In the upper Midwest, nearly 1000 Canvasbacks and Redheads have been found dead along Lake St. Clair (often referred to as "the sixth Great Lake"). This is a conservative number, considering how many may have died out in the lake that were not detected by shoreline observers. Toxicology reports have been coming back negative. The die-off has been attributed to malnutrition due to a larger-than-usual number of ducks wintering in the lake because it didn't freeze over as early as it usually does. Diving ducks that typically feed on invertebrates or mollusks (Bufflehead, scaup) don't seem to be impacted, while plant-eating Canvasbacks and to a lesser extend Redheads are most effected. My husband took this photo of a dead Canvasback on the Detroit River, where he's seen a few every weekend the past month. This phenomena has been occurring for at least several years, although perhaps not at this scale.
This situation has provoked kind of an odd reaction on the local bird forums, with people saying that they really hope it is malnutrion and not botulism, which also periodically claims a lot of waterfowl in the Great Lakes. I have to say I disagree, since botulism only occurs under certain circumstances and is often self-limiting. On the other hand, if malnutrition is is really due to too many ducks overwintering on Lake St. Clair (and other Great Lakes), we are likely in for a chronic and increasing problem.
A recent paper  looked at trends in ice duration in 65 waterbodies in the Great Lakes and found average rates of change in ice freeze and breakup dates were 5.8 and 3.3 times faster, respectively, than historical rates from 1846 to 1995 for the Northern Hemisphere. The following chart  looks specifically at ice trends from the three basins of Lake Erie.
Since around 1990, the central and eastern basins of the lake have not been freezing over until later than they have since the late 1950s. The western basin is much shallower and tends to freeze earlier, and is most comparable to Lake St. Clair.
If migrating waterfowl increasingly encounter open water during fall migration, this may cause higher numbers to overwinter in these areas. I have heard anecdotal reports to this effect, but was unable to put my hands on much hard data. I produced the chart below using numbers from the annual early-fall survey of Canvasbacks done by the Michigan Dept. of Natural Resources . These numbers only represent the Michigan side of the lake, but indicate increasing numbers of Canvasbacks on the lake in early November.
The trend for wintering Canvasbacks on a portion of the Detroit River near Lake Erie, from the Rockwood (MI-ON) Christmas Bird Count (CBC), which is held annually in mid-December, is also positive and significant. These data are in conflict with the long-term (1966-2003) population trends for Canvasbacks determined by the North American Breeding Bird Survey and continental CBCs, both of which show declines (although significant only for the CBCs) .
The idea being floated is that more Canvasbacks (and other waterfowl) are staying in the Great Lakes, and when the lakes do freeze up, they are unable to obtain food and starve to death. This seems to me to be only part of the story. The following chart looks at the number of days Lake Erie stays frozen.
Since the 1980s, the trend for the number of days the entire lake has ice cover has been steadily decreasing. It seems possible to me that the increased number of waterfowl might in fact have ice-free foraging areas, but too much competition for too little food. Canvasbacks rely on aquatic vegetation more than any other food source in winter — especially wild celery (Vallisneria americana) .
Wild celery did increase in the Detroit River (and Lake St. Clair) in the 1990s. But something else has been increasing in the region since the wild celery has been recovering — Mute Swans. The chart below shows CBC trends on the Detroit River CBC, held annually on January 1. And a little further downriver, the situation is the same on the previously mentioned Rockwood CBC.
Mute Swans also feed on submerged vegetation, up to 8 pounds a day. This added competition for food just compounds the problem, and in fact, some Mute Swans have also been found dead. Mute Swans are resident (although in this area they retreat to inland marshes to nest), and consume submerged vegetation year round, eventually damaging the reproductive capacity of the plants.
This could be a perfect storm of climate change contributing to less ice cover, attracting more overwintering Canvasbacks, combining with a population explosion of non-native Mute Swans to create a substantial alteration in food availability. If that's the case, things will likely only get worse, not better.
 Jensen, O. P., B. J. Benson, J. J. Magnuson, V. M. Card, M. N. Futter, P.A. Soranno, and K. M. Stewart. 2007. Spatial analysis of ice phenology trends across the Laurentian Great Lakes region during a recent warming period. Limnology and Oceanography 52: 2013-2026.
 Assel, R. A. 2004. Lake Erie ice cover climatology – basin averaged ice cover: winters 1898-2002. NOAA Technical Memorandum GLERL-128. NOAA Great Lakes Environmental Research Laboratory, Ann Arbor, MI, 15 pp. ftp://ftp.glerl.noaa.gov/publications/tech_reports/glerl-128/
 Soulliere, G.J., T.E. Maples, and E.N. Kafcas. 2000. Twenty-five Years of Canvasback Inventory in Michigan, 1974-1999 . Michigan Department of Natural Resources, Wildlife Division Report No. 3332.
 Butcher G. S.. K. Niven, and J. R. Sauer. 2005. Using Christmas Bird Count data to assess population dynamics and trends of waterbirds. American Birds 59: 23-25.
 Varro, F. 2003. Restoring Vallisneria americana in the Upper Mississippi River for canvasback habitat. Restoration and Reclamation Review 8. <http://horticulture.cfans.umn.edu/vd/h5015/03papers/varro.pdf>