1. ROMS User Workshop, October 2, 2007, Los Angeles
A Three-Dimensional Variational Data Assimilation in Support of Coastal Ocean Observing Systems
Zhijin Li and Yi Chao
Jet Propulsion Laboratory
Jim McWilliams and Kayo Ide
UCLA
As I was preparing my talk today during the past few days, it is clearly very difficult to decide how broad my talk should cover and how details I should penetrate on individual topics. In the end, I decide to give a relatively broad talk covering the variety of topics that I am working on. If you are interested in hearing more about a particular topic, I will be more than happy to talk to you after the seminar. Here is my and phone number, so feel free to call me to follow up. JPL is only 3 miles from here, so you are certainly welcome to visit us up the hill.
Before I go into my talk, I want to emphasize that this is very important and exciting time for oceanography, although many of my colleagues expressed great frustrations during this transition and uncertain time.
Oceanography before 1960s involves many energetic individuals who charted ships into the restless sea for weeks and sometimes months trying to observe the ocean. They came back from these long cruises, very different from the vacation cruises you and I are familiar with, very tired and spend the next few years processing the data, and eventually publish them in journal papers. Once the ship data are documented and papers are published, they wrote the next proposals to pick another piece of the world ocean to explore.
The next three decades from 1960s-1990s, oceanographers are gradually getting organized into field programs or expeditions. Easy to get funded, simultaneous measurements are required and have to be funded at the same time. Many oceanographers are working together, involving a fleet of large ships, centralized data center, and the associated modeling program. During the same time, there are major breakthroughs in satellite oceanography, measuring a number of ocean surface parameters from space.
Now in the 21st century, the field of oceanography is facing another major challenge. To answer many of the important questions, there is a need for a much closer coordination between various disciplines in oceanography, ranging from physical, chemical to biological oceanography. Global sea level rise, carbon cycle. Ocean scientists need also to work with experts in the information technology ranging from data management, scientific computing, and visualization. Advanced sensors are also needed to measure those parameters that have never been measured before, so scientists need to work with engineers and technologists.
There is an emerging need for shared infrastructure, in the same way as the astronomers sharing the telescope, and physicists sharing the accelerator. So the next few decades, oceanographers will learn how to develop this community-based observatory. What I am going to talk about today is the companion piece of the ocean observatory, also known as the virtual observatory, and the new term within the National Science Foundation is called Cyberinfrastructure. My talk has certain bias toward the coastal ocean simply because of its strong link to people and societal applications.
ROMS User Workshop, October 2, 2007, Los Angeles

2. Feedback & Adaptive Sampling
Integrated Ocean Observing System (IOOS): Modeling, data assimilation, forecasting and adaptive sampling
Theoretical
Understanding &
Numerical
Models
Users:
Managers
Education &
Outreach
Observations
(satellite, in situ)
Data
Assimilation
Products
Information
Observing
System Design
Feedback & Adaptive Sampling

3. Outline Real-time Regional Ocean Modeling System (ROMS)
Three-dimensional variational data assimilation
Assimilated observations
Evaluation of analyses and forecasts
Observing system experiments (OSE)
Relocatability

4. Coupled with Tides Sea Surface M2 Tidal Currents RMS Error of SSHs
Tide Gauge
Sea Surface M2
Tidal Currents
ROMS Simulation
HF Radar Obs
RMS Error of SSHs

5. Regional Ocean Modeling System (ROMS): From Global to Regional/Coastal
Modeling Approach
15-km
5-km
1.5-km
Regional Ocean Modeling System (ROMS): From Global to Regional/Coastal
12-km
Multi-scale (or “nested”) ROMS modeling approach is developed in order to simulate the 3D ocean at the spatial scale (e.g., 1.5-km) measured by in situ and remote sensors

6. Data Assimilation
Data Assimilation
Model
When the numerical model is so good as its prediction
is superior to the climatological (almanac) forecast

7. ROMS Analysis and Forecast Cycle:
Incremental 3DVAR
y: observation
x: model
3-day
forecast xf
6-hour
forecast
6-hour
assimilation
cycle xa
Initial
condition
Time
Aug.1
00Z
Aug.1
06Z
Aug.1
12Z
Aug.1
18Z
Aug.2
00Z

8. Why a There-Dimensional Variational Data Assimilation
Real-time capability
Implementation with sophisticated and high resolution model configurations
Flexibility to assimilate various observation simultaneously
Development for more advanced scheme

9. 3DVAR: Weak Geostrophic Constraint and Hydrostatic Balance
Geostrophic balance
Vertical integral of the hydrostatic equation
ageostrophic streamfunction and velocity potential
The basic variables for an oceanic model are sea surface height, velocity u/v, temperature and salinity. You can compare them to atmospheric variables, surface presure, u/v, temperature and moisture. We consider two weak constraints: geostrophic and hydrostatic balance. With the geostrophic balance, we have the variables from u/v to ageostrophic streamfunction and velocity potential. With hydrostatic balance, we have the variable of non-steric sea surface height. Non-steric SSH variation is purely due to the mass convergence.

10. Inhomogeneous and anisotropic 3D Global Error Covariance
Cross-shore and vertical
section salinity correlation
SSH correlations
Kronecker Product

11. Assimilated observations: Satellite infrared SSTs
Infrared, High resolution
Cloud contamination
Microwave, Low resolution (25km)
No cloud contamination
NOAA
GOES
NASA Aqua
AMSR-E
NASA TRIMM
TMI
NOAA
AVHRR
Cross validation. Another important aspect is the difference btewwen the bulk SST and skin/subskin SST. Currently, We use winds. When wind peed is weaker that 3.5m/s, no satellite SSTs
Are assimiulated.

12. Assimilated observations: Satellite SSHs along track
JASON-1
Resolution: 120km cross track, 6km along track

13. Integrated Ocean Observing Systems
AOSN-II
T/S profiles from gliders
Ship CTD profiles
Aircraft SSTs
AUV sections

14. Assimilated Current Observations
Shipboard
Bottom
Buoy
Acoustic Doppler Current Profiler (ADCP)
High Frequency Radar
Mapped 2D surface current

15. 3DVAR with First Guess at Appropriate Time (FGAT)If
FDAT 3DVar is equivalent to 4DVar

16. ROMS Performance Against Assimilated Data
August 2006 Mean
All Gliders
Mean Diff
RMS Diff
Temperature (C)
-0.3
0.3
0.0
0.75
Salinity (PSU)
-0.1
0.1
0.0
0.20

17. Comparison of Glider-Derived Currents (vertically integrated current)
AOSN-II, August 2003
Black: SIO glider; Red: ROMS
Off shore, T/S can constrait u/v much better. MB06 is near shore. This is AOSN From Russ Davis.

18. Predictability during AOSN-II
Forecast Correlation
RMSE
We calc correlation. It is about forcast going wown. Again, anaysis to 24h forcast are marginally useful.
Note: because gliders are moving, one cannot estimate the persistence

19. Observing System Experiment (OSE)
– Typically aimed at assessing the impact of a given existing data type on a system
– Using existing observational data and operational analyses, the candidate data are either added to withheld from the forecast system
– Assessing the impact

20. Observing System Experiment (OSE): Glider Data Denial Experiment
Temperature
RMS Error
Salinity
w/o CalPoly glider
with CalPoly glider
CalPoly
SIO
WHOI
1st week
2nd week

21. ROMS without HF radar data assimilation ROMS with
Impact of
HF Radar
HF radar
ROMS without HF radar data assimilation
ROMS with
HF radar data assimilation

22. Southern California Coastal Ocean Observing System (SCCOOS)
Southern California Bight
US WEST COAST

23. Real-Time SCCOOS Data Assimilation and Forecasting System

24. Evaluation with HF radar velocities

25. Toward a Relocatable ROMS Forecasting System: Demonstration for Prince William Sound, Alaska
9-km
3-km
1-km