Sea Ice Modeling: A Mini-Workshop
4. Rapporteur Summaries
Session 2. Sea Ice and Climate Processes
(Rapporteur: Mark Serreze)
The four talks and subsequent discussion in this session revealed considerable hurdles to be overcome regarding issues of scale and the level of detail that must be modeled. While each speaker emphasized the need to address local to regional processes, it was frequently countered that this cannot yet be done in global models. The problem can be summarized in the following question: what is the minimum level of detail that global models must treat to simulate sea ice in a reasonably realistic manner?
Judith Curry started the session with a lively overview of her recent work. She pointed out that a "first order" requirement is that models treat sea ice dynamics using a realistic rheology-thermodynamics alone are insufficient. She further stressed that correct simulation near the coasts is the most important, for, to paraphrase, "that is where the action is". Curry further argued that models must address sub-grid-scale dynamics, such as ridging and ridge breakoff. Regarding thermodynamics, models must treat processes such as snow cover and its redistribution, sea ice age and internal temperatures, and surface albedo in a way that responds to ice characteristics, leads, ridges and ice biota, the latter through its influence on radiation fluxes. While acknowledging that most of these processes are not being treated at present (and certainly not in global models), Curry showed results from her modeling studies employing SHEBA data that incorporate such treatments. She remarked that with today's computing power, it is increasingly difficult to argue that computing requirements preclude detailed treatments. This perspective generated considerable discussion.
Curry provided a four-point strategy to tackle the issue of which processes are the most important:
- Evaluate parameterizations against observations.
- Evaluate ensembles of parameterizations in a single-column ice thickness model.
- Use the theory of feedback control systems to evaluate the impact of parameterizations on feedback processes.
- Re-evaluate as needed in 3-D and coupled models.
Curry stressed within this framework that one cannot simply compare time series of different model variables - one must compare processes.
David Rind asked how one should treat snow aging over sea ice. Curry said that this needed some thought. It was also queried as to how representative the SHEBA data are of Arctic conditions. Curry responded that the data should be quite representative, although there is a need for a complementary data set over Antarctica. She also stressed the need for better data on melt ponds.
Mark Serreze and James Maslanik followed with a summary of their efforts to understand processes leading to observed sea ice anomalies and the extent to which these anomalies can be modeled. They started with an overview of the observed sea ice record, noting that the recent downward trend in Arctic sea ice extent does not reflect a general northward retreat of the ice but rather the integrated effects of large regional anomalies during late summer and early autumn. Since the observed record points to the importance of regional anomalies, they argue that global models must address regional processes.
Serreze and Maslanik focused their modeling studies on the extreme anomaly observed in the Laptev and East Siberian seas in the late summer of 1990. Observational studies indicate that the anomaly resulted from persistent offshore winds, which promoted early and extensive melt and advected ice away from the coast. In turn, proper simulation of the anomaly using a stand-alone Hibler type model requires treatment of both ice dynamics and thermodynamics. Model runs using thermodynamics or dynamics alone resulted in poor simulations. They also examined results from a regional coupled ice-atmosphere model (ArcSym). The simulated sea level pressure field in the ArcSym model for 1990 was close to observations, but differed in details of the location of pressure features and pressure gradients. However, even these small errors were sufficient to yield a poor simulation of the ice anomaly.
They then summarized results from their observational study of the record sea ice anomaly in the Beaufort Sea in 1998. It was stressed that while 1990 and 1998 were radically different in terms of the location of ice anomalies and associated anomalies in atmospheric circulation, the relevant physical processes were the same - in both cases, strong offshore winds promoted melt and advected sea ice away from the coast. Although reinforcing their view of the importance of regional atmospheric processes, they also noted that Miles McPhee predicted an ice anomaly based on the fact that extensive open water developed in this area in 1997. This resulted in a large coverage of thin, first year ice at the beginning of the 1998 melt season, which predisposed the ice to extensive decay. This argues that global models may need to have a long "memory" of ice conditions. A possible link between the summer 1998 anomaly and the strong 1997/1998 El-Nino event was also suggested.
At the end of their presentation, it was pointed out that while Serreze and Maslanik are likely correct that even small shifts in the position/strength of atmospheric pressure features can strongly impact the sea ice, global models presently do not have the resolution to adequately address regional circulation features with high detail.
Tom Grenfell presented Don Perovich's talk and his own, both talks addressing sea ice albedo. Perovich's paper summarized the temporal evolution and spatial variability of albedo based on SHEBA data. Results were based on measurements along a transect approximately 200 m long, along which albedo was measured every few days. Both broadband and spectral albedo were measured. A rapid drop in albedo was observed in early June due to melt. After this date, large variations in albedo were found along the transect, related to differences in ice type, ice characteristics and melt ponds. Grenfell used this as ammunition to justify the need for sub-grid treatment of albedo. He also remarked on the significant "flickering" of albedo over time, related to synoptic weather events. Grenfell also summarized results from aerial surveys - less variability was noted in albedo because of the areal integration.
Grenfell's own paper turned to the problem of albedo parameterization in global models. Many problems were noted. These include the wavelength dependence of albedo, the co-dependence of albedo and transmissivity, the need to know the inherent optical properties of different snow and ice types, and requirements for data on the coverage of these different surface types. Grenfell reviewed several existing albedo parameterizations. He noted that Ebert and Curry have developed parameterizations with significant improvements over existing efforts. The problem of scale was again raised. Curry eloquently argued the need to incorporate many different ice types, each with its own wavelength-dependent albedo. However, it was countered that global models cannot treat albedo at this level of detail. For example, the Community Climate Model 3 (CCM3) uses two spectral bands and four ice types.