Large-scale research project PalMod: What have we learned?

The PalMod project (short for paleo-modeling), funded by the German Federal Ministry of Education and Research (BMBF), investigates the climate system and its variability during the last glacial cycle using complex Earth system models. However, in such models, some components, such as ice shields, are usually prescribed and are not computed interactively, i.e., they do not adjust to the modeled climate. While such an approximation is quite suitable for short simulations of several decades, it is not appropriate for simulations of glacial time scales. Therefore, new components must be added to the model systems to describe the interactions of the relevant physical and biogeochemical processes in the Earth system, including the ice-covered regions. This further development allows the reproduction of the entire climate evolution of the last glacial cycle in a long, continuous (i.e. transient) simulation, to better understand the interactions and feedback effects between individual components of the climate system, and thus, to be able to better assess the uncertainties in long-term climate projections.

Within the PalMod project, in which the Max Planck Institute for Meteorology (MPI-M) is a lead partner, up to 18 scientific institutions are pursuing the goal of investigating the climate system and its variability during the last glacial cycle using these complex Earth system models. During the last ice age, about 20,000 years ago, the global mean temperature was about 4 to 5 degrees lower than today, and large ice sheets lay on top of North America and Scandinavia. As a result, sea level was about 120 meters lower than today. The atmospheric concentration of CO2 was more than one-third lower than the pre-industrial value. Due to changing solar irradiance, an increase in atmospheric CO2 concentration, and the climate warming associated with it, most of the ice shields disappeared within about 10,000 years. To directly calculate such climate changes and climate fluctuations over many centuries and millennia is a huge challenge for climate modelers. So far, transient simulations with complex climate models have only been accomplished over several centuries, in very few cases over a few millennia, or in the form of equilibrium simulations, where the climate forcing does not change.

In the first, four-year project phase (August 2015 - July 2019), the focus was on technical model development as well as on the simulation of the last termination, i.e., the transition from the last glacial period to the present warm period and the associated thawing of the large land ice masses. For this purpose, first transient experiments of the last termination were carried out with a subsystem of the newly developed model, which prescribes the ice shields using ice shield reconstructions from observational data and contains some worldwide firsts. The land-sea mask and river directions were interactively adjusted depending on topography, extent of ice shields, and the change in sea level. An interactive methane module with terrestrial methane emissions was integrated into the land-surface model; a methane sink model was developed in collaboration with the Max Planck Institute for Chemistry (MPI-C) to determine atmospheric methane concentrations in transient experiments. The results of these transient experiments were published in phase two of the project, and reveal some climate characteristics observed in the paleo-data, such as the greening of the Sahel/Saharan region in the early Holocene and the doubling of atmospheric methane between the Last Glacial Maximum (LGM) and pre-industrial times.

The second phase of PalMod (October 2019 - September 2022) addressed coupled climate and carbon cycle experiments during three eras of the last glacial cycle: the transition from the last warm period to the last glacial (120,000 to 86,000 before present), a phase of the glacial with large climate variability (46,000 to 26,000 years before present), and the last termination. The third phase of PalMod is expected to start in the summer of 2023.


Findings from the first two phases

Ice Sheets, Atmosphere and Ocean

First transient simulations of the end of the last glacial (26,000 years ago to pre-industrial times) with ice shields prescribed from reconstructions were used to test the performance of the model and to understand important physical interactions between ice shields and climate. Changes in solar irradiance and greenhouse gases led to a substantial increase in melting at the edges of the ice shields, which was accompanied by a massive retreat of the northern hemispheric ice shields between 15,000 and 9,000 years before present (Kapsch et al., 2021). Episodes of increased ice melt during the end of the last ice age led to increased freshwater runoff into the ocean and a significant weakening of the Atlantic Meridional Overturning Circulation (AMOC). A consequence of this weakening of the AMOC is the occurrance of periods of enhanced cooling in the Northern Hemisphere. Sensitivity experiments allowed us to examine the uncertainties in the simulations that result from using different boundary conditions (Kapsch et al., 2022). The results show that on long time scales, the time and amplitude of meltwater discharge – prescribed according to the reconstructions – dominate climate variability because they influence when and how much the AMOC changes. In addition, differences in the topography of the ice shield reconstructions used for the time of the LGM led to shifts in the atmospheric and oceanic circulation, which in turn influenced the variability of the ocean. These results thus highlight uncertainties in transient model simulations related to model implementation and underline the challenges in interpreting differences between different models participating in international model intercomparisons (e.g., PMIP4 - Paleo Model Intercomparison Project Phase 4).

These simulations also clarify the limitations of experiments with prescribed ice shields, since the ice shields do not respond directly to changes in the model climate. For example, if the ice shield reconstruction shows a period of increased melting, this will result in cooling due to freshwater discharge into the North Atlantic. Such cooling would in turn affect the ice shields. However, since these are prescribed, such a feedback effect is not occurring. This underscores the relevance of fully coupled simulations with interactive ice shields. The first of these fully coupled simulations with interactive ice shields were performed at MPI-M. For this purpose, the MPI-ESM1.2-CR was successfully coupled to the ice shield model mPISM (10 to 15 km resolution) and to the dynamical model of interaction between ice shield, Earth’s crust and upper mantle, VILMA. A first ensemble of fully coupled simulations shows the main features of the transition from the last glacial to the interglacial, including the massive retreat of the ice shield and the weakening of the AMOC due to enhanced ice melt. These transitions match well with observational data. The ensemble also shows that Heinrich events occur in the simulations. Heinrich events describe periods of accelerated ice advances and their discharge into the sea. They are associated with an increased meltwater and iceberg input to the ocean in the Northern Hemisphere and lead to a weakening of the AMOC. However, the timing and magnitude of these events are also strongly dependent on the initial conditions as well as the model parameters. This suggests that a direct comparison of simulated abrupt climate events with observational data will be difficult, and new approaches to model-data comparisons will need to be developed that focus more on the underlying physical processes than on the exact timing of such events.

Figure: Transient model response to different ice-sheet boundary conditions and implementations of meltwater release for the simulations conducted for the study. (a) Global meltwater release, (b) Atlantic Meridional Overturning Circulation (AMOC) at 1,000 m depth and 26°N, (d) North Atlantic sea-surface temperature (SST) for simulations with GLAC-1D (black) and ICE-6G (red) ice-sheet boundary conditions as well as for ICE-6G ice sheets but a globally homogenous distribution of meltwater (blue) and no meltwater release (green). Vertical shadings mark approximate timings of the Bølling-Allerød warm period (left) and Younger Dryas cold period (right) according to proxy evidence.

Heinrich Events

With the newly implemented ice shields in the Earth system model, simulations of Heinrich events – climate fluctuations during the ice age – have also been carried out. These events were shown to be caused by the succession of two mechanisms: iceberg calving and elevation loss of the Laurentide Ice Sheet, which extended over much of North America. The iceberg calving affected the ocean circulation, and the elevation loss of the Laurentide Ice Sheet affected the atmospheric circulation. By using the new model configuration, the interplay of the effects could be studied, and the sequence of the two effects could be observed in a simulation for the first time (Ziemen et al., 2019).

To better understand the sensitivity of Heinrich events to specific climate factors, an extensive ensemble of coupled “ice shield–solid Earth” simulations has been performed (Schannwell et al., 2022). These simulations show that the number and strength of simulated Heinrich events are determined by climate boundary conditions. For example, cyclically recurring conditions can lead to a synchronization of Heinrich events in different regions of the Laurentide Ice Sheet, such as the region around the MacKenzie River drainage basins in northern Canada and the Hudson Bay drainage basin in eastern Canada. A temporal shift in boundary conditions changes the time at which Heinrich events occur, however this is highly dependent on the region. Overall, the Mackenzie region is much more sensitive to perturbations in climate conditions than the Hudson region. For the occurrence of periodic Heinrich events, a favorable combination of ice dynamics and climate forcing is required. Outside this region, ice streams can either become very slow or develop into a permanent stream.

Video: Temporal evolution of the ice sheets during the last termination in a coupled simulation with interactive ice sheets. The simulation begins 18,000 years before present and shows vertically averaged ice flow velocity in meters per year over the ice sheets, vegetation fraction over land, and meltwater input from icebergs into the ocean in meters per year. The contours indicate the effect of ice sheets on the Earth's crust (green contours show a lowering of the Earth's crust compared to today, brown lines show an uplift). In the course of the animation, ice streams in several locations cycle between their active (high velocities) and quiescent phases. During the phases of increased ice stream activity large amounts of icebergs are released, which successively melt and release meltwater into the ocean.


The warming climate and the increase of carbon dioxide during the postglacial led to changes in the vegetation composition. For the Northern Hemisphere, the MPI-M in cooperation with experts from the Research Unit Potsdam of the Alfred Wegener Institute and the University of Heidelberg compared the development of extratropical forests with pollen-based biome-reconstructions. The results showed that the expansion of forests, averaged over the Northern Hemisphere, takes place about 4000 years earlier in the MPI-ESM than the reconstructions suggest (Dallmeyer et al., 2022). The possible reasons for this temporal difference are manifold. Unfortunately, the technical possibilities are currently not sufficient to check all reasons. However, the simulated climate agrees well with reconstructions made independently of pollen data. Because vegetation in the model responds to changes in climate within a few decades, the differences between model and reconstructions indicate a lag in forest response to climatic changes of several millennia: pollen-based reconstructions lag the actual climate in time. Such an imbalance would have far-reaching consequences. Pollen-based climate reconstructions are widely used to evaluate paleo-simulations, including for single time-slice experiments. They are also a substantial component for calculating global and hemispheric climate trends. Results show that this can lead to erroneous conclusions. Therefore, a critical approach to pollen-based reconstructions is recommended. The study further suggests that forests are responding much more slowly to ongoing global warming than Earth system models calculate. The study is currently examining whether similar results will be seen for the Southern Hemisphere.

Marine Carbon Cycle

Changes to the ocean circulation and its impact on the marine carbon cycle are critical to the CO2 sink in glacial climate – and to the postglacial rise in CO2. To quantify the mechanisms of sequestration of atmospheric CO2 in the glacial ocean, HAMOCC, the MPI model of marine biogeochemistry, was extended by introducing the stable carbon isotope 13C, which is essential for studying marine ventilation and biogeochemistry (Liu et al., 2021). This model extension explicitly calculates 13C in all carbon sinks and comprises temperature-dependent fractionation during photosynthesis and the air-water exchange. While model results for the LGM and the pre-industrial state generally agree well with carbon isotope ratios derived from proxy data, they also show high sensitivity of marine biogeochemical tracers to changes in freshwater fluxes. Sensitivity experiments with different circulation states, adjusted by changes in vertical diffusivity, and with different representations of the sinking of marine biogenic particles show that a shallow and weak AMOC does not necessarily result in an improvement compared to proxy data, and that the effects of different representations of sinking depend strongly on the circulation state. Thus, the 13C data do not provide confirmation of the frequently postulated weak and shallow AMOC.

Interaction between terrestrial and marine Biogeochemistry

At MPI-M, PalMod is advancing the integration of climate and biogeochemistry components within the framework of a complex Earth system model. The terrestrial and marine biogeochemistry groups are working closely together to tackle the challenge of interactive coupling between land surface hydrology, ocean circulation, and carbon sinks driven by the postglacial sea level rise. Thus, a new type of system for the transfer of terrestrial carbon, including its stable isotopes, and nutrients to the ocean through coastal flooding during the postglacial has been developed and tested (Extier et al., 2022). Such a coupling between the JSBACH land model and HAMOCC allows, for the first time, a consistent study of the global carbon cycle considering the land-ocean mass transfer in MPI-ESM simulations on glacial time scales. These new results suggest that oceanic CO2 outgassing is triggered by terrestrial organic carbon fluxes during postglacial floods, but only on a regional scale. Regional CO2 outgassing (mainly in the Indonesian region) is sustained by wood input during flooding. This is supported by observations suggesting that prior to the meltwater pulse, the tropical rainforest in this region supported the storage of carbon-rich materials on land. These materials became available to the ocean during the postglacial flooding. The new approach is also an important step toward a fully coupled ESM for the carbon cycle, which is being sought for simulations of the last glacial cycle in PalMod. In addition, the MPI-ESM meanwhile accounts for climate-dependent weathering fluxes that result in inputs of carbon, nutrients, and alkalinity that enter the ocean via rivers. Here, the response of the marine carbon cycle to changes in these fluxes is investigated. The scientists are therefore developing an implementation of the empirical weathering model formulated at the Universität Hamburg, which can be used online in transient model experiments. First results indicate that changes in weathering in the postglacial are mainly driven by changes in surface runoff. However, the effects of these changes are counterbalanced by erosion of loess deposits and a decrease in the area of continental shelves.

Methane Cycle

In collaboration with the MPI-C in Mainz, MPI-M researchers have added a full methane cycle to the MPI-ESM, which includes a new component for terrestrial methane emissions combined with a fast model for atmospheric methane decomposition. This combination permits unique interactive simulations of atmospheric methane in the framework of the coupled ESM. As a world premiere, this has allowed the full evolution of atmospheric methane to be studied in transient model experiments from the glacial maximum to the present (Kleinen, Mikolajewicz, and Brovkin, 2020; Kleinen et al., in prep.). In addition, the scientists have continued one of the MPI-ESM experiments further into the future and have shown that atmospheric methane concentrations are significantly underestimated in current future scenarios (Kleinen et al., 2021). In particular, they found that natural methane emissions increase particularly sharply in the high warming scenarios, up to four times more than previously assumed. This is especially true for the centuries after 2100 AD, when full warming will occur. In the experiments, global temperatures and methane concentrations remain at high levels for a long time (the experiments extended to 3000 AD) and decrease only slowly, as can also be seen in the ice and the air bubbles trapped in it from past interglacials.

About PalMod

Besides researchers from the MPI-M, up to 17 German groups of researchers and one Canadian group are involved in the project, which is funded by the German Federal Ministry of Science and Research (BMBF). PalMod is scheduled to run for 10 years and is expected to achieve its research goals in three project phases. The first phase started in August 2015 and ended in July 2019. The second phase started in October 2019 and ended in September 2022. Martin Claussen, director emeritus at the MPI-M and former head of the department “Land in the Earth System”, was one of the three scientific leaders of the PalMod project, along with Mojib Latif (GEOMAR, Kiel) and Michael Schulz (MARUM, Bremen). Since Martin Claussen’s retirement, Tatiana Ilyina has taken over his part of leadership. Victor Brovkin, Uwe Mikolajewicz, Thomas Kleinen (all MPI-M) and further scientists from other institutions form the PalMod steering group. Other MPI-M scientists involved in PalMod are and were: Anne Dallmeyer, Thomas Extier, Mathias Heinze, Chetankumar Jalihal, Marie-Luise Kapsch, Fanny Lhardy, Bo Liu, Virna Meccia, Thomas Riddick, Clemens Schannwell, Katharina Six and Florian Ziemen.


More Information:

Project webpage:



Phase I

Meccia, V.L., Mikolajewicz, U. (2018) Interactive ocean bathymetry and coastlines for simulating the last deglaciation with the Max Planck Institute Earth System Model (MPI-ESM-v1.2). EGU, Geosci. Model Dev., 11, 4677–4692. DOI:

Riddick, T., Brovkin, V., Hagemann, S., Mikolajewicz, U. (2018) Dynamic hydrological discharge modelling for coupled climate model simulations of the last glacial cycle: the MPI-DynamicHD model version 3.0. EGU, Geosci. Model Dev., 11, 4291–4316. DOI:

Ziemen, F., Kapsch, M.-L., Klockmann, M., Mikolajewicz, U. (2019) Heinrich events show two-stage climate response in transient glacial simulations. Climate of the Past, 15, 153-168.

Phase II

Brovkin, V., Brook, E., Williams J.W., Bathiany, S., Lenton, T.M., Barton, M., DeConto, R.M., Donges, J.F., et al. (2021) Past abrupt changes, tipping points and cascading impacts in the Earth system. Nature Geoscience,

Dallmeyer, A., Kleinen, T., Claussen, M., Weitzel, N., Cao, X., Herzschuh, U. (2022) The deglacial forest conundrum. Nat Commun 13, 6035.

Extier, T., Six, K., Liu, B., Paulsen, H., Ilyina, T. (2022) Local oceanic CO2 outgassing triggered by terrestrial carbon fluxes during deglacial flooding. Climate of the Past, 18, 273-292.

Kapsch, M.-L., Mikolajewicz, U., Ziemen, F. A., Rodehacke, C. B., Schannwell, C. (2021) Analysis of the surface mass balance for deglacial climate simulations. The Cryosphere, 15, 1131–1156, doi.org10.5194/tc-15-1131-2021.

Kapsch, M.-L., Mikolajewicz, U., Ziemen, F., Schannwell, C. (2022) Ocean response in transient simulations of the last deglaciation dominated by underlying ice-sheet reconstruction and method of meltwater distribution. Geophysical Research Letters, 49.

Kleinen, T., Gromov, S., Steil, B., Brovkin, V. (2021) Atmospheric methane underestimated in future climate projections. Environmental Research Letters, 16: 094006.

Kleinen, T., Mikolajewicz, U., Brovkin, V. (2020) Terrestrial methane emissions from the Last Glacial Maximum to the preindustrial period, Climate of the Past, 16, 575–595,

Liu, B., Six, K. D., Ilyina, T. (2021) Incorporating the stable carbon isotope 13C in the ocean biogeochemical component of the Max Planck Institute Earth System Model. Biogeosciences, 18, 4389-4429,

Schannwell, C., Mikolajewicz, U., Ziemen, F., Kapsch, M.-L. (2022) Sensitivity of Heinrich-type ice-sheet surge characteristics to boundary forcing perturbations, EGUsphere [preprint],



Prof. Dr. Victor Brovkin
Max Planck Institute for Meteorology
Email: victor.brovkin@we dont want


Prof. Dr. Martin Claussen
Max Planck Institute for Meteorology
Email: martin.claussen@we dont want


Dr. Tatiana Ilyina
Max Planck Institute for Meteorology
Email: tatiana.ilyina@we dont want


Dr. Marie-Luise Kapsch
Max Planck Institute for Meteorology
Email: marie-luise.kapsch@we dont want


Dr. Thomas Kleinen
Max Planck Institute for Meteorology
Email: thomas.kleinen@we dont want


Uwe Mikolajewicz
Max Planck Institute for Meteorology
Email: uwe.mikolajewicz@we dont want