Calibration of Traffic Flow Models under Adverse Weather and Application in Mesoscopic Network Simulation

Tian Hou, Hani S. Mahmassani, Roemer M. Alfelor, Jiwon Kim, Meead Saberi
2013 Transportation Research Record  
Saint Paul, Minnesota, metropolitan area and showed that adverse weather causes clear reductions in traffic speed: up to 6% for rain, 13% for snow, and 12% for reduced visibility (1). Ibrahim and Hall (5) analyzed the effects of adverse weather on the speed-flow and flow-occupancy relationships for Canadian travelers and found the effects of snow to be much larger than those of rain and to cause a reduction in free-flow speed of 38 to 50 km/h. The effects of weather on traffic volume are also
more » ... ident from empirical data. The research conducted by Datla and Sharma indicates that the impact of cold and snow on traffic volume varies with the type of trip and hour of the day (6). From traffic data collected in Canada, they observed that commute trips experience the lowest reductions in volume because of snowy weather, of up to 14%, while recreational trips experience the highest reductions, of up to 31%. They also found that reductions in commute trips during off-peak hours (−10% to −15%) were generally greater than those during peak hours (−6% to −10%); however, an opposite pattern was observed for recreational trips. All these studies show that inclement weather may have a significant and comprehensive impact on the transportation system that cannot be ignored by planners and decision makers. To mitigate the impacts of adverse weather on highway travel, the FHWA Road Weather Management Program has been involved in research, development, and deployment of strategies and tools for weather-responsive traffic management. In a project completed in 2006, the Road Weather Management Program used data from Seattle, Washington; Minneapolis, Minnesota; and Baltimore, Maryland; to develop statistical models and adjustment factors to quantify the impacts of weather on traffic flow (7). One of the challenges remaining is to integrate those models into decision support systems to help improve the performance of the transportation system during inclement weather conditions. The traffic estimation and prediction system (TrEPS) is a tool currently available for traffic planners and operators to assist with evaluating and implementing weatherresponsive traffic management strategies. Weather-sensitive TrEPS capabilities aim for accurate estimation and prediction of the traffic states under inclement weather conditions. Mahmassani et al. identified several key components within the TrEPS framework for which the impact of weather must be incorporated on both the supply and demand sides (8). One such element on the supply side consists of well-calibrated weather-integrated traffic flow models. Successful application of weather-sensitive TrEPS requires detailed calibration of weather effects on the underlying traffic flow models. The main objectives of this paper are (a) to develop systematic procedures for calibrating traffic flow models under inclement The weather-sensitive traffic estimation and prediction system (TrEPS) aims for accurate estimation and prediction of the traffic states under inclement weather conditions. Successful application of weathersensitive TrEPS requires detailed calibration of weather effects on the traffic flow model. In this study, systematic procedures for the entire calibration process were developed, from data collection through model parameter estimation to model validation. After the development of the procedures, a dual-regime modified Greenshields model and weather adjustment factors were calibrated for four metropolitan areas across the United States (Irvine, California; Chicago, Illinois; Salt Lake City, Utah; and Baltimore, Maryland) by using freeway loop detector traffic data and weather data from automated surface-observing systems stations. Observations showed that visibility and precipitation (rain-snow) intensity have significant impacts on the value of some parameters of the traffic flow models, such as free-flow speed and maximum flow rate, while these impacts can be included in weather adjustment factors. The calibrated models were used as input in a weather-integrated simulation system for dynamic traffic assignment. The results show that the calibrated models are capable of capturing the weather effects on traffic flow more realistically than TrEPS without weather integration. Driving behaviors and the resulting traffic flow characteristics during inclement weather are different from those observed during so-called normal conditions. On the basis of type (rain, snow, frog, wind, etc.), duration, and intensity of the weather, its impact on the performance of traffic networks may vary under different scenarios. Maze et al. identified three predominant categories of variables that are affected by inclement weather: traffic safety, traffic flow relationships, and traffic demand (1). Andrey et al. found that, in Canadian cities, collision rates increase during precipitation by 50% to 100% relative to normal seasonal conditions (2). Similar findings are presented in the literature for cities in the United States (3, 4) and indicate that the duration and intensity of rainfall and snowfall have a positive and statistically significant relationship on the number of crashes. Maze et al. studied the freeway system in the Minneapolis-
doi:10.3141/2391-09 fatcat:b755oen4vvd7hn6o4huyjlgz4e