regime (Kerry and Roughan, 2020). This downstream trans-
port is constrained primarily by observations over the eddy
field but is also impacted by the EAC array, the northern
XBT lines, and the HF radar observations (downstream
impacts, Figure 3b).
Normalizing the impacts by the number of observations
(e.g., Figure 3a,b) reveals that observations over the eddy
field make the greatest contribution to volume transport
estimates along the coast. SSH, SST, and Argo observa-
tions made in the region of high eddy variability (33°–37°S)
have more impact than the same observations made else-
where as they provide information to constrain the variable
region. Even for volume transport estimates where the jet
is mostly coherent, satellite and Argo observations of the
(downstream) eddy field have greater impact than the
same observation types upstream (Figure 3a). The eddy
field observation impact exceeds the impact of observa-
tions local to 28°S.
Subsurface observations that sample hydrography within
EAC eddies, such as those from Argo, gliders, and XBTs,
are also particularly impactful (Figure 3a,b). Observations
made in the upper 500 m of the water column contribute
more to changes in the circulation estimates than deeper
observations (Figure 4a,b). When glider observations sam-
ple eddies offshore of the continental shelf (Figure 2c), they
have large impacts on EAC transport and EKE (contribut-
ing to 28%–36% of transport increments, and 38% for EKE;
Kerry et al., 2018).
METHOD 2: OBSERVING SYSTEM EXPERIMENTS
Observing System Experiments (OSEs) compare the results
of a DA system that withholds certain observations with a
system that includes them (e.g., Chang et al., 2023). Using
the EAC-ROMS configuration for 2012–2013, we compared
the impact of assimilating only the more traditional obser-
vations (satellite-derived SSH and SST, and vertical profiles
from Argo and XBTs: the TRAD experiment), versus also
including data from more novel observation platforms (HF
radar, deep and shallow moorings, and gliders: the full
suite of all available observations, the FULL experiment;
Siripatana et al., 2020).
While the overall surface and subsurface properties
FIGURE 4. Summary of subsurface observation impacts. (a) Observation impacts using the adjoint-based method on transport through the shore
normal section crossing the coast at 28°S (upstream) grouped into depth bins and normalized by the number of observations. (b) Same as (a) but
for transport through section crossing the coast at 34°S (downstream). Adapted from Kerry et al. (2018) (c) OSSEs show the depth region of greatest
variability (>500 m) benefits most from subsurface observations (Gwyther et al., 2022). (d) OSEs show improvement in shelf velocities with mooring
data assimilated (Siripatana et al., 2020). (e) Example of glider data (glider path shown in red) constraining the subsurface temperature and velocity
structure of a cold core eddy off Sydney (Siripatana et al., 2020). Text in the black boxes summarizes parallels between the information in panels a–b
and that in panels c–e. SEQ = Southeast Queensland. CH and COFFS = Coffs Harbor. SYD = Sydney.
Absolute value of impact per observation (Sv)
Dense subsurface observations
constrain eddy subsurface structure
Data from shelf moorings improve
velocity representation
FULL
TRAD
Transport through 28oS
Transport through 34oS
Depth of observations
i)
iii)
ii)
vi)
Observations that constrain
the variable upper ocean
(>500m) are most impactful
SEQ200m
CH100m
SYD100m
28oS