Adapting data assimilation


Adapting data assimilation methods to study decadal to multi-century climate variability



The goal of data assimilation is to directly combine the information provided by proxy records (which are indirect measures of past climate variability), the estimates of past climate forcing, and the physics embodied in climate models in order to provide the best reconstruction of past changes. Data assimilation has been successfully applied in many fields but is still not standard in paleoclimatology. Our goal here is to test and adapt the available data assimilation methods to study the decadal to multi-century variability observed during the past millennia. We focus on ensemble methods, in particular the particle filters.

Time series (in °C) of annual mean surface temperature in a simulation assimilating proxy data (black) for the European region (defined here as the continental region between 0–50°E, and 30–70° N). The green curve is the mean over the same region derived from instrumental data. The grey lines represent a range of uncertainty of the simulation. The temperature is given as the difference with the values for the reference period 1850–1995. All the time series have been decadally smoothed. For more details see, Goosse et al. (2012)