Prediction of the Kuroshio Meander
with JMA Operational Ocean Assimilation-Prediction System (COMPASS-K)
Masafumi Kamachi*, Satoshi Sugimoto, Kumi Yoshita, Toshiyuki Sakurai, Toshiya Nakano
and Tsurane Kuragano
* JMA/MRI, 1-1 Nagamine, Tsukuba, 305-0052, JAPAN (mkamachi@mri-jma.go.jp)
An ocean data assimilation system (COMPASS-K) has been operated in the Japan Meteorological Agency
from 2001. Using the COMPASS-K, 84 cases of 90-days prediction experiments are conducted to assess
short-range (i.e., one month) prediction properties of the model, from 1993 to 1999, for the Kuroshio axis
variability. Each prediction is started from an initial condition that is obtained in a reanalysis experiment. The
predictions represent transition from straight to meander (and vice versa) of the Kuroshio axis. Real time
operational prediction is also conducted and it succeeds in predicting the Kuroshio large meander in August,
2004.
1. Introduction
persistency and to variability of observation. A
time scale when the error of the single trajectory
Observations so far have revealed that the
reaches to the valiability of observation or errors
Kuroshio axis has significant variabilities south of
of climatology/persistency has been adopted as a
Japan. Previous studies (e.g., Kawabe, 2003)
predictability. US Navy group showed the time
showed that the Kuroshio takes three typical
scale is about 30 days (e.g., Hurlburt et al., 2000).
axes: nearshore non-large-meander, offshore
Kyoto university group obtained the time scale,
non-large-meander, and typical large-meander.
which depends on the transition stages from
Many observational, theoretical and numerical
straight to meander (90 days) and vice versa (30
studies
and
days) in their shallow water model. They clearly
e.g.,
showed that the difference depends on the
have
mechanisms
of
reported
the
descriptions
variabilities
(see,
Kawabe, 2003 for a review and references of the
vorticity balance (Komori et al., 2003).
studies of the Kuroshio axis). Under consideration
We predict the Kuroshio axis with all model
of the results of the previous studies, modeling
state variables in one month using an ocean
and data assimilation techniques may be able to
general circulation model (OGCM) used in the
predict the variability of the ocean state south of
operational data assimilation system described in
Japan, especially Kuroshio axis.
Kamachi et al. (2004b). We also carried out a
Some
studies
have
conducted
single
trajectory prediction experiments in which a single
successful operational prediction of the 2004
Kuroshio large meander.
prediction run started from an initial condition. The
error of the prediction results has been compared
to error growths of climatological condition and
2. System and performance
The COMPASS-K system consists of an
OGCM,
analyses
of
observed
data,
and
height anomalies (SSHAs) were extracted from
the TOPEX altimeter data by an along-track filter.
assimilation techniques (Kamachi et al., 2004b).
Space
and
time
The model is an eddy permitting version of the
calculated from each meso- and large-scale
OGCM of the Meteorological Research Institute.
SSHAs. The decorrelation scales are different for
The area covered by the model calculation is from
meso- and large-scale SSHAs and also for
119˚E to 109˚W, and 12.5˚N to 55.5˚N in the North
different regions. Each meso- and large-scale
Pacific. The grid spacing gradually changes from
SSHAs
1/4˚x1/4˚ in the region from 23˚N-45˚N and 120˚E
horizontal model grids with the space and time
-180˚ to 0.5˚ in latitude and to 1.5˚ in longitude
decorrelation
outside of the region. The model has 21 vertical
(Kuragano and Kamachi, 2000). The SSHAs were
levels. There are 5 levels in the upper 200m.
then projected to subsurface temperature and
Coastal and bottom topography data of ETOPO5
salinity anomalies using regression coefficients.
was used to specify the model bottom topography.
Finally,
A slope-advective bottom topography scheme is
subsurface
also adopted for the correct calculation of flows
temperature and salinity, temperature and salinity
around a steep bottom topography (Ishizaki and
in each grid are obtained from the surface to 1750
Motoi, 1999). The generalized Arakawa scheme
m every 5 days for 1993 to 2001. Calculated
is adopted for the momentum advection terms
5-day mean values of temperature and salinity
(Ishizaki and Motoi, 1999). The values of the
are adopted for the assimilation experiments in
coefficients of the viscosity and diffusivity are
the study. A time-retrospective nudging method is
changed according to the Richardson’s 4/3-power
adopted for the assimilation (see Kamachi et al.,
low. The model uses a rigid lid approximation.
2004b for the method).
were
decorrelation
optimally
scales
adding
the
values
in
interpolated
each
mesoto
scales
model
and
the
to
are
the
grid
large-scale
climatological
The model was spun up for 180 years to
The assimilation (reanalysis) experiment is
obtain a statistical equilibrium state. After the
conducted with historical observation data from
spin-up, the model was again spun up for 7 years
1993 to 2001, and Japan Meteorological Agency
using a seasonal wind stress averaged for
(JMA) started an operational nowcasting from
1980-1998 of NCEP daily wind stress (Kalnay et
2001. The reanalysis data from 1993 to 2001 are
al., 1996). The model was then integrated from
examined with observed data (e.g., Kamachi et al.,
1980 to 2001 using the NCEP daily mean wind
2004b).
stress.
When we use a rigid-lid OGCM and an
3. Prediction experiments and results
assimilation method such as nudging, we need a
statistical
evaluation
of
the
subsurface
We produced outputs every 5 days from the
temperature and salinity fields. Temperature and
reanalysis experiment, from 1993 to 1999. We
salinity fields were calculated from satellite
selected the initial conditions once a month for
altimetry and in situ data using a four-dimensional
this study from the outputs. The prediction
space-time optimum interpolation and vertical
experiments started from the initial conditions and
projection. Meso- and large-scale sea surface
were integrated for 90 days. We, therefore, had
84 cases of 90-days prediction experiments for
the Kuroshio axis variability.
We conducted
a
real
time
operational
prediction of the 2004 Kuroshio large meander
The predictions represent transition from
using
the
operational
assimilation-prediction
straight to meander of the Kuroshio axis, and the
system. We show an example of the successful
results
the
predictions. Initial condition of the prediction is an
mechanisms, submitted by previous studies, of
output of the reanalysis: June 4th, 2004 (Fig. 1a).
the
and
The prediction period is 60 days (to August 3rd).
interaction that act as a trigger of the meander
Figure 1 (b) shows the predicted state that shows
and self-sustained oscillation (Kamachi et al.,
the large meander of the Kuroshio (white square
2004a). The reanalysis shows the meander
area in the figure). The predicted Kuroshio axis is
evolution by the eddy activity. Simulation (no
similar to the observed one (here Fig. 1 (c) shows
assimilation) shows no meander state even with
an assimilated result as a real state). According to
the same atmospheric forcing as the prediction.
a series of the prediction experiments, JMA had a
Therefore, the initial condition contains the
press release and warned the change of the
information of the meander and the system can
ocean state (esp., temperature and surface
represent
(standard
velocity fields) in May 11th. In August the
deviation) values of the axis error for all 84 cases
Kuroshio large menader occurred. It is the first
are 13, 17, and 20 (10, 10, and 12) km, in 138.5E,
time to predict the large meander successfully.
in
are
transition
the
the
30-,
analyzed
with
according
eddy
evolution.
60-,
and
to
propagation
Mean
90-days
predictions
The nowcasting data calculated with the
respectively. Observed mean deviation from
system are opened through the Japan GODAE
seasonal variation is 30 km. Then the predictive
server (http://godae.kishou.go.jp)
limit of the system is about 80 days. The gradual
decrease of the amplitude in a stage from
meander to straight axes is also predicted. The
predictive limit is about 20 days (see Kamachi et
al., 2004a).
(a)
(b)
(c)
Fig.1 Comparison of surface velocity fields. (a): Initial Condition, June 4 (straight path);
(b): 60-days Prediction, August 3 (meander path); (c): Observation (assimilation), August
3 (meander path).
References:
Hurlburt, H. E., O. M. Smedstad, R. C. Rhodes, J.-F.
Kamachi, M., T. Kuragano, H. Ichikawa, H. Nakamura,
Cayula, C. N. Barron and E. J. Metzger (2000): A
A. Nishina, A. Isobe, D. Ambe, M. Arai, N.
feasibility
model
Sugimoto, K. Yoshita, T. Sakurai, and F. Uboldi
eddy-resolving nowcast/forecast skill using satellite
(2004b): Operational data assimilation system for the
altimeter data. Naval Research Laboratory Report
Kuroshio south of Japan: Reanalysis and validation.
NRL/MR/7320—00-8235, 23 pp.
J. of Oceanogr., 60, 303-312
demonstration
of
ocean
Goda, S.
Ishizaki, H. and T. Motoi, (1999): Reevaluation of the
Kawabe, M. (2003): Study on variations of current path
Takano-Oonishi scheme for momentum advection on
and transport of the Kuroshio. Oceanography in
bottom relief in ocean models. J. of Atmos. Ocean.
Japan, 12, 247-267 (in Japanese).
Tech., 16, 1994-2010.
Kalnay, E., and Coauthors (1996):
Komori, N., T. Awaji, Y. Ishikawa, and T. Kuragano
The NCEP/NCAR
(2003): Short-range forecast experiments of the
40-Year Reanalysis Progect. Bull. Amer. Meteor.
Kuroshio path variabilities south of Japan using
Soc., 77, 434-471.
TOPEX/Poseidon altimetric data. J. Geophys. Res.,
Kamachi, M., T. Kuragano, S. Sugimoto, K. Yoshita, T.
108, N0. C1, 3010, doi:10.1029/2001JC001282.
Sakurai, T. Namano, N. Usui, and F. Uboldi (2004a):
Kuragano, T. and M. Kamachi, (2000): The global
Short-range prediction experiments with operational
statistical space-time scales of oceanic variability
data assimilation system for the Kuroshio south of
estimated from the TOPEX/POSEIDON altimeter. J.
Japan. J. Oceanogr., 60, 269-282.
Geophys. Res., 105, 955-974.