Schematic representation of the procedure leading to a reanalysis (i.e. a reconstruction of the state of a system) using data assimilation (modified from Goosse 2016).
The Paleoclimate Reanalyses, Data Assimilation and Proxy System modeling (DAPS) working group was designed to stimulate the development and the application of methods for the joint use of observations and models in paleoscience. That required that models simulate the variables measured for a direct comparison and to apply techniques that optimally combine both sources of information, taking into account the uncertainties.
- Review and evaluate existing methodologies.
- Develop synergies and joined activities where gaps are identified.
- Provide practical applications and training to potential users.
- Stimulate the expansion of proxy system modeling and data assimilation in new areas.
DAPS was a PAGES working group active from 2016-2019.
Reanalyses combine observations with the knowledge of the dynamics of a system, as represented in a model, to obtain an estimate of the state of this system. They have some clear advantages compared to more traditional methods. In particular, the data assimilation techniques that allow blending observations and model results do not rely on the stationarity of a statistical relationship between the record and the reconstructed variable. Reanalysis provide physically-consistent estimates for different variables such as temperature, precipitation, atmospheric and oceanic circulations. Furthermore, they take into account explicitly the uncertainties on all the available sources of information in one single process in order to reduce the uncertainty of the reanalysis itself.
Nevertheless, many challenges needed to be addressed to apply them more systematically in paleosciences. Specifically, the data assimilation techniques needed to be adapted to observations with large and poorly known systematic uncertainties arising from resolution, chronology as well as to the complex response in those records to climatic and non-climatic factors, and to biased observing networks. To obtain unbiased results, a model-data fusion required the development and inclusion in the process of forward (proxy system) models that explicitly reproduced the observed quantity from model outputs allowing the measured variable to be directly assimilated into simulations.
Schematic representation of the procedure leading to a reanalysis (i.e. a reconstruction of the state of a system) using data assimilation (modified from Goosse, 2016).
During the DAPS workshop in May 2019, the group performed a metadatabase hackathon. Feng Zhu, an ECR participant, kindly turned this into a website, a "nexus" for data assimilation and proxy system modeling, where people interested in the subject can find information of the data needed, the types of existing models, their characteristics, and how to find the code: https://daps-pages.github.io/
Hugues Goosse (Université catholique de Louvain, Belgium) (mailing list administrator) Mike Evans (University of Maryland, USA)
Samar Khatiwala (Oxford University, UK)