Advanced Data Assimilation Strategies in Modern

1
Advanced Data Assimilation Strategies in
Modern Observational Networks For Real-Time, High
Resolution Applications AMS-Albuquerque-2001
Hernan G. Arango Rutgers University, New
Brunswick, NJ Pierre J.F. Lermusiaux Harvard
University Scott M. Glenn Rutgers University
2
What is Data Assimilation??
A Melded Estimate of Data and Dynamics
3
Overall Goal Determine, Develop and validate a
comprehensive data assimilation system for the
estimation of real oceanic fields
4
(No Transcript)
5
Main ESSE Components
6

ESSE Flow Diagram
DE0/N

DP0/N
-
-

Most Probable Forecast

Synoptic Obs
A Posteriori Residules dr ()
Historical, Synoptic, Future in Situ/Remote
Field/Error Observations d0R0

-
-
Data Residuals
Measurement Error Covariance

d-CY(-)
Ensemble Mean



eqYj(-)
Gridded Residules

Y(-)

-


j1
Y()
Y()
Y1 Yj Yq

-
Y1 Yj Yq

0

-
E(-) P(-)

-
0



-
/-

E0 P0
0
jq
uj(o,Ip) with physical constraints
Continuous Time Model Errors Q(t)
Key
Ea() Pa()

E() P()
Field Operation Assumption
7
meters
State Vector
241 x 111 x 25
Z u v u
v T S
26,751 26,510
26,751 662,750 666,000
668,775 668,775 2,746,201
-
-
8
Observational Network July 2000
9
CODAR Normalized Dominant Error Covariance
x10-2
10
Navy Products
NOAA Rutgers
Global Atmospheric Forecasts
NOGAPS
NCEP
I.C. B.C.
I.C. B.C.
Atmosphere- Ocean Nowcasting/ Forecasting System
Local Atmospheric Forecasts
COAMPS 27 km 6 hours
RAMS 4 km 30 min
Atm. Forcing
Atm. Forcing
Ocean Models
MODAS (POM)
ROMS
PBL SBL BBL WBL
I.C. B.C.
Waves
WAM
Wave Models
11
Validation at Thermistor String CS2
RMS0.6865
RMS0.7972
RMS1.4863
Observations Model Estimate
RMS3.7447
RMS1.2795
12
A-Line Temperature Cross Section
CS2 CS4
13
Model Observation Comparison at Thermistor
String CS2
Low Wind Forcing High Wind
Forcing
Real Data Model Estimate
14
Conclusions The ESSE approach
is a nonlinear, robust and efficient
data assimilation scheme for the realistic
estimation of oceanic fields and
their associated error and variability
covariances. As expected, in
coastal regions, the dominant error
covariance is non-isotropic with larger
along-shore scales. In
shallow coastal regions, the success of ocean
data assimilation is affected by the
accuracy of atmospheric forecasts,
specially during high wind events.
15
ROMS/COAMPS ROMS/RAMS
MODAS COAMPS
16
For more information contact Hernan G.
Arango Rutgers University, New Brunswick,
NJ 732-932-6555 x266 arango_at_imcs.rutgers.edu h
ttp//marine.rutgers.edu/cool http//marine.rutger
s.edu/po/models/roms.html