Sea Ice Modelling and Data Assimilation at CIS Tom Carrieres
Sea Ice Models - Model developments began around First coupled ice-ocean models available in Currently models cover all CIS operational areas and run daily in support of ice operations at around 99% reliability - Now considered an integral part of CIS operations and some products are now totally model generated
Coupled Sea Ice-Ocean Models - CIOM east coast (BIO) - CIOM nested grid (BIO) - Gulf of St Lawrence (IML) - Arctic Ocean (IOS 2004) - Archipelago (IOS 2005) - CIOM Great Lakes (CIS 2005)
Coupled Sea Ice-Ocean Models Model Products Existing Products ice drift ice concentration ocean currents New Products - forecast ice charts - ice drift charts
Coupled Sea Ice-Ocean Models Model Verification/Case Studies Routine verification concentration vs daily ice charts ice drift vs Tracker-Radarsat and beacons Conclusions presentation/utility essential operations involvement essential output sometimes extremely helpful and sometimes misleading
Future IICWG-3 Direction Move Ice Services from labour intensive subjective analyses to NWP-like mode Automated analyses based on as many data sources as possible Allow analysts to focus on operationally critical areas Emphasis on national and international collaboration Limited resources Build on existing expertise Shared benefits Additional CIS Direction Move models to CMC Promote model utility and development with DFO Raise awareness of CIS needs with ACSD/CMC
Data Assimilation Operational Analysis ice concentration/thickness – averaging from ice charts ocean and other ice fields from model forecast Initialization Ice concentration/thickness – nudging or insertion All other fields – insertion Development Analysis ice concentration/thickness – statistical interpolation using ice charts or SSM/I SST - statistical interpolation using AVHRR/AMSR
Statistical Interpolation Subscript A=analysis field B=background field (model) O=observed field observed data are located at irregularly spaced points r k analysis and model data are situated on model grid r i weights W ik balance the influence of the observation and background values to minimize the expected error in the analysis field
Determination of Weights the background and observation error correlation for the locations r k and r i the ratio of observation error variance to background error variance If we assume errors are homogeneous:
Error Correlations h k h i bottom depths (in m) Lhorizontal decorrelation length (10 km) L h depth decorrelation length (50 m) the Kronecker delta Background error correlation: Observation error correlation:
Optimizing S.I. parameter Input parameters to statistical interpolation 1. ε 2 – relative weighting parameter 2. L – horizontal decorrelation length 3. L h – depth decorrelation length (50m) We want to isolate the values of these parameters that will generate the best analysis field with the lowest error
Optimizing Parameters for Ice Chart Data Background – initial “best guess” field 24-hour forecast from model run of previous day Observation Daily ice charts Verification Radarsat image analysis
Source Data Ice charts valid left - Radarsat image analysis chart valid 09:56 UTC centre - daily ice chart valid 18:00 UTC right - Radarsat image analysis chart valid 21:18 UTC
Starting Point for o 2 Determined value:
Results -- Optimal ε 2 Optimum value 0.95
Results -- Optimal L Optimum Value ~30 km
Optimizing Parameters for SSMI-NASA Team2 Total Ice Background: model 24 hour forecast based on previous analysis Observations: total ice concentration analysis only, ice types/thickness partitioning from model NSIDC daily gridded data assumed valid for 18 UTC Model ocean T&S adjusted in areas of analysis increments SST approaches T f for increased ice SST approaches T f +0.5 for decreased ice S adjusted in balance with above Increments drop off linearly with depth Validation: Radarsat image analysis charts and daily ice charts
Preliminary Results
Ocean (over-) Adjustment
SSM/I vs Ice Charts
Sea Ice DA Planning Workshop Montreal March 2004 Goals: to develop a sea ice data assimilation plan for Canada to forge stronger relationships between remote sensing, modelling and data assimilation experts Participants Ice and ocean modellers, remote sensing experts, NWP modellers and data assimilation experts, managers and 5 US sea ice experts
Sea Ice DA Planning Workshop Key results Data: Start with easiest (ice charts) and build PM observation operator Models Level of complexity required is influenced by end products and data to be assimilated Will use existing dynamic-thermodynamic model and build additional capabilities as required DA technique 3Dvar will be used: allows assimilation of direct observations Currently in use for Canadian NWP Incremental change from current CIS OI developments Allows for development path to 4Dvar
MSC Coordination Meeting Goals To develop a coordinated approach to sea ice research but more specifically sea ice modelling and data assimilation Participants MSC research (models, DA, climate) CMC (NWP) CIS Applied Research
MSC Coordination Meeting Key Results Coupled atmosphere-ice-ocean modelling CIS is viewed as an important client and participant CIS participation in development team CIS membership on CPOP CIS membership on ACSD management team Sea ice data assimilation Led by MSC data assimilation group Focus on sea ice for CIS operations CIS participation
Future work Migrate to 3Dvar for ice charts, PM and SST Development of PM observation operator Migrate ice model operations to CMC Development of single ice model Verifications and simulations of coupled A-I-O system