Ocean and sea-ice data assimilation and forecasting
in the TOPAZ system
L. Bertino, K.A. Lisæter, I. Kegouche, S. Sandven
NERSC, Bergen, Norway
Arctic ROOS meeting, 18th Dec. 2007
Motivation
Objective: Provide short-term (10 days) forecasts of physical
and biogeochemical ocean parameters to the public at large and intermediate users.
Strategy Focus on advanced data assimilation techniques Gradual increase of resolution (as affordable…) Nesting on regions of higher interest
Method
The ice-ocean system has two sources of information A nonlinear ice-ocean model A regular flow of observations
Uncertainties arise primarily from The initial state Surface boundary conditions Measurements errors
Monte Carlo methods can handle non-linear dynamics. Provide the best estimate Provide the residual uncertainty
Each source of uncertainty must be simulated realistically.
Sequential data assimilationRecursive Monte Carlo method
Forecast Analysis
Observations
1. Initial uncertainty
2. Model uncertainty
3. Measurement uncertainty
12
3
Member1
Member2
……
Member99
Member100
The TOPAZ model system TOPAZ: Atlantic and Arctic
HYCOM EVP ice model coupled 11- 16 km resolution 22 hybrid layers
EnKF 100 members Sea Level Anomalies (CLS) Sea Surface Temperatures Sea Ice Concentrations (SSM/I) Sea ice drift (CERSAT)
Runs weekly since Jan 2003 ECMWF atmos. forcing
Model upgrade
Doubling the horizontal resolution
• TOPAZ2: 18 to 36 km
• TOPAZ3: 11 to 16 km
TOPAZ2
TOPAZ3
System Validation
Consistency?Accuracy?
Performance?
Consistency: Against Climatology
TOPAZ2 TOPAZ3
Temperature anomalies at 30 m depths
Accuracy: against ice concentrations
TOPAZ2 TOPAZ3
Model minus obs.
AccuracyAgainst in-situ profiles from NPEO
Aerial CTD casts
Temperature Salinity
Assimilation on 4th and 11th April
Up to +10 days forecast
Forecast skills: Barents Sea - ice concentrations
Average Winter 2007 Average Summer 2007
Ice drift validation
In-situ Ice drifting buoys (Statoil/CMR) Manned expeditions
Remote sensing ASAR (NERSC) WP2 QuickSCAT (Ifremer)
Modelling TOPAZ V1, class 1
A good agreement
[ J. Wåhlin]
Historical minimum Arctic sea-ice area, summer
2007Observed sea-ice from SSM/I, NORSEX algorithm
Forecasting the ice minimum in TOPAZ
• Overlay of successive forecasts
• TOPAZ catches the freeze-up
Products
StandardsDelivery
Timeliness
What products?
MERSEA products Class 1:
3D daily fields ocean and sea-ice
Anomalies to climatolgy Class 2:
Predefined sections Predefined moorings
Class 3: Volume fluxes through sections Salt and heat transports
Class 4: Differences with observations, Forecast skills
Other products (targeted) Ensemble uncertainties, Predicted drift Icebergs
Class 2 metrics
Sections stored daily Moorings stored daily
Uncertainty estimatesexample sea-ice thickness
Ensemble average
13th March 2007
Ensemble standard dev.
13th March 2007
Forecasting the drift of Tara
TOPAZ successive forecasts in red Actual positions of Tara from DAMOCLES in black Updated on Google Earth [ K. A. Lisæter]
8m draft
Iceberg simulations
An iceberg is sensitive to Winds Waves Currents Ice drift Ice thickness Iceberg shape Tides Melting …
13m draft 18m draft
[ I, Keghouche, NERSC ]
Availability
Forecast updated every Thursday 10 days forecast horizon
Available freely via Webpage http://topaz.nersc.no (static pictures) OPeNDAP http://topaz.nersc.no/thredds (data) No password required But feedback is welcome
Available to date TOPAZ2: October 2005 to October 2007 TOPAZ3: July 2007 to present
Plans
Ongoing projects (MERSEA, BOSS4GMES) Assimilation of additional data (Argo) Inclusion of ecosystem model
NORWECOM from IMR, Bergen. RT exploitation of TOPAZ at met.no
Developments of TOPAZ at NERSC Exploitation at met.no (ongoing)
Planned project MyOcean (2008-2011) 30-years reanalysis