1st FAMOS-2012 Workshop
Sea ice retreat and forecasting
“In response to the unprecedented arctic sea ice retreat in summer 2012, and to the one of key recommendations of the National Academies report “Seasonal-to-Decadal Predictions of Arctic Sea Ice: Challenges and Strategies”, 2012” (“The most important step to advance sea ice prediction over seasonal to decadal time scales is to establish sustained and coordinated collaboration among the sea ice data user, modeling, and observation communities. A commitment to establishing better communication could help reveal further gaps in understanding of the Arctic environment, and inform effective research activities to generate more accurate, timely, and useful sea ice forecasts”), the agenda for the first morning session of the 2012 main FAMOS workshop was re-organized to focus on this topic. Don Perovich gave an invited talk summarizing the key events of the summer. This led to lively discussions in the break-out groups. A result was two proposed new collaborative projects. The first will examine the possibility of short-term forecasting of the rapid sea ice loss that occurred in early August 2012 in response to the large low pressure system that developed at this time, using various atmospheric forcings and initial conditions (leader: R. Allard). The second project will consider the possibility of a seasonal forecast of the summer 2012 sea ice season, focusing on the role of initial conditions, i.e. on observational requirements for model forecasting (leader: T. Martin).”
2nd FAMOS-2013 Workshop
Sea Ice Session
Don Perovich summarized the sea ice extent summer minimum of 2012 and gave an outlook for 2013; winners of the 2012 sea ice minimum forecast challenge were presented with their awards, the winners were … (minimum extent), … (date of minimum extent), … (minimum extent map); announced the 2013 sea ice minimum challenge.
Rick Allard presented updates to the Navy’s sea ice forecast model system and an assessment of the Navy’s outlook for the 2013 sea ice minimum.
Danny Feltham presented new physics available with the new version 5 of the CICE sea ice model, such as variable drag coefficients, melt ponds, and anisotropic rheology; changing ice type (more first year ice) affects surface topography, affect melt pond size, and finally surface albedo, accelerating melt.
Sinead Farrell reported on sea ice conditions in the Canada Basin as observed by NASA IceBridge flights in late winter/spring 2013: mean thickness of first year ice 1.6 m (modal thickness 1.2 m), multi year ice mean thickness 3.0 m (mode at ~2 m).
Andrew Roberts stressed the importance of running sea ice models at high spatial resolution (at least 10 km) forced by atmospheric fields (or coupled to an atmosphere model) at short time steps (hourly or less) to simulate the scale dependency of sea ice deformation as it has been observed; in this case the classic viscous-plastic rheology also shows scale-dependency of deformation.
Torge Martin showed that the momentum influx into the Arctic Ocean increased over the last 30 years in fall, winter and spring because of ice weakening but decreased in summer due to reduced ice area.
Bruno Tremblay presented an Arctic-wide model for land-fast ice and showed differences between the East-Siberian and Laptev sea land-fast ice extent, which are related to wind direction, topography, and shear stresses.
Frederic Dupont reported on improvements achieved by coupling the CICE sea ice model instead of LIM2 to the ORCA ocean model.
Dmitry Dukhovskoy gave a brief introduction to a new topological approach improving the evaluation of sea ice models and their quality by not only accounting for absolute area but also shape of the ice cover.
- For the first time the AOMIP/FAMOS program included 2-3 hours devoted to posters only; the sea ice session sported 21 of 53 posters.
- New general theme: How does the changing sea ice cover affect the Arctic Ocean?
– sea ice thins, weakens and thus drifts faster
– changes in surface roughness and hence changes in drag coefficients
– increased doming of the Beaufort Gyre
– Atlantic water circulation
- Increase complexity of sea ice models, new physics are being implemented, e.g. variable drag coefficients, melt ponds, …
- Predictability of Arctic sea ice coverage
- New IceBridge observations (LASER for ICEsat2) allow derivation of ridge and lead statistics
- More studies combine model and observational data
- Wide range of model types are used: from simplified basin set ups to fully coupled CMIP5 models
- Role of sea ice and changing ice characteristics on bioproductivity in the Arctic
Discussion (Round 1, entire audience, Wednesday)
Question: At what point can we declare that models are perfect enough? When do we stop developing? Answer: Models aren’t perfect, and there still remain:
- Physical deficiencies
- Resolution and numerical limitations
- Strong parameter sensitivities
- Poorly understood dependence on forcing data and coupling configurations
There remain difficulties in evaluating the performance of sea ice models:
- Simple metrics, such as sea ice extent, seldom reveal much about the capacity of models to correctly simulate physics
- More sophisticated metrics of existing observations, such as Dmitry’s topological approach for extent and concentration, may reveal more about model skill than simple statistical approaches
- Physically-based metrics, such as the timing and nature of melt, as well as dynamical scaling, were proposed to help address the problem of understanding model physics
- Assessment of observable quantities is as important as inter-comparison of models, and hence the role of the observational community is important in helping to design metrics.
- Conversely, the modeling community needs to help the observational community in identifying gaps in the observational record.
Broader issues surrounding the performance of current generation sea ice models should be taken into account:
- The sensitivity of sea ice cover to atmospheric radiation is significant. While atmospheric modeling may be beyond the scope of FAMOS, feedbacks to atmospheric radiation resulting from sea ice model physics are well within the scope of sea ice projects.
- To that end, the influence of biogeochemistry on the surface conditions of sea ice, including albedo, remains relatively unexplored in models, and is an important avenue for investigation.
- More broadly, the importance of ocean-ice-atmosphere coupling should not be underestimated, and so the analysis of a hierarchy of models, from assimilated, surface-forced ice-ocean simulations, through to fully coupled global models, is more likely to shed light on physical deficiencies in models than by only comparing different finely-tuned stand-alone ice-ocean model configurations.
An important starting point is to seek out common threads among the research groups involved in FAMOS, and begin to explore ways we can work together to better understand deficiencies and improve sea ice models.
Discussion (Rounds 2 & 3, sea ice breakout group only, Thursday & Friday)
The sea ice discussion was dominated by modeling topics (which mostly was caused by the lack of participants focusing on observations).
Some general results of the discussion:
- Increased exchange of experience is highly desired. To continue this after the meeting, a separate FAMOS sea ice e-mail list or—even better—a discussion/chat forum is needed. This could be created in using either Google Hangouts, EarthCube, LinkedIn.
- In particular for young scientists a list of available observations for Arctic research and model evaluation would be great, also to share experience with observational data. Such information could be posted on the FAMOS web page.
- Develop a list of common model output (similar to the CMIP protocols), which would increase the possibility of across-model group studies within FAMOS; [A first draft of the proposed standard output can be found HERE].
Please read and suggest corrections (use file FAMOS_Sea_Ice_Model_Output_List. Xlsx for your suggestions/corrections)
- Separate documentation of the models used in the FAMOS community, including information on model physics, settings, initial conditions and forcing (be as detailed as possible). [A first draft of the requested model descriptions, divided into model class, is provided HERE]
Please complete information about your model using template file FAMOS_Model_Description_Template.xlsx) and send it to us.
The following group projects emerged
- Model intercomparison focusing on coupling and forcing (sub-topics: sea ice physics packages, albedo, radiation, snow, boundary layer formulation) [A. Roberts]
- Variability and modeling of Sea ice surface drag [M. Tsamados]
– air-ice stress (boundary layer stability, ice topography)
– ice-ocean stress (near-surface ocean stratification)
[possible collaboration (observations): J. Toole, S. Cole, C. Lee]
- Investigate the partitioning of sea ice surface, bottom, and lateral melt and its possible long-term change; How does this shape the local ice thickness trajectory from winter to summer? Does this affect the summer sea ice minimum (area and volume) in a simulation? [T. Martin]
- Study the relationship of strain, land-fast ice, and sea ice drift (and drift trend) [B. Tremblay]
Observational needs for Arctic modeling:
- Wind speed is highly undersampled (only few buoys carry wind speed sensors); wind speed information necessary for ice kinematic studies
- Snow is a big issue; has great effect on thermodynamics; only sparse observations; precipitation in reanalysis products has great caveatsis off; maybe use snow thickness climatology from satellite observations instead of reanalysis precipitation?
- Can procedures used to link ground, air-borne, and satellite observations be applied to model-observation comparisons?
- Create a wish list of desired observations (quantities, locations, season)