Second Phase of PMIP (PMIP 2 Workshop 1 Summary)


Note: what follows was originally published in EOS as:

Second phase of Paleoclimate Modelling Intercomparison Project,
EOS, Vol 86, No. 28, 12 July 2005


Second phase of PMIP

M. Crucifix1, P. Braconnot2, S.P. Harrison3 and B. Otto-Bliesner4

1Hadley Centre, Met Office
Exeter, Devon, UK
Click for email

2Laboratoire des Sciences du Climat et de l'Environnement
DSM / CEA Saclay, L'Orme des Merisiers / Bat. 709, F-91191 Gif-sur-Yvette, France
Click for email

3Macquarie University, Australia
Click for email

4Climate and Global Dynamics Division
National Center for Atmospheric Research, Boulder, CO, USA
Click for email


By devoting an entire chapter of the fourth assessment report (AR4) to paleoclimates, the Intergovernmental Panel on Climate Change (IPCC) has recognized that studying past climates is essential for comprehending the climate system, and hence for predicting its future evolution.

In particular, numerical simulations of past climates, different from today, constitute an independent test of the performance and reliability of the general circulation models commonly used for climate predictions. This idea is the motivation for the Paleoclimate Modelling Intercomparison Project (PMIP), a long-standing initiative endorsed by the World Climate Research Programme and the International Geosphere and Biosphere Programme. The focus has so far been on two reference periods, chosen as being the most representative of cold and warm climate intervals in the recent past: the Last Glacial Maximum (LGM, 21,000 years ago) and the mid-Holocene (6000 years ago).

The first phase of PMIP, launched in 1992, was based on analysis of atmospheric general circulation models. PMIP 1 met with considerable success, with 67 model experiments archived in a central database and over 70 papers published in the refereed literature. One of the greatest achievements of PMIP 1 was to provide a framework for paleoclimate modeling by bringing a community of climate modelers and field experts together. In 2002, the PMIP steering committee decided, therefore, that the time had come to define new numerical experiments that take advantage of the new generation of atmosphere-ocean-vegetation coupled models [Harrison et al., 2002]. With these new simulations PMIP 2 also hopes to extract quantitative information on climate sensitivity in order to better predict the consequences of climate change on human societies.

The first PMIP 2 workshop, held in April 2005, gathered most of the scientists involved in the project. The attendance spanned the different representative disciplines (climate modeling, geophysics, biology, palynology, geochemistry, and paleoceanography) with 65 scientists from five continents. On the agenda were a discussion of model results obtained so far (including a comparison with paleodata), an overview of newly available paleoclimate data syntheses, and plans to widen the scope of PMIP 2 by defining experimental protocols for other periods in the past.

A Series of New Paleoclimate Simulations

Compared with PMIP 1, the PMIP 2 project considers, for each period, a suite of numerical experiments that differ by the number of components of the Earth system that are interactively taken into account: AGCM simulations use atmosphere general circulation models with prescribed sea surface temperatures; OAGCMs have a component calculating the dynamics of the ocean; and OAVGCMs are coupled to a dynamic vegetation model.

The high scientific potential of this approach was highlighted in a series of presentations describing recent OAGCM and OAVGCM simulations. Y. Zhao showed how the shoaling of the mixed layer in the Indian Ocean enhances the response of the Indian monsoon to the changes in orbital forcing. This is a good example of positive feedback of the ocean response on the atmosphere dynamics. Having interactive vegetation may further enhance the response of the atmosphere with, in turn, consequences to the structure of the ocean mixed layer. Such an ocean-vegetation synergy needs to be taken into account to understand the abundant monsoon rainfall in the Sahel during the mid-Holocene [Braconnot et al., 2004].

Another goal of PMIP 2 is to maintain a database of model results in which long-term means and individual years are archived. Information on interannual and interdecadal variability can therefore be extracted. Specifically, preliminary comparisons of the El Niño Southern Oscillation and North Atlantic Oscillation signals in PMIP 2 experiments were shown by U. Merkel and R. Gladstone. The first PMIP 2 model outputs also provided the opportunity to illustrate some innovative analysis techniques (quantification of cloud feedback based on regime analysis, simulations of potential vegetation, lake-level modeling).

Apart from using more sophisticated climate models, PMIP 2 uses up-to-date information to define the experiments, including the latest reconstruction of the Last Glacial Maximum topography (ICE-5G) and a river routing scheme derived from geological evidence complemented by hydrological modeling. It is therefore essential to determine to what extent these refinements improve the ability of state-of-the-art climate models to reproduce past climatic data.

On the one hand, a series of posters was devoted to specific sensitivity experiments analyzing the role of river routing, horizontal resolution, and freshwater fluxes on simulation results. On the other hand, a scheme was developed during the meeting (S. Harrison) that objectively determines the level of agreement between models and data on the basis of a series of qualitative but robustly documented paleoclimatic features, for example, abundant monsoon rainfall in the Sahel throughout the mid-Holocene. It was also shown (A. Abe-Ouchi) how quantitative information produced from paleoclimate records, such as the amplitude of the tropical Pacific cooling at the Last Glacial Maximum, may be used to infer constraints on climate sensitivity on the basis of a large ensemble of climate simulations. This technique could provide a valuable contribution to the IPCC.

Promising Advances in Data Reconstructions

The meeting was equally rich on the data side. Pollen data and macrofossils (M. Kerwin, M. Edwards) show that the warmest epoch between the Last Glacial Maximum and today (the Holocene thermal optimum) occurred as early as 10,000 years ago in the Alaskaeastern Siberia sector (Beringia), while northeastern America summer temperatures hit their maximum less than 6000 years ago. The late thermal optimum in northeastern Canada is partly linked to the presence of remainders of the Laurentide Ice Sheet until about 7000 years ago. Perhaps in connection with the disappearance of the Laurentide Ice Sheet, cysts of dynoflagellates reveal a sharp increase in surface salinity and mixed-layer depth around 7500 years ago in the Labrador Sea (A. de Vernal).

When possible, a multi-proxy approach is preferable for inferring robust information on past climate evolution. A good example was given for southeastern Brazil (B. Turcq), where lake-level reconstructions, speleothemes (e.g., stalagmites), and pollen fossils allow the documentation of a continuous increase in precipitation since the Last Glacial Maximum.

Progress has also been made about the knowledge of past ice sheets. Geophysical constraints on the history of the Laurentide Ice Sheet (W. R. Peltier) allow the inferring of the existence of large freshwater meltwater runoff flow into the Arctic during the Younger Dryas (12,700-11,700 years ago). There are also an increasing number of attempts to extract indices of past climate variability from high-resolution paleoclimatic records, although it has not been possible so far to infer information on interannual variability that is spatially consistent (S. Brewer). Finally, detailed analysis of modern climate and pollen data remains essential, as illustrated by an extensive survey of modern bioclimatic relationships presented by B. Thompson.

Widening the Scope: New Targets for PMIP Experiments

Although the mid-Holocene and LGM continue to be of interest, other periods pose interesting challenges. Specifically, PMIP 2 has expanded the set of standard experiments to include simulations of the previous interglacial/glacial transition (this is the glacial inception, 115,000 years ago), the early Holocene (9000 years ago), the Younger Dryas (the cold period observed in different regions of the Northern Hemisphere between 12,700 and 11,700 years ago), and the abrupt cooling event 8200 years ago.

The focus on glacial inception is motivated by some earlier experiments suggesting that vegetation and ocean feedbacks are essential for explaining year-to-year accumulation of snow in northeastern America at the end of the previous interglacial period. With the PMIP community focusing on fully coupled OAVGCMs, it is appropriate to revisit this issue. Younger Dryas and 8.2 kyr experiments are an opportunity to obtain information on the stability of the ocean circulation, and the consequences of possible changes in its structure, by comparing model outputs with paleoclimatic data. The meeting has allowed the defining of the corresponding experimental setups.

Several contributions also highlighted the necessity of using Earth system models of intermediate complexity to develop analysis methods, explore the parameter space, and analyze the response of climate over long timescales. For example, it was shown that several thousands of years are needed by the climate system to recover from a perturbation of the North Atlantic freshwater balance. Furthermore, the characteristic response time differs for glacial and interglacial conditions (I. Ross, E. Bauer).

A. Koutavas (LDEO, Columbia) and A. Ganopolski (Potsdam Institute for Climate Impact Research, Germany) were awarded the best poster prizes. The meeting also provided the opportunity to tighten social links; and an excursion to the nearby Porquerolles Island nicely complemented the program. Full details about the meeting, the project, and how to be involved in result analysis are available at

The Paleoclimate Modelling Intercomparison Project Workshop was held on 3-8 April 2005, in Giens (Var, France).


Financial support to the meeting was provided by Centre National de la Recherche Scientifique (CNRS, France), Commissariat l'Energie Atomique (CEA, France), the international programs on climate variability and predictability (CLIVAR) and Past Global Changes (PAGES). The local organization and the program were settled by P. Braconnot and B. Otto-Bliesner with the help of the PMIP 2 steering committee and the Laboratoire des Sciences de Climat et de l'Environnement (CEA-CNRS, France) climate modeling group. S. Galot cared for the administration and French language support. Travel funding for many of the U.S. participants was provided by the U.S. National Science Foundations Earth System History program and U.S. National Oceanic and Atmospheric Administrations Office of Global Programs.


Braconnot, P., S. P. Harrison, S. Joussaume, C. D. Hewitt, A. Kitoh, J. E. Kutzbach, Z. Liu, B. Otto-Bliesner, J. Syktus, and N. Weber (2004), Evaluation of PMIP coupled ocean-atmosphere simulations of the mid-Holocene, in Past Climate Variability Through Europe and Africa, edited by R. W. Batterbee et al., pp. 515-533, Springer, New York.

Harrison, S. P., P. Braconnot, S. Joussaume, C. D. Hewitt, and R. J. Stouffer (2002), Comparison of paleoclimate simulations enhances confidence in models, Eos Trans. AGU, 83(40), 447.

Workshop Links

Home Top Site Map @ Last updated \2011/01/12 15:52:00