Road condition survey quality control: the visual and automated routes

Road condition survey quality control is the set of checks that show a visual rating or an instrument reading can be trusted before it enters a road authority's register. In TMH (Technical Methods for Highways), the South African road manual series, TMH 9 (Committee Draft Final, May 2016) proves the people who rate pavements by eye. TMH 13 (Committee Draft Final, May 2016), for automated road condition assessment, proves the devices that measure. The two quality routes share no metric. (COTO, 2016b) (COTO, 2016d)

Key facts

• The South African road manual TMH 9 (Committee Draft Final, May 2016), Part A section A.5.4, splits a quality management plan for visual assessment into quality control and quality acceptance.
• TMH 9 Part A, section A.5.3, allows only assessors with a valid accreditation for the pavement type to carry out visual assessments.
• TMH 13 Part A (Committee Draft Final, May 2016), section A.4.4, says validation sections are selected and profiled before the survey, ideally by the network agency or another contractor.
• COTO (Committee of Transport Officials) Standard Specifications Chapter 20 (Draft Standard, October 2020), clause A20.1.5.5, sets a maximum validation age for texture, roughness and rutting unless the contract says otherwise.
• TMH 9 compares ratings between people and TMH 13 compares measured values with a reference device, so neither South African road manual gives a shared quality score.
• TMH 13 Part G (Committee Draft Final, May 2016) applies TMH 9 assessor and quality-plan requirements to people rating images and asks for TMH 9 ratings of automatically detectable distresses.

The concept map

The connections in text

1. Assessor accreditation (TMH 9 Part A) → Quality management plan (TMH 9 Part A): accredited assessors only
2. Quality management plan (TMH 9 Part A) → TMH 9 reference ratings (TMH 9 Part A): sample re-assessed
3. Quality management plan (TMH 9 Part A) → Rating verification routine (TMH 22 App. J-4): ratings then verified
4. TMH 9 reference ratings (TMH 9 Part A) → Data acceptance (TMH 22): agreement test
5. Rating verification routine (TMH 22 App. J-4) → Data acceptance (TMH 22): flags conflicting ratings
6. Device class specified (TMH 13 Part A) → Contract quality plan (TMH 13 Part A): plan with tender
7. Device class specified (TMH 13 Part A) → Validation section test (TMH 13 Part C): validated before survey
8. Device class specified (TMH 13 Part A) → Imaging validation (TMH 13 Part G): class sets criteria
9. Contract quality plan (TMH 13 Part A) → Control testing (TMH 13 Part C): covers control checks
10. Validation section test (TMH 13 Part C) → Control testing (TMH 13 Part C): re-measured in survey
11. Validation section test (TMH 13 Part C) → Validation age limit (COTO Specs): dated certificate
12. Validation age limit (COTO Specs) → Data acceptance (TMH 22): age within limit
13. Control testing (TMH 13 Part C) → Data acceptance (TMH 22): control records
14. Imaging validation (TMH 13 Part G) → TMH 9 reference ratings (TMH 9 Part A): images checked against ratings
15. Data acceptance (TMH 22) → Road asset register (TMH 22): accepted data

What does quality control mean for road condition data?

Quality management of road condition data has two parts: quality control, the checks made while data are collected, and quality acceptance, the checks made before data are accepted for use. TMH 9 Part A, section A.5.4, uses both terms for visual assessment, and Draft TMH 22 (March 2013), the road asset management manual, defines them for all road asset data in sections D.5.7 and D.5.8. D.5.8 adds that acceptance is often called quality assurance and declares the data acceptable, or not, against the data collection appointment's requirements. (COTO, 2016b) (COTO, 2013)

These manuals are issued under COTO, the Committee of Transport Officials. A road condition survey records pavement state in one of two ways: an accredited assessor rates each pre-defined road segment (the length of road that receives one rating) by eye, or an instrument measures a quantity such as the international roughness index (IRI), in m/km.

How is a visual rating proved trustworthy?

A visual rating is proved trustworthy by proving the person who rates and then sampling the work, in four checks (three in TMH 9, one in Draft TMH 22). A rating records each distress by degree (how severe it is) and extent (how much of the segment it affects), judged by eye, so no physical reference exists to calibrate the assessor. The rating procedure is in how to run a TMH 9 visual assessment. (COTO, 2016b)

• Accreditation. Section A.5.3 allows only assessors with a valid accreditation to carry out assessments. It covers one pavement type, lasts a fixed period and stays valid only while the assessor keeps assessing regularly; section A.5.2 describes the training.
• Quality management plan. Section A.5.4 splits the plan into quality control, an internal responsibility of the authority carrying out the assessment, and quality acceptance. Control starts with a calibration session on network-specific issues and a follow-up check of the first assessments.
• Agreement test. For quality acceptance, experienced accredited assessors independently re-assess a representative sample and compare their results with the originals. Annexure A is one worked example, from the Western Cape Government: it converts each degree and extent pair to a deduct value and then a rank value, and applies a two-tailed paired t-test, which compares two sets of ratings made on the same segments, to each distress type.
• Verification routine. Draft TMH 22, section D.5.11 and Appendix J-4, lists pairs of ratings that conflict on one segment. Draft TRH 22, Pavement Management Systems (1994; TRH is Technical Recommendations for Highways), printed the same routine as Technical Addendum A.2, whose techniques A.1 says were not yet accepted practice; J-4 carries it forward without citing TRH 22. (COTO, 2013) (CSRA, 1994)

Narrated film · 1:31

Two routes to trusted condition data

A two-branch map: the visual route and the instrument route run in parallel towards data acceptance and the road asset register, joined by one bridge at TMH 13 Part G.

Read the film transcript

How is a road condition rating, or an instrument reading, proved trustworthy? TMH 9 proves the people who rate pavements by eye, and TMH 13 proves the devices that measure. The two quality routes share no metric, so a delivery needs two kinds of evidence. On the visual route, only accredited assessors rate, under a quality management plan, and an independent sample is re-assessed and compared. On the instrument route, the system is validated on reference sections, retested during the survey, and held to a maximum validation age. The routes join in TMH 13 Part G, on imaging, which ties image validation to TMH 9 reference ratings. Both routes end at data acceptance, which feeds the road asset register. One route proves the people, the other the devices, and imaging connects them. In a synthetic delivery, fifty-eight of sixty profiler records arrive, one segment conflicts, and the validation is fifty-two days old. None of these checks shows that the ratings or roughness values are right. A validated device is not validated data, and agreement between assessors shows consistency, not accuracy. The manuals guide the visual acceptability limits, validation criteria and validation age, but the road authority sets them. Keep each check's evidence in one pack, because neither TMH 18 version has a field for assessor, device or validation date. Next, TMH 18 sets out the file formats, and TMH 22 takes accepted records into the register.

How is an instrument reading proved trustworthy?

An instrument reading is a physical quantity, so the instrument route proves the device against a reference, in six checks. TMH 13 Part A (sections A.4.3 and A.4.4) and Part C, Roughness, set out the pieces; how to check a roughness survey on a validation section walks through the survey check. (COTO, 2016d) (COTO, 2016a)

• Device class. Part C defines classes of roughness device (Table C.1), and Draft TMH 22, section D.4.3, ties the permitted class to the road class. (COTO, 2016a) (COTO, 2013)
• Contract quality plan. Section A.4.3.6 says the contractor should ideally submit, with the tender, a plan covering equipment, staff training, scheduled equipment checks and positioning checks. (COTO, 2016d)
• Validation. Component calibration deals with parts such as lasers. Validation checks the whole system by measuring sections that a reference device has already profiled (section A.4.3.3). The sections are chosen and profiled before the survey, ideally by the network agency or a contractor other than the one surveying (section A.4.4). (COTO, 2016d)
• Operator and protocol. Part C says a profiler is validated for a specific operator and measurement protocol, never adjusted to match a benchmark, and repaired by its manufacturer if it is not valid (section C.4). (COTO, 2016a)
• Control testing. During the survey, validation sections are measured again. For a profiler, a failed control test means data collected since the last successful one are discarded and measured again (sections A.4.3.4 and C.4.4). (COTO, 2016d) (COTO, 2016a)
• Validation age. Part A says the specification should state how long a validation stays valid (section A.4.3.3). COTO Standard Specifications for Road and Bridge Works for South African Road Authorities, Chapter 20: Quality Assurance (Draft Standard, October 2020), clause A20.1.5.5, sets a maximum age for texture, roughness and rutting validations unless the contract says otherwise. (COTO, 2016d) (COTO, 2020a)

Chapter 20 lists TMH 9 among its published pavement test methods (clause A20.1.4.11) but gives only TMH 13 a clause of its own. (COTO, 2020a)

Why do visual and instrument quality checks share no measure, and where do they meet?

Visual and instrument quality checks share no common measure because they answer different questions. The visual route compares ratings between people: do two accredited assessors, rating the same segments, differ by more than chance allows? The instrument route compares a measured value with a reference device: are the bias (average error) and precision (scatter of repeats) of a value such as IRI acceptable? Neither TMH 9 nor TMH 13 converts one result into the other, so a delivery needs two kinds of evidence. (COTO, 2016b) (COTO, 2016a)

The routes meet in TMH 13 Part G, Imaging (Committee Draft Final, May 2016). Where people rate pavement images (a Class 2 system), section G.3.3.3 says TMH 9's procedure, assessor training and quality-plan requirements apply, and allows reference ratings to be a consensus estimate. Where software detects cracks (Class 1), validation is split into detection, checked against field measurements, and analysis; TMH 9 degree and extent ratings are reported for the distresses the software can detect, and manual review records the rest (sections G.3.3.3 and G.3.4.3, Tables G.6 and G.7). (COTO, 2016h)

TMH 13 Part C, Appendix C-3 (Situation C-3.4), suggests a visual assessment to explain a roughness survey that departs from the last one: a plausibility check, not a shared metric. (COTO, 2016a)

Which quality check applies to which road condition record?

Each of ten records has a check and a clause; rows one to four are visual, five to nine instrument, and ten both.

Record	Check	Clause	Evidence file
Assessor	Valid accreditation for the pavement type	TMH 9 A.5.3	Accreditation register with dates
Rating forms	Calibration session and follow-up check	TMH 9 A.5.4	Session notes, follow-up record
Sampled segments	Independent re-assessment compared	TMH 9 A.5.4, Annexure A	Control sheets, test workings
Each segment	Conflicting ratings flagged	TMH 22 D.5.11, J-4	Verification output, outcomes
Contractor's plan	Quality plan submitted with the tender	TMH 13 A.4.3.6	Contract quality plan
Device and operator	Class permitted; components calibrated; tested on reference sections	TMH 22 D.4.3; TMH 13 A.4.3.3, A.4.4, C.4	Pre-validation and validation reports with date
Survey period	Validation within the clause or contract age	COTO Ch. 20 A20.1.5.5	Certificate and survey dates
Control tests	Sections re-measured during the survey	TMH 13 A.4.3.4, C.4.4	Control test records
Image-derived distress	Reference ratings and manual review	TMH 13 G.3.3.3, G.3.4.3	Imaging validation, review log
Whole delivery	Validity, accuracy, integrity; missing segments	TMH 22 D.5.8 to D.5.12	Acceptance record

Last reviewed on 21 October 2026 against the editions cited in this guide.

What does a delivery check of a road condition survey look like?

Synthetic example. Every value is invented. A contractor delivers data for synthetic Road S2 from km 10.0 to km 16.0 (6,000 m): three visual segments of 2,000 m each, and profiler records at 100 m intervals along one lane.

1. Count the instrument records. 6,000 m divided by 100 m gives 60 records. The file holds 58, so two are missing. TMH 22 sections D.5.8 and D.5.12 list checks for missing segments and record counts. (COTO, 2013)
2. Run the verification routine. Segment S2-02 is rated for severe bleeding and very good skid resistance, the conflicting pair D.5.11 uses as its example, so the form returns to the assessor. (COTO, 2013)
3. Date the validation. The certificate is dated 9 March and the last survey day is 30 April, so the validation was 52 days old on that day. Compare 52 days with the contract and clause A20.1.5.5. (COTO, 2020a)

Check value: 58 of 60 records, one flagged segment and a 52-day-old validation. None of these checks shows that the ratings or the IRI are right; that needs the sample comparison and the validation and control records.

What do TMH 9, TMH 13 and the COTO Standard Specifications leave to the road authority?

The manuals guide two decisions without fixing them; the road authority sets both in the survey specification and the contract.

• Visual acceptability limits. TMH 9 Part A describes the sample and the comparison; Annexure A shows one authority's limits as an example. (COTO, 2016b)
• Validation criteria and age. TMH 13 Part A says both are to be specified; Part C offers guideline criteria (Table C.9) the network manager may relax, and Chapter 20 a default age the contract can change. (COTO, 2016d) (COTO, 2016a) (COTO, 2020a)

Which misreadings of road condition quality checks should you avoid?

• A validated device means validated data. Validation shows a device, operator and protocol were valid on a date. Control tests and the age limit check that it stays true during the survey. (COTO, 2016a)
• Agreement between assessors means accuracy. Agreement shows consistency: TMH 9 checks the sample "for accuracy" against another assessor's rating, and Part G says reference ratings may rest on consensus. The insight on repeatability and uncertainty explains the difference. (COTO, 2016h)
• Image-derived distress is a TMH 9 rating. Part G asks for validation, TMH 9 ratings for detectable distresses and manual review for the rest. (COTO, 2016h)
• The data file carries its own evidence. In the road-data formats of TMH 18 (Draft Version 4, October 2014; Committee Draft 5.0, February 2016), neither the visual summary file nor the combined instrument data file has a field for assessor, device, class or validation date. (COTO, 2014) (COTO, 2016i)

What evidence should you keep after a road condition survey?

Keep the evidence file listed in the table for each check in one pack, indexed by road and survey date, with the acceptance record and the manual editions used. The insight on survey acceptance evidence shows how a reviewer reads such a pack.

Frequently asked

What quality control does TMH 9 set out for a visual assessment?

TMH 9 (Committee Draft Final, May 2016, also written TMH9), the South African manual for visual assessment of road pavements, describes in Part A, section A.5.4, a quality management plan with two parts. Quality control is an internal responsibility of the authority carrying out the assessment, starting with a calibration session and a follow-up check of early assessments. Quality acceptance is independent re-assessment of a representative sample by experienced accredited assessors, compared with the original ratings.

How do you check a consultant's road condition data before payment?

Check a consultant's road condition survey data by asking for two kinds of evidence. For visual ratings, ask for the assessors' accreditation, the quality management plan and the independent sample comparison. For instrument data, ask for the dated validation report, the control test records and the device class. Then run the data quality checks in Draft TMH 22 (March 2013), the road asset management manual, sections D.5.8 to D.5.12; the appointing authority decides whether the data are accepted.

How long does a TMH 13 validation stay valid?

Not indefinitely. The South African road manual TMH 13 Part A (Committee Draft Final, May 2016) says the specification should state how long a validation, the test of a whole measuring system on sections a reference device has already profiled, stays valid. COTO (Committee of Transport Officials) Standard Specifications Chapter 20 (Draft Standard, October 2020), clause A20.1.5.5, sets a maximum age for texture, roughness and rutting validations unless the contract says otherwise.

Can automated crack detection replace a TMH 9 visual rating?

Not by itself. In Part G, Imaging, of the South African road manual TMH 13 (Committee Draft Final, May 2016), crack-detection software is validated against field measurements, TMH 9 degree and extent ratings are reported for the distresses the software can detect, and manual review records the rest. Where people rate the images, TMH 9 assessor training and quality-plan requirements apply.

Do visual and automated surveys share one quality measure?

No. In South African road condition surveys, visual quality checks compare ratings between people, using degree and extent, while instrument quality checks compare measured values, such as the international roughness index (IRI) in m/km, with a reference device on a validation section. Neither TMH 9 nor TMH 13 converts one result into the other, so a delivery needs two sets of evidence; TMH 13 Part G is where the routes connect.

Does a TMH 18 road-data file record the assessor, device or validation date?

No. In TMH 18, the South African manual for road asset data exchange formats, neither Draft Version 4 (October 2014) nor Committee Draft 5.0 (February 2016) gives the visual summary file or the combined instrument data file a field for assessor, device, class or validation date. Quality evidence therefore travels separately, in a pack indexed by road and survey date.

Where this connects

TMH 18 sets out the road-data file formats in which accepted records travel to the authority, and Draft TMH 22, Parts D and E, takes accepted data into the register and its condition indices. (COTO, 2014) (COTO, 2016i) (COTO, 2013)

References

1. COTO (2016b). TMH9: Manual for Visual Assessment of Road Pavements. Part A: General. Committee Draft Final, May 2016. Original source.Section A.2: degree and extent
2. COTO (2016a). TMH13: Automated Road Condition Assessments. Part C: Roughness. Committee Draft Final, May 2016. Original source.Sections C.2–C.5; Table C.5 is for calibration-section selection
3. COTO (2016d). TMH 13: Automated Road Condition Assessments. Part A: General. Committee Draft Final, May 2016. Original source.Sections, tables and clauses are cited inline in each guide
4. COTO (2016h). TMH 13: Automated Road Condition Assessments. Part G: Imaging. Committee Draft Final, May 2016. Original source.Sections, tables and clauses are cited inline in each guide
5. COTO (2013). Draft TMH 22: Road Asset Management Manual. Draft, March 2013. Original source.Sections, tables and clauses are cited inline in each guide
6. CSRA (1994). Draft TRH22: Pavement Management Systems. Pretoria: Department of Transport, 1994. Original source.Technical Addendum A.3, equations A.3.1–A.3.3 and Table A.3.1, printed pp. 125–126; Section 6.2.2, printed pp. 55–56, for alternative-weight normalisation
7. COTO (2020a). Standard Specifications for Road and Bridge Works for South African Road Authorities. Draft Standard, October 2020. Original source.Sections, tables and clauses are cited inline in each guide
8. COTO (2014). TMH 18: Road Asset Data Electronic Exchange Formats. Draft Version 4, October 2014. Original source.Sections, tables and clauses are cited inline in each guide
9. COTO (2016i). TMH 18: Road Asset Data Electronic Exchange Formats. Committee Draft 5.0, February 2016. Original source.Sections, tables and clauses are cited inline in each guide

Use the manual edition and project requirements that apply to your work. Worked examples use synthetic values and are not site-specific design advice.