GEOS/Ecco Coupled Model and Data Assimilation System for Decadal Prediction

The societal importance of long-range climate predictions out to decadal scales has motivated increased
research activity and development of dynamical models for decadal prediction. Predictability over these
timescales relies on the decadal-scale variability in the earth system, generally characterized independently
in the Atlantic, Pacific, and Indian Oceans. Recent reviews and studies discuss the needed ‘ingredients’
for accurate decadal prediction. These include a model that can properly capture the propagation of the
low frequency modes, and the proper initialization of the phase of the low frequency modes in the coupled state,
particularly of the ocean. The final ‘ingredient’ for skill in decadal prediction is related to the fidelity
of the surface turbulent heat fluxes.

The study presented here reports on the development of a Goddard Earth Observing System/Estimating the Circulation
and Climate of the Ocean (GEOS/ECCO) Atmosphere-Ocean Data Assimilation System (AODAS). The GEOS/ECCO AODAS combines
the GEOS and ECCO modeling and assimilation capabilities in order to improve initialization of low-frequency modes. The GEOS/ECCO AODAS structurally resembles the GEOS-S2S AODAS, and includes a predictor sequence, during which the
atmosphere is constrained by an atmospheric assimilation and "drops off" the near-surface atmospheric ‘forcing’ fields,
the net radiation and the precipitation and discharge needed to drive an offline ocean model. At the end of a 2-month
assimilation window, an offline GEOS/ECCO ocean model and adjoint estimation sequence is conducted driven by the forcing
from the predictor sequence, which is restricted to allow for changes in the surface turbulent fluxes only. When this
is completed a corrector sequence follows, constraining the surface turbulent fluxes in the coupled model to the fluxes
computed during the adjoint estimation.

The approach here relies on the ability of the ECCO state estimation to properly represent the low frequency
modes of variability that reside in the ocean, the importance of the surface fluxes generated by ECCO estimation for the
representation of the low frequency modes in the coupled system, and the ability of the proper initialization of the
phase of the decadal modes to result in increased decadal prediction skill. A series of decadal forecasts initialized
from the GEOS/ECCO AODAS are conducted to evaluate the resulting decadal forecasts. The evaluation of the these forecasts
hinges on the behavior of the low frequency modes that are responsible for predictability at those time scales.