Several large ensemble model simulations from General Circulation Models (GCM), Earth System Models (ESM), or Regional Climate Models (RCM), have been generated over the recent years to investigate internal variability and forced changes of the climate system - and to aid the interpretation of the observational record by providing a range of historical climate trajectories that could have been. The increased availability of large ensembles also enables broadening their application to new and inter-disciplinary fields.

This session invites studies using large GCM, ESM, or RCM ensembles looking at the following topics: 1) forced changes in internal variability and reinterpretation of observed record; 2) development of new approaches to attribution of observed events or trends; 3) impacts of natural climate variability; 4) assessment of extreme and compound event occurrence; 5) use of large ensembles for robust decision making; 6) large ensembles as testbeds for method development; and 7) novel methods for efficient analysis and post-processing of large ensembles.

We welcome research across the components of the Earth system and particularly invite studies that apply novel methods or cross-disciplinary approaches to leverage the potential of large ensembles.

Public information:
Announcement: Note that this session will be conducted in two parts:

1) The official #ShareEGU20 live chat. May 8, 14:00-15:45.
Agenda: https://meetingorganizer.copernicus.org/EGU2020/displays/36913

2) An additional live streaming of oral presentations. Friday, May 8, 16:15-18:00.
Agenda: bit.ly/2RX1hd9.
RSVP form: bit.ly/3bvzqZ4

Order of discussion for display items during the live chat:

Nathalie Schaller
2. Bin Yu
3. Benoit Hingray
4. Laura Suarez Gutierrez
5. Flavio Lehner
6. Bo Christiansen
7. Karsten Haustein
8. Renate Wilcke
9. Lea Beusch
10. Andrea Böhnisch
11. Satoshi Watanabe
12. Peter Watson
13. Tamas Bodai
14. Gabor Drotos
15. Gerhard Smiatek
16. Ralf Hand (1st display)
17. Ralf Hand (2nd display)
18. Mátyás Herein

19. Joel Zeder
20. Sebastian Milinski
21. Anna Merrifield
22. Raul R. Wood
23. Shipra Jain
24. Aaron Spring

Co-organized by CL2/HS13
Convener: Flavio LehnerECSECS | Co-conveners: Andrea DittusECSECS, Ralf Ludwig, Laura Suarez-GutierrezECSECS, Karin van der WielECSECS
| Attendance Fri, 08 May, 14:00–15:45 (CEST)

Files for download

Session materials Download all presentations (80MB)

Chat time: Friday, 8 May 2020, 14:00–15:45

Chairperson: Flavio Lehner, Andrea Dittus, Ralf Ludwig, Laura Suarez-Gutierrez, Karin van der Wiel
D3001 |
| solicited
| Highlight
Nathalie Schaller

Large ensembles are key to investigate climate and weather extremes and their impacts, as they, by definition, rarely occur. One field that relies heavily on them is probabilistic event attribution, i.e. where one tries to quantify how human influence affects the probability of occurrence of the extreme event in question. An ensemble of over 130’000 members allowed us to quantify that human influence increased the probability of heavy precipitation by around 40% in the January 2014 floods in southern England. By using a hydrological model, we could then quantify that the probability of 30-day peak river flows of the Thames river was increased by around 20%. However, it was unclear whether the number of properties at risk in the catchment was affected. This study also showed how uncertainty increases at each step of the modelling chain and how some factors, like the characteristics of the Thames catchment in this case, might play a bigger role in assessing impacts than potentially the size of the ensemble.

Large ensembles are also useful to understand the physical mechanisms behind extreme events. In another study about the relationship between atmospheric blocking and heatwaves, we used three large ensembles from different climate models. While we found that the 2003 European heatwave and blocking conditions were well contained within the 3 ensembles’ envelope, and that the models simulated even more extreme events, the 2010 Russian event was outside the ensembles’ envelope, except for one single ensemble member.

Finally, I will present two projects, one on floods in Norway and one about the health impacts of having a heatwave combined with high air pollution, where large ensembles would be useful, but are competing with the need for high spatial resolution for computational resources.

How to cite: Schaller, N.: Using large ensembles to investigate the impacts of climate extremes, EGU General Assembly 2020, Online, 4–8 May 2020, EGU2020-19303, https://doi.org/10.5194/egusphere-egu2020-19303, 2020

D3002 |
| Highlight
Joel Zeder and Erich M. Fischer

The scientific understanding of changes in climate extremes is mostly limited to moderate definitions of extreme events occurring every few years, due to a lack of long-term observational daily data sets. In order to estimate return levels beyond observed time periods and event magnitudes, extreme events are typically modelled statistically based on extreme value theory. This is challenging since the short observational record may be affected by low-frequency natural internal variability and limits the block size that can be used.

Here we test some common assumptions in the statistical modelling of extremes based on indices of climatic extremes (Tx7d, Rx1d, Rx5d) using long pre-industrial control runs and initial-condition large ensembles with thousands of years of model data.

The tail of a distribution fitted to temperature and precipitation maxima is known to be highly sensitive to the compliance with statistical assumptions and choices such as the block size. Typically, 1-year block maxima are extracted from observational time series due to short record length. It is unclear whether these maxima are already in the domain of true extremes suitable for an extreme value analysis. Furthermore, the observational record is too short to sample low-frequency regional variability and potential transient changes in the mean climate. Standard uncertainty estimates (confidence intervals and hypothesis tests) are generally not accounting for potential biases introduced by a dominant mode of climate variability or violated modelling assumptions.

Based on a 4700-year pre-industrial control simulation and an 84-member ensemble performed with CESM 1.2.2 model, we systematically extend the statistical modelling of temperature and precipitation extremes to larger block-sizes and longer synthetic observational periods. This analysis reveals a considerable influence of climate variability on tail estimates. Furthermore, the use of too small block sizes can induce substantial random as well as systematic biases. Statistical model complexity and thus uncertainty further increases for extremes retrieved from transient large-ensemble members, as non-stationarity has to be accounted for in the model formulation. Thus, the potential of spatial pooling or conditioning on further climatic variables as proxies for a specific climatic mode to derive more robust tail estimates is also evaluated. Findings based on the CESM ensemble are compared with pre-industrial control runs performed with other models in CMIP6 and other initial-condition large ensembles of the CLIVAR large ensemble working group.

How to cite: Zeder, J. and Fischer, E. M.: The challenge of estimating high return levels with short records under large internal variability, EGU General Assembly 2020, Online, 4–8 May 2020, EGU2020-18105, https://doi.org/10.5194/egusphere-egu2020-18105, 2020

D3003 |
Sebastian Milinski, Nicola Maher, and Dirk Olonscheck

Initial-condition large ensembles with ensemble sizes ranging from 30 to 100 members have become a commonly used tool to quantify the forced response and internal variability in various components of the climate system. However, there is no consensus on the ideal or even sufficient ensemble size for a large ensemble.

Here, we introduce an objective method to estimate the required ensemble size. This method can be applied to any given application. We demonstrate its use on the examples that represent typical applications of large ensembles: quantifying the forced response, quantifying internal variability, and detecting a forced change in internal variability.

We analyse forced trends in global mean surface temperature, local surface temperature and precipitation in the MPI Grand Ensemble (Maher et al., 2019). We find that 10 ensemble members are sufficient to quantify the forced response in historical surface temperature over the ocean, but more than 50 members are necessary over land at higher latitudes. 

Next, we apply our method to identify the required ensemble size to sample internal variability of surface temperature over central North America and over the Niño 3.4 region. A moderate ensemble size of 10 members is sufficient to quantify variability over North America, while a large ensemble with close to 50 members is necessary for the Niño 3.4 region.

Finally, we use the example of September Arctic sea ice area to investigate forced changes in internal variability. In a strong warming scenario, the variability in sea ice area is increasing because more open water near the coastlines allows for more variability compared to a mostly ice-covered Arctic Ocean (Goosse et al., 2009; Olonscheck and Notz, 2017). We show that at least 5 ensemble members are necessary to detect an increase in sea ice variability in a 1% CO2 experiment. To also quantify the magnitude of the forced change in variability, more than 50 members are necessary.

These numbers might be highly model dependent. Therefore, the suggested method can also be used with a long control run to estimate the required ensemble size for a model that does not provide a large number of realisations. Therefore, our analysis framework does not only provide valuable information before running a large ensemble, but can also be used to test the robustness of results based on small ensembles or individual realisations.

Goosse, H., O. Arzel, C. M. Bitz, A. de Montety, and M. Vancoppenolle (2009), Increased variability of the Arctic summer ice extent in a warmer climate, Geophys. Res. Lett., 36(23), 401–5, doi:10.1029/2009GL040546.

Olonscheck, D., and D. Notz (2017), Consistently Estimating Internal Climate Variability from Climate Model Simulations, J Climate, 30(23), 9555–9573, doi:10.1175/JCLI-D-16-0428.1.

Milinski, S., N. Maher, and D. Olonscheck (2019), How large does a large ensemble need to be? Earth Syst. Dynam. Discuss., 2019, 1–19, doi:10.5194/esd-2019-70.

How to cite: Milinski, S., Maher, N., and Olonscheck, D.: How large does a large ensemble need to be?, EGU General Assembly 2020, Online, 4–8 May 2020, EGU2020-13843, https://doi.org/10.5194/egusphere-egu2020-13843, 2020

D3004 |
| Highlight
Anna Merrifield, Lukas Brunner, Ruth Lorenz, and Reto Knutti

Multi-model ensembles can be used to estimate uncertainty in projections of regional climate, but this uncertainty often depends on the constituents of the ensemble. The dependence of uncertainty on ensemble composition is clear when single model initial condition large ensembles (SMILEs) are included within a multi-model ensemble. SMILEs introduce new information into a multi-model ensemble by representing region-scale internal variability, but also introduce redundant information, by virtue of a single model being represented by 50–100 outcomes. To preserve the contribution of internal variability and ensure redundancy does not overwhelm uncertainty estimates, a weighting approach is used to incorporate 50-members of the Community Earth System Model (CESM1.2.2), 50-members of the Canadian Earth System Model (CanESM2), and 100-members of the MPI Grand Ensemble (MPI-GE) into an 88-member Coupled Model Intercomparison Project Phase 5 (CMIP5) multi-model ensemble. The weight assigned to each multi-model ensemble member is based on the member's ability to reproduce observed climate (performance) and scaled by a measure of historical redundancy (dependence). Surface air temperature (SAT) and sea level pressure (SLP) diagnostics are used to determine the weights, and relationships between present and future diagnostic behavior are discussed. A new diagnostic, estimated forced trend, is proposed to replace a diagnostic with no clear emergent relationship, 50-year regional SAT trend.

The influence of the weighting is assessed in estimates of Northern European winter and Mediterranean summer end-of-century warming in the CMIP5 and combined SMILE-CMIP5 multi-model ensembles. The weighting is shown to recover uncertainty obscured by SMILE redundancy, notably in Mediterranean summer. For each SMILE, the independence weight of each ensemble member as a function of the number of SMILE members included in the CMIP5 ensemble is assessed. The independence weight increases linearly with added members with a slope that depends on SMILE, region, and season. Finally, it is shown that the weighting method can be used to guide SMILE member selection if a subsetted ensemble with one member per model is sought. The weight a SMILE receives within a subsetted ensemble depends on which member is used to represent it, reinforcing the advantage of weighting and incorporating all initial condition ensemble members in multi-model ensembles.

How to cite: Merrifield, A., Brunner, L., Lorenz, R., and Knutti, R.: A weighting scheme to incorporate large ensembles in multi-model ensemble projections, EGU General Assembly 2020, Online, 4–8 May 2020, EGU2020-4524, https://doi.org/10.5194/egusphere-egu2020-4524, 2020

D3005 |
Raul R. Wood, Flavio Lehner, Angeline Pendergrass, Sarah Schlunegger, and Keith Rodgers

Identifying anthropogenic influences on climate amidst the “noise” of internal climate variability is a central challenge for the climate research community. In recent years, several modeling groups have produced single-model initial-condition large ensembles (SMILE) to analyze the interplay of the forced climate change and internal climate variability under current and future climate conditions. These simulations help to improve our understanding of climate variability, including extreme events, and can be employed as test-beds for statistical approaches to separate forced and internal components of climate variability.

So far, most studies have focused on either an individual or a  limited number of SMILEs. In this work we compare seven large ensembles to disentangle the influence of internal variability and model response uncertainty for multiple precipitation indices (e.g. wettest day of the year, precipitation with a return period of 20 years). What can we learn from intercomparison of SMILEs, how similar are they in terms of spatial patterns and forced response, and what if they aren’t? How does the forced response of an ensemble of SMILEs compare to the CMIP5 multi-model ensemble? By assessing multiple SMILEs we can identify robust signals for regional and global precipitation properties and revealing anthropogenic responses that are inherent to our current representations of the Earth system.

How to cite: Wood, R. R., Lehner, F., Pendergrass, A., Schlunegger, S., and Rodgers, K.: What can we learn from single model initial-condition large ensembles (SMILEs)? A Comparison of Multiple SMILEs for Precipitation, EGU General Assembly 2020, Online, 4–8 May 2020, EGU2020-19202, https://doi.org/10.5194/egusphere-egu2020-19202, 2020

D3006 |
| Highlight
Shipra Jain, Adam A Scaife, Nick Dunstone, Doug Smith, Saroj K Mishra, and Ruth Doherty

India suffers from severe social-economic losses due to floods and droughts during boreal summer (June-September) and therefore there is a growing interest in the current risk of extreme monsoon rainfall. In this analysis, we estimate the risk of flood, drought and unprecedented (outside the range of present observational record) rainfall over India using UNprecedented Simulated Extremes using ENsembles (UNSEEN) method. The UNSEEN is a statistical framework under which the risk of unprecedented rainfall extremes can be estimated using a large ensemble of initialized climate simulations to sample a broad range of internal variability. This is the first application of the method to the hindcasts from multiple coupled atmosphere-ocean models. Under this method, we first test individual models against the observed rainfall record over India and select models that are statistically indistinguishable from observations. The risk of floods, droughts and unprecedented rainfall is then estimated using a large ensemble of summer precipitation simulated by the selected set of models. We note that in present climate the risk of drought is higher than the flood, with droughts being more frequent and intense than the floods. This asymmetry in rainfall extremes is found to be partly due to the asymmetry in El-Nino Southern Oscillation (ENSO) phase, with El Nino reaching higher magnitude more frequently than La Nina. The current risk of record breaking drought (>23% deficit w.r.t climatological mean) is 1.6% whereas the risk for record-breaking flood (>16% excess) is 2.6%. There is even a risk of 30% rainfall deficit that could occur around once in two centuries, which is not yet seen in observations and would have a catastrophic influence on India.

How to cite: Jain, S., Scaife, A. A., Dunstone, N., Smith, D., Mishra, S. K., and Doherty, R.: Current Risk of Extreme Monsoon Rainfall over India using Large Ensemble Simulations, EGU General Assembly 2020, Online, 4–8 May 2020, EGU2020-7674, https://doi.org/10.5194/egusphere-egu2020-7674, 2020

D3007 |
| Highlight
Aaron Spring, Tatiana Ilyina, and Jochem Marotzke

On inter-annual time scales the growth rate of atmospheric CO2 is largely driven by the response of the land and ocean carbon sinks to climate variability. Therefore, climate mitigation in terms of emission reductions can be disguised by internal variability.
However, the probability that emission reductions induced by a policy change caused reductions in atmospheric CO2 growth trend is unclear.
We use 100 historical MPI-ESM simulations and interpret mitigation in 2020 as a policy shift from Representative Concentration Pathway 4.5 to 2.5 in a comprehensive causation attribution framework.
Here we show that five-year CO2 trends are higher in 2021-2025 than over 2016-2020 in 30% of all realizations in the mitigation scenario, compared to 52% in the non-mitigation scenario. Therefore, mitigation is sufficient or necessary to cause these trends by 42% or 31%, respectively and therefore far from certain.
A stronger increase in atmospheric CO2 trends despite emission reductions is possible when the global carbon cycle triggered by internal climate variability releases more CO2 than mitigation saves. Such trends might occur for of up to ten years. Certainty that mitigation causes trend reductions is only reached after ten or fifteen years, respectively of the type of causation.
Our analysis showcases the inherent uncertainty of near-term CO2 projections. Assessments of the efficacy of mitigation in the near term are incomplete without quantitatively considering internal variability.

How to cite: Spring, A., Ilyina, T., and Marotzke, J.: Inherent Uncertainty Disguises Attribution of Reduced Atmospheric CO2 Growth to Mitigation for up to a Decade, EGU General Assembly 2020, Online, 4–8 May 2020, EGU2020-2180, https://doi.org/10.5194/egusphere-egu2020-2180, 2020

D3008 |
Xuewei Fan

Surface air temperature outputs from 16 global climate models (GCMs) participating in the sixth phase of the Coupled Model Intercomparison Project (CMIP6) were used to evaluate agreement with observations over the global land surface for the period 1901–2014. Projections of Bayesian model averaging (BMA) multi-model ensembles under four different Shared Socioeconomic Pathways (SSPs) were also examined. The results reveal that the majority of models reasonably capture the dominant features of the spatial changes in observed temperature with a pattern correlation typically greater than 0.98. However, most models underestimate annual temperature over northeastern North America and overestimate it over central Eurasia. In addition, most CMIP6 models overestimate the warming trend in most regions. The BMA multi-model ensembles show more agreement than individual models do in simulating the spatial patterns of the temperature, but with less spatial variability compared with the observations. In the 21st century, temperature is generally projected to increase over the global land surface under all four SSP scenarios. By the end of the 21st century, temperature is projected to increase by 1.35 °C/100 yr, 3.61 °C/100 yr, 6.39 °C/100 yr and 8.03 °C/100 yr under the SSP1-2.6, SSP2-4.5, SSP3-7.0 and SSP5-8.5 scenarios, respectively, with greater warming projected over the high latitudes of the northern hemisphere and weaker warming over the tropics and the southern hemisphere.

How to cite: Fan, X.: Global surface air temperatures in CMIP6: Historical performance and future, EGU General Assembly 2020, Online, 4–8 May 2020, EGU2020-5436, https://doi.org/10.5194/egusphere-egu2020-5436, 2020

D3009 |
Bin Yu, Guilong Li, Shangfeng Chen, and Hai Lin

Recent studies indicated that the internal climate variability plays an important role in various aspects of projected climate changes on regional and local scales. Here we present results of the spreads in projected trends of wintertime North American surface air temperature and extremes indices of warm and cold days over the next half-century, by analyzing a 50-member large ensemble of climate simulations conducted with CanESM2. CanESM2 simulations confirm the important role of internal variability in projected surface temperature trends as demonstrated in previous studies. Yet the spread in North American warming trends in CanESM2 is generally smaller than those obtained from CCSM3 and ECHAM5 large ensemble simulations. Despite this, large spreads in the climate means as well as climate change trends of North American temperature extremes are apparent in CanESM2, especially in the projected cold day trends. The ensemble mean of forced climate simulations reveals high risks of warm days over the western coast and north Canada, as well as a weakening belt of cold days extending from Alaska to the northeast US. The individual ensemble members differ from the ensemble mean mainly in magnitude of the warm day trends, but depart from the ensemble mean in conspicuous ways, including spatial pattern and magnitude, of the cold day trends. The signal-to-noise ratio pattern of the warm day trend resembles that of the surface air temperature trend; with stronger signals over north Canada, Alaska, and the southwestern US than the midsection of the continent. The projected cold day patterns reveal strong signals over the southwestern US, north Canada, and the northeastern US. In addition, the internally generated components of temperature and temperature extreme trends exhibit spatial coherences over North America, and are comparable to the externally forced trends. The large-scale atmospheric circulation-induced temperature variability influences these trends. Overall, our results suggest that climate change trends of North American temperature extremes are likely very uncertain and need to be applied with caution.

How to cite: Yu, B., Li, G., Chen, S., and Lin, H.: The role of internal variability in climate change projections of North American surface air temperature and temperature extremes in CanESM2 large ensemble simulations, EGU General Assembly 2020, Online, 4–8 May 2020, EGU2020-5854, https://doi.org/10.5194/egusphere-egu2020-5854, 2020

D3010 |
Benoit Hingray, Guillaume Evin, Juliette Blanchet, Nicolas Eckert, Samuel Morin, and Deborah Verfaillie

The quantification of internal variability and model uncertainty sources in Multi-scenario Multi-model Ensembles of climate experiments (MMEs) is a key issue. It is expected to both help decision makers to identify robust adaptation measures and scientists to identify where their efforts are needed to narrow uncertainty. The setup of available MMEs makes however uncertainty analyses difficult. In the popular single-time ANOVA approach for instance, a precise estimate of internal variability requires multiple members for each simulation chain (e.g. each emission scenario/climate model combination) but multiple members are typically available for a few chains only (Hingray et al. 2019). In almost all ensembles also, the matrix of available scenario/models combinations is incomplete making a precise estimate of the main effects of each model difficult (e.g. projections are typically missing for some GCM/RCM combinations) (Evin et al. 2019).

We present QUALYPSO, a Bayesian approach developed to assess the different sources of uncertainty in incomplete MMEs (Evin et al. submitted). It is based on the quasi-ergodic assumption for transient climate projections and uses data augmentation (Hingray and Said, 2014). The climate response of each available simulation chain is first estimated with a trend model fitted to raw climate projections. Residuals from the climate change response are used to estimate the internal variability of the chain. Scenario uncertainty and the different components of model uncertainty (e.g. GCM uncertainty, RCM uncertainty) are then estimated with a Bayesian ANOVA model applied to the climate change responses of all available chains. The different parameters of the ANOVA model and the missing quantities associated to the missing chains (e.g. missing scenario/GCM/RCM combinations) are jointly estimated using data augmentation techniques.

QUALYPSO presents many advantages over classical estimation approaches. It first exploits all available experiments, avoiding a dramatic loss of information (the classical case when standard approaches are applied; where the typical solution is to select a complete subset of climate experiments). Along with the estimation of missing data, it also provides an assessment of the estimation uncertainty and adequately propagates the uncertainty due to missing chains. With the explicit treatment of missing experiments, it is then expected to produce unbiased estimates of all parameters, in contrast to direct empirical estimates.

QUALYPSO can be applied to any kind of climate variable and any kind of MMEs. We present examples of application for different hydroclimatic variables from different ensembles of projections including EUROCORDEX and CORDEX-Africa.

Hingray, B., Saïd, M., 2014. Partitioning internal variability and model uncertainty components in a multimodel multireplicate ensemble of climate projections. J.Climate.

Hingray, B., Blanchet, J., Evin, G. Vidal, J.P. 2019. Uncertainty components estimates in transient climate projections. Precision of estimators in the single time and time series approaches. Clim.Dyn.

Evin, G., Hingray, B., Blanchet, J., Eckert, N., Morin, S., Verfaillie, D. 2019. Partitioning uncertainty components of an incomplete ensemble of climate projections using data augmentation. J.Climate.

Evin, G., Hingray, B. Blanchet, J., Eckert, N., Menegoz, M. Morin, S. (revision). Partitioning uncertainty components of an incomplete ensemble of climate projections using smoothing splines. J.Climate.

How to cite: Hingray, B., Evin, G., Blanchet, J., Eckert, N., Morin, S., and Verfaillie, D.: Partitioning uncertainty components of an incomplete ensemble of climate projections using data augmentation, EGU General Assembly 2020, Online, 4–8 May 2020, EGU2020-21864, https://doi.org/10.5194/egusphere-egu2020-21864, 2020

D3011 |
Nicola Maher, Laura Suarez-Gutierrez, and Sebastian Milinski

We evaluate how large ensembles of ten coupled climate models represent the observed internal variability and response to external forcings in historical surface temperatures based on a novel methodological framework. This framework allows us to directly attribute whether discrepancies between models and observations arise due to biases in the simulated internal variability or rather in the forced response, without relying on assumptions to separate both signals in the observations. The largest discrepancies occur due to overestimated forced warming in some models during recent decades. The areas where most models, a maximum of nine, adequately simulate observed temperatures are the North Atlantic, Tropical Eastern Pacific, and the Northern Hemisphere land areas. In contrast, none of the models considered offers an adequate representation over the Southern Ocean. Our evaluation shows that CESM-LE, GFDL-ESM2M, and MPI-GE perform best at representing the internal variability and forced response in observed surface temperatures both globally and regionally. 

How to cite: Maher, N., Suarez-Gutierrez, L., and Milinski, S.: Which climate models capture the variability and forced response in observed temperatures: a large ensemble comparison, EGU General Assembly 2020, Online, 4–8 May 2020, EGU2020-6081, https://doi.org/10.5194/egusphere-egu2020-6081, 2020

D3012 |
Flavio Lehner, Clara Deser, Nicola Maher, Jochem Marotzke, Erich Fischer, Lukas Brunner, Reto Knutti, and Ed Hawkins

Partitioning uncertainty in projections of future climate change into contributions from internal variability, model response uncertainty, and emissions scenarios has historically relied on making assumptions about forced changes in the mean and variability. With the advent of multiple Single-Model Initial-Condition Large Ensembles (SMILEs), these assumptions can be scrutinized, as they allow a more robust separation between sources of uncertainty. Here, we revisit the framework from Hawkins and Sutton (2009) for uncertainty partitioning for temperature and precipitation projections using seven SMILEs and the Climate Model Intercomparison Projects CMIP5 and CMIP6 archives. We also investigate forced changes in variability itself, something that is newly possible with SMILEs. The available SMILEs are shown to be a good representation of the CMIP5 model diversity in many situations, making them a useful tool for interpreting CMIP5. CMIP6 often shows larger absolute and relative model uncertainty than CMIP5, although part of this difference can be reconciled with the higher average transient climate response in CMIP6. This study demonstrates the added value of a collection of SMILEs for quantifying and diagnosing uncertainty in climate projections.

How to cite: Lehner, F., Deser, C., Maher, N., Marotzke, J., Fischer, E., Brunner, L., Knutti, R., and Hawkins, E.: Partitioning climate projection uncertainty with multiple Large Ensembles and CMIP5/6, EGU General Assembly 2020, Online, 4–8 May 2020, EGU2020-5991, https://doi.org/10.5194/egusphere-egu2020-5991, 2020

D3013 |
Bo Christiansen

Ensembles of model experiments have become the standard tool both in studies of climate change and in studies of prediction on many different time-scales. When analyzing such ensembles the mean of the ensemble is often interpreted as the best estimate and the spread of the ensemble as an estimate of the uncertainty.  Naively we might argue that the error of the ensemble mean would approach zero as the size of the ensemble increases. However, this argument is based on the assumption that the ensemble is centered around the observations - the truth-plus-error interpretation. A competing assumption - the indistinguishable interpretation -- holds that the observations and the models are all drawn from the same distribution.

The rationale for the truth centered interpretation is that it is the situation that would be expected after calibration of statistical models. However, for multi-model ensembles of climate models there is an increasing amount of evidence pointing towards the indistinguishable interpretation. But why should the indistinguishable interpretation hold for an ensemble that basically is a representation of our incomplete knowledge of the climate system?

Here we analyze CMIP5 ensembles focusing on three measures that separate the two interpretations: the error of the ensemble mean relative to the error of individual models, the decay of the ensemble mean error for increasing ensemble size, and the correlations of the model errors. To get more freedom in our analysis we use a simple statistical model where observations and models are drawn from distributions with different variances and which include a bias. The two interpretations can be found as limits of this more comprehensive model  for which analytical results can be found using the simplifying properties of high dimensional space (the blessing of dimensionality).

We find that the indistinguishable interpretation becomes an increasingly better assumption when the errors are based on smaller and smaller temporal and spatial scales. Building on this, we present a simple conceptual mechanism for the indistinguishable interpretation based on the assumption that the climate models are calibrated or tuned on large scale features such as, e.g., annual means or global averages.

How to cite: Christiansen, B.: Understanding the distribution of multi-model ensembles, EGU General Assembly 2020, Online, 4–8 May 2020, EGU2020-7807, https://doi.org/10.5194/egusphere-egu2020-7807, 2020

D3014 |
Karin van der Wiel and Richard Bintanja

Weather or climate extreme events disproportionately affect societies and ecosystems. Physical understanding of the impact of global climate change on the occurrence of such extreme events is therefore crucial. Here we separate changes in the occurrence of high-temperature and heavy-precipitation events in a part caused by climatic changes of the mean state and a part caused by climatic changes in variability. We extend the frequently used Probability Ratio (PR) framework, used to quantify changes in the occurrence of extreme events, such that it produces a 'PRmean' value for changes due to a change in mean climate and a 'PRvar' value for changes due to changes in climate variability. Large ensemble climate model simulations are used to quantify changes in extreme events in a 2C warmer world. It is found that the increased occurrence of high-temperature extremes is predominantly caused by the increase of mean temperatures, with a much smaller role for changes in variability (PRmean >> PRvar). The spatial differences are considerable, however, with the polar regions standing out as regions where changes in temperature variability do have a considerable limiting effect on extreme event occurrence. Changes in heavy-precipitation extremes are generally due to changes in both mean climate and variability (PRvar ≈ PRmean). Despite complex feedbacks in the global climate system, the ratio of PRmean to PRvar is largely independent of the event threshold and the climate scenario. These results help to quantify robustness of projected changes in climate extremes, given that projections of changes in the mean state are in many cases much better constrained than projections of changes in variability.

How to cite: van der Wiel, K. and Bintanja, R.: Contribution of climatic changes in means and variability to temperature and precipitation extremes, EGU General Assembly 2020, Online, 4–8 May 2020, EGU2020-7852, https://doi.org/10.5194/egusphere-egu2020-7852, 2020

D3015 |
Fabian von Trentini, Emma E. Aalbers, Erich M. Fischer, and Ralf Ludwig

Single model large ensembles are widely used model experiments to estimate internal climate variability (here: inter-annual variability). The underlying assumption is that the internal variability of the chosen model is a good approximation of the observed natural variability. In this study, for the first time over Europe, we test this assumption based on the comparison of three regional climate model large ensembles (16 members of an EC-EARTH-RACMO ensemble, 21 members of a CESM-CCLM ensemble, 50 members of a CanESM-CRCM ensemble) for four European domains (British Isles, France, Mid-Europe, Alps). Simulated inter-annual variability is evaluated against E-OBS and the inter-annual variability and its future change are compared across the ensembles. Analyses comprise seasonal temperature and precipitation, as well as indicators for dry periods and heat waves. Results show a large consistency of all three ensembles with E-OBS data for most indicators and regions, validating the abilities of these ensembles to represent natural variability on the annual scale. EC-EARTH-RACMO shows the highest inter-annual variability for winter temperature and precipitation, whereas CESM-CCLM shows the highest variability for summer temperature and precipitation, as well as for heatwaves and dry periods. Despite these model differences, the sign of the future changes in internal variability is largely the same in all models: for summer temperature, summer precipitation and the number of heat waves, the internal variability increases, while it decreases for winter temperature. While dry periods reveal a tendency to increase in variability, the changes of winter precipitation remain less conclusive. The overall consistency across single model large ensembles and observations strengthens the concept of large ensembles, and underlines their great potential for understanding and quantifying internal climate variability and its role in climate change dynamics.

How to cite: von Trentini, F., Aalbers, E. E., Fischer, E. M., and Ludwig, R.: Comparing inter-annual variabilities in three regional single model initial-condition large ensembles (SMILE) over Europe, EGU General Assembly 2020, Online, 4–8 May 2020, EGU2020-7119, https://doi.org/10.5194/egusphere-egu2020-7119, 2020

D3016 |
Nathaniel Cresswell-Clay, Caroline C. Ummenhofer, Diana L. Thatcher, Alan D. Wanamaker, and Rhawn F. Denniston

The Azores High is a subtropical high-pressure ridge in the North Atlantic. During boreal winters, anticyclonic winds rotate around the Azores High, transporting moisture to Western Europe. Variability in the size and intensity of the Azores High thus corresponds to variability in hydroclimate across Western Europe. We use the Last Millennium Ensemble (LME), which is run using the Community Earth System Model (CESM) and features thirteen transient simulations covering the period 850 to 2005 A.D. with prescribed external forcing (e.g. greenhouse gas, solar, volcanic, land use, orbital, and aerosol). The LME is shown to accurately simulate the variability and trends in the Azores High when compared to observational records from the 20th century. The Azores High has grown in size during the Industrial Era. This growth is most dramatic when observing the frequency of winters during which the Azores High is extremely large. The LME shows more winters with an extremely large Azores High in the past 100 years than any other 100-year period in the last millennium. Using LME as well as other simulations from the Paleoclimate Modelling Intercomparison Project Phase III, the recent expansion of the Azores High is shown to be well outside the range of natural variability since 850 A.D. Individual forcing simulations within the LME provide smaller ensembles in which only one external forcing is varied. These experiments attribute Azores High expansion to the recent increase in atmospheric greenhouse gas concentrations. Recent hydroclimatic signals across Western Europe consistent with the Azores High variability are also discussed.

How to cite: Cresswell-Clay, N., Ummenhofer, C. C., Thatcher, D. L., Wanamaker, A. D., and Denniston, R. F.: Unprecedented Expansion of the Azores High due to Anthropogenic Climate Change, EGU General Assembly 2020, Online, 4–8 May 2020, EGU2020-6015, https://doi.org/10.5194/egusphere-egu2020-6015, 2020

D3017 |
Felicitas Hansen, Danijel Belusic, and Klaus Wyser

The large-scale atmospheric circulation is one of the most important factors influencing weather and climate conditions on different timescales. Its short- and long-term changes considerably determine both mean and extreme values of surface parameters like temperature or precipitation rates. Future changes of circulation patterns are of particular interest as these may significantly alter or amplify the expected thermodynamic changes due to changing concentrations of greenhouse gases, albedo and land use. We analyse both historical as well as future climate simulations of the SMHI large ensemble (S-LENS) performed with the EC-Earth3 global climate model to examine large-scale circulation situations and their association to extremes in precipitation and temperature over Sweden. Various methods exist to classify mostly sea level pressure or geopotential height fields into characteristic circulation types, and we compare several of these methods for their applicability to represent precipitation and temperature variability over our region of interest. S-LENS consists of a 50-member ensemble for a historical period (1970-2014) and four 50-member climate change scenario ensembles covering the 21st century differing in terms of assumptions made for future radiative forcing development. We study the efficiency of circulation types in the historical period to give rise to extremes, and examine further the frequency and within-type changes of those circulation types associated with extremes by the middle and the end of the 21st century under the different climate change scenarios. S-LENS with its comparatively large number of both multi-decadal scenarios and realizations for each scenario serves as a perfect testbed to study potential changes in events of low frequency within the environment of a single model.

How to cite: Hansen, F., Belusic, D., and Wyser, K.: Future changes of circulation types associated with extremes over Sweden, EGU General Assembly 2020, Online, 4–8 May 2020, EGU2020-17931, https://doi.org/10.5194/egusphere-egu2020-17931, 2020

D3018 |
| Highlight
Karsten Haustein, Benjamin Strauss, Sihan Li, and Friederike Otto

In order to streamline observational and global climate model based extreme event attribution techniques, we propose a multi-method framework which drastically increases the robustness of rapid attribution studies, hence further facilitating the communication of extreme weather related risks across the globe.

We use advanced observational datasets for temperature (Berkeley Earth) and rainfall (CPC), together with CMIP5 simulations and the large HadRM3P ensemble from the weather@home project (W@H) Recent (Climatology) and current/future warming scenarios (1°C, 1.5°C, 2°C, 3°C and 4°C) are juxtaposed to pre-industrial (Natural) baseline conditions.

Two scaling approaches are applied to the observational data to estimate the statistics of future warming scenarios. One in which percentiles of the metric of interest (Tmax, Tmin, Precip) are scaled with Global Mean Surface Temperature (GMST) and another in which the mean is scaled against GMST. Model subsetting (similar to the HAPPI experiment) as function of GMST is applied to the CMIP5 data in order to assign the warming thresholds. W@H scenarios are prescribed to achieve the desired warming threshold. We analyse the results in terms of classes of events, using percentiles, absolute and return-time based thresholds. Before the subsetting, model biases are removed means of quantile-mapping (both for CMIP5 and W@H).

The results between both scaling methods and model subsetting are mostly consistent across many regions and virtually for all temperature thresholds under consideration. The percentile-based scaling method does, however, reveal that the tail of the distributions (highest Tmax, lowest Tmin) has potentially widened with warming. Overall, we find that historically rare extreme events become increasingly common in the future as far as Tmax and Precip is concerned. In contrast, cold extremes become increasingly rare.

How to cite: Haustein, K., Strauss, B., Li, S., and Otto, F.: A consistent multi-method global extreme event attribution framework, EGU General Assembly 2020, Online, 4–8 May 2020, EGU2020-8760, https://doi.org/10.5194/egusphere-egu2020-8760, 2020

D3019 |
| Highlight
Renate Wilcke, Erik Kjellström, Anders Moberg, and Changgui Lin

Long-lasting high-pressure dominated weather resulting in remarkably warm and dry conditions in large parts of northern Europe during summer 2018. As a consequence, Sweden experienced a very long warm period with an unusual high number of warm days, which could be felt in many parts of the society. Groundwater shortage, many extensive forest fires (requiring assistance on European scale), health impacts on people, drought related shortage of food for livestock leading to emergency slaughter in many regions.According to SMHIs weather observations the average over Sweden for the four-month period May-August was on average 3.3K warmer than the 1961-1990 climatological mean.

Here, we evaluate climate conditions in Sweden during the summer 2018 in relation to the historical climate, reaching back to pre-industrial times. Basing the evaluation on long observation time series (150 years for some station across Sweden, and 250 years for Stockholm) as well as on 5 large ensembles from different global models, we want to assess to what extent an extreme event like the summer of 2018 may have changed as a result of global warming.

To grasp the character of summer 2018, not only daily values are considered, but also periods of heat days and heat indices describing the amplitude and length of an event.

With the extended length of the summer season, on account of an exceptional warm May, 2018 sets its record for many heat related indices and would have very unlikely been observed in pre-industrial times according to the given model data.

How to cite: Wilcke, R., Kjellström, E., Moberg, A., and Lin, C.: The extreme warm summer 2018 in Sweden - set in a historical context, EGU General Assembly 2020, Online, 4–8 May 2020, EGU2020-7855, https://doi.org/10.5194/egusphere-egu2020-7855, 2020

D3020 |
Lea Beusch, Lukas Gudmundsson, and Sonia I. Seneviratne

Earth System Models (ESMs) are invaluable tools to study the climate system’s response to a specific greenhouse gas emission scenario, but their projections are associated with internal climate variability and model uncertainty. To account for these uncertainties, large single-model initial-condition ensembles and multi-model ensembles are created and observations are used to constrain their projections. However, ensemble size is usually limited since ESM simulations are computationally costly. Climate change impact and integrated assessment models, on the other hand, could profit from more realizations which are consistent with observations and the associated improved sampling of the constrained phase space.

Here, we employ MESMER, a Modular Earth System Model Emulator with spatially Resolved output, to generate stochastic realizations of land temperature field time series at a yearly resolution at a negligible computational cost (Beusch et al., 2019). MESMER successfully approximates large multi-model initial-condition ensembles on grid-point to regional scales if it is trained with runs from each contained ESM. Here, we create 1000 emulations per ESM for models of the 6th phase of the Coupled Model Intercomparison Project (CMIP6) covering the historical time period and the high-end emission scenario SSP585 (1870 – 2100) (Beusch et al., submitted). The resulting ensemble is referred to as a “superensemble”.

The modular framework of MESMER opens new avenues for validating and constraining ESM ensembles (Beusch et al., submitted). Within the emulator, the local warming signal is expressed as a combination of the global mean temperature trend and the local response to this global trend. These two features can be validated separately by comparison to observations. It is found that ESMs which perform well in terms of global mean temperature trend do not necessarily perform well in terms of local response and vice versa. Additionally, different ESMs perform well in different regions. The most naive approach would be to base temperature projections solely on ESMs which perform well on both global and regional scales. However, this would result in discarding valuable information from many ESMs which perform well at only one of the scales. To circumvent this issue, we therefore propose to use MESMER to combine all global mean temperature trends with all local modules that are consistent with observations. Thereby, we obtain a regionally-optimized “crossbred” superensemble which constitutes a large recombined multi-model initial-condition ensemble and makes full use of all ESM features which are consistent with observations. The regionally diverse behavior of the crossbred superensemble highlights the importance of considering spatially resolved temperature projections.


L. Beusch, L. Gudmundsson, and S. I. Seneviratne: Emulating Earth System Model Temperatures: from Global Mean Temperature Trajectories to Grid-point Level Realizations on Land, doi: 10.5194/esd-2019-34, 2019 (accepted for ESD).

L. Beusch, L. Gudmundsson, and S. I. Seneviratne: Crossbreeding CMIP6 Earth System Models with an Emulator for Regionally-optimized Land Temperature Projections, submitted.

How to cite: Beusch, L., Gudmundsson, L., and Seneviratne, S. I.: Crossbreeding Earth System Models with an Emulator for Regionally-optimized Land Temperature Projections, EGU General Assembly 2020, Online, 4–8 May 2020, EGU2020-9835, https://doi.org/10.5194/egusphere-egu2020-9835, 2020

D3021 |
Andrea Böhnisch, Ralf Ludwig, and Martin Leduc

The ClimEx-project ("Climate change and hydrological extreme events"; www.climex-project.org) provides a single-model initial-condition ensemble that is unprecedented in terms of size, resolution and domain coverage: 50 members of the Canadian Earth System Model version 2 (CanESM2 Large Ensemble, 2.8° spatial resolution) are downscaled using the Canadian Regional Climate Model version 5 (CRCM5 Large Ensemble, 0.11° spatial and up to hourly temporal resolution) over two domains, Europe and northeastern North America. The high-resolution climate information serves as input for hydrological simulations to investigate the impact of internal variability and climate change on hydrometeorological extremes.

This study evaluates the downscaling of a teleconnection which affects northern hemisphere climate variability, the North Atlantic Oscillation (NAO), within the nested single-model large ensemble of the ClimEx project. The overall goal of this study is to assess whether the range of NAO internal variability is represented consistently between the driving global climate model (GCM, i.e., the CanESM2) and the nested regional climate model (RCM, i.e., the CRCM5).

The NAO pressure dipole is quantified in the CanESM2-LE; responses of mean surface air temperature and total precipitation sum to changes in the NAO index are evaluated within a Central European domain in both the CanESM2-LE and the CRCM5-LE. NAO–response relationships are expressed via Pearson correlation coefficients and the change per unit index change for historical (1981–2010) and future (2070–2099) winters.

Results show that statistically robust NAO patterns are found in the CanESM2-LE under current forcing conditions, and reproductions of the NAO flow pattern present in the CanESM2-LE produce plausible temperature and precipitation responses in the high-resolution CRCM5-LE. The NAO–response relationship is more strongly evolved in the CRCM5-LE than in the CanESM2-LE, but the inter-member spread shows no significant differences: thus internal variability expressed as inter-member spread can be seen as being represented consistently between the GCM and RCM. NAO–response relationships weaken in the future period in both the CanESM2-LE and CRCM5-LE, suggesting that the NAO influence on Central European temperature and precipitation decreases.

The results stress the advantages of a single-model ensemble regarding the evaluation of internal variability. They also strengthen the validity of the nested ensemble for further impact modelling using RCM data only, since important large-scale teleconnections present in the driving GCM propagate properly to the fine scale dynamics in the RCM.

How to cite: Böhnisch, A., Ludwig, R., and Leduc, M.: Using a nested single-model large ensemble to assess the internal variability of the North Atlantic Oscillation and its climatic implications for Central Europe, EGU General Assembly 2020, Online, 4–8 May 2020, EGU2020-4925, https://doi.org/10.5194/egusphere-egu2020-4925, 2020

D3022 |
Satoshi Watanabe

In this study, a methodology that uses super ensemble simulation with appropriate bias correction for river planning was proposed. The Database for Policy Decision-Making for Future Climate Change (d4PDF) is a super ensemble experiments that comprise over 1000-year output have been conducted. The d4PDF provides regional downscaling simulation that focuses around Japan. It is expected that the impact assessments of climate changes on various fields considering uncertainly are conducted.

The impact of climate change on floods is a serious issue. In Japan, all class A river has design rainfall for the river planning that is defined considering historical observations of precipitation that happens once in several hundred years, which the planning year is different depending on the situation of a river. The design rainfall provides the fundamental information for planning river management.  The Ministry of Land, Infrastructure, Transportation and Tourism defines the value of the rainfall in the planning year in each class A river basin by considering the hydro-meteorological and social characteristics of each basin. As the design rainfall was defined in the mid-1900s for most of the rivers, the method to estimate precipitation in the planning year was conducted with limited observation data using extreme statistical value. The super ensemble simulation data is expected to contribute for the decision making with appropriate setting of design rainfall.

We proposed a method to correct the bias of super ensemble simulation and estimated the design rainfall in 47 river basins selected from class A river basins. The estimated design rainfall was compared between the one estimated with super ensemble simulation and the one estimated with conventional approach. The spread of results oriented from super ensemble simulation indicated that uncertainly of design rainfall estimated with conventional approach was so high that the consideration of uncertainty is necessary for river planning. The experiments indicated that the use of super ensemble simulation with appropriate bias correction could provide knowledge that aids us in understanding the hydrological extremes.

How to cite: Watanabe, S.: An application of super ensemble simulation with appropriate bias correction for river planning in Japan, EGU General Assembly 2020, Online, 4–8 May 2020, EGU2020-12875, https://doi.org/10.5194/egusphere-egu2020-12875, 2020

D3023 |
Peter Watson, Sarah Sparrow, William Ingram, Simon Wilson, Drouard Marie, Giuseppe Zappa, Richard Jones, Daniel Mitchell, Tim Woollings, and Myles Allen

Multi-thousand member climate model simulations are highly valuable for showing how extreme weather events will change as the climate changes, using a physically-based approach. However, until now, studies using such an approach have been limited to using models with a resolution much coarser than the most modern systems. We have developed a global atmospheric model with 5/6°x5/9° resolution (~60km in middle latitudes) that can be run in the climateprediction.net distributed computing system to produce such large datasets. This resolution is finer than that of many current global climate models and sufficient for good simulation of extratropical synoptic features such as storms. It will also allow many extratropical extreme weather events to be simulated without requiring regional downscaling. We will show that this model's simulation of extratropical weather is competitive with that in other current models. We will also present results from the first multi-thousand member ensembles produced at this resolution, showing the impact of 1.5°C and 2°C global warming on extreme winter rainfall and extratropical cyclones in Europe.

How to cite: Watson, P., Sparrow, S., Ingram, W., Wilson, S., Marie, D., Zappa, G., Jones, R., Mitchell, D., Woollings, T., and Allen, M.: Multi-thousand member ensemble atmospheric simulations with global 60km resolution using climateprediction.net, EGU General Assembly 2020, Online, 4–8 May 2020, EGU2020-10895, https://doi.org/10.5194/egusphere-egu2020-10895, 2020

D3024 |
Tamas Bodai, Gabor Drotos, Matyas Herein, Frank Lunkeit, and Valerio Lucarini

We study the teleconnection between the El Niño–Southern Oscillation (ENSO) and the Indian summer monsoon (IM) in large ensemble simulations, the Max Planck Institute Earth System Model (MPI-ESM) and the Community Earth System Model (CESM1). We characterize ENSO by the JJA Niño 3 box-average SST and the IM by the JJAS average precipitation over India, and define their teleconnection in a changing climate as an ensemble-wise correlation. To test robustness, we also consider somewhat different variables that can characterize ENSO and the IM. We utilize ensembles converged to the system’s snapshot attractor for analyzing possible changes in the teleconnection. Our main finding is that the teleconnection strength is typically increasing on the long term in view of appropriately revised ensemble-wise indices. Indices involving a more western part of the Pacific reveal, furthermore, a short-term but rather strong increase in strength followed by some decrease at the turn of the century. Using the station-based SOI as opposed to area-based indices leads to the identification of somewhat more erratic trends, but the turn-of-the-century “bump” is well-detectable with it. All this is in contrast, if not in contradiction, with the discussion in the literature of a weakening teleconnection in the late 20th century. We show here that this discrepancy can be due to any of three reasons: ensemble-wise and temporal correlation coefficients used in the literature are different quantities; the temporal moving correlation has a high statistical variability but possibly also persistence; MPI-ESM does not represent the Earth system faithfully.

How to cite: Bodai, T., Drotos, G., Herein, M., Lunkeit, F., and Lucarini, V.: The forced response of the El Niño–Southern Oscillation-Indian monsoon teleconnection in ensembles of Earth System Models, EGU General Assembly 2020, Online, 4–8 May 2020, EGU2020-12061, https://doi.org/10.5194/egusphere-egu2020-12061, 2020

D3025 |
Gabor Drotos

The availability of a large ensemble enables one to evaluate empirical orthogonal functions (EOFs) with respect to the ensemble without relying on temporal variability at all. Variability across the ensemble at any given time is supposed to represent the most relevant probability distribution for climate-related studies, and this distribution is presumably subject to temporal changes in the presence of time-dependent forcing. Such changes may be observable in spatial patterns of ensemble-based EOFs and associated eigenvalues. Unfortunately, estimates of these changes come with a considerable error due to the finite size of the ensemble, so that associating a significance level with the presence of a change (with respect to a null hypothesis about the absence of any change) should be the first step of analyzing the time evolution.

It turns out, however, that the conditions for the applicability of usual hypothesis tests about stationarity are not satisfied for the above-mentioned quantities. What proves to be feasible is to estimate an upper bound on the significance level for nonstationarity. This means that the true significance level would ideally be lower or equal to what is estimated, which would prevent unjustified confidence in the detection of nonstationarity (i.e., falsely rejecting the null hypothesis could not become more probable than claimed). Most importantly, one would avoid seriously overconfident conclusions about the sign of the change in this way. Notwithstanding, the estimate for the upper bound on the significance level is also affected by the finite number of the ensemble members. It nevertheless becomes more and more precise for increasing ensemble size and may serve as a first guidance for currently available ensemble sizes.

The details of the estimation are presented in the example of the EOF-based analysis of the El Niño–Southern Oscillation (ENSO) as it appears in the historical and RCP8.5 simulations of the Max Planck Institute Grand Ensemble. A comparison between results including and excluding ensemble members initialized with an incomplete spinup in system components with a long time scale is also given.

How to cite: Drotos, G.: On associating significance levels with temporal changes in empirical orthogonal function analysis: a case study for ENSO, EGU General Assembly 2020, Online, 4–8 May 2020, EGU2020-20730, https://doi.org/10.5194/egusphere-egu2020-20730, 2020

D3026 |
Gerhard Smiatek and Harald Kunstmann

The summer 2018 was extremely dry and hot in Germany and many parts of Europe. We investigate to which extend SST increases in the North Atlantic Ocean and sea ice extent decreases in the polar sea influence such extremes. We simulate in total the four years 1998, 2003, 2014 and 2015 as years with cool, extremely warm, warm and average SST by multiple integrations of the Model for Prediction Across Scales (MPAS). For each year we perform 30 global MPAS runs in approximately 60 km resolution with SST and sea ice extent from ERA-Interim data as boundary condition. The runs are initialized on different days in December and run until the following September 1st.

The contribution investigates the results obtained from the total of 120 simulations. It discusses the resulting probability density functions (PDF) and changes in the summer precipitation and temperature in connection to changes in the summer North Atlantic Oscillation (sNAO). The results indicate that the SST and sea ice extent influence the range and mean values of the precipitation and temperature distribution functions.  Extreme values, however, occur with both cool and warm SST.

How to cite: Smiatek, G. and Kunstmann, H.: Influence of the SST increase and sea ice extent decrease on extreme summer temperatures and precipitation in Central Europe, EGU General Assembly 2020, Online, 4–8 May 2020, EGU2020-19888, https://doi.org/10.5194/egusphere-egu2020-19888, 2020

D3027 |
Ralf Hand, Jürgen Bader, Daniela Matei, Rohit Ghosch, and Johann Jungclaus

The question, whether ocean dynamics are relevant for basin-scale North Atlantic decadal temperature variability is subject of ongoing discussions. Here, we analyze a set of simulations with a single climate model, consisting of a 2000-year pre-industrial control experiment, a 100-member historical ensemble, and a 100-member ensemble forced with an incremental CO2 increase by 1%/year. Compared to previous approaches, our setup offers the following advantages: First, the large ensemble size allows to robustly separate internally and externally forced variability and to robustly detect statistical links between different quantities. Second, the availability of different scenarios allows to investigate the role of the background state for drivers of the
variability. We find strong evidence that ocean dynamics, particularly ocean heat transport variations, form an important contribution to generate the Atlantic Multidecadal Variability (AMV) in the Max Planck Institute Earth System Model (MPI- ESM). Particularly the Northwest North Atlantic is substantially affected by ocean circulation for the historical and pre-industrial simulations. Anomalies of the Labrador Sea deep ocean density precede a change of the Atlantic Meridional Overturning Circulation (AMOC) and heat advection to the region south of Greenland.
Under strong CO2 forcing the AMV-SST regression pattern shows crucial changes: SST variability in the north western part of the North Atlantic is strongly reduced, so that the AMV pattern in this scenario is dominated by the low-latitude branch. We found a connection to changes in the deep water formation, that cause a strong reduction of the mean AMOC and its variability. Consequently, ocean heat transport convergence becomes less important for the SST variability south of Greenland.

How to cite: Hand, R., Bader, J., Matei, D., Ghosch, R., and Jungclaus, J.: Changes of decadal SST Variations in the subpolar North Atlantic under strong CO2 forcing as an indicator for the ocean circulation’s contribution to Atlantic Multidecadal Variability, EGU General Assembly 2020, Online, 4–8 May 2020, EGU2020-10091, https://doi.org/10.5194/egusphere-egu2020-10091, 2020

D3028 |
Stefan Brönnimann, Ralf Hand, Jörg Franke, and Andrey Martynov

The recently started PALAEO-RA project aims at creating a new global monthly 3-dimensional reanalysis dataset of the past 600 years' climate. Large spatial and temporal gaps in the available historical data on these time scale make the climate history being an under-determined problem when using observations only. In PALAEO-RA we will addionally use information from an ensemble of simulations with an atmospheric general circulation model (AGCM). The model offers additional physical constraints. The model reproduces teleconnection patterns and reflects typical large-scale modes of variability to set the historical data into a physically consistent regional to global context.

In brief, the method that we plan to use consists of two steps: First, we are currently producing an  ensemble of historical simulations with the atmospheric general circulation model ECHAM6. Once finished, it will have a size of ca. 30 members, covering the period fom 1420 to present. The ensemble is supposed to reflect the range of realistic climate states under prescribed historical radiative forcings (based on the PMIP4 setup) and ocean boundary conditions (HadISST.2 & SST reconstructions by Samakinwa et al., see abstract EGU2020-8744).

Secondly, we will apply Ensemble Kalman Fitting, a technique for the offline assimilation of historical observations (instrumental observations, documentary data, tree ring width and other proxies), basing on the assumption that the occurrence of a distinct observation has a different probability depending on the meso- and large-scale circulation patterns of the atmosphere.

Our poster will give a brief overview on the project with a focus on introducing the AGCM ensemble, also to allow for discussions on further applications of the latter.

How to cite: Brönnimann, S., Hand, R., Franke, J., and Martynov, A.: PALAEO-RA: Combining an intermediate-size AGCM ensemble with historical observations and proxies to create a new dataset of the past 600 years of climate history, EGU General Assembly 2020, Online, 4–8 May 2020, EGU2020-21389, https://doi.org/10.5194/egusphere-egu2020-21389, 2020

D3029 |
Tímea Haszpra, Gábor Drótos, Dániel Topál, and Mátyás Herein

Different teleconnection index time series are obtained even within a single member of a large ensemble simulation if different base periods are chosen. This also has an effect on the apparent strength of teleconnections. In this study, the reasons behind this caveat are discussed analytically and exemplified for the Arctic Oscillation (AO). Additionally, a solution is presented in the so-called snapshot framework using large ensemble simulations.

The AO is the leading mode of atmospheric variability in the Northern Hemisphere winter. Traditionally, its loading pattern is defined as the leading mode of the empirical orthogonal function (EOF) analysis of sea-level pressure (SLP) from 20° to 90° N for a given base period. The AO index (AOI) time series is constructed by projecting the SLP anomalies on this loading pattern and is standardized for the base period. The strength of the linkages related to AO is generally defined by a correlation coefficient between time series of the AOI and another meteorological variable.

Using the CESM-LE and the MPI-GE, we show that the utilization of different base periods within a single member often results in AOI time series differing by as much as 0.5–0.8. We reveal why such differences can arise in any EOF-based quantity: (1) The loading pattern represents a standing oscillation pattern, treated as constant within the studied time interval, and the time evolution of the corresponding index (e.g. AOI) shows how this pattern oscillates in time. However, when the climate changes, stationarity cannot be assumed: whether the oscillation pattern and its amplitude remain the same within the given time interval is dubious. (2) Any shift in the index time series originates from a change in the mean state of the climate system, e.g., from the change in the temporal mean of the SLP field, which is the center of the oscillation described by a given EOF mode. Beyond the meteorological reasons we also give analytically derived results for the shift and for the change in the oscillation amplitude.

To avoid the problems resulting from the assumption of a constant pattern and climatological mean, the traditional EOF-based description should be replaced by the recently developed snapshot EOF (SEOF) analysis if an ensemble is available (Haszpra et al. 2019). This method carries out the EOF analysis across the ensemble at each time instant, instead of the time dimension within each member. As a consequence, instantaneous anomalies originating from internal variability are compared only to the set of states permitted by the climate system at the given time instant. Therefore, beyond a correct characterization at each time instant, the time-dependence of an oscillation pattern and the corresponding amplitude can also be monitored. Furthermore, instantaneous correlation coefficients between the instantaneous index and another variable can be computed across the ensemble to reveal the correct teleconnection strengths and their time-dependence. 

Haszpra et al. 2019: On the time evolution of the Arctic Oscillation and related wintertime phenomena under different forcing scenarios in an ensemble approach. J. Clim. (submitted)

How to cite: Haszpra, T., Drótos, G., Topál, D., and Herein, M.: The role of the base period in evaluating teleconnection indices and strengths, and how to eliminate it in the snapshot framework using large ensembles, EGU General Assembly 2020, Online, 4–8 May 2020, EGU2020-2894, https://doi.org/10.5194/egusphere-egu2020-2894, 2020