Data Mining and Gap Analysis for Weather Responsive Traffic Management Studies

5.0 Weather Responsive Transportation Management (WRTM) Survey Summary

A web-based survey was conducted to reach out to the research community, primarily Universities, to learn about projects related to analysis of weather and traffic data, as well Weather-Related Traffic Management strategies transportation researchers may be evaluating.

E‑mails were sent to 170 addresses and 39 responses were received. Respondents were asked if they had conducted analysis using weather data and the type of analysis they did. Twenty-five (64 percent) out of the 39 respondents indicated that they have conducted research using weather data. Section 5.1 summarizes the results of the initial survey while Section 5.2 summarizes the follow-up survey. The full set of results is included in Appendix C.

5.1 Initial WRTM Research Institution Survey

Table 5.1 summarizes the type of analysis these researchers conducted. In addition to the predefined categories, 21 percent of researchers chose “Other.” The Other category included “descriptive statistics of weather pattern,” “road maintenance,” “performance measures for winter operations,” “resources used during emergency events,” “behavior of different sensor systems,” in the “other” category.

Respondents were asked how the data analysis was conducted. More than half of the analysis was done by “collected field data.” Table 5.2 exhibits the details of the responses. Data falling into the “other” category includes “archival weather data,” “material withdraws from accounting system, e.g., salt,” “GPS data,” “crash reports,” “video detection,” and “wireless magnetometers.”

When being asked what sources of weather data they used, the respondents indicated a wide range of sources as shown in Table 5.3.

As for the weather variables analyzed, 62 percent of the respondents responded “precipitation.” Other commonly used variables included “Temperature,” “Ice,” and “Visibility.” Variables listed in the “other” category included “sunshine duration,” “thunder storm,” “humidify,” “weather event duration,” “status of road surface,” “water film thickness,” etc.

About one-third of the respondents, 11 out of 38, indicated that they did not have problems or restrictions at all in accessing the weather data. When being asked “What were the limitations that you encountered with the weather data,” the respondents stated issues, including accessibility of data, accuracy in locating weather station with respect to the considered road section, unnecessary data, reliability of data, and comprehensiveness (lack of data covering a whole city).

We also asked “what types of data analysis would you like to conduct using weather data in the future.” The answers from the respondents are listed in Table 5.5. The largest number of respondents reported that either Traffic Flow Impacts or Safety were the focus of their study. The types of analysis listed in “other” include “road maintenance,” “pavement design,” “performance measures,” and “correct material usage.”

At the end of the survey, suggestions for future weather data collection were asked. Twenty-one respondents answered this question. The suggestions are mainly focused on the following aspects: Open source data; improve the reliability, and finer grained weather data in order to match with the frequency of traffic counting.

In summary, majority of the respondents have conducted research or are interested in conducting research using weather data. The research involving weather data has a wide range of topics. However, quality of current weather data is not quite satisfactory. Researchers are expecting weather data to be more accessible, more reliable, more accurate, and to cover more locations.

5.2 Follow-up WRTM Research Institution Survey

Due to the strong response to the original survey a follow-up survey was conducted to provide more detail on research efforts. The follow up survey was sent to respondents who offered to provide more information and 18 responses were received. A summary of results is presented in this section and full details are presented in Appendix C.

Sixteen of the 18 respondents, 89 percent, would like to use the weather information to study the impact of inclement weather on traffic stream behavior. Majority of them, 11 out of 18 (61 percent), would like to develop inclement weather traffic control strategies using the weather information. Details are exhibited in Table 5.6. In addition to the listed research topics, respondents propose research topics such as road surface conditions, driving behavior, including speed and acceleration, etc.

The level of resolution in weather data required by the researchers varies. Most of the researchers prefer more detailed weather information than currently available data. Some required the capability of aggregating weather data as they need (Table 5.7).

Respondents were asked about the type of traffic data they need to conduct their analysis. Most of them, 13 out of 18, need aggregate loop detector data. Sixty-one percent (11 out of 18) wants disaggregate vehicle trajectory data such as naturalistic data or GPS data. Four respondents specified their requirements for traffic data as “travel survey data,” “disaggregate detector data,” or “ADT by route.”

While being asked about the level of accuracy and resolution required in the traffic data to conduct analysis, researchers indicated several levels of details. They believed that level of accuracy for traffic data should depend on the type of analysis or the focus of the analysis. The accuracy of traffic data should be comparable to weather data. In general, for traffic data (speed, density, and flow), a minimum of two-to-five minutes aggregated data is required by most of the researchers. For vehicle trajectory data, a higher resolution, one-to-five seconds, is desired.

Respondents were asked the type of data required to integrate weather and traffic data. According to the answers, location and time information are indispensable. Researchers also asked for data on such items as traffic intensity, speed limit, probe-vehicle data, road surface condition and IntelliDrive-based weather information, etc.

At the end of the follow up survey, researchers were asked about the sampling size issues they need to address to conduct their research. The following issues were highlighted:

• Distance between weather station and traffic count station is too far;
• The sampling time length is not long enough;
• The data resolution is not high enough;
• The sites are too few; and
• Weather types are not covered comprehensively.

In summary, researchers are looking forward to more detailed weather data. An efficient linkage of weather data with accidents, work zones, video, and other types of data is highly desired. RWIS data, road surface condition reports, and probe data were other needs identified by the respondents. Researchers also made suggestions on further weather data collection in the aspect of a more open data warehouse, fewer errors in data sensors, and more reliable vehicle-based data.

Previous | Next