A look back

In May 2017, a community of Earth system modellers from eight Helmholtz centres across Germany came together to initiate a four-year long adventure under the Helmholtz-funded initiative Advancing Earth System Modelling Capacity - the ESM project.

The group had a clear objective: they would have joined forces to develop and apply innovative Earth system modelling capacity to contribute to solving some of the grand challenges mankind is facing. The efforts concentrated on enhancing the representation of Earth system model compartments, for which world-leading expertise was available within the group. Another cornerstone of the project has been the development of a flexible framework for the effective coupling of Earth system model components, which is challenging given the diverse nature of the components and models being coupled. Furthermore, Earth system data assimilation capacity needed to be advanced, as the most effective way of combining models and observations.

Today, four years into the project and close to a transition into a new form, the project and its participants look back at the hard work of these years and reflect on the achievements and advancements of the consortium.

Enhanced Earth system compartments—Ocean-cryosphere-biosphere systems (Arne Biastoch, GEOMAR – Task 1.2)

The development of high-resolution ocean models was an important aspect in ESM, both to establish ocean-only1,2 and coupled3 configurations. A key feature is the two-way nesting approach that allows to host a high-resolution grid into a global model. With 1/20° resolution, VKING20X (Fig. 1) represents a flagship configuration that allows to simulate mesoscale pattern (boundary currents, frontal zones, eddies) in the whole Atlantic Ocean for integration of multi-decadal periods. The hindcast experiment covering the past 60 years under prescribed atmospheric forcing und land-ice melting is able to realistically simulate the different convection patterns in the Labrador and Irminger Seas4.

A series of experiments revealed a good correspondence with ship-board sections and moored transports, hence enables to expand the interpretation of spatially and temporally limited observations into a basin-scale context. An important scientific result was the robust decline of the Atlantic Meridional Overturning Circulation (AMOC) in the subpolar North Atlantic by about 2.5 Sv in the period 2000-2017, that is consistently seen in experiments and long-term moorings at 53°N2.


1. Schwarzkopf, F. U. et al. The INALT family - A set of high-resolution nests for the Agulhas Current system within global NEMO ocean/sea-ice configurations. Geosci. Model Dev. 12, 3329–3355 (2019).

2. Biastoch, A. et al. Regional Imprints of Changes in the Atlantic MeridionalOverturning Circulation in the Eddy-rich Ocean ModelVIKING20X. Ocean Sci. Discuss. in prep. (2021).

3. Matthes, K. et al. The Flexible Ocean and Climate Infrastructure version 1 (FOCI1): Mean state and variability. Geosci. Model Dev. 13, 2533–2568 (2020).

4. Rühs, S. et al. Changing spatial patterns of deep convection in the subpolar North Atlantic. J. Geophys. Res. 126, e2021JC017245 (2021).

Earth System Modelling for renewable energy (Naveed Akhtar, Hereon – Task 1.2)

The increasing demand for renewable energy production has stimulated the rapidly increasing deployment of offshore wind farms (OWFs).  To simulate the impact of these large offshore wind farms on power production and regional climate at the first step, a wind farm parameterization has been implemented into the regional climate model (COSMO-CLM)1. The wind farm parameterization in COSMO-CLM has been validated with the in situ airborne measurements2. Furthermore, we simulate near future wind farm scenarios considering existing and planned OWFs in the North Sea and assess power generation losses and wind variations due to wind farm wake. The annual mean wind speed deficit within a wind farm can reach 2–2.5 ms-1 depending on the wind farm geometry (Fig. 1). The large size of wind farms and their proximity affect not only the performance of its downwind turbines but also that of neighboring downwind farms, reducing the capacity factor by 20 % or more, which increases energy production costs and economic losses2.

Akhtar, N. & Chatterjee, F. Wind farm parametrization in COSMO5.0_clm15 (2020) doi:https://doi.org/10.35089/WDCC/WindFarmPCOSMO5.0clm15.

Akhtar, N., Geyer, B., Rockel, B. et al. Accelerating deployment of offshore wind energy alter wind climate and reduce future power generation potentials. Sci Rep 11, 11826 (2021). doi: https://doi.org/10.1038/s41598-021-91283-3

Enhancing Terrestrial Systems model compartments (Luis Samaniego, UFZ - Task 1.3)

One major application of land-surface models (LSM), that simulate the energy and water cycle over land, is to provide the lower boundary condition in coupled land-atmosphere simulations. For this purpose, various processes are represented in LSMs such as water transport through the soil or evapotranspiration among others. The parameters within these processes are often derived from small-scale experiments or have been tuned to provide realistic results in coupled land-atmosphere simulations. However, inverse estimation of these parameters by comparing simulated fluxes to observed ones has not been done at large scale. In this project, we are estimating soil parameters of the HTessel LSM by comparing simulated and observed discharge at a wide range of river catchments spread over the entire globe. We are using the Multiscale Parameter Regionalization (MPR) to reduce the dimensionality of the parameter space. This is achieved by relating high-resolution geophysical attributes to model parameters through transfer functions. The number of transfer functions parameters is then independent of the spatial domain of the model simulation. By this means, automatic parameter estimation can be applied efficiently. We find high improvements in Kling-Gupta efficiency (KGE) by comparing simulations based on calibrated parameters compared to simulations based on default parameters. On average, there is a 70 % improvement in KGE. These results are promising that performance of LSMs can be further improved. Next steps are to include other parameters than soil parameters in the parameter estimation and to also investigate the impact of the chosen transfer function in MPR.

Towards a joined, coupled ESM (Patrick Jöckel, DLR - Task 1.5)

Within this task, a roadmap towards a joined, coupled Earth System Model (ESM) for the entire community was developed, and first major steps towards its implementation have been undertaken.  The guiding principle was to achieve the highest possible degree of interoperability between components, which are already in use and under development at the different centers, but with a minimized need of interference with existing codes. In other words, the focus was on the implementation of interfaces, which enable new coupled systems for the investigation of scientific questions that cannot be answered without.

We aim at a multi-institutional, overarching system for a maximum benefit for all partners by optimally exploiting synergies and the specific expertises at the different centers.

The figure shows how such a joined, coupled Earth System Model could look like, naturally not implying that all shown components and domains are required for all scientifically meaningful model setups. Nevertheless,
the structure is such that a variety of model setups could be realized within the same framework. In essence, the main characteristic is the three-fold hierarchy: the internal coupling for tightly coupled processes on similar time and grid scales within a physical domain, the classical external coupling for connecting models of different physical domains, and a second layer of external coupling for executing multiple instances, as for instance required for on-line nested setups and parallel data assimilation frameworks.

The success of this overall strategy is becoming scientifically visible through the "frontier" simulations that could be conductes within this project.

The ICON implementation in the coupled TSMP (Stefan Poll, Abouzar Ghasemi, FZJ – Task 1.5)

As part of the ongoing development of the Terrestrial System Modeling Platform (TSMP, www.terrsysmp.org) towards a regional Earth system model, the ICON model in its numerical weather prediction and large eddy modes has been implemented into TSMP, alongside the COSMO atmospheric model. TSMP is a scale-consistent, highly modular, fully integrated soil-vegetation-atmosphere modeling system, composed of an atmospheric model (ICON or COSMO), the Community Land Model (CLM), and the hydrologic model ParFlow (https://www.parflow.org) coupled via OASIS3-MCT. The energy conserving ICON coupling with the land surface model CLM is done by directly using the turbulent surface fluxes exchanged by the OASIS3-MCT coupler. Different spatio-temporal resolutions between the coupled compartments are possible. Given the enhanced physical treatment of the land surface processes, the computational overhead due to the coupling of the ICON simulation with CLM is moderate. Compared with the default land surface model TERRA-ML it is about 5%. Currently, tests are ongoing with the latest TSMP configuration that also incorporates the GPU version of ParFlow (Hokkanen et al., 2021). Hence, TSMP can make full use of the heterogeneous, modular CPU-GPU setup of the JUWELS tier-0 HPC system Booster Module. This allows for the compute-intensive, high-resolution ParFlow hydrological simulations to be efficiently accelerated.

Inferring Mantle Viscosity Through Data Assimilation of Relative Sea-Level Observations in a Glacial Isostatic Adjustment Model (R. Schachtschneider, J. Saynisch-Wagner, V. Klemann, M. Bagge, M. Thomas, GFZ - Task 2)

Glacial isostatic adjustment is largely governed by rheological properties of the Earth's mantle. Large mass redistributions in the ocean-cryosphere system and the subsequent response of the visco-elastic Earth have led to dramatic sea level changes in the past. In order to understand and predict current and future sea level changes the knowledge of mantle viscosity is essential. During ESM, we used PDAF to couple a particle filter to the visco-elastic Earth model VILMA. With this setup we successfully infer mantle viscosity profiles (Fig.1) from paleo sea level observations.

We investigate the sensitivity of the assimilation’s success on spatiotemporal observation distribution and the respective uncertainties. We could show, to correctly recover Earth’s lower mantle viscosities the observations around the Laurentide ice shield are much more important than respective observations from Fennoscandia.

Impacts of the assimilation of satellite sea surface temperature data on volume and heat budget estimates for the North Sea (Wei Chen, Hereon – Task 2)

A 3-D (GALATON3DVAR) data assimilation technique is applied, which contains assumed model error correlations that depend on the mixed layer depth (MLD) derived from a coupled circulation/ocean wave model within the framework of GCOAST model system developed in Helmholtz-Zentrum Hereon. Mechanisms controlling the heat budget of the North Sea are investigated based on a combination of satellite sea surface temperature measurements (OSI SAF SST) and numerical model simulations. The study reveals that even with balanced pressure gradients there is a significant response of the three-dimensional model circulation to sea surface temperature (SST) assimilation in the North Sea (Figure 1). Mechanisms controlling the heat budget of the North Sea are different at west, north, and east boundaries. The largest DA impact on volume/heat transport is in the Norwegian Channel, where the dominant process is Eulerian transport, followed by tidal pumping and wind pumping. Volume transports through the Norwegian Channel are modified by SST assimilation, due to an acceleration of the along-shelf current at the northern edge of the North Sea, a decrease in the horizontal pressure gradient from the Atlantic to the North Sea, which leads to changes in lateral heat advection.

Frontier Simulations with a focus on the Asian summer monsoon (ASM) (Sabine Brinkop, DLR - Task 3.2)

Within the ESM project, we found that formaldehyde is quickly hydrated within warm clouds. The quick outgassing and gas-phase oxidation of its hydrated form lead to a four times higher formation of formic acid than any other known gas-phase source (Franco et al., 2021). In order to represent this new diol mechanism for other aldehydes, we developed the Jülich Aqueous-phase Mechanism of Organic Chemistry (JAMOC, Rosanka et al., 2020a). By applying JAMOC within the global atmospheric model ECHAM/MESSy (EMAC), we find that during the Asian monsoon, the in-cloud OVOC oxidation leads to a reduced impact of the 2015 Indonesian peatland fires on tropospheric ozone (O3) (Rosanka et al., 2020b,c). Short lived pollutants from these fires are efficiently transported by the Asian monsoon anticyclone into the lower stratosphere where they lead to a depletion of O3. This indicates that these fires contribute to the variability in lower stratospheric ozone that was observed in recent years (Rosanka et al. 2020c).

In a joint cooperation between FZJ and DLR we used the chemistry-climate model EMAC and two Lagrangian models together: ATTILA (Atmospheric Tracer Transport in a LAgrangian Model, Brinkop and Jöckel, 2019), and CLaMS (Chemistry Lagrangian Model of the Stratosphere, McKenna et al., 2002) coupled within the MESSy software framework (Modular Earth Sub-model System, Jöckel et al., 2010, 2016) as our modelling infrastructure. This model configuration is novel in that both Lagrangian models - which are different with respect to the representation of mixing and convection - run consistently together with EMAC in one single simulation with a focus on the ASM.

We were interested in the pathways of trajectories showing the vertical and horizontal climatological transport from defined source regions in the boundary layer that contribute to the composition of the air in the Asian monsoon anticyclone. First results show that the transport from the Tibetan plateau becomes important in certain months, and that the effect of convection on the distribution of tracers is not only dependent on the respective emission region but also on the life-time of the tracer.



  • Brinkop, S. and Jöckel, P.: ATTILA 4.0: Lagrangian advective and convective transport of passive tracers within the ECHAM5/MESSy (2.53.0) chemistry–climate model, Geosci. Model Dev., 12, 1991–2008, doi.org/10.5194/gmd-12-1991-2019, 2019.
  • Franco, B. , T. Blumenstock, C. Cho, L. Clarisse, C. Clerbaux, P-F. Coheur, M. de Mazière, I. de Smedt, H.-P. Dorn, T. Emmerichs, H. Fuchs, G. Gkatzelis, D. W. T. Griffith, S. Gromov, J. W. Hannigan, F. Hase, T. Hohaus, N. Jones, A. Kerkweg, A. Kiendler-Scharr, E. Lutsch, E. Mahieu, A. Novelli, I. Ortega, C. Paton-Walsh, M. Pommier, A. Pozzer, D. Reimer, S. Rosanka, R. Sander, M. Schneider, K. Strong, R. Tillmann, M. van Roozendael, L. Vereecken, C. Vigouroux, A. Wahner, D. Taraborrelli (2021): Ubiquitous atmospheric production of organic acids mediated by cloud droplets. Nature, 593, pages 233–237, https://doi.org/10.1038/s41586-021-03462-x, 2021
  • P. Jöckel, A. Kerkweg, A. Pozzer, R. Sander, H. Tost, H. Riede, A. Baumgaertner, S. Gromov, and B. Kern, Geosci. Model Dev., 3, 717–752, https://doi.org/10.5194/gmd-3-717-2010, 2010
  • Patrick Jöckel, Holger Tost, Andrea Pozzer, Markus Kunze, Oliver Kirner, Carl A. M. Brenninkmeijer, Sabine Brinkop, Duy S. Cai, Christoph Dyroff, Johannes Eckstein, Franziska Frank, Hella Garny, Klaus-Dirk Gottschaldt, Phoebe Graf, Volker Grewe, Astrid Kerkweg, Bastian Kern, Sigrun Matthes, Mariano Mertens, Stefanie Meul, Marco Neumaier, Matthias Nützel, Sophie Oberländer-Hayn, Roland Ruhnke, Theresa Runde, Rolf Sander, Dieter Scharffe, and Andreas Zahn. Geosci. Model Dev., 9, 1153–1200, https://doi.org/10.5194/gmd-9-1153-2016, 2016
  • McKenna, D. S., Konopka, P., Grooß, J.-U., Günther, G., Müller, R., Spang, R., Offermann, D., and Orsolini, Y.: A new Chemical Lagrangian Model of the Stratosphere (CLaMS): 1. Formulation of advection and mixing, J. Geophys. Res., 107, 4309, https://doi.org/10.1029/2000JD000114, 2002. ab
  • Rosanka, S., Sander, R., Wahner, A., and Taraborrelli, D.: Oxidation of low-molecular weight organic compounds in cloud droplets: development of the JAMOC chemical mechanism in CAABA/MECCA (version 4.5.0gmdd), Geosci. Model Dev. Discuss. [preprint], https://doi.org/10.5194/gmd-2020-337, accepted for publication, 2020a
  • Rosanka, S., Sander, R., Franco, B., Wespes, C., Wahner, A., and Taraborrelli, D.: Oxidation of low-molecular weight organic compounds in cloud droplets: global impact on tropospheric oxidants, Atmos. Chem. Phys. Discuss. [preprint], https://doi.org/10.5194/acp-2020-1041, accepted for publication, 2020b

EU Frontier Simulations (Rakovec, Kaluza, Thober, Samaniego, UFZ - Task 3.3)

We have established a framework of the frontier simulations for the high-resolution, century-long multi-model (n=3) hydrological modelling over Europe using a large ensemble of bias-corrected and downscaled EURO-CORDEX dataset (n=88). This framework represents an interface to the applicants of the climate simulations in climate change impact, adaptation, and mitigation studies over Europe. So far, we have completed the model simulations with the mHM model (Figure T3.3_1), while two more land surface model simulations (Jules, HTESSEL) will be launched in JUWELS starting this summer. A consistent setup employing identical geophysical information to reduce uncertainty that might stem from the use of different ancillary dataset. In our system, differences between models are solely due to differences in process descriptions. This allows us to attribute any difference in the outcomes directly to the difference in process description of the models within the multi-model ensemble. We use existing parametrisations and our experience of delivery of continental-wide/global modeling from previous projects (i.e. EDgE; Samaniego et al. 2019} and ULYSSES (https://www.ufz.de/index.php?en=47367) to guarantee timely output.  All HM/LSMs share a common (highly parallelized and efficient) streamflow routing algorithm (mRM; Thober et al. 2019) to minimize uncertainties derived from the river network characterisation and its parameterization. Results of this project will provide the most comprehensively derived large ensemble (n=264; 3 HM/LSMs x 88 EURO-CORDEX simulations, Figure T3.3_2) hydrologic variables at high-resolution (3 km) available over Europe, which will better describe the projected uncertainties in terrestrial hydroclimatic regimes, making a giant leap over previous assessment studies (Samaniego et al. 2019, Samaniego et al. 2018, Thober et al. 2018, Marx et al. 2018).