Impact of glider data assimilation on model predictions during AOSN II and MB2006 experiments. Igor

1
Impact of glider data assimilation

on model predictions
during
AOSN II and MB2006 experiments. Igor Shulman,
Clark Rowley, Jim Cummings, John Kindle,
Stephanie AndersonNaval Research Laboratory,
SSCSergio DeRada, Jacobs-Sverdrup, Inc., Peter
Sakalaukus, USM Acknowledgments Support from
NRL Base, ONR 32Collaborations with AOSN and
MURI ASAP groupsWorkshop on Environmental
Modeling of California Central Coast, MBARI,
August 2007.
2
Equipment Sampler from the AOSN Program
3
Goals and Motivation
NRL RO 6.1 BIOSPACE (FY08-12) Kindle-PI,
Co-PIsArnone, Shulman, Teague Objective combine
satellite, in-siitu and glider observations for
1-5 days BSE predictions.
NAVO and NRL have been acquiring fleet of
gliders
NRL LBSFI (FY06-10) (Jacobs-PI) Objective
Glider observations are being integrated into
relocatable modeling systems at NAVO
Glider data assimilation Impact of glider data
on BSE predictions
NRL 6.2 Variational Data Assimilation
(FY07-09) PINgodock, Co-PIs Cummings, Shulman,
Smith Objective Compare sequential and
variational data assimilation approaches
Commitment to West Coast Modeling Test and
evaluation of techniques and approaches are being
conducted in Monterey Bay and California Current
System
4
Modeling Approach
MONTEREY BAY NCOM ICON
REGIONAL NCOM CCS
GLOBAL NCOM
1-4 km

• Utilize Ocean Models which are being transitioned
  to NAVO
• NCOM (global and relocatble NCOM), HYCOM (global
  is planned to be transitioned)
• Couple Across scales
• Global Regional Coastal Local
• Force with COAMPS Fluxes 81/27/9/3 km grid
• Observational data from gliders, aircraft, AUVs
  and ship are assimilated by using the NRL Coupled
  Ocean Data Assimilation System

MONTEREY BAY NCOM frsICON
0.5- 1.5 km
5
Approach (continued)
Ocean Obs
Utilize Data Assimilation systems which are being
transitioned to NAVO
NRL Coupled Ocean Data Assimilation system
(NCODA)
Ocean QC
Innovations
3D MVOI
Sequential Incremental Update Cycle
Analysis-Forecast-Analysis
Increments
Ocean Model
First Guess
Forecast Fields Prediction Errors
Model forecast fields and prediction errors are
used in the QC of newly received ocean
observations
6
Approach (continued)NCODA Error Covariances

• Separable Error Covariances product of a
  variance and a correlation
• Background Error Variances
• vary by position, depth and analysis variable
• evolve with time, updated continuously using
  analyzed increments
• Background Error Correlations
• total correlation product of a horizontal and a
  vertical correlation
• horizontal correlations are multivariate in mass
  and velocity
• horizontal length scales are non-homogenous and
  anisotropic
• vertical correlations can be computed from
  background vertical density gradients, evolve
  with time
• length scales large (small) when stratification
  weak (strong)

7
Impact of glider data assimilation on SSTs at M1
Relaxation (Aug. 20-22)
BIAS(oC) RMS(oC) No assimilation
1.7 2.15 With assimilation
-0.5 1.12
AOSN II, 2003
M1
Upwelling
Relaxation
Upwelling
Relaxation
Assimilation of gliders improves SST predictions
during relaxation events
End of glider assimilation. The model returns
to model predictions without assimilation in
1-1.5 days
8
Impact of glider data assimilation
AOSN II
No assimilation
Upwelling, August 29
Assimilation of gliders
Aircraft SSTs
SST
SST
M1
AUGUST 15
M2
SSS
SSS
Assimilation of glider data provided a better
agreement of the model predicted and observed
spatial distributions of SSTs. Assimilation
resulted in fresher water masses closer to shore,
which improved surface salinity predictions at
moorings locations.
9
Impact of glider data assimilation on surface
salinity at mooring locations
AOSN II, 2003
M1
M2
10
Impact of glider data assimilation.
AOSN II, Aug. 2 Sept. 3, 2003
Temperature
Salinity
During AOSNII glider tracks covered area of
moorings locations. For this reason, the model
with glider data assimilated is able to reproduce
the observed changes in thermocline and halocline
depths during upwelling/relaxation events.
M2
Observations M2
No Assimilation
With assimilation glider data
Model predictions with only glider data
assimilated are comparable with model predictions
when other observations were also assimilated.
With assimilation glider, ship, aircraft data
11
Impact of glider data assimilation.
AOSN II, Aug. 2 Sept. 3, 2003
M2
Salinity
Temperature
With assimilation glider data as profiles
With assimilation gliders as dives
12
Impact of glider data assimilation.
MB2006
gliders
Upwelling
Relaxation
Upwelling
BIAS(oC) RMS(oC)
Correlation No assimilation
- 1.2 1.34
0.66 Glider data assimilation
-1.0 1.17
0.68 Glider, aircraft, ship assimilation
-0.25 0.6 0.85
13
Adjoint sensitivity maps at M1
Upwelling
Relaxation
gliders
14
Adjoint sensitivity maps at M2
Upwelling
Relaxation
gliders
15
Impact of glider data assimilation.
gliders
MB2006, Aug. 1 Aug. 17, 2006
Temperature
Salinity
During MB2006 glider tracks were to the north of
moorings, for this reason glider data
assimilation has minimal impact on model
predictions at mooring locations.
M1
Observations M1
No Assimilation
In opposite to ASON Model predictions with only
glider data assimilated are different from model
predictions when other observations were also
assimilated.
With assimilation glider data
With assimilation glider, ship, aircraft data
16
Impact of glider data assimilation.
gliders
MB2006, Aug. 1 Aug. 17, 2006
M1
BIAS(oC) RMS(oC) Correlation No
assimilation - 1.2
1.34 0.66 Gliders as profiles
-1.0 1.17
0.68 Gliders as dives -0.72
1.01 0.77
Salinity
Temperature
With assimilation gliders as profiles
With assimilation gliders as dives
17
Conclusions

• Optimization of sampling strategies with gliders
  is critical
• marked improvement in model skill with glider
  data assimilation when gliders tracks covered
  the area of moorings location (AOSNII experiment)
  
• minimal improvement during MB2006 experiment
  when glider tracks were to the north of mooring
  locations.
• The model forecast degrades to the level of the
  model predictions without assimilation in 1-1.5
  days.
• Subsurface T/S and velocity structure were
  greatly improved with assimilation especially
  during transition between upwelling and relaxed
  wind regimes (AOSNII period).

18
Anomalous Conditions during MB2006Currents at M1
upwelling
MB2006
AOSN II, 2003
Northward flow during upwelling
Southward flow during upwelling
relaxation
Southward flow during relaxation
Northward flow during relaxation
19
Anomalous Conditions during MB2006SSH anomaly
Observed strong positive SSH anamoly
Observations
Global NCOM
Kelvin Wave
AOSN II, 2003
MB2006
20
FUTURE PLANSNRL 6.2 Variational Data
Assimilation for Ocean Prediction (FY07-FY09)

• PIHans E. Ngodock
• Co-PIs J. Cummings, I. Shulman, S. Smith and N.
  Baker

OBJECTIVES Implement variational data
assimilation methods for the NAVY operational
ocean models (NCOM, HYCOM) Compare sequential
(NCODA) and variational data assimilation systems
21
Future Plans

• NRL 6.1 RO Bio-Optical Studies of Predictability
  and Assimilation for the Coastal Environment
  (BIOSPACE , FY08 FY12)
• PI J. Kindle
• Co-PIs B. Arnone, I. Shulman, W. Teague, P. Lee
• Objectives Improve our understanding of
• Coupled bio-optical and physical processes in the
  coastal zone
• The variability and predictability of the coastal
  oceans properties on time scales of 1-5 days,
  i.e., the time scales of accurate atmospheric
  forecasts.

22
NRL RO BIOSPACE Field Programs, FY 08
and FY10

• NRL observational assets
• Slocum Gliders (4)
• Continuous profiles of T, S, fluorescence,
  backscattering, attenuation, current speed
• Scanfish
• CTD, spectral backscattering, absorption.
  Complements high spatial and temporal sampling.
  Will participate in both survey and adaptive
  sampling modes
• SEPTR Real-time bottom-mounted Profiler
• Examine
• Vertical structure of physical, bio-optical
  properties
• Relationship to satellite vertically integrated
  measurements
• Space-time variability of physical-bio-optical
  properties
• Sampling strategies
• Expected collaborations MBARI, Rutgers (ONR
  MURI Rapid Environmental Assessment Using an
  Integrated Coastal Ocean Observation-Modeling
  System), MURI ASAP group.