Document Type

Technical Report

Publication Date


Organizational Units

College of Natual Science and Mathematics, Chemistry and Biochemistry


The University of Denver remote sensor for on-road motor vehicle carbon monoxide emissions was used for eleven days in the Los Angeles Basin in December, 1989. The remote sensor has been incorporated into the 1990 Clean Air Act Amendments as "on-road emissions testing". The device measures the CO/CO2 ratio for one-half second behind each vehicle, from which the exhaust %CO is calculated. Vehicles were measured in a mix of many driving modes and speeds ranging from deceleration coming up to a red traffic light through idling in heavy congestion up to accelerations and cruises entering a freeway ramp at highway speeds. The results have been validated by both EPA and CARB blind comparisons. The calculated %CO is analogous to that which would have been measured had the vehicle been equipped with a tailpipe probe. The mass emissions in grams CO per gallon of gasoline used can also be derived. Eight of the days monitored normal urban street driving; three monitored freeway ramps. Over 27,000 valid CO emission measurements were made. When the videotapes had been read and returned to California authorities for matching the license plates, the total number of vehicles both measured and matched with the license plate database was over 16,000. Because of the poor contrast of older California license plates and the sun angles, more plates were readable when the front of the vehicles were imaged. With this arrangement a significant number of vehicles without front plates could not be identified. The license plate matched fleet was 0.15 %CO cleaner (3/4 of year on average newer) than the total fleet. This probably arises because older vehicles have older style plates which are both intrinsically harder to read (lower contrast), and often in poorer condition.

Overall for the driving modes and vehicles tested more than fifty percent of the CO was emitted by eleven percent of the vehicles with %CO equal to or greater than five (gross polluters). New vehicles were so clean (gross polluters were less than 1% for the 1989 and 90 model years) that their emissions were almost negligible. The percentage of gross polluters rises from 4% (328 vehicles) of the 83-90 model year vehicles through 17% for the 75-80 model year vehicles to 30% (504 vehicles) of the 1974 and older fleet. If the whole measured fleet could maintain the 1989 and 1990 measured emissions then the total on-road pollution from the 16,000 vehicles measured would decrease more than fivefold. Despite the fact that the new vehicles are on average clean, the dirtiest 20% of the one year old fleet was dirtier than the cleanest 20% of any model years regardless of age. Because old vehicles are not numerous, and most new vehicles are low emitters, most of the carbon monoxide came from emissions of the dirtiest 20% of the vehicles with model years between 1976 and 1988.

An analysis of the data indicates that a conservative upper limit of fifteen percent of the measured CO emissions arises from vehicles in either a cold start or an off-cycle acceleration mode. Forty three percent of the fleet of 77 vehicles measured four or more times were always in the clean (<1 %CO) category. These emit 4% of the total CO from all 77 vehicles. One quarter of the fleet of 77 showed emissions consistently between one and five percent CO. These vehicles emitted 18% of the CO An additional 25% of the fleet were over the five percent CO cut point at least twice. These vehicles emitted 70% of the emissions. Only a small fraction (5 vehicles, 7% of the fleet of 77 vehicles) jumped into the high category only once. The emissions variability observed in this data set is similar to the emissions variability observed when vehicles are repetitively subjected to conventional I/M testing. These results imply that an inspection and maintenance program incorporating remote sensing, which targets gross polluters with multiple violations, has the potential to identify a significant fraction of the CO emissions while inconveniencing only a small fraction of the vehicle owners. Our analysis concludes that on-road remote sensing as a component of an I/M program has the advantages of being representative of the on-road emissions of the vehicle in question, being an emissions test which is almost impossible to circumvent, and incorporates a "fairness factor" such that the more a vehicle is driven, the more frequently it will be tested. When age related factors are eliminated the findings in California are essentially identical to findings from on-road CO studies of large fleets of vehicles in Denver, Chicago and Toronto.

Forty-seven vehicles out of a fleet of 387 vehicles registered as diesels show emissions greater than 2%CO. Of these vehicles, thirty-nine are 1975-84 General Motors vehicles. The vehicles are such high emitters that the only sub-fleet found to be dirtier are 1955-1970 vehicles. Three lines of evidence point to the conclusion that more than half of the vehicles listed in this category are not diesel powered and are incorrectly registered thereby avoiding the California Smog-Check program.

There were differences in average CO emissions between the sites measured, and to a lesser extent between different days at the same sites. To aid in understanding this phenomenon, all remote sensing data available at the University of Denver from a variety of US cities with altitudes lower than 7,000 ft were analyzed in terms of hourly average CO emissions compared to hourly average fleet age. From this analysis a linear model was developed which demonstrated that almost all of the observed differences could be accounted for by differences in average age. This results because of the previously shown influence of the gross polluters which increases with fleet age. Smaller, load induced average emission increases between an uphill but slow cruise-mode freeway off-ramp and a flat but high speed acceleration on-ramp were discernable after the age differences had been eliminated. The linear model predicts average %CO for all fleets measured in the USA to better than 0.5 %CO with a knowledge of only the average fleet age.

The important conclusions are that a few vehicles (gross polluters) emit most of the CO A few vehicles are always measured in the gross polluter category, a few are frequently in that category, and most are never gross polluters. The fraction of gross polluters increases from one in one hundred new vehicles up to one in three old ones. Although new vehicle standards and technology changed from the early seventies to the early eighties, no sharp breaks are observed for the transition model years. The evidence suggests that on-road CO emissions increase linearly with average age of the fleet, and that the linear increase is dominated by the steady increase in the fraction of gross polluters with age. This increase with age appears to be caused in large part by improper (in some cases illegal) maintenance practices.

Publication Statement

This report was originally published as:

Stedman, D.H., Bishop, G.A., Peterson, J.E., & Guenther, P.L. (1991). On-Road Remote Sensing of CO Emissions in the Los Angeles Basin. Final report to the California Air Resources Board under Contract No. A932-189.

This document, commissioned by the California state government, is public domain.