transport · family: the missing yardstick
counts of unknown accuracy
Highway decisions rest on traffic counts whose accuracy is rarely known, checked by methods that vary agency to agency, and may not hold after installation
Problem statement
Traffic volume counts from automated counters are a key input to the management systems highway agencies run — HPMS, pavement and bridge management and maintenance, congestion and safety management — and to road and bridge design and transportation planning. Yet, as the problem statement puts it, "All highway agencies use traffic count data, but it is rarely known how accurate said data are." Agencies evaluate accuracy by methods that differ from agency to agency, and data quality changes over time as equipment ages and as the environment affects how sensors perform, so "the ground truth about traffic count accuracy obtained immediately after equipment installation may not remain representative over time." The unsolved problem is a practical, statistically sound, uniformly applied way to establish ground truth for in-service counters — how long to count, how many lanes and vehicles, what accuracy criteria — that agencies will actually run, so that count accuracy is a known quantity rather than an assumption.
Why this matters
The lack of uniformity in the methods agencies use to evaluate count accuracy "leads to challenges with analyses of national traffic data sets, comparison of traffic trends between different agencies, or even using traffic data within a given agency." The statement says traffic count programs are being asked to evaluate maturing detection technologies and third-party traffic data sources, that the need is even more relevant as new technologies such as AI-based video recognition are deployed, and that common standards and procedures are necessary to quantify the accuracy of traffic counts. The TRB Highway Traffic Monitoring Committee (ACP70) first identified this need in a 2016 survey of professionals; it was submitted as a research need for the FY 2023 NCHRP program, in a compendium dated February 2022.
What’s been tried and why it hasn’t worked
FHWA's Traffic Monitoring Guide recommends calibrating every permanent and portable counter annually, and a family of ASTM standards nominally covers evaluation — E2300-09 (device specification), E2532-09 (test methods for evaluating device performance), E2759-10 (truth-in-data practice), and E177-20 (precision and bias terminology). But the standards have "limited implementation" and agencies have built their own procedures as well. The problem statement says the reasons need to be explored and documented, and it names several issues with E2532, the most critical being its acceptance clause (Sections 7.2.9 and 7.3.7), which declares a device nonfunctional or inaccurate if its calculated difference exceeds the specified tolerance "for all values of the data item measured." The authors argue that this wording makes no provision for random error via a confidence interval, and that including one would allow precision and bias to be determined, contrary to what Sections 7.2.10 and 7.3.8 now state. The statement also says the standard omits alternative ground-truth methods, and it asks the research to look at methods for evaluating both the counter technology itself and its implementation at a particular site, considering both accuracy and reliability (consistent accuracy over time). NCHRP has since funded Project 03-145 ($600,000, TTI, 2024–2027) to set evaluation criteria and test methods for traffic sensors that could seed a national evaluation program. Its project page frames the need as independent evaluation of sensor systems and lists a "Ground truth baseline" among its example criteria; it does not mention the D-21 statement, and it does not say whether the work will cover ground truth for counters already in service.
What would unlock progress
Progress needs a ground-truth protocol that is cheap enough to repeat: a defined sample size and duration derived from the statistics of counting error, a reference method (the submitter's response in the compendium names permanently installed continuous counters, video, and manual counts as references for the companion statement D-20) that itself has known accuracy, and a software tool that turns a short reference count into accuracy, precision, and bias estimates with confidence intervals for a given site and technology. An adjacent field to borrow from — this write-up's own suggestion, not something the sources say — is measurement-system analysis in manufacturing quality (gauge R&R and ISO 5725 precision/bias studies), which deals with separating bias from noise through repeated measurement; the highway-monitoring field has the standards vocabulary (E177) but, on the statement's account, no widely implemented protocol.
Entry points for student teams
A team could build an open-source video reference-count tool with a human-in-the-loop verification interface and quantify its own error, then use it to run side-by-side ground-truth tests on 2–3 counter technologies at a partner agency's permanent site, deriving the count duration and vehicle sample needed to bound accuracy at a stated confidence for high-volume (>2,000 vph/lane) and low-volume (500–5,000 AADT) roads. A statistics team could draft a revised E2532 acceptance clause with confidence intervals and simulate its behavior against realistic error distributions. A data team could analyze a state's historical validation records to estimate how fast accuracy drifts by sensor type and environment. Relevant skills: statistics, traffic engineering, computer vision, and standards writing.
Genes
Structural cousins: stuck the same way, in other fields
Sources
"Practical Ground Truth Method and Tools for Evaluating Accuracy, Precision, and Bias of Traffic Volume Counting Equipment," NCHRP FY2023 Problem Statement 2023-D-21 (Olga Selezneva, ARA; Lawrence A. Klein; Steven Jessberger, FHWA; submitted by Kent L. Taylor, NCDOT), with NCHRP evaluation, in *NCHRP FY 2023 Program: Compendium of Problem Statements* (TRB, February 2022), and accessed 2026-08-17; NCHRP Project 03-145 "National Traffic Sensor System Evaluation Program" (TTI, in progress 2024–2027), accessed 2026-08-17 go to source 1 ↗ go to source 2 ↗ go to source 3 ↗
Working notes
Notes made while writing and checking this brief, shown as written. They use working shorthand.
The problem statement was authored by traffic-monitoring specialists including an FHWA staff member and submitted by NCDOT; the NCHRP evaluator recommended funding at $500,000/36 months (confirmed in the compendium PDF; the FHWA reviewer suggested combining it with D-20) but no NCHRP project number for it was found on 2026-08-17. All quoted statement language ("rarely known how accurate", "may not remain representative over time", the E2532 7.2.9/7.3.7 clause, ACP70 2016 survey) was confirmed against the D-21 text at verification. NCHRP 03-145 was verified on its TRB project page as related-but-distinct (product evaluation, not in-service ground truth). `failure:ignored-context` (deployment/operational sub-pattern): existing standards and practice assume accuracy established at type approval and installation, ignoring aging and environmental drift; `constraint:data` because the binding gap is a missing reference measurement, not sensor technology. `failure:not-attempted` was rejected because standards and agency procedures constitute serious prior attempts. Related collection briefs: none on traffic monitoring; nearest neighbors are the data-quality briefs in `domain:digital`. Verified at intake 2026-08-17: gate (net) + adversarial source check + contested-tag second coding.
Source type: Self-articulated (traffic-data practitioners articulating a validation gap in their own measurement systems)
Reconciliation 2026-10-01: Fact repair after reading the whole write-up against the three sources it names, all fetched directly on 2026-10-01 and read as text. Read in full: the D-21 problem statement (https://apps.trb.org/nchrpballoting/BallotingDocs/D-21prob.htm) and the NCHRP 03-145 project page (https://apps.trb.org/cmsfeed/TRBNetProjectDisplay.asp?ProjectID=5307). Read in part: the FY 2023 compendium (https://onlinepubs.trb.org/Onlinepubs/nchrp/docs/FY2023NCHRPProblemStatementCompendium.pdf; the full text was searched, and the cover, the D-21 pages with their reviews and submitter response, and the D-20 submitter response were read). (1) Both titles changed. The long title said accuracy is "Checked Once at Installation" and "Then Assumed Forever," and the display title was "Counts Nobody Re-Checks." The statement does not say that. It says it is rarely known how accurate the data are, that the ground truth obtained right after installation may not remain representative over time, and that FHWA's guide recommends calibrating every counter each year. The long title was "Every Highway Decision Rests on Traffic Counts Whose Accuracy Is Checked Once at Installation, by Methods That Vary Agency to Agency, and Then Assumed Forever" and is now "Highway Decisions Rest on Traffic Counts Whose Accuracy Is Rarely Known, Checked by Methods That Vary Agency to Agency, and May Not Hold After Installation." The display title is now "Counts of Unknown Accuracy." (2) The opening listed inductive loops, piezo strips and radar, and called counts the denominator of crash rates and project prioritization. The statement names no counter technologies except video recognition, and does not mention crash rates or project prioritization; the sentence now lists the uses the statement gives (HPMS, pavement and bridge management and maintenance, congestion and safety management, road and bridge design, planning). (3) "Accuracy is typically established once, when a device is installed and type-approved" and drift from "weather, pavement condition, and congestion" are not in the statement, which says quality changes with equipment aging and environmental effects on sensor performance; the sentence now says that. "What tolerance and confidence" now reads "what accuracy criteria," the statement's words. (4) That counts "cannot be compared or pooled with known error," "probe-based" estimates, and comparison "against their own equipment" are not in the statement; the two sentences now follow its wording (programs are asked to evaluate maturing technologies and third-party data sources; the need is even more relevant with AI-based video recognition; common standards and procedures are necessary). (5) "Six years later it was still being resubmitted" is not in the sources: nothing says the statement was submitted before, and the compendium is dated February 2022. The sentence now gives the program year and the compendium date. (6) The text said the statement "pinpoints why" the standards are little used. The statement says the reasons need to be explored and documented, and separately lists issues with ASTM E 2532; the text now keeps those apart. (7) The text said the clause "precludes determining precision and bias (contrary to what Sections 7.2.10 and 7.3.8 state)." The statement says the reverse about those sections: including a confidence interval would allow precision and bias to be determined, contrary to what the two sections now state. Corrected. (8) That E2532 "addresses type approval of a device, not the accuracy of a particular installation as it ages" is not in the statement and was removed; the statement's own request, that the research look at methods for evaluating both the technology and its implementation at a particular site, and at both accuracy and reliability, replaces it. (9) The judgment that Project 03-145 is "distinct from" this problem is not on the project page, which lists a ground truth baseline among its example criteria and does not mention D-21; the sentence now reports what the page says. The line in the paragraph above that calls 03-145 related-but-distinct is the intake reader's judgment, not the page's. (10) "Most likely reviewed video" as the reference method is not in the sources; the text now gives the three references the submitter's response in the compendium names (there, as the references used in the companion statement D-20). The comparison with gauge R&R and ISO 5725 is not in any source and is now labelled as this write-up's own suggestion. (11) "An FHWA data lead" now reads "an FHWA staff member"; the statement gives no job title. Confirmed in the sources and left as written: the Source line (title, three authors, submitter, compendium title and date, project number, agency and dates); the three quotations from the statement and the phrase "limited implementation," each found word for word; the annual calibration recommendation; the four ASTM designations and what each covers; Sections 7.2.9 and 7.3.7; the 2016 ACP70 survey (the compendium gives the committee's name as Highway Traffic Monitoring in a neighbouring statement); $600,000, Texas A&M Transportation Institute, September 2024 to March 2027 for Project 03-145; the reviewer's $500,000 and 36 months and the FHWA suggestion to combine with D-20; and, in the entry points, the 2–3 technologies side by side, over 2,000 vehicles per hour per lane, and 500 to 5,000 AADT. Could not be checked: whether D-21 was ever funded under a project number (the sources do not say), and the text of the ASTM standards themselves, which the write-up describes only through the statement. A second reader fetched the same three sources independently and went over these repairs on the same day; all eleven were kept, and two wordings were brought closer to the sources: the opening sentence now says "HPMS" rather than "federal HPMS reporting" (the statement says only "HPMS"), and the Project 03-145 sentence now says "sensor systems" rather than "sensor products" (the project page's term). Genome tags untouched.