A Data Assimilation System for Costal Ocean RealTime
- Slides: 29
A Data Assimilation System for Costal Ocean Real-Time Predictions Zhijin Li and Yi Chao Jet Propulsion Laboratory, California Institute of Technology James C. Mc. Williams (UCLA), Kayo Ide (UMD) ROMS Meeting, April 5 -8, 2010, Hawaii 1
Outline 1. Developed costal ocean data assimilation and forecasting systems 2. Recap on the three-dimensional variational data assimilation 3. A multi-scale three-dimensional variational data assimilation 4. Summary 2
2003 Autonomous Ocean Sampling Network (AOSN) Experiment 3
Southern California Bight Real-Time System HF Radar Observation Data Assimilation 4 http: //ourocean. jpl. nasa. gov
Prediction of Drifter Trajectories in the Prince William Sound L 0 10 km L 1 3. 6 km L 2 1. 2 km Oil Spill: 1989 Exxon Tanker Wreck , Prince William Sound, Alaska 5
PWS 2009 Field Experiment Ensemble of Co-located ROMS Simulated Trajectories 6
Data Assimilation and Forecasting Cycle 3 -day forecast xf Initial condition Aug. 1 00 Z Time scales comparable with those of the atmosphere 7 6 -hour forecast 6 -hour assimilation Time cycle xa Aug. 1 06 Z Aug. 1 12 Z Aug. 1 18 Z Aug. 2 00 Z
A There-Dimensional Variational Data Assimilation (3 DVAR) 1. Real-time capability 2. Implementation with sophisticated and high resolution model configurations 3. Flexibility to assimilate various observation simultaneously 4. Development for more advanced scheme (Li et al. , 2006, MWR; Li et al. , 2008, JGR, Li et al. , 2008, JAOT) 8
Weak Geostrophic Constraint: Decomposition of Balanced and Unbalanced Components Geostrophic balance Geostrophic sea surface level Ageostrophic streamfunction and velocity potential 9
Kronecker Product Formulation of 3 D Error Correlations 10
Inhomogeneous and anisotropic 3 D correlations Non-steric SSH correlations (Li et al. , 2008, JGR) 11 Cross-shore and vertical section salinity correlation
Assimilation of Multi-Satellite SSTs and SSHs JASON-1 Infrared and Microwave SST 12 Sea Surface Heights
Assimilation of Real-Time High Frequency Radar Velocities 2008 -12 -08 Short distance: 100 km, res of 1 km, 5 MHz Long distance: 200 km, res of 5 km, 25 MHz 13 http: //www. cocmp. org/ http: //www. sccoos. org/
Performance of ROMS 3 DVAR: AOSN-II, August 2003 TEMP(C) Comparison of Glider-Derived Currents (vertically integrated current) SALT(PSU) Glider temperature/salinity profiles 14 (Chao et al. , 2009, DSR) Black: SIO glider; Red: ROMS
Southern California Coastal Ocean Observing System (SCCOOS) SIO Glider Tracks Motivation: assimilating sparse vertical profiles along with high resolution observations for a very high resolution model 15
Multi-Scale Data Assimilation: Concept Background Observation Multi-scale DA (Boer, 1983, MWR) 16
Multi-Scale Data Assimilation: Scheme Large Scale Sparse Obs Small Scale High Resolution Obs 17
Twin Experiments: Observations • Model resolution of 1 km • SSTs and surface velocities at 2 km by 2 km • T/S profiles 1. at 10 km by 60 km (ideal) 2. at 10 km by 180 km (real) 18
Root-Mean Squared Errors (RMSEs) at 30 m 19
Root-Mean Squared Errors (RMSEs)at 50 m 20
RMSEs NO-DA 3 DVAR MS 3 DVAR 21
RMSEs NO-DA 3 DVAR MS 3 DVAR 22
SCB Operational System: 3 DVAR vs MS 3 DVAR 23
HF Radar and Data Assimilation Analysis Velocities Standard 3 DVAR 24 MS-3 DVAR Correlation RMSE U 0. 62 0. 13 m/s 0. 75 0. 11 m/s V 0. 68 0. 11 m/s 0. 82 0. 08 m/s
Summary Ø A 3 DVAR system has been developed with unique formulations for coastal oceans. Ø The MS 3 DVAR system has been demonstrated significantly better skill and computational efficiency, and it has been implemented in operational applications. Ø For more information on real-time data assimilation and forecasting systems: http: //ourocean. jpl. nasa. gov 25
Backup 26
MS 3 DVAR Work Flow Obs (Glider, Satellite, HF radar, etc) Forecast Large Scale (LS) LS-3 DVAR Increment 27 Small Scale (SS) SS-3 DVAR
ETKF vs MS-3 DVAR in Twin experiments • Observations: HF radar velocities and SSTs, along with Sparse T/S profiles • ETKF continuously reduces RMSEs because of the predicted error covariance, while MS-3 DVAR more effectively fit to high resolutions observations at the early stage RMSE, ETKF 28 RMSE, MS-3 DVAR 28
A Hybrid Ensemble MS-3 DVAR (Lorenc 2003) Applied to the small-scale components 29