Automated Data Collection and Quality Management for Pavement Condition Reporting (2022)

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.

35   A questionnaire was developed to determine agency practices for pavement condition surveys, use of pavement performance measures, and pavement condition reporting requirements. The questionnaire was provided to highway agencies in all 50 U.S. states, the District of Columbia, and Puerto Rico. Agencies were asked several questions related to practices regarding pavement condition surveys, APCS use and challenges, use of various performance measures for pavement condition, and modifications and challenges in meeting national reporting requirements. The intended recipients of the agency questionnaire were pavement management engineers (or comparable positions) responsible for conducting, analyzing, and reporting pavement con- dition information. The detailed questionnaire is provided in Appendix A and agency responses are summarized in Appendix B. As of May 2021, 48 SHAs (92%), including Puerto Rico and the District of Columbia, responded to the questionnaire. The following provides a summary of the questionnaire results. Pavement Condition Data Quality Management Plans As described in Chapter 2, each SHA was required to develop and submit a pavement condi- tion DQMP to FHWA by May 2018. Pierce and Weitzel (2019) provide a summary of agency practice related to pavement condition survey methods, the types of distress and conditions collected, and data quality control and acceptance requirements. Rather than ask SHAs to respond to additional questions or provide DQMPs for current pavement condition, agencies were asked to indicate if any updates had occurred since the original submittal of the pavement condition DQMP to FHWA. In total, 45 agencies indicated no significant changes were made and 3 agencies indicated updates have been conducted on the originally submitted DQMP: • California: Updated on the basis of FHWA 2018 guidelines (FHWA 2018a), • Florida: Modified to include statistical validation for comparing field verification site data (rut depth, percent cracking, and faulting), and • New Hampshire: Updated due to the outsourcing of the pavement condition survey to a data collection vendor. Survey Methods Agencies indicated the use of various methods for conducting the pavement condition survey (Figure 15). Of the 48 responding agencies, 17 use a combination of fully and semiautomated surveys, 13 use fully automated surveys, 7 use a combination of manual and fully automated surveys, 6 use semiautomated surveys, 3 use a combination of manual and semiautomated surveys, and 2 use manual pavement condition surveys. C H A P T E R   3 State of the Practice

36 Automated Data Collection and Quality Management for Pavement Condition Reporting Automated Pavement Condition Surveys For agencies that conduct semi- and fully automated APCSs, the majority (24 agencies) have been doing so for more than 10 years (Figure 16). An approximately equal number of agencies have conducted APCSs for less than 5 years (9 agencies) and from 5 to 10 years (10 agencies). When asked who conducts data collection for the pavement condition survey, 24 agencies indicated that it is conducted by a vendor, 14 that it is conducted by the agency using agency equipment, and 10 that it is conducted by both the agency and the vendor (Figure 17a). Agencies were also asked to indicate who processes and analyzes the pavement condition survey. Eighteen agencies indicated that they process and analyze the condition data, 16 that both the agency and the vendor process and analyze the condition data, and 14 that the vendor processes and analyzes the data (Figure 17b). Figure 18 provides a summary of who conducts data collection (e.g., vendor or agency) and who conducts data processing and analysis for the pavement condition survey. The pre dominant Number of responses: 48 2 3 6 7 13 17 0 5 10 15 20 Manual Manual and semiautomated Semiautomated Manual and fully automated Fully automated Fully and semiautomated No. Agencies Figure 15. Method used for conducting pavement condition surveys. Number of responses: 43 Figure 16. Number of years conducting APCSs.

State of the Practice 37   methods include using agency staff and equipment to conduct data collection, processing, and analysis (14 agencies) and contracting with vendors to conduct data collection, processing, and analysis (14 agencies). The remaining agencies use the following combinations of methods: • Agency and vendor both conduct data collection, processing, and analysis—9 agencies. • Vendor conducts data collection and both the agency and the vendor conduct data processing and analysis—7 agencies. • Vendor conducts data collection and the agency conducts data processing and analysis— 3 agencies. • Both the agency and the vendor conduct data collection and the agency conducts data pro- cessing and analysis—1 agency. Number of responses: 48 a. .bn.oitcelloc ataD Data processing and analysis. Vendor, 14 Agency, 18 Both vendor and agency, 16 Figure 17. Pavement condition data collection and processing and analysis. Number of responses: 48 Figure 18. Entity conducting pavement condition data collection, processing, and analysis.

38 Automated Data Collection and Quality Management for Pavement Condition Reporting The information illustrated in Figure 18 is summarized below and in Table 24 according to the method of pavement condition survey (i.e., fully automated, semiautomated, manual, and various combinations): • Of the agencies that use agency staff and equipment to conduct the pavement condition survey, 5 use fully automated methods, 3 use a combination of fully and semiautomated methods, 2 use manual methods, and the remaining 4 use a combination of manual and fully automated (2 agencies) or manual and semiautomated (2 agencies) methods. • Of the agencies that contract with vendors to complete the pavement condition survey, 6 use fully and semiautomated methods, 4 use fully automated methods, 2 use semiautomated methods, and 2 use manual and fully automated methods. • Of the agencies that reported that both the agency and the vendor collect, process, and ana- lyze the pavement condition survey, 4 use fully and semiautomated methods, 3 use manual and fully automated methods, 1 uses fully automated methods, and 1 uses semiautomated methods. • Of the agencies that indicated that the vendor collects the data and both the agency and ven- dor process and analyze it, 3 use fully automated methods, 2 use fully and semiautomated methods, and 2 use semiautomated methods. • Of the agencies that indicated that the vendor conducts the data collection and the agency processes and analyzes it, 2 use fully and semiautomated methods and 1 uses manual and semiautomated methods. • The single agency that indicated that both it and the vendor collect the data and that the agency processes and analyzes it uses a semiautomated method. Automated Data Collection Vehicles The reported number of data collection vehicles for conducting the APCS versus the total number of lane miles included in the annual pavement condition survey is illustrated in Fig- ure 19. While there is a considerable range in the number of vehicles used to collect an APCS, for most agencies, one data collection vehicle is used for every 10,000 lane miles surveyed per year. In general, the number of vehicles used for a vendor-collected APCS is higher than that in agency-collected and agency-and-vendor-collected APCSs; however, several agencies indicated the number of vehicles used by the vendor is increased to meet the data collection schedule. Implementation of Automated Pavement Condition Surveys The SHA transition from manual surveys to APCSs has been the trend over the past several decades (Pierce and Weitzel 2019). However, adoption of an APCS typically includes additional Entity Conducting Pavement Condition Survey Method Total Data Collection Processing & Analysis Fully Fully & Semi. Manual & Fully Semi. Manual & Semi. Manual Agency Agency 5 3 2 0 2 2 14 Vendor Vendor 4 6 2 2 0 0 14 Agency & vendor Agency & vendor 1 4 3 1 0 0 9 Vendor Agency & vendor 3 2 0 2 0 0 7 Vendor Agency 0 2 0 0 1 0 3 Agency & vendor Agency 0 0 0 1 0 0 1 Total responses 13 17 7 6 3 2 48 Note: semi. = semiautomated. Table 24. Responsibility for data collection, processing, and analysis.

State of the Practice 39   activities. Figure 20 provides a summary of agency responses to a variety of activities in prepara- tion of APCS implementation. Approximately half the agencies indicated assessing data quality issues (24 agencies), modifying the pavement index or rating calculations (20 agencies), and modifying distress definitions (19 agencies). Approximately a third of the agencies indicated adding index or rating calculations (15 agencies), adding distress definitions (14 agencies), developing a correlation with manual pavement condition surveys (13 agencies), and modify- ing pavement treatment decision trees (12 agencies). Eight agencies indicated no modifications were needed, eight added additional “branches” to the treatment decision tree, and four noted modifying historical manual pavement condition surveys. Other agency responses on modification required to implement APCS included the following: • Developed distress indices to include patching and raveling data and implemented an over- all level of service calculation. • Retired all historical distress data as it became irrelevant or incomplete for comparison with APCS results. • Developed an agency-specific APCS rating manual for data collection. • Have conducted APCS for a long time and annually adjust process for the betterment of the data. • Developed workflow for processing and reporting raw data from data collection vehicles. • Annually review decision trees and, every few years, compare automated cracking results (mainly the calculation of cracking index values) with engineering judgment. With PM2, started looking more at the raw cracking data (length and width). • Currently researching alternative methods for quantifying crack surveying. The impetus for this research has been the introduction of more fully automated survey technology (includ- ing 3D) within the collection vendor community. The traditional crack and severity-level method is old, in need of updating, and cumbersome to execute relative to the technology readily available from the vendor pool currently. Number of responses: 34 R² = 0.8408 - 2 4 6 8 10 12 - 20,000 40,000 60,000 80,000 100,000 N um be r o f D at a Co lle cti on V eh ic le s Annual Lane Miles Surveyed per Year Agency Vendor Both Figure 19. Number of data collection vehicles per annual lane mile surveyed.

40 Automated Data Collection and Quality Management for Pavement Condition Reporting Of those agencies that have transitioned to APCS, the majority (30 agencies) noted challenges with validation and verification of data quality (Figure 21). Approximately half the agencies also indicated having challenges with the consistency of results (22 agencies), significant dif- ferences compared with manual pavement condition surveys (19 agencies), integrating auto- mated pavement condition results in the pavement management system (18 agencies), and training staff on data analysis (17 agencies). Several agencies also noted that coordination of data col- lection activities (14 agencies), training staff on data collection (12 agencies), and reduced data quality (2 agencies) were also challenging. Six agencies indicated no challenges with imple- menting an APCS. Other noted agency challenges included • Historical manual pavement condition data that did not correlate with the data collected using APCS, • Current evaluation of year-to-year consistency of APCS results, • Storage problems, and • No indication from business process owner to move away from manual surveys. Agencies were asked to select from a list of potential disadvantages to APCS (Figure 22): • Approximately half the agencies indicated that APCS disadvantages include missing a direct correlation with historical data (18 agencies) and increased costs for data collection (17 agencies). • About a third of the agencies noted disadvantages associated with year-to-year data variability (14 agencies), increased processing costs (13 agencies), costs associated with performance model modification (12 agencies), and additional cost and personnel hours for procuring equipment and vendors (12 agencies). Number of responses: 41 4 8 8 12 13 14 15 19 20 24 0 10 20 30 Modify historical manual surveys Add decision trees (branches) No modifications needed Modify decision trees Develop correlation with manual surveys Add distress definitions Add index/rating calculations Modify distress definitions Modify index/rating calculations Assess data quality issues Number of Agencies Figure 20. Agency modifications to implement APCS.

Number of responses: 42 2 6 12 14 17 18 19 22 30 0 10 20 30 40 Reduced data quality No challenges Training staff on data collection Coordination of data collection activities Training staff on data analysis Integrating results into the pavement management system Significant difference compared to manual surveys Consistency of results Validation/verification of data quality Number of Agencies Figure 21. Challenges with transitioning to APCS. Number of responses: 41 2 4 5 7 8 9 9 9 11 12 12 13 14 17 18 0 5 10 15 20 Added costs for modifying decision trees Breakdowns and long repair delays Added costs for modifying distress ratings Added costs for modifying distress manual Added costs for modifying pavement management software Difficulties with operational changes Added costs/personnel for calibration, validation, and verification Technology evolution, forcing early equipment replacement Dependence on a single vendor Added costs and personnel for equipment or vendor procurement Added costs for modifying performance models Increased processing costs Year-to-year variability of results Increased collection costs Direct correlation with historical data Number of Agencies Figure 22. Disadvantages of APCS.

42 Automated Data Collection and Quality Management for Pavement Condition Reporting • Other disadvantages included dependence on a single vendor (11 agencies); technology evaluation forcing early equipment replacement (9 agencies); additional costs and personnel for calibration, validation, and verification (9 agencies); difficulties with operational changes (9 agencies); additional costs for modifying pavement management software (8 agencies); additional costs for modifying distress manuals (7 agencies); additional costs for modifying distress ratings (5 agencies); equipment breakdowns and long repair delays (4 agencies); and additional costs for modifying decision trees (2 agencies). Other agency-noted disadvantages of APCS included • Staff training and retention; • Many items appearing inevitable with newer 3D technology, although they should be one- time local costs; • Changing vendors and software; and • Additional data storage requirements and costs. The agencies noted a number of advantages with APCS (Figure 23). More than half the agencies responded to all listed advantages of APCS, including the ability to collect data com- patible with HPMS and PM2 (36 agencies); increased rater safety (31 agencies); the ability to collect sensor (e.g., IRI, rutting, faulting) and distress data with a single device (30 agencies); the ability to easily track, review, and reproduce historical data and images (28 agencies); well- defined data collection methods (27 agencies); improved accuracy of distress identification (27 agencies); 100% roadway coverage (27 agencies); enhanced timeliness of data processing (25 agencies); enhanced timeliness of data collection (25 agencies); and access to ancillary data collection (19 agencies). Number of responses: 43 19 25 25 27 27 27 28 30 31 36 0 10 20 30 40 Access to ancillary data collection Enhanced timeliness of data collection Enhanced timeliness of data processing 100% roadway coverage Improved accuracy of distress identification Well-defined data collection methods Ability to easily track, review, and reproduce historical data and images Ability to collect sensor and distress with a single device Increased rater safety Ability to collect data compatible with HPMS and PM2 Number of Agencies Figure 23. Advantages of APCS.

State of the Practice 43   Additional agency-noted advantages to APCS include • Ease of use within the geographic information system (GIS); • More fully automated identification of crack type and severity from 3D technology and, possibly, more consistency in the annual data set and from year to year; • Provision of frequent updates of right-of-way (ROW) imagery, surface macrotexture, local- ized roughness assessments, and many other transportation asset management benefits; • Improved data quality control and acceptance; and • Accurate results and efficient identification of certain distress types achieved with automated routines. Lane Miles and Cost The agencies were asked to provide the total number of annual lane miles included in the annual pavement condition survey along with the estimated total annual cost (i.e., from start-up to integration of results into the pavement management system). Figure 24 summarizes the number of lane miles included in the annual pavement condition survey. Of the 46 respond- ing agencies, 28 agencies conduct pavement condition surveys on less than 15,000 lane miles, 11 agencies evaluate 15,000 to 30,000 lane miles annually, and 7 agencies evaluate more than 30,000 lane miles annually. As noted, the agencies were asked to provide the total annual cost for collecting, analyzing, and reporting the results of the pavement condition survey. The estimated cost is based on the agency-reported annual cost divided by annual lane miles collected. Figure 25 represents all agency responses regardless of the survey method (e.g., manual, semiautomated, fully auto- mated) or who conducts the work (e.g., agency versus vendor). Cost is categorized by annual lane miles because of potential differences in the economy of scale. In general, as the number of lane miles surveyed increases, the cost per lane mile decreases. For annual pavement condition surveys of less than 5,000 lane miles per year, there is a large spread in lane mile costs, ranging from $43 to $429 per lane mile, with a median value of Figure 24. Annual lane miles surveyed. Number of responses: 46 10 10 8 4 7 7 5 ≤ 5 5 - 10 10 - 15 15 - 20 20 - 30 ≥ 30 N o. o f A ge nc ie s Annual Lane Miles Surveyed (x 1,000)

44 Automated Data Collection and Quality Management for Pavement Condition Reporting $131 per lane mile (10 agencies). For pavement condition surveys ranging in length from 5,000 to 10,000 lane miles, the range in cost is $38 to $188 per lane mile with a median value of $96 per lane mile (7 agencies). For pavement condition surveys of more than 10,000 lane miles, the range of values from the first to third quartile is significantly reduced as compared with pavement con- dition surveys of less than 10,000 annual lane miles. In addition, the median cost per lane mile for more than 10,000 surveyed lane miles ranges from $53 to $68 per lane mile (24 agencies). Table 25 summarizes the average weighted cost per lane mile according to the pavement con- dition survey method. The highest average weighted cost is $104 per lane mile for manual pave- ment condition surveys; however, this is based on responses from only 2 agencies with a very broad range in cost per lane mile. The average weighted costs per lane mile for semi- and fully automated APCSs are relatively close, at $72 per lane mile and $69 per lane mile, respectively. The average weighted costs were $67 per lane mile for agencies that conduct both manual and fully automated surveys, $56 per lane mile for agencies that conduct semi- and fully automated surveys, and $51 per lane mile for agencies that conduct manual and semiautomated surveys. Regardless of the method of pavement condition survey, the average weighted cost of all the responding agencies was $62 per lane mile. Data Storage Requirements With the use of the APCS comes a large volume of data and images. Agencies were asked to indicate how they were storing data and images and what size of data storage drive they needed to store the results of the annual pavement condition survey. Figure 26 illustrates the number of agencies that use data storage drives (34 agencies), cloud storage (7 agencies), and a combination of both data drive and cloud storage (6 agencies). Additional agency comments in relation to data storage included the following: • “Hybrid approach is used for now; short term includes on-premise storage, while future stor- age will be cloud based (anticipated in 2022). Agency has had challenges with implementation of cloud-based storage due to slow upload speeds.” • “Raw pavement distress data [are] stored on storage drives, while pavement images are stored [in] the cloud.” Number of responses: 41 $- $50 $100 $150 $200 $250 $300 $350 $400 $450 ≤ 5 5 - 10 10 - 15 15 - 20 20 - 30 ≥ 30 Co st p er L an e M ile Lane Miles Surveyed (x 1,000) Figure 25. Pavement condition survey cost per lane mile.

State of the Practice 45   Survey Method Data Collection Data Analysis No. of Agencies Weighted Cost per Lane Mile ($) Average Total Average Manual Agency Agency 2 104 104 Semiautomated Both Agency 1 60 72 Both Both 1 100 Vendor Both 2 83 Vendor Vendor 2 84 Fully automated Agency Agency 2 26 69 Both Agency 1 43 Vendor Both 3 65 Vendor Vendor 3 114 Manual and fully automated Agency Agency 2 33 67 Both Both 2 58 Vendor Vendor 2 124 Semi- and fully automated Agency Agency 3 88 56 Both Both 2 65 Vendor Agency 2 28 Vendor Both 2 43 Vendor Vendor 5 69 Manual and semiautomated Agency Agency 2 45 51 Vendor Agency 1 58 Total All All 40 na 62 Note: na = not applicable. Table 25. Average weighted cost per lane mile by condition survey method. Number of responses: 47 Data storage, 34 Cloud storage, 7 Both, 6 Figure 26. Storage requirements for pavement condition survey data. Figure 27 illustrates the agencies’ responses for the number of drive storage needs based on the annual lane miles collected. In general, for every 10,000 lane miles collected annually, approximately 10 terabytes of data storage is needed. Sampling and Acceptance As noted in Chapter 2, PM2 required agencies to establish requirements for data quality and acceptance for the pavement condition survey. Agencies were asked to provide the sampling

46 Automated Data Collection and Quality Management for Pavement Condition Reporting rate as a percentage of total annual lane miles collected and the estimated hours to complete the activities for acceptance of the annual pavement condition survey. Table 26 summarizes agency- reported sample sizes versus total annual lane miles surveyed. Of the 41 responding agencies, 14 use a sample size of less than 5% of the total annual lane miles, 10 use a sample size of 5% to 15%, 4 use a sample size of 15% to 25%, 1 uses a sample size of 50% to 75%, and 12 use a sample size of 75% to 100%. The agency questionnaire did not request details related to acceptance practices, since these activities differ by agency and were previously summarized by Pierce and Weitzel (2019). However, it should be noted that agencies’ acceptance testing practices included both data and image review activities. Figure 28 illustrates the agency-reported hours necessary to complete acceptance testing divided by the number of lane miles evaluated for acceptance by survey method. The responses exclude the one responding agency that conducts a manual survey and the one agency that uses all three survey methods. The range in responses is quite broad for agencies using both semi- and fully automated, and manual and fully automated methods. The time to conduct acceptance Number of responses: 31 R² = 0.7595 - 10,000 20,000 30,000 40,000 50,000 60,000 0 10 20 30 40 50 60 An nu al L an e M ile s S ur ve ye d Data Storage Requirements (TB) Figure 27. Storage requirements for pavement condition survey data. Total APCS Lane Miles Number of Agencies, by Sample Size <5% 5%–15% 15%–25% 50%–75% 75%–100% Total <5,000 5,000–10,000 10,000–15,000 15,000–20,000 20,000–30,000 >30,000 Total 1 1 1 1 — — 4 — — — — 1 — 1 1 3 4 1 1 2 12 5 10 8 4 7 7 41 — 3 — 1 2 4 10 3 3 3 1 3 1 14 Note: Sample size = percentage of total annual lane miles. Table 26. Acceptance testing sampling rate.

State of the Practice 47   testing is similar for agencies that use a combination of semi- and fully automated methods (12 agencies) compared with agencies that use only fully automated methods (10 agencies), with median values of 0.27 hours per lane mile and 0.36 hours per lane mile, respectively. The lowest time to complete acceptance testing was found for agencies that use a combination of manual and fully automated methods (6 agencies) at a median value of 0.07 hours per lane mile. Agencies that use a combination of manual and semiautomated surveys (3 agencies) complete acceptance testing at a rate of 0.09 hours per lane mile. Five agencies that conduct semiautomated surveys complete acceptance testing at a rate of 0.25 hours per lane mile. Agencies were also asked to provide any noted challenges with quality control and acceptance of data and images. A summary of agency responses is provided in Figure 29. The majority of agencies (27 agencies) indicated correct crack detection as a challenge with quality control and acceptance of images. Other common challenges included collection and reporting of crack- ing data (17 agencies), adjusting algorithms to meet requirements (17 agencies), missing data (15 agencies), correct condition data by surface type (13 agencies), poor quality of pavement images (11 agencies), meeting control and verification site requirements (10 agencies), and incorrect location (10 agencies). Less-common challenges included collection and reporting of faulting data (8 agencies), data out of range (7 agencies), collection and reporting of IRI data (5 agencies), collection and reporting of rutting data (4 agencies), and meeting random sample requirements (4 agencies). Two agencies reported no challenges and one reported a challenge with the correct format. Other agency-reported challenges to data and image quality included the following: • Per the DQMP, left and right IRI values should not differ by more than 50 inches/mile. At many locations, the difference between left and right IRI may be due, for example, to unstable permafrost or poor curb and gutter conditions. Determining the cause of the outliers results in increased investigation time. • For variable slab length, accurate determination of faulting values can be challenging. • The agency added the capability to measure faulting in 2015 but has had trouble with this metric and will establish validation testing against ground truth measurements this year. • Faulting data are dependent on joint identification. Number of responses: 36 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 Full and Semi Full Manual and Full Manual and Semi Semi Ho ur s p er L an e M ile Survey Method Figure 28. Time needed to complete acceptance testing.

48 Automated Data Collection and Quality Management for Pavement Condition Reporting • The image location is incorrect. • Semiautomated rated distresses such as patching and raveling are the biggest challenge. • Occasionally the location is incorrect because of issues with the linear referencing system and occasional poor quality of the pavement image; however, these are not widespread or major issues. Reporting Pavement Condition Performance Measures Agencies determine a variety of pavement performance measures to assist in describing the current pavement condition. Figure 30 provides a summary of agency pavement performance measures. As anticipated, due to PM2 requirements, the highest-reported performance mea- sures were IRI (42 agencies), average rut depth (36 agencies), percent cracking (35 agencies), and faulting (35 agencies). The next two highest-reported indices (30 agencies each) were asphalt pavement cracking (percent area) and JPCP cracking (percent slabs). More than half the agencies reported that they determined an agency-specific combined index (26 agencies) Number of responses: 44 1 2 4 4 5 7 8 10 10 11 13 15 17 17 27 0 10 20 30 Correct format No challenges Meeting random sample requirements Rutting data collection and reporting IRI data collection and reporting Data out of range Faulting data collection and reporting Incorrect location Meeting control and verification site requirements Poor pavement image quality Correct condition data by surface type Missing data Adjusting algorithms to meet requirements Cracking data collection and reporting Correct crack detection Number of Agencies Figure 29. Challenges with quality control and acceptance of data and images.

State of the Practice 49   and an agency-specific cracking index (21 agencies). More than one-third of the agencies indi- cated determining a rutting index (18 agencies), percent area of CRCP cracking (16 agencies), faulting index (12 agencies), and load- and non-load-related distress indices (10 agencies each). Lesser-determined indices included the PCI (9 agencies), remaining service life or interval (5 agencies), PSR (4 agencies), and agency-modified PCI (3 agencies). Other pavement performance measures and indices included • Collection of data on patching and raveling, with plans to incorporate these into existing indices; • Multi-cracked slabs for JPCP; • Work on the development of a new pavement performance index based largely on remaining service life; • Cracking-severity index (“Distress Index”) by year, but agency is in the process of researching alternatives for future use; • Agency-specific combined index with indices for alligator, block, longitudinal, and transverse cracking; rutting; patching; and IRI; and • Travel weighted average condition (condition index weighted by annual average daily truck traffic). Number of responses: 47 3 4 5 9 10 10 12 16 18 21 26 30 30 35 35 36 42 0 10 20 30 40 50 Modified PCI PSR Remaining service life or interval PCI Non-load-related distress index Load-related distress index Faulting index CRCP cracking (% area) Rutting index Agency-specific cracking index Agency-specific combined index JPCP cracking (% slabs) Asphalt pavement cracking (% area) Average faulting (in.) Percent cracking Average rut depth (in.) IRI (in./mi) Number of Agencies Figure 30. Performance measures and indices.

50 Automated Data Collection and Quality Management for Pavement Condition Reporting PM2 Requirements Considering that the primary focus of PM2 was to improve and ensure the quality of pave- ment condition, most agency responses were in relation to development, implementation, and documentation of the components of the DQMP (e.g., quality control, calibration, staff train- ing). As shown in Figure 31, to comply with PM2 requirements, the majority of agencies devel- oped a DQMP (42 agencies). More than half of the responding agencies noted they developed data quality control requirements (29 agencies); established control, verification, and blind site sections (27 agencies); developed data acceptance requirements (26 agencies); and developed a process for equipment calibration and certification (21 agencies). Twenty agencies added PM2 definitions, 18 added PM2 indices, and 10 changed their distress definitions for the pave- ment condition survey. Sixteen agencies indicated implementing APCS in response to PM2 requirements. Approximately one-quarter of the agencies developed a rater certification pro- gram (8 agencies) and rater training program (7 agencies). Seven agencies indicated the need to change how indices were calculated, and 3 agencies indicated no modifications were needed to comply with PM2 requirements. Number of responses: 47 3 7 7 8 10 16 18 20 21 26 27 29 42 0 5 10 15 20 25 30 35 40 45 No modifications were required Changed indices calculation Developed a rater training program Developed a rater certification program Changed distress definitions Implemented automated pavement condition surveys Added indices for PM2 Added PM2 definitions Established a process for data collection equipment calibration and certification Established data acceptance requirements and procedures Established control, verification, and/or blind site sections Developed data quality control requirements and procedures Developed a DQMP Number of Agencies Figure 31. Activities to comply with PM2 requirements.

State of the Practice 51   Other agency responses included the following: • All data collection and management is done with GIS. • An annual rater training for manual cracking is used for verification comparisons for HPMS and agency data collection needs. • Calibration, processes, and procedures were in place prior to the PM2 requirements; how- ever, they were not formalized in one document until the DQMP was developed. Calibration procedures had to be expanded and a training log maintained for all staff involved with data collection and processing. • Data collection process met most of the PM2 requirements, but some elements have been added. • Established equipment calibration and certification requirements have been enhanced. • Blind sites were added to the existing control sites. • Automated distress is used as a starting point; all automated routines are manually reviewed. Agencies identified a number of challenges in implementing the PM2 requirements (Fig- ure 32). Approximately half the agencies noted challenges with establishing SHA performance targets (22 agencies), needing to adjust performance targets (12 agencies), and establishing pavement performance targets for MPOs (11 agencies). Fifteen agencies had difficulties defin- ing baseline condition, 9 noted having insufficient data, 8 had challenges meeting performance targets, 7 had challenges meeting new Interstate requirements, and 3 noted extenuating circum- stances. Eleven agencies reported they had no challenges meeting the PM2 requirements. Additional PM2 reporting challenges included • An increased level of quality control and acceptance, but challenges in meeting the require- ments with the current staffing level; • Moving from manual to semiautomated systems for measuring pavement distress; Number of responses: 44 3 7 8 9 11 11 12 15 22 0 5 10 15 20 25 Extenuating circumstances New interstate requirements Meeting targets Insufficient data Establishing MPO performance targets No challenges Needing to adjust targets Determining baseline condition Establishing SHA performance targets Number of Agencies Figure 32. Challenges with PM2 reporting requirements.

52 Automated Data Collection and Quality Management for Pavement Condition Reporting • Insufficient data initially—mostly for the local NHS; only IRI had been collected; • Time and effort for collecting new data specific to the HPMS (e.g., percent cracking); • A disconnect between agency and NHS performance measures; • Agency performance measures that are significantly different from NHS performance mea- sures with regard to percent good and percent poor; and • Separate pavement performance models based on FHWA performance measures (i.e., IRI, faulting, rutting, percent cracking) and on agency models that are based on agency pave- ment condition measures (e.g., IRI, faulting, rutting, alligator cracking, longitudinal cracking, transverse cracking, raveling). HPMS Requirements As shown in Figure 33, the majority of responding agencies (24 agencies) indicated no chal- lenges in reporting HPMS requirements. Other agencies reported challenges in determining percent cracking (15 agencies), reporting percent cracking (6 agencies), reporting faulting (4 agencies), and determining PSR and IRI (2 agencies each). Additional agency challenges with HPMS reporting requirements included the following: • A difference between the agency and HPMS definitions of fatigue cracking and determination of percent cracking; • Limitations of current APCS technology; notably, spalling on CRCP and other surface dis- tresses is difficult for current technology to identify accurately; • Data collection challenges; missing required data; • HPMS requirements that are different from agency internal reporting metrics; • Reporting requirements and additional workload are not useful for pavement management; • HPMS change in width of wheel path to 1 meter caused challenges, including discontinuity with more than 14 years of historical cracking index data; • Agency standard for asphalt pavements is percent length, not percent area, which required reprocessing the data for the HPMS; • Additional cost and time requirements; • Challenge in determining the baseline data for percent cracking and faulting due to qualita- tive versus quantitative cracking values and changes in how the faulting values are calculated; Number of responses: 41 0 2 2 4 6 15 24 0 5 10 15 20 25 30 Reporting rutting Reporting IRI Determining PSR Reporting faulting Reporting percent cracking Determining percent cracking No reporting challenges Number of Agencies Figure 33. Agency HPMS reporting challenges.

State of the Practice 53   • Challenge in accurate collecting of IRI data in urban areas due to speed requirements; • Level of detail required by PM2 (0.1 mi) and reporting around bridges; • Delay in collection of pavement data in 2020 due to the COVID-19 pandemic, which added challenges in completing the required NHS and HPMS pavement condition surveys; and • Reporting construction locations. Agency Reporting Agencies provide the results of the pavement condition survey to a variety of agency offices. The majority of responding agencies indicated that pavement condition results are provided to upper management (40 agencies), asset management (39 agencies), districts (38 agencies), pavement design (35 agencies), and maintenance (32 agencies) (Figure 34). To a slightly lesser degree, pavement condition data are also provided to the following offices: transportation plan- ning (27 agencies), materials (25 agencies), budget (19 agencies), and construction (15 agencies). Following are other agency responses regarding the reporting of the results of pavement condition surveys: • “Everyone through ArcGIS Online®, but . . . report our survey results specifically to [just] a few groups.” • “Pavement management.” • “Data are available to the entire agency using tools and a video log.” • “Research efforts related to pavement design and asset management . . . [and] to certain con- struction warranty monitoring efforts.” • “We publish an annual report and post it on our website along with pavement condition maps . . . [that] are available to anyone. . . . [W]e make presentations to upper management, planners, and districts each year.” Number of responses: 44 15 19 25 27 32 35 38 39 40 0 10 20 30 40 50 Construction Budget Materials Transportation Planning Maintenance Pavement Design Districts Asset Management Upper Management Number of Agencies Figure 34. Agency offices to which results of pavement condition survey are reported.

54 Automated Data Collection and Quality Management for Pavement Condition Reporting • “Upon request, to anyone within the agency, universities for research, and private consultants.” • “Design.” • “Division offices.” • “Highway Systems Office.” • “Performance measures.” • “State legislature.” • “Annual pavement condition report available online.” • “Pavement conditions (map and data) are freely available to everyone through a public GIS web application.” To capture examples of agency pavement condition results, the research team asked the agencies whether they produce an annual report and whether the report is readily available. Of the 44 responding agencies, 36 indicated a pavement condition status report is prepared; of these, 11 agencies indicated the status report was unavailable (i.e., internal document only), 4 provided a copy of the status report, and 21 provided a link for accessing the status report (Figure 35). A list of agency links for pavement status reports is provided in Appendix B. Additional Comments and Suggestions Finally, agencies were asked to provide any additional suggestions or comments related to APCS technology, data quality management, or reporting requirements. A summary of agency suggestions and comments is provided in Table 27. Summary of Chapter 3 Over the past 30 to 40 years, SHAs have been transitioning from manual to semiautomated to fully automated surveys. This transition was initiated through collection of transverse and longitudinal profiles for determining IRI, rut depth, and faulting. Automated quantification of surface distress was initiated by capturing video of the pavement surface, which transitioned to digital images, then to 2D technology, and, finally, to 2D and 3D technology. However, a number of challenges in transitioning to an APCS were noted. These included a lack of method Number of responses: 44 Yes (can provide a file), 4 Yes (can provide a link), 21 Yes (unavailable), 11 No, 8 Figure 35. Availability of pavement condition status report.

State of the Practice 55   standardization; lack of information on how agencies have successfully transitioned to an APCS; compatibility issues with historical records; an impact on performance models, cost, and time to conduct the survey; and data quality concerns. The primary benefits of transitioning from manual to an APCS included improved efficiency and safety. Today, the majority of SHAs quan- tify pavement condition through an APCS. SHAs are required to collect, analyze, and submit pavement condition measures to several national programs, including PM2, HPMS, and GASB-34. For PM2 reporting, SHAs are required to submit IRI, rutting, faulting, and percent cracking annually for the Interstate system and every 2 years for the non-Interstate NHS. For low-speed highways (<40 mph), agencies may submit PSR values in lieu of IRI, faulting, cracking, and percent cracking. For HPMS reporting, SHAs are required to submit the same pavement condition measures and meet other reporting require- ments. For GASB-34, SHAs report general infrastructure assets along with depreciation or pres- ervation cost. PM2 requires SHAs to develop, submit, and use a DQMP for the pavement condition survey, regardless of the type of survey method. In general, the DQMP includes criteria for equipment calibration and certification, quality control, and acceptance criteria. Of the 48 SHA DQMPs reviewed, all include criteria for IRI, rut depth, faulting, and percent cracking, as required by PM2. Agencies also include criteria for other distress types (e.g., block cracking, patching, raveling), profile measures (e.g., cross slope, macrotexture), and image (e.g., clarity, perspective). A web-based questionnaire with questions related to pavement condition surveys and agency and national requirements for reporting pavement condition was developed and sent to 52 SHAs. Of the 48 responding agencies, 13 reported using fully automated methods to quantify pavement Table 27. Agency suggestions and comments. Topic Agency Suggestions and Comments Automated pavement condition surveys • There is a lot to know and learn; asking the right questions is a great start (Indiana). • In general, agency has noted increased speed in data collection, improved worker safety, and improved data quality (California). Data analysis • Agency has had issues with determining cracking versus concrete pavement joints. This not only affects the distress index but also the average faulting (Minnesota). • Currently, technological limitations and accurate cracking determination remain major challenges for the agency (California). Data consistency and repeatability • Agency moved to APCS for consistency and repeatability (New Jersey). • Agency found automated data to be more consistent year to year than previous manual surveys (Florida). Lidar • Agency is looking to purchase a vehicle-based lidar system; data storage and analysis are now less of an issue and return on investment looks to be quite high (Ohio). Quality control and acceptance • 100% NHS and HPMS pavement condition surveys are an agency challenge based on the level of manual pavement condition survey and quality control required for acceptable level of accuracy in reporting (Wisconsin). Reporting • Agency has noted measurable improvements with managing and disseminating data due to the adoption of GIS; agency reports are provided on web-based maps and dashboards (Arizona). Standards • Agency suggests promoting the use of a universal data format, minimizing or eliminating the need for a vendor license (Florida). • Agency recommended the need to develop standards to keep pace with current and future APCS technology (South Dakota). • Pavement management community often uses the same words and requirements but implements them in different ways. Agency recommends development of standards of definition (Kansas). • Efforts underway with FHWA TPF-299/399 to standardize 2D/3D pavement image data format will help improve data analysis and quality control and acceptance efforts (California).

56 Automated Data Collection and Quality Management for Pavement Condition Reporting condition, 6 reported using semiautomated methods, 17 reported using a combination of semi- and fully automated methods, 3 reported using a combination of manual and semiautomated methods, and 7 reported using a combination of manual and fully automated methods. The majority of the agencies have been conducting APCS for 10 or more years (24 agencies). The predominant number of agencies use either vendor contracts or agency-purchased equipment and agency staff (14 agencies each) for conducting the APCS. The remaining 20 agencies use a combination of vendor and agency equipment and staff for data collection, analysis, and reporting. Predominant modifications and challenges with APCS implementation included assessing, vali- dating, and verifying data quality (30 agencies); obtaining consistent results (22 agencies); modify- ing index and rating calculations (20 agencies); modifying distress definitions (19 agencies); and finding significant differences between manual and APCS results (19 agencies). Noted disadvantages of APCS implementation included the lack of a direct correlation with historical data (18 agencies), increased cost of data collection (17 agencies), and year-to-year variability of results (14 agencies). Noted advantages of APCS included data compatibility with national reporting requirements (36 agencies), increased rater safety (31 agencies), and the ability to use a single device to collect sensor and distress data (30 agencies). Specific to data collection, agencies reported the correct crack detection (27 agencies), cracking data collection and report- ing (17 agencies), and adjusting algorithms to meet requirements (17 agencies) as the primary challenges with data and image quality control and acceptance. More than half of the responding agencies (28 agencies) conduct pavement condition surveys on 15,000 lane miles or less. While the cost of data collection varies according to the number of lane miles, survey method, vendor versus agency data collection and data analysis, and extent of included distress types, the average weighted cost for a pavement condition survey is approximately $62 per lane mile. Manual surveys of pavement condition resulted in the highest weighted cost of $104 per lane mile, while the combination of semi- and fully automated surveys resulted in the lowest cost at $56 per lane mile. The majority of agencies store the results of the pavement condi- tion survey on data storage drives, requiring approximately 10 terabytes of storage for every 10,000 lane miles collected. The requirements for data quality and acceptance were met by all SHAs, per PM2 requirements. Of the 48 responding agencies, 14 sample less than 5% of the collected lane miles, 12 sample 75% to 100%, and 10 sample 5% to 10%. Similar to agency costs for conducting the pavement condition survey, the time spent for conducting the data quality and acceptance review is depen- dent on the survey method, sample size, and distress types included in the evaluation. Noted data quality issues included correct crack detection (27 agencies), collection and reporting of cracking data (17 agencies), adjusting algorithms to meet requirements (17 agencies), and missing data (15 agencies). As anticipated due to PM2 requirements, the predominant distress types reported included IRI (41 agencies), rutting (35 agencies), faulting (34 agencies), percent cracking (35 agencies), and other agency-specific distress types and indices (36 agencies). The majority of agencies (42 agencies) indicated developing DQMPs (or individual components) to meet the PM2 requirements. Approximately half of the responding agencies (22 agencies) indicated challenges with establish- ing PM2 pavement performance targets; less than one-third of the agencies reported challenges with determining baseline conditions (15 agencies) and adjusting targets (12 agencies). Eleven agencies reported no challenges at all. Since HPMS has been in place for a number of years, the majority of agencies (24 agencies) indicated no challenges in providing the needed information. For internal agency reporting, the majority of agencies provide information on pavement con- dition to upper management (40 agencies), the asset management office (39 agencies), districts (38 agencies), pavement design (35 agencies), and maintenance offices (32 agencies).