HOW MIGHT CLIMATE CHANGE EFFECT LAKE-EFFECT SNOW STORMS

presented by: Kenneth Kunkel Illinois State Water Survey Champaign, Illinois

(SLIDE 2] The Great Lakes not only are affected by climatic forcing on a large scale, they have effects on localized climate conditions, particularly downwind of the lakes. Locations in these areas experience enhanced snowfall in winter, warmer winters, and cooler summers compared to locations distant from the shoreline. [SLIDE 3] The climate conditions at two nearby locations, Green Bay, Wisconsin and Frankfort, Michigan, illustrate this. With prevailing northwesterly flow during the wintertime, Frankfort is located downwind of Lake Michigan while Green Bay is upwind. [SLIDE 4] This graph shows monthly average snowfall at these two locations. Frankfort receives much more snowfall, particularly in the winter months of December, January, February, due to the lake-effect snowfall phenomenon. For example, in January snowfall at Frankfort is more than double that of Green Bay. [SLIDE 5] The next slide illustrates this on a larger scale. This map shows the spatial distribution of average annual snowfall for the period 1971-2000. Snowfall is enhanced on the southern shoreline of Lake Superior, the eastern shoreline of Lake Michigan, the southeastern shoreline of Lake Erie, and the southeastern shoreline of Lake Ontario. Snowfall in the Upper Peninsula exceeds 200 inches in some places. Snowfall on the east shore of Lake Michigan is around 100 inches or more

The lakes have these effects on local climates because the water temperatures of the lakes respond more slowly to changes in seasonal solar forcing than do the surrounding land areas. [SLIDE 6] This graph shows the air temperature at Milwaukee compared to the Lake Michigan surface water temperature near Muskegon. During the spring and summer, the air temperature over the land warms more quickly than the lake temperature. However, in the fall and winter the lake temperature drops more slowly than the air temperature and thus remains warmer than the surrounding land.

[SLIDE 7] The lake-effect snowfall phenomenon results directly from these differences in temperature between the land and the lake surface waters. In the late fall and early winter, cold frontal passages from the north and northwest bring in air that is much colder than the lakes. As this cold air passes over the lakes, it is warmed and moistened by the warm lake surface waters. Snow forms in this warmer and moister air, which is still cold enough for snow, and is then deposited on the lee shores of the lakes

[SLIDE 8] This slide shows a typical weather map associated with lake-effect snow. A low pressure center is located in eastern Canada, with a cold front draped across the eastern and southern U.S. A cold high pressure center is located over the central U.S. In between these low and high pressure centers, cold northwesterly flow sweeps across the Great Lakes, causing lake-effect snow.

[SLIDE 9] (click on image and slideshow to view, then press escape to exit) I will now briefly describe the results of the study that we did as part of the U.S. National Assessment of the Consequences of Climate Variability and Change. Our study was on lake-effect snow, but in particular focused on heavy lake-effect snowstorms. We defined these as storms in which 8 inches or more of snowfall occurs over a 24-hour period. The initial focus of the study was on Lake Erie, but we extended the results to Lakes Superior and Michigan.

[SLIDE 10] The first part of the study identified the weather conditions that were associated with historical heavy lake-effect snowstorms. We found that four weather conditions were nearly always present with these storms. The first condition was that the surface air temperature was in the range of 14 to 32EF. It is obvious that we need temperatures at or below freezing to have snow. However, we also need temperatures that are not too cold. When temperatures are very cold, the atmosphere cannot hold as much water and is not as favorable for heavy snowstorms. As noted before, lake-effect snows occur because the water temperatures are warmer than the air temperatures. We found that the difference between lake surface temperature and air temperature was greater than 13EF during these heavy snowstorms. The third condition was high wind speeds, specifically greater than 14 miles per hour. Finally, the wind direction needed to be such that a long passage of air over the lake occurred, allowing for a long period of time in which the air could be moistened and heated.

[SLIDE 11] As described by Peter in the previous talk, we used two future climate simulations, one from the Hadley Centre climate model and the other from the Canadian Climate Center model. We examined differences in the frequency of those four weather conditions that were associated with heavy lake-effect snowstorms. We examined the differences in frequency between the present period and the end of the 21st Century.

[SLIDE 12] There are differences in the way these two models are constructed and I will describe a couple of these. Climate models break the atmosphere into boxes and determine values of weather conditions for each box. The boxes in the Hadley Centre model for the atmosphere are smaller than those in the Canadian model. Thus, for the Hadley Centre model there are more boxes representing the Great Lakes and thus somewhat more detail is available. [SLIDE 13] However, the models also need to simulate circulation patterns in the ocean because of the importance of coupling between the ocean and the atmosphere in the climate system. The boxes in the Canadian model for the ocean are smaller than those of the Hadley Center model. The models also need to treat the land surface and this treatment is more sophisticated in the Hadley Center model than in the Canadian model. These model differences lead to different projections for the future.

[SLIDE 14] We found that the Hadley Centre Model projects a more than 50% reduction in the frequency of heavy lake-effect snowstorms by 2100 while the Canadian Model projects a more than 90% reduction. This reduction is almost entirely due to warmer winter temperatures. The Hadley Center Model projects a 5EF warming for winter by 2100 while the Canadian Model projects a 10EF warming. We found that the frequency of the other conditions that are associated with heavy lake-effect snow did not change very much in the future. Because these other conditions did not change but temperature did change, we expect that there could be an increase in lake-effect rain that would occur instead of lake-effect snow.

[SLIDE 15] These were the results for Lake Erie. When we extrapolated the results to the other lakes, we found that the decrease in heavy lake-effect storms was about the same for southern Lake Michigan as for Lake Erie. However, we found a much smaller decrease for the Lake Superior snow belt, only about 10% for the Hadley Centre model projection.

[SLIDE 16] What confidence do we have in these results? Although there are differences in the model projections, these two models (and all other climate model simulations that have been done to date) indicate significant warming over the Great Lakes. Since temperature was found to be the key factor in this study, it seems that the eventual decline in the frequency of heavy lake-effect snowstorms is highly likely.

[SLIDE 17] There are potential compensating effects, particularly related to ice cover. Lake-effect snow is greatly diminished once ice cover develops. With a future warmer climate, it is highly likely that the duration of ice cover will decrease. Thus, there may be a longer season for lake-effect snow which could wholly or partially compensate for the warmer temperatures. Our results for the end of the 21st Century did take this into account. However this could be an important compensation in the early portion of the 21st Century, which we did not examine.

[SLIDE 18] Since this study indicated that air temperature is the key element, we performed a simple analysis of what could happen in the future based on historical climate data for the Great Lakes region. [SLIDE 19] This provides some insight into possible future outcomes. This analysis assumes that, as temperatures rise, the day-to-day variations remain the same. This assumption may not be valid but the analysis of the two climate model projections did not indicate significant changes in variations. [SLIDE 20] In this analysis we used a north-south transect of sites from Benton Harbor near southern Lake Michigan to Traverse City near northern Lake Michigan to Iron Mountain in the Upper Peninsula to Houghton near the shores of Lake Superior. [SLIDE 21] This graph shows the day-to-day variations in daily air temperature at Benton Harbor for a typical winter season for November through February. The horizontal green lines show the temperature range in which heavy lake-effect snow typically occurs. During December through February there are many days that are within this range. There are very few days that are too cold for heavy lake-effect snow while there are many days that are below freezing by only a few degrees. One very simple scenario for future warming is that the day-to-day variations in temperature remain the same but every day simply becomes warmer by some increment. The example I will show is for a future warming of 7EF. In this scenario each day in this record simply becomes 7EF warmer. [SLIDE 22] The next graph shows what happens when we add 7EF to each day. In this case, many of the days that were within the heavy lake-effect snow band are now warmer than 32EF. The few days that were too cold for heavy snow now become favorable for heavy snow but they are much fewer in number than the days which now are too warm for heavy snow. Thus the number of days favorable for heavy snow has decreased substantially.

[SLIDE 23] The next graph shows a typical year at Houghton. Winters are colder at Houghton. There are many days during the winter when temperatures fall within the range associated with heavy snow. But there are quite a few days that are colder than this range. [SLIDE 24] When we add 7EF of warming, some days that were in the heavy snow range are now too warm for heavy snow. However, this is largely compensated for by the days that were too cold that are now within the heavy snow range. Overall the number of days in the heavy snow range does not change very much. [SLIDE 25] This is illustrated in the next graph which also shows the results for Iron Mountain and Traverse City. This shows the number of days in the heavy snow range for current climate conditions (0 on the horizontal scale), and then for various degrees of warmth, up to 10EF. At Benton Harbor, each degree of warming decreases the number of favorable days; by the time we reach 10EF of warming the number of days has decreased by more than half. By contrast, at Houghton, the number of days decreases very slowly for every degree of warming; by the time weÕve reached 10EF of warming, there has been a decrease, but only about 20 percent. The results for Traverse City and Iron Mountain are intermediate between the Benton Harbor and Houghton decreases.

[SLIDE 26] Since this study was performed, new climate model results have become available, one from a newer version of the Hadley Center model and another from a U.S. model. These projections are consistent in indicating significant warming in the Great Lakes region. [SLIDE 27] This slide shows two of the temperature projections from these models for the winter. Both project substantial warming by 2100. One way of thinking about climate change is that a warming is equivalent to a movement towards the south to warmer climates. Taking one of the warmer projection from the newer model results, how far south would we need to move in todayÕs climate to find similar temperatures? [SLIDE 28] This next graph shows the answer. For this projection of the future, this is equivalent to Michigan moving south to Arkansas, a radical change in MichiganÕs climate.

[SLIDE 29] In our assessment study, we did not examine the effects of climate change on total snowfall. However, an examination of past historical data can provide some guidance. This examination shows in general that warmer winters are generally less snowy. [SLIDE 30] This is illustrated by historical climate data for Traverse City. There are about 100 years of data for Traverse City and in this graph each dot shows the data for one year. November through March snowfall is plotted against November through March air temperature. Although there is a great deal of scatter in the historical data, in general the low snowfall years tend to occur with warmer temperatures while the higher snowfall years tend to occur with colder temperatures. The straight line is the best fit to these data and it shows that there is about a 25% decrease in average snowfall for every 5EF warming.

[SLIDE 31] To conclude, the models suggest that during the 21st Century heavy lake-effect snowstorms will dramatically decrease in frequency in the southern portions of the Great Lakes basin. However, changes in the northern portions of the basin may be much less. These changes are caused almost entirely by warmer temperatures; other conditions favorable for heavy snow do not change much in the climate models. [SLIDE 32] We were not able to examine the speed which these changes will take place. So there remains an important question. Will decreased frequencies begin to occur early in the 21st Century or later? Even if the decreased frequencies do not occur until later, the models do show warming in the early part of the 21st Century, so mid-winter melting events may become more frequent, even if total snowfall does not change.