PI: Edwin K. Schneider (COLA)
Co-PI: Lennart Bengtsson (MPI)
Co-PI: Ian Watterson (CSIRO)
The research will investigate the extent to which the differences in the higher latitude (especially North Atlantic) responses of some of the coupled models in the CMIP2 model intercomparison can be related either to forcing from the tropics or internal atmospheric dynamics. The experimental procedure will be to compare simulations of several of the CMIP2 AGCMs (CCM3, ECHAM4, CSIRO) forced by various tropical SST distributions, including those taken from observations and produced in the CMIP2 integrations. This is a pilot project to test the usefulness of the procedure. If the results are promising, we will attempt to involve other CMIP2 models and groups.
There are substantial differences in the geographical structure of the equilibrium response to doubling of CO2, even with the simple slab mixed layer ocean. It has been found that the difference in the geographical structure of the high latitude warming, as well as the difference in the global mean, between two of the more extreme models (MRI and CCSR2) in a WGCM model intercomparison (LeTreut and McAvaney, 2001) was related to the response of the tropical SST. Large high latitude and global mean warming was related to large warming in the eastern equatorial Pacific, while the low warming in high latitudes and the global mean was related to small eastern equatorial Pacific SST warming. The difference between coupled atmosphere-ocean GCM versions of the same models shows similar characteristics (Noda, 2001, private communication). Results from the CMIP2 intercomparison show that the CCSR2 model produces stronger tropical warming, but also much larger northern hemisphere high latitude warming at the time of doubling of CO2 compared to the other models (J. R�is�nen, 2001, private communication). A mechanism for the structure of the tropical SST response to influence the global and higher latitude regional sensitivity has been discussed by Schneider et al. (1997). Hoerling et al. (2001) demonstrated that SST forcing between 30°S and 30°N probably influences the low frequency changes in the NAO.
From a practical point of view, understanding regional differences in the model sensitivity is perhaps even more important than the question of global sensitivity differences. The proposed work is designed to compare, in a series of controlled simulations, how different models respond in the extratropics and higher latitudes to potential structures of tropical warming.
The geographical distribution of the warming at the time of doubling (J. R�is�nen, 2001, private communication) shows that the 19 CMIP2 the models agree on certain features, such as warming in the eastern Pacific and equatorial Atlantic. However, the model responses vary widely in the North Atlantic and the Southern Ocean. For obvious reasons, it is important to understand why the models do not agree in the geographical distribution of their responses. The North Atlantic is of special interest to the countries of North America and Europe, and we will also be particularly interested in this region.
a) Background experiments
There is a line of evidence, including papers cited in the introduction, that influences of tropical SST changes on the extratropics and higher latitudes could be important in simulating and predicting global warming. We have produced an example of the feasibility of this influence reduced to its simplest elements, using the COLA AGCM (Bengtsson et al., 2001). 50 year integrations of a T21 version of the COLA AGCM have been performed, forced with [climatological annual cycle of SST] plus [a constant in space and time SST anomaly between 10°N and 10°S that represents the structure and amplitude of the observed trend of SST over the past 50 years]. The SST anomaly will be referred to below as the TSSTA (for Tropical SST Anomaly). The model winter SLP anomaly is highly significant and resembles to the observed 1948-1999 winter SLP trend from the NCEP reanalysis. Also of interest is the result from forcing the AGCM with the negative of TSSTA. The result in this case is not just the negative of the result found with TSSTA. This lack of symmetry shows that there is strong nonlinearity in the response to TSSTA anomaly, even in a single model. If nonlinearity of the response is due to changes in the wave propagation properties of the atmosphere, for example, then difference in regional responses might be explained by differences in the models' climatological mean simulation. The surface heat flux response in the TSSTA minus control experiment in a coupled model gives an indication of how the SST will tend to respond (in the absence of changes in the ocean heat transport and atmospheric feedbacks) to the SST forcing outside the forcing region. There are strong regional responses to the relatively small TSSTA, order 20 W m-2. Some of these responses occur in regions where models agree. Other large responses occur in regions where models disagree (e.g. the North Atlantic south of Greenland). Differences in the tropical response of models to greenhouse forcing could then be responsible for differences in the extratropical response.
b) Proposed experiments
The proposed experiments use the same technique as those performed by Bengtsson et al. (2001). The equilibrium response of an AGCM to time-independent SST anomalies superimposed on a climatological (periodic) SST will be calculated, using long (order 50 year) simulations. The AGCMs and the SST distributions will vary in the different simulations. The experiments will find to what extent the differences in high latitude responses and other regional aspects of the CMIP2 coupled models can be explained as intrinsic to the manner in which the atmospheric models respond to a similar tropical forcing, and to what extent the differences can be explained as model-consistent responses originating from different tropical forcings. As a pilot project, our experiments will use only the AGCMs in a limited model suite (CCM3, ECHAM4, and CSIRO). These models were chosen to span the range of global sensitivities in CMIP2. At the time of doubling, CCM3 (as part of CSM/PCM) has low temperature sensitivity and intermediate precipitation sensitivity; ECHAM4 has intermediate temperature sensitivity and low precipitation sensitivity; CSIRO has high temperature and precipitation sensitivity. The experiments or ones suggested by results from this pilot project could in principle be done independently by the full set of modeling groups, as an AMIP-type (specified SST) intercomparison. The experimental procedure is chosen because the signal can easily be separated from the noise by simple time means, and statistical significance of the signal can be made arbitrarily high by extending the length of the simulations. The questions that are relevant to the problem of extratropical/tropical coupling and the experiments that will be performed to answer them are:
1. Do the models have a significant extratropical response to tropical greenhouse-related SST anomalies? Do the responses of different models differ or are they similar?
The specified SST experiments described in Sec. 2.b will be repeated for each AGCM: 50 years with climatological annually varying SST, 50 years with TSSTA, and 50 years with -1×TSSTA.
2. Are processes internal to the atmospheric models responsible for the differences in extratropical SST at time of doubling in the CMIP2 experiments?
50 year simulations of each AGCM will be run forced by the time-of-doubled CO2 climatological annual cycle of SST averaged over all of the CMIP2 experiments. These experiments will also use doubled CO2 concentration in the models. The atmospheric circulation and surface heat flux (a predictor of SST difference when coupled) will be compared. Differences between results produced by the different AGCMs with the same forcing are due to processes internal to the models.
3. Is different tropical forcing responsible for differences in extratropical SST at time of doubling in the CMIP2 experiments?
Experiments as described in the preceding bullet will be repeated, but with the time-of-doubled CO2 tropical SST of the respective CMIP2 integration for each AGCM added. This will provide a measure of the importance of the tropical forcing for each model. Additional experiments will be carried out using common tropical SST anomalies, including the extremes of tropical SST found in the full set of CMIP2 cases. Each model will have its own response to the tropical forcing. Differences between models in this response will be another measure of the role of differences internal to the models.
These experiments should show whether the differences in the regional (especially North Atlantic) behavior of the transient coupled greenhouse simulations can be attributed to differences in the atmospheric response to tropical forcing, to internal differences in the models, or to neither. If the response of the three atmospheric models is similar in experiments 1 and 2, then internal differences in the model are probably not important. If the response of (one or more of) the models to tropical forcing is vigorous and can explain differences in the extratropical responses in the CMIP2 experiments, then tropical forcing is probably important in the explanation.
c) Objectives for proposed work and expected significance
The objective is to design an experimental procedure to implicate or rule out the tropical behavior of coupled models as being responsible for disagreement between CMIP2 transient projections of regional higher latitude climate change (for a limited subset of the CMIP2 models). The procedure outlined above will be tested using the atmospheric components of three of the coupled models. If successful, the procedure could be applied in a full model intercomparison. However, this is exploratory research. We expect that the procedure may require modification as the research progresses.
d) CMIP2 data requirements
Access to the SST data from the control and transient CMIP2 is required for this project.
Bengtsson, L., E. K. Schneider, and Z. Z. Hu, 2001: An influence of tropical warming in high latitudes. In preparation.
Covey, C., K. M. AchutaRao, U. Cubasch, P. Jones, S. J. Lambert, M. E. Mann, T. J. Phillips, and K. E. Taylor, 2001: An overview of results from the Coupled Model Intercomparison Project (CMIP). Submitted to Global and Planetary Change. (ms. available at https://pcmdi.llnl.gov/mips/cmip/overview_ms/ms_text.html).
Hoerling, M. P., J. W. Hurrell, and T. Xu, 2001: Tropical origins for recent North Atlantic climate change. Submitted to Science.
LeTreut, H. and B. J. McAvaney, 2001: Model intercomparison: slab ocean 2xCO2 equilibrium experiments (in preparation).
Meehl G. A., G. J. Boer, C. Covey, M. Latif, and R. J. Stouffer, 2000: The Coupled Model Intercomparison Project (CMIP). Bull. Amer. Met. Soc., 81, 313-318.
Schneider, E. K., 2001: Diagnosing the causes of differences of simulations of the equatorial Pacific by two coupled ocean- atmosphere general circulation models. In preparation. Schneider, E. K., R. S. Lindzen, and B. P. Kirtman, 1997: A tropical influence on global climate. J. Atmos. Sci., 54, 1349- 1358.