4 PAGES News • Vol.16 • No 2 • April 2008 Special Section: Data-Model Comparison Linking data and models GERRIT LOHMANN Alfred Wegener Institute for Polar and Marine Research, Bremerhaven, Germany; gerrit.lohmann@awi.de Paleo-data and models have the poten- tial for a truly symbiotic relationship. One application of paleoclimate data is to validate state-of-the-art coupled climate models for past time slices and speciic climate transitions. Analyzing proxy-re- constructed paleoclimate records and models in tandem allows for the evalu- ation of climate transitions through the analysis of forcing and feedback mecha- nisms in past and future climate changes. In return, model simulations can aid in the interpretation of the causes of observed variations in paleoclimate data. Climate simulations enable a separation of the externally forced climate signal from in- ternal variability (to the extent that the signal is distinguishable from the noise), something that cannot be achieved us- ing proxy data alone. To become efective, these mechanisms require that data and model simulations can be compared in a meaningful way. This special section of PAGES News highlights the importance of validating the results of individual reconstructions and simulations, and showcases a vari- ety of methods that can be used for their comparison. Spatial and temporal obstacles Comparisons between paleoclimate and model data are hindered by the dif- ferent characteristics of each data set. Model output is less reliable at small spa- tial scales, while some proxy data can be representative of only single sites. An im- portant task is to develop methodologies for coping with these characteristics. This will allow subsequent, unbiased compari- sons between simulations and proxy data that explicitly take into account the esti- mated errors in the proxy reconstructions and the small-scale resolution issues with models. In general, proxy data are sampled at discrete spots in the areal dimensions of the Earth's surface and record the temporal dimension well. The vertical component of the environmental signal is only marginally captured, for example, by near-surface, thermocline, and near- bottom living marine organisms, or by sample transects across topographic gra- dients. While the reconstructions usually contain many points in time, the data will only be available at a limited number of discrete spatial locations. On the other hand, the spatial scale of recent global cli- mate models (less than 300 km) enables the combined spatio-temporal domain to be explored. Model-data advances The comparison of paleoclimate and model data can be carried out in a num- ber of ways. The simplest approach is to subsample the model output ields, pick- ing out data only from those locations and seasons for which paleo-reconstructions exist for comparison. However, before a valid comparison can be made, it must be conirmed that the climate response is being compared on similar spatial and temporal scales. There are two key statistical methods that synchronize the spatial scale of the model simulation and proxy reconstruc- tion. The upscaling technique identiies the underlying large-scale processes, e.g., the teleconnections that control low- frequency variations observed in many proxy records. In this case, the statistical method brings the climate modes and shifts, such as those observed in the past 100 years, into a long-term context. The upscaling method can even be used to reconstruct synoptic conditions related to reconstructed ice core data (Rimbu et al., p. 5). The large-scale signal, which can be remote from the local proxy record, can then be compared with model simula- tions. In a similar direction, a variational approach can be used to connect the dif- ferent scales of local paleo-information together with a dynamically consistent spatial smoothing (Kühl et al., p. 8). This is in contrast to the downscaling technique, where large-scale information obtained by a model simulation can be “zoomed in” to the smaller scale of proxy climate information (Raible et al., p. 10; Meyer and Wagner, p. 12). Comparison of model simulations and data reconstructions can additionally aid in clarifying the true environmental sig- nals recorded by proxies. It has been pro- posed that proxy parameters be included as tracers in models and create numerical simulations of proxy generation and burial in all available archives (i.e., synthetic ma- rine, terrestrial and ice core), to compare them with real archives. This idea devel- oped from the need for a mechanistic un- derstanding of how environmental condi- tions are transferred to the archive. Proxy parameters that are found in a number of diferent archives, e.g., 18 O in foraminifera and ice cores, are particularly useful in this approach, as they provide informa- tion on how the climate signal is recorded within the one proxy in diferent climate components (ocean, ice). An example of modeling of marine ∆ 14 C, and subsequent comparison with marine reservoir age, is presented by Butzin et al. (p. 13). The idea of generating pseudo-proxy records is also an important component of the new Paleo-Reconstruction Challenge (see Am- mann, PAGES News 2008, 16(1): 4). A variety of comparison techniques allow many additional objectives to be achieved through comparison of model simulations and proxy reconstructions. Identifying key regions where speciic cli- mate signals are likely to be recorded can optimize sampling eforts for future proxy reconstructions. It would be useful, for example, to explore targeted climate hot spots through a combined use of paleocli- mate reconstructions and model-simulat- ed climate data. Signal-to-noise ratios, or inverse methods, may be used to evaluate locations where climate phenomena can best be detected. Comparison can also be made be- tween diferent climate simulations, in or- der to assess whether additional forcings raise the levels of variability to a similar extent in both the simulated and recon- structed climates (Goosse et al., p. 15). Model results may also provide informa- tion on the onset, duration and magnitude of climate events, similar to how they may be expressed in proxy-based reconstruc- tions (Wiersma et al., p. 16). Another ap- proach is to test the robustness of model results through comparison of multiple model outputs with data (Otto-Bliesner and Brady, p. 18). Data-model compari- sons can also enable the estimation of cli- mate sensitivity (Schneider von Deimling et al., p. 20), which is important for predic- tions of future temperature rise based on current CO 2 projections. Climate sensitivity estimates are reli- ant on estimates of past climate forcings. Data-model comparison can addition- ally be useful in reducing uncertainties in these past forcings (Crowley et al., p. 22). A slightly diferent approach is to run the model using the proxy data and to then use the model to obtain a dynamically consistent interpretation of the data. This approach has been used, for example, to clarify the temperature and ocean circula- tion regime during the Pliocene (Chandler