Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
10 C h a p t e r 3 3.1 Introduction The research team built the RID from two primary sources of data. These sources included new roadway data collected through the mobile data collection project (about 12,500 cen- terline miles), which was quality assured to meet minimum accuracy requirements, and acquired roadway and supple- mental data provided by the departments of transportation (DOTs), local agencies, and private vendors within each state and study area. Of these data sources, the mobile collection provided the highest level of attribution, granularity, and spatial accuracy. However, as illustrated in Figure 3.1, the cost of the mobile data collection presented a limitation on extent (coverage) due to time and resource limitations. The research team developed a plan with SHRP 2 to use the available funding for mobile data collection while ensuring ade- quate coverage. Between NDS sites, the criteria for determining what percentage of road miles were collected included the num- ber of NDS participants at a given site, the availability of second- ary sources of data, the study area size, and economies of scale. Within each study area, allocation was based on GPS traces of the NDS participants (which roads were being used and at what frequency), recruit maps (participantsâ self-reporting of typical travel patterns), and road classification. The final allocation focused on a mix of urban and rural roads with an emphasis on rural two-lane roadways, a high priority for safety research. Given that the mobile data collection was concurrent with the NDS data collection, the roads driven by NDS participants and the total number of miles of road frequently driven by the recruits was estimated from available, though limited, GPS trace data. Considering this, SHRP 2 and the research team developed a strategy to collect the data in two primary phases, with the first phase constrained to less than half of the total coverage of 12,500 centerline miles (or 25,000 data collection miles because data was collected in both directions of travel). The following sections describe the principles followed in the selection of roads in these two phases of mobile collection (years 2012 and 2013). However, prior to these phases of data collection, an initial pilot data collection phase of about 900 centerline miles (or 1,800 data collection miles) was completed in 2011 to assess the collection process and the minimum accu- racy requirements developed for the data collection. This pilot phase was critical to the success of the mobile data collection effort. In addition to the pilot and two primary phases, a third phase, or clean-up round, was programmed by SHRP 2 to ensure that mobile data was collected on those sections of roads where NDS participants were involved in a crash. 3.2 allocation Guidance: Between NDS Sites (how Many Miles?) This section describes the criteria used to determine the allo- cation of total mileage for mobile data collection in each of the six NDS sites. The following criteria were used: ⢠Other things being equal, if a study area had more NDS participants, mileage was proportionally increased com- pared to other NDS sites. ⢠Other things being equal, if a study area had little or no secondary-source road data (e.g., highway authority data- bases), especially for rural two-lane roads, mileage was increased. In general, more detailed secondary informa- tion was available for the state-owned or state-maintained roads in most states. Table 3.1 provides the percentage of state-owned or state-maintained roads in each of the NDS sites. ⢠Other things being equal, if an NDS site was larger in area (a proxy for the average length of trips), more miles were collected. Table 3.2 shows the size of each NDS site. This resulted in North Carolina having more miles collected than New York, even though there were fewer participants. Even though crash history was not used to determine the mileage split between sites, crashes occurring during the NDS Data Collection Plan for the Mobile Data Collection Project
11 study period were identified, and roads where crashes occurred were collected in a clean-up round in fall 2013. 3.3 allocation Guidance: Within Sites (Which roads?) As stated previously, the assignment of roads for mobile data collection was to be on those roads where NDS participants were driving, as well as a focus on rural two-lane roads. As GPS traces from the NDS participantsâ vehicles became available, mobile data collection routes were assigned accord- ingly. The following process was used to determine the roads to be collected at each site: 1. Total miles covered by NDS participants; 2. Miles of two-lane rural roads; and 3. NDS participant route maps. Depending on the amount of mobile data collection mileage available in each study area, all commonly used rural, two-lane routes were covered. In other cases, NDS participantsâ route maps (collected by the site contractors) were used to identify/ confirm collection routes. An example participant route map is shown in Figure 3.2. The figure shows routes regularly used (work or school) in pink and routes that are used for less fre- quent activities in blue. To select the individual routes at each site, the research team used a combination of initial GPS traces from Virginia Tech Transportation Institute (VTTI) and participant route maps. The following describes the route selection process for the three years of data collection: ⢠Pilot phase. In 2011, available GPS traces from the NDS par- ticipants were limited. Given this situation, the research team and SHRP 2 decided to do a pilot data collection to assess the data collection process and assess minimum accuracy requirements. Limited GPS trace data (see Fig- ure 3.3 for an example) were provided for three sites: Tampa, Florida (FL), Raleigh/Durham, North Carolina (NC), and Buffalo, New York (NY). NDS participantsâ maps were also used to confirm/select additional routes in 2011. In total, 1,800 miles (or 900 centerline miles) were collected in these three sites. ⢠Phase 1. In 2012, more extensive GPS traces were available for FL, NC, NY, and Seattle, Washington (WA), while lim- ited GPS trace data were provided for Bloomington, Indi- ana (IN) and State College, Pennsylvania (PA). A second mobilization of the data collection vans to FL, NC, and NY finished the collection effort for these three sites, except for the clean-up round. One mobilization to WA completed all the data collection for that site, likewise except for the clean-up round. About half the miles in IN were collected during this phase of data collection. ⢠Phase 2. In 2013, GPS traces for PA and additional GPS traces for IN were provided. The data collection plan cov- ered the rest of the miles in IN and all the collection miles in PA. ⢠Phase 3 (clean-up round). In 2013, and after the data collec- tion plans for Phases 1 and 2 data collection efforts were finalized, an analysis of NDS participant crashes was con- ducted by Virginia Tech Transportation Institute (VTTI) to Figure 3.1. Data detail, coverage, and accuracy by source. Table 3.1. Percentage of State-Maintained Roads in Each NDS Site Area Percentage Tampa, Florida 21% Central Indiana 10% Durham, North Carolina 64% Erie County, New York 11% Central Pennsylvania 33% Seattle, Washington 7% Table 3.2. Geographic Area for Each NDS Site Study Site Area (mi2) Tampa, Florida 4,865 Central Indiana 3,590 Durham, North Carolina 6,490 Erie County, New York 5,456 Central Pennsylvania 6,722 Seattle, Washington 10,060
12 Figure 3.2. Sample NDS participant route map. Pink highlight indicates typical work/ school routes; blue highlight indicates other common routes.
13 assess whether the mobile data were collected on roads where these crashes occurred. Several crashes in five sites (IN, NC, NY, PA, and WA) did not have roadway data. A clean-up round of data collection was conducted and additional miles in the five sites were collected to com- plete the mobile data collection effort for the SHRP 2 Safety study. Note these were crashes that were known to VTTI at the time of the analysis, not the final crash count for the NDS. 3.4 Data Collection This section covers the actual data collection of the mobile data project. Once the data collection plan for each of the NDS sites had been approved by SHRP 2, this information was provided to the mobile data collection vendor, Fugro Roadware (Fugro). Fugro then developed the initial routing package and worked with the research team on finalizing the routes to be driven before mobilizing equipment to a specific NDS site and proceeding with the actual data collection. The research team continued to coordinate with Fugro through- out the data collection period. Figures 3.4 through 3.9 present the roadways collected using the mobile van for each study site over the 3-year data collec- tion period. The figures show the percentage of rural and urban roads collected. A rural segment is geographically located out- side of a 2010 U.S. Census Bureauârecognized incorporated place or designated place. An urban segment is geographically located within a 2010 U.S. Census Bureauârecognized incor- porated place or designated place. See http://www.census.gov/ geo/reference/gtc/gtc_place.html for formal descriptions of incorporated places and census-designated places (CDPs). The following gives the total number of miles collected for each site, for a total of just above 25,000. Figure 3.10 shows the six study sitesâ data collection plan in comparison to the total miles in the United States. ⢠Florida: 4,366 mi. ⢠Indiana: 4,635 mi. ⢠North Carolina: 4,558 mi. ⢠New York: 3,570 mi. ⢠Pennsylvania: 3,670 mi. ⢠Washington: 4,277 mi. Figure 3.3. Example GPS traces from North Carolina in 2011. (text continues on page 18)
14 Total Miles = 4,366 Rural = 45% Urban = 55% Figure 3.4. Florida data collection plan.
15 Total Miles = 4,635 Rural = 64% Urban = 36% Figure 3.5. Indiana data collection plan.
16 Total Miles = 4,558 Rural = 63% Urban = 37% Figure 3.6. North Carolina data collection plan. Total Miles = 3,570 Rural = 68% Urban = 32% Figure 3.7. New York data collection plan.
17 Total Miles = 3,670 Rural = 83% Urban = 17% Figure 3.8. Pennsylvania data collection plan. Total Miles = 4,277 Rural = 31% Urban = 69% Figure 3.9. Washington data collection plan.
18 3.5 Data Collection requirements The mobile data collection contractor, Fugro Roadware, was required to collect the following roadway attributes. Fugro drove the right lane in each direction. ⢠Horizontal curvature: 44 Radius; 44 Length; 44 PC; 44 PT; 44 Direction of curve (left or right based on driving direction). ⢠Grade. ⢠Cross slope/superelevation. ⢠Lanes: Number, width, and type (e.g., through, turn, pass- ing, acceleration, carpool). ⢠Shoulder type/curb (and paved width, if it exists). ⢠All MUTCD signs. ⢠Guardrails/barriers. ⢠Intersection: Location, number of approaches, and control (uncontrolled, all-way stop, two-way stop, yield, signalized, roundabout). Ramp termini are considered intersections. ⢠Median presence: type (depressed, raised, flush, barrier). ⢠Rumble strip presence: location (centerline, edgeline, shoulder). ⢠Lighting presence. ⢠Videolog. Table 3.3 provides a summary of the initial required minimum accuracy requirements that were developed in the RFP for the mobile data collection project. Curve mini- mum accuracy requirements were further modified after the pilot data collection in 2011 (see Table 6.1). The pilot data collection conducted in 2011 was critical toward Figure 3.10. Miles collected per NDS site. Table 3.3. Initial Required Minimum Accuracy Data Element Accuracy Requirement Curvature length 50 ft Curvature radius 50 ft PC 25 ft PT 25 ft Grade (+ or -) 1.0% Cross slope/superelevation 1.0% Lane width 1 ft Paved shoulder width 1 ft Inventory feature location (signs and barriers) 7 ft (continued from page 13)
19 establishing the data collection process, determining feasi- bility of minimum accuracies, and testing the quality assur- ance process. 3.6 project Management plan The two contracting teams (Iowa State University and Fugro) followed a management plan, approved by SHRP 2, that included a communication plan, delivery schedule, data deliv- ery format and procedures, an asset-marking guide, and a data dictionary to ensure smooth communication among project stakeholders and timely delivery of data by the mobile data contractor. The communication plan outlined the requirements based on roles for each of the members of the project team, which was composed of one key member from SHRP 2, the Iowa State University research team, and Fugro Roadware, supported by their respective internal teams. The communi- cation matrix clearly identified the who, what, when, and how in terms of how specific communications were to take place in order to keep all interested parties well informed and clear of any ambiguities that could render the informa- tion confusing or inaccurate. Figure 3.11 shows the commu- nication matrix. In addition to the communication plan, the two contrac- tors developed a data collection and delivery format manual. Type of Information Prepared By Distribution List Purpose of Communication Frequency Transmittal Method Kick Off Meeting S04A, S04B Project Team, Project Sponsor Clarify goals and objectives, individual roles and responsibilities Once at project start-up Meeting in D.C. Monthly Reports S04B SHRP 2 Discuss status, issues and concerns related to the Project Monthly Email Quarterly Reports S04B SHRP 2 Communicate issues/concerns, schedule, deliverables, milestones, etc. Quarterly Email Project Schedule S04B S04A, SHRP 2 Document and monitor key tasks, milestones, issues/concerns. Document site data acquisition, delivery, and re-runs Monthly Email, Project Web Page Quality Control Plan S04B S04A, SHRP2 Document the quality control procedures for data collection including calibration, collection procedures, and retrospective examination prior to submission Once at project start-up Email, Project Web Page Data Collection Routes S04A SHRP 2, S04B Database of routes to be collected, in a GIS enviroment, for each of the study sites. Delivered a minimum of 3 weeks in advance of data collection per site.2011 (NY, NC, FL), 4-6 weeks in advance of collection 2012 (All), 2013 Email, Project Web Page Control Site Identification S04A SHRP 2, S04B List of control sites for each study site to be evaluated according to S04B's contract Minimum of one week In advace of the first data collection for each site Email, Project Web Page Data Collection Schedule S04B S04A, SHRP 2 Document site data acquisition, delivery, re-runs, and control site survey information Weekly Updates Initially this will be by a scheduled conference call. As the project progresses, Control Site Data Submission S04B S04A, SHRP 2 Collected data from control sites One week after start of data collection. Email, Project Web Page Control Site Quality Assessment S04A SHRP 2, S04B Feedback on control site data quality One-weeks after submittal from S04B Email, Project Web Page Random Assessment S04A SHRP 2, S04B Feedback on the quality of collected data As needed Email, Project Web Page Data Delivery S04B S04A Delivery of all data collected and formatted according to the roadway information database Minimum of 500 miles every two weeks based on the data delivery schedule Email, Project Web Page Initial Overall Data Quality Assessment S04A SHRP 2, S04B Feedback on quality for all collected data based on automated processes Two-weeks after submittal from S04B Email, Project Web Page Overall Data Quality Assessment S04A SHRP 2, S04B Feedback on quality for all collected data including control sites and random checks Two-months after submittal from S04B Email, Project Web Page Invoice Request S04B SHRP 2, S04A Submit invoice for payment based on overall data quality assessment Once per month Email, Project Web Page Change Requests S04B SHRP 2, S04A Communicate, receive approval and document status of all change requests. As Needed Email, Conference call Figure 3.11. Mobile data collection project communication matrix.
20 Figure 3.12 shows an example for capturing speed limit sign location and message information. Data deliveries from the mobile data collection vendor to the Iowa State University research team were provided in batches of approximately 500-mile increments every two weeks, based on a preestablished delivery schedule. These deliveries were in Figure 3.12. Mobile data collection manual. Microsoft Access format and were uploaded directly to the proj- ect FTP site. Right-of-way (ROW) images from the videolog were delivered on portable media to the research team. Follow- ing each delivery, the research team proceeded with the data quality assurance process, which is discussed later in the report under the chapter on Mobile Data Quality Assurance.
