MODEL-BASED OBSERVING SYSTEM EVALUATION IN A WESTERN
BOUNDARY CURRENT: OBSERVATION IMPACT FROM THE COHERENT
JET TO THE EDDY FIELD
By Colette Kerry, Moninya Roughan, Shane Keating, and David Gwyther
ABSTRACT
Ocean forecast models rely on observations to provide
regular updates in order to correctly represent dynamic
ocean circulation. This synthesis of observations and mod-
els is referred to as data assimilation. Since initial condi-
tions dominate the quality of short-term ocean forecasts,
accurate ocean state estimates, achieved through data
assimilation, are key to improving prediction. Western
boundary current (WBC) regions are particularly challeng-
ing to model and predict because they are highly variable.
Understanding how specific observation types, platforms,
locations, and observing frequencies impact model esti-
mates is key to effective observing system design.
The East Australian Current (EAC), the South Pacific’s
WBC, is a relatively well-observed current system that
allows us to study the impact of observations on prediction
across different dynamical regimes, from where the current
flows as a mostly coherent jet to the downstream eddy
field. Here we present a review of the impact of observa-
tions on model estimates of the EAC using three different
methods. Consistent results across the three approaches
provide a comprehensive understanding of observation
impact in this dynamic WBC. Observations made in regions
of greater natural variability contribute most to constrain-
ing the model estimates, and subsurface observations
have a high impact relative to the number of observations.
Significantly, sampling the downstream eddy-rich region
constrains the upstream circulation, whereas observing the
upstream coherent jet provides less improvement to down-
stream eddy field estimates. Studies such as these provide
powerful insights into both observing system design and
modeling approaches that are vital for optimizing observa-
tion and prediction efforts.
INTRODUCTION
Accurate estimates of past, present, and future ocean
states are crucial to effective management of our ocean
environment and marine industries. Short-term ocean
predictions (days to weeks) are vital to myriad environ-
mental, societal, and economic applications, including
facilitating the adaptive management of marine ecosys-
tems, forecasting extreme weather events, predicting the
onset and persistence of marine heatwaves, providing
accurate ocean forecasts for shipping and military opera-
tions, predicting the fate of pollutants, and guiding search
and rescue operations.
Ocean state estimates require the combination of
numerical models and ocean observations, referred to as
data assimilation (DA). Observations provide sparse data
points while the model provides dynamical context. The
goal of DA is to combine the model with observations to
reduce uncertainty in the model estimate. For forecasting
purposes, model estimates are updated through assim-
ilation when observations become available and provide
improved initial conditions for the next forecast (Figure 1).
Due to the dynamic nature of the ocean circulation, ocean
models must be regularly updated through DA to, for exam-
ple, correctly represent the timing and locations of oceanic
eddies (e.g., Thoppil et al., 2021; Chamberlain et al., 2021).
A critical component of the DA problem is the way by
MODEL-BASED DESIGN AND EVALUATION
OF OBSERVING NETWORKS