When pavement temperatures drop below freezing and there is sufficient moisture in the air, frost—transformed into black ice by the action of traffic—can form on the pavement surface.
This roadway hazard occurs frequently in cold climates. Aided by road condition forecasts, departments of transportation (DOTs) across North America have been combating the formation and persistence of black ice through anti-icing and deicing measures. A recent innovation—night icing potential prediction—cost-effectively revolutionizes road-condition forecasting, focusing winter road maintenance efforts where they are needed most to enhance driver safety.
Pavement conditions are widely forecasted utilizing road weather information systems (RWIS), which has been the primary road-condition forecasting tool for many DOTs for more than a decade. RWIS technology utilizes in-pavement sensors to predict pavement temperatures and road conditions over the coming 24-hour period. These systems have the capability to predict the timing of potential road icing using point thermodynamic models that track anticipated precipitation types and amounts. Examining pavement-sensor records, maintainers can surmise when the existing liquid and/or solid material applications will wear away and no longer counteract the icing.
RWIS stations are very powerful tools when predicting black-ice and frost formation on the pavement overnight. However, this forecast is valid only at the exact RWIS station location, leaving operators with little knowledge regarding the pavement-condition forecast between stations. Solely using RWIS stations, it takes a vast network to get an accurate view of pavement conditions across an entire region. Each tower represents a significant capital investment and requires frequent maintenance. Most DOT budgets limit the placement of RWIS stations to no closer than 19 miles apart, except in densely populated areas. RWIS networks, while incredibly useful in predicting pavement conditions for spot locations, cannot detect large variations in road-pavement temperature between towers.
Enabled by Global Positioning System (GPS) deployments in recent years, automatic vehicle location (AVL) technology has been developed and implemented in many fleets across the country. Through a combination of AVL and vehicle-mounted infrared temperature sensors, it is now possible to thermally map the road surface between RWIS stations. AMEC Earth & Environmental (AMEC), in collaboration with the Nova Scotia Department of Transportation and Infrastructure Renewal (NSTIR), utilized high-tech thermal mapping to develop a night icing potential prediction system that proved highly effective in improving safety on Nova Scotian roadways. With the support of Paul Richard, NSTIR’s manager of maintenance and operations, who championed this endeavor, AMEC devised an innovative approach to accurately predict the earliest time for the potential formation of ice along continuous stretches of road to a resolution of 1 km.
This application of thermal mapping can be utilized worldwide to generate road-condition forecasts for entire routes rather than at particular points where pavement data is continuously collected. For DOTs with RWIS networks, night icing potential forecasting presents a cost-effective approach to improving safety across every kilometer of roadway they maintain.
The test case
Night icing potential, known as NIP, is the potential that ice will form on a particular section of pavement overnight. The NIP prediction system developed by AMEC is an operational application of infrared thermal mapping for cold climates. The end product accurately predicts the earliest hour of potential icing for every 1-km segment of roadway. The final color-coded NIP forecast product is sent from AMEC forecasters directly to NSTIR operators, who can quickly identify the segments of their route that are likely to ice up in the early morning hours and which sections do not need to be treated at all.
The infrared thermal maps of the roadway were obtained through a collaboration between meteorological consultants at AMEC and road-maintenance personnel at NSTIR. NSTIR provided patrol vehicles equipped with AVL and infrared pavement temperature sensors, as well as drivers, which together are the largest investment required to thermally map the road segments.
AMEC provided an orientation to NSTIR operators, detailing the procedure to collect infrared data through an outfitted patrol vehicle run. AMEC then coordinated the run timing with NSTIR according to the forecasted atmospheric conditions to develop thermal fingerprints, which will be discussed in some detail. Once thermal fingerprints were obtained, meteorologists established the relationship between RWIS station pavement temperatures and the pavement temperatures along each segment of road. This collection of relationships was then integrated daily with RWIS forecasts to predict NIP.
Each patrol vehicle had an infrared road surface temperature measuring sensor attached to the side of the vehicle. The monitoring industry produces several sensors suitable for this application. The important factor in selecting a surface temperature monitoring instrument is its response time to road surface temperature changes. The Sprague RoadWatch used by NSTIR could accurately sense a 1°C change in road surface temperature in 1/10 of a second. Additionally, this sensor has an RS-232 connection that sent a signal directly into the AVL unit, which made it a practical choice for this application.
As the patrol staff proceeded at a designated speed, AVL utilized GPS to fix the position and time, while cellular communications and Internet data display and recovery logged the data every two seconds. Runs were performed in the early morning hours, and atmospheric and road conditions were observed by the drivers during the runs. At a speed of 35 km per hour, NSTIR was able to collect data points with position, time, air and road temperatures every 20 meters.
Between February and March 2007, 23 runs were completed across a 42-km section of Trans Canada Highway 104 (TCH 104) in northern Nova Scotia just west of New Glasgow. There were two RWIS stations located within this section of road. Once the data was sent to AMEC, meteorologists categorized the runs into classes based on observed weather conditions along the route at the time of each run. This portion of the project involved multiple GIS technicians and meteorologists at AMEC, an effort coordinated by Senior Associate Paul Delannoy.
AMEC generated a thermal profile for each run, then filtered and averaged all of the runs within each class to establish a static thermal fingerprint for each class. The thermal fingerprints identified consistent relative cold or warm spots along the section of road, with a new data point every 20 meters. In consultation with NSTIR, AMEC split the roads into 1-km segments, assigning the coldest data point to the entire segment. This cold-biasing errs on the side of caution: When a NIP forecast identifies 6 a.m. as the earliest icing potential for a segment of road, it means that the coldest part of that 1-km section may freeze as early as 6 a.m. In GIS format, the thermal fingerprints can be shown as a highly complex function of elevation and land cover.
After determining the length of road segments, AMEC set up an association for each weather class between road segments, thermal fingerprints and available RWIS sites. The association is an intricate process that requires an advanced meteorological assessment of local climatology to decide which RWIS station is most relatable to each 1-km road segment. It is imperative that the road segments are matched not to the nearest RWIS station, but to the RWIS station observations that will be most representative of the associated route portion.
For example, in the demonstration project, Upper Mount Thom has an RWIS station near the crest of the hill at an elevation of nearly 450 meters. For a particular segment on the lee side of the hill, thermal mapping revealed that it was consistently 2°C warmer than the RWIS station reading at the top of the hill when extreme class conditions (clear, calm nights) were observed. Every time those atmospheric conditions are met in the future, it is reasonable to assume that that segment of roadway would again display the tendency to be 2°C warmer than the RWIS station. Once these types of associations were determined for each class, AMEC could forecast the earliest potential for icing along each road segment based on their RWIS forecast for the next morning.
Forearmed and forewarned
To forecast icing at an RWIS station, meteorologists utilize station data, atmospheric model guidance and the METro pavement model to determine potential for frost formation at the RWIS site. Previously, this represented the upper limit of detail that forecasters could provide. However, utilizing the road’s established set of thermal fingerprints, together with RWIS-site forecasts, AMEC forecasters are able to indicate the earliest possible onset time for icing for every 1-km segment of roadway.
The operational product is a color-coded forecast map that designates the earliest possible icing time for each road segment. The areas that are anticipated to be colder than the RWIS station indicate the earliest icing potential. The thermally indicated warmer segments reflect the latest icing times and in some cases are forecast never to ice. The earliest icing onset times are utilized to better plan overnight winter road maintenance, scheduling patrols near those times and anticipating the need for materials application. Operators do not waste effort on sections of road that are forecast to remain ice-free, providing more time for them to focus on the areas that are forecast to have significant ice buildup. Valuable time and materials are saved by intelligent and informed decision making, made possible through NIP forecasts.
This type of forecast is now used operationally by NSTIR on a daily basis. NIP forecasts are issued every afternoon in the winter to indicate the next morning’s potential for icing. Earliest possible icing times are indicated spatially on a map which shows a particular road, the RWIS stations along it, and each 1-km road segment. The road segments are labeled with the NIP onset hour, distinguishable also by color.
Because of the cold-biasing of the segment temperatures during thermal-profile development, NIP can overforecast icing events. During a two-month operational test period during which NSTIR performed all patrol runs of the roadway, it was determined that NIP forecasts did not miss a single icing event. The forecasts were correct nearly three-quarters of the time. There were occasions where icing was forecast and it was not observed. By design, NIP errs on the side of safety, providing confidence that although icing is sometimes overpredicted, icing events will not be missed.
It is important to consider that NIP forecasts are only as accurate as the RWIS forecasts from which they are generated. If a given segment of road is forecast to be 1° colder than its associated RWIS station and the RWIS forecast has an error of 2°, the error is transferred to the NIP forecast. Accurate RWIS forecasting is critical to the success of NIP.
Since the NIP test run in early 2007, NSTIR has developed permanent thermal maps of additional routes and extended NIP forecasting to nearly 450 km of highways across the province. Paul Richard and NSTIR are strong advocates for the service. Through further thermal mapping product development, AMEC has continued to elevate road condition forecasting to more advanced levels.
For any DOT with a pavement sensor network, NIP forecasting is the next logical step to expand and improve services. What was previously an RWIS-centric, site-specific forecast can now become a spatially continuous forecast product, with the operational application of thermal-mapping technology.
By utilizing existing personnel and resources, thermal fingerprints can be achieved cost-effectively with the assistance of a NIP-experienced weather forecast provider. The NSTIR-AMEC collaboration has paved the way for other DOTs to utilize the same methods to establish their own NIP forecasting service. The initial development of this technology and technique involved extensive research and development. Now that the details have been tested and proven, this procedure is ready to be applied to other road networks and road segments without significant ramp-up time.
Many of the NIP forecast features are customizable. For example, the Nova Scotia NIP product indicates forecasted freezing times for each 1-km segment. It is possible to provide a NIP forecast for finer (i.e., shorter) road segments using the same thermal fingerprinting equipment and approach, if desired for urban areas. It also is possible to acquire infrared data with much faster vehicle speeds.
Thermal-mapping forecasting techniques are not limited to icing potential, but rather can be expanded to forecast other pavement-related parameters. For example, rather than simply forecasting the earliest onset of ice, AMEC has recently developed a similar forecast product that predicts the lowest overnight road temperature for each segment. This is particularly useful when some form of precipitation is expected, since it provides an indication of whether the rain or snow will bond to the roadway. The possibilities for utility of thermal-mapping data are numerous, and AMEC is investigating other partnerships to test these interesting possibilities to improve road safety and reduce the cost of road-condition forecasting.