Dashboards for Real-time Monitoring of Winter Operations Activities and After-action Assessment

The Indiana Department of Transportation (INDOT) operates a fleet of nearly 1100 snowplows and spends up to $60M annually on snow removal and de-icing as part of their winter weather operations and maintenance activities. Systematically allocating resources and optimizing material application rates has the potential to save revenue that can be reallocated for other roadway maintenance operations. Modern snowplows are beginning to be equipped with a variety of Mobile Road Weather Information Sensors (MARWIS) that provide a host of analytical data characterizing on-the-ground conditions during periods of wintry precipitation. The state of Indiana operates 14 such snow plow trucks that have been equipped with MARWIS devices. These devices collect geo-location data including latitude, longitude and altitude along with speeds and heading. In addition, they estimate road conditions from the measured parameters, which include dew point, surface temperature, friction, ice percent, relative humidity and water film height. Traffic speeds fused with road conditions and precipitation data from weather stations can provide a uniquely detailed look at the progression of a winter event and the performance of the fleet. This research thus used a combination of traffic speeds, MARWIS and North American Land Data Assimilation System (NLDAS) data to develop real-time dashboards characterizing the mobility impact of precipitation and pavement surface temperature.

Inclement weather significantly impacts traffic operations, reducing efficiencies of state maintenance vehicles and creating unsafe driving conditions for the public. Tracking the performance of snow and ice control operations has hence assumed prime importance for highway agencies and contractors due to the attached stakeholder expectations. A number of performance metrics for the same are in use worldwide however no single metric exists that applies to a variety of road types, storms and traffic conditions. Improvements in weather prediction and forecasting models carry immense value for state agencies who can use these advances for efficient resource allocation, scheduling and maintaining roadways during heavy precipitation events. This study leverages traffic condition, road condition, weather data and geo-location of fleets to develop dashboards for real-time monitoring of winter operations activities and after-action assessment. The dashboards will potentially provide agencies with a comprehensive look at traffic, weather and pavement conditions in a single interface, making resource allocation and scheduling significantly more streamlined when operating off of a unified data source.

Twenty heavy snow events were identified for the state of Indiana from November 2018 through April 2019. These storms were classified as being of observable impact if the peak congested miles (miles of interstate operating below 45 mph) on the event days were over 100 miles. Two particular instances with moderate observable impacts on interstates of 182 miles and 231 miles at their peaks occurred in January and March, respectively, and were used as a case study for this research. The stark contrast seen in the winter weather maintenance strategy for these dates made them ideal candidates for this study. A suite of dashboards has been developed that compile traffic, road condition and weather data all into one unique solution for agencies towards real-time tracking for winter weather operations (Figure 1). Speed profile heat maps highlight congestion on stretches of interstate roadway, while MARWIS trajectories for snow plows plotted on these heat maps pinpoint the whereabouts of maintenance vehicles with respect to traffic conditions giving agencies a dynamic look at their fleet’s geolocation. A direct comparison in the dashboards in Figure 1 in the left and right columns, contrasts the agency’s response to two winter events and highlights the visible differences that can be caused in winter maintenance activity by a less than ideal weather forecasting model. The snow plows covered the January 12, 2019 storm extremely well by pre-treating as well as plowing the roads for the entire duration of the storm. They however lagged behind the March 30, 2019 storm by a couple of hours and only began plowing activity after wintry precipitation had already begun impacting the interstate roadway system. The above can be verified from the observation that snow plows were in operation on January 12, much before the wintry precipitation amounts began to rise, while that was not the case for March 30, 2019.

These web dashboards can clearly show the impact and efficiency of winter weather maintenance activity before, during and after a storm and can prove to be a valuable medium for tracking an agency’s snow plow fleet in real-time with the progression of a storm. Modern snow plows can report back data from on-board controllers (plowing and salting) as well as from externally fitted mobile road weather devices. Using the broad spectrum of enhanced probe-vehicle data at their disposal, agencies and Departments of Transportation (DOT) can now have a complete picture of the road conditions without having to depend solely on weather forecast models. Although weather forecasts will never be 100% accurate, dashboards such as these provide a framework for winter operations staff to monitor conditions, perform after-action assessment, and provide improved feedback to forecast modeling colleagues.

Moving forward, these dashboards can also be used to quantify a number of performance metrics. The recovery from event performance metric, for example, can be computed with the help of the MARWIS trajectories as they track pavement friction values in real-time and can exactly pinpoint the start and end of a snow event based on the pavement friction value dipping below a pre-defined threshold. Similarly, safety, travel delays, speed reduction are all performance metrics that could easily be deduced using the dashboards developed in this research. Finally, multi-variate Autoregressive Integrated Moving Average (ARIMA) forecasting models could be developed based on the three data sets that constituted this research – traffic (speeds), weather (surface skin temperature, precipitation amount) and MARWIS (trajectories, weather data) to be able to forecast traffic conditions for a winter event and how they would be impacted by precipitation using historical data. This holds promise as an effective weather prediction tool for agencies eventually helping them better allocate and schedule their assets in advance of a snow event leading to an overall improvement in winter weather mobility.