Ten-Year Averages 1999-2008
Based on:
http://www.nws.noaa.gov/om/hazstats/images/69-years.pdf
Lightning 42
Tornadoes 63
Hurricanes 117 (Katrina in 2005 greatly inflates this figure compared to normal values)
Floods 65
Wind 43
Cold 21
Winter 33
Heat 162
If you count 'severe weather' deaths as all directly caused by convective and tropical phenomena, the number would be:
Convective + Tropical = 222
Floods and wind may or may not be convective in origin and therefore not considered severe weather (I believe they should be however), here is the count with wind+floods added:
Convective + Tropical + Wind + Floods = 330
Heat has the highest death rate- but unlike the other hazards, it tends to threaten a very narrow demographic of the population - the elderly or in poor health who do not have access to air-conditioned shelter.
Winter total:
Winter + cold = 54
Fatalities due to road icing have traditionally not been counted as winter weather deaths. I believe they should, as the icing - a direct weather phenomenon - is the primary factor in many of the crashes. The ice gets under vehicle tires, separating the car from the roadway and causing it to leave the road without abnormal driver behavior - just as if a tornado had blown it off the road.
Last winter, there were at least 477 road ice fatalities. This figure is independent of the other winter death categories. Combined with the 10-year winter death average, that would give a figure of 531 for all winter phenomena, compared to 330 severe weather deaths. Excluding heat deaths, road icing alone beats out all other weather phenomena combined - 477 versus 384 (winter + severe weather). This season's road ice death toll is already outpacing last year.
There is debate on what to count and what not to count as weather-related. I believe if a death is a direct result of a weather phenomena, it should be counted. Many of the arguments against counting road deaths, if applied to other severe weather hazards, would mean we would not count most flood, severe storm and hurricane deaths. I can go deeper into that if the discussion warrants.