Field notes / Emissions evidence / 30 May 2014

Road readings call for an investigation.

The measured road/laboratory gap for Vehicles A and B should not be dismissed as ordinary driving variation. Investigate possible test-dependent emissions controls. The readings do not establish a defeat device, intent, or a legal certification failure.

The dated call

Demand an investigation, not a misconduct finding. Two vehicles had weighted FTP-75 laboratory results below the study's printed Tier 2 Bin 5 benchmark, while their warm/hot-start road-route means were many times that benchmark. Vehicle C provides a third, different pattern, including a rural route elevation. Neither on-road PEMS measurements nor these route/lab comparisons are formal certification-cycle verdicts.

Measurements: VW1 study, pp. iii-iv, 63-65. Rule: VW2 Tier 2 rule, p. 6734. Decision is this analysis, not a statement by either source.

One common comparison basis

Units before multiples

The rule lists Bin 5 NOx at 0.07 g/mi. At 1 mile = 1.609344 km, that is 0.043495983 g/km. The study prints a rounded 0.043 g/km benchmark. Every route multiple below divides its PEMS mean in g/km by 0.043 g/km, not by the unrounded conversion. Sources: VW2 rule, p. 6734; VW1 study, p. 59.

Different test contexts

Road results are PEMS averages over two consecutive runs per route, in warm/hot-start condition. The two laboratory values are weighted FTP-75 chassis-dynamometer results with their own repeat counts. A road mean above the numerical benchmark is an observation warranting scrutiny, not proof of failure on the prescribed certification test. VW1 study, pp. iii, 63-65.

Route comparison

All bars use the same scale: multiple of the study's 0.043 g/km benchmark. Zero is the left baseline; a blank reading is marked missing, never zero. Choose a vehicle to highlight its bars and inspect its five route records. The complete table remains below.

NOx route means as multiples of the 0.043 g/km study benchmarkThree vehicle groups, five routes each. Shared axis from zero to 36 times. A route without a reading is marked missing. Exact measurements and multiples are in the table below. NOx mean / 0.043 g/km (times benchmark)0122436 Vehicle A Route 1: 14.279 timesR1 highway Route 2: 23.000 timesR2 urban LA Route 3: 35.000 timesR3 rural Route 4: 34.419 timesR4 San Diego Route 5: missingR5 San Franciscomissing Vehicle B Route 1: 8.000 timesR1 highway Route 2: 18.814 timesR2 urban LA Route 3: 15.605 timesR3 rural Route 4: 15.698 timesR4 San Diego Route 5: 18.953 timesR5 San Francisco Vehicle C Route 1: 1.116 timesR1 highway Route 2: 1.628 timesR2 urban LA Route 3: 9.512 timesR3 rural Route 4: missingR4 San Diegomissing Route 5: 1.233 timesR5 San Francisco R1 highway; R2 LA; R3 uphill/downhill;R4 San Diego; R5 San Francisco.VW1 route data, p. 65VW2 rule, p. 6734Scale: 0 to 36 times; common benchmark 0.043 g/km.
Figure 1. Route means, not laboratory results. Color identifies vehicle; bar length shows the same study-benchmark multiple for all three. Source links are inside the chart.

Complete road-route record

Mean and sigma in g/km NOx. Sigma is the reported standard deviation over two consecutive runs, not a confidence interval. The multiples are calculated from the mean / 0.043 g/km and rounded to three decimals. Missing means or sigmas are not zeros. Every row: VW1 study, p. 65, Table 4.3.

Five PEMS routes by three anonymized vehicles, including unmeasured combinations
VehicleRouteMean (g/km)Sigma (g/km)Times 0.043 g/kmSource
A1 highway0.6140.28314.279VW1 p. 65
B1 highway0.3440.0968.000VW1 p. 65
C1 highway0.048missing1.116VW1 p. 65
A2 urban (LA)0.9890.11423.000VW1 p. 65
B2 urban (LA)0.8090.07518.814VW1 p. 65
C2 urban (LA)0.0700.0411.628VW1 p. 65
A3 rural-up/downhill1.5050.18135.000VW1 p. 65
B3 rural-up/downhill0.6710.01615.605VW1 p. 65
C3 rural-up/downhill0.4090.0299.512VW1 p. 65
A4 urban (San Diego)1.4800.33534.419VW1 p. 65
B4 urban (San Diego)0.6750.05715.698VW1 p. 65
C4 urban (San Diego)missingmissingmissingVW1 p. 65
A5 urban (San Francisco)missingmissingmissingVW1 p. 65
B5 urban (San Francisco)0.815missing18.953VW1 p. 65
C5 urban (San Francisco)0.0530.0211.233VW1 p. 65

Laboratory FTP-75, separate category

Reported weighted average NOx during chassis-dynamometer testing at CARB El Monte, not road-route observations. Vehicle C has no FTP-75 result in this record. VW1 study, pp. iii, 63-64.

Reported laboratory means and one-sigma repeat variation (g/km)
VehicleMean (g/km)Sigma (g/km)RepeatsSource
A0.0220.0062VW1 pp. 63-64
B0.0160.0023VW1 pp. 63-64

What might account for the gap?

Road conditions and uncertainty

Route 3 is rural uphill/downhill. Higher load and terrain plausibly contribute to its elevations, including Vehicle C's 9.512 times benchmark; this packet has no matched flat-route counterfactual. Vehicle A's Route 1 DPF regeneration run also had evening stop-and-go traffic. The reporters say these factors confound attribution of the 97% increase between that run and the run without regeneration. Run-to-run sigmas, especially A's 0.283 g/km on Route 1 and 0.335 g/km on Route 4, show variation; missing sigmas remain unknown. Instrument uncertainty is not quantified in the supplied excerpts, so no numerical instrument-error allowance can be inferred. VW1 study, pp. 64-65.

Investigation threshold, not mechanism

The reporters observed no fault codes before vehicle selection. That limits a simple malfunction account but does not identify the cause of the gap. Only three anonymized vehicles, each with a different manufacturer or after-treatment technology, were studied. A and B repeatedly have large road-route means while their reported weighted FTP-75 means fall below the printed benchmark; C shows lower means on most measured routes but rises on the rural route. My inference: terrain, regeneration, traffic and measurement/run variability merit targeted checks, yet this pattern is substantial enough to demand investigation of possible test-dependent controls. There is no evidence here proving such controls, intent, model-wide behavior, or failure of a formal certification test. VW1 study, pp. ii-iv, 63-65.

Source ledger

  1. VW1: WVU CAFEE / ICCT study, final report 15 May 2014, ICCT publication 30 May 2014; excerpts pp. ii-iv, 9-10, 59, 63-65 in cafee-study.txt. The source URL contains a later upload-directory name; it is not later evidence.
  2. VW2: EPA Tier 2 rule, Federal Register 10 February 2000, p. 6734, in tier2-rule.txt. Source registry: SOURCES.csv.