• Andrey, J., Mills B. , Leahy M. , and Suggett J. , 2003: Weather as a chronic hazard for road transportation in Canadian cities. Nat. Hazards, 28, 319343.

    • Search Google Scholar
    • Export Citation
  • Call, D. A., 2005: Rethinking snowstorms as snow events: A regional case study from upstate New York. Bull. Amer. Meteor. Soc., 86, 17831793.

    • Search Google Scholar
    • Export Citation
  • Call, D. A., 2010: Changes in ice storm impacts over time: 1886–2000. Wea. Climate Soc., 2, 2335.

  • Changnon, S. A., 1996: Effects of summer precipitation on urban transportation. Climatic Change, 32, 481494.

  • Datla, S., and Sharma S. , 2008: Impact of cold and snow on temporal and spatial variations of highway traffic volumes. J. Transp. Geogr., 16, 358372.

    • Search Google Scholar
    • Export Citation
  • Eisenberg, D., 2004: The mixed effects of precipitation on traffic crashes. Accid. Anal. Prev., 36, 637647.

  • Hall, F. L., and Barrow D. , 1988: Effects of weather and the relationship between flow and occupancy on freeways. Traffic Flow Theory and Highway Capacity, Transportation Research Record 1194, Transportation Research Board, 55–63.

    • Search Google Scholar
    • Export Citation
  • Hanbali, R. M., and Kuemmel D. A. , 1993: Traffic volume reductions due to winter storm conditions. Snow Removal and Ice Control Technology, Transportation Research Record 1387, Transportation Research Board, 159–164.

    • Search Google Scholar
    • Export Citation
  • Hassan, Y. A., and Barker D. J. , 1999: The impact of unseasonable or extreme weather on traffic activity within Lothian region, Scotland. J. Transp. Geogr., 7, 209213.

    • Search Google Scholar
    • Export Citation
  • Keay, K., and Simmonds I. , 2005: The association of rainfall and other weather variables with road traffic volume in Melbourne, Australia. Accid. Anal. Prev., 37, 109124.

    • Search Google Scholar
    • Export Citation
  • Knapp, K. K., and Smithson L. D. , 2000: Winter storm event volume impact analysis using multiple-source archived monitoring data. Winter Maintenance Innovations: Vehicle Rental Rates, Transportation Research Record 1700, Transportation Research Board, 10–16.

    • Search Google Scholar
    • Export Citation
  • Schmidlin, T. W., 1993: Impacts of severe winter weather during December 1989 in the Lake Erie snowbelt. J. Climate, 6, 759767.

  • View in gallery
    Fig. 1.

    Map showing the locations of the toll barriers and the sites with weather data.

  • View in gallery
    Fig. 2.

    Scatterplot of daily traffic counts and snow amounts for the Ripley toll barrier. Scatterplots for the other barriers were similar in appearance.

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 252 118 7
PDF Downloads 198 98 5

The Effect of Snow on Traffic Counts in Western New York State

David A. CallDepartment of Geography, Ball State University, Muncie, Indiana

Search for other papers by David A. Call in
Current site
Google Scholar
PubMed
Close
Full access

Abstract

While most people know anecdotally that weather affects traffic, relatively little research has examined the correlation between snow and traffic in great detail. Most studies have also been difficult to generalize for other areas and regions where drivers may be accustomed to more (or less) snow.

This study examines the relationship between snow and traffic volumes in western New York State, an area that is regularly inundated by snow (more than 225 cm most seasons). Total daily traffic counts for the New York State Thruway (Interstate 90) showed a moderate negative correlation to snow for the period of study (2003–10). However, this correlation is caused by the large number of passenger cars and other similar vehicles on the road. Most other vehicle types, such as tractor trailers, had no correlation whatsoever. Additionally, the results for all vehicle classes were similar for both suburban and rural areas. Finally, it was observed that the ratio of large to small vehicles increases during snow events.

Corresponding author address: David A. Call, Department of Geography, Ball State University, 2000 W. University Ave., Muncie, IN 47306. E-mail: dacall@bsu.edu

Abstract

While most people know anecdotally that weather affects traffic, relatively little research has examined the correlation between snow and traffic in great detail. Most studies have also been difficult to generalize for other areas and regions where drivers may be accustomed to more (or less) snow.

This study examines the relationship between snow and traffic volumes in western New York State, an area that is regularly inundated by snow (more than 225 cm most seasons). Total daily traffic counts for the New York State Thruway (Interstate 90) showed a moderate negative correlation to snow for the period of study (2003–10). However, this correlation is caused by the large number of passenger cars and other similar vehicles on the road. Most other vehicle types, such as tractor trailers, had no correlation whatsoever. Additionally, the results for all vehicle classes were similar for both suburban and rural areas. Finally, it was observed that the ratio of large to small vehicles increases during snow events.

Corresponding author address: David A. Call, Department of Geography, Ball State University, 2000 W. University Ave., Muncie, IN 47306. E-mail: dacall@bsu.edu
Keywords: Snow

1. Introduction

Anyone who has ever driven a car in winter weather is aware that snow and ice affect traffic volume. Prior scholarship has established this relationship, but most studies have been for small, limited areas, and there are numerous gaps. For example, researchers have not examined the relationship between snowfall and vehicle type. This study explores the relationships between snowfall and daily traffic counts in a very snowy area: western New York State. Traffic counts were examined for three toll barriers on the New York State Thruway (Interstate 90) where all vehicles must pay a distance-based toll. Two of these barriers are in suburbs adjacent to Buffalo, New York, while the third is located approximately 120 km southwest of Buffalo in a rural area just east of the Pennsylvania state line (see Fig. 1). The goals of this study are to see if the prior established relationships between snow and traffic apply in an area regularly inundated by snow, to see whether or not snow affects different types of vehicles in a similar manner, and to see if snow affects a rural area differently than a highly suburbanized one. The primary finding is that snow does affect traffic counts in western New York, but the effects vary greatly by vehicle type.

Fig. 1.
Fig. 1.

Map showing the locations of the toll barriers and the sites with weather data.

Citation: Weather, Climate, and Society 3, 2; 10.1175/WCAS-D-10-05008.1

2. Literature review

Many researchers have asked questions about how weather affects traffic both in North America (Hall and Barrow 1988; Changnon 1996; Eisenberg 2004) and overseas (Hassan and Barker 1999; Keay and Simmonds 2005). One reason for such extensive research is that the effects of weather can vary widely within relatively small areas such as U.S. states (Knapp and Smithson 2000) or between similar-size cities in the same region (Andrey et al. 2003). This can make it challenging to apply the results from one area to another.

Studies examining the relationship between winter weather and traffic occupy a portion of the literature on weather and traffic. Hanbali and Kuemmel (1993) found reduced traffic volumes in connection with snowstorms, while Call (2010) noted similar effects with major ice storms. Knapp and Smithson (2000) used a regression analysis to observe that winter storm events with at least a 4-h duration and 2-cm total accumulation caused traffic reductions in Iowa, but also found that such reductions varied widely from as little as 16% to as much as 47%. Schmidlin (1993), however, noted only a minor reduction in traffic when examining traffic data on a monthly time scale. More recently, Datla and Sharma (2008) correlated cold and snowfall with traffic volumes in Alberta, Canada, and found that the effects varied with road type (e.g., local roads versus recreational roads).

3. Methodology

Eight years of traffic count data (2003–10) for the month of January were obtained from the New York State Thruway Authority for each barrier. (The author requested additional years of data, but these either do not exist or were not provided.) January is typically the snowiest month for Buffalo, with a monthly average of 66 cm for 1971 through 2000. December, nearly as snowy, was not included because of distortions caused by the Christmas holiday season and National Football League (NFL) games at a proximate stadium.

The data were highly detailed and included hourly traffic counts. These were then aggregated to daily counts for each vehicle class and then summarized into total daily counts for each barrier. Data analyses were affected by systemwide changes that occurred during the study period. In May 2005, tolls increased significantly (typically 25%) and vehicle classes were simplified. The old system used 10 perplexing categories, while the new system classified vehicles based solely on the number of axles (including axles on trailers) and height (under or over 7 feet 6 inches; 228.6 cm). Thus, daily counts for each vehicle class were separately analyzed for the 2003–05 and 2006–10 time periods.

Daily temperature and snowfall data were obtained from the National Climatic Data Center (NCDC) for first-order stations in Buffalo, New York, and Erie, Pennsylvania, and from cooperative stations located in Dunkirk (temperature only), Fredonia (snowfall only), and Westfield (snowfall only); all three of those stations are located in New York. Snow in this area is primary lake effect and varies greatly with elevation and distance. Stations south of Interstate 90 (e.g., Jamestown, New York) were at much higher elevations and too far away to be included.

Williamsville (exit 50) is located 3 km from the Buffalo observing site, so the Buffalo data were used for this location without modification. Lackawanna (exit 55) is 15 km from the Buffalo observing site and 60 km from Dunkirk–Fredonia. A simple distance-weighted average was used to estimate the temperature and snowfall data for this location. To test the validity of this approach, the data analysis was also performed with the Buffalo data only and differences were minimal. Ripley (exit 61) is located 13 km from Westfield, 40 km from Dunkirk–Fredonia, and 67 km from Erie. Unfortunately, Westfield data were available for 2003 only. The Westfield and Erie data were averaged for this year, while for the other years the data from Dunkirk–Fredonia and Erie were averaged. Similar to Lackawanna, a distance-weighted average was used and compared to other approaches using single stations, and the results did not vary from the weighted-average results by more than a few hundredths. Data from the first-order stations was for the midnight-to-midnight period, while data from the cooperative stations was for a slightly different “day” (e.g., 9 p.m. to 9 p.m.). Similar tests involving various combinations of stations were run and it appeared that these time variations also had a negligible effect.

The author also assumed that the roads were maintained in a consistent manner throughout the course of this study and ignored any changes to them from road construction projects, such as widening or repaving. It is, however, unlikely that road construction projects were started, stopped, or active during the month studied (January).

4. Results

Snowfall amounts and total daily traffic counts were negatively correlated at all three barriers as shown in Table 1. Correlations ranged from −0.25 to −0.32 and all were significant at the α = 0.01 level. A positive correlation between mean temperature and total daily traffic counts was also observed (refer again to Table 1). This correlation was consistent across locations but much smaller in magnitude (0.11 to 0.15) and in all three cases was significant at the α = 0.05 level but not the α = 0.01 level. The temperature correlation results were similar to those found in other studies (e.g., Datla and Sharma 2008). A scatterplot illustrating the relationship between snowfall amounts and daily traffic counts is shown in Fig. 2. This plot is for Ripley only, but it is representative of the plots generated for the other sites as well. The plot also shows insufficient heavy snowfall events to conclusively establish the nature of the snow–traffic relationship (e.g., linear) or whether there are “thresholds” where snow changes from being a nuisance to a serious disruption.

Table 1.

Correlations between mean number of vehicles per day, mean temperature, and snow. Boldface indicates that a correlation was significant at the 0.01 level and an asterisk indicates that a correlation was significant at the 0.05 level.

Table 1.
Fig. 2.
Fig. 2.

Scatterplot of daily traffic counts and snow amounts for the Ripley toll barrier. Scatterplots for the other barriers were similar in appearance.

Citation: Weather, Climate, and Society 3, 2; 10.1175/WCAS-D-10-05008.1

When daily traffic counts by class were correlated with snowfall, results varied considerably between classes but were similar for all three locations. As shown in Table 2, relatively large (−0.3 to −0.4) negative correlations were observed between snowfall and class 1 (2003–05) and class 2L (2006–10) vehicles, which include most passenger cars, sport utility vehicles, pickup trucks, and the like. These vehicles constitute an overwhelming majority at Williamsville and Lackawanna, accounting for more than double the number of all other vehicles combined. At the rural Ripley barrier, these are also the most common type but constitute a much smaller proportion of the total daily vehicle count.

Table 2.

Correlations between snow and vehicle counts for traffic classes with more than 1000 vehicles passing through per day. Boldface indicates that a correlation was significant at the 0.01 level, an asterisk indicates that a correlation was significant at the 0.05 level, and an ampersand indicates a correlation that was significant at the 0.10 level.

Table 2.

The second most common group of vehicles is five-axle tractor trailers. The majority of these vehicles fell into classes 3 or 5 under the 2003–05 classification system and classes 5H and 5S under the 2006–10 system. Regardless of how tractor trailers are classified, all correlations between snowfall and traffic count are close to 0 and almost none are statistically significant. In exception, a marginally significant (p = 0.09) and very small negative correlation (−0.14) was observed during the 2006–10 period at Williamsville for 5S vehicles. This may be a spurious result, or it may have resulted from a snowstorm on 30 January 2008, in which authorities closed the Thruway for a 55-mile stretch just east of the Williamsville Barrier (exits 46–49). This data point was flagged as an outlier in preliminary data analysis.

In May 2005, class structures changed and toll rates increased considerably. While automobile (class 1/2L) traffic counts did not show much change, there was notable reduction in tractor trailer traffic following the increase, especially at Ripley, where the decline was greater than 40%. Nonetheless, the correlations between snow and vehicle class remained similar following the class and rate structure changes.

Compared to the passenger car–truck and tractor trailer classes, there are few vehicles in the other vehicle classes. In most cases, the correlations are insignificant; the exceptions are shown in Table 3. For all three barriers, snow had significant negative correlations for 2006–10 classes 3L and 4L, which are generally cars or pickup trucks towing trailers; these vehicles account for fewer than 1% of the vehicles on a typical day. At Williamsville, two additional small and marginally significant correlations were observed. One was for two-axle high (2H) vehicles, typically referred to as “box trucks.” Another weak correlation was observed for four-axle high (4H) vehicles at this barrier. At exit 61, a positive correlation was observed between snow and three-axle high (3H) vehicles. An average of 182 of these vehicles pass through the barrier each day. Because this number of vehicles is relatively small, it is unclear if this is a real correlation or a data artifact. Several other vehicle classes also had positive correlations at Ripley, but these classes (8, 9, and 7S) all comprise very few vehicles and may also be data artifacts.

Table 3.

Other significant results for all exits. See Table 1 for symbol legend.

Table 3.

5. Conclusions

In summary, total daily vehicle counts and daily vehicle counts by class were correlated with daily snowfall amounts at three toll barriers in western New York State. Total daily vehicle counts had statistically significant large (~0.30) negative correlations with snowfall and smaller (~0.15) positive correlations with mean temperature. Snowfall had dramatically different effects on the various vehicle classes. Snowfall greatly diminished passenger car and similar traffic but had almost no effect on tractor trailer vehicle counts. Except for cars towing trailers, most other types of vehicles also had minimal correlations. These results occurred under both classification systems and before and after the toll increases of 2006. There was little variation in the results from barrier to barrier even though many fewer vehicles pass through the rural Ripley barrier on a daily basis.

Tractor trailers are driven long distances by professionals under tight deadlines, and these drivers are less likely to postpone travel because of inclement weather. Thus, it seems logical that no correlation was found between snowfall and tractor trailer traffic. The drop in the number of passenger vehicles could result from people choosing to stay home during inclement weather, or it could result from school and business closings, or perhaps a combination of both factors. Of perhaps greater interest is that this decrease was observed at both the suburban and rural locations. The Ripley barrier is more than 40 km from any sizeable (>10 000 person) city and near a state boundary. It seems unlikely that most vehicles here are commuters given the low population density and long distances between nearby Thruway exits (typically 16 km). Perhaps long-distance travelers are modifying their travel plans because of snow. Accepting this hypothesis raises an entirely new set of research questions, such as “where do these travelers go instead?” and “how much long-distance travel is essential as opposed to optional?” Or perhaps there are a significant number of long-distance commuters traveling more than 50 miles each day. At this time, it is not clear why passenger car traffic shows a strong correlation to snowfall at Ripley.

Another notable finding is that the ratio of “high” to “low” vehicles on the Thruway increases greatly during a snow event. This could increase the risk of death or injury for occupants of passenger cars and similar vehicles as multiple vehicle accidents are more likely to involve large vehicles. To examine this speculation further, researchers would need to analyze accident and fatality data for the Thruway to see if these values change when it is snowy.

Several reviewers pointed out that the limited snow data used may have affected the results of this study. Indeed, narrow bands of intense snow (often found with lake effect) may have occurred between stations and affected traffic counts. Incorporating “between-station” snowfall would probably improve correlation strength, but it would require a thorough analysis of hourly data. A related limitation is the use of daily snowfall amounts instead of hourly amounts. The use of daily data does not discriminate light, long-duration snowfalls from intense, brief snowfalls (“snowbursts”). For Buffalo and other cities in upstate New York, Call (2005) suggested that such snowbursts are often more disruptive than light, long-duration events. Thus, a natural follow up to this project would examine the relationship between hourly traffic counts and hourly snowfall. Although hourly snowfall data may be difficult to obtain, radar and/or visibility data could possibly be used as a proxy.

Finally, future researchers could examine additional related questions, such as whether or not these correlations vary with the day of the week, if they occur elsewhere along the New York State Thruway (such as Albany, New York, in the eastern part of the state), or if they occur at exits where people can transfer onto local roadways as opposed to systemwide barriers. Revenue figures could also be analyzed to estimate the economic losses associated with snowfalls. This would help with quantifying the effects of snowfall—a useful extension of this research.

REFERENCES

  • Andrey, J., Mills B. , Leahy M. , and Suggett J. , 2003: Weather as a chronic hazard for road transportation in Canadian cities. Nat. Hazards, 28, 319343.

    • Search Google Scholar
    • Export Citation
  • Call, D. A., 2005: Rethinking snowstorms as snow events: A regional case study from upstate New York. Bull. Amer. Meteor. Soc., 86, 17831793.

    • Search Google Scholar
    • Export Citation
  • Call, D. A., 2010: Changes in ice storm impacts over time: 1886–2000. Wea. Climate Soc., 2, 2335.

  • Changnon, S. A., 1996: Effects of summer precipitation on urban transportation. Climatic Change, 32, 481494.

  • Datla, S., and Sharma S. , 2008: Impact of cold and snow on temporal and spatial variations of highway traffic volumes. J. Transp. Geogr., 16, 358372.

    • Search Google Scholar
    • Export Citation
  • Eisenberg, D., 2004: The mixed effects of precipitation on traffic crashes. Accid. Anal. Prev., 36, 637647.

  • Hall, F. L., and Barrow D. , 1988: Effects of weather and the relationship between flow and occupancy on freeways. Traffic Flow Theory and Highway Capacity, Transportation Research Record 1194, Transportation Research Board, 55–63.

    • Search Google Scholar
    • Export Citation
  • Hanbali, R. M., and Kuemmel D. A. , 1993: Traffic volume reductions due to winter storm conditions. Snow Removal and Ice Control Technology, Transportation Research Record 1387, Transportation Research Board, 159–164.

    • Search Google Scholar
    • Export Citation
  • Hassan, Y. A., and Barker D. J. , 1999: The impact of unseasonable or extreme weather on traffic activity within Lothian region, Scotland. J. Transp. Geogr., 7, 209213.

    • Search Google Scholar
    • Export Citation
  • Keay, K., and Simmonds I. , 2005: The association of rainfall and other weather variables with road traffic volume in Melbourne, Australia. Accid. Anal. Prev., 37, 109124.

    • Search Google Scholar
    • Export Citation
  • Knapp, K. K., and Smithson L. D. , 2000: Winter storm event volume impact analysis using multiple-source archived monitoring data. Winter Maintenance Innovations: Vehicle Rental Rates, Transportation Research Record 1700, Transportation Research Board, 10–16.

    • Search Google Scholar
    • Export Citation
  • Schmidlin, T. W., 1993: Impacts of severe winter weather during December 1989 in the Lake Erie snowbelt. J. Climate, 6, 759767.

Save