( はてなダイアリーから移動しました)

Impressions of a meeting on paleoclimate modeling

[This note is still to be revised. I have also written in Japanese.]

A scientific meeting on paleoclimate modeling was held in Kyoto last week (from Monday 6th to Friday 10th December). I attended the sessions from Tuesday to Friday.

I used to review and promote paleoclimate modeling, but I have been inactive in this area for about a decade, and I do not have a plan to be active again. This time I was requested by the organizers to be there.

This was not a meeting covering all aspects of paleoclimate, but a more focused one. It was a workshop of a project called PMIP3: "Paleoclimate Modeling Intercomparison Project, Phase 3".

As for methods, PMIP mainly uses coupled ocean-atmosphere general circulation models. Optionally, biogeochemical cycles, vegetation dynamics, and ice sheet dynamics etc. may be coupled. On the other hand, in order to simulate many actual years in a limited computer time, some models use very simplified atmospheric parts.

As for time frames, PMIP focuses on the Quaternary (the last two million years of the history of the earth), though some extension back is included. Paleoclimates of Mesozoic, Paleozoic and Proterozoic times are not subjects of the meeting, though some participants seem to be interested in them.

More in particular, PMIP plans to make the following coordinated numerical experiments which applies the same condition to various models of the participants, and to check correspondence of the results with evidence of actual climate in the past. Namely:

  1. The Last Glacial Maximum ("LGM", 21000 years ago) time slice
  2. Mid-Holocene ("MH", 6000 years ago) time slice
  3. The Last Millennium ("LM", Year 850 C.E. to present) history

The first two have been included in the design of PMIP since the 1st phase which started in 1994. Those two climates are conceived as (statistically) steady states, and their differences from the present climate are considered basically as responses to boundary conditions. Principal conditions of the LGM was extents of continental ice sheets and CO2 concentration, as demonstated by Broccoli and Manabe (1987). A principal condition of the MH was orbital parameters of the earth, in particular, periherion in the summer of the northern hemisphere, as demonstrated by COHMAP (1988). (Orbital forcing in comparison with the present was stronger at 10000 years ago, but then ice sheet largely remained. So the case of 6000 years ago which has simpler configuration was chosen.)

The third one, LM, probably emerged from curiosity about relationship between climate and human history, but also became under focus in the debates about how anomalous the current global warming is in a millennial context.

In addition, such experiments were considered to be done in the context of PMIP3 as:

  • Simulation of events of rapid climate changes during the glacial period (Dansgaard-Oeschger oscillations and Heinrich events) and during the last deglaciation (the Younger Dryas event and the event 8200 years ago).
  • Deglaciation (21000 years ago to ca. 10000 years ago) history
  • Holocene (ca. 10000 years ago to present) history
  • The Last Interglacial (LIG or Eemian, ca. 125000 years ago) a few time slices and/or history
  • Mid-Pliocene (3 million years ago) time slice
  • Eocene (This was not precisely specified yet. Some participants presented results of simulation aimed at understanding the Paleocene-Eocene Thermal Maximum (PETM) event 55.8 million years ago.)

The participants were interdisciplinary in terms of scientific disciplines: meteorology, oceanography, glaciology, physical geography, ecology, geology, paleontology, geochemistry, isotope geoscience, computational science, database management, etc.

On the other hand, it seemed to me that all major participants were well known each other and they form a group that share scientific issues (even though they probably compete one another as peers in the same community).

It was recognized that this subject of science is relevant to society, but the actual discussions seemedto be limited to contribution to IPCC with improved scientific knowledge. (Maybe I miss points raised in the sessions on the first day which I did not attend.) Some of the participants do discuss the relevance of climate studies to society elsewhere, but even those people seemed to concentrate in conducting the project. As if a sole outsider, I expressed some opinions from a perspective related to anthropogenic global warming (AGW). I reiterate them in the following section.

Comments from a perspective related to anthropogenic global warming

The last millennium (LM)

Was climate really "Hockey-stick-like"? Approached from modeling, is it certain that solar and volcanic forcing and internal variability of climate cannot produce global warming comparable to the one of the latter half of the 20th century in the setting of the last millennium?

To address this question, perhaps some groups have to do LM historical runs of climate models.

But, magnitudes of solar and volcanic forcing are not so large compared with their uncertainties. (As I think in the light of a workshop on solar variability and climate which was held last month in Japan [See article on 2010-11-21]), it is now thought (e.g. Gray et al. 2010) that the difference of total solar irradiance (TSI, so-called "solar constant") between the Maunder Minimum (17th century) and present is not so large as considered by Lean et al. (1995) whose reconstruction was used in many "Climate of the 20th Century" experiments of CMIP3 (so-called IPCC AR4 experiment series).

PMIP3 LM forcing data set has been prepared, and it contains multiple reconstructions of solar and volcanic forcing. It was suggested that we should do experiments using not only various models but also various sets of forcing, to produce a multi-dimentional enensemble covering uncertainties in models, uncertainties in forcing and uncertainty in evidence of paleoclimate.

While the suggestion is reasonable, historical runs are time-consuming, and if we need runs with multiple sets of conditions, we would need compromise in spatial resolution of the models.

I think a few teams should conduct LM historical runs including those with maximum plausible solar and volcanic forcing, to demonstrate the range of plausible variability of global climate in the last millennium. This is something like showing the noise level to which the current anthropogenic signal emerges. The result may be that there is some discernible difference of noise level between runs with plausible values of known forcing and runs without forcing, or it may be that there is none. This seems to be a duty of the community of modelers, but it can be recommended only to those few who are really eager to know the noise level. We should not put concentrated efforts of the whole paleoclimate modeling community in this part.

On the other hand, I think that global synthesis of evidence in this time scale (not too much concentrating on validation of this particular series of experiments which seem dull to many) seems very interesting and desirable.

To advance understanding of the response of climate to solar and volcanic forcing, I think that we need models which can reproduce stratospheric dynamics realistically. It is well known that the major volcanic forcing is via stratospheric aerosols (mainly sulfates) which scatter solar radiation. There may be another chain of effects from heating of the stratosphere due to absorption by these aerosols. Also, relative variations of the ultraviolet part of solar radiation is much larger than that of TSI, which can change distribution of stratospheric temperature, which in turn affect propagation of planetary-scale dynamic waves. Good expression of stratospheric dynamics requires high vertical resolution, so I suggest mechanistic experiments with idealized forcing rather than historical millennium runs. Response of the (upper) ocean would also be important, so each run need to have time frames of a few decades.

(Another suggestion about the solar influence, connection via cosmic rays and cloud formation, does not yet seem to be understood quantitatively enough to be included in coordinated modeling experiments.)

The Last Interglacial (LIG)

My comments on this item and the next item of rapid climate changes are not requests of changing specifications of experiments, but requests about how to present the results of the experiments to people at large.

It seems important to me to demonstrate how similar the climate of LIG to the projected global warming due to enhanced greenhouse effect, and how different. I sometimes find such simplistic discourse:
"The warming of the Eemian was less than 2 degrees C from the pre-industrial level, but the sea level was about 5 m higher than present. If 2 deg. C warming would sustain, the sea level will rise by 5 m, and it is certainly not desiable for the global human society. So 2 deg. C warming can be considered as dangerous climate change."
I do not think that the sea level is a simple function of the global mean surface air temperature. I expect that the contrast between two hemisphere and the contrast between seasons are not similar between LIG (where orbital forcing seems important) and the case of enhanced greenhouse effect.
In order to use LIG as an analog to be used in climate policy, we need to know the situation more precisely.

Rapid climate changes, especially Younger Dryas and the event 8200 years ago
It seems that the known events of rapid climate changes are related to either (partial) collapse of continental ice sheets or breaking of lakes which store meltwater of ice sheets, so that the same thing cannot occur in the current climate. (Collapses of Greenland and West Antarctic ice sheets may occur, however, so we want to know similarity and dissmilarity of setting between then and now.)

Even though the structures are different, I think it important to know the rate of changes of sea level, climate, vegetation etc. in those events as hints for adaptation to climate change (both to projected anthropogenic global warming and to surprise scenarios). We need to combine modeling and evidence. For adaptation, people want information in regional changes rather than global. Though it is not always possible to confidently reconstruct regional changes, it will be helpful where it is possible.

On the rationale of data synthesis and mapping
(This section is what I think for many decades, but reinforced during I attend the meeting. It seemed that major participants already recognize this issue, but I felt that it must be emphasized for a broader community.)

Paleoclimate studies need evidence of actual climate which are not recorded in real-time by modern instruments, but reconstructed from various kinds of evidence which are usurally called "proxy" records.

Evidence-based paleoclimatologists usually study samples at certain fixed locations in detail, and give some time series of characteristic variables. Then, if part of the time period of the data overlap that of modern instrumental records, they evaluate corralation between the characteristic variables and the modern climatic variables, and claim that the former can be used as proxy of the latter. They also often discuss periodicity in the time series, or synchroneity of events with time series of evidence in remote locations, But such discussions often remain speculative if they are left alone.

In order to have more organized knowledge of climate, evidence-based knowledge must be combined with theoretical thoughts and modeling. To do this, compilation and mapping of evidence is crucial.

From a theoretical starting point, it is easier to discuss something global than something local. On the other hand, empirical evidence is usually local. In order to compared with theoretical thoughts, global characteristics must be estimated by combining many local pieces of evidence. For example, theoretical thought may give some "prediction" (shall we rather call it "retrodiction"?) of global mean temperature. To check it, pieces of evidence which can be considered as proxies of temperature are conveted to values of temperature, and then spatially averaged taking account of spatial and seasonal representativeness as well as uncertainty of proxies.

We can sometimes (but not always) consider that climate change has globally the same sign in terms of temperature. But it is unlikely in terms of hydrological variables such as precipitation. If climate change happens as enhancement of large-scale atmospheric circulation including the vertical motions, it is likely that the downward branch will became drier while the upward branch will be moister. We need to look at the spatial pattern of the changes to understand them.

Also, spatial contrasts are important as causes. In both atmospheric and oceanic circulations, locations of upward and downward motions depend more on the spatial contrast of temperature and other variables (humidity in the atmosphere and salinity in the ocean) than the global average of such elements. To check such cause-and-effect relationships, we need evidence of spatial distribution of temperature etc.


  • A.J. Broccoli and S. Manabe, 1987: The influence of continental ice, atmospheric CO2, and land albedo on the climate of the last glacial maximum. Climate Dynamics, 1, 87 - 99.
  • COHMAP members, 1988: Climatic changes of the last 18,000 years: Observations and model simulations. Science, 241, 1043 - 1052.
  • L.J. Gray, J. Beer, M. Geller, J.D. Haigh, M. Lookwood, K. Matthes, U. Cubasch, D. Fleitmann, G. Harrison, L. Hood, J. Luterbacher, G.A. Meehl, D. Shindell, B. van Geel and W. White, 2010: Solar influence on climate. Reviews of Geophysics, 48, RG4001, doi:10.1029/2009RG000282, 58 pp.
  • J. Lean, J. Beer and R. Bradley, 1995: Reconstruction of solar irradiance since 1610: Implications for climate change. Geophysical Research Letters, 22, 3195 - 3198.