Empirical Studies on Traffic Flow in Inclement Weather

2.0 Literature Review

This section synthesizes the research that has been conducted on the impacts of weather on traffic performance. The literature on research into this area can be divided into three general categories, although it is important to note that there are overlaps between them:

• Secondary Data – Secondary data studies use archived data collected by parties other than the researchers. This information is generally collected for other purposes than the objective of the research. Many research efforts have collected data from traffic count databases and weather stations in an attempt to identify the impact of adverse weather on traffic flow. Early efforts go back to the 1960s and 1970s when the University of Michigan used an early ITS system to collect traffic and speed data during different weather conditions on Detroit area freeways. Some efforts have looked only at speed impacts, while others have attempted to estimate changes in capacity and speed-density relationships. An increasing number of these research efforts are using microsimulation models to add value to the data. These models permit the analyst to fill in data gaps and test different assumptions regarding driver behavior. Much of the work to date has used tools such as CORSIM but increasingly sophisticated models are allowing researchers to disaggregate the traffic stream based on driver response. Different driver profiles can be developed based on their reactions to adverse weather (aggressive, cautious, moderate) and their interactions can be modeled. Some of the research projects identified in this category are listed in Appendix A.
• Primary Data Collection – Primary data studies use aggregate data collected by researchers specifically for the study. A number of researchers have been able to collect primary data on traffic speed and volumes during adverse weather. In some cases, data have been collected as part of traffic flow studies, while in other cases they have been collected to test driver response to traffic management system devices or other ITS technologies. Many of these studies have been conducted in Northern Europe, where both driving habits and roadway configurations will probably differ from those in the U.S. Increasingly there is recognition that primary, focused data collection efforts will provide a clearer picture of traffic flow impacts. This approach is still limited, however, as is secondary data collection, in its ability to disaggregate the market and identify how different classes of drivers will respond. A sampling of these research efforts is presented in Appendix A.
• Human Factors Studies – Human factors studies generally involve primary data collection, focused on individual reactions to driving conditions. There is increasing focus on the behavior of individual drivers or groups of drivers based on the theory that gender, age, region of residence, and other factors impact driver behavior. Studying the response of individual drivers to adverse weather has a number of advantages. Field observations of decreased speeds and greater headways can be better understood by observing how individual drivers respond to certain stimuli, and how specific categories of drivers respond. This information is particularly helpful in light of the growing deployment of ITS technologies. Motorist warning systems that notify drivers through roadside Dynamic Message Signs (DMS) about weather-related conditions such as fog, flooding, or slippery pavement are deployed by some transportation agencies. The benefits of these systems can be maximized if agencies can optimize the location and provide the most effective warning on the sign itself. Human factors research is an important element in the analysis and design of these systems. A summary of these research efforts is presented later in this chapter.

The remainder of this chapter summarizes the results found in many of these studies. A discussion of macroscopic effects of weather on capacity, delay, volume, and speed is followed by a summary of much less understood microscopic behavioral and human factors impacts. Microscopic analysis looks at individual vehicle movements and their interactions while human factors research investigates driver response, which may or may not be manifested in vehicle response.

2.1 Operations and Macroscopic Traffic Research

This section describes the impact of weather on the relationships between traffic speed, flow, and density, and other macroscopic measurements. Following are key freeway traffic parameters that are impacted by adverse weather:

• Capacity – Capacity is the maximum number of vehicles that can pass a point during a specified time period and is a characteristic of the roadway.
• Delay – Delay is the time lost by a vehicle due to causes (i.e., adverse weather) beyond the control of the driver.
• Traffic Volume/Demand – Traffic volume is the number of vehicles passing a point during a specified time period or the number of vehicles desiring to pass that point during a specified period. Flow is traffic volume normalized to an hour, or vehicles per hour. Demand is the number of vehicles that desire to travel past a point during a specified period.
• Speed – Freeway speed is the average speed of vehicles in the absence of traffic control devices. Note that unless otherwise specified, speeds are averaged over time and over a number of traffic conditions. Free-flow speed refers to the average speed of vehicles under low-volume conditions.

Capacity

The Highway Capacity Manual (2000) asserts that adverse weather can significantly reduce capacity and operating speeds, and addresses the issues of when and how to take these effects into account. The manual references several studies in its discussion of weather effects. Lamm, Choueiri, and Mailaender (1990) conclude that speeds are not affected by wet pavement until visibility is affected, which suggests that light rain does not impact operating speeds, but heavy rain does and can be expected to have a noticeable effect on traffic flow.

Similarly, Ibrahim and Hall (1994) found minimal reductions in maximum observed flows and operating speeds (detailed in the discussion on speed) in light rain, but significant reductions in heavy rain. There is a 14 to 15 percent capacity reduction during heavy rain, compared to clear and dry conditions. The HCM does not define the intensity ranges associated with light and heavy rain and light and heavy snow classifications.

Similarly to rain, light snow was found to have minimal effects, while heavy snow was found to have a potentially large impact on capacity and operating speeds in the Ibrahim and Hall study. Light snow resulted in a five to 10 percent reduction in maximum observed flows (midway between the effects of light and heavy rain). Heavy snow resulted in a 30 percent drop in capacity. Snow accumulation obscures lane marking, which can cause drivers to seek greater lateral clearance in addition to longer headways.

The HCM notes that no studies have quantified the effects of fog on capacity, but several European studies have examined the effectiveness of fog warning systems. These studies, however, do not report how speeds or capacities are affected by fog. A 1995 Brilon and Ponzlet study examined the effect of environmental conditions and variability in capacity at 15 Autobahn sites in Germany. While the results cannot be directly translated to North American conditions, the study extends the research results cited above and identifies other environmental conditions that impact capacity. Combinations of daylight, darkness, dry, wet, weekend, and weekday, as well and four- and six-lane configurations were examined as shown in Table 2.1.

Table 2.1 Reduction in Capacity from Daylight and Dry Conditions

Number of Lanes	Weekday/ Weekend	Dark and Dry	Daylight and Wet	Dark and Wet
Six-lane	Weekday	13%	12%	38%
Six-lane	Weekend	21%	27%	[no value]
Four-lane	Weekday	19%	18%	47%
Four-lane	Weekend	25%	29%	[no value]

Source: Brilon and Ponzlet, 1995.

Given that the winter peak-periods often occur in darkness, these capacity reductions are important to recognize and study further. These research findings can be incorporated with the HCM methodology by using the speed-flow curve to model the effects of adverse weather and evaluate the expected traffic performance for certain sections of uninterrupted flow facilities.

In a University of Virginia study, Smith, et al. (2004), investigated the impact of rainfall at varying levels of intensity on freeway capacity and operating speeds to gain an understanding of the impact of weather conditions on key traffic parameters. Traffic and weather data were collected for a one-year period between August 1999 and July 2000, on two freeway links in Hampton Roads, Virginia. Traffic data (volume, time mean speed, and occupancy) were collected from the Smart Travel Laboratory at two-minute intervals. Average speed and flow rate were compiled at 15-minute intervals. Hourly weather data were collected from the weather station at Norfolk International Airport, three miles from the study freeway segments. Rainfall data were collected in inches per hour, and intensity was assumed to be the same for every 15-minute interval over the course of an hour. Rainfall was classified into light rain (0.01 to 0.25 inch per hour/0.25 to 6.4 mm per hour) and heavy rain (greater than 0.25 inch per hour/6.4 mm per hour), based on guidelines provided by the Swedish Meteorological and Hydrological Institute and the Philippine Atmospheric Geophysical and Astronomical Services Administration. Records during periods of darkness were removed from the study.

Analysis of the data began with plotting speed-flow curves. A maximum observed throughput approach was used to estimate freeway link capacity; the mean of the highest five percent flow rates was used to determine the percent changes in capacity due to rainfall. Capacity reduction was evident and statistically significant, as rainfall intensity became greater. Light rain decreased freeway capacity by 4 to 10 percent, and heavy rain decreased capacity by 25 to 30 percent.

Prevedouros and Chang (2004) analyzed video surveillance data of freeway and arterial roadways in Honolulu recorded between 1996 and 2000, focusing on measurements from traffic platoons collected during busy but fluid conditions. Headways were measured at identical locations under dry, wet pavement (no rain), and light-to-moderate rain conditions. Analysis of the data finds that, on average, freeway capacity is reduced by 8.3 percent.

Table 2.2 Summary of Rain Effects on Capacity

Capacity Reduction	Capacity Reduction	Capacity Reduction	Capacity Reduction	Capacity Reduction
Researcher	Ibrahim and Hall	Brilon and Ponzlet	Smith	Prevedouros and Chang
Location	Toronto, Ontario	Germany	Hampton Roads, Virginia	Honolulu, Hawaii
Year	1994	1995	2004	2004
Light Rain	[no value]	12-47%	4-10%	8.3%
Heavy Rain	14-15%	12-47%	25-30%	8.3%

Delay

A 2003 Mitretek study (Stern, et al., 2003) attempted to quantify the amount of travel delay imposed upon drivers due to the effects of inclement weather. Using a metropolitan Washington D.C. network spanning from the eastern suburbs in Maryland to the western suburbs in Virginia and containing 33 roadway segments, researchers utilized travel time data for each weekday between December 1999 and May 2001. Traffic data were collected in five-minute increments between 6:30 a.m. and 6:30 p.m.

Researchers used two models to quantify the travel delay: a regression analysis using surface observations and an analysis by means of precipitation category using radar data. The regression analysis used weather data from Automated Surface Observing System (ASOS) stations at three International Airports in the Washington D.C. area. Air temperature, dew point, wind speed and direction, and rainfall accumulation were measured in hourly increments. In a two-step linear regression process, the travel time was regressed against weather variables for all ASOS sites. Final variables in the analysis were precipitation type and intensity, wind, visibility distance, and pavement condition (which was inferred from the other weather information). Then, the linear regression models were reduced for each segment to predict a base travel time and the increase due to weather.

Despite limitations in data (primarily, the absence of other variables affecting travel time and the temporal and spatial differences between weather and traffic data), the study found that, when weather phenomena occur, the average regional increase in travel time is 14 percent, with pavement condition the most frequent explanatory variable, followed by precipitation. Wind speed and surface visibility appeared in a few models.

Policy prevented the posting of travel times indicating speeds greater than the speed limit. This has tangible implications on the analysis of delay during congestion-free traffic periods, and may result in the study underestimating the effects of weather on travel times. Researchers recognized the need for an accurate travel time data source and recommended future studies focus on evaluating the relative impacts of weather and incident/congestion variables.

Traffic Volume

Adverse weather can reduce demand on the transportation system, as drivers postpone discretionary trips or activities get canceled. On the other hand, there may be increases in vehicular demand because some who travel by bicycle or on foot switch to motorized mode when adverse weather is forecast. Adverse weather can also have more complex effects on demand, such as peak-hour demand shifting as drivers leave early or late to avoid driving in dangerous conditions.

Hanbali and Kuemmel conducted a study in 1992 to measure reductions in traffic volumes during snowstorms on highways and freeways outside of urban areas in Illinois, Minnesota, New York, and Wisconsin. Automatic vehicle detectors collected traffic data during the first three months of 1991, including annual average daily traffic and actual 24-hour counts. Other data used in the study included highway characteristics, level of service (in terms of snow and ice removal), and road treatment to achieve bare pavement. Climate data, based on the National Climatic Data Center, included the storm period (start and end time and date), temperature range, snow depth, and type of snow.

Researchers measured the hourly traffic volumes during every snowstorm and compared it to the normal hourly traffic volumes corresponding to the same type of day/time/season. From that, hourly reduction factors were derived for each snowstorm, and compiled and correlated for each categorized group (shown in Table 2.3). Researchers concluded that volume reductions increased with total snowfall, but that the reductions were smaller during peak travel hours and on weekdays, likely due to the nondiscretionary nature of most weekday trips.

Table 2.3 Snowstorm Impacts on Volumes

Snowfall	Weekdays (Range of Volume Reduction)	Weekends (Range of Volume Reduction)
< 25 mm	7-17%	19-31%
25-75 mm	11-25%	30-41%
75-150 mm	18-34%	39-47%

Source: Hanbali and Kuemmel, 1992.

Knapp, et al. (2000) conducted a study to evaluate winter weather impacts on traffic volume and safety. Hourly traffic and weather information was collected and analyzed along interstates in Iowa for 1995, 1996, 1997, and 1998. Seven RWIS stations and nearby automatic vehicle detectors were used to approximate storm and nonstorm weather event parameters and traffic volumes.

The goal was to limit the research to relatively significant winter storm events, so event time periods were defined as those where an RWIS station recorded precipitation, air temperature below freezing, wet pavement surface, and a pavement temperature below freezing for at least four hours with an estimated snowfall of at least 5.1 mm per hour (0.2 inch per hour). The research compared and statistically analyzed volume and crash data from winter storm and nonstorm event time periods.

The traffic volume analysis covered 64 winter storm events (618 hours). There was large variability in winter storm traffic volume impacts, ranging from 16 percent to 47 percent reduction. The overall average reduction was approximately 29 percent, with a 95 percent confidence interval of 22.3 percent to 35.8 percent. A regression analysis indicated that percent volume reduction had a statistically significant relationship with total snowfall and the square of the maximum wind speed. Storm event duration, snowfall intensity, and minimum and maximum average wind speed either were correlated to the explanatory variables or were not found to be statistically significant. The coefficients indicated that volume reduction increases with each variable and the adjusted R-squared value indicates that the model had some explanatory power.

Ibrahim and Hall’s (1994) flow reduction results of 10 to 20 percent for heavy rain were consistent with a previous study by Jones and Goosby (1970). Little or no effect on flow was observed under light rain conditions. Light snow resulted in a 5 to 10 percent reduction in maximum observed flows (or midway between the effects of light and heavy rain).

Freeway Speed

A Federal Highway Administration (FHWA) study found that interstate speeds decrease in inclement weather. Table 2.4 describes the average percent speed reduction for a variety of weather conditions.

Table 2.4 Speed Reductions in Inclement Weather

Condition	Percent Speed Reduction
Dry	0%
Wet	0%
Wet and snowing	13%
Wet and slushy	22%
Slushy in wheel paths	30%
Snowy and sticking	35%
Snowing and packed	42%

Source: FHWA, 1977.

Ibrahim and Hall’s (1994) study discussed in the Highway Capacity Manual used traffic data from the freeway traffic management center for the Queen Elizabeth Way (QEW) in Mississauga, Ontario, with volume, occupancy, and speed data recorded in 30-second intervals during the months of October, November, and December 1990; and January and February 1991. Weather records for Pearson International Airport were obtained and it was confirmed that the records accurately reflected the weather conditions at the QEW freeway location.

The study site was selected so that it would not be influenced by ramp or weaving sections, and the time period from 10:00 a.m. to 4:00 p.m. was chosen to focus on uncongested data and eliminate periods of darkness, which literature had shown to impact driver behavior and operating speed.

Regression analyses were conducted on the clear weather data to select models for the uncongested flow-occupancy and speed-flow relationships. A quadratic model was the best fit for the flow-occupancy relationship, while a simple linear model with dummy variables was used for the speed-flow relationship.

The comparison analysis showed that both the difference in slope and the difference in intercept of the speed-flow function within the rainy (snowy) condition (i.e., difference between light and heavy conditions) were more important than the differences between clear and rainy (snowy) weather. Comparing rainy and snowy conditions allowed the researchers to conclude that, while light rain and light snow have nearly the same effect on traffic operations, heavy snow has a much greater impact than heavy rain.

In light rain, a 1.9 km/hr (1.2 mph) and 6.4 to 12.9 km/hr (4 to 8 mph) reduction in operating speeds can be expected during free-flow conditions and at a flow of 2,400 vehicles per hour, respectively. In heavy rain, a 4.8 to 6.4 km/hr (3 to 4 mph) and 12.9 to 16.0 km/hr (8 to 10 mph) reduction in speed can be expected. Light snow resulted in a significantly significant drop of 0.96 km/hr (0.6 mph) in free-flow speeds, while heavy snow resulted in a 37.0 to 41.8 km/hr (23 to 26 mph) (35 to 40 percent) free-flow speed reduction.

Ibrahim and Hall concluded that, while adverse weather affects both the flow-occupancy and speed-flow relationships, other factors, including the driver’s familiarity with rainy and snowy conditions, may affect these relationships. Regional differences are expected to be a factor in speed reductions. Additionally, facilities and capabilities to deal with adverse weather (i.e., quality of drainage or effective plowing operations) can also affect the magnitude of speed reduction and flows.

A study examining the importance of weather in performing a capacity or level of service analysis was conducted on a rural interstate in Idaho (Kyte, et al., 2001). Data were collected from the same location on a four-lane, level-grade freeway between 1996 and 2000, with high-truck volumes and flow rates almost always less than 500 passenger cars per hour per lane (pcphpl). Traffic data (time, speed, and length of vehicle), visibility distance, and weather data (wind speed and direction, air temperature, relative humidity, roadway surface condition, and type and amount of precipitation) were recorded in five-minute intervals.

In good weather, there was a nearly constant relationship between vehicle speed and flow rate, resulting in an estimated free-flow speed of 122 kilometers per hour (76 miles per hour) in ideal conditions (9 km/hr greater than the value computed using the HCM method). The study evaluated visibility reduction, wind, and pavement condition (wet or snow covered) factors. In this field study, for visibility levels greater than one kilometer (0.625 mi), speeds were nearly constant and near the ideal condition free-flow speed. As visibility decreased below one kilometer, speed decreased, with a substantial decline when visibility dropped below 0.3 kilometer (0.18-mile). Wind conditions did not have as clear a relationship, perhaps due to wind gust that are not reflected in average wind speed or variable driver responses to wind. Despite these limitations, data indicated that the critical wind speed was 24 kilometers per hour (14.9 miles per hour).

To determine the individual effects of weather variables on speed, speed was regressed against pavement condition, wind speed, and visibility, using the critical values described above. The results indicated that all coefficients were statistically significant, but with a high-degree of variability in the results as shown in Table 2.5.

Table 2.5 Impact of Environmental Conditions on Speed

Factor	Speed Reduction (km/h)	Speed Reduction (mph)
Wet	9.5	5.9
Snow	16.4	10.2
Wind > 24 km/h	11.7 (variation of speed drop is high)	7.3 (variation of speed drop is high)
Visibility < 0.28 km	0.77 per 0.01 km below critical	0.48 per 33 ft below critical

Source: Kyte, et al., 2001.

The results for wet pavement and snow correspond closely with other studies. Researchers concluded that, since these factors indicated different free-flow speeds and since one-third of cities experience rain at least 125 days per year, weather should be considered in capacity and level of service analysis.

Research in the University of Virginia study (Smith, et al., 2004) has shown that speed is relatively insensitive to increasing flow rates until congestion sets in. Therefore, the mean speeds were calculated for uncongested conditions for each station/weather combination, and percent changes in speeds based on rainfall intensity were tested for statistical significance.

Operating speed reductions were not quite as dramatic as the capacity reductions discussed previously. Results indicated that the presence of rain is a more important factor than the difference in intensity, and decreased operating speeds by three to five percent. The difference between speed reductions during light and heavy rain is not statistically significant. This conclusion contradicts the results of the studies performed by Lamm, Choueiri, and Mailaender; and Ibrahim and Hall utilized in the HCM. Perhaps the low speed reduction is a result of an upper limit on free-flow speed not exceeding the speed limit. This data restriction may result in underestimating the effects of rain on operating speed. Researchers find that the study indicates rainfall has a greater impact on capacity than is currently presented in the HCM, and that the impact of heavy rainfall may be overstated.

Summary of Existing Literature on Weather Impacts on Macroscopic Traffic Parameters

Low Visibility

There is very limited research on the impact of low visibility on traffic flow. The Brilon and Ponzlet study in Germany indicated a 13 to 47 percent reduction in capacity in darkness, relative to daylight conditions. Additionally, the Kyte study found a 0.77 kilometer per hour (0.48-mile per hour) reduction in speed for every 0.01 kilometer (0.0062-mile) below the critical visibility of 0.3 kilometer (0.18-mile). There is anecdotal evidence that, in reduced visibility, drivers unconsciously increase their speed as they acclimate to foggy conditions. In a laboratory simulation, the more foggy environmental conditions were, the more drivers underestimated their speed (Snowden, et al., 1998).

Rain

There have been a number of studies investigating the impact of rain on traffic flow, particularly freeway speed and capacity. Table 2.6 summarizes the results of the studies.

Table 2.6 Summary of Rain Effects on Speed

Speed Reduction	Speed Reduction	Speed Reduction	Speed Reduction
Researcher	Ibrahim and Hall	Kyte	Smith
Location	Toronto, Ontario	Idaho	Hampton Roads, Virginia
Year	1994	2001	Smith
Light Rain	1.9-12.9 km/h (1.2-8 mph)	9.5 km/h (15.3 mph)	3-5%
Heavy Rain	4.8-16.1 km/h (3-10 mph)	9.5 km/h (15.3 mph)	3-5%

The Mitretek study indicated that travel time was increased between 3.4 to 25 percent during rain.

Snow

Similar to rain, there have been a significant number of studies investigating the impact of snow on freeway flow as presented in Table 2.7 and 2.8.

Table 2.7 Summary of Snow Effects on Volume

Volume Reduction	Volume Reduction: Freeway	Volume Reduction: Freeway	Volume Reduction: Arterial
Researcher	Hanbali and Kuemmel	Knapp	Maki
Location	Illinois, Minnesota, New York, Wisconsin	Iowa	Minneapolis, Minnesota
Year	1992	1995-1998	1999
Light Snow	7-31%	[no value]	[no value]
Heavy Snow	11-47%	16-47%	15-30%

 

Table 2.8 Summary of Snow Effects on Speed

Speed Reduction	Speed Reduction: Freeway	Speed Reduction: Freeway	Speed Reduction: Arterial	Speed Reduction: Arterial
Researcher	Ibrahim and Hall	Kyte	Maki	Perrin
Location	Toronto, Ontario	Idaho	Minneapolis, Minnesota	Salt Lake City, Utah
Year	1994	2001	1999	2001
Light Snow	0.97 km/h (0.6 mph)	16.4 km/h (26.4 mph)	[no value]	13%
Heavy Snow	37.0-41.8 km/h (23-26 mph)	16.4 km/h (26.4 mph)	40%	25-30%

Heavy snow was also found by Ibrahim and Hall to decrease capacity by 30 percent, and light snow was found to decrease flows by five to 10 percent.

Table 2.9 summarizes the findings of research studies. A blank cell indicates that there are no empirical studies investigating a particular relationship. Most of these results reflect localized, mostly small sample research studies.

Table 2.9 Summary of Research Results – Freeway

Condition	Volume	Maximum Observed Flow	Capacity	Speed
Low Visibility	[no studies]	[no studies]	[no studies]	reduced 13%
Rain	[no studies]	reduced 0-20%	reduced 4-47%	[no studies]
Snow	reduced 7-47%	reduced 5-10%	reduced 30%	reduced 13-40%
Wind	[no studies]	[no studies]	[no studies]	reduced 10%

2.2 Microscopic Behavior Research

Macroscopic impacts on traffic flow resulting from adverse weather are the aggregate results of microscopic driver behavior. Microscopic driver behavior includes acceleration, deceleration, car-following, lane changing behavior, and gap acceptance. Though microscopic driving behavior models have been used for decades, knowledge of microscopic driving behavior remains limited, because human behavior is so complex and microscopic data collection is expensive.

To date, scarce research has been conducted on how weather events impact microscopic driving behavior logic, such as vehicle following and lane changing. While it is logical to conclude that adverse weather results in a more challenging driving environment, the exact mechanisms for a motorist’s response to weather events are limited. Knowing which critical parameters within a driving behavior model should be changed under various weather conditions would aid in the development of weather-responsive traffic management strategies.

Colyar, et al. (2003) conducted a study to identify and assess the sensitivity of a range of model parameters that could be affected by weather events and which most impact the quality of traffic flow. However, this study did not research field data of microscopic driver behavior under different weather conditions.

Sterzin (2004) used aggregate weather and traffic data to refine and enhance microscopic driving behavior models used in a traffic simulator. Field data from a case study indicated that the presence of precipitation was significant in reducing speeds, and was incorporated into the acceleration and lane changing models with aggregate calibration. A sensitivity analysis of eight parameters within four key components of driving behavior – free-flow acceleration, car-following acceleration, lane changing, and gap acceptance – was performed. The sensitivity analysis indicated that car-following parameters had the most impact on modeling driving behavior in inclement weather, with desired speed and gap acceptance also playing a role. Calibration results found that car-following acceleration and deceleration were negatively affected by the presence of precipitation, while the critical gap (in lane changing behavior) was increased, indicating more cautious driving behavior. Additionally, the mean of desired free-flow speed decreased, while the spread of the distribution around the mean increased, in adverse weather. Sterzin reasoned that the range of aggressiveness of drivers increases the variability of desired speed in inclement weather, as some modify their behavior due to environmental conditions, while others do not.

The FHWA is currently sponsoring the Next Generation Simulation (NGSIM) program, which has a primary focus on microscopic modeling, including supporting documentation and validation data sets that describe the interactions of multimodal travelers; vehicles and highway systems; and interactions presented to them from traffic control devices, congestion, and other features of the environment. As part of this program, a variety of traffic microsimulation stakeholders were surveyed about the parameters and mechanisms used to capture a variety of influencing factors (weather and environment included). Responses showed that the models used to simulate acceleration and lane changing generally do not have any parameters that explicitly relate to weather and its effects.

This NGSIM effort is expected to provide enhanced understanding of microscopic driving behavior. However, evaluating the impacts of weather on this behavior is complicated by the fact that efforts to collect the trajectory data – detailed, subsecond vehicle position data that are typically used in simulation models development, estimation, and validation – are hindered by inclement weather. To date, the video recording technology is not sophisticated enough to capture vehicle trajectory data in low visibility or precipitation.

2.3 Human Factors Research

Studies that look at individual driver response to adverse weather are limited in number. Techniques used include use of driving simulators, use of video to observe individual vehicles in the traffic stream or installation of video in the vehicle to observe driver behavior under varying conditions. Driving simulators have a long history and have been used extensively in automobile design and safety evaluations. Use of video is increasing and has a variety of applications. The Next Generation Simulation project (NGSIM) being sponsored by FHWA is currently using video to help refine car-following and lane changing algorithms for the next generation of simulation models. The University of Michigan Transportation Research Institute (UMTRI) is currently videotaping a group of drivers to evaluate their reactions to curve and lane departure warning systems. The UMTRI application could also be matched with weather data, from either outside sources or the vehicle’s “black box” to help evaluate driver response to adverse road weather conditions.

The Vehicle Infrastructure Integration (VII) Initiative is another source of research on driver behavior in adverse weather. Among the many applications of VII will be utilization of data available from the vehicle’s “black box,” and location data, to identify unsafe situations and address them through driver warnings. Automated control of the vehicle is the next possible step in extreme circumstances. Field tests currently getting underway will look at various uses of weather data collected directly through the vehicle. Temperature data as well as data on antilock braking system engagement and windshield wiper usage, for example, may be used in combination or individually to collect data on driver response to weather events.

The sources below document a variety of techniques to evaluate driver response to adverse weather. Driving simulators and surveys are the techniques that have been used most often. Driving simulators are a valuable tool with numerous applications but it is important to note that the driver is not under the same pressure as on the road. There are no consequences to mistakes and as a result drivers are not likely to behave in exactly the same manner. Surveys are based on recall or perception that may not match the actual performance of a driver in the field. One would anticipate that respondents might consider themselves better drivers than they actually are, and answer questions accordingly. However, surveys can be useful when targeted toward specific types of information.

Because the goals and methods of these studies vary so widely it is hard to draw clear conclusions regarding driver behavior in adverse weather. A number of the surveys conducted and some of the observational studies do indicate that drivers have a realistic view of the risks involved in adverse weather. While they do modify their behavior to some extent, the research indicates that the changes do not reflect the level of risk involved. This indicates a need for better driver education and also additional research into management and information systems that will help to modify driver behavior. There are some indications that real-time information systems placed in the field can help reduce the speed differential during times of adverse weather. There are also hopeful indications that high-quality pretrip information may influence driver behavior.

The findings have implications for research into the impact of adverse weather on traffic flow. It is important to obtain a better understanding of driver behavior during adverse weather and it is also critical to understand what strategies will influence behavior in a way that will improve safety.

A summary of research efforts targeted toward individual drivers is included in Table 2.10.

Table 2.10 Literature and Research on Focused on Individual Drivers

Reference	Authors	Year	Comment
Effects of Variable Message Signs on Driver Speed Behavior on a Section of Expressway under Adverse Fog Conditions	V. Ganesh Babu Kolisetty, Takamasa Iryo, Yasuo Asakura and Katsuhiko Kuroda, Kobe University Japan	2006	The authors used a driving simulator in a laboratory setting to examine the effect of VMS on driver speed behavior while viewing the information provided through VMS. The simulation focused on 8.5 km of an Expressway in Japan under foggy conditions in cases with and without VMS. Results showed that 40% of subjects were clearly impacted by the VMS, 40% were marginally impacted and 20% were not impacted at all.
Road Safety Evaluation Using a Driving Simulation Approach: Overview and Perspectives	A. Benedetto and L.V. Sant’Andrea, University of Roma, Italy	2005	This paper reviews current practice in the use of driving simulators and identifies an approach for using simulators to investigate the role of human factors in a multidisciplinary approach. Three promising areas of study are identified 1) theoretical investigations 2) road project validation 3) road safety audit and assessment. Weather is one of the suggested areas of research.
Analysis of ATIS Effect on Mode and Route Choice	Mohamed Abdel Aty and Fathy Abdella	2005	The objectives of this paper were to evaluate driver behavior in response to traveler information. Issues evaluated included mode choice, route diversion and adherence to pretrip route plan. A travel simulator was used as a data collection tool. The simulator uses a realistic network, two modes of travel, actual historical volumes and different weather conditions. Among the findings were that as the level of information is increased, including both pretrip and en-route, drivers are more likely to divert from their normal route.
Impact of Traveler Advisory Systems on Driving Speed: Some New Evidence	Linda Ng Boyle and F. Mannering	2004	This study used a full sized driving simulator to collect information on the effect of real-time weather/incident hazard information provided by VMS and in-vehicle information systems. The study found that these systems were effective in reducing speeds under adverse conditions but that drivers tended to increase their speed downstream of the hazardous area in an attempt to make up the lost time.
Road Safety and Weather Information: Weather and Transportation in Canada	J. Andrey, B.N. Mills and J. Vandermolen, University of Waterloo, Canada	2003	This paper summarizes current knowledge regarding weather-related crash risks and the role of weather information in road safety. It notes the lack of information due to the difficulties in monitoring weather-related driver behavior. A review of research found that most drivers access weather information prior to their trip but do not change their travel patterns. More research is needed to determine what level of information is needed to influence traveler decisions.
Motorists Perceptions of and Responses to Weather Hazards: Weather and Transportation in Canada	J. Andrey and C. Knapper, University of Waterloo, Canada	2003	This study used group interviews and a large sample public survey to explore driver reaction to weather hazards in southern Ontario. A parallel study was carried out with driving instructors. The survey results were then compared to objective measurements of risk, based on accident statistics. The study found that both drivers and instructors have a realistic view of driving risks in hazardous weather but that drivers do not tend to modify their behavior. Another problem identified was that driver education does not provide helpful strategies for coping with hazardous weather.
Integrating Human Factor Evaluation in the Design Process of Roads – A Way to Improve Safety Standards for Rural Roads	S. Cafiso, G. LaCava, R. Heger, R. Lamm, University of Catania, Italy	2003	The objective of this project was to improve highway design standards with respect to human factors needs. A field data collection effort used an instrumented car traveling in traffic on two-lane roads. The car was equipped with GPS equipment, speed and acceleration sensors, a video camera recording a view of the driver and a suite of equipment to collect psycho-physiological responses. Based on the data a procedure was developed to categorize good and poor driving conditions.
Field Operational Test of the Freightliner/Meritor Wabco Roll Stability Advisor and Control at Praxair	C. Winkler, J. Sullivan, S. Bogard, R. Goodsell and M. Hagan, University of Michigan Transportation Research Institute	2002	The report documents the experience with a Field Operational Test of the Freightliner/Meritor Roll Stability Advisor and Control. The system is intended to reduce rollover risk and improve driver performance through in-cab messages. Where necessary it can slow the vehicle automatically. The system was tested under actual operating conditions with 14 drivers completing the study. Weather was one of the factors considered in system performance.
The “Darwin” Driver Vision Support System: Its Potential Impact on Driving Behavior and Road Safety in Conditions of Reduced Visibility	Phillip Barham, Luisa Andreone, Xiang Hua Zhang and Maxime Vache, University of Leeds, Oskar Faber TPA	2000	This paper described the Darwin project which tested a vision support system used to aid drivers during periods of low visibility. The system included the use of an onboard infrared camera for detecting objects and a virtual image to present the objects to drivers. A driving simulator was used to conduct a number of human factors evaluations. Useful information was provided on system design and the evaluation indicated that the system could encourage drivers to reduce headways.
Naturalistic Driving Studies: A Tool for the Development and Evaluation of In-Vehicle Systems	Robert Llaneras, University of Minnesota	1999	This study summarizes the results of an expert panel on naturalistic analysis methods, defined as those employing unobtrusive in-vehicle instrumentation to record driver behavior and vehicle performance over extended periods of time. The paper lists a number of research topics suitable for these types of studies.
Estimates of Driving Abilities and Skills in Different Conditions	T. Galski, T.H. Ehle and J.B. Williams, Kessler Institute for Rehabilitation, East Orange, New Jersey	1998	This research was a preliminary effort to determine whether various driving situations required different driving skills and abilities. Experienced driver evaluators and trainers estimated the magnitude of driving abilities and skills for different photographed driving situations. Results found some driving situations more demanding than others, but interestingly did not find a difference between good weather and adverse weather.
Driver Acceptance of Weather-Controlled Road Signs and Displays	P. Rama and J. Louma	1997	The study was designed to investigate driver acceptance of weather controlled signs on Finland’s south coast. A series of VMS and Variable Speed Limit Signs are used to provide information and modify speeds based on information gathered from road weather monitoring systems. 590 drivers were interviewed at various intervals after implementation to assess their reactions. Survey responses indicated that only a small percentage of respondents modified their behavior. However the study pointed out the need for study based on real-time monitoring rather than recall.
Estimates of Driver Mental Workload: A Long-Term Field Trial of Two Subsidiary Tasks	L.R. Zeitlin, City University of New York	1995	This human factors study evaluated driver mental workload in two vanpools over a four-year period in New York. The test took place on a variety of roadways and different conditions were recorded, including time of day, traffic conditions, vehicle density, speed, weather, and brake applications. It was noted that weather influenced drivers’ ability.
Car-Following Measurements, Simulations, and a Proposed Procedure for Evaluating Safety	S. Chen,	1995	This paper describes how video data of highway traffic were used to study the effect of illumination, weather and traffic density on driver behavior. Results showed that environmental factors influence driver behavior mainly in congested traffic but not in free flow traffic.
The Advanced Transportation Weather Information System (ATWIS)	Mark S. Owens, University of North Dakota	1995	The ATWIS project was designed to provide a current road and forecasted weather report to the traveling public and commercial vehicles across the Interstate system in the Dakotas and Minnesota. This five-year project helped demonstrate the feasibility of in-vehicle traveler information.
Commuters’ Propensity to Change Transportation Decisions in Adverse Travel Conditions: Results from Behavioral Survey in Brussels	Asad Khattak and Andre de Palma, PATH Program, University of California Berkeley	1994	This study used a comprehensive behavioral survey to help understand traveler behavior under normal and unexpected travel conditions.
Weather Hazards: The Motorist’s Perspective	J. Andrey and C. Knapper	1993	The study is based on a telephone survey of 200 drivers in Hamilton, Ontario and 200 drivers in Ottawa. The survey focused on whether various driver groups differed in their perceptions of weather hazards and their adjustments to them. The survey results indicated that most drivers have a realistic understanding of the risks of driving in adverse weather but that the adjustments they reported making did not reflect the magnitude of the hazard. The results indicated a need for both improved education and active measures to encourage more cautious behavior.

24. Research Needs and Approach

The preceding literature review shows that there is an initial base of research regarding the impacts of precipitation (both rain and snow) and other weather events on macroscopic traffic flow parameters and system performance. However, the results of these studies in terms of weather impacts on speed, capacity, and volumes are variable and often apply only to certain traffic states, on particular types of facilities, and in specific locations. Therefore, there is a need to improve and/or expand our understanding of how these common weather events impact driving behavior and traffic flow under varying demands, on heterogeneous facilities, and in different locations.

Research Focus

The next sections of this report describe the research work undertaken to better understand the impacts of weather on traffic flow. The research is intended to accomplish the following specific objectives:

1. Study the impact of precipitation on macroscopic traffic flow parameters over a full-range of traffic states.
2. Study the impact of precipitation on macroscopic traffic flow parameters using consistent, continuous weather variables.
3. Study the impact of precipitation on macroscopic traffic flow parameters on a wide-range of facilities.
4. Study regional differences in reaction to precipitation.
5. Study macroscopic impacts of reduced visibility.

The work was divided into two phases: Phase I involved developing a data collection and analysis plan, and Phase II focused on conducting the analysis and interpreting the results.

This report also describes the work that needs to be done for Phase III research that will examine human factors and their influences on microscopic traffic parameters such as desired speed, acceleration, and minimum gap during common weather conditions such as rain and snow precipitation. The proposed research includes human studies of individual characteristics and behavior.