPM). One preliminary outcome appears to be that there is no one-size-fits-all solution, and more advanced methods do not necessarily provide more realistic outcomes for all evaluation measures. Yet, advanced methods such as machine learning clearly have a better range of applicability, for instance by providing daily time series with realistic variability (which cannot be done by most simple methods). Besides comparing different methods, another aim of the common framework is to develop simple statistical measures of the quality of emulated climate. This work is in progress and more outcomes are expected later in 2021.

Figure 1. Schematic of a set of GCM/RCM simulations called “the matrix” (not based on real data) showing the response in temperature and precipitation in a system with 6 GCMs (circles, 1-6) and 5 RCMs (triangles, a-e). The matrix shows the available simulations in green, whereas missing simulations are in red. Note that this figure only shows the concept, and that there are many more GCM runs, but also downscaled experiments available (e.g. >70 in the EURO-CORDEX set). Also the “emptiness” of the matrix varies considerably per region and emission scenario.

One of the ways EUCP is providing guidance to the scientific community is through research into the concept of climate storylines and their applications. This guidance comes in the form of scientific publications on specific methods or applications related to the concept of storylines, and two deliverable reports, which bring together findings from several pieces of work across the project. We demonstrate that storylines can be potentially useful as a tool for; i) framing and communicating complex information, ii) simplifying/navigating large climate model ensembles for a particular purpose, and iii) investigating specific types of events and impacts.

What do we mean by storylines?

The term ‘storylines’ is gaining popularity within climate science and services discourse, with several attempts to define and categorise what this means and motivate the approach. The concept also features in the IPCC AR6 Working Group I report.

In EUCP we refer to ‘climate storylines’, and broadly define them as narrative-based portrayals of past, current or future weather and climate which aim to improve the usefulness or useability of climate information and data for a range of applications. The term can also be used to refer to steps in the analysis process followed when sub-selecting the most appropriate climate models for further analysis. This can be necessary to minimise the number of simulations required when running high resolution climate models or performing impact modelling.

What are the outcomes and benefits for the user and scientific communities?

Despite the growing body of academic literature, there is a lack of real-world examples showcasing their enhanced utility and the role of cutting-edge climate science, such as the research that EUCP is generating. Ongoing work investigates the potential of storylines to meet the EUCP objective of delivering ‘consistent, authoritative and actionable climate information’.

In September 2021, EUCP scientists participated in a workshop to present and discuss the range of scientific research and applications related to storylines across the project with the aim of finalising the content of the related deliverable in WP5. This report will include case studies of storylines co-produced and assessed with users from the water resources and heritage management sectors (see Figure 2), and examples of storylines-related work from across the project with a reflection on their potential applications (such as the recently published storylines of the 2018 European drought). Results from these studies and activities will continue to be published beyond the end of the project.

Figure 2. a): The engagement process followed to produce trial storylines which explore uncertainty for planning and decision applications, b): one page from an example product for the heritage management sector and c): feedback gathered from a focus group with sector representatives from around Europe.

Projections of northern hemisphere extratropical climate underestimate internal variability and associated uncertainty, O’Reilly et al., Sep 2021, communications earth & environment

Natural variability in the climate has a strong influence on climate trends, alongside anthropogenic forcings such as rising levels of greenhouse gases. Representing the internal variability of the climate system in models is key to producing reliable projections, particularly regional projections. However, studies have shown that climate models do not simulate this variability accurately and especially underestimate the fluctuations occurring on decadal timescales. This paper compensates for the missing variability by creating synthetic projections with variability consistent with the large-scale observed atmospheric circulation. Using these synthetic ensemble data, the team found increased uncertainty in the projected 21st century temperature and rainfall changes over the northern extratropics, particularly over Europe. These are also associated with increased probabilities of extreme seasons in the future. These findings will inform climate policies aimed at helping communities mitigate or adapt to the future impacts that climate change may bring at the regional scale.

Increased risk of near term global warming due to a recent AMOC weakening, Bonnet et al., Oct 2021, Nature Communications

Climate sensitivity is a key part of climate model projections, helping define how much warming at the global scale we might expect to see in future. The latest models, dubbed CMIP6, generally show higher climate sensitivity than CMIP5, raising concerns about how to remain below key thresholds defined by the Paris Agreement. This study investigates the internal variability from a few large ensembles of simulations. In one of the models, the team found that the simulations that best matched observed temperatures in the historical period also had significant weakening of Atlantic Ocean currents due to internal multi-centennial variability. This may mask global warming and lead us to underestimate climate sensitivity when using recent observed global warming to constrain climate projections. These findings are important for informing the latest climate policies aimed at mitigating and adapting to the impacts of future climate.

Spatial extent of precipitation events: when big is getting bigger, Matte et al., Oct 2021, Climate Dynamics

Extreme rainfall is one of the most damaging climate hazards, and it is expected to increase in frequency and intensity as the climate continues to warm. How extreme rainfall events may change in size, however, is not as well understood. This study addresses this, using an ensemble of climate simulations to look at how the size of these systems might change over Europe at 1°C, 2°C and 3°C of warming. The team found that the spatial extent of areas associated with the most intensive daily rainfall events are set to become larger and more intense, particularly the most intense, core areas of these systems. Smaller systems, however, become less frequent, reflecting a shift towards larger, more intense rainfall systems across Europe. A better understanding of these future changes is a key part of informing future climate policies aimed at limiting the impact of extreme rainfall events and protecting communities and infrastructure.

On November 24th 2021, we held the latest instalment in our series of workshops with our Multi-User Forum. This forum gathers climate information users and lets them talk directly with climate scientists. Among the attendees were representatives from many different sectors including agriculture, water supply and urban planning. This collaboration helps users shape the climate prediction system at the core of EUCP, ensuring it meets their needs and is as useful and usable as possible.

The goal of this workshop was to present new EUCP products to users and discuss their usability. Almost 50 participants joined in learning about EUCP breakthroughs, including improved decadal climate predictions, new methods for merging climate projections on different timescales, and the first pan-European ensemble of high-resolution convection-permitting climate models (CPMs). The latter are particularly useful for simulating extreme climatic events.

New advances in EUCP’s climate science and data infrastructure were presented in the form of ‘storyboards’, each illustrating to users the benefits of a key result or product from the EUCP project. For example, one storyboard covers work led by Marjanne Zander on convection-permitting models, like the ones mentioned above. The team used these models to simulate flash flooding in the Alps under a warmer future climate, finding that although the frequency of these events is set to fall in the summer months, extremes will be more severe. Autumn extremes will also be more severe, but here the frequency is set to increase as well.

Delegates had the opportunity to pose questions directly to the scientists involved in each storyboard on their work, its usability, and further developments. This yielded key insights into different users’ perspectives, such as how (un)comfortable they are with technical language and complex datasets. The discussions also highlighted how important it is to merge information from multiple lines of evidence into a single, coherent output.

Within small breakout groups there were discussions about the EUCP products from the perspectives of business and adaptation strategies, raising the need to align the science with the required output, and users’ desire for concise explanations of the main parts of the EUCP legacy. Users found the storyboards helpful and made a number of suggestions to help EUCP develop them further to widen their audience. This will prove very useful in refining EUCP’s web platform for storyboards ahead of its full release. In the meantime, more feedback is always welcome.

We are already looking forward to our third Multi-User Forum workshop next year, when we will see the benefits of these insights and suggestions.

Continuing our series of project assemblies, we held a Mini-Assembly on the morning of November 26^th 2021. This gave us all the opportunity to share our latest highlights and discuss the future of the EUCP project. More than 70 delegates joined in the virtual session.

Prof. Jason Lowe of the Met Office opened the meeting by highlighting some of our recent successes, including:

●      explanatory storyboards in development on our science and its key benefits.

●      the number of excellent recent papers we have published in the scientific literature (find out more on our Publications page).

●      our recent communications, including our involvement at COP26, our latest Newsletter, and a video we have helped produce on Europe’s climate in 2050, which has already attracted over 100,000 views on YouTube!

Work Package 1 opened the next part of the meeting, highlighting their work on determining the impact of using climate models of different resolutions on simulations of North Atlantic temperatures, and the effect of volcanic eruptions on the climate. Work Package 2 have been working on comparing different ways of weighting climate simulation ensemble members in order to emphasise the ones most consistent with observations. This work is particularly relevant to important IPCC messages on the plausible range of global climate sensitivity to regional projections. WP2 are also developing an atlas which will display the results of several different constraint methods, as well as unconstrained simulations, helping users decide which is best for them.

Work Package 3 told us about the latest progress on their new, very high-resolution projections for European domains, generated using convection-permitting regional models. These simulations offer the first multi-model projections of this kind, allowing us for the first time to assess the uncertainty in the very high-resolution characteristics of these projections, such as extreme sub-hourly rainfall. These simulations are also useful to other teams, including Work Package 4 They described their work applying these new projections to assess flood risk in the Alpine region, as well as other case studies of the risks of drought and storm surge across Europe and the Caribbean. Making these types of projections more usable by those who need them has also been a key focus of this team.

Work Package 5’s work has been all about merging climate projections on different spatial and temporal scales. They have been looking at how simulations of extreme rainfall events scale at different model resolutions, and at how the complexity of a high-resolution model can be replicated in a low-resolution simulation using a neural net-based emulation system. Finally, Work Package 6 gave us all a great overview of the recent Multi-User Forum workshop, in which climate information users had the opportunity to hear directly from scientists about their work and how it could be useful to them in their sector. You can read more about this event here.

Following these science highlights, Assembly delegates had the opportunity to discuss new ideas and initiatives for exploiting these advances. We discussed how we can make the data and science products from the project easy to find and use, as well as how we can continue to engage our users to co-develop products that meet their needs. We had a lively discussion around the development of climate services on a decadal timescale, an extremely collaborative programme of work.

To close the meeting, we devoted time to discussing the closing stages of the EUCP project. As we near the final six months of the project, the legacy of work that will be produced is of great importance and ensuring that it is used to its fullest will be an important task. We also began planning for our final event to close the project, and share and showcase the project’s important developments. Keep looking for further announcements in our Newsletter and social media feeds.