Modelling Mechanisms of Past Change
The scientific community’s ability to forecast the rates and magnitudes of future change is limited by numerical models of climate change, which in turn are limited by the lack of data on how past climate has varied over time and the mechanisms that drove these changes. This COST Action will optimize methodologies to evaluate model simulations with the aid of suitable palaeodata. Model-data comparisons are considered a challenge because the characteristics of model output and paleodata are very different and many sources of uncertainty exist (e.g. Kohfeld and Harrison, 2000; Renssen and Osborn, 2003; Lohmann, 2008). Climate is represented at different spatial scales: local in proxy records, several hundred kilometres or more in models. Further, the registered variability in proxies is only partly caused by climatic variations, so that it is necessary to isolate the climatic signal using statistical methods and to represent the non-climatic residuals by a suitable stochastic model. Finally, the responses of the proxies to the local or large-scale climate may be non-invertible.
The activities will be focused on:
1. Transient simulations: Several modelling groups (NCAR, Univ Bristol, LSCE, VU Univ Amsterdam) are now working on transient simulations of the last termination using coupled atmosphere-ocean-vegetation models (Kahana and Valdes, 2009; Roche et al., 2009; Liu et al., 2009). This is particularly challenging because dynamical ice-sheets are not yet included in these models, implying that the deglaciation of ice-sheets has to be prescribed.
2. Compilation of forcings for transient simulations: Such transient simulations require high quality data-sets with information on the forcing. Some forcings are well-established such as orbital forcing (independently computed) and greenhouse gas forcing (derived from ice cores). Other forcings, however, are not as well constrained, especially the detailed response of the ice sheets in time and space (e.g., Tarasov and Peltier, 2004). For instance, climate models require detailed information on the timing, location and magnitude of meltwater flows into the oceans.
3. Evaluation of experiments: Evaluation of these transient simulations requires high-quality climate reconstructions from key locations, in different environments (e.g., Renssen et al. 2001). To capture the change in time, continuous records with an excellent chronological control are required (see WG1 and 2). The simulations can be employed to establish the key locations.
4. Exploring downscaling techniques: To bridge the difference in spatial scales between proxy records and climate models, downscaling techniques could be very useful (e.g. Vrac et al. 2007). Downscaling assumes that the state and statistics of smaller scales are a function of the state and statistics of the larger scales. The details of the “downscaling” function are determined by the physiographic details of the considered region or locality. Different techniques will be assessed
5. Evaluating different forward modelling efforts: Great potential for improvement of model-data comparisons is offered by the forward-modelling approach, where appropriate process-based (physical, biological, chemical) or empirical models are driven by climate model output to simulate a proxy value or time series, which is then compared with the actual proxy data (e.g. Renssen and Osborn, 2003). This approach can deal explicitly with non-linear and non-invertible proxy response to multiple climate drivers, and can also aid our understanding of the processes responsible for the proxy behaviour (e.g. LeGrande et al., 2006). The state-of-the art will be evaluated.
6. Assessing the potential of data assimilation: Data assimilation also has the prospective to improve model-data comparisons. In data-assimilation, the climate model is driven by data, as is commonly applied in modern weather forecasting. In this type of simulations, meteorological observations are used to constrain the climate model toward the state presented by the data. This technique thus provides a climate state that has been reconstructed from a set of palaeodata, while remaining consistent with the model’s physics and the applied external forcings. Recently, this method has recently been successfully applied with proxy-based reconstructed temperature time-series covering the last millennium (e.g. Goosse et al., 2008, Crespin et al., 2009). The assimilation of palaeodata for earlier periods is still under development.
7. Carbon cycle modelling: The last glacial cycle was characterized by varying amplitudes and different patterns of changes in climate, vegetation cover, and atmospheric CO2. Models of the carbon cycle have been incorporated in climate models to study the links between the carbon cycle and climate (e.g. Joos et al., 2004). These models could provide important new insights on the role of carbon cycle feedbacks during past changes.