CHAPTER 5. STUDY DATA

This chapter presents four types of data that were gathered, processed, and used in this study to apply and validate the methods presented in Chapter 4.

• CCS data were used in all methods.
• Probe data were used in Methods 18 and 19.
• Census data were used in Methods 13–17.
• OpenStreetMap (OSM) network data were used in Methods 18 and 19.

Each data type is described in a separate section of this chapter.

CCS DATA

The research team downloaded 2011–2022 CCS data for all states from FHWA’s Travel Monitoring Analysis System. For each CCS and year, five quality control checks, as recommended and applied by the FHWA Office of Highway Policy Information, were used to identify and exclude potential erroneous values and atypical patterns and volumes:

• 10 or more consecutive hours with 0 volume for all CCSs within functional classes 1R, 2R, 3R, 4R, and 1U–7U.
• 100-fold change in volume from one hour to the next.
• Hour with volume of 50 or greater adjacent to an hour with 0 volume.
• Hourly volume of greater than 2,500 vehicles per lane.
• Directional split greater than 60–40 at the daily level.

Records meeting at least one of these conditions were excluded from the analysis. The next step was to select the states, years, and CCSs to be used in the analysis based on the following criteria:

• The CCSs met the minimum data completeness requirements for calculating AADT, as recommended in the TMG (FHWA 2022). Specifically, for the higher FCs 1R–4R and 1U–4U, all selected CCSs had at least three hourly volumes for all 24 hours of every day of the week in any given month and year. For the lower FCs, the selected CCSs had at least two hourly volumes for all 24 hours of every day of the week in any given month and year.
• Each state had 10 or more CCSs/year for a minimum of three years. This criterion was necessary to apply the clustering methods and produce meaningful results that would contain at least two clusters per state and year. Analyzing three or more years for each state was necessary to increase the sample size of the analysis and examine potential differences in the performance of the assignment methods over time.
• CCSs from all roadway functional classes and traffic volumes were selected, as stated in the project objective.
• The selected states had different traffic, weather, and geographical characteristics.

The research team selected 45 states and three years per state to be used in the analysis. The selected states and the number of CCSs per state, year, functional class, and area type are shown in Table 20. The records are sorted first by state in alphabetical order, and then within each state, the records are sorted by the total number of CCSs (last column) in descending order. For each state, the table shows the number of CCSs for the three years with the highest number

of stations—these are the CCSs and years that were used to apply and validate Methods 1–17. In the case of Minnesota, Ohio, and Texas, CCSs from more than three years were used to match the probe data obtained in this project for examining Methods 18 and 19.

Table 20. Selected 45 States—Number of CCSs by State, Year, Functional Class, and Area Type.

^*Data from these years were used only in Methods 18 and 19 to match the years of the probe data that were analyzed in this study.

No CCS was available for functional class 6U (urban minor collectors). All other functional classes were included in the analysis. The research team used data from over 10,000 CCS and year combinations for 45 states.

PROBE DATA

As previously mentioned, the research team obtained probe data from three states—one vendor per state. The probe data refer to the raw (unadjusted) number of probe device trips along short segments where CCSs exist. Table 21 shows the main characteristics of the probe data, including the three states, the vendors, the temporal resolution, the FHWA vehicle class groups, the years, and the number of CCS locations per year.

Table 21. Main Characteristics of Probe Data Used to Apply Methods 18 and 19.

The probe data from Texas and Ohio contained 365 daily counts per CCS, with each count accounting for all vehicle classes combined (1–13). On the other hand, the probe data from Minnesota included three sets of 12 monthly probe counts per CCS: (a) one set for all 13 vehicles classes combined, (b) a second set for medium-duty trucks (vehicle classes 4–6), and (c) a third set for heavy-duty trucks (vehicle classes 7–13).

Table 22 shows the number of CCSs for which probe counts were provided for all 13 vehicle classes together. The number of CCSs is broken down by functional class and area type. The CCSs shown in Table 22 were used to apply Method 18, which involved annualizing SDCs using segment-specific probe-based adjustment factors developed for all 13 vehicle classes as one group.

Table 22. Number of CCSs Used to Apply and Validate Method 18.

Table 23 shows the number of classification CCSs for which monthly probe counts were separately provided for medium-duty trucks (vehicle classes 4–6) and heavy-duty trucks (vehicle classes 7–13). The CCSs shown in Table 23 were used to apply Method 19, which involved

annualizing counts using segment-specific probe-based adjustment factors developed for medium- and heavy-duty trucks.

Table 23. Number of CCSs Used to Apply and Validate Method 19.

CENSUS DATA

For all 45 states, CCSs, and years included in Table 20, the research team downloaded (from https://api.census.gov) and processed the following American Community Survey (ACS) attributes: total population, number of occupied housing units, employment, number of workers who did not work at home, and land area of the census tract where each CCS is located. ACS data include annual estimates of census variables. As explained in the previous chapter, after downloading the data, the geographical density of each attribute was calculated at the census tract level.

OSM TRANSPORTATION NETWORK

The research team downloaded (from https://download.geofabrik.de/), processed, and used the latest OSM transportation networks for Minnesota, Ohio, and Texas to identify the OSM segments where CCSs exist and then obtain probe data for these segments. In the case of Minnesota and Ohio, a polygon was created around each segment where CCSs were located. The probe data were obtained for the polygons of interest. In the case of Texas, the research team mapped waypoint probe data along each polygon and then determined the total number of daily trips of probe devices along each polygon.