In this chapter, we have discussed some recent developments that aim to optimize the combination of proxy data and climate model results, with a focus on the climate of the past millennium.
One development is to evaluate temperature reconstructions based on proxy records by using independent "pseudo-proxy records" that are derived from model experiments. An important application of this method is the validation of the reliability of the statistical CFR method that has been used for several temperature reconstructions of the past millennium and improved since its introduction.
Another recent development is to run climate models to compute the climatic "fingerprint" associated with a particular forcing, and subsequently to apply multiple regression to find the combination of fingerprints that gives the best match with proxy-based temperature reconstructions. This technique has provided a detailed analysis of the impact of different forcings on the climate of the past millennium, including the important effect of anthropogenic forcings on the late 20th century climate. Model and proxy data could also be used to constrain model parameters as well as important characteristics of the climate system.
Progress is also made by the application of data assimilation techniques that use both proxy data and models' results in order to obtain a climate state that is consistent with the proxy records and model physics. Once such a climate state is obtained, aspects of this state that are not registered in proxies can be analyzed in detail, thus providing additional information on the climate of the past millennium. In addition, this climate state can be used to find key locations for data collection. At the moment, the number of suitable proxy data and the computation time involved still limit the application of data assimilation methods for the past millennium.
The examples displayed here have shown that the combined analysis of proxy and model results is certainly an interesting way to improve our understanding of past climate changes. The examples also show that several difficulties remain that need intense collaboration between scientists from different backgrounds to solve them. Any improvement in the representation of past changes by models or in the number and quality of proxy records would certainly be very beneficial. The method used to combine proxy data and models, however, needs also to be refined and different methods could be required if one is interested in reconstructions of past changes or in understanding the causes of those changes. In particular, estimating precisely the impact on the results of the analysis of the various hypotheses or prior assumptions involved in the method, sometimes hidden in complex formulations that could be assessed by only a few specialists, is certainly an important task.
Was this article helpful?