Coupled atmosphere-ocean data assimilation experiments with a low-order climate model

Decadal climate prediction has emerged as an important yet challenging component of the seamless weather-climate prediction paradigm. Two important challenges to improving predictions on intra-annual and longer timescales are: (1) obtaining accurate joint representations of atmospheric and oceanic states, including errors, over a wide range of time scales; and (2) dealing with relatively sparse oceanic observations, particularly prior to the late 20th century, in producing initial conditions for historical forecasts that may be verified. We propose that developing this ability to verify historical forecasts is a key aspect supporting the development of improved initialization strategies.

In order to explore the utility of different strategies for coupled ensemble data assimilation over long periods of time, we use a simplified, idealized, climate model. Specifically, fundamental issues are explored using an ensemble data assimilation framework with a modified version of the model described in Roebber (1995). The model simulates the behavior of the North Atlantic climate system, involving interactions between large-scale atmospheric circulation and ocean states driven by the variability of the meridional overturning circulation (MOC). A wide range of exploratory experiments are performed to determine the importance of joint assimilation of atmosphere and ocean observations, and to assess the extent to which the deep ocean can be initialized given only atmospheric observations. Results are discussed in light of guidance they provide toward the development of next-generation decadal climate prediction systems, and verification of forecasts from these systems.