Working Group 2: Biogeochemistry

Coordinators:

	Dr. Victor Brovkin MPI for Meteorology, Bundesstr. 53, 20146 Hamburg
	Dr. Tatiana Ilyina MPI for Meteorology, Bundesstr. 53, 20146 Hamburg
	Dr. Peter Köhler Alfred-Wegener-Institut (AWI) Helmholz-Zentrum für Polar- und Meeresforschung, Am Alten Hafen 26, 27568 Bremerhaven

 

 

Participating institutions
1. Alfred-Wegener-Institut (AWI) Helmholtz-Zentrum für Polar- und Meeresforschung, Bremerhaven
2. Potsdam-Institut für Klimafolgenforschung (PIK) e. V., Potsdam
3. MARUM - Zentrum für Marine Umweltwissenschaften der Universität Bremen, Bremen
4. Max-Planck-Institut für Meteorologie (MPI-M), Hamburg
5. GEOMAR Helmholtz-Zentrum für Ozeanforschung Kiel, Kiel
6. Institut für Biogeochemie und Meereschemie der Universität Hamburg (UHH), Hamburg
7. Institut für Geowissenschaften der Universität Kiel (CAU), Kiel
8. Leibniz-Institut für Troposphärenforschung (TROPOS) e.V., Leipzig
9.  Max-Planck-Institut für Chemie (MPI-C), Mainz

 

Summary

WG2 aims at understanding and quantifying feedbacks between biogeochemistry and climate during glacial cycles. Four work packages are focused on marine carbon cycle, terrestrial procresses, CH4 cycle, and the dust cycle. Scientific challenges include reproducing glacial CO2 cycle with comprehensive ESMs, understanding of rapid changes in atmospheric greenhouse gases concentrations during abrupt climate changes, and reconstructing atmospheric lifetime of CH4 using a coupled atmospheric chemistry model. The main tasks during the first 4 years are to simulate the full glacial CO2, CH4 and dust cycles using EMICs, and to prepare comprehensive terrestrial and marine biogeochemistry components of ESMs to fully interactive, orbitally forced glacial cycle simulations planned for phase 2 of the project. Time-slice sensitivity and transient deglaciation simulations with ESMs will be used to investigate a contribution of individual biogeochemical mechanisms. The dust cycle, an important driver for the ocean biogeochemistry, provides a link to WG1 as an interactive component of the physical climate system. The amplitude and timing of terrestrial and marine processes will be constrained using proxies for biogeochemistry including carbon isotopes together with WG3. Calibration of poorly known parameters as well as an optimization of the computational performance of the biogeochemistry models, including usage of acceleration techniques, will be performed in collaboration with WG4.
 

 

WP2.1: Marine carbon cycle

Project leader: Dr. Peter Köhler, Dr. Tatiana Ilyina

Principal Investigators (PIs):

Dr. Peter Köhler Dr. Christoph Völker Prof. Dr. Dieter Wolf-Gladrow	Alfred-Wegener-Institut (AWI) Helmholz-Zentrum für Polar- und Meeresforschung Am Handelshafen 12, 27515 Bremerhaven
Dr. Tatiana Ilyina	Max Planck Institute for Meteorology (MPI-M) Bundesstr. 53, 20146 Hamburg
Prof. Dr. Birgit Schneider	Christian-Albrechts-Universität Kiel (CAU), Institut für Geowissenschaften Ludewig-Meyn-Str. 10, 24118 Kiel
Dr. André Paul Dr. Ute Merkel Prof. Dr. Michael Schulz	MARUM – Zentrum für marine Umweltwissenschaften, Universität Bremen Universität Bremen, Postfach 33 04 40, 28334 Bremen
Prof. Dr. Andreas Oschlies Prof. Dr. Klaus Wallmann Dr. Markus Pahlow	GEOMAR Helmholtz-Zentrum für Ozeanforschung Kiel Düsternbrooker Weg 20, 24105 Kiel
Dr. Andrey Ganopolski	Potsdam Institute for Climate Impact Research (PIK) Telegrafenberg, Postfach 601203, 14412 Potsdam

 

Overarching Goals
Understanding the role and mechanisms of the feedbacks between climate and marine biogeochemical processes during glacial times

 

 

WP 2.2: Terrestrial carbon cycle

Dr. Victor Brovkin	MPI for Meteorology (MPI-M) Bundesstr. 53, 20146 Hamburg
Prof. Jens Hartmann, Dr. Nils Moosdorf	University of Hamburg (UHH) Bundesstr. 55, 20146 Hamburg
Dr. Andrey Ganopolski	Potsdam Institute for Climate Impact Research (PIK) Telegrafenberg, 14412 Potsdam

 

Overarching Goals

Understanding and quantifying mechanisms and feedbacks of the terrestrial carbon cycle and climate during glacial cycles.

 

 

WP 2.3: Methane cycle

Principal Investigators (PIs):

Dr. Victor Brovkin, Dr. Thomas Kleinen	Max-Planck-Institut für Meteorologie (MPI-M) Bundesstr. 53, 20146 Hamburg
Dr. Georg Feulner, Prof. Dr. Wolfgang Lucht	Potsdam-Institut für Klimafolgenforschung (PIK) Telegraphenberg A31 14473 Potsdam
Dr. Benedikt Steil	Max-Planck-Institut für Chemie (MPI-C) Hahn-Meitner-Weg 1, 55128 Mainz

 

Overarching Goals

The overarching goal of WP2.3 during the first four years is to interactively simulate the full methane cycle during deglaciation within the models CLIMBER-LPJmL and MPI-ESM. This includes the simulation of natural wetlands as the largest natural source of methane within the land surface models, as well as the simulation of the atmospheric sink where methane is oxidized to CO2. Using these two components, the atmospheric concentration of methane will be determined, allowing a comparison to proxy data from ice cores.

 

 

WP 2.4: Simulating the mineral dust cycle on glacial-interglacial timescales

Principal Investigators (PIs):

Prof. Dr. Ina Tegen, Dr. Kerstin Schepanski	TROPOS Leibniz Institute for Tropospheric Research, Permoser Str. 15, 04318 Leipzig
Dr. Ute Merkel	Zentrum für Marine Umweltwissenschaften (MARUM) MARUM, Leobener Strasse, 28359 Bremen

 

Overarching Goals

Understanding the role and mechanisms of the mineral dust cycle and its feedbacks during glacial-interglacial cycles and improving the representation of the dust cycle in Earth System Models
 

Working Group 3: Synthesis and Analysis of Proxy Data

Coordinators

 

	Prof. Dr. Michal Kucera MARUM - Zentrum für Marine Umweltwissenschaften der Universität Bremen Leobener Strasse, 28359 Bremen
	Dr. Stefan Mulitza MARUM - Zentrum für Marine Umweltwissenschaften der Universität Bremen Leobener Strasse, 28359 Bremen

 

Participating Institutions
1. MARUM - Zentrum für Marine Umweltwissenschaften der Universität Bremen, Bremen
2. Karlsruher Institut für Technologie (KIT), Karlsruhe
3. Alfred-Wegener-Institut (AWI) Helmholtz-Zentrum für Polar- und Meeresforschung, Bremerhaven, & Potsdam
4. Helmholtz-Zentrum Potsdam - Deutsches GeoForschungsZentrum (GFZ), Potsdam
5. Institut für Meteorologie, Freie Universität Berlin, Berlin
6. Meteorologisches Institut der Rheinischen Friedrich-Wilhelms-Universität Bonn; Hans-Ertel-Zentrum für Wetterforschung, Deutscher Wetterdienst (DWD)
7. Insitut für Geowissenschaften, Johannes Gutenberg-Universität, Mainz
8. Helmholtz-Zentrum Geesthacht (HZG) Zentrum für Material- und Küstenforschung GmbH, Geesthacht

 

Summary

The aim of WG3 is to generate a synthesis of proxy records to quantify how individual elements of the global climate system and associated biogeochemical processes changed during the last glacial cycle. This synthesis will be used to verify and diagnose global and regional model simulations carried out within the project and contribute to an improved understanding of the relevant processes. To this end, we will develop data synthesis tools and concepts allowing harmonized recording of paleoclimatic data, use these to synthesize paleoclimatic records representing conditions at the sea surface, deep ocean, and the land surface including polar regions and finally develop and apply approaches for proxy data analysis that facilitate optimum comparison with model simulations and process understanding. The main scientific challenge will be to quantify and reduce paleodata uncertainties due to proxy attribution and chronology, to regionalize model output to facilitate comparison with locally representative proxy records and to develop informative model/climate diagnostics. During the first 4 years, the focus will be on conceptual development of the data synthesis and analysis tools and their application on the period covering the last deglaciation. Analysis of sensitivity to uncertainty and boundary conditions will be used to adapt and evaluate regional modeling approaches whilst sensitivity analysis based on existing time-slice compilations will be carried out to develop ways of quantifying uncertainty in state estimation based on proxy data. The second phase of the project will complete the data synthesis for the entire glacial cycle and focus on data-model comparison, including the assessment of the variance spectra in proxy records, compared to model simulations. Choice of proxy parameters, development of model diagnostics will be coordinated closely within the project (parameters and diagnostics: WG1 and WG2; diagnostics: WG4). Such close collaboration represents a significant added value, leading to optimum data/model compatibility and allowing us to make the most of the highly explicit treatment of data uncertainties.

 

WP3.1: Paleodata synthesis tools and concepts

Principal Investigators (PIs):

Dr. S. Mulitza Prof. Dr. M. Kucera	MARUM-Zentrum für Marine Umweltwissenschaften der Universität Bremen Leobener Str., 28359 Bremen
Dr. T. Laepple	Alfred-Wegener-Institut Helmholtz-Zentrum für Polar- und Meeresforschung Forschungsstelle Potsdam ECUS Young Investigator Group Telegrafenberg A 43, 14473 Potsdam
Prof. Dr. A. Brauer	Deutsches GeoForschungsZentrum  Telegrafenberg, 14473 Potsdam
Prof. Dr. F. Sirocko	Johannes Gutenberg-Universität Mainz Becher-Weg 21, 55099 Mainz
Dr. C. Ohlwein Prof. Dr. A. Hense	Meteorologisches Institut der Rheinischen Friedrich-Wilhelms-Universität Bonn Auf dem Hügel 20, 53121 Bonn
Dr. Eduardo Zorita Prof. Dr. Hans von Storch	Helmholtz-Zentrum Geesthacht Zentrum für Material- und Küstenforschung Institut für Küstenforschung Max-Planck-Straße 1, 21502 Geesthacht

Overarching Goals

Paleoclimatic records can provide a wealth of quantitative information on the state of specific climatic subsystems. At present, however, an extensive synthesis and regionalization of paleoclimatic data is hindered by chronological inconsistencies, uncertainties related to proxy attribution, calibration and non-climate influences and inconsistent data structures that do not allow fast, automated processing of individual data from different sources. The overarching goal of WP3.1 is to provide the software tools, techniques and concepts that allow an efficient homogenization of existing paleoclimatic records into data collections tailored to serve for climate model validation and for a quantification of global and regional climate change during the past 130 kyrs. This includes the construction of a common integrated marine, land and ice chronology, allowing quantitative assessment of age uncertainty, as well as the development of tools to quantify, constrain and make optimal use of the uncertainty of individual proxies to provide data that can be used for model validation

 

 

WP3.2: Paleodata compilation and synthesis

Principal Investigators (PIs):

Prof. Dr. A. Brauer Dr. J. Mingram Dr. habil N. Nowaczyk Dr. B. Plessen	Helmholtz-Zentrum Potsdam Deutsches GeoForschungsZentrum GFZ Telegrafenberg Haus A 20, D-14473 Potsdam
Prof. Dr. F. Sirocko	Johannes-Gutenberg-Universität Mainz Institut für Geowissenschaften Becher-Weg 21, 55099 Mainz
Dr. Peter Köhler Dr. Frank Lamy	Alfred-Wegener-Institut (AWI) Helmholz-Zentrum für Polar- und Meeresforschung Am Handelshafen 12, 27515 Bremerhaven
Dr. Bernhard Diekmann Prof. Dr. U. Herzschuh	Alfred-Wegener-Institut (AWI) Helmholz-Zentrum für Polar- und Meeresforschung Telegrafenberg, 14473 Potsdam
Prof. Dr. M. Kucera Dr. S. Mulitza	MARUM - Zentrum für Marine Umweltwissenschaften der Universität Bremen Leobener Str., 28359 Bremen

 

Overarching Goals

In the last decades tremendous advances have been achieved in reconstructing past climatic variability on different time scales by exploiting various types of geological archives including high latitude ice cores, ocean sediments, speleothems and lake sediments. However, most of the obtained proxy time series of past climate variability largely reflect regional signals, with the size or definition of a ‘region’ depending on the type of archive. The next logical step, a compilation of all these valuable information into a global or at least hemispheric synthesis is still lacking and remains a major challenge for the paleoclimate community. However, a global proxy synthesis from such a variety of different archives is not trivial and faces two main challenges, (1) fundamentally different types of chronologies (e.g. incremental, radiometric, astronomical tuning), and (2) a broad variety of different proxy data reflecting to different climatic parameters (e.g. temperature, precipitation, winds) with different sensitivities toward climate changes. The overarching goal of this WP to adequately address these challenges, develop schemes for quality control, classification, and, harmonization of existing proxy records as fundamental prerequisite for generating a robust compilation of the key proxy records from key geological archives in order to quantify variations of individual elements of the global climate system and associated biogeochemical processes during the last glacial cycle.

 

WP3.3: Data-model interface and data analysis

Principal Investigators (PIs):

Dr. M. Werner Prof. Dr. Gerrit Lohmann	Alfred-Wegener-Institut (AWI), Helmholtz-Zentrum für Polar- und Meeresforschung Bussestraße 24, 27515 Bremerhaven
Prof. Dr. U. Cubasch	Meteorologisches Institut, Freie Universität Berlin (FUB) Carl-Heinrich-Becker-Weg 6-10, 12165 Berlin
Prof. Dr. Ch. Kottmeier Dr. G. Schädler	Institut für Meteorologie und Klimaforschung, Karlsruher Institut für Technologie (KIT) Wolfgang-Gaede-Weg 1, 76128 Karlsruhe
Dr. A. Paul	MARUM-  Zentrum für Marine Umweltwissenschaften der Universität Bremen Leobener Str., 28359 Bremen

 

Overarching Goals

According to the main conclusion reached by the PMIP/PAGES/MARUM workshop COMPARE2012 (Kucera et al., 2012), paleoclimatic data-model comparison needs to be quantitative and “intelligent”, in the sense that it allows to identify and evaluate the processes that caused past climate changes. To achieve this, we want to develop and test methods that facilitate data-model comparison and data analysis and thereby enable an assessment of the Earth system models (ESMs) used in WG1 and the homogenized paleoclimatic data synthesis generated in WP3.1/WP3.2. Because water isotope data are among the most abundant paleoclimatic data, we focus the development of data-model interfaces on the implementation of stable water isotope diagnostics in global and regional climate models. Furthermore, we aim to assimilate time slice and time series products from WP3.1/WP3.2  for selected time in global and regional climate models, in order to provide steady-state as well as transient state estimates periods that are consistent with both data and models and their uncertainties. Finally, we intend to make use of regional climate models for dynamical downscaling of the results obtained by WG1 to facilitate local-scale data-model comparison.

Working Group 4: Optimization

Coordinators:

	Prof. Dr. Thomas Slawig Christian-Albrechts-Universität Kiel, Institut für Informatik und Mathematisches Seminar Christian-Albrechts-Platz 4, 24118 Kiel
	Dr. Joachim Biercamp Deutsches Klimarechenzentrum DKRZ Bundesstraße 45a, 20146 Hamburg

 

 

Participating Institutions:
1. Alfred-Wegener-Institut, Helmholtz-Zentrum für Polar- und Meeresforschung, Bremerhaven
2. Christian-Albrechts-Universität Kiel, Institut für Informatik und Mathematisches Seminar
3. Deutsches Klimarechenzentrum, Hamburg
4. Max-Planck-Institut für Meteorologie, Hamburg
5. MARUM - Zentrum für Marine Umweltwissenschaften Universität Bremen
6. Universität Hamburg, Klimacampus

Summary

Working Group 4 focuses on optimization of Earth system modeling, and specifically on the ESM configurations that will be used in multi-millennia runs by the PalMod initiative. Using the notion of optimization, we refer to two different aspects: optimization of quality and optimization of performance. Both aspects are crucial for the intended initiative as long-term Earth system simulations only make sense if the quality of the results can be assured and the results can be obtained in an acceptable amount of wall clock time. The aim of the PalMod initiative, namely to perform multimillenia runs in order to draw conclusions also for future climate, relies on the quality of the achieved results. On the other hand, the ambitious goal of long-time simulations requires a sophisticated and optimized combination of algorithms, software and hardware that enable a high simulation performance. All progress made regarding runtime reduction of ESMs, will have immediate benefit on the whole PalMod initiative.

It has been shown in numerical weather prediction as well as climate simulations that an essential prerequisite for accurate model results is a well-calibrated data set-up of the model. While in other context extensive experience exists in quantifying and optimizing the parameter set, extreme long time paleo climate simulations have not been optimized in this way so far. The development of suitable metrics and calibration strategies is therefore a necessary prerequisite to enable and improve the planned extreme long-time simulations with uncertain data input (WP4.1). Since it appears unfeasible to achieve this task for the whole complexity of the ESM at once, we will focus in a first step on the ocean biogeochemistry, leaving the optimization of other components as well as the coupled system to a later project phase.

A second paramount objective of the work package is the rigorous analysis and accurate mathematical description of coupling processes and their optimization with respect to computational performance. While in shorter period climate simulations, the representation of interfaces between coupled subsystems can be considered a solved problem, in long-term simulations the accuracy and structure properties of this coupling is not sufficiently well studied. Thus investigating accurate mathematical representations of multi-scale interfaces between subsystems in the ESM is a central objective of work package WP4.2 on coupling strategies. Further investigations and developments are necessary to enhance the subsystems to consider time-dependent topographical and geographical changes in the boudaries between subsystems, e.g. due to largely changing ice coverage or sea level changes. Finally, optimizing the coupling algorithms is essential to achieve scalable ESM performance.

Optimization of runtime performance for the intended long-term simulations will be addressed by two strategies, corresponding to two different time horizons: (i) research into potentially disruptive approaches that might substantially change the way ESMs are formulated and (ii) exploitation of known optimization methods that will speed up the existing model and thus be immediately beneficial to the project. Since the goal in PalMod is ambitious and the consortium unifies Germany’s expertise in Earth system modeling, we think that we shall take the opportunity to also follow ambitious and innovative research paths. In WP 4.3 we will thus study the options of algorithms based on parallelization in time and model order reduction. Both have the potential to significantly reduce the runtime of long-term simulations. In WP 4.4 we will leverage optimization efforts w.r.t. runtime that are more gradual and incremental, but all together should provide significant speed-up. This will take place at several model components, concentrating on the most time-consuming parts. Besides treating algorithmic and software engineering aspects, this work package will also reflect upon hardware related optimizations. Since today’s high performance hardware environment is rapidly changing and will do so in the future, we will keep the hardware-focussed research and development as generic as possible, always respecting the maintainability and portability of source code modifications.

WG4 has a truly interdisciplinary topic and focus: It combines the expertise of Earth system modeling and simulation with those of applied mathematics and computer science. The research will be performed in interdisciplinary work packages in cooperation with the other working groups. The work of WG4 will be an important contribution to the scientific community of (disciplinary and mathematical) modelling, simulation and optimization. We take this into account by providing positions for both PhD students and PostDoc researcher, thus also laying a basis for future interdisciplinary work in this and similar areas of research.

 

 

WP4.1: A model calibration framework exemplified for the ocean biogeochemical component

Principal Investigators (PIs):

Dr. Iris Kriest, Dr. Markus Schartaut	GEOMAR Helmholtz-Zentrum für Ozeanforschung Kiel Düsternbrooker Weg 20, 24105 Kiel
Prof. Dr. Thomas Slawig	CAU Kiel Institut für Informatik Christian-Albrechts-Platz 4, 24118 Kiel

 

Overarching Goals

Aim of the project is to develop a mathematical and algorithmic framework for model calibration, parameter estimation and uncertainty analysis. It will be exemplified for an important component of Earth system models, namely the ocean biogeochemistry. The framework will consist of (i) problem- adapted quality criteria as metrics, cost or misfit functions, (ii) a software interface for the coupling of different models with standardized forcing data, and (iii) numerical optimization algorithms taking into account the specific properties of quality criteria, models and data. The project combines methods from marine biogeochemical modeling, computer science and applied mathematics. In a later phase of the PalMod initiative, these methods and results will be beneficial for similar tasks in other Earth system model components.

 

 

WP4.2: Improvement of coupling strategies

Principal Investigators (PIs):

Prof. Dr. Jörn Behrens	Universität Hamburg Num. Methoden in den Geowissenschaften Grindelberg 5 20144 Hamburg
Dr. Dirk Barbi	AWI Helmholtz-Zentrum für Polar- und Meeresforschung Am Handelshafen 12 27570 Bremerhaven
Dr. Victor Brovkin Dr. Stefan Hagemann Dr. Tatiana Ilyina Dr. Uwe Mikolajewicz	Max-Planck-Institut für Meteorologie Bundesstraße 53, 20146 Hamburg
Dr. André Paul Dr. Matthias Prange	MARUM - Zentrum für Marine Umweltwissenschaften Universität Bremen Leobener Straße, 28359 Bremen

 

Overarching Goals

This work package is concerned with providing mathematical, computational and hardware tools for a core task in earth system modelling: interfacing different components – the coupling within the system. The aim of this work package is therefore to optimize coupling strategies for extremely long simulation times in terms of computational efficiency as well as numerical accuracy. This work will be thoroughly coordinated with the efforts in WP1. In particular, a capability of the coupling strategies to capture spatial and temporal scale interaction and sub-grid scale phenomena (e.g. grounding line migration in ice sheets, or topographic features of importance) will be developed. This capability appears crucial in the representation of subtle movements of boundaries (e.g. land-sea interfaces with changing sea level, ice sheet grounding lines, etc.).

During the initial phase of the project, the opportunity of studying coupling strategies in a more fundamental way will be taken. Since very delicate balances govern the solutions in earth system models, they rely heavily on accurate and mathematically consistent coupling between sub-systems. It is therefore necessary to ensure the equivalence of continuous (however unknown) and discrete solution branches. This will be studied together with algorithmic improvements for handling structure preservation of the coupled quantities.

We will test implementations of the newly developed algorithms on different emerging hardware platforms and will extract generic optimization methods for efficient data structures and data propagation through hierarchical memory HPC architectures.

 

 

WP4.3: Strategies for parallelization in time and model order reduction

Principal Investigators (PIs):

Prof. Dr. Thomas Slawig Prof. Dr. Steffen Börm	CAU Kiel, Institut für Informatik und Mathematisches Seminar Christian-Albrechts-Platz 4, 24118 Kiel
Prof. Dr. Michael Hinze	Universität Hamburg, FB Mathematik, Optimierung und Approximation Bundesstr. 55, 20146 Hamburg

 

 

Overarching Goals

This project aims to investigate the potential of acceleration techniques based on a parallelization w.r.t. time and on model order reduction for long-time ESM runs. Parallelization in time shall be used in addition to the already implemented spatial parallelization that is based on domain decomposition. Model order reduction shall be used to construct a transient model with a reduced basis set of functions generated by snap shots of time slices of an ESM. This reduced model can be used in a parallel-in-time integration to generate initial states at the beginning of each time interval.

 

WP4.4: Algorithmic and Implementation Performance Optimization

 

Principal Investigators (PIs):

Dr. Hendryk Bockelmann Dr. Joachim Biercamp	DKRZ Hamburg Bundesstraße 45a, 20146 Hamburg
Prof. Dr. Steffen Börm Prof. Dr. Thomas Slawig	CAU Kiel, Institut für Informatik und Mathematisches Seminar Christian-Albrechts-Platz 4, 24118 Kiel
Prof. Dr. Jörn Behrens	Universität Hamburg, Numerische Methoden in den Geowissenschaften Grindelberg 5, 20144 Hamburg
Dr. Tatiana Ilyina	Max-Planck-Institut für Meteorologie Bundesstraße 53, 20146 Hamburg

 

Overarching Goals

The major objective of this work package is to provide continuous performance optimization of the global fully coupled comprehensive Earth System Model MPI-ESM suitable for long-term transient paleoclimate simulations. The paleo-version of the MPI-ESM will be built upon the existing coarse resolution configuration MPI-ESM-CR and include all features required for performance and proxy- based validation of the transient multi-millennia simulations like interactive coupling with ice sheet and solid earth models, time-varying land-sea-mask and ocean topography, embedded dust cycle and carbon isotopes models etc.

It is obvious that, besides the physical adequateness of the model, high computational efficiency of the code enabling throughput rates of the order of several simulated centuries per day is essential for the success of the project. The emphasis of this work package will therefore be on the incremental update of different modules, algorithms and numerical schemes to more efficient computational solutions. We are strongly focussed to encourage the interplay between model improvement, numerical needs and computer science allowing for measurable improvements in the 1st project phase.

According to our previously gained knowledge on the model performance, HAMOCC and ECHAM6 will be the main bottlenecks for high throughput in paleo climate simulations. Hence, we will pursue a strategy of four tasks to optimize model specific components and its implementation. Novel modules and improved numerical schemes will be developed to such maturity that they can be deployed into the production type ESM for glacial-interglacial simulations.

The first task will focus on the biogeochemistry model HAMOCC, where grid coarsening strategies and an improved sediment acceleration tool will be developed. As a second task, several bottlenecks within the atmospheric component ECHAM6 need to be eliminated such as the implementation of new transport methods for tracer constituents based on semi-Lagrangian advection schemes. These two tasks aim at a significant reduction of the possible runtime of model years per day while keeping the model code still maintainable and understandable for researchers.

The aforementioned developments will be done closely related to two more technically motivated optimization tasks: based on an evaluation of existing ESM configurations with detailed performance analysis, we will quantify the potential for performance gain that might be achieved by the use of specialized hardware of different types or by new algorithms in selected ESM components. This leads to the third optimization task, where more classical implementation improvements are used to optimize the current model for already available HPC systems. This will encompass optimizations like refactoring for hybrid MPI-OpenMP parallelization or improved single core performance. The fourth task is motivated by the fact that accelerator adoption reaches a critical mass in the HPC community (75 systems on the Nov2014 TOP500 list are using accelerator/co-processor technology and more than half of new HPC systems in 2015 are expected to contain some type of accelerator). Therefore, we will investigate and exploit the potential of hardware accelerators, like general purpose GPUs or the Intel Xeon Phi accelerator cards. We will pay particular attention to numerical algorithms, like interpolation schemes, and how model components need to be reorganized for porting to GPUs and/or Intel Xeon Phi as well as to vector computers like NEC SX-ACE.