Search

We develop a data assimilation scheme with the Icosahedral Non-hydrostatic Earth System Model (ICON-ESM) for operational decadal and seasonal climate predictions at the German weather service. For this purpose, we implement an Ensemble Kalman Filter to the ocean component as a first step towards a weakly coupled data assimilation. We performed an assimilation experiment over the period 1960–2014. This ocean-only assimilation experiment serves to initialize 10-year long retrospective predictions (hindcasts) started each year on 1 November. On multi-annual time scales, we find predictability of sea surface temperature and salinity as well as oceanic heat and salt contents especially in the North Atlantic. The mean Atlantic Meridional Overturning Circulation is realistic and the variability is stable during the assimilation. On seasonal time scales, we find high predictive skill in the tropics with highest values in variables related to the El Niño/Southern Oscillation phenomenon. In the Arctic, the hindcasts correctly represent the decreasing sea ice trend in winter and, to a lesser degree, also in summer, although sea ice concentration is generally much too low in both hemispheres in summer. However, compared to other prediction systems, prediction skill is relatively low in regions apart from the tropical Pacific due to the missing atmospheric assimilation. Further improvements of the simulated mean state of ICON-ESM, e.g. through fine-tuning of the sea ice and the oceanic circulation in the Southern Ocean, are expected to improve the predictive skill. In general, we demonstrate that our data assimilation method is successfully initializing the oceanic component of the climate system.

Ocean heat content observations in the Indian Ocean have revealed distinctive periods of significant multi-decadal trends — for example a cooling between 1990 and 1999 followed by an unprecedented warming between 2000 and 2009. However, a systematic assessment of the relative importance of anthropogenic forcings versus natural variability in driving such trends is still missing. Here, we utilise four state-of-the-art Single Model Initial- Condition Large Ensembles with MPI-ESM1.2-LR containing different factual and counterfactual forcing scenarios to address the problem. We are able to robustly attribute the unprecedented warming of the Indian Ocean between 2000 and 2009 to the increasing anthropogenic greenhouse gas emissions. Our results also reveal that the preceding cooling is likely to be intrinsic to Indian Ocean heat content variability, since none of the applied counterfactual scenarios exhibits such an observed decrease in Indian Ocean heat content. Furthermore, we trace the underlying reasons for the observed inherent cooling between 1990 and 1999 to a significant reduction in heat transported into the Indian Ocean from the Pacific Ocean by the Indonesian Throughflow. These results have implications for decadal predictions of Indian Ocean heat content.

Earth system models are now regularly being used in inter-annual to decadal climate prediction. Such prediction systems based on CMIP5-generation Earth system models had demonstrated an overall positive impact of initialization, i.e. deriving initial conditions of retrospective forecasts from a separate data assimilation experiment, on decadal prediction skill. This view is now being increasingly challenged in the context of improvements both in CMIP6-generation Earth system models and CMIP6-evaluation of external forcing as well as in the context of ongoing transient climate change. In this study we re-evaluate the impact of atmospheric and oceanic initialization on decadal prediction skill of North Atlantic upper ocean heat content (0-700m) in a CMIP6-generation decadal prediction system based on the Max Planck Institute Earth system model (MPI-ESM). We compare the impact of initial conditions derived through full-field atmospheric nudging with those derived through an additional assimilation of observed oceanic temperature and salinity profiles using an ensemble Kalman filter. Our experiments suggest that assimilation of observed oceanic temperature and salinity profiles into the model reduces the warm bias in the subpolar North Atlantic heat content, and improves the modelled variability of the Atlantic meridional overturning circulation and ocean heat transport. These improvements enable a proper initialization of model variables which leads to an improved decadal prediction of surface temperatures. Our results reveal an important role of subsurface oceanic observations in decadal prediction of surface temperatures in the subpolar North Atlantic even in CMIP6-generation decadal prediction system.

Slowly varying ocean heat content is one of the most important variables when describing cli-mate variability on interannual to decadal time scales. Since observation-based estimates ofocean heat content require extensive observational coverage, incomplete observations are oftencombined with numerical models via data assimilation to simulate the evolution of oceanic heat.However, incomplete observations, particularly in the subsurface ocean, lead to large uncertain-ties in the resulting model-based estimate. As an alternative approach, Kadow et al (2020) haveproven that artificial intelligence can successfully be utilized to reconstruct missing climate in-formation for surface temperatures. In the following, we investigate the possibility to train theirthree-dimensional convolutional neural network to reconstruct missing subsurface temperaturesto obtain ocean heat content estimates with a focus on the North Atlantic ocean.The network is trained and tested to reconstruct a 16 member Ensemble Kalman Filter assimi-lation ensemble constructed with the Max-Planck Institute Earth System Model for the periodfrom 1958 to 2020. Specifically, we examine whether the partial convolutional U-net representsa valid alternative to the Ensemble Kalman Filter assimilation to estimate North Atlantic sub-polar gyre ocean heat content.The neural network is capable of reproducing the assimilation reduced to datapoints with ob-servational coverages within its ensemble spread with a correlation coefficient of 0.93 over theentire time period and of 0.99 over 2004 – 2020 (the Argo-Era). Additionally, the network isable to reconstruct the observed ocean heat content directly from observations for 12 additionalmonths with a correlation of 0.97, essentially replacing the assimilation experiment by an extrap-olation. When reconstructing the pre-Argo-Era, the network is only trained with assimilationsfrom the Argo-Era. The lower correlation in the resulting reconstruction indicates higher un-certainties in the assimilation outside of its ensemble spread at times with low observationaldensity. These uncertainties are highlighted by inconsistencies in the assimilation’s represen-tations of the North Atlantic Current at times and grid points without observations detectedby the neural network. Our results demonstrate that a neural network is not only capable ofreproducing the observed ocean heat content over the training period, but also before and aftermaking the neural network a suitable candidate to step-wise extend or replace data assimilation.

Variable fluxes of anthropogenic CO2 emissions into the land and the ocean and the remaining proportion in the atmosphere reflect on the global carbon budget variations and further modulate global climate change. A more accurate reconstruction of the global carbon budget in the past decades and a more reliable prediction of the variations in the next years are crucial for assessing the effectiveness of climate change mitigation policies and supporting global carbon stocktaking and monitoring in compliance with the goals of the Paris Agreement.In this study, we investigate reconstructions and predictions of the CO2 fluxes and atmospheric CO2 growth from ensemble prediction simulations using 5 Earth System Model (ESM) - based decadal prediction systems. These novel prediction systems driven by CO2 emissions with an interactive carbon cycle enable prognostic atmospheric CO2 and represent atmospheric CO2 growth variations in response to the strength of CO2 fluxes into the ocean and the land, which are missing in the conventional concentration-driven decadal prediction systems with prescribed atmospheric CO2 concentration.The reconstructions generated by assimilating physical ocean and atmosphere data products into the prediction systems are able to reproduce the annual mean historical variations of the CO2 fluxes and atmospheric CO2 growth. Multi-model ensemble means best match the assessments of CO2 fluxes and atmospheric CO2 growth rate from the Global Carbon Project with correlations of 0.79, 0.82, and 0.98 for atmospheric CO2 growth rate, air-land CO2 fluxes, and air-sea CO2 fluxes, respectively. The CO2 emission-driven prediction systems with an interactive carbon cycle still maintain the predictive skill of CO2 fluxes and atmospheric CO2 growth as found in conventional concentration-driven prediction systems, i.e., about 2 years for the air-land CO2 fluxes and atmospheric CO2 growth, the air-sea CO2 fluxes have higher skill up to 5 years. The ESM-based prediction systems are capable to reconstruct and predict the variations in the global carbon cycle and hence are powerful tools for supporting carbon budgeting and monitoring, especially in the decarbonization processes. Furthermore, we investigate the contribution of uncertainty in the predictions of CO2 fluxes and atmospheric CO2 growth rate from internal climate variability, different model responses, and emission-forcing reductions to identify the prominent challenge in limiting the skill of CO2 predictions.

We present the CMIP6 version of the Max Planck Institute-Grand Ensemble (MPI-GE CMIP6) with 30 realisations for the historical period and five emission scenarios. The power of MPI-GE CMIP6 goes beyond its predecessor ensemble MPI-GE by providing high-frequency model output, the full range of emission scenarios including the highly policy relevant scenarios SSP1-1.9 and SSP1-2.6, and the opportunity to compare the ensemble to high resolution simulations of the same model version. We demonstrate with six novel application examples how to use the power of MPI-GE CMIP6 to better quantify and understand present and future extreme events in the Earth system, to inform about uncertainty in approaching Paris Agreement global warming limits, and to combine large ensembles and artificial intelligence. For instance, MPI-GE CMIP6 allows us to show that the recently observed Siberian and Pacific North American heat waves are projected to occur every year in 2071-2100 in high-emission scenarios, that the storm activity in most tropical to mid-latitude oceans is projected to decrease, and that the ensemble is sufficiently large to be used for infilling surface temperature observations with artificial intelligence.

Current sub-seasonal prediction systems are traditionally based on models developed for numerical weather prediction. We present a different approach wherein we develop a sub-seasonal prediction system using a coupled Earth system model, the Max-Planck-Institute Earth system model (MPI-ESM), developed primarily for the use in climate prediction. We present results from initialized sub-seasonal reforecasts for the time period 1993-2017 from a 1st generation (CMIP5) seasonal-turned-sub-seasonal prediction system based on MPI-ESM including different components of the Earth system: atmosphere, land surface, ocean, and marine ecosystems. With our system we find (1) that atmospheric variables can be predicted with a quality and prediction horizon similar to what is found within the range of current sub-seasonal to seasonal prediction systems, (2) that extreme events as diverse as heatwaves over land, storm severity over Europe, and sudden stratospheric warmings can be skilfully predicted one to a few weeks ahead, (3) that sea surface temperatures can be skilfully predicted in the majority of large marine ecosystems for several weeks ahead, and (4) that sea ice area in the majority of Arctic seas can be skillfully predicted several weeks ahead. Our findings indicate that a coupled Earth system model like MPI-ESM can already be seamlessly used for sub-seasonal to seasonal (to decadal) climate predictions of different domains of the Earth system. Ultimately these results ask for the seamless approach to be embedded into the development of future coupled Earth system models.

Decadal climate predictions use state-of-the-art climate models and combine them with initialisation procedures to create information about our future. Their development has proven successful in the past years and offer a wide range of applications. One of them is the option to learn about the used climate models. With predictions usually aiming at time periods up to ten lead years it is often assumed that initialisation will wear off over time and the model will regress to results comparable to uninitialised simulations.This contribution investigates decadal predictions over lead times of up to twenty years. The decadal prediction system is based on the Max Planck Institute Earth system model (MPI-ESM), uses atmospheric nudging and an oceanic Ensemble Kalman filter for initialisation and is applied for periods from 1960 onwards. We demonstrate that the effect of initialisation within the prediction can be found for long lead years and does not necessarily regresses back to the uninitialised simulation.We show that in some areas the prediction skill varies over time, while in others it persists or drops quickly. Examples are a consistently increased prediction skill compared to historical simulations in the North East Pacific or decreased prediction skill for lead years longer than ten in the South Atlantic. We also take a look at the Atlantic Meridional Overturning Circulation (AMOC) and its development over time. We show that the AMOC drifts on short time scales towards a new state, which is reached after about ten lead years. For decadal predictions with MPI-ESM we find that for large areas of the globe the correct determination of future developments of external forcings plays an important role. This asks the question whether the current approach to hindcasts is appropriate to determine our capability to predict the future.

We evaluate the prediction skill of the Max Planck Institute Earth System Model (MPI-ESM) decadal hindcast system for German Bight storm activity (GBSA) on a multiannual to decadal scale. We define GBSA every year via the most extreme 3-hourly geostrophic wind speeds, which are derived from mean sea-level pressure (MSLP) data. Our 64-member ensemble of annually initialized hindcast simulations spans the time period 1960–2018. For this period, we compare deterministically and probabilistically predicted winter MSLP anomalies and annual GBSA with a lead time of up to 10 years against observations. The model produces poor deterministic predictions of GBSA and winter MSLP anomalies for individual years but fair predictions for longer averaging periods. A similar but smaller skill difference between short and long averaging periods also emerges for probabilistic predictions of high storm activity. At long averaging periods (longer than 5 years), the model is more skillful than persistence- and climatology-based predictions. For short aggregation periods (4 years and less), probabilistic predictions are more skillful than persistence but insignificantly differ from climatological predictions. We therefore conclude that, for the German Bight, probabilistic decadal predictions (based on a large ensemble) of high storm activity are skillful for averaging periods longer than 5 years. Notably, a differentiation between low, moderate, and high storm activity is necessary to expose this skill.

Ensemble forecasting has become a standard means to obtain information about forecast uncertainties in meteorological centres across the world. The large datasets generated by ensemble prediction systems carry much information that is difficult to analyse manually – here, techniques from the field of artificial intelligence can be beneficial to aid the analysis. Cluster analysis is one commonly used (unsupervised machine learning) approach to automatically determine distinct scenarios in numerical weather forecasting ensembles, both in atmospheric research and operational forecasting. Typically, a cluster analysis focusses on a selected meteorological forecast variable, a specific region, and time (or a time window). The dimensionality of the data is reduced by techniques like principal component analysis, and a clustering algorithm – typically k-means – is applied to the reduced data set. Challenges with such an approach arise through the determined clusters often being sensitive to factors including the selected region, forecast variable, and algorithm parameters, and also through the employed algorithms often appearing as a “black box” to the user. In our work, we attempt to make the clustering process more transparent by providing a visual analysis framework to analyse the sensitivity of generated clusters with respect to various factors. The presented framework is coupled to the open-source meteorological ensemble visualization software Met.3D, allowing for interactive specification of clustering parameters and for interactive visual analysis, including 3-D elements. A case study using ensemble prediction data of sudden stratospheric warmings (SSWs) is presented, demonstrating how visualizing similarity between clusterings with different parameters can aid the interpretation of the data.

The Atlantic Meridional Overturning Circulation (AMOC) is a crucial feature of Earth’s climate, and the possible recent AMOC weakening in association with anthropogenic emissions is a topic of current scientific discussion. To assess whether anthropogenic emissions affect the circulation, we examine low AMOC strengths with the framework of the causal counterfactual theory. We compare the occurrence of low strengths in a factual world with all forcings present against a counterfactual world, equivalent except for anthropogenic forcings. To represent these two worlds, we use past and future 30-member ensembles of simulations from the Max Planck Institute Earth System Model (MPI-ESM1.2-LR), whose AMOC characteristics at 26.5°N are in overall agreement with observations from the RAPID program.Considering the whole historical period (1850-2020), our results show that causation probabilities for anthropogenic forcings on low AMOC strengths at 26.5 °N are generally low. However, within future projections, we find that anthropogenic forcings will very likely be a necessary cause of any AMOC strength below 16.0 Sv by about 2040. These results are sensitive to the choice of attribution window: if the starting year is chosen closer to the present, all probabilities substantially increase. Within the analyzed simulations, the main attribution results found at 26.5°N are confirmed for other latitudes in the North Atlantic.

We investigate the potential for enhancing the seasonal prediction skill of mid-latitude cyclonic activity, represented by eddy kinetic energy (EKE) at 250 hPa over the North Atlantic and Europe, in hindcast simulations with the Max Planck Institute Earth System Model (MPI-ESM) against the ECMWF ERA5 reanalysis. Our analysis focuses on wintertime months (December-March) from 1982 to 2019, with a 30-member seasonal hindcast ensemble initialized every November 1st. Based on the initial confirmation that in both ERA5 reanalysis and MPI-ESM hindcasts, the eddy-driven jet stream and the wintertime North Atlantic Oscillation (NAO) play a significant role in wintertime's spatial and temporal distribution of mid-latitude cyclonic activity, we perform ensemble subsampling.Specifically, we sample each winter so that a northern position of the jet stream is consistent with a positive phase of the NAO and represents poleward enhanced EKE activity. In contrast, a southern position of the jet stream is consistent with a negative phase of the NAO and represents equatorward enhanced EKE activity. Preliminary analysis of the predictive skill of MPI-ESM hindcasts with respect to ERA5 shows that such subsampling with respect to a consistent representation of the jet stream position and the NAO phase leads to improvements over the skill from the 30-member ensemble mean, with significant correlations concentrated over areas of major frequency of storm tracks. Our results put into practical use that an enhanced representation of the large-scale climate variability plays a crucial role in the long-term prediction of high-frequency events such as mid-latitude cyclones.

The correct representation of wind speeds at hub height (e.g., 100 m above ground) is becoming more and more important with respect to the expansion of renewable energy. In this study, a post-processing of the wind speed of the regional reanalysis COSMO-REA6 in Central Europe is performed based on a combined physical and statistical approach. The physical basis is provided by downscaling wind speeds with the help of a diagnostic wind model, which reduces the horizontal grid point spacing by a factor of 8 compared to COSMO-REA6 and considers different vertical atmospheric stabilities. In the second step, a statistical correction is performed using a neural network, as well as a generalized linear model based on different variables of the reanalysis. Although only a few measurements by masts or lidars are available at hub height, an improvement of the wind speed in the root-mean-squared error of almost 30 % can be achieved. A final comparison with radiosonde observations confirms the added value of combining the physical and statistical approaches in post-processing the wind speed.

The purpose of this study is to perform verification of the structural characteristics of high-resolution spatial forecasts without relying on an object identification algorithm. To this end, a wavelet approach developed for image texture analysis is applied to an ensemble of high-resolution quantitative precipitation forecasts. The forecasts are verified against estimates from a high-resolution regional reanalysis with a similar model version. The wavelet approach estimates an averaged wavelet spectrum for each spatial field of the ensemble forecasts and the reanalysis, thereby removing all information on the localization of precipitation and investigating solely the overall structure of forecasts and reanalysis. In order to assess skill using a multivariate score, an additional reduction of dimensionality is needed. This is performed using singular vectors from a linear discriminant analysis as it favors data compression in the direction where the ensemble is most discriminating. We discuss implications of this strategy, show that the averaged wavelet spectra give valuable information on forecast performance. The skill difference between a so-called perfect forecast using for verification a member of the ensemble, and the non-perfect forecast using the reanalysis points to significant deficiencies of the forecast ensemble. Overall, the discriminating power solely based on global spectral information is remarkable, and the COSMO-DE-EPS is a quite good forecast ensemble with respect to the reanalysis.

In November 2020, the new version of the German Climate Forecast System, GCFS2.1, became operational at Deutscher Wetterdienst (DWD), providing new seasonal forecasts every month. The system is based on the Max Planck Institute for Meteorology Earth-System Model (MPI-ESM-HR) and is developed jointly by DWD, the Max Planck Institute for Meteorology and Universität Hamburg.In GCFS2.1, ERA5 and ORAS5 reanalyses are assimilated using atmospheric, oceanic and sea ice nudging, respectively. From the assimilation, 50-member 6-month forecast ensembles are initialized at the start of each month. Prediction skill is assessed with a 30-member 6-month hindcast ensemble covering the time period 1982-2019 for February, May, August and November start months, and 1990-2019 for the remaining start months. Both the forecast and hindcast ensembles are generated by oceanic bred vectors with additional physical perturbations applied to the upper atmospheric model layers.Here, we investigate the performance of GCFS2.1 summer and winter forecasts over Europe. While our main focus is on the prediction of large scale patterns that control the weather regimes during these two seasons, e.g. European blockings, special emphasis is paid on the impact of the January 2021 sudden stratospheric warming (SSW) event on the performance of GCFS2.1. The inclusion of the early phases of the January 2021 SSW event in the forecast initialisation significantly changes the GCFS2.1 forecast for February 2021 European surface climate. Prediction skill of GCFS2.1 for summer European blocking events will be also compared to the previous version GCFS2.0.

We analyse the connections between the wintertime North Atlantic Oscillation (NAO), the eddy-driven jet stream with the mid-latitude cyclonic activity over the North Atlantic and Europe. We investigate, through the comparison against ECMWF ERA5 and hindcast simulations from the Max Planck Institute Earth System Model (MPI-ESM), the potential for enhancement of the seasonal prediction skill of the Eddy Kinetic Energy (EKE) by accounting for the connections between large-scale climate and the regional cyclonic activity. Our analysis focuses on the wintertime months (December-March) in the 1979-2019 period, with seasonal predictions initialized every November 1st. We calculate EKE from wind speeds at 250 hPa, which we use as a proxy for cyclonic activity. The zonal and meridional wind speeds are bandpass filtered with a cut-off at 3-10 days to fit with the average lifespan of mid-latitude cyclones. Preliminary results suggest that in ERA5, major positive anomalies in EKE, both in quantity and duration, are correlated with a northern position of the jet stream and a positive phase of the NAO. Apparently, a deepened Icelandic low-pressure system offers favourable conditions for mid-latitude cyclones in terms of growth and average lifespan. In contrast, negative anomalies in EKE over the North Atlantic and Central Europe are associated with a more equatorward jet stream, these are also linked to a negative phase of the NAO. Thus, in ERA5, the eddy-driven jet stream and the NAO play a significant role in the spatial and temporal distribution of wintertime mid-latitude cyclonic activity over the North Atlantic and Europe. We extend this connection to the MPI-ESM hindcast simulations and present an analysis of their predictive skill of EKE for wintertime months.

North Atlantic climate variability is dominated by two important subsystems, the Atlantic Meridional Overturning Circulation (AMOC) and the Sub-Polar Gyre (SPG). While the AMOC is responsible for the transport of mass and heat into higher latitudes, SPG has been linked with large-scale changes in the subpolar marine environment. The changes in strength, intensity and positions of the constituent currents of the SPG impose variabilities in the distribution of heat and salt in the North Atlantic Ocean. Consequently, the predictability on decadal scales of the two subsystems is of huge importance for the understanding of variability in the North Atlantic.Our contribution investigates the decadal and multi-decadal predictability of these subsystems within the Max Planck Institute for Meteorology Earth System Model (MPI-ESM). We analyse the model’s capability to predict these subsystems as well as the dependence of the two subsystems on each other. These investigations open new opportunities for a better understanding of the impact of the North Atlantic onto important marine ecosystems and its changes in the upcoming decade.

The North Atlantic Oscillation (NAO) has over the year a major influence on European weather. In many applications, being it in modern or paleo climate science, the NAO is assumed to varying in strength, but otherwise often understood as being a constant feature of the pressure system over the North Atlantic. In recent years investigations on the seasonal-predictability of the winter NAO has shown that the prediction skill is varying over time. This opens the question, why this is the case and how well models are able to represent the NAO in all its variability over the 20th century.To investigate this further we take a look at a seasonal prediction of the NAO with the Max Planck Institute Earth System Model (MPI-ESM) seasonal prediction system, with 30 members over the 20th century. We analyse its dependence of prediction skill on various features of the NAO and the North Atlantic system, like the Atlantic Multidecadal Variability (AMV). As such we will demonstrate, that the NAO is a much less stable system over time as currently assumed and that models may not be in the position to predict its full variability appropriately.

Current state-of-the-art decadal ensemble prediction systems are run with an ensemble size of 10 to 40 members, their retrospective forecasts of the past are used to assess the system's prediction skill. Here, we present an attempt for a large ensemble decadal prediction system for the time period 1960-today, with an ensemble size of 80 members, based on the low resolution version of the Max Planck Institute Earth system model (MPI-ESM-LR). The ensemble is forced with CMIP6 conditions and initialized every year in November through a weakly coupled assimilation using atmospheric reanalyses via nudging and observed oceanic temperature and salinity profiles via a 16-member ensemble Kalman filter. To generate ensemble members beyond 16, we use additional physical perturbations at stratospheric height. The analysis of our large ensemble prediction system presented here aims for answering two questions: (1) How does the ensemble mean deterministic prediction skill for global and North Atlantic key climate indices change with ensemble size? (2) How well may the 80-member ensemble serve as a basis for a robust statistical analysis of probabilities of extremes in the North Atlantic sector? Preliminary results for global and regional air surface temperature show that in terms of ensemble mean ACC and full ensemble CPRSS with reference data, the 80-member ensemble leads to similar prediction skill as the 16-member ensemble. This indicates that the additional ensemble members may lead to a better sampling of the distribution of model trajectories, paving the way for a more robust statistical probabilistic analysis.

Understanding the variability and drivers of air-sea CO2 fluxes on seasonal timescales is critical for resolving the ocean carbon sink's evolution and variability. Here, we investigate whether discrepancies in the representation of air-sea CO2 fluxes on a seasonal timescale accumulate to influence the representation of CO2 fluxes on an interannual timescale in two important ocean CO2 sink regions – the North Atlantic basin and the Southern Ocean. Using an observation-based product (SOM-FFN) as a reference, we investigate the representation of air-sea CO2 fluxes in the Max Planck Institute's Earth System Model Grand Ensemble (MPI-ESM GE). Additionally, we include a simulation based on the same model configuration, where observational data from the atmosphere and ocean components is assimilated (EnKF assimilation) to verify if the inclusion of observational data alters the model state significantly and if the updated modelled CO2 flux values better represent observations.We find agreement between all three observation-based and model products on an interannual timescale for the North Atlantic basin. However, the agreement on a seasonal timescale is inconsistent with discrepancies as large as 0.26 PgC/yr in boreal autumn in the North Atlantic. In the Southern Ocean, we find little agreement between the three products on an interannual basis with significant seasonal discrepancies as large as 1.71 PgC/yr in austral winter. However, while we identify regional patterns of dominating seasonal variability in MPI-GE and EnKF, we find that the SOM-FFN cannot demonstrate robust conclusions on the relevance of seasonal variability in the Southern Ocean. In turn, we cannot pin down the problems for this region.

Marine Heatwaves (MHWs) are Sea Surface Temperature (SST) extremes that can have devastating impacts on marine ecosystems but can also impact circulation patterns in the ocean and the atmosphere. The variability of MHWs has been studied in historical observations and longterm climate projections, but predictability has only been analyzed on seasonal timescales. Here, we we present the first attempt to study the decadal predictability of MHW days per year in an ensemble of decadal hindcasts based on the Max Planck Institute Earth System Model (MPI-ESM-LR).Our results show that there are strong regional differences in prediction skill. While many regions show little to no skill, we find in the Subpolar North Atlantic correlation coefficients up to 0.7 for MHW days up to lead year 8. We demonstrate that these correlations mainly arise from correctly predicting the absence of MHWs in individual years. MHW days per year might be successfully predicted by only using yearly mean SST as a proxy, which also demonstrates that in the Subpolar North Atlantic, any increase in SST is accompanied by more MHWs and vice versa.

The north Atlantic subpolar gyre (SPG) has been widely implicated as the source of large-scale changes in the subpolar marine environment. However, inconsistencies between indices of SPG-strength have raised questions about the active role SPG-strength and size play in determining water properties in the eastern subpolar North Atlantic (ENA). Here, by analyzing various SPG indices derived from observations and a global coupled model, we show that the choice of the SPG index dictates the interpretation of SPG strength-salinity relationship in the ENA. Variability in geostrophic currents derived from observed hydrography and model based Lagrangian trajectories reveal zonal shifts of advective pathways in the ENA and meridional shifts in the western intergyre region. Such shifts in advective pathways are manifestations of variability in the size and strength of the SPG, and they impact salinity by modulating the proportion of subpolar and subtropical waters reaching the ENA. SPG indices based on subsurface density and principal component analysis of sea surface height variability capture these shifts in advective pathways, and are therefore best suited to describe SPG-salinity relationship in the ENA. Our results establish the dynamical constraints on the choice of the SPG index and emphasize that SPG indices should be cautiously interpreted.

The North Atlantic Subpolar Gyre (SPG) has been widely implicated as the source of large-scale changes in the subpolar marine environment. However, inconsistencies between different indices of SPG strength based on Sea Surface Height (SSH) observations have raised questions about the active role SPG strength and size play in determining water properties in the eastern subpolar North Atlantic (ENA). Here, by analyzing SSH-based and various other SPG-strength indices derived from observations and a global coupled model, we show that the interpretation of SPG strength-salinity relationship is dictated by the choice of the SPG index. Our results emphasize that SPG indices should be interpreted cautiously because they represent variability in different regions of the subpolar North Atlantic. Idealized Lagrangian trajectory experiments illustrate that zonal shifts of main current pathways in the ENA and meridional shifts of the North Atlantic Current (NAC) in the western intergyre region during strong and weak SPG circulation regimes are manifestations of variability in the size and strength of the SPG. Such shifts in advective pathways modulate the proportions of subpolar and subtropical water reaching the ENA, and thus impact salinity. Inconsistency among SPG indices arises due to the inability of some indices to capture the meridional shifts of the NAC in the western intergyre region. Overall, our results imply that salinity variability in the ENA is not exclusively sourced from the subtropics, instead the establishment of a dominant subpolar pathway also points to redistribution within the SPG.

We analyse central equatorial Pacific inter-annual prediction skill of sea surface temperature (SST) and net primary productivity (NPP) using initialized retrospective forecasts with the Max Planck Institute Earth system model over the time period 1998-2014. We find significant NPP predictability for up to 5 lead years, which is far beyond the SST predictability of less than 1 year in this area. While El-Nino-Southern-Oscillation (ENSO) limits SST predictability, we find the origin of the high NPP prediction skill to be in the tropical upwelling zones of the eastern Pacific, i.e., the Peru-Chile current system offshore South America. Off-equatorial Rossby waves are initiated off the coast of Chile and travel towards the central tropical Pacific on a time scale of 4 to 5 years. On their arrival, the Rossby waves modify the depth of the nutricline, which is fundamental to the availability of nutrients in the euphotic layer in the central tropical Pacific.We further demonstrate that the seasonal upwelling in the central equatorial Pacific, which is mainly driven by ENSO, transports nutrients, i.e. nitrate and phosphate, from below the nutricline into the euphotic zone, effectively transferring the Rossby wave signal from depth to the near-surface ocean. A shallower than normal nutricline leads to larger primary production, and vice versa, a deeper than normal nutricline to smaller primary production. The Rossby waves also modulate the SST, however, these changes are damped on the daily to weekly time scale due to surface heat fluxes at the atmosphere-ocean boundary. Therefore, the off-equatorial Rossby waves maintain the high predictability of NPP but not the SST. We conclude that NPP predictions in the central equatorial Pacific benefit from the memory contained in properly simulated off-equatorial Rossby waves.

On interannual to decadal time scales, memory in the Earth's climate system resides to a large extent in the slowly varying heat content of the ocean, which responds to fast atmospheric variability and in turn sets the frame for large-scale atmospheric circulation patterns. This large-scale coupled atmosphere–ocean feedback is generally well represented in today's Earth system models. This may fundamentally change when data assimilation is used to bring such models close to an observed state to initialize interannual to decadal climate predictions. Here, we review how the large-scale coupled atmosphere–ocean feedback is preserved in common approaches to construct such initial conditions, with the focus on the initialized ocean state. In a set of decadal prediction experiments, ranging from an initialization of atmospheric variability only to full-field nudging of both atmosphere and ocean, we evaluate the variability and predictability of the Atlantic meridional overturning circulation, of the Atlantic multidecadal variability and North Atlantic subpolar gyre sea surface temperatures. We argue that the quality of initial conditions for decadal predictions should not purely be assessed by their closeness to observations, but also by the closeness of their respective predictions to observations. This prediction quality may depend on the representation of the simulated large-scale atmosphere–ocean feedback. This article is categorized under: Climate Models and Modeling > Knowledge Generation with Models

More than one hundred days were simulated over very large domains with fine (0.156 km to 2.5 km) grid spacing for realistic conditions to test the hypothesis that storm (kilometer) and large-eddy (hectometer) resolving simulations would provide an improved representation of clouds and precipitation in atmospheric simulations. At scales that resolve convective storms (storm-resolving for short), the vertical velocity variance becomes resolved and a better physical basis is achieved for representing clouds and precipitation. Similarly to past studies we found an improved representation of precipitation at kilometer scales, as compared to models with parameterized convection. The main precipitation features (location, diurnal cycle and spatial propagation) are well captured already at kilometer scales, and refining resolution to hectometer scales does not substantially change the simu-lations in these respects. It does, however, lead to a reduction in the precipitation on the time-scales considered – most notably over the ocean in the tropics. Changes in the distribution of precipitation, with less frequent extremes are also found in simulations incorporating hectometer scales. Hectometer scales appear to be more important for the representation of clouds, and make it possible to capture many important aspects of the cloud field, from the vertical distribution of cloud cover, to the distribution of cloud sizes, and to the diel (daily) cycle. Qualitative improvements, particularly in the ability to differentiate cumulus from stratiform clouds, are seen when one reduces the grid spacing from kilometer to hectometer scales. At the hectometer scale new challenges arise, but the similarity of observed and simulated scales, and the more direct connection between the circula-tion and the unconstrained degrees of freedom make these challenges less daunting. This quality, combined with already improved simulation as compared to more parameterized models, underpins our conviction that the use and further development of storm-resolving models offers exciting opportunities for advancing understanding of climate and climate change.

High-resolution simulations (grid spacing 2.5 km) are performed with ICON-LEM to characterize convective organization in the Tropics during August 2016 over a large domain ranging from northeastern South America, along the tropical Atlantic to Africa (8,000×3,000 km). The degree of organization is measured by a refined version of the wavelet-based organization index (WOI), which is able to characterize the scale, the intensity and anisotropy of convection based on rain rates alone. Exploiting the localization of wavelets both in space and time, we define a localized version of the convective organization index (LWOI). We compare convection observed in satellite-derived rain rates with the corresponding processes simulated by ICON-LEM. Model and observations indicate three regions with different kinds of convective organization. Continental convection over West Africa has a predominantly meridional orientation and is more organized than over South America, because it acts on larger scales and is more intense. Convection over the tropical Atlantic is zonally oriented along the ITCZ and less intense. ICON and observations agree on the number and intensity of the African easterly waves during the simulation period. The waves are associated with strong vorticity anomalies and are clearly visible in a spatiotemporal wavelet analysis. The central speed and the wavelength of the waves is simulated well. Both the scale and intensity components of LWOI in ICON are significantly correlated with environmental variables. The scale of precipitation is related to wind shear, CAPE and its tendency, while the intensity strongly correlates with column-integrated humidity, upper-level divergence and maximum vertical wind speed. This demonstrates that the LWOI components capture important characteristics of convective precipitation.

The long term variability and its predictability of the monthly mean oceanic surface net freshwater fluxes is compared in a set of retrospective predictions. All are using the same model setup, and only differ in the implemented ocean initialisation method and ensemble generation method. The basic aim is to deduce the differences between the initialization/ensemble generation methods in view of the uncertainty of the verifying observational data sets. The analysis will give an approximation of the uncertainties of the net freshwater fluxes, which up to now appear to be one of the most uncertain products in observational data and model outputs. All ensemble generation methods are implemented into the MPI-ESM earth system model in the framework of the ongoing MiKlip project (www.fona-miklip.de). Hindcast experiments are initialised annually between 2000–2004, and from each start year 8 ensemble members are run for 10 years forward. Four different ensemble generation methods are compared: (i) a method based on the Anomaly Transform method in which the initial oceanic perturbations represent orthogonal and balanced anomaly structures in space and time and between the variables taken from a control run, (ii) one-day-lagged ocean states from the MPI-ESM-LR Baseline 1 system, (iii) one-day-lagged ocean and atmospheric states with preceding full-field nudging to re-analysis in the atmospheric and anomaly nudging in the oceanic component of the system – the Baseline MPI-ESM-LR system, (iv) an Ensemble Kalman Filter (EnKF) implemented into oceanic part of MPI-ESM, assimilating monthly subsurface oceanic temperature and salinity using the Parallel Data Assimilation Framework and full-field nudging in the atmosphere. The hindcasts are evaluated probabilistically using freshwater flux data set from NCEP-R2. On the global scale the physically motivated methods (i) and (iv) provide probabilistic hindcasts to some extent higher correlation and reliability than the lagged initialization methods (ii)/(iii) despite the large uncertainties in the verifying observations and in the simulations. We suggest similar approaches for further evaluations of other variables of decadal hindcasts systems.

Climate and weather variability in the North Atlantic region is determined largely by the North Atlantic Oscillation (NAO). The potential for skillful seasonal forecasts of the winter NAO using an ensemble‐based dynamical prediction system has only recently been demonstrated. Here we show that the winter predictability can be significantly improved by refining a dynamical ensemble through subsampling. We enhance prediction skill of surface temperature, precipitation, and sea level pressure over essential parts of the Northern Hemisphere by retaining only the ensemble members whose NAO state is close to a “first guess” NAO prediction based on a statistical analysis of the initial autumn state of the ocean, sea ice, land, and stratosphere. The correlation coefficient between the reforecasted and observation‐based winter NAO is significantly increased from 0.49 to 0.83 over a reforecast period from 1982 to 2016, and from 0.42 to 0.86 for a forecast period from 2001 to 2017. Our novel approach represents a successful and robust alternative to further increasing the ensemble size, and potentially can be used in operational seasonal prediction systems. Plain Language Summary Predicting Northern Hemisphere winter conditions, which are controlled largely by fluctuations in the pressure filed over the North Atlantic (North Atlantic Oscillation, NAO), for the next season is a major challenge. Most state‐of‐the‐art seasonal prediction systems show a correlation between observed and predicted NAOs of less than 0.30. Our novel approach uses dynamical links (teleconnections) between the autumn state of sea surface temperature in the North Atlantic, Arctic sea ice, snow in Eurasia, and stratosphere temperature over the Northern Hemisphere as predictors of the NAO in the subsequent winter to subsample a dynamical reforecast ensemble. We select only the ensemble members that consistently reproduce winter NAO states that evolve in accordance with the autumn state of these predictors. As a result the winter NAO prediction skill increases to a correlation value of 0.83. Considering these well established NAO teleconnections in our Earth system model leads to an improved prediction skill of European winter conditions, that is, surface temperature, precipitation, and sea level pressure. Our results advance seasonal prediction of European weather to a level that is usually limited to tropical regions and are relevant for a variety of societal sectors, such as global and national economies and energy and water resources.

We analyze the time dependency of decadal hindcast skill in the North Atlantic subpolar gyre within the time period 1961–2013. We compare anomaly correlation coefficients and temporal interquartile ranges of total upper ocean heat content and sea surface temperature for three differently initialized sets of hindcast simulations with the global coupled model MPI-ESM. All initializations use weakly coupled assimilation with the same full value nudging in the atmospheric component and different assimilation techniques for oceanic temperature and salinity: (1) ensemble Kalman filter assimilating EN4 observations and HadISST data, (2) nudging of anomalies to ORAS4 reanalysis, (3) nudging of full values to ORAS4 reanalysis. We find that hindcast skill depends strongly on the evaluation time period, with higher hindcast skill during strong multiyear trends, especially during the warming in the 1990s and lower hindcast skill in the absence of such trends. Differences between the prediction systems are more pronounced when investigating any 20-year subperiod within the entire hindcast period. In the ensemble Kalman filter initialized hindcasts, we find significant correlation skill for up to 5–8 lead years, albeit along with an overestimation of the temporal interquartile range. In the hindcasts initialized by anomaly nudging, significant correlation skill for lead years greater than two is only found in the 1980s and 1990s. In the hindcasts initialized by full value nudging, correlation skill is consistently lower than in the hindcasts initialized by anomaly nudging in the first lead years with re-emerging skill thereafter. The Atlantic meridional overturning circulation reacts on the density changes introduced by oceanic nudging, this limits the predictability in the subpolar gyre in the first lead years. Overall, we find that a model-consistent assimilation technique can improve hindcast skill. Further, the evaluation of 20 year subperiods within the full hindcast period provides essential insights to judge the success of both the assimilation and the subsequent hindcast quality.

International audience; We demonstrate in this paper that conventional time-averaged decadal hindcast skill estimates can overestimate or underestimate the credibility of an individual decadal climate forecast. We show that hindcast skill in a long period can be higher or lower than skill in its subperiods. Instead of using time-averaged hindcast skill measures, we propose to use the physical state of the climate system at the beginning of the forecast to judge its credibility. We analyze hindcasts of North Atlantic sea surface temperature (SST) in an initialized prediction system based on the MPI-ESM-LR for the period 1901-2010. Subpolar North Atlantic Ocean heat transport (OHT) strength at hindcast initialization largely determines the skill of these hindcasts: We find high skill after anomalously strong or weak OHT, but low skill after average OHT. This knowledge can be used to constrain conventional hindcast skill estimates to improve the assessment of credibility for a decadal forecast. Plain Language Summary Credible predictions of climate up to 10 years into the future, so-called decadal climate predictions, can be a potent tool for decision makers. However, previous work indicated that such predictions are sometimes credible and sometimes not. This study illustrates that knowing the physical state of the climate system at the start of a decadal climate prediction helps assessing the credibility of that prediction. Analyzing sea surface temperature in the North Atlantic as a case study, we show that northward heat transport in the ocean provides a good indicator of the credibility of decadal predictions in the North Atlantic. Unlike previous studies, we do not only analyze time-averaged prediction credibility, but look at individual predictions in the past. This makes our findings particularly relevant for individual forecasts and decision makers.

We present results from the assimilation of observed oceanic 3-D temperature and salinity fields into the global coupled Max Planck Institute Earth system model with the SEIK filter from January 1996 to December 2010. Our study is part of an effort to perform and evaluate assimilation and prediction within the same coupled climate model without the use of re-analysis data. We use two assimilation setups, one where oceanic observations over the entire water column are assimilated, and one where only oceanic observations below 50 m depth are assimilated. We compare the results from both assimilations with an unconstrained control experiment. While we do not find significant improvements due to assimilation in terms of the root-mean-square error of simulated temperature, 0–700 m heat content, sea surface height (SSH), and the Atlantic meridional overturning circulation (AMOC) against observations, we find the variability in terms of correlation with observations significantly improved due to assimilation, most prominently in the tropical oceans. Improvements over the control experiment are stronger in the sub-50 m assimilation experiment and in integrated quantities (SSH, AMOC).