An International Collaborative Effort Towards Automated Sea Ice Chart Production

1. An International Collaborative Effort Towards Automated Sea Ice Chart Production Tom Carrieres Ice Modelling Manager Applied Science Division Canadian Ice Service

2. Tom Carrieres Doug Lamb Lars-Anders Breivik Rashpal Gill Dean Flett Mark Buehner Bruce Ramsay Mike Van Woert Mike Manore Contributors

3. Background State of the Art Review Data Models Data Assimilation Research and Development Steps Outline

4. Current Ice Chart Production - ice analysts with extensive experience - increasing problem of information overload - visible, infrared and passive microwave data - SAR, scatterometers, aerial reconnaissance - ice drift from beacons, models, algorithms - derived geophysical fields very limited - operational time constraints subjective labour intensive analyses Background

5. NWP/data assimilation benefits - objective, optimal integration of varied data - allows forecasters more time to focus on critical areas IICWG 3 - are similar benefits feasible for sea ice? - provide first guess for analysts to build on - expert judgement focuses on critical areas - objective weighting of data leads to more consistent products - possible avenue for international collaboration - science workshop and White Paper on data assimilation Background

6. - observations that resolve phenomena - models that adequately predict future state - consistent, inclusive, objective analysis Requirements

7. State of the Art Review - DataSatellite Data

8. Ice motion - beacon/buoy derived motion is sparse but almost continuous - sequential image derived motion is sensor dependent with variable resolution and interval - algorithms perform poorly in marginal or featureless ice areas State of the Art Review - Data

9. Data Issues - should revisit algorithm development within the context of objective, automated use - characterization of data errors almost non-existent - variety of data sources is optimal - errors can offset each other - take advantage of higher resolution data - data management becomes an issue - mix of derived fields and direct satellite measurements may provide the most useful combination of information State of the Art Review - Data

10. Components of an ice model - drift - thickness distribution and redistribution - strength and rheology - thermodynamics - ocean coupling Typical model - resolution: 5-20 km - forecast period: hours to days - areal extent: hundreds to thousands of km State of the Art Review - Models

11. Issues - less direct coordination/cooperation has occurred in model development - most models designed for climate change and or engineering design - operational ice forecasting is more of an initial value problem and scale is different - many complex processes have been modelled but very few ice characteristics are observed - focus on assimilative models? - simple models? - constrained by data - for use in 4Dvar? - model skill is dependent on accuracy of forcing fields State of the Art Review - Models

12. - combines observed and model data - statistically optimal manner - constrained by model physics - accounts for relative errors - constrains models to reality overcoming - uncertainties in forcing - parameterization limitations - finite spatial and temporal scale - improves on data by - filling observation gaps - combining disparate data into coherent products - adding temporal consistency State of the Art Review Data Assimilation

13. Analysis segment - simplest technique is to accept observations from individual sources as truth and interpolate or average the data onto a grid - to combine observations from different sources, statistical analysis or optimal interpolation methods are employed, making use of the error variances of each of the data sources - model forecasts may be used as complete background fields with their own applicable error statistics State of the Art Review Data Assimilation

14. Model initialization segment - simplest way to initialize a model with analysis fields is by insertion or replacing model fields but this can cause numerical instabilities - nudging techniques reduce these problems by inserting data in an asymptotic process over a number of model timesteps - NWP has moved far beyond this with variational techniques or Ensemble Kalman filters State of the Art Review Data Assimilation

15. NWP situation is different - analysis is an interpolation and filtering process with analysis weights to determine the relative contribution from the various observations and the model first guess - few observations compared to the degrees of freedom in the models, and variations in scales between what is resolved by the models and what is actually observed - weights are defined in terms of the expected errors variances and error correlations of the model first guess and the observations - horizontal error correlations are assumed to be smooth, isotropic and homogenous - multivariate error correlations are defined by assuming balanced constraints on the synoptic scale (e.g. geostrophy). State of the Art Review Data Assimilation

16. Issues - lack of in-situ/direct observations - incomplete and inconsistent data sets - difficulties of air/sea/ice interface - interactions - time and space scales - ice is a discontinuous, deformable medium so assumption of isotropy and homogeneity in the error variance fields is less valid - multivariate treatment is important e.g. sea temperature consistent with ice extent - observation operator that relates model variables to quantities observed by satellites is key State of the Art Review Data Assimilation

17. Research and Development StepsRequirements Definition - Canada

18. - co-development of better passive microwave ice algorithms using digital ice charts as validation and tuning in conjunction with use of more detailed unique datasets - development of forward and inverse algorithms to mesh direct observations (eg. radiance) with model predictions - development of simpler information extraction algorithms for SAR - for all of the above, it is essential to characterize the errors in a manner suitable for data assimilation Research and Development StepsData

19. Models - development of adequate or modification of existing models for a data assimilation system, with investigation of appropriate level of model complexity - characterization of model errors Data Assimilation - determination of the best techniques for automated analysis of ice data - determination of best techniques for incorporating automated and manual analysis fields into models - determination of most useful data to be used within a sea ice data assimilation system Research and Development Steps

20. Other Issues - can/should we develop a common framework (with each other and with NWP community) - expertise in all areas does not exist at individual Ice Centers and may not exist within the entire operational ice community - resources: this work will reduce the workload of analysts in the future but where do the resources come from now in order to develop such systems? - what is the best way to engage the entire ice community in a collaborative effort? - how do we engage other expertise particularly in the NWP, oceanography and climate change communities? - verification has not been specifically discussed but is important and methodologies should be shared Research and Development Steps

21. Strategies - information exchange - joint research (project based) - operational model and/or data exchange - ice integrated in national NWP programs - common system approach (regional, hemispheric) - others Research and Development Steps

22. Next Steps - IICWG science meeting on modelling/data assimilation - IICWG approved/funded workshops on data assimilation etc. - identification of national leads Longer Term - Coordinated "working groups"? - Data Group - develop data products and assess data errors - Model Group - assemble model components - Assimilation Group - develop and test assimilation approaches - Evaluation Group - obtain comparison data for models, data, and assimilated products; implement and test operationally - Coordinate to share data, models, assimilation algorithms, and results - Consistent grids for models, data, forcing - Interchangeable modular components - Follow example of model intercomparison projects Research and Development Steps