Displays

CL3.3

Information on the future climate is an essential basis for managing the risks, as well as potential opportunities, arising from a changing climate. Typically, this information comes from state-of-the-art numerical simulations of the climate in the form of climate predictions and climate projections. For many decision-makers and policymakers the information available from climate simulations is not at the appropriate spatial and temporal scales they need to form the basis for their climate-related risk assessments or for climate action plans. Also, some decision-makers require information that spans a range of time scales from a few months or a year ahead to decades into the future. Observational and emerging constraints can help evaluate and possibly constrain model-based uncertainty ranges.

This session aims to cover the advances in providing usable and reliable climate information for Europe over the next 40 or so years. It welcomes, without being restricted to, presentations on:

• Improved methods to quantify and understand uncertainty in climate predictions and projections for Europe. This could be on spatial scales from convectively resolving to global.
• Processes which bridge time scales from beyond a season to multiple decades and methodologies to blend the output from initialised predictions and non-initialised projections
• Demonstration of added value of initialised vs non-initialised near-term climate predictions and projections using innovative verification tools
• Illustration of the value of such climate information system through applications

The session will bring together research scientists and users from a range of projects including EUCP and national initiatives with the aim of sharing experiences, novel results and initiating discussions on this emerging topic.

Solicited speakers:
David Sexton (Met Office)
James Murphy (Met Office)
Carlo Buontempo (Copernicus Climate Change Service C3S)

Public information:
For a related Zoom session in parallel to the chat see: https://www.eucp-project.eu/eucp-updates/join-the-eucp-related-egu-session-bringing-together-future-climate-predictions-and-projections-for-europe/
https://zoom.us/j/94246127256?pwd=dWJpV1c4amtkWDhWMndFOW5hNk1sZz09
Meeting ID: 942 4612 7256
Password: 854513

Share:
Convener: Jason A. Lowe | Co-conveners: Daniel BefortECSECS, Christopher O'ReillyECSECS, Albrecht Weerts, Antje Weisheimer
Displays
| Tue, 05 May, 16:15–18:00 (CEST)

Files for download

Download all presentations (76MB)

Chat time: Tuesday, 5 May 2020, 16:15–18:00

Chairperson: Chris O'Reilly
D3447 |
EGU2020-7297
David Sexton, Jason Lowe, James Murphy, Glen Harris, Elizabeth Kendon, Fai Fung, Carol McSweeney, John Rostron, Kuniko Yamazaki, Hazel Thornton, Giorgia Fosser, Simon Tucker, and Philip Bett

UK Climate Projections 2018 (UKCP18) included land and marine projections and were published in 2018 to replace UKCP09. The land projections had three components, and all were designed to provide more information on future weather compared to UKCP09. The first component updated the UKCP09 probabilistic projections by including newer CMIP5 data and focussing on seasonal means from individual years rather than 30-year averages. The probabilistic projections represent the wider uncertainty. The second two components (global and regional projections) both had the aim of providing plausible examples of future climate, but at different resolutions.

The global projections were a combination of 13 CMIP5 models and a 15-member perturbed parameter ensemble (PPE) of coupled simulations for 1900-2100 using CMIP5 RCP8.5 from 2005 onwards. The PPE was provided at 60km atmosphere, quarter degree ocean and the large-scale conditions from twelve of the members were used to drive the regional model at both 12km and 2.2km resolution. These plausible examples are more useful for providing information about weather in a future climate to support a storyline approach for decision making.

The talk will present examples of new ways to use UKCP18 compared to UKCP09.  We will show how the global projections can be used to understand that the recent record winter daily maximum temperature in the UK in 2019 had a large contribution from internal variability and what this means for breaking the record in a warming climate. We also use an example from China to demonstrate one way to exploit information at different time scales, looking at how a circulation index, which is predictable and related to tropical cyclone landfall, changes in a future climate.

Finally, we show that while the enhanced resolution of the global and regional projections has improved our capability to provide climate information linked to the better representation of circulation, they lack diversity in some of the key drivers of future climate. Therefore, a key way forward will be to find an appropriate balance between the need for better diversity (e.g. multiple ensembles such as CMIP or PPEs) and the need for an appropriate resolution to retain this new capability.

D3448 |
EGU2020-7313
James Murphy

The challenge of combining initialised and uninitialised decadal projections

James Murphy, Robin Clark, Nick Dunstone, Glen Harris, Leon Hermanson and Doug Smith

During the past 10 years or so, exploratory work in initialised decadal climate prediction, using global climate models started from recent analyses of observations, has grown into a coordinated international programme that contributes to IPCC assessments. At the same time, countries have continued to develop and update their national climate change scenarios.  These typically cover the full 21st century, including the initial decade that overlaps with the latest initialised forecasts. To date, however, national scenarios continue to be based exclusively on long-term (uninitialised) climate change simulations, with initialised information regarded as a separate stream of information.

We will use early results from the latest UK national scenarios (UKCP), and the latest CMIP6 initialised predictions, to illustrate the potential and challenges associated with the notion of combining both streams of information. This involves assessing the effects of initialisation on predictability and uncertainty (as indicated, for example, by the skill of ensemble-mean forecasts and the spread amongst constituent ensemble members). Here, a particular challenge involves interpretation of the “signal-to-noise” problem, in which ensemble-mean skill can sometimes be found which is larger than would be expected on the basis of the ensemble spread. In addition to initialisation, we will also emphasise the importance of understanding how the assessment of climate risks depends on other features of prediction system design, including the sampling of model uncertainties and the simulation of internal climate variability.

D3449 |
EGU2020-18610
Carlo Buontempo
Climate adaptation often requires high resolution information about the expected changes in the statistical distribution of user-relevant variables. Thanks to targeted national programmes, research projects and international climate service initiatives  this kind of information is not only becoming more easily available but it is also making its way into building codes, engineering standards as well as the risk assessments for financial products.  If such an increase in the use of climate data can be seen as a positive step towards the construction of a climate resilient society, it is also true that the inconsistencies that exist between the information derived from different sources of information, have the potential to reduce the user uptake, increase the costs of adaptation and even undermine the credibility of both climate services and the underpinning climate science.
This paper offers a personal reflection on the emerging user requirements in this field. The presenation also aims at suggesting  some prelimimary ideas in support of the development of appropriate methodologies for extracting robust evidence from different sources in a scalable way.
D3450 |
EGU2020-5777
| Highlight
Lukas Brunner, Carol McSweeney, Daniel Befort, Chris O'Reilly, Ben Booth, Glen Harris, Jason Lowe, Marianna Benassi, Erika Coppola, Rita Nogherotto, Gabriele Hegerl, Reto Knutti, Geert Lendrink, Hylke de Vries, Said Qasmi, Aurelien Ribes, and Sabine Undorf

Political decisions, adaptation planning, and impact assessments need reliable estimates of future climate change and related uncertainties. Different approaches to constrain, filter, or weight climate model simulations into probabilistic projections have been proposed to provide such estimates. Here six methods are applied to European climate projections using a consistent framework in order to allow a quantitative comparison.  Focus is given to summer temperature and precipitation change in three different spatial regimes in Europe in the period 2041-2060 relative to 1995-2014. The analysis draws on projections from several large initial condition ensembles, the CMIP5 multi-model ensemble, and perturbed physics ensembles, all using the high-emission scenario RCP8.5.  
The methods included are diverse in their approach to quantifying uncertainty, and include those which apply weighting schemes based on baseline performance and inter-model relationships, so-called ASK (Allen, Stott and Kettleborough) techniques which use optimal fingerprinting to scale the scale the response to external forcings, to those found in observations and Bayesian approaches to estimating probability distributions. Some of the key differences between methods are the uncertainties covered, the treatment of internal variability, and variables and regions used to inform the methods. In spite of these considerable methodological differences, the median projection from the multi-model methods agree on a statistically significant increase in temperature by mid-century by about 2.5°C in the European average. The estimates of spread, in contrast, differ substantially between methods. Part of this large difference in the spread reflects the fact that different methods attempt to capture different sources of uncertainty, and some are more comprehensive in this respect than others. This study, therefore, highlights the importance of providing clear context about how different methods affect the distribution of projections, particularly the in the upper and lower percentiles that are of interest to 'risk averse' stakeholders. Methods find less agreement in precipitation change with most methods indicating a slight increase in northern Europe and a drying in the central and Mediterranean regions, but with considerably different amplitudes. Further work is needed to understand how the underlying differences between methods lead to such diverse results for precipitation. 

D3451 |
EGU2020-2654
Leonard Borchert, Holger Pohlmann, Johanna Baehr, Nele-Charlotte Neddermann, Laura Suarez Gutierrez, and Wolfgang A. Müller

We use a decadal prediction system with the Coupled Model Intercomparison Project Phase 6 version of the coupled Max Planck Institute Earth System Model to predict the probability of occurrence for extremely warm summers in the Northern Hemisphere. An assimilation run with Max Planck Institute Earth System Model shows a robust response of summer temperature extremes in northern Europe and northeast Asia to North Atlantic sea surface temperature via a circumglobal Rossby wavetrain. When the North Atlantic is warm, warm summer temperature extremes occur with a probability of 20% and 24% in northern Europe and northeast Asia, respectively. In a cold North Atlantic phase, these probabilities are 0% and 8%. A similar dependence of the probability of occurrence for summer temperature extremes in these regions to North Atlantic SST can be found in observations.
To examine this effect in decadal predictions, we pool all available realizations for any given year in the decadal prediction system. Using 10 ensemble members and 10 lead years, we therefore end up with 100 realizations for any year between 1970 and 2018. We find that the probability of occurrence for summer temperature extremes in the pooled initialized climate predictions shows good agreement with the observations and the assimilation run. This agreement is related to high skill of the model system in predicting North Atlantic SST. Consequently, the likelihood of a warm summer temperature extreme occurring in the examined regions in the next 10 years can be inferred from predictions of North Atlantic temperature.

D3452 |
EGU2020-22378
| Highlight
Nikolina Ban, Erwan Brisson, Cécile Caillaud, Erika Coppola, Emanuela Pichelli, Stefan Sobolowski, Marianna Adinolfi, Bodo Ahrens, Antoinette Alias, Ivonne Anders, Sophie Bastin, Danijel Belusic, Ségolène Berthou, Rita Cardoso, Steven Chan, Ole Christensen, Jesus Fernandez, Lluis Fita, Thomas Frisius, and Klaus Goergen and the Erika Coppola

Here we present the first multi-model ensemble of climate simulations at kilometer-scale horizontal resolution over a decade long period. A total of 22 simulations, performed by 21 European research groups are analyzed. Six different regional climate models (RCMs) are represented in the ensemble. The simulations are compared against available high-resolution precipitation observations and coarse resolution (12 km) RCMs with parameterized convection. The model simulations and observations are compared with respect to mean precipitation, precipitation intensity and frequency, and heavy precipitation on daily and hourly timescales in different seasons.

The results show that kilometer-scale models produce more realistic representation of precipitation than the coarse resolution RCMs. The most significant improvements are found for heavy precipitation and precipitation frequency on both daily and hourly time scales in the summer season. In general, kilometer-scale models tend to produce more intense precipitation and reduced wet-hour frequency compared to coarse resolution models. Although differences between the model simulations at the kilometer-scale and observations exist, it is evident that they are superior to the coarse-resolution RCMs in the simulation of precipitation in the present-day climate, and thus offer a promising way forward for investigations of climate and climate change at local to regional scales.

D3453 |
EGU2020-19471
Sanne Muis, Maialen Irazoqui Apecechea, Job Dullaart, Joao de Lima Rego, Kristine S. Madsen, Jian Su, Kun Yan, and Martin Verlaan

Climate change will lead to increases in the flood risk in low-lying coastal areas. Understanding the magnitude and impact of such changes is vital to design adaptive strategies and create awareness. In  the  context  of  the  CoDEC  project  (Coastal  Dataset  for  Evaluation  of  Climate  impact),  we  developed a consistent European dataset of extreme sea levels, including climatic changes from 1979 to 2100. To simulate extreme sea levels, we apply the Global Tide and Surge Model v3.0 (GTSMv3.0), a 2D hydrodynamic model with global coverage. GTSM has a coastal resolution of 2.5 km globally and 1.25 km in Europe, and incorporates dynamic interactions between sea-level  rise,  tides  and  storm surges. Validation of the dataset shows a good performance with a mean bias of 0-.04 m for the 1 in 10-year water levels. When analyzing changes in extreme sea levels for the future climate scenarios, it is projected that by the end of the century the 1 in 10-year water levels are likely to increase up to 0.5 m. This change is largely driven by the increase in mean sea levels, although locally changes in storms surge and interaction with tides can amplify the impacts of sea-level rise with changes up to 0.2 m in the 1 in 10-year water level.

The CoDEC dataset will be made accessible through a web portal on Copernicus Climate Data Store (C3S). The dataset includes a set of Climate Impact Indicators (CII’s) and new tools designed to evaluate the impacts of climate change on different sectors and industries. This data service will support European coastal sectors to adapt to changes in sea levels associated with climate change. In this presentation we will also demonstrate how the C3S coastal service can be used to enhance the understanding of local climate impacts.

D3454 |
EGU2020-3027
Segolene Berthou, Elizabeth Kendon, Malcolm Roberts, and Benoit Vanniere

D3455 |
EGU2020-5511
Tamzin Palmer, Carol Mc Sweeney, and Ben Booth

An alternative approach to constraining climate projections based on a probabilistic approach with observational constraints, is to select a subset of models from the ensemble based on their ability to represent key physical processes, along with some indicators of model performance. The method that is presented here is based on the assumption that if a model is unable to reproduce the key factors important for determining the regional climate, the projections from this model are not considered reliable. The projection range for CMIP5 for the three EUCP European regions is assessed using two different subsampled model ensembles.

The first sub-sampling method presented uses the approach of Mc Sweeney et al. (2015), which assessed the models based on their performance for the UK climate. Each model in the CMIP5 ensemble (where data is available), is firstly assessed against these key performance indicators and poor performers eliminated from the selection. Several models also share large portions of code and therefore have similar errors and projections, Sanderson et al 2015a and 2015b quantifies these similarities. This analysis was used identify ‘near-neighbours’ and further reduce the selection. The applicability of a sub-selection of models based on their performance for the UK climate is assessed for the wider European area and found to reduce the projected range for the Northern European Area (NEU), for precipitation and near surface temperature considerably. The impact on the projected ranges for the Central European Area (CEU) and the Mediterranean (MED) was not as large, suggesting that a different set of physical processes are of primary importance for these regions.

To further investigate the effect of subsampling based on physical processes, a subset of CMIP5 models identified by the approach of Vogel et al. (2018) has been applied for the EUCP European areas. Vogel et al. (2018) looked at the ability of the CMIP5 models to reproduce the correlation between the hottest day of the year and precipitation within the same range as that found in the observations. This approach is designed to subsample the ensemble based on the ability of the model to represent soil moisture feedback processes with the atmosphere. It is thought that these processes are likely to be increasingly important for determining the projected climate in the CEU and MED regions.  

Finally, the projection range for the CMIP6 ensemble in the EUCP regions for precipitation and the near surface temperature will be presented and compared with those for CMIP5.

D3456 |
EGU2020-9183
Paolo Stocchi, Emanuela Pichelli, Erika Coppola, Jose Abraham Torres Alvarez, and Filippo Giorgi

The recent increase in climate modeling activities at convection permitting scales (grid spacing under 4 km) has strongly been motivated by the increased computer capacities in the last years with the aim to reduce the model errors associated with parameterized convection and a more detailed representation of present and future regional climate. Some Regional climate projects addressing on convection permitting modeling simulations and projections have been recently implemented to make more robust conclusions on the added value of convection permitting simulation to future climate projections. Here, we present convection resolving climate simulations performed in the framework the European Climate Prediction System (EUCP) project, using the non-hydrostatic version of the RegCM model. The RegCM simulations have a grid spacing of 3 km, over three different regions (Pan-Alpine, Central Europe, and South-East Europe). These simulations were driven by initial and boundary conditions built from intermediate 12 km simulations driven by the global climate model (GCM) HadGEM2-ES. We considered three time slices each one of them covering a 10-year period, the historical (1996-2005), the near future (2041-2050) and the far future (2090-2099) under the RCP8.5 scenario. The high resolutions (3 km) simulations, over the historical period, are evaluated through comparison with available observations data sets (including in-situ and satellite-based observation of precipitation) and coarse resolution (12 km) simulation is used as benchmark. The kilometer-scale RegCM4.7 scenario (RCP8.5) simulations, driven by HadGEM2-ES, near future (2041-2050) and the far future (2090-2099), are also analyzed and presented, focusing on the future change in terms of mean precipitation, precipitation intensity and frequency and heavy precipitation on daily and hourly timescales in different seasons.

D3457 |
EGU2020-9263
Frederiek Sperna Weiland, Pety Viguurs, Marjanne Zander, and Albrecht Weerts

Flash floods are a significant natural hazard in the Alpine region (FOEN, 2010). With changing rainfall regimes and decreased snow accumulation due to climate change, the risk of flash flood occurrence and timing thereof could change as well (Etchevers et al., 2002).

In this study the frequency and occurrence of flash floods in the Alpine region is estimated for current and future climate (RCP8.5) using state-of-the-art high-resolution convection permitting climate models (CP-RCMs). For the historical period and far future (2100), data from an ensemble of convection permitting climate models (Ban et al., submitted 2019) was used to drive a high-resolution distributed hydrological model, i.e. the wflow_sbm model (Imhoff et al., 2019, Verseveld et al., 2020). The model domains cover the mountainous parts of the Danube, Rhone, Rhine and Po located in the Alps.  The CP-RCM time-series available are of limited length due to computational constrains. At the same time the locations of flash floods vary per year therefore a regional scale analysis is made to assess whether in general the severity, frequency and timing of flash floods in the Alps will likely change under changing climate conditions.

This research is embedded in the EU H2020 project EUCP (EUropean Climate Prediction system) (https://www.eucp-project.eu/), which aims to support climate adaptation and mitigation decisions for the coming decades by developing a regional climate prediction and projection system based on high-resolution climate models for Europe.

References:

Etchevers, P., Golaz, C., Habets, F., and Noilhan, J., Impact of a climate change on the Rhone river catchment hydrology, J. Geophys. Res., 107( D16), doi:, 2002.

Federal office for the environment FOEN (2010) Environment Switzerland 2011, Bern and Neuchatel 2011. Retrieved from www.environment-stat.admin.ch

Imhoff, R.O., W. van Verseveld, B. van Osnabrugge, A.H. Weerts, 2019. Scaling point-scale pedotransfer functions parameter estimates for seamless large-domain high-resolution distributed hydrological modelling: An example for the Rhine river. Submitted to Water Resources Research, 2019.

N. Ban, E. Brisson, C. Caillaud, E. Coppola, E. Pichelli, S. Sobolowski, …, M.J. Zander (submitted 2019): “The first multi-model ensemble of regional climate simulations at the kilometer-scale resolution, Part I: Evaluation of precipitation”, manuscript submitted for publication.

D3458 |
EGU2020-10048
Rita Nogherotto, Paolo Stocchi, Erika Coppola, and Filippo Giorgi

The Reliability Ensemble Averaging (REA) method calculates average, uncertainty range and a measure of reliability of simulated regional climate changes from ensembles of different model simulations. The REA method is applied to mean seasonal temperature and precipitation changes in three different European spatial regimes in the period 2041-2060 and 2081-2100 relative to the reference period 1995-2014. Regional ensemble results of 55 scenario simulations for the RCP8.5 and RCP2.6 at 0.11 degree resolution over the common EURO-CORDEX domain, using 8 GCMs and 11 RCMs, are compared with the driving CMIP5 global models. For each region we show the median and the 25th-75th and 5th-95th percentile spreads of the weighted temperature and precipitation change. The spread of the changes (both 25th-75th and 5th-95th percentiles) are strongly reduced by the weightening as expected, while the best estimate changes (median) of the projection ranges varies according to the region and the season. The method is also applied to evaluate the reliability of the extreme precipitation simulations.

 

D3459 |
EGU2020-10149
Panagiotis Athanasiou, Ap van Dongeren, Alessio Giardino, Michalis Vousdoukas, Roshanka Ranasinghe, and Jaap Kwadijk

Climate change driven sea level rise (SLR) is expected to rise with even higher rates during the second half of the present century. This will exacerbate shoreline retreat of sandy coasts, which comprise one third of the global coastline. Sandy coasts have high touristic and ecological value while they are the first level of defense against storms, protecting valuable infrastructures and buildings. Therefore, in recent years, large scale risk assessments are considered useful tools for the guidance of policy makers to identify high risk hotspots.  Reliable input data at this scale are required in order to make useful estimations. Among others, crucial data to assess the impact of SLR on shoreline retreat are the detection of different coastal types and, in particular, of sandy erodible beaches, and the nearshore slope, which is usually assumed to be uniform.

The important issue of input data uncertainty and spatial variation and consequent impact on predictions has been so far ignored in most large-scale studies. Estimates of shoreline retreat are however very sensitive to the variation in these inputs. Here we quantify SLR driven potential shoreline retreat and consequent land loss in Europe during the 21st century by employing different combinations of geophysical datasets for (a) the location of sandy beaches and (b) their nearshore slopes. For the estimation of the shoreline retreat, the Bruun Rule is used, which offers a suitable approach for a first approximation of erosion impacts at large scales. Sea level rise projections associated with the moderate-emission- mitigation-policy (RCP4.5) and the high-end, business-as-usual scenario (RCP8.5) are used as boundary conditions. The location of sandy beaches is determined from two different datasets. One is based on manual visual estimation from satellite images and the other on automatic detection from satellite images using machine learning techniques. For nearshore slopes we apply the commonly used constant slope assumption of 1:100 and a newly produced global dataset which captures the spatial variation of coastal slopes.

With this approach, we create four different combinations for each SLR scenario, for which we estimate and compare land loss at EU, country and NUTS3 regional level. We find that the land loss estimations for each combination can differ significantly, especially at the regional and local level. At the European or country level, even though differences in total land loss projections can be significant, they can be concealed by the spatial aggregation of the results. Using data-based spatially-varying nearshore slope data, a European averaged median shoreline retreat of 97 m (54 m) is projected under RCP 8.5 (4.5) by year 2100, relative to the baseline year 2010. This retreat would translate to 2,500 km2 (1,400 km2) of land loss. A variance-based global sensitivity analysis indicates that the uncertainty associated with the choice of geophysical datasets can contribute up to 45% (26%) of the variance in land loss projections for Europe by 2050 (2100).

D3460 |
EGU2020-2618
Daniel J. Befort, Christopher H. O’Reilly, and Antje Weisheimer

There is an increasing demand for robust, reliable and actionable climate information for the near-term future (1-40 years). Seamless information on these time-scales can only be derived from uninitialized climate projections, which however are not aligned with the observed, internal state of the climate system. Another source of information are Initialized predictions for which the observed state is taken into account, but these are only available for the upcoming decade. 

In this study, we test in how far decadal predictions can be used to constrain uninitialized projections to obtain skillful predictions on time-scales beyond decades. This is done by selecting a sub-ensemble of uninitialized projections, which are chosen by their proximity to the decadal predictions ensemble mean. This framework is applied to surface air temperatures from CMIP5 simulations over the North Atlantic Gyre region, as decadal predictions show largest added value over projections for this region. Skill is measured using anomalous correlation coefficient (ACC) and root-mean-square-error (RMSE). Results show that ACC values for forecast years 10-15 for the constrained sub-ensemble are similar to those derived for the non-constrained uninitialized ensemble. However, RMSEs are significantly decreased for the constrained sub-ensemble, not only for the first 10 forecast years but also beyond. This is mainly due to the fact that the constrained sub-ensemble has a much higher ability to capture the observed warming trend during the end of the 20th century compared to the uninitialized ensemble mean. Further to these results, the limitations of this framework are discussed, including an assessment of the potential upper limit of added value and it’s dependency on the skill of the decadal forecast system. 

This easy-to-apply framework can be used to provide crucial climate information for mitigation and adaptation strategies by filling the gap between initialized decadal predictions and uninitialized projections.

D3461 |
EGU2020-5852
Christopher O'Reilly, Daniel Befort, and Antje Weisheimer

In this study methods of calibrating the output of large single model ensembles are examined. The methods broadly involve fitting seasonal ensemble data to observations over a reference period and scaling the ensemble signal and spread so as to optimize the fit over the reference period. These calibration methods are then applied to the future (or out-of-sample) projections. The calibration methods are tested and give indistinguishable results so the simplest of these methods, namely Homogenous Gaussian Regression, is selected. An extension to this method, applying it to dynamically decomposed data (in which the underlying data is separated into dynamical and residual components), is found to improve the reliability of the calibrated projections. The calibration methods were tested and verified using an “imperfect model” approach using the historical/RCP8.5 simulations from the CMIP5 archive. The verification indicates that this relatively straight-forward calibration produces more reliable and accurate projections than the uncalibrated (bias-corrected) ensemble for projections of future climate over Europe. When the two large ensembles are applied to observational data, the 2041-2060 climate projections for Europe for the RCP 8.5 scenario are more consistent between the two ensembles, with a slight reduction in warming but an increase in the uncertainty of the projected changes. 

D3462 |
EGU2020-10990
Hylke de Vries and Geert Lenderink

Following the huge increase in computer power over the past fifty years, it is now possible to conduct regional climate simulations at the convective-permitting (CP) scale (horizontal grid spacing O(1-3km)). Societal relevance is evident, as a large fraction of high-impact weather operates on these scales (e.g., intense summer convection, extreme wind gusts, hail storms) and it is only partly known how these processes change as the planet warms. However, because the computational demands of running a CP regional climate model (CP-RCM) are formidable, most CP-RCM simulations to date are relatively short (O(10) years), with future trends being derived as differences between two time slices. Consequently, internal variability is considerable in these decadal simulations for most variables, but is difficult to estimate, especially for the extremes. Nevertheless, these simulations are the best we have at the moment. 

For some variables such as summer rainfall, spatial pooling might be a way to increase the sample size. However we believe a careful analysis of the internal variability is still necessary to provide a context for the future changes. Here we examine different methods to estimate the amplitude of the internal variability of precipitation and temperature over Europe. Special focus will be on comparing the differences between such estimates obtained from O(10) years time-slice experiments and those obtained from long transient RCM simulations (which in contrast to the CP-RCMs are available!), with the expectation that some of the lessons learned in the "RCM-world" carry over to that of the CP-RCM.

D3463 |
EGU2020-11554
Gabriele Hegerl, Andrew Ballinger, and Sabine Undorf

Quantifying and reducing the uncertainty of climate projections will benefit both mitigation and adaptation decisions. Observed climate change provides evaluation of climate model simulated change, but the contribution by different external forcing factors needs to be reliably separated in order to use observational constraints. We revisit this ASK (for Allen et al., 2000; Stott and Kettleborough, 2002) approach to use attributed responses to greenhouse gas forcing to constrain future predictions.

We derive constraints on the projected near-surface summer temperature change over Europe as well as over three European subregions. The temperature responses to different external forcings (natural and greenhouse-gas (GHG) or combined anthropogenic) are estimated as the multi-model means of historical simulations from the Coupled Model Intercomparison Project 5 and incoming CMIP6, and the range of factors by which they can be scaled and still be consistent with observations since 1950 (E-OBS) given internal variability is calculated and applied to future RCP8.5 simulations.

Results show that both the response to GHG-only and to the combined anthropogenic (including aerosols etc.) forcing are detectable in the observed temperature change over Europe, and that the response over the Mediterranean region might be underestimated. Observed precipitation changes over Europe are also detected over some regions, although the confounding effects of the North Atlantic Oscillation need to be considered carefully. The results demonstrate the successful application of the ASK method for constraining projections of regional change over Europe.

 

D3464 |
EGU2020-15892
Peter Greve, Peter Burek, Renate Wilcke, Lukas Brunner, Carol McSweeney, Ben Booth, Geert Lenderink, and Yoshihide Wada

Global hydrological models (GHMs) have become an established tool to simulate water resources on continental scales. To assess the future of water availability and various impacts related to hydrological extreme events, these models usually use sets of atmospheric variables (such as e.g., precipitation, humidity, temperature) obtained from (regional) climate model simulations as input data. The uncertainty associated with the climate projections is transferred onwards into the impact simulations and is usually accounted for through the use of large model ensembles. These ensembles thus enable assessments addressing the robustness of projected hydrological changes and impacts. Given recent efforts within the European Climate Prediction (EUCP) project to test existing and develop new techniques to constrain/weight climate model ensembles, we use here different methods to specify the large-scale meteorological input to an ensemble of regional climate models that provide the input data for a state-of-the-art GHM. The climate models are weighted/constrained based on the key large-scale climatic and meteorological drivers shaping the hydrological characteristics in different regions and large river basins across Europe. To assess the potential benefits of the different techniques, we compare simulation ensembles using unweighted input data obtained from the full ensemble of regional climate models against an ensemble based on constrained/weighted forcing data. Given the large uncertainties usually associated with hydrological impact simulations forced by the full range of available climate models, processing the ensemble output of GHMs based on uncertainty assessments of the underlying climate forcing could lead to more robust projections of water resources in general and hydrological extreme events in particular.

D3465 |
EGU2020-16822
Danijel Belusic, Petter Lind, Oskar Landgren, Dominic Matte, Rasmus Anker Pedersen, Erika Toivonen, Felicitas Hansen, and Fuxing Wang

Current literature strongly indicates large benefits of convection permitting models for subdaily summer precipitation extremes. There has been less insight about other variables, seasons and weather conditions. We examine new climate simulations over the Nordic region, performed with the HCLIM38 regional climate model at both convection permitting and coarser scales, searching for benefits of using convection permitting resolutions. The Nordic climate is influenced by the North Atlantic storm track and characterised by large seasonal contrasts in temperature and precipitation. It is also in rapid change, most notably in the winter season when feedback processes involving retreating snow and ice lead to larger warming than in many other regions. This makes the area an ideal testbed for regional climate models. We explore the effects of higher resolution and better reproduction of convection on various aspects of the climate, such as snow in the mountains, coastal and other thermal circulations, convective storms and precipitation with a special focus on extreme events. We investigate how the benefits of convection permitting models change with different variables and seasons, and also their sensitivity to different circulation regimes.

D3466 |
EGU2020-18478
Carol McSweeney, David Sexton, Philip Bett, Hazel Thornton, Ruth McDonald, Marie Drouard, Tim Woollings, John Rostron, Kuniko Yamazaki, and James Murphy

European climate is influenced by a number of large-scale phenomena which are typically poorly represented by global climate models.  A key motivation in generating a coupled perturbed parameter ensemble (PPE) for use in the latest UK climate projections (UKCP18) was to exploit the significant improvements in regional dynamics that have been demonstrated at higher vertical and horizontal resolutions.

The UKCP18 package includes a number of products, including a set of 28 global model ‘realizations’ comprising a 15-member PPE and a filtered sub-set of 13 CMIP5 members. These physically coherent, spatially and temporally complete scenarios of future change provide a flexible tool for exploring plausible future changes and their likely impacts.

We present an assessment of the PPE’s ability to represent key aspects of the regional large-scale circulation and its implications for the realistic simulation of UK and European climate and its variability. These include the large-scale circulation climatology, frequencies of weather types determined by clustering of north Atlantic MSLP anomalies, latitude and strength of the north Atlantic jet, location and frequency of north Atlantic and European storms and the frequency of blocking events. We show that the PPE members perform at least at as well as, or better than, the filtered 13-member CMIP5 subset with respect to these circulation characteristics.  This realistic behaviour offers a good basis for UK and European climate impacts studies, as well as the further development of ‘storylines’ approaches.

D3467 |
EGU2020-19475
Jason A. Lowe, Carol McSweeney, and Chris Hewitt

There is clear evidence that, even with the most favourable emission pathways over coming decades, there will be a need for society to adapt to the impacts of climate variability and change. To do this regional, national and local actors need up-to-date information on the changing climate with clear accompanying detail on the robustness of the information. This needs to be communicated to both public and private sector organisations, ideally as part of a process of co-developing solutions.

EUCP is an H2020 programme that began in December 2017 with the aim of researching and testing the provision of improved climate predictions and projections for Europe for the next 40+ years, and drawing on the expertise of researchers from a number of major climate research institutes across Europe. It is also engaging with users of climate change information through a multiuser forum (MUF) to ensure that what we learn will match the needs of the people who need if for decision making and planning.

The first big issue that EUCP seeks to address is how better to use ensembles of climate model projections, moving beyond the one-model-one-vote philosophy. Here, the aim is to better understand how model ensembles might be constrained or sub-selected, and how multiple strands of information might be combined into improved climate change narratives or storylines. The second area where EUCP is making progress is in the use of very high-resolution regional climate simulations that are capable of resolving aspects of atmospheric convection. Present day and future simulations from a new generation of regional models ae being analysed in EUCP and will be used in a number of relevant case studies. The third issue that EUCP will consider is how to make future simulations more seamless across those time scales that are most relevant user decision making. This includes generating a better understanding of predictability over time and its sources in initialised forecasts, and also how to transition from the initialised forecasts to longer term boundary forced climate projections.

This presentation will provide an overview of the challenges being addressed by EUCP and the approaches the project is using.



 

D3468 |
EGU2020-21533
Saïd Qasmi, Aurélien Ribes, and Hervé Douville

Observational constraints involve the combined use of models and observations in order to assess their consistency, and aim to reduce the uncertainty on future climate using past information. Several constraints are investigated with the CMIP5 models and re-examined in the light of newly available CMIP6 data. This includes constraints based on detection-attribution approaches and physically-based constraints, in particular those related to the water cycle (e.g. soil moisture, clouds, snow cover). A wide range of methods is used to provide a probabilistic description of future changes address the issue of combining together multi-model ensembles of projections, and a large number of observational constraints. Uncertainty quantification techniques are used to assess the sensitivity of the results (i) to the used method and (ii) to the internal variability.

D3469 |
EGU2020-3666
Deborah Verfaillie, Francisco J. Doblas-Reyes, Markus G. Donat, Nuria Pérez-Zanón, Balakrishnan Solaraju-Murali, Verónica Torralba, and Simon Wild

Decadal climate predictions and forced climate projections both provide potentially useful information to users for the next ten years. They only differ in the former being initialised with observations, while the latter is not. Bringing together initialised decadal climate predictions and non-initialised climate projections in order to provide seamless climate information for users over the next decades is a new challenging area of research. This can be achieved by comparing the forecast quality of global initialised and non-initialised simulations in their common prediction time horizons (up to 10 years ahead), and quantify in how far initialisation improves the forecast quality. Forecast quality has been usually explored through skill assessment. However, the impact of initialisation on the reliability, which quantifies the agreement between the predicted probabilities and observed relative frequencies of a given event, of decadal predictions has not yet been investigated sufficiently. Hence, users of probabilistic predictions are particularly sensitive to the potential lack of reliability which would imply that the probabilities are not trustworthy and this can have negative consequences for decision-making. In this communication, initialised decadal hindcasts (or retrospective forecasts) from 12 forecasting systems of the Coupled Model Intercomparison Project Phase 5 are compared to the corresponding non-initialised historical simulations in terms of reliability over their common period 1961-2005. We show that reliability varies greatly depending on the region or model ensemble analysed and on the correction applied. In particular, the North Atlantic and Europe stand out as regions where there is some added-value of initialised decadal hindcasts over non-initialised historical simulations in terms of reliability, mainly because of smaller biases and/or a better representation of the trend. Furthermore, we show that post-processed data display more reliable results, indicating that bias correction and calibration are fundamental to obtain reliable climate information.

D3470 |
EGU2020-4914
Alessio Bellucci, Marianna Benassi, Silvio Gualdi, and Annarita Mariotti

Understanding processes and mechanisms which contribute to decadal climate variability is a crucial step in the development of a reliable prediction system, and as such it constitutes an important segment of the activities carried forward by the EU-funded Horizon 2020 EUCP project.

Sea surface temperature (SST) variability in the North Atlantic is known to be a key source of decadal predictability for the Euro-Atlantic sector. However, the nature of the observed variability is at the core of a long-standing debate.

In this work, we investigate the origins of North Atlantic SST variability, focusing on a specific event: the mid-20th century (1940-1975) “warm-to-cold” transition. This event is particularly interesting as it represents a well documented decadal-scale fluctuation of the observed climate record and can be used as a suitable test-bed to evaluate the relative skill of initialized versus non-initialized (historical) climate simulations.

Several mechanisms and processes have been taken into account to explain the cooling in the middle of 20th century, ranging from a slowdown of the Atlantic Meridional Overturning Circulation (AMOC) to an increase in anthropogenic aerosol. Here the 1940-1975 transition is examined firstly in the NCAR Large Ensemble (NCAR-LENS), aiming to further explore the role of the possible drivers. Despite the lack of a realistic model state initialization, the NCAR-LENS shows some skill in capturing the North Atlantic SST transition, suggesting a non-negligible influence of the external forcing. Some lag between observations and model results is found, with the ensemble mean SST leading the onset of the observed transition by about ten years. This is consistent with previous studies, where some evidence was found of the driving role of anthropogenic aerosol and greenhouse gas forcing. In contrast, the simultaneous ocean dynamic response (AMOC) exhibits a large intra-member spread. This finding corroborates the hypothesis of a non-oceanic driver for the decadal-scale SST fluctuation. The same episode is then analysed in the NCAR Decadal Prediction Large Ensemble (NCAR-DPLE), which shares the same model code, configuration details, component resolutions, and external forcing datasets as for the non-initialized LENS ensemble. This allows a rigorous attribution of the relative roles of initialization, (mainly constraining the ocean-driven internal variability) and external forcing conditions on the overall skill in reproducing the Atlantic decadal variability, with clear implications for decadal predictability and predictions.

 

D3471 |
EGU2020-8985
Marjanne Zander, Frederiek Sperna Weiland, and Albrecht Weerts

In this study a methodology is developed and tested to delineate homogeneous regions of extreme rainfall around a city of interest using meteorological indices from reanalysis data.

Scenarios of future climate change established with numerical climate models are well-established tools to help inform climate adaptation policy. The latest generation of regional climate models is now employed at a grid resolution of 2 to 3 kilometers. This enables the simulation of convection; whereby intensive convective rainfall is better represented (Kendon et al., 2017). However, the associated large computational burden limits the simulation length, which poses a challenge for estimating future rainfall statistics.

Rainfall return periods are a commonly used indicator in the planning, design and evaluation of urban water systems and urban water management. In order to estimate potential future rainfall for return periods larger than the length of the simulation length, regional frequency analysis (RFA) can be applied (Li et al., 2017).  For applying RFA, time series from nearby locations are pooled, the locations considered should fall within the same hydroclimatic climate. This is a region which can be assumed statistically homogeneous for extreme rainfall (Hosking & Wallis, 2009).

Traditionally, these homogeneous regions are defined on geographical region characteristics and rain gauge statistics (Hosking & Wallis, 2009).  To make the methodology less dependent on rain gauge record availability, Gabriele & Chiaravalloti (2013) used meteorological indices derived from reanalysis data to delineate the homogeneous regions.

Here we evaluate the methodology to delineate homogeneous regions around cities. Meteorological indices are calculated from the ERA-5 reanalysis dataset (Hersbach et al., 2018) for days with extreme rainfall. The variation herein is used as a measure of homogeneity. The derived homogeneous regions will in future work be used for data pooling of convection-permitting regional climate model simulations datasets to enable the derivation of future extreme rainfall statistics.

This study is embedded in the EU H2020 project EUCP (EUropean Climate Prediction system) (https://www.eucp-project.eu/), which aims to develop a regional climate prediction and projection system based on high-resolution climate models for Europe, to support climate adaptation and mitigation decisions for the coming decades.

References:

Gabriele, S., & Chiaravalloti, F. (2013). “Searching regional rainfall homogeneity using atmospheric fields”. Advances in Water Resources, 53, 163–174. https://doi.org/https://doi.org/10.1016/j.advwatres.2012.11.002

Hersbach, H., de Rosnay, P., Bell, B., Schepers, D., Simmons, A., Soci, C., …, Zuo, H. (2018). “Operational global reanalysis: progress, future directions and synergies with NWP”, ECMWF.

Hosking, J. R. M., & Wallis, J. R. (2009). “Regional Frequency Analysis: An Approach Based on L-Moments”. The Edinburgh Building, Cambridge CB2 2RU, UK: Cambridge University Press. ISBN: 9780511529443.

Kendon, E. J., Ban, N., Roberts, N. M., Fowler, H. J., Roberts, M. J., Chan, S. C., … Wilkinson, J. M. (2017). “Do Convection-Permitting Regional Climate Models Improve Projections of Future Precipitation Change?” BAMS, 98(1), 79–93. https://doi.org/10.1175/BAMS-D-15-0004.1

 Li, J., Evans, J., Johnson, F., & Sharma, A. (2017). “A comparison of methods for estimating climate change impact on design rainfall using a high-resolution RCM.” Journal of Hydrology, 547(Supplement C), 413–427. https://doi.org/https://doi.org/10.1016/j.jhydrol.2017.02.019

D3472 |
EGU2020-18519
Torben Schmith, Peter Thejll, Fredrik Boberg, Peter Berg, Ole Bøssing Christensen, Bo Christiansen, Marianne Sloth Madsen, and Jens Hesselbjerg Christensen

Severe precipitation events occur rarely and are often localized in space and of short duration, but are important for societal managing of infrastructure such as sewage systems, metros etc. Therefore, there is a demand for estimating expected future changes in the statistics of these rare events. These are usually projected using RCM scenario runs combined with extreme value analysis to obtain selected return levels of precipitation intensity. However, due to RCM imperfections, the modelled climate for the present-day usually has errors relative to observations. Therefore, the RCM results are ‘error corrected‘ to match observations more closely in order to increase reliability of results.

In the present work we evaluate different error correction techniques and compare with non-corrected projections. This is done in an inter-model cross-validation setup, in which each model in turn plays the role of observations, against which the remaining error-corrected models are validated. The study uses hourly data (historical & RCP8.5 late 21st century) from 13 models covering the EURO-CORDEX ensemble at 0.11 degree resolution (about 12.5 km), from which fields of selected return levels are extracted for 1 h and 24 h duration. The error correction techniques applied to the return levels are based on extreme value analysis and include analytical quantile-quantile matching together with a simpler climate factor approach.

The study identifies regions where the error correction techniques perform differently, and therefore contributes to guidelines on how and where to apply calibration techniques when projecting extreme return levels.