Abstract

Research over the past several decades has indicated that snowfall has increased dramatically over portions of the past century across those areas of the Great Lakes region of North America that are subject to lake-effect snowfall. Within this study, time series of annual midwinter snowfall within lake-effect areas show evidence of a clear increase in both snowfall and snowfall frequency through a 40-yr period beginning in the early 1930s and ending in the early 1970s. The goal of the work presented here is to determine to what extent the apparent increases in lake-effect snowfall actually modified the winter hydroclimate of the areas.

Simple hydroclimatic analysis of midwinter precipitation to the lee of Lakes Erie and Ontario for the period of significant snowfall increases suggests that the changes were a product of 1) a shift toward more precipitation events that were snowfall rather than rainfall, 2) an associated decrease in midwinter rainfall, 3) an increase in the intensity of individual snowfall events, and 4) an increase in the snowfall/snow water equivalence ratio. The balance was a small increase in total precipitation confined to areas in close proximity to the lakes across northeastern Ohio and western New York, while areas outside the regions generally experienced an overall decrease in midwinter precipitation. While the cause(s) of the snowfall trends remains elusive, the results of the work presented here suggest that no great long-term regional change occurred in the true wintertime seasonal hydroclimate of the lake-effect areas. Rather, much of the touted snowfall increase simply came at the expense of rainfall events to produce only small changes in total precipitation over the time period of significant snowfall increase.

1. Introduction

Each winter, lake-effect snowfall routinely influences the lives of the inhabitants of areas downwind of any of the five Great Lakes of North America (Fig. 1a). The phenomenon by which snowfall is produced from the passage of cold air masses over the relatively warm waters of the lakes has been generally understood for decades (e.g., Holroyd 1971; Chagnon and Jones 1972; Baker 1976; Dewey 1979; Eichenlaub 1979; Braham and Kelly 1982; Niziol et al. 1995; Kristovich and Braham 1998). Cold air masses of low water vapor content originating in Canada are warmed and moistened from below as they pass over the Great Lakes as a result of efficient energy and moisture fluxes. The product is lower-atmospheric instability and upward motion that is enhanced by boundary layer turbulence, friction from downwind land surfaces, and upslope flow across higher terrain (Changnon and Jones 1972; Dewey 1979; Niziol et al. 1995). The resultant lake-effect snowfall on the leeward side of any of the five Great Lakes can be quite dramatic. One-day snowfalls of greater than 175 cm have occurred, as have 5-day events of nearly 300 cm, while seasonal totals have approached 1200 cm in the lake-effect snowfall region of New York (Dewey 1977). The lake bodies are also known to influence the speed and intensity of midlatitude cyclones to produce snowfall that is lake enhanced (Angel and Isard 1997; Angel and Isard 1998). As such, lake-effect and lake-enhanced snowfall is at the center of the wintertime hydroclimate of the lake-effect snowbelts in the Great Lakes region.

Fig. 1.

(a) The study region, (b) elevation across the region, and (c) distribution of daily precipitation stations for the 50- and 70-yr study periods

Fig. 1.

(a) The study region, (b) elevation across the region, and (c) distribution of daily precipitation stations for the 50- and 70-yr study periods

Climatological research over the past several decades has indicated that snowfall seems to have increased rather dramatically over much of the past century across those portions of the Great Lakes region that are subject to lake-effect snowfall (Namias 1960; Thomas 1964; Eichenlaub 1970; Braham and Dungey 1984; Harrington et al. 1987; Leathers et al. 1993; Norton and Bolsenga 1993; Ellis and Leathers 1996; Leathers and Ellis 1996). Monotonic 70-yr snowfall increases of up to 120% have been found across the lake-effect snowbelts of Michigan (Braham and Dungey 1984). Similarly, 60-yr snowfall trends (1931–90) of 0.5–2.6 cm yr−1 across the lake-effect regions of western Pennsylvania and New York have been documented (Leathers and Ellis 1996). Explanation of the apparent increases in lake-effect snowfall has been elusive. Leathers and Ellis (1996) found that an increase in the frequency of the synoptic weather patterns conducive to lake-effect snowfall explained 30%–60% of the total snowfall increases that occurred east of Lakes Erie and Ontario over a 40-yr period. An additional increase in the intensity of the snowfall associated with the synoptic weather patterns explained much of the remaining increase in total snowfall, but without an explanation of why the intensity increase occurred.

Clearly there is evidence that the lake-effect snowfall mechanism appeared to become more productive in terms of measured snowfall within the past century. The importance of the question of why is matched with the question, To what extent has the general hydroclimate actually changed? The work presented here is aimed at establishing a better understanding of the nature of the winter lake-effect snowfall increases by addressing such questions as 1) Did total precipitation increase in lake-effect regions? and, What fraction of any trends is attributable to snowfall rather than rainfall? 2) Did the frequencies of snowfall and rainfall events change through time, or in other words, did the fraction of precipitation events that were snowfall change? 3) Did the intensity of precipitation events change? 4) Did the ratio snowfall/snow water equivalence (SWE) change through time? The specific goal of answering these questions is to determine if lake-effect snowfall increases acted to modify the winter hydroclimate of lake-effect areas or if the snowfall increases were simply a representation of a trend toward a greater fraction of the typical precipitation falling as snow rather than rain.

2. Data and methods

A broad area of study was chosen to ensure capture of the lake-effect snowfall areas within the United States (Fig. 1a). The snowbelts of Canada were not considered to avoid problems associated with international inconsistencies in snowfall data (primarily U.S. snow depth versus Canadian SWE). It is assumed that general patterns and trends in the hydroclimate of U.S. lake-effect areas translate to similarities with the same types of areas within Canada. Elevation changes across the study area (Fig. 1b) are important to note because the enhancement of lake-effect snowfall in areas of higher terrain downwind of the lakes leaves an imprint on the regional snowfall pattern.

Daily total precipitation and snowfall data were taken from the National Climatic Data Center (NCDC) Summary of the Day database constructed of measurements from first-order weather stations and the U.S. Cooperative Observer Network. Only stations with at least 90% data completeness were considered (based on Stooksbury et al. 1999). Two subsets of stations were constructed from the NCDC database—527 stations with at least 50 yr of data (1952–2001) and, to extend the record, 105 stations with at least 70 yr of data (1932– 2001; Fig. 1c). The two periods were logically chosen based upon dramatic reductions in the number of stations when considering longer periods of study. The number of available stations only decreased from 568 for a 40-yr study period to 527 for the 50-yr period referenced above; however, the number decreased to 139 stations when considering a 60-yr period. The sample size only decreased to 105 for the 70-yr period referenced above, but decreased significantly to 65 when considering an 80-yr period.

Using daily snowfall data, daily total precipitation data were stratified into amounts and frequencies of rainfall and SWE by assuming that a daily total precipitation value was entirely SWE on a day in which snowfall was recorded and entirely rainfall when no snowfall was recorded. This crude assumption is necessary given the nature of the database within which daily events of both rainfall and snowfall (a rain–snow mix) are not distinguishable. Although likely to not be of great importance, this fact should not be forgotten when considering the results of this work. The stratified data were used to calculate per event (daily) intensities of snowfall, precipitation, SWE, and rainfall.

Daily values of total precipitation, rainfall, SWE, snowfall, and individual event intensities of each were totaled to monthly values, as were frequencies of precipitation, rainfall, and snowfall for the period November through April, and those months with less than 90% data completeness were removed from the record. Mean monthly snowfall/SWE ratios were constructed using monthly totals of snowfall and SWE. Monthly frequency values for precipitation and snowfall were used to calculate the fraction of precipitation events that were snowfall. Mean monthly values of individual event intensity for each of the precipitation types were constructed from the daily calculation explained above. As lake-effect snowfall is most prominent during the heart of the winter season, monthly values for the midwinter period December through February were translated into 3 month seasonal values. Any season missing 1 month of data was stricken from the record. Monthly climatological means for the 30-yr period 1972–2001 were mapped by interpolating station data to a 30 (longitudinal) × 14 (latitudinal) grid across the entire study area (0.75° latitude by 0.75° longitude grid cells).

A linear regression line was fit through the monthly and seasonal values associated with the precipitation data for specified periods of record at each station in order to approximate the change in each variable with time. The slope of the regression line represents the trend in the variable. After examination of the monthly and seasonal snowfall data at one reliable station (long record, data completeness) within the general snowbelts of each of Lakes Erie, Michigan, Superior, and Ontario (Fig. 1a), monotonic trends were most evident across the first 40 yr (from 1932/33 through 1971/72) of the 70-yr period. As such, trend analyses were confined to that 40-yr period of interest. However, this made analysis to the lee of Lakes Michigan and Superior difficult because of the lack of stations with long records within Michigan (Fig. 1c). Therefore, spatial analyses of precipitation trends were confined to the lee of Lakes Erie and Ontario. Trends in midwinter precipitation were mapped by interpolating point data to a 1° latitude by 1° longitude grid across the smaller eastern portion of the study region.

3. Results and discussion

a. Snowfall climatology

Using data from the 527 stations with at least 50 yr of data, a 30-yr (1972–2001) climatology of monthly snowfall for the region was produced. Lake-effect snowfall is evident in the pattern of mean monthly snowfall across the region from November through April. Magnitudes are greatest and the signature of the lakes is most clearly defined during the period December through February. Here, mean January values are used to illustrate the spatial pattern and magnitude of snowfall to the lee of each lake relative to the rest of the region (Fig. 2a). Evident is the influence of the lakes themselves, the influence of elevation (Fig. 1b), and the fact that Lake Ontario seems to be the most effective snowfall producer, at least within the United States. With the advection of cold air across the lakes predominantly from central Canada, westerly and northwesterly winds produce a snowfall signature to the east and southeast of the lakes. Relatively high snowfall amounts are also evident in the high elevations of the Appalachian Mountains in southern Pennsylvania, western Maryland, and West Virginia—more remote areas that are subject to lake-effect snowfall events (McMillan and McGinnis 1975).

Fig. 2.

(a) Mean Jan snowfall, (b) precipitation, (c) SWE, (d) rainfall, (e) percentage of precipitation as SWE, and (f) snowfall/SWE ratio

Fig. 2.

(a) Mean Jan snowfall, (b) precipitation, (c) SWE, (d) rainfall, (e) percentage of precipitation as SWE, and (f) snowfall/SWE ratio

In order to gain perspective of the contribution of snowfall to the general wintertime hydroclimate across the region, mean monthly liquid precipitation values (total precipitation, SWE, rainfall) for the midwinter month of January are presented. To the lee of the western lakes of Superior and Michigan the effect of the lakes, predominantly lake-effect snowfall, is evident in the mean total precipitation for January (Fig. 2b). To the lee of Lakes Erie and Ontario the signature of the lakes is clouded a bit by the relatively large amount of total precipitation in January near the coastline of the Atlantic Ocean (Fig. 2b). The low water content of lake-effect snowfall produced in a relatively cold, dry atmosphere prevents a replication of the snowfall pattern in the mean precipitation pattern. Still, a high amount of precipitation to the lee of each of the lakes evidences their effects. Much of the January precipitation to the lee of the lakes is associated with snowfall, as is evident by the pattern of mean January SWE (Fig. 2c). This is more the case to the lee of Lake Superior (inland, northerly location) and less the case to the lee of Lake Erie (less inland, southerly location). The pattern of mean January rainfall (Fig. 2d) clearly reflects the effects of latitude and proximity to the relatively warm Atlantic Ocean to the east.

On the whole, SWE accounts for 50%–60% of the mean January precipitation to the lee of Lake Erie, 55%– 80% to the lee of Lake Michigan, 65%–75% to the lee of Lake Ontario, and 90%–95% to the lee of Lake Superior (Fig. 2e). As lake-effect snowfall typically is of low water content, snowfall/SWE ratios are typically high to the lee of the lakes (Fig. 2f). Values are generally greater than 18–20:1 to the lee of the lakes with values as low as 10–14:1 across the extreme southern and eastern portions of the study region.

The pattern of mean precipitation frequency in January (Fig. 3a) clearly reflects the influence of the lakes and the many days on which snowfall is recorded to the lee of the lakes (Fig. 3b). Monthly patterns of mean snowfall frequency exhibit the lake-effect signature and the fact that the phenomenon is most active during the period December through February (not shown). Within some lake-effect areas snowfall is typically recorded on nearly one-half of the days during January (Fig. 3b). Rare are days of rainfall in January across the region except across the southern and eastern portions of the region (Fig. 3c). Within the pattern there is indication of an increased rainfall frequency from West Virginia northward into western Pennsylvania compared to just east of this area. This most likely represents common scenarios of warm air advection along the western portion of the Appalachian Mountains in association with a midlatitude cyclone, contrasted by cold air damming along the eastern portion of the Appalachians. Within the lake-effect areas, the percentage of January precipitation events that are snowfall ranges from 55% to 70% east of Lake Erie to greater than 90% to the lee of Lake Superior (Fig. 3d).

Fig. 3.

(a) Mean Jan precipitation frequency, (b) snowfall frequency, (c) rainfall frequency, and (d) percentage of precipitation events that are snowfall

Fig. 3.

(a) Mean Jan precipitation frequency, (b) snowfall frequency, (c) rainfall frequency, and (d) percentage of precipitation events that are snowfall

The spatial climatology of mean snowfall for the study region indicates 1) rather clear patterns of the effects of Lakes Erie, Michigan, Superior, and Ontario, 2) the large majority of midwinter precipitation within the lake-effect areas is in the form of snowfall (SWE), with some areas averaging approximately a snowfall event every other day in January, and 3) snowfall/SWE ratios within the lake-effect areas are rather high because of the low water content of lake-effect snowfall.

b. Snowfall trends

In order to determine a specific period over the past century during which snowfall within the snowbelts of the Great Lakes seemed to increase, annual snowfall at four long-record stations was examined: Lowville and Fredonia, New York, and Traverse City and Ironwood, Michigan (Fig. 1a). In order to focus on lake-effect snowfall, trends of the total midwinter seasonal snowfall are examined. The season is the 3-month period December through February when snowfall is most prominent across the region. Snowfall at Lowville, Fredonia, and Traverse City seemed to increase rather steadily across roughly a 40-yr period beginning in the early 1930s through the early 1970s (Figs. 4a–c). At Ironwood (Fig. 4d), little change occurred through the first one-quarter of that 40-yr period, but a marked increase is evident from the early 1950s through the early 1970s. Over the most recent 2 decades, snowfall amounts at each station have generally remained higher than the long-term mean but do not show a continued increase.

Fig. 4.

Mean midwinter (Dec–Feb) snowfall at (a) Lowville, (b) Fredonia, (c) Traverse City, and (d) Ironwood. Thick line is a 5-yr running mean

Fig. 4.

Mean midwinter (Dec–Feb) snowfall at (a) Lowville, (b) Fredonia, (c) Traverse City, and (d) Ironwood. Thick line is a 5-yr running mean

The 3-month frequency of snowfall shows a marked increase over much of the 40-yr period from the early 1930s to the early 1970s at Lowville, Traverse City, and Ironwood (Figs. 5a, 5c, and 5d). At Fredonia, the frequency of snowfall decreased from the early 1930s through the late 1950s before steadily increasing through the late 1980s (Fig. 5b).

Fig. 5.

Mean midwinter (Dec–Feb) snowfall frequency at (a) Lowville, (b) Fredonia, (c) Traverse City, and (d) Ironwood. Thick line is a 5-yr running mean

Fig. 5.

Mean midwinter (Dec–Feb) snowfall frequency at (a) Lowville, (b) Fredonia, (c) Traverse City, and (d) Ironwood. Thick line is a 5-yr running mean

If at Fredonia the frequency of snowfall decreased over much of the period for which an increase in snowfall occurred, then the intensity of individual snowfall events had to increase. Such a change is evident at Fredonia (Fig. 6b), but only slightly at Lowville beginning in the early 1950s (Fig. 6a). In contradiction to these trends, the stations to the lee of the two western lakes showed a decrease in snowfall event intensity during much of the period from the early 1930s through the early 1970s (Figs. 6c and 6d).

Fig. 6.

Mean midwinter (Dec–Feb) snowfall event intensity at (a) Lowville, (b) Fredonia, (c) Traverse City, and (d) Ironwood. Thick line is a 5-yr running mean

Fig. 6.

Mean midwinter (Dec–Feb) snowfall event intensity at (a) Lowville, (b) Fredonia, (c) Traverse City, and (d) Ironwood. Thick line is a 5-yr running mean

The time series of annual midwinter snowfall and snowfall frequency at long-record stations associated with Lakes Erie, Michigan, Ontario, and Superior generally show evidence of a rather monotonic increase in both snowfall and snowfall frequency through a 40-yr period beginning in the early 1930s and ending in the early 1970s. The exceptions are with snowfall to the lee of Lake Superior and snowfall frequency to the lee of Lake Erie, where in each case a clear increase did not begin until the 1950s. Over the last 3 decades, snowfall and snowfall frequency remained rather high at each of the stations but did not show a clear trend. Therefore, the general period over which a trend in snowfall seemingly took place within a lake-effect snowbelt of each of the four Great Lakes examined was the early 1930s through the early 1970s. Over that period, snowfall appears to have increased to the lee of each of the lakes and it did so primarily as a function of an increase in the frequency of snowfall events. The increased frequency produced higher snowfall totals with the aid of an increase in event intensity to the lee of the eastern Great Lakes and despite an event intensity decrease to the lee of the western lakes.

c. Hydroclimatic trends

Increasing snowfall within the lake-effect areas of the Great Lakes most certainly had significant impacts on the social and economic aspects of the region, and much has been made of the snowfall increases from a general climatic standpoint. To examine whether or not the snowfall increases were associated with a significant change in the actual hydroclimate of the lake-effect areas, the temporal patterns of liquid precipitation and its two components, SWE and rainfall, were examined. Initially, this was done for the four stations of long record for the period of significant change in snowfall over the past century, or the early 1930s through the early 1970s. Specifically, the period examined was the winter season (December through February) of the 40-yr period from 1932/33 through 1971/72.

Precipitation at Lowville, Fredonia, and Traverse City (Figs. 7a–c) exhibited a similar trend of an increase through approximately 1950, or the first one-half of the 40-yr period, and then little trend over the latter half of the period, with values near the mean of the first half of the period. At Ironwood (Fig. 7d), a slow decline in precipitation occurred through the early 1960s, or approximately the first three-quarters of the period, prior to a small increase over the last decade.

Fig. 7.

Mean midwinter (Dec–Feb) precipitation at (a) Lowville, (b) Fredonia, (c) Traverse City, and (d) Ironwood. Thick line is a 5-yr running mean

Fig. 7.

Mean midwinter (Dec–Feb) precipitation at (a) Lowville, (b) Fredonia, (c) Traverse City, and (d) Ironwood. Thick line is a 5-yr running mean

The precipitation increases across the first one-half of the period at Lowville and Traverse City (Figs. 7a and 7c) were largely the product of increases in SWE (Figs. 8a and 8c), while mean seasonal SWE leveled off to show no trend at each station over the second one-half of the period. At Fredonia, the modest increase in precipitation across the first one-half of the period (Fig. 7b) was not marked by an increase in SWE (Fig. 8b), which showed very little trend throughout the 40-yr period. At Ironwood, the decline in precipitation through approximately the first 30 yr of the period (Fig. 7d) was largely the product of a decrease in SWE (Fig. 8d), which, like precipitation, increased over the last decade of the period.

Fig. 8.

Mean midwinter (Dec–Feb) SWE at (a) Lowville, (b) Fredonia, (c) Traverse City, and (d) Ironwood. Thick line is a 5-yr running mean

Fig. 8.

Mean midwinter (Dec–Feb) SWE at (a) Lowville, (b) Fredonia, (c) Traverse City, and (d) Ironwood. Thick line is a 5-yr running mean

At Lowville, a slight increase in rainfall contributed to the overall increase in precipitation across the first one-half of the 40-yr period (Fig. 9a), yet very little if any of the similar precipitation increase at Traverse City is attributable to rainfall (Fig. 9b). The drop-off in precipitation after the early 1950s at each station (Figs. 7a and 7c) was almost entirely the product of lessened amount of mean seasonal rainfall (Figs. 9a and 9c), as SWE showed very little trend over the latter one-half of the period (Figs. 8a and 8c). At Fredonia, the similar increase in precipitation over the first one-half of the period (Fig. 7b) was almost entirely the product of an increase in rainfall (Fig. 9b), as SWE changed little (Fig. 8b). As with Lowville and Traverse City, the drop-off in precipitation after the early 1950s was much more a product of decreased rainfall, with SWE showing little trend. At Ironwood, the steady decline in precipitation through the early 1960s and the increase over the last decade of the period (Fig. 7d) both seem to be entirely the product of similar trends in SWE (Fig. 8d), as rainfall showed little trend through either period (Fig. 9d).

Fig. 9.

Mean midwinter (Dec–Feb) rainfall at (a) Lowville, (b) Fredonia, (c) Traverse City, and (d) Ironwood. Thick line is a 5-yr running mean

Fig. 9.

Mean midwinter (Dec–Feb) rainfall at (a) Lowville, (b) Fredonia, (c) Traverse City, and (d) Ironwood. Thick line is a 5-yr running mean

In summary, three of the four stations examined showed an overall precipitation increase from the early 1930s through the early 1950s, with smaller mean seasonal magnitudes from the 1950s through the early 1970s but with no decreasing trend. The fourth station exhibited a decreasing trend in precipitation from the early 1930s through the early 1960s, with a slight increase over the decade into the early 1970s. The varying trends in precipitation across the period of study are in association with rather monotonic increases in snowfall through the full 40-yr period at each of the four stations. This suggests that the hydroclimatic implications of the snowfall increases are not straightforward, as total liquid precipitation showed no steady pattern of an upward trend as with snowfall. In fact, the liquid equivalent of snowfall (SWE) showed a clear positive trend at only two stations (Lowville and Traverse City), and in each case the trend only appears for the first one-half of the 40-yr period. At another station SWE shows no trend (Fredonia), while at the fourth station SWE decreases through the first 30 yr of the 40-yr period from 1932/ 33 through 1971/72.

The time series of snowfall and total precipitation at the four long-record stations do not suggest a simple relationship between the two, meaning that the much-discussed lake-effect snowfall increase within the past century may not have produced a clear trend in the true hydroclimate of the lake-effect regions. Using multiple stations in a spatial analysis should provide a clearer, more reliable indication of the effects of snowfall increases on the hydroclimate of the regions. However, it is first appropriate to consider other factors that may have unnaturally altered snowfall and precipitation totals through time, such as changes in measurement techniques, station locations, and observation times.

d. Metadata considerations

In considering station and measurement changes that may have influenced the historical precipitation records within the lake-effect regions of the Great Lakes, the station histories of the four long-record stations were examined against the temporal patterns of precipitation. Given the rather smooth trends in the data for the four stations, it is doubtful that an abrupt station change, such as a station move, change in measurement technique, or change in observation time, will express itself as the culprit of the trends. However, knowledge of station changes is a reminder of the caution with which precipitation trends, especially snowfall trends, should be viewed.

Changes in the time of observation at the stations should have a marginal effect on seasonal precipitation totals, as any differences between a morning and evening 24-h observation (e.g., daytime heating, evaporation, and windiness) should be largely reduced through the coarse seasonal totaling used here. Through 1967, metadata for U.S. cooperative stations only list the observing time as morning or afternoon, and thereafter through 1981 all observation times are listed as afternoon as a result of the standardization to afternoon observations (specific observation times were not recorded until January 1982). At Ironwood, afternoon observations were made for the entire 40-yr period from 1932/ 33 through 1971/72. Obviously, the generally common observation time eliminates concerns with this type of station change. At both Fredonia and Lowville, morning observations were made through the winter of 1947/48 while afternoon observations began in the subsequent winter. At each station, the record of snowfall, snowfall frequency, snowfall event intensity, and snow water equivalence showed no marked change at or shortly after the change in observation time. Rainfall showed somewhat of an abrupt decrease in magnitude, but approximately 5–7 yr after the change in observation time. Total precipitation showed the same decrease, but of a smaller magnitude. At Traverse City, afternoon observations were made through 1940/41 before changing to morning observations from 1941/42 through December 1946. In January 1947 observations reverted back to an afternoon time until December 1948 when midnight became the standard 24-h observing time at Traverse City (first-order weather station). No abrupt changes in any of the precipitation variables occurred near the time of any of the observation changes at Traverse City.

Concerns over station location changes are typically associated with the effects of elevation on precipitation, especially snowfall. Station location history for U.S. cooperative stations exists for the period 1948–present and somewhat earlier for first-order weather stations (Traverse City). Although the time of observation did not change at Ironwood, the station underwent several location changes during the period from 1932/33 through 1971/72. Six location changes of the Ironwood station have occurred, never more than 2 min of latitude or longitude, and with associated elevation changes of between 5 and 28 m. The two largest elevation changes occurred in 1955/56 (+28 m) and 1963/64 (−28 m). The only significant change in precipitation occurred with snowfall, snowfall frequency, and SWE shortly after 1963/64. However, the change was toward snowier conditions, which is opposite what a decrease in elevation might suggest, but in accord with what a small move north and west toward the lake might suggest. Still, the change in precipitation is not abrupt, but continues on an upward trend that suggests a continuous change rather than an abrupt change as might be expected from the relocation of a station.

Lowville underwent two station location changes during the 40-yr period from 1932/33 through 1971/72. A small westward movement of the station with an increase in elevation of 11 m occurred prior to the 1968/ 69 winter season, and a small southeastward movement with a decrease in elevation of 25 m occurred prior to the 1971/72 season. The latter relocation is prior to the last year of the 40-yr period of record and therefore has little bearing on the time series that are examined here. In the 4 yr of the period after the first relocation, many of the snowfall variables did show an increase, which could conceivably be associated with the increase in elevation and the closer proximity of the station to Lake Ontario. However, the small changes over the few years at the end of the record should have very little influence on the trends of the full 40-yr period.

At Fredonia, the station was relocated once over the period of study, moving slightly southeast and gaining 3 m of elevation prior to the 1960/61 winter season. As with Ironwood, small trends in the precipitation record for Fredonia occurred several years after the relocation, but the trend was rather continuous and not abrupt as what might result from a station move. The station at Traverse City underwent two very minor relocations over the course of the study period, neither of which was large enough to produce a latitude–longitude change on a resolution of minutes. Prior to the 1944/ 45 winter season the station was relocated to gain 7 m of elevation, and prior to the 1948/49 season a relocation of the station resulted in a loss of 11 m of elevation. Neither change was associated with a noticeable change in the precipitation time series for Traverse City, but rather tended to occur within a larger continuous trend pattern.

There appears to be no clear indication of a signal of station relocations within the temporal patterns of precipitation at the four long-record stations. The patterns tend to be rather continuous with no abrupt changes. Still, the general precipitation trend statistics presented later in this paper should be viewed with some caution given the likelihood of multiple station relocations through time and the potential impact of the relocation on precipitation records.

Likely to be the greatest concern with measurement of snowfall through time is potential changes in measurement practice. Station history data within the database of the Cooperative Observer Program of the National Weather Service as managed by NCDC does not provide indication of measurement techniques. This means that there is a lack of evidence of observer changes and alteration of measurement techniques of snow (snow board versus multiple sampling; once per day versus 6-hourly) and the liquid equivalent (gauge collection and melting; core sample). Extensive quality control of data at NCDC provides significant assurance of error elimination but does not account for vagaries that might be introduced by changes in measurement technique. However, as with a change in station location, it would seem that a significant change in the manner by which snowfall is measured would produce a systematic and relatively abrupt change in the temporal record. This is clearly not the case in the precipitation records examined here, as the trends discussed are continuous and rather monotonic in nature. This provides a significant measure of confidence that the trends are not artifacts of operational changes to the stations used, and an additional amount of confidence is added through spatial analysis that relies on the records of many stations.

e. Spatial patterns of hydroclimatic trends

To examine whether or not the snowfall increases were associated with a significant change in the actual hydroclimate of the lake-effect areas, spatial analyses of liquid precipitation types were constructed. In order to capture the period of significant change over the past century, the early 1930s through the early 1970s, the analyses were based upon the database containing stations with precipitation records of 70 yr and longer. Specifically, the period examined was the winter season (December through February) of the 40-yr period from 1932/33 through 1971/72. Critically problematic is a lack of stations within Michigan that continue a 70-yr or greater record of daily precipitation and snowfall measurements (n = 3; Fig. 1c). As such the spatial analyses were confined to the eastern lakes, where a reasonable distribution of long-record stations exists, especially in New York and Pennsylvania (Fig. 1c).

The spatial pattern of the trend in snowfall for the 40-yr period to the lee of Lakes Erie and Ontario clearly shows the influence of the lakes (Fig. 10a). Trends are positive throughout the region, and the greatest trends are logically located in the areas of greatest annual snowfall to the lee of the lakes and in the area of higher elevation associated with the Appalachian Mountains (Fig. 1b). The greatest trends to the lee of Lakes Erie and Ontario are greater than 1.5 and 3 cm yr−1 respectively, equaling 40-yr changes of 60 and 120 cm of snowfall in midwinter.

Fig. 10.

Trend in midwinter (Dec–Feb) (a) snowfall, (b) SWE, (c) rainfall, (d) precipitation, (e) percentage of precipitation that is SWE, and (f) snowfall/SWE ratio. Shading indicates positive trends

Fig. 10.

Trend in midwinter (Dec–Feb) (a) snowfall, (b) SWE, (c) rainfall, (d) precipitation, (e) percentage of precipitation that is SWE, and (f) snowfall/SWE ratio. Shading indicates positive trends

As would be expected given the increase in snowfall, the amount of liquid precipitation measured on days for which snowfall was recorded, SWE, increased across the entire region (Fig. 10b). As with snowfall trends, the greatest trends in SWE were associated with each of the lakes and the higher elevations of the Appalachian Mountains. Despite the greatest snowfall increases associated with Lake Ontario being centered south of the lake, the greater SWE increases occurred over the higher elevations to the east of the lake. To the lee of Lakes Erie and Ontario the greatest trends in SWE are larger than 0.10 and 0.18 cm yr−1, respectively, extrapolating to 40-yr changes of 4 and greater than 7 cm of SWE. Seemingly, these types of changes would impact the wintertime hydroclimate of the lake-effect areas with some significance.

Opposite to the trends in snowfall, trends in midwinter rainfall were negative across the entire region (Fig. 10c), as great in magnitude as −0.16 cm yr−1 in some areas. Across the lake-effect areas, trends ranged between approximately −0.04 and −0.12 cm yr−1, equaling 40-yr decreases of 1.6 to 4.8 cm of rainfall in midwinter. Trends in total precipitation (Fig. 10d) show that the increase in SWE outweighed the decrease in rainfall across a significant portion of the area to the lee of Lake Ontario, but only within a small area to the lee of Lake Erie. Across the majority of the remainder of the region, the trend in total precipitation in midwinter across the 40-yr period was negative, or dominated by the decrease in rainfall.

Despite large trends in snowfall across the entire region east of Lakes Erie and Ontario, the trend in the water equivalent of the snow was large enough only within areas just to the lee of the lakes to outweigh regionwide decreases in rainfall during midwinter over the 40-yr period. This possibly suggests that the overall wintertime hydroclimate of the lake-effect areas changed less significantly than the snowfall trends alone would indicate and that the significant increase in snowfall may have been a product of simply more precipitation events that were snowfall rather than rainfall. For the greater region east of Lakes Erie and Ontario, the wintertime hydroclimate was marked by a decrease in precipitation, which would likely be the result if the greater region was indeed subject to more lake-effect scenarios—regionally cold, dry air masses.

Trends in the percentage of precipitation that was recorded as SWE support the idea of an increase in the source of the precipitation within the region being snowfall (Fig. 10e). The percentage of midwinter precipitation falling as snow increased across the entire region over the 40-yr period, especially so to the lee of Lakes Erie and Ontario, with values as high as 0.6% yr−1 for a 40-yr change as great as 24%. This equates to nearly a 50% difference between the fraction of precipitation that was snowfall and the fraction that was rainfall. Potentially adding to the dramatic snowfall increase was a modest increase in the midwinter snowfall/SWE ratio over the 40-yr period to the lee of Lakes Erie and Ontario (Fig. 10f). Trends ranged from 0.1 per year to the southeast of Lake Erie and southward into the higher elevations of the Appalachians to as great as 0.3 per year to the south of Lake Ontario. This equals 40-yr increases in the snowfall/SWE ratio by amounts of 4– 12:1, meaning that the depth of snow became ever greater than the amount of liquid precipitation that it contained.

Across the entire region east of Lakes Erie and Ontario, the frequency of snowfall in midwinter increased over the 40-yr period (Fig. 11a), although in a configuration that is not clear evidence of a lake effect. The greatest trends extend southward from the easternmost end of each lake, and the largest trends range from 0.3 to 0.4 days yr−1 for a 40-yr difference between 12 and 16 days in midwinter. The less-clear pattern of frequency trends compared to magnitude trends may be a product of the difficulty in defining a clear trend in a variable of potentially small numbers, such as snowfall frequency across some areas of the region. Consistent with a decrease in rainfall across the entire region is a decrease in the frequency of rainfall (Fig. 11b). Cumulatively, the frequency of midwinter precipitation increased over the 40-yr period only in two areas inland from each of the two lakes (Fig. 11c), providing further evidence that possibly no significant change in the ultimate hydroclimate of the region occurred. With an increase in the frequency of snowfall over the entire region and a general decrease in the frequency of rainfall, the trend in the percent of precipitation events that were snowfall increased (Fig. 11d).

Fig. 11.

Trend in midwinter (Dec–Feb) (a) snowfall frequency, (b) rainfall frequency, (c) precipitation frequency, and (d) percentage of precipitation events as snowfall. Shading indicates positive trends

Fig. 11.

Trend in midwinter (Dec–Feb) (a) snowfall frequency, (b) rainfall frequency, (c) precipitation frequency, and (d) percentage of precipitation events as snowfall. Shading indicates positive trends

Seemingly, the increases in snowfall across the regions to the lee of Lakes Erie and Ontario are the product of increases in the frequency of snowfall at the expense of rainfall events, and to some extent the product of an increase in the amount of snowfall per amount of associated precipitation (SWE). A final potential contributor to the increase in snowfall and SWE within the lake-effect areas is an increase in the intensity of individual snowfall events. Over the 40-yr period, the intensity of snowfall events increased to the lee of both Lakes Erie and Ontario (Fig. 12a). Away from the lakes to the east of the region, the intensity of snowfall events showed very little change and even decreased across much of the area. To the lee of the lakes, event intensity trends generally ranged between 0.04 and 0.06 mm per event per year, equaling 40-yr differences of between 1.6 and 2.4 mm per event (0.16–0.24 cm per event).

Fig. 12.

Trend in midwinter (Dec–Feb) (a) snowfall event intensity, (b) SWE event intensity, (c) rainfall event intensity, and (d) precipitation event intensity. Shading indicates positive trends

Fig. 12.

Trend in midwinter (Dec–Feb) (a) snowfall event intensity, (b) SWE event intensity, (c) rainfall event intensity, and (d) precipitation event intensity. Shading indicates positive trends

The translation of snowfall to SWE produces a trend in SWE event intensity, the pattern of which again shows an increase across the 40-yr period to the lee of each of the lakes with negative trends eastward (Fig. 12b). To the lee of the lakes, the pattern is similar for rainfall event intensity (Fig. 12c) and possibly indicates a lake enhancement of rainfall, with positive trends to the lee of each lake and negative trends inland from the lakes. The largest trends in rainfall event intensity occurred near the coastline of the Atlantic Ocean, possibly indicative of an increase in coastal storms that are synoptically conducive to lake-effect snowfall events due to cold air advection to the west of the midlatitude cyclone. The product of the changes in the intensity of rainfall events and the liquid equivalent from snowfall events was an increase in the overall intensity of precipitation events to the lee of Lakes Erie and Ontario (Fig. 12d).

All things considered, the increases in snowfall across the regions to the lee of Lakes Erie and Ontario seem to be the product of 1) increases in the frequency of snowfall at the expense of rainfall events, 2) an increase in the intensity of snowfall events, and 3) to some extent an increase in the snowfall/SWE ratio.

4. Conclusions

The specific goal of the work presented here was to determine if apparent increases in lake-effect snowfall within the past century acted to modify the winter hydroclimate of the susceptible areas of the North American Great Lakes. The alternative is that snowfall increases were simply a representation of a trend toward a greater fraction of the typical precipitation falling as snow rather than rain, with little change in the general hydroclimate.

Time series of annual midwinter snowfall and snowfall frequency at long-record stations associated with Lakes Erie, Michigan, Ontario, and Superior showed evidence of a rather monotonic increase in both snowfall and snowfall frequency through a 40-yr period beginning in the early 1930s and ending in the early 1970s. From the early 1970s through 2001, snowfall and snowfall frequency remained rather high at each of the stations but did not show a clear continuation of the upward trend. Over the 40-yr period of interest, the intensity of individual snowfall events increased to the lee of the two eastern lakes, but decreased to the lee of the two western lakes. Therefore, the snowfall increases seemed to be primarily a function of an increase in the frequency of snowfall events, with the aid of an increase in event intensity to the lee of the eastern Great Lakes and despite an event intensity decrease to the lee of the western lakes.

Spatial hydroclimatic analyses were confined to the eastern lakes of Erie and Ontario. The influence of the lakes was evident in the spatial pattern of 40-yr snowfall trends, as the greatest trends occurred to the lee of Lakes Erie and Ontario and in places exceeded 1.5 and 3 cm yr−1, respectively. Still, snowfall trends were positive throughout the region. Snow water equivalence (SWE) increased across the entire region as well, but again the greatest trends in SWE were associated with each of the lakes and higher elevations to the lee of the lakes. To the lee of Lakes Erie and Ontario the greatest trends in SWE were larger than 0.10 and 0.18 cm yr−1, respectively.

Opposite to the trends in snowfall, trends in midwinter rainfall were negative across the entire region, as large as −0.16 cm yr−1 in some areas. Across the lake-effect areas, trends ranged between approximately −0.04 and −0.12 cm yr−1. Trends in total precipitation indicated that the increase in SWE outweighed the decrease in rainfall across a significant portion of the area to the lee of Lake Ontario, but only within a small area to the lee of Lake Erie. Across the majority of the remainder of the region the trend in total precipitation in midwinter across the 40-yr period was negative, or dominated by the decrease in rainfall. This suggests that the overall wintertime hydroclimate of the lake-effect areas changed by an amount much less than the snowfall trends would suggest and that the significant increase in snowfall may have been more related to a shift toward more precipitation events that were snowfall rather than rainfall.

Trends in the percentage of precipitation that was recorded as SWE support the idea of an increase in the source of the precipitation within the region being snowfall. The percentage of midwinter precipitation falling as snow increased across the entire region over the 40-yr period, especially so to the lee of Lakes Erie and Ontario, with values as high as 0.6% per year. This equates to a 40-yr change as great as 24% (or nearly a 50% difference between the fraction of precipitation that is snowfall and the fraction that is rainfall). Potentially adding to the dramatic snowfall increase was a modest increase in the midwinter snowfall/SWE ratio. Trends ranged from 0.1 per year to the southeast of Lake Erie and southward into the higher elevations to as great as 0.3 per year to the south of Lake Ontario. This equals 40-yr increases in the snowfall/SWE ratio by amounts of 4–12:1, meaning that the depth of snow became ever greater than the amount of liquid precipitation that it contained.

Across the entire region east of Lakes Erie and Ontario the frequency of snowfall in midwinter increased over the 40-yr period. The greatest trends ranged from 0.3 to 0.4 days yr−1. Consistent with a decrease in rainfall across the entire region was a decrease in the frequency of rainfall such that, cumulatively, the frequency of midwinter precipitation increased over the 40-yr period only in two areas inland from each of the two lakes. This provides further evidence that possibly no great change in the ultimate hydroclimate of the region occurred. With an increase in the frequency of snowfall over the entire region and a general decrease in the frequency of rainfall, the trend in the percent of precipitation events that were snowfall increased markedly. Finally, the intensity of snowfall events increased to the lee of both Lakes Erie and Ontario over the course of the 40-yr period, by as much as 0.04–0.06 mm per event per year. Away from the lakes to the east of the region, the intensity of snowfall events showed very little change and even decreased across much of the area.

As a whole, the midwinter hydroclimatic analysis across the region to the lee of Lakes Erie and Ontario for the period of significant snowfall increases suggests that the increases were very much a product of 1) a shift toward more precipitation events that were snowfall rather than rainfall, 2) an increase in the intensity of individual snowfall events, and 3) an increase in the snowfall/SWE ratio. This suggests a greater frequency of cold air advection across the lakes (addressing points 1 and 3 above), as indicated by Leathers and Ellis (1996), and possibly colder air masses advecting over the lakes (addressing points 2 and 3 above). More comprehensive analysis of lower atmospheric (850 mb) air temperature in combination with lake surface temperature would be ideal in shedding additional light on the cause(s) of the snowfall increases. However, daunting data limitations are associated with each, making a concrete explanation difficult at best.

Whatever the cause, the results of the work presented here suggest that a change in the true wintertime hydroclimate of the lake-effect snowfall areas to the lee of the eastern Great Lakes did not occur on the scale of the snowfall increases during the past century. Rather, much of the snowfall increase came at the expense of rainfall events to produce only small changes in total precipitation.

While the overall regional hydrological regime appears to have changed little during the time period of the lake-effect snowfall increases, there may have been hydroclimatic consequences to specific locations. Additionally, the intraseasonal hydroclimatology of the region may have changed significantly in that increased snowfall at the expense of rainfall likely meant a deeper and/or more persistent snowpack that delayed runoff that would have been more immediate with rainfall. This is particularly a point for further research that may add further validity to the trend of increased snowfall at the expense of rainfall to the lee of the eastern Great Lakes within the past century.

Acknowledgments

This work was partially funded by a National Science Foundation Graduate Research Fellowship.

REFERENCES

REFERENCES
Angel
,
J. R.
, and
S. A.
Isard
,
1997
:
An observational study of the influence of the Great Lakes on the speed and intensity of passing cyclones.
Mon. Wea. Rev
,
125
,
2228
2237
.
Angel
,
J. R.
, and
S. A.
Isard
,
1998
:
The frequency and intensity of Great Lakes cyclones.
J. Climate
,
11
,
61
71
.
Baker
,
D. G.
,
1976
:
The mesoscale temperature and dew point fields of a very cold airflow across the Great Lakes.
Mon. Wea. Rev
,
104
,
860
867
.
Braham
,
R. R.
, and
R. D.
Kelly
,
1982
:
Lake-effect snowstorms on Lake Michigan, USA.
Cloud Dynamics, E. M. Agee and T. Asai, Eds., D. Reidel, 87–101
.
Braham
,
R. R.
, and
M. J.
Dungey
,
1984
:
Quantitative estimates of the effect of Lake Michigan on snowfall.
J. Climate Appl. Meteor
,
23
,
940
949
.
Chagnon
Jr.,
S. A.
, and
D. M. A.
Jones
,
1972
:
Review of the influences of the Great Lakes on weather.
Water Resour. Res
,
8
,
360
371
.
Dewey
,
K. F.
,
1977
:
Lake-effect snowstorms and the record breaking 1976–77 snowfall to the lee of Lakes Erie and Ontario.
Weatherwise
,
30
,
228
231
.
Dewey
,
K. F.
,
1979
:
Lake Erie induced mesosystems—An operational forecast model.
Mon. Wea. Rev
,
107
,
421
425
.
Eichenlaub
,
V. L.
,
1970
:
Lake-effect snowfall to the lee of the Great Lakes: Its role in Michigan.
Bull. Amer. Meteor. Soc
,
51
,
403
451
.
Eichenlaub
,
V. L.
,
1979
:
Weather and Climate of the Great Lakes.
The University of Notre Dame Press, 335 pp
.
Ellis
,
A. W.
, and
D. J.
Leathers
,
1996
:
A synoptic climatological approach to the analysis of lake-effect snowfall: Potential forecasting applications.
Wea. Forecasting
,
11
,
216
229
.
Harrington
,
J. A.
,
R. S.
Cerveny
, and
K. F.
Dewey
,
1987
:
A climatology of mean monthly snowfall for the conterminous United States: Temporal and spatial patterns.
J. Climate Appl. Meteor
,
26
,
897
912
.
Holroyd
,
E. W.
,
1971
:
Lake-effect cloud bands as seen from weather satellites.
J. Atmos. Sci
,
28
,
1165
1170
.
Kristovich
,
J. R.
, and
R. R.
Braham
,
1998
:
Mean profiles of moisture fluxes in snow-filled boundary layers.
Bound.-Layer Meteor
,
87
,
195
215
.
Leathers
,
D. J.
, and
A. W.
Ellis
,
1996
:
Synoptic mechanisms associated with snowfall increases to the lee of Lakes Erie and Ontario.
Int. J. Climatol
,
16
,
1117
1135
.
Leathers
,
D. J.
,
T. L.
Mote
,
K. C.
Kuivinen
,
S.
McFeeters
, and
D. R.
Kluck
,
1993
:
Temporal characteristics of USA snowfall 1945–46 through to 1984–85.
Int. J. Climatol
,
13
,
65
76
.
McMillan
,
M. C.
, and
D. F.
McGinnis
,
1975
:
Snowfall patterns near the Great Lakes.
Mon. Wea. Rev
,
103
,
837
838
.
Namias
,
J.
,
1960
:
Snowfall over the eastern United States: Factors leading to its monthly and seasonal variations.
Weatherwise
,
13
,
238
247
.
Niziol
,
T. A.
,
W. R.
Snyder
, and
J. S.
Waldstreicher
,
1995
:
Winter weather forecasting throughout the eastern United States. Part IV: Lake-effect snow.
Wea. Forecasting
,
10
,
61
77
.
Norton
,
D. C.
, and
S. J.
Bolsenga
,
1993
:
Spatiotemporal trends in lake-effect and continental snowfall in the Laurentian Great Lakes, 1951–1980.
J. Climate
,
6
,
1943
1956
.
Stooksbury
,
D. E.
,
C. D.
Idso
, and
K. G.
Hubbard
,
1999
:
The effects of data gaps on the calculated monthly mean maximum and minimum temperatures in the continental United States: A spatial and temporal study.
J. Climate
,
12
,
1524
1533
.
Thomas
,
M. K.
,
1964
:
A survey of Great Lakes snowfall,.
Proceedings of the 7th Conference on Great Lakes Research, Publ. 11, Great Lakes Research Division, University of Michigan, 294–310
.

Footnotes

Corresponding author address: Dr. Andrew W. Ellis, Department of Geography, Arizona State University, Box 870104, Tempe, AZ 85287-0104. Email: andrew.w.ellis@asu.edu