This task will focus on improving a sea ice model to ensure better and more consistent thermodynamic estimates in the integrated observing system. Also the quality of the forcing fields will be assessed.
Ideally, the numerical models used in an integrated observing system should be as good and reliable as possible, and the simulated probability distribution for the state variables should be as close as possible to that observed in nature. These requirements are of concern to climate research, seasonal forecast and short range forecast models, although the differences in application will set different bounds to what errors are tolerable. If the models have large drifts towards an unrealistic climate, the assimilation methods have to do a brutal job correcting the model state, and the results will most likely not be physical consistent and reliable in the end.
One of the large uncertainties in the future projections of the Arctic climate is the fate of the sea ice. Several of the climate models included in the ACIA (Arctic Climate Impact Assessment, http://www.acia.uaf.edu/) predict an ice free Arctic Ocean during summer within the end of this century. However, most of these models include simplified ice models with underlying assumptions that may influence the sensitivity of the model in a climate change situation. These uncertainties are due to thermodynamic and dynamic assumptions in the models (Fichefet & Maqueda 1997; Zhang & Rothrock 2005). Although some of the new climate models (e.g. NCAR CCSM3, GFDL CM2) used in the next IPCC-report (Forth Assessment Report (AR4), expected in 2007) amend many of the past deficiencies, there are still assumptions made in these models that are poorly tested, with parameters and constant that at present are subject to tuning to get the overall climate as realistic as possible. Even in hindcast simulations of the sea ice state during last half of the 20th century the uncertainties due to differences in models are considerable (Rothrock et al., 2003).
The thermodynamic formulation of the sea ice model is important in an operational integrated observing and monitoring system. Ice thickness is not easily available in near real-time for the operational models, so thickness estimates must to a large degree rely on the sea ice model forecast. Most of the present day models divide ice into two categories: thin ice/open water on one hand and thick ice on the other, based on an approach after Hibler (1979). Although such a model can be tuned to give reasonable results, several studies point out the importance of including a full ice thickness distribution that resolves and captures the thin sea ice in a realistic way (Hibler 1980; Flato & Hibler 1995; Schramm et al. 1997; Zhang & Rothrock 2001; Bitz et al., 2001). This is because most of the sea ice volume in the Arctic freezes as thin fast growing ice, which subsequently is mechanically deformed into thick ice. Freezing of the thin, young ice and the implied changes in insulation also account for most of the heat exchange between the ice-ocean interface and the atmosphere during winter. Hence, the thin young ice is of great importance for the overall Arctic climate (Maykut 1982). Another feature of sea ice is the inclusion of sea salt into the ice in brine pockets. This has a large impact on the effective heat capacity of the ice and makes it important to model the total heat content of the ice in a realistic and heat conserving way (Bitz & Lipscomb 1999; Bitz et al., 2001; Røed & Debernard 2004). Neglecting the variations in the heat capacity due to changes in brine volume causes the model to melt ice too rapidly during summer, and freeze too rapidly during autumn and early winter. At present, most models set the ice salinity to a constant value despite the fact that the salinity varies a lot from salty young ice to almost fresh multiyear ice. Prognostic sea ice salinity is required to account for these variations in a conserving manner. The timing of the salt release during freezing, and a proper treatment of the early stages of ice growth in leads has a significant effect on the atmosphere-ice-ocean heat balance (Wettlaufer et al. 2000). The inclusion of prognostic sea ice salinity should also facilitate a salt conserving treatment of the melt water in ponds at the ice surface during the melting season. These ponds have a pronounced influence of the sea ice albedo during summer (Perovich et al., 2002), and are therefore of great importance to the ice-albedo feedback that are believed to be one of the key mechanisms behind the observed strong reduction in sea ice (Lindsay & Zhang, 2005). The global influence of changes in the ice-albedo in a global coupled climate model has been shown by Dethloff et al., 2006.
Methods and tools
The met.no ice model MI-IM is described in detail in Røed and Debernard (2004). The thermodynamics are modeled with one ice layer that includes fully prognostic internal energy accounting for the changes in salt brine volume, ice concentration and ice mass, while the heat capacity of the snow layer is neglected. This is similar to the thick ice / thin ice – open water formulation of Hibler (1979), but with a fully, heat-conservative treatment of the prognostic total heat content of the sea ice. The snow is insulating, reflective and has a latent heat contribution to the total heat budget of the model. The momentum equations in MI-IM are discretized on an Arakawa C grid with the elastic viscous plastic rheology of Hunke and Dukowicz (1997). The model is successfully coupled with met.no’s operational ocean model MIPOM (based on the sigma-coordinate Princeton Ocean Model), and to a local version of the Miami Isopycnic Coordinate Ocean Model (MICOM). MI-IM is also acting as the flux-coupler interface in the Oslo Regional Climate Model (ORCM; Debernard and Køltzow, 2005), where it is coupled with MICOM and the HirHam atmosphere model. The present operational ice-ocean model runs in a configuration with a horizontal resolution of 20-km for a domain that covers the Arctic Ocean: http://met.no/kyst_og_hav/northern_anim.html.
Work plan
In the present task we will enhance the treatment of thin ice and the overall description of the ice thickness in the sea ice model. Furthermore, we will implement a realistic description of the changes in the sea ice heat capacity that are due to changes in the ice salinity. The seasonal cycle of the surface albedo should be more realistically modeled with the inclusion of melt ponds. Operational ice-ocean forecasting, weather forecasting, climate modeling, and the integrated observing system will all greatly benefit from these improvements. At last, evaluation of the forcing fields from the weather forecasts models that are utilized in the iAOOS will improve the sea ice-ocean forecast, but may also be used to improve the atmosphere models at later stages.
Subtask 1: Model development
1. Include and test a full thickness and enthalpy (internal energy) distribution thermodynamic sea ice model. Evaluate the benefit of using a full thickness distribution in an ice-ocean monitoring system.
2. Include prognostic salinity in the ice model. What are the effects on the simulated sea ice and ocean state?
3. Include the melt pond and albedo parameterizations developed in Task 3.7 and evaluate the effects on the overall simulated ice state.
4. Assess the quality of the atmospheric forcing (deduced from operational weather forecast models) used in the ice-ocean model by comparing model fields with satellite based OSISAF radiation products (Task 3.8). Also in-situ measurements from WP 1 and other IPY resources can be utilized directly. Are there serious deficiencies in the atmospheric forcing-fields that may ruin the analysed ice-ocean state? And if so, how should we correct the forcing?
