The objectives of the Idaho Storm Warning System field operational test were to investigate the feasibility of using an advanced weather sensing system to provide low visibility warnings to Idaho Transportation Department (ITD) personnel and motorists, and to assess the degree to which driver behavior was modified through display of roadway condition data via dynamic message signs (DMS). The test location is a section of Interstate 84 in southeast Idaho and northwest Utah. The interstate highway; which runs from Interstate 86 near Rupert, Idaho to Interstate 15 near Tremonton, Utah; was equipped with systems measuring traffic, road surface, and weather data. Three visibility sensor systems (Handar-Belfort, SSI-Belfort, and SSI-WIVIS) were installed to measure visibility, wind speed and direction, precipitation type and rate, air temperature, relative humidity, and pavement condition. Automatic traffic counters recorded the lane number, time, speed, vehicle type (passenger car or truck), and length of each vehicle passing the sensor site. A closed circuit television (CCTV) camera system was installed to confirm sensor visibility readings. Data generated by these systems were transmitted to a master computer located at the Cotterell control center. Because existing phone lines and power supply facilities in this rural area were not reliable, three uninterruptible power supplies (UPS) were installed at the site and a dedicated twisted pair telephone cable was installed from the site to the control center. Four DMS were also installed along the interstate to provide information to drivers. Two DMS had a direct impact on motorists in the test area: one located five miles west of the I-84/I-86 interchange and another located near Sublett, Idaho. The other signs were located further south, near the Idaho-Utah boarder and near the I-84/I-15 junction in Utah.
(Source: Idaho Storm Warning System Operational Test - Final Report, Figure 2, page 9)
The operational test was conducted in two phases from 1993 to 2000. In phase one, sensor data collected on sixteen days during the winter of 1995/1996 were analyzed. These data, covering three clear “normal days” and 13 low visibility “event days,” were evaluated with three different methodologies. CCTV videotapes were used to estimate visibility distance based on views of five target signs placed along the interstate at known distances (i.e., 250, 500, 850, 1200 and 1500 feet). Video image estimates were then compared to visibility readings from sensors. Data from the sensor systems were also compared to determine if they concurrently identified low visibility events (i.e., periods with visibility below 0.23 miles or 0.37 km). The third method compared visibility readings and vehicle speed. Analysis results indicated that the three sensors agreed 82.9 percent of the time, in determining whether or not a low visibility event had occurred. The SSI-WIVIS sensor was not as accurate as the Handar-Belfort or SSI-Belfort sensors. The Handar-Belfort and SSI-Belfort sensors tracked visibility consistently and their measurements were correlated with vehicle speeds. These sensors agreed on whether or not a low visibility event had occurred 96.8 percent of the time. In 82.5 percent of cases, they agreed that there was no event. The Handar-Belfort and SSI-Belfort sensors consistently identified visibility distances below 0.23 miles or above 0.23 miles, and were deemed effective in identifying low visibility events.
Phase two evaluation was carried out utilizing weather and traffic data collected from 1997 to 2000. These data spanned 19 days in 5,790 five-minute intervals and were used to develop a vehicle speed profile for "ideal" conditions. Driving conditions were considered “ideal” when visibility exceeded 0.23 miles (0.37 km), wind speed was below 10 mi/h (16 km/h), there was no precipitation, and the pavement was dry. Under “ideal” conditions, the average speed was 67.7 mi/h (108.8 km/h) for all lanes, 69.1 mi/h (111.1 km/h) in southbound lanes, and 66.1 mi/h (106.3 km/h) in northbound lanes. Vehicle speeds were examined under various weather conditions to assess speed changes based on driver perception of conditions with no supplemental information from DMS or other sources. The results showed that when visibility was very low (i.e., below 0.10 miles or 0.16 km) and all other conditions were “ideal,” average vehicle speeds dropped by eight to ten mi/h (12.9 to 16.1 km/h) or roughly 13 percent. When wind speeds were high (i.e., over 30 mi/h or 48.2 km/h) and all other conditions were "ideal," drivers reduced speed from 67.7 mi/h (108.8 km/h) to 63.8 mi/h (102.6 km/h). Average speeds drop substantially to 50.0 mi/h (80.4 km/h) during periods with heavy precipitation and wet pavement, representing a 26 percent reduction. Drivers reduced speeds to 40 mi/h (64.3 km/h) under severe conditions when winds are high, visibility was low, snowfall was heavy, and pavement was snow covered. Under these severe conditions, speeds were reduced by nearly 41 percent. This analysis indicated that drivers reduced their speeds in adverse weather conditions without advisory information from DMS.
To determine whether use of message signs encouraged drivers to further reduce their speeds, another analysis was conducted with and without DMS warnings under various weather conditions. The DMS near the I-84/I-86 interchange was visible to southbound motorists, who were potentially influenced by displayed messages. Drivers traveling north could not view the sign’s messages. Thus, differences in northbound and southbound travel speeds were attributed to DMS usage. ITD personnel manually operated DMS based upon sensor system data, low visibility warnings generated by the master computer, and their judgment. When all weather conditions were considered together, the analysis revealed that DMS had little or no effect on driving speeds. Average speed without DMS was 61.5 mi/h (98.9 km/h), while average speed with DMS display was 56.2 mi/h (90.4 km/h). However, under certain conditions substantial speed reductions were observed in southbound lanes where drivers benefited from advisory information.
When winds were high and DMS were not used, average speed was 54.8 mi/h (88.1 km/h) in southbound lanes and 50.8 mi/h (81.7 km/h) in northbound lanes. When DMS were used during high winds, average speeds dropped by 12.5 mi/h (20.1 km/h) or 23 percent in southbound lanes, compared to a reduction of 3.7 mi/h (6.0 km/h) or seven percent in northbound lanes. Speed variability also decreased in southbound lanes.
When high winds and moderate to heavy precipitation occurred simultaneously, mean speeds in southbound lanes drop from 47.0 mi/h (75.6 km/h) without DMS to 41.2 mi/h (66.2 km/h) with DMS warnings—or by roughly 12 percent. When high winds occurred with snow-cover pavement, mean speeds in southbound lanes dropped 35 percent from 54.7 mi/h to 35.4 mi/h (87.9 km/h to 56.9 km/h) compared to a nine percent decline from 48.4 to 44.1 mi/h (77.8 to 70.9 km/h) in northbound lanes. Average speed in southbound lanes was 8.7 mi/h (14.0 km/h) lower—or nearly 20 percent lower—than speed in northbound lanes, where drivers were unaffected by DMS displays.
Researchers concluded that when travel conditions deteriorate due to poor visibility, high winds, wet or snow-covered pavement, and/or heavy precipitation; drivers reduced speeds even without condition information from DMS. Drivers further reduced speeds when advisories were issued during high winds, high winds with moderate to heavy precipitation, and high winds with snow-covered pavement. Under these conditions, speeds were nearly 20 mi/h (32.2 km/h) lower when warnings were displayed on DMS. A noteworthy qualitative benefit of the Storm Warning System test was ITD's increased awareness of road weather information systems (RWIS) and their applications.
- On April 11, 1996, the speed limit in the project area was increased from 65 to 75 mi/h for cars and trucks. In 1998, truck speeds were reduced to 65 mi/h. This change did not affect the data collected during phase two, when the speed limit was 75 mi/h throughout the evaluation period.
- The time-lapse video camera proved to be a useful tool for confirming significant changes in visibility levels, but was not effective in determining precise visibility distances. Thus, the results of the video analysis were inconclusive and could not be used to substantiate whether the visibility sensors provide accurate estimates of visibility.
- It should be recognized that the DMS analysis was not based on highly controlled experimental conditions and the results of the analysis cannot be considered conclusive.