Excess Pollution from Vehicles—A Review and Outlook on Emission Controls, Testing, Malfunctions, Tampering, and Cheating

Abstract

Although the transition to electric vehicles (EVs) is well underway and expected to continue in global car markets, most vehicles on the world’s roads will be powered by internal combustion engine vehicles (ICEVs) and fossil fuels for the foreseeable future, possibly well past 2050. Thus, good environmental performance and effective emission control of ICE vehicles will continue to be of paramount importance if the world is to achieve the stated air and climate pollution reduction goals. In this study, we review 228 publications and identify four main issues confronting these objectives: (1) cheating by vehicle manufacturers, (2) tampering by vehicle owners, (3) malfunctioning emission control systems, and (4) inadequate in-service emission programs. With progressively more stringent vehicle emission and fuel quality standards being implemented in all major markets, engine designs and emission control systems have become increasingly complex and sophisticated, creating opportunities for cheating and tampering. This is not a new phenomenon, with the first cases reported in the 1970s and continuing to happen today. Cheating appears not to be restricted to specific manufacturers or vehicle types. Suspicious real-world emissions behavior suggests that the use of defeat devices may be widespread. Defeat devices are primarily a concern with diesel vehicles, where emission control deactivation in real-world driving can lower manufacturing costs, improve fuel economy, reduce engine noise, improve vehicle performance, and extend refill intervals for diesel exhaust fluid, if present. Despite the financial penalties, undesired global attention, damage to brand reputation, a temporary drop in sales and stock value, and forced recalls, cheating may continue. Private vehicle owners resort to tampering to (1) improve performance and fuel efficiency; (2) avoid operating costs, including repairs; (3) increase the resale value of the vehicle (i.e., odometer tampering); or (4) simply to rebel against established norms. Tampering and cheating in the commercial freight sector also mean undercutting law-abiding operators, gaining unfair economic advantage, and posing excess harm to the environment and public health. At the individual vehicle level, the impacts of cheating, tampering, or malfunctioning emission control systems can be substantial. The removal or deactivation of emission control systems increases emissions—for instance, typically 70% (NO_x and EGR), a factor of 3 or more (NO_x and SCR), and a factor of 25–100 (PM and DPF). Our analysis shows significant uncertainty and (geographic) variability regarding the occurrence of cheating and tampering by vehicle owners. The available evidence suggests that fleet-wide impacts of cheating and tampering on emissions are undeniable, substantial, and cannot be ignored. The presence of a relatively small fraction of high-emitters, due to either cheating, tampering, or malfunctioning, causes excess pollution that must be tackled by environmental authorities around the world, in particular in emerging economies, where millions of used ICE vehicles from the US and EU end up. Modernized in-service emission programs designed to efficiently identify and fix large faults are needed to ensure that the benefits of modern vehicle technologies are not lost. Effective programs should address malfunctions, engine problems, incorrect repairs, a lack of servicing and maintenance, poorly retrofitted fuel and emission control systems, the use of improper or low-quality fuels and tampering. Periodic Test and Repair (PTR) is a common in-service program. We estimate that PTR generally reduces emissions by 11% (8–14%), 11% (7–15%), and 4% (−1–10%) for carbon monoxide (CO), hydrocarbons (HC), and oxides of nitrogen (NO_x), respectively. This is based on the grand mean effect and the associated 95% confidence interval. PTR effectiveness could be significantly higher, but we find that it critically depends on various design factors, including (1) comprehensive fleet coverage, (2) a suitable test procedure, (3) compliance and enforcement, (4) proper technician training, (5) quality control and quality assurance, (6) periodic program evaluation, and (7) minimization of waivers and exemptions. Now that both particulate matter (PM, i.e., DPF) and NO_x (i.e., SCR) emission controls are common in all modern new diesel vehicles, and commonly the focus of cheating and tampering, robust measurement approaches for assessing in-use emissions performance are urgently needed to modernize PTR programs. To increase (cost) effectiveness, a modern approach could include screening methods, such as remote sensing and plume chasing. We conclude this study with recommendations and suggestions for future improvements and research, listing a range of potential solutions for the issues identified in new and in-service vehicles.

1. Introduction

The subject of pollution emissions from ICE vehicles has faded into the background in policy discussions in Europe and the US, given the attention to hydrogen and battery electric vehicles (EVs). And while vehicle electrification is well underway, it is a slow process that can be hampered by specific barriers to EV adoption. Thus, a large number of ICE vehicles will remain in use for the foreseeable future. This suggests that attention to exhaust gas emissions performance is still necessary, especially when considering that, in practice, these can be much higher than intended. Governments are the appropriate actors to enforce robust emission standards and to monitor vehicle performance and compliance. Vehicle emissions regulations are not infallible policy interventions, and vehicle manufacturers have found loopholes in the requirements. For their part, regulators are often playing catch-up and, in many instances, applying compliance and enforcement practices that are deficient. To help, this paper provides insights into the nature of the problem and suggestions for improvement.

Electrification of the road transport sector (including multi-mode electric mobility) is the most appropriate and cost-effective approach to reduce greenhouse gas (GHG) emissions, improve energy efficiency and air quality, and promote energy security. However, there are significant barriers to EV adoption. These include (i) higher upfront costs of EVs compared to conventional ICE vehicles; (ii) low consumer preference and resistance to change, (iii) the lack of access to EVs in all markets; (iv) lack of ubiquitous charging and fueling infrastructure to win over range-anxious drivers; and (v) a lack of strong emission regulations, subsidies, and scrappage programs in most jurisdictions. The relatively long useful life of current ICE vehicles, typically 15 to 20 years for a first-time buyer, will lead to a slow natural fleet turnover. The rate of vehicle fleet electrification is country-specific, and like other innovations when first introduced, EVs have had to rely heavily on early adopters and public subsidies.

It is therefore not surprising that predictions of EV penetration vary widely. Studies report expected EV shares in new vehicle sales in 2030 from less than 10% to 100% (e.g., [1,2,3]). This suggests that the internal combustion engine will dominate the transport sector in many markets, particularly in emerging economies, in 2050 and beyond, making good performance and effective emission control of paramount importance. Robust policies that prevent deterioration and discourage cheating and tampering through in-service (real-world) emission testing and maintenance and repair programs will need to be present in any jurisdiction that is interested in abating pollution from the ICE vehicle sector.

The differences in emission performance of an ICE vehicle in the real world versus official regulatory tests are well established. Conventional vehicle emission control systems tend to be substantially less effective and more emissive in real-world driving, and this seems to be particularly evident in diesel vehicle performance. The reasons for this difference include emission system aging and deterioration, the limitations of dynamometer tests, the use of Auxiliary Emission Control Devices (AECDs) for engine protection, and poor maintenance. However, it can also be factors such as tampering by vehicle owners or the presence of defeat devices used by vehicle manufacturers.

Intentionally disabling a vehicle’s emissions control system during real-world operation is illegal because it can result in a large increase in emissions well above the applicable emission limits. For instance, exceedances of NO_x limits by about 5 to 40 times have been reported (e.g., [4,5,6,7,8,9,10,11,12,13]). In the case of the notorious Volkswagen (VW) dieselgate scandal, in some instances, California authorities observed and documented excess NO_x emissions 80 times the allowable limits [14].

Addressing intentional cheating, tampering, and malfunctions is an important and necessary action for the effective implementation of vehicle emission standards because these behaviors can undermine the expected benefits of such policies. Here, we provide a broad outlook by (i) offering a comprehensive review of the published literature, (ii) take historical stock of the issues, (iii) articulate the current state of play in the world of vehicle emission control (or lack thereof), and (iv) formulate questions for potential future research.

The authors conducted an in-depth literature search using university library catelogues and Google in combination with relevant keywords and search strings, i.e., cheating, tampering, emission(s), PTI, I&M, vehicle inspection, and OBD. Moreover, several reports that have not been widely published but have been collected by the authors over time were included in the review process. A total of 228 publications were reviewed in this study.

2. Vehicle Emission Controls

Engine designs and emission control systems in ICE vehicles are complex, computer-controlled systems generally optimized to balance performance, efficiency, and pollution abatement. Post-combustion emission controls in modern vehicles today are akin to chemical reactors in the tailpipe. Engine management and on-board diagnostics play a key role. When these systems malfunction due to natural deterioration or intentional tampering, it can have a large negative effect on performance and emissions. For instance, removing emission controls from a diesel pick-up truck reportedly increased NO_x and PM_2.5 emissions by as much as 300 times [15].

There are, however, important differences in these issues between compression ignition (CI), spark-ignition (SI), and gasoline direct injection (GDI) engines. Generally, the SI engine dominates the light-duty vehicle (LDV) sector in countries like the US and Australia, while CI technology is the engine of choice in heavy-duty vehicle (HDV) and equipment applications. The exception is Europe, where the diesel passenger car has been popular. For environmental performance, these engines should all be equipped with modern and proven post-combustion emission controls such as oxidation/reduction catalysts and particle filters.

2.1. Spark-Ignition Engines and Vehicles

Emission control of CO, HC, and NO_x for SI engine vehicles using petrol, CNG, or LPG is based on precise air–fuel ratio control and catalytic after-treatment using a three-way oxidation/reduction catalyst (TWC). Particle emissions from SI engines are generally assumed to be relatively low in indirect injection engines (IDIs). Therefore, early regulations essentially ignored them. This is no longer the case, especially now that direct injection (DI) SI engines have a dominant market presence. In addition, particle emissions from CNG and LPG DI engines are now more readily recognized as a concern similar to diesel particles. For this reason, emission regulations in the American, European, and the Japanese–Korean markets include particulate matter limits equally for all engine types.

Initially, extensive field testing in Europe in the 1990s showed that the long-term stability of a number of emission-relevant components for ignition, mixture, and lambda ratio control proved to be poorer than anticipated [16]. A damaged catalytic converter or a malfunctioning oxygen sensor increased HC and CO emissions by a factor of 20, often without seriously affecting driveability [17]. Since then, SI engine emission controls have matured and combine both emission reduction efficiency and durability [18,19]. These improvements have resulted in modest incremental costs for automakers, which are typically passed onto consumers.

Experience with SI engines in the last three decades shows that sophisticated control systems have led to a major reduction in emissions of air pollutants. Previous on-road measurement studies (e.g., remote sensing) have confirmed that real-world emission control for SI engines is generally effective and robust [20,21]. This technology is also less affected by cheating or tampering, as will be discussed.

Nevertheless, SI technology can be prone to high emissions levels during transient operation (i.e., the effect of time lag in closed-loop control) and during extremely high- or low-load conditions (e.g., over-fueling at full throttle, poor combustion, and low exhaust gas temperature at idle). Petrol vehicles can also be vulnerable to aging. Typically, the durability and failure of sensitive components such as sensors and actuators are the main issue that requires regular maintenance and repair [18]. Recent research of 50 high-mileage petrol vehicles reports a doubling of average NO_x emissions due to aging [22,23]. A small portion (e.g., 5%) of on-road petrol vehicles were found to have malfunctioning emission control systems, which increased NO_x emissions by a factor of 10 [24]. Unfortunately, with the current European On-board diagnostics (EOBD) systems and periodic technical inspection (PTI) test methods, these high-emission vehicles are difficult or even impossible to detect. Furthermore, there are no regular programs in Europe that map the emission behavior of older vehicles. Given the ever-increasing average age of these vehicles, such test programs are sorely lacking.

2.2. Compression-Ignition Engines and Vehicles

Diesel exhaust emissions contribute to ground-level ozone pollution, and prolonged exposure to diesel particles can lead to premature death because diesel particles are carcinogenic according to the International Agency for Research on Cancer (IARC) and do not have a safe exposure threshold. IARC based its action on evidence of the causality of lung cancer from exposure to diesel exhaust [25]. Similarly, in 1998, the state of California identified diesel exhaust as a toxic air contaminant due to its carcinogenicity in humans [26]. Hundreds of compounds have been identified in diesel exhaust, and at least 40 substances are considered toxic [27]. Diesel exhaust particles (DEPs), the majority of which are ultrafine particles (UFP, with an aerodynamic diameter < 0.1 mm), are composed of solid aggregates of an elemental carbon and/or a metal core and a large surface area for the adsorption of various (toxic) chemicals, including polycyclic aromatic hydrocarbons (PAHs) and transition metals [28]. Gas-to-particle formation in diesel exhaust from heterogeneous and homogenous nucleation is also a source of UFPs [29]. Combustion-generated nanoparticles (<0.05 mm) were the subject of active research worldwide, starting in the mid-1990s, when modern diesel engines were discovered to have lower PM emissions but an enormous increase in the number of UFP emissions. Diesel particle filtration (i.e., the diesel particle filter, DPF) gained popularity for addressing concerns regarding occupational exposure for workers involved in tunnelling operations in Switzerland [30,31]. Due to those early applications, the DPF is now the ubiquitous emission control technology in all modern diesel CI engines, beating the most stringent PM emission standards.

Emission control for CI engines in diesel fuel vehicles has also become increasingly sophisticated and requires the use of relatively new technologies. Technology has shifted from open-loop, mechanically controlled engines with no aftertreatment to complex controls utilizing high-pressure injection, variable geometry turbochargers, exhaust gas recirculation, and a variety of sophisticated exhaust gas aftertreatment devices, such as DPFs and selective catalytic reduction (SCR) (e.g., [32]). A modern light-duty diesel vehicle (car, SUV, and LCV) typically is equipped with a diesel oxidation catalyst (DOC) for CO and HC emission control, a DPF for PM and particle number (PN) emission control, and a combination of NO_x reduction technology such as (cooled) exhaust gas recirculation (EGR), SCR, and/or lean-NO_x trap (LNT) [33].

Given the stricter USA emission standards for NO_x, the modern diesel LDVs found on American roads are typically fitted with a combined system of EGR, SCR, and LNT controls; in Europe, these may only be equipped with SCR or LNT controls [34,35]. For trucks, the Euro IV standard required the use of SCR and/or EGR, possibly in combination with a DPF, whereas the Euro V standard forced the introduction of combined EGR/DOC, SCR/EGR, or DPF/EGR technology. The integration of SCR, EGR, and DPFs is necessary to comply with the Euro VI limits and beyond (e.g., [36,37]).

Previous studies suggest that real-world emission control for diesel CI engines is less effective, less robust, and more susceptible to cheating or tampering when compared with SI engines [20,21]. Moreover, there are higher costs associated with diesel controls compared to petrol and LPG engines. Whereas the estimated additional cost to a manufacturer to produce a Euro 6 LDV is about USD 400 for a petrol engine, the cost is three to five times higher for a diesel engine [34]. Cost and durability aspects are relevant in relation to cheating or tampering, as will be discussed later. The review suggests that defeat devices are primarily a concern with diesel vehicles (both light and heavy), where emission control deactivation in real-world driving can lower manufacturing costs (i.e., the manufacturer can install inferior and cheaper systems and engine calibrations), improve fuel economy, reduce engine noise, improve vehicle performance, and extend refill intervals for diesel exhaust fluid (DEF such as AdBlue), if present [38].

2.3. GDI Engines and Vehicles

Gasoline direct injection (GDI) technology is a recent entrant into the LDV market, penetrating the on-road petrol vehicle fleet as emphasis shifted in the EU, US, and other markets toward improved fuel efficiency and lower carbon emissions. The technology has rapidly gained popularity [39]. For instance, for the model year 2022’s new vehicle stock, over 70% of all car manufacturers relied on GDI technology in new vehicles sold in the US, and for several, the penetration of GDI technology exceeded 90% [40]. PN emissions from GDI, most of which are UFPs, are a concern because, in some circumstances, they can be much higher than other conventional alternatives.

Gasoline particulate filters (GPFs) have been applied to GDI engines in Europe since 2019 (Euro 6d-temp) to control for PM and PN emissions. In contrast, in the US, where the most recent vehicle emission standards continue to focus solely on PM rather than on PM and PN, we may see this technology applied to a lesser extent [41]. Insufficient and inconclusive health evidence linking morbidity and mortality outcomes to chronic and acute exposures to UFP emissions is the reason given by US federal regulators for excluding PN emissions from the regulations [42]. However, interestingly, the same regulators have promulgated a PN emission standard for civil aviation aircraft engines to align the US with international standards. In Europe and the many jurisdictions around the world where the Euro standards have been adopted, the use and proper functioning of GPFs is therefore of increasing interest. Emission testing suggests that GPFs have similar filtration behavior as DPFs. However, the PN emissions of petrol engines can be up to 50 times lower than diesel engines (e.g., [43]).

2.4. Engine Management System

Vehicle manufacturers conduct extensive research and development programs in search of the optimization of driveability, performance, fuel consumption, and emission control. They rely on a variety of engine parameters (e.g., engine speed, coolant temperature, throttle position, vehicle speed, etc.) and complex software in the engine management system. A modern car can contain more than 70 computers (electronic control units, ECUs) and 100 million lines of code [13,44]. In this trend, we see nearly all aspects of vehicle and engine operation being controlled by the ECU, creating an embedded closed-loop control system for engine sensors and actuators (e.g., fuel injection and air valves, including the EGR valve). This allows manufacturers to precisely control all aspects of engine operation and thus drive significant improvements in performance, reliability, and fuel economy. But, at the same time, this complexity and sophistication creates ample room for cheating and tampering.

The ECU is also responsible for ensuring that the engine and emission control systems operate as designed to comply with emissions standards. While systems like the DOC or the DPF can be passive, others require active control by the ECU, sometimes after sacrificing performance or emissions. These trade-offs are particularly challenging for diesel engines, which in their native form are noisier, run on lower-quality fuel, operate at lower exhaust temperatures, and notoriously generate more PM and NO_x emissions than their petrol and gaseous fuel counterparts [13].

When SCR is used, diesel engines can be calibrated to produce higher engine-out NO_x emissions and lower fuel consumption, resulting in better fuel efficiency and lower engine-out PM (soot) emissions [45]. NO_x is then reduced downstream by the SCR. Similarly, the Continuously Regenerating Trap (CRT) uses a DOC installed upstream of the DPF to oxidize approximately 25% to 55% of the NO_x to NO₂. Elevated NO₂ concentrations promote filter regeneration (i.e., DPF self-cleaning) at lower exhaust temperatures through continuous combustion of the collected soot [46].

However, these optimization strategies also mean that when the SCR, DOC, or DPF systems are not working properly, deactivated, or removed altogether, artificially elevated excess engine-out NO_x/NO₂ or uncontrolled PM emissions are released in the exhaust and straight into the environment. Regulation targeting some of these emission excursions can help. For instance, in the mid-2000s, as CRT technology was gaining popularity, California imposed explicit limits on the amount of NO₂ created for DPF regeneration to avoid excess emissions and potential near-source receptor exposures.

2.5. On-Board Diagnostics (OBD)

Real-time OBD (US), a California policy innovation, and EOBD (European OBD) requirements were first introduced in the 1980s and 1990s, respectively, to guarantee primarily the integrity of the engine exhaust emission control system and, in Europe, vehicle safety [47,48,49]. For HDVs, OBD was first introduced in 2005 (Euro IV) in the EU and phased in between 2005 and 2010 in the US [50]. OBD is a complex and expansive network of sensors installed to continuously monitor, record, and store parameters indicating the state of performance of various emission-related components. As the name implies, OBD was intended to be only a diagnostic tool to identify for the repair technician components that require maintenance to prevent excess emissions. It was not a direct measurement of emissions. However, as sensor technology evolves toward emission measurement, the potential of OBD has expanded.

The system runs diagnostics software in the background and can detect abnormal and abrupt changes in performance, notify the driver (via a Malfunction Indicator Lamp or MIL light on the dashboard), and indicate the cause of the fault (fault codes, diagnostic trouble code, or DTC). OBD uses preset threshold limits (OTL) for activating the MIL light for selected pollutants. In Europe, these are PM and NO_x. In the US, these also include NMHC; CO; and aspects of fueling, such as fuel leaks and purging issues.

The threshold limits are defined as multipliers (1.5 to 5.0 times) or additional offsets to the relevant vehicle emission standard [48,50,51]. OBD provides important feedback to vehicle drivers and technicians about maintenance status and potentially urgent repairs [50]. Experience in the US suggests that most drivers (about 80%) will seek diagnosis and repair when the MIL is on [48]. The remaining 20% of drivers who ignore the warning are addressed through required routine inspections and maintenance (I/M). Like any other system, OBD can deteriorate due to normal aging or factors such as the use of poor-quality fuel [52]. Today, most routine I/M checks in the US are conducted using the OBD system rather than a traditional “sniff” test of the tailpipe. In Europe, both OBD tests and PTI emissions tests are required, and member states have a certain freedom of choice. In other countries, like Australia, no formal PTR program exists.

3. Cheating and Tampering

Cheating (by manufacturers) and tampering (by owners) refer to intentional and unlawful modifications of a vehicle’s emission control system resulting in environmentally harmful effects from sub-optimal performance. Tampering is simpler and can be performed casually by a vehicle owner or operator. It usually involves “re-flashing” the ECU with unauthorized software (often procured online); removal of vehicle hardware like the TWC, DPF, or other emission control device; or both. In contrast, we refer to cheating as a more complex endeavor, usually involving corporate malfeasance on the part of the original equipment manufacturer (OEM) because the modifications are sophisticated and completed deep in the vehicle’s software and hardware inner workings. Cheating involves a defeat device.

A defeat device is any physical (hardware) or digital (software) modification, a calibration change, or any other modification or a combination of these to a vehicle for a specified reason that reduces the effectiveness of the original emission control system. The defeat device is also an intervention that is not openly and truthfully disclosed by the vehicle OEM to (and allowed by for instance, emergency vehicles [53]) the relevant approval authorities. Thus, the device intentionally “defeats” the emissions control system, typically during real-world driving conditions. In contrast, when these modifications are transparently declared during the emission certification or homologation process between the OEM and regulators, the modifications are legally allowed as AECDs or Auxiliary Emission Strategies (AES), as they are known in the US and the EU, respectively.

Most commonly, AECDs are temporary changes to the operating characteristics of an engine under specific conditions, such as cold start operation. The changes are for safety and engine protection, including the emission control aftertreatment system (e.g., extended operation at high engine load). The use of inferior technology for emission control or otherwise cannot be used as a justification for an AECD.

In this paper, we will distinguish between the impacts of legally approved AECDs granted to an OEM, an illegal defeat device employed by a vehicle manufacturer, and tampering by vehicle owners.

3.1. Legally Allowed Adjustments for Engine Performance While Still Meeting All Applicable Emission Standards

The differences between real-world vehicle emissions and those measured during compliance tests are well established. After all, laboratory tests are only a snapshot of nominal performance. Testing requirements are spelled out in detail in the regulation to ensure a common baseline and repeatability. Automakers can employ complex engine calibrations to optimize emission performance during these tests. In-use emissions are allowed to deviate to a degree from those observed in the laboratory.

In Europe, the type approval (TA) process is conducted by recognized TA authorities or certifying testing bodies and currently includes (1) laboratory engine and chassis dynamometer tests and on-road Real Driving Emission (RDE) tests of a representative pre-model or prototype vehicle; (2) durability tests of emission control devices, prior to type-approval; (3) conformity of production (CoP) tests of new vehicles, sampling the factory output; and (4) In-Service Conformity (ISC) dynamometer and RDE tests by the manufacturer (and from 2020 also independently verified by the TA authority) of a limited number of in-use vehicles with a maximum mileage of 100,000 km [54].

In the US, the equivalent conformity process is commonly referred to as vehicle emission certification because the primary concern is engine operation and exhaust pollution emission control. Safety and consumer protections are ensured under the purview of other authorities. Although the approaches in Europe and the US share many similarities, there are some important differences. In the US, only two government authorities can issue emission certifications to a car or engine maker: the state of California through its California Air Resources Board (CARB) or the US federal government through action by the US Environmental Protection Agency (US EPA). The rest of the states in the US can choose to adopt and accept either certification or both.

Similar to Europe, vehicle emission certification in the US rests largely on an analysis of (1) compliance with OBD system operational requirements and (2) with all applicable regulatory emission limits. But the actual certification process is mostly self-administered. Each car maker conducts its own emission tests of a representative test vehicle for a specified vehicle category and submits those results to the relevant authorities. Car makers are compelled by law to submit only accurate and truthful information. This information is subject to audit, and authorities can conduct at will any compliance testing to verify emission test results or assessment of OBD system performance. During the certification process, automakers and regulators are in constant dialogue. Once authorities are satisfied that a vehicle or an engine type meets all requirements, then an emissions compliance certificate is issued, and the car maker is then free to legally sell the certified vehicle model anywhere in the US.

The technical need to allow for deviations in emission performance in the real world under certain operating conditions (i.e., commanded high engine load, which may not be fully captured during dynamometer tests) is fully acknowledged in the design of the emission standards and the required tests to show compliance. This is the reason AECDs exist. They are an important flexibility afforded to car manufacturers. When properly disclosed to authorities during the certification process, car manufacturers are free to use them in complete adherence with the law. And although AECDs may temporarily reduce the effectiveness of the emission control system, the excess emissions are fully accounted for in the air quality management plans underpinning the standard-setting process [55]. As discussed in the next section, Volkswagen failed to disclose its use of AECDs and submitted false information to the government authorities, which is an unlawful act [55]. Similarly, in Europe, vehicle operation can also legally include several AESs [56]. An AES becomes active and replaces the Base Emission Strategy (BES) temporarily for a specific purpose in response to a change of conditions, such as ambient temperature [57].

The US EPA has provided guidance (e.g., [58,59]) regarding the shared responsibility of both manufacturers and the agency for the submission of information, evaluation of requirements, and associated timelines [60]. Similarly, the European Commission has released two guidance documents (C/2017-352 for pre-RDE and 2023/C-68-01 for post-RDE Euro 5 and 6 vehicles) that are not legally binding but clarify good practices, test protocols, and associated testing thresholds for the detection of defeat devices for different categories of testing. Moreover, the new policy establishes the expected use of best available control technologies and specifies that wear and malfunctioning cannot be used as justification for the use of a defeat device. It outlines extended documentation requirements that need to be provided by vehicle manufacturers to the EU TA authorities for examination. The guidance also relevantly points out that all types of emissions should be considered in defeat device detection (e.g., evaporative emissions). A key point is that approved legal deviations should be limited in general, narrowly applied to specific operating conditions, and should not lead to extended periods of high emissions in real-world driving conditions [38].

3.2. Cheating by Vehicle Manufacturers

As vehicle manufacturers attempt to comply with increasingly more stringent emissions standards [61], emission control technologies can become more complex and expensive. These emission controls add costs to vehicle production, and their use generally reduces overall fuel efficiency, hence increasing the emissions of another heavily regulated pollutant, CO₂. Technological limitations can also put compliance with emission regulations in conflict with other important engine design parameters, such as performance, durability, and efficiency. This creates an incentive for vehicle manufacturers to evade the regulatory requirements by exploiting the general peculiarities of official tests [13]. Then, there is also the human tendency to bend or circumvent the rules. Corporations, large and small, may not be immune to temptation.

3.2.1. History

Cheating by vehicle manufacturers is neither a new phenomenon, nor is it restricted to specific vehicle types. The practice and extent of cheating vary between makes and models, and vehicle manufacturers can choose to implement more or less robust and fail-safe engine and emission control systems [62,63,64].

The notorious Volkswagen “dieselgate” scandal (including its subsidiaries Audi and Porsche) in 2015 received worldwide attention and raised awareness about cheating [10]. The estimated number of affected VW vehicles worldwide was 11 million [12], and 40% of the VW passenger cars sold in the EU were fraudulent cars [62]. Volkswagen had to pay more than USD 25 billion in fines and remedies (e.g., [38]). Reuters reported VW stating that the scandal had cost it nearly USD 33 billion worldwide in fines and settlements [65]. The heaviest fines on the company were levied by California and the US federal government. Ironically, these were the two jurisdictions with the fewest number of fraudulent vehicles, only approximately 85,000 and 450,000, respectively. The rest of the world was left to contend with the excess pollution coming from the VW vehicles. One of the remedies imposed on VW was handing out “goodwill packages” to affected customers, including gift cards, roadside assistance extensions, and other benefits in an attempt to repair relationships with customers [12]. US regulators responsible for the settlement agreements were keenly interested in consumer protection and pushed the company to offer real options to the affected consumers: buybacks or repairs. In some instances, VW owners chose to be left alone, but in all cases, repairing the environmental harm from every single fraudulent vehicle on US roads was the sole responsibility of VW. It is noted that Volkswagen had already reached settlements regarding an illegal defeat device with the US EPA back in the 1970s (temperature sensors to disable EGR) and later on in 2005 regarding a defective oxygen sensor [53,66,67].

Cheating scandals are not limited to Volkswagen. Since 2015, many other car and engine makers have been caught. Several vehicle manufacturers seem to have been deploying increasingly sophisticated strategies to disable or reduce the effectiveness of emission controls on the road. In fact, almost as soon as vehicle emission compliance testing began, a wide range of vehicle manufacturers were reportedly caught cheating. These include General Motors, Ford, Opel, Honda, Hyundai, Jeep, Kia, Mercedes, Fiat, Renault, BMW, Suzuki, Mitsubishi, Audi, Porsche, RAM, and Volkswagen [9,13,58,60,67,68,69,70,71,72,73,74,75].

Chrysler (now Stelantis), for instance, was accused in 1974 by the US EPA of using three small valves on their 1973 models that would briefly disable the emission control system during specific conditions [69,76]. In 1995, General Motors agreed (without admitting wrongdoing) to pay more than ten million dollars in fines and recall affected vehicles after being accused of circumventing pollution controls on 470,000 Cadillac luxury sedans by using air-conditioning status for official test detection [13,67,68,69]. In 1998, settlements were reached with Honda and Ford in the US for charges that both companies illegally sold vehicles and engines equipped with defeat devices that prevented emission-control systems from working properly [58].

In the same year, the largest civil penalty ever at the time for a violation of environmental laws (USD 83.4 million) was handed down to seven manufacturers of heavy-duty diesel engines (HDDE) [69]. The companies sold 1.3 million HDDEs in the US containing defeat devices (software code, like VW), which allowed an engine to pass the regulatory emission certification test but then turn off emission controls at cruise conditions to optimize fuel economy by sacrificing lower NO_x emissions [77]. As a result, these engines emitted up to three times the allowable level for NO_x. Importantly, in the settlement, engine manufacturers were also forced to “pull ahead” and introduce cleaner engines (i.e., with lower PM and NO_X emissions) three years earlier than required. In total, the consent decree reached more than USD 1 billion in penalties and remedies [55,78,79]. Lastly, it also included the first-ever agreement for in-use emission compliance in the Not-to-Exceed (NTE) limits applicable and fully enforceable within agreed-upon measurement allowances developed by US regulators and industry. The importance of the NTE regulation cannot be overstated. It led to, among other advances, the rapid development of portable emission measurement systems (PEMS) that are now commonly used and that could meet regulatory-level test criteria when used on the road during real-world operation. NTE would be the prelude to Europe’s new RDE requirements for light-duty vehicles, which started in 2007, and the new in-use compliance checks in the US in the post-VW dieselgate era. Specifically, California’s enhanced testing for certification and in-use compliance was described formally a week after the VW violations were issued in September 2015 [80]. It took advantage of the advances in PEMS technology and in-use measurement approaches.

While there have been previous defeat device cases, the 2015 Volkswagen scandal is unique. No previous case involved shutting off emission controls permanently while driving on the road, and no other manufacturer reportedly lied to regulatory agencies and tried to cover up their defeat device for almost two years [63]. The case also led to the largest environmental penalties in the history of the automotive industry and, importantly, criminal convictions that resulted in jail time for some company executives. California regulators gave VW only one option to remain in the automotive business—switch its focus completely to zero-emission electric vehicles. VW agreed, and this is how the dieselgate scandal became a catalyst for electric vehicles in the US and around the world [55]. Electrify America was created in the US by VW to execute its newly found “wisdom”. The scandal has been a case study of corporate governance malfeasance [44,71,80,81,82,83].

Recent cheating in the EU undertaken by vehicle manufacturers could relate to artificially inflating CO₂ emissions during legislative tests in anticipation of new EU legislation that will require reductions of fleet average emissions by 15% in 2025 and 30% in 2030 based on 2021 levels [84]. The stringency of the reductions may have effectively been eroded by artificially inflating the test results over the World-harmonized Light-duty Test Procedure (WLTP) and increasing the 2021 baseline. Over-declaration of WLTP CO₂ emissions by manufacturers was recently found to be, on average, 5% in 2020 [85]. Consequently, EU emission regulations will be based on measured WLTP emissions to check compliance with the 2025 and 2030 targets, rather than manufacturer-declared values [85].

Despite the undesired global attention automakers receive when caught cheating, the associated damage to brand reputation, the temporary drop in vehicle sales and stock value [44], and the costs of fines and forced recalls, cheating may continue. The profit margin proposition for automakers may be too powerful for some, if not many. The evidence lies in the actions of major auto companies who have been caught taking socially irresponsible action in the recent past (i.e., action that has led to adverse environmental impacts but also compromised other aspects, such as safety) and seem to have weathered the storm [12]. One factor that seems to encourage cheating is consumers who are willing to buy and pay a premium for allegedly sustainable vehicles [71].

Corporate culture and governance appear to be important factors in relation to the occurrence of misconduct. For instance, it has been stated that VW’s misconduct was rooted in the company’s unique autocratic corporate culture, senior management with variable salaries, its ambitious diesel car targets for the US market, and the time and budget constraints imposed on employees to reach those targets [44,71,81,82,83].

The world has likely not witnessed the last cheating case. The latest cheating scandal reportedly involves RAM trucks in the US with more than USD 2 billion in fines to recall and fix almost one million MY 2019–2023 pick-up trucks, which were fitted with emission-cheating devices [75].

3.2.2. Enforcement

There is a long history of enforcement action against cheating in the US, but, importantly, not in other parts of the world. Enforcement action in the US is supported by specific US EPA regulations and guidance (e.g., [58,59]), which provides clarity and reduces the possibility of manufacturers exploiting loopholes [56]. It can be argued that regulatory enforcement in the US is world-leading. Clearly, US authorities are not afraid to penalize manufacturers that are caught cheating. Penalties have included fines, consumer compensation, imposition of corporate “community service” (e.g., buy-back of older cars, buying electric school buses, and rebuilding older engines), and recalls [63,68]. In the US, California leads the way. When dealing with the VW cheating, it demonstrated a robust regulatory program and an approach based on three core principles: (i) a commitment to underpin policy decisions with scientific evidence generated from extensive vehicle emissions research and development activities; (ii) a comprehensive analysis of all test evidence, including OBD system performance; and (iii) a vigorous compliance and enforcement program based on clear expectations that holds all manufacturers, large and small, accountable. This approach appears to have worked well, since the state was the first to detect the emission anomalies of the VW diesel cars. Then, the state moved to resolve the violations in a way that fully repaired the damage done to the environment and consumers [55].

Similar emissions malfeasance has been observed in other parts of the world, but without the associated clarity and enforcement action seen in the US [38]. For instance, high NO_x emission behavior has been observed but not addressed for cars manufactured in Australia, which employed lean-burn fuel injection strategies during freeway driving, likely to improve fuel economy [86]. Specific manufacturers like BMW appear to have properly functioning emission control systems in the US [64], but not in Europe [60].

In Europe, member states have conducted testing and found high in-use emissions, but regulators there have struggled to prove and act against automakers apparently utilizing defeat devices [38,56]. It is worth noting that California regulators first learned of potential excess in-use emissions from diesel cars from European researchers. A specific factor in the EU is that car manufacturers can choose from 28 type approval authorities and could select the one that appears to employ the least robust homologation approach [56]. Another EU requirement for software or hardware fixes of defeat devices prevents effects on fuel consumption or CO₂ emissions, thus constraining the software corrections [63]. In a letter to a Committee of Inquiry of the European Parliament, California regulators discussed the strengths and weaknesses of the Californian and European approaches to emissions testing for certification and enforcement [87]. These differences between the EU and US approaches may be disappearing, as recent EC guidance documents (C/2017-352 for pre-RDE and 2023/C-68-01) and rulings by the Court of Justice of the European Union (e.g., C-693/18) now provide a strict and narrow definition of what is considered an allowable use of an AES (e.g., only sudden, immediate, and exceptional damage to the engine impacting safety).

3.2.3. Inferior Technology

The design of emission control technology—or rather, intentional sub-optimal design to save costs and maximize profit—has been blamed for poor durability and performance. For instance, NYSOAG [88] reported that under normal soot loading conditions, DPFs used by Porsche, VW, and Audi could only last for 50,000 miles before needing replacement, which is significantly shorter than the 120,000-mile warranty VW was required to offer, compelling it to reduce wear on the DPF. Smoke issues with new diesel SUVs have also been raised in the Australian media [89,90,91]. Studies using remote sensing devices (RSD) suggest that a few percent of these new Australian light-duty vehicles have soot levels, possibly indicating that DPFs may be of inferior quality [21]. DPFs can also suffer from undetected manufacturing faults, e.g., lacking a good seal between the can and substrate, which causes leakage and reduces the expected level of emission control [92].

In vehicles that use SCR for NO_x control, Porsche, VW, and Audi appeared to have installed defeat devices to compensate for DEF tanks of insufficient size. The size of the DEF tank in some of the VW diesel cars implicated in the dieselgate scandal in California reportedly became one of the tell-tale signs of the cheating. Analysis of ECU software suggested that defeat devices were created by Bosch, an ECU manufacturer, and enabled by VW and Fiat for their respective vehicles [13]. Eventually, Bosch was formally accused by California and US authorities and entered into a settlement agreement for its role in the VW violations [93].

Importantly, inferior non-durable technology cannot be used in the US by vehicle manufacturers to argue for the need of an AECD [59]. Until recently, this was not the case in Europe. There, manufacturers used technological arguments for the need for reduced emission control under real-world conditions. These arguments were accepted by the authorities, but the problem may have been related to poorly calibrated or inferior emission control systems [38]. Recent rulings in the EU now align requirements with the US. The use of a defeat device because of inferior technology or for engine protection is not allowed [94].

3.2.4. Finding Defeat Devices

Suspicious real-world emissions behavior suggests that the use of defeat devices by vehicle manufacturers may currently be widespread [94,95]. Given the cost and benefit to the manufacturer and the difficulty of detection, it may not be surprising that corporate misconduct and cheating appears to continue to this day. For instance, recent on-road remote sensing analysis reports that simple hardware “fixes” and software “upgrades” in the UK resulted in a 30–40% reduction in on-road NO_x emissions for “dieselgate”-affected engines [96].

Knowledge of the precise conditions of the regulatory emissions certification test makes it possible for manufacturers to intentionally alter their vehicles during the test, a practice called “cycle beating” in reference to the drive cycles used on the dynamometer for the test [13]. Critics of this approach say demonstration of emission compliance today is akin to a teacher sharing with students the questions on a final exam prior to the test. There is some truth to that. However, authorities are compelled by law to specify in detail the conditions for a compliance test. Car maker representatives subject to the test are also free to witness while the test is conducted by the authorities. In practice, there is extensive interaction between regulators and car makers on all technological, scientific, policy, and legal aspects of the certification process. They rely on and depend on each other. Hence, integrity and honesty are of paramount importance in the ethical compass guiding the execution of duty for both parties.

Typically, a defeat device detects if a vehicle is running on a dynamometer for an emissions test, and if so, it then switches to a special low-emission operational mode. In the real world, engine performance, fuel economy, and drivability are then optimized by the defeat device at the expense of excess exhaust emissions. Examples of potential defeat devices include temperature sensors, engine hood sensors (i.e., the emissions test is conducted with an open hood), timers (i.e., the emissions test is of a specified duration), integrators (i.e., the relationship of distance versus time, which is fixed for an emissions test), and sensors that can detect the operation of the power steering pump or air conditioning (i.e., these are not used during the emissions test).

Defeat devices are generally easy to design and difficult to detect [38]. They are typically embedded in proprietary computer code that is difficult to access and interpret by a third party. New methods are required to find them. Software verification techniques are an example where code is used to scan for intentional attempts to alter a system’s behavior under specific test conditions [13,74]. This type of analysis is quite powerful and can provide strong evidence if a defeat device is present, but it is also time-consuming and requires unique expertise. Extensive knowledge and understanding of the OBD system are also important. In the VW dieselgate scandal, inquiries into the OBD performance by California regulators led to the earliest indications of cheating during compliance testing. Since September 2020, the EU type approval framework grants type-approval authorities and technical services access to vehicle software, which enables the possible detection of defeat devices [38].

Another new approach to detection is to subject the vehicle to an enhanced test program by authorities or third-party experts that includes repeated testing using both a laboratory chassis dynamometer and on-road tests with a PEMS [97,98]. This way, irregularities in emissions can be detected and independently quantified. Further analysis of emissions data can pinpoint suspicious behavior and identify possible triggers for a defeat device. The impact depends on a range of factors, but it is clear that a defeat device typically results in a large and substantial step-change increase in emissions [38]. For instance, real-world NO_x emissions were observed to be 5 to 35 times higher than the US emission limits [5]. In the VW dieselgate, the highest NO_x emissions observed from fraudulent vehicles were 80 times the applicable limits [14]. Test conditions with known elevated emissions, such as cold starts or DPF regeneration, are excluded from this additional analysis. The new RDE approach in Europe and the new in-use enhanced testing in the US [80] are examples of these new approaches for emission compliance into the future. However, it is noted that despite the RDE legislation, emissions still can sometimes be high in practice [99].

A clearly established and agreed upon set of (international) criteria are not yet available (e.g., under continuous review in the EU) and a possible test program for defeat device detection will require determination of the appropriate categories of test data for analysis (e.g., driving conditions) and the emission thresholds for each pollutant (e.g., [38,94,100]). The assessment of emissions strategies and the search for defeat devices remains a complex, resource-intensive, and time-consuming endeavor [57]. But any future regulatory approach must allow for random vehicle tests that reflect real-world conditions.

3.2.5. Cheating and “The Gap”

It has been well-established that the gap in emissions between real-world fuel consumption and the regulatory test has been growing over the last two decades or so [101,102]. This “gap” has been most pronounced in the EU and was particularly large when EU legislation was based on the low-load and artificial New European Driving Cycle (NEDC) test cycle, where studies report differences of up to 60–100% for CO₂ emissions for individual vehicles [103,104] and significantly larger differences for other air pollutants. On average, the gap between fuel consumption and GHG emissions increased from about 10% in 2000 to over 40% in 2015 [105,106].

Vehicle manufacturers increasingly exploited the legally allowed tolerances and flexibilities during chassis dynamometer testing to artificially minimize emission results. These flexibilities related to minimizing road load coefficients, the use of low-resistance and overinflated tires, the use of high-performance lubricants, turning off auxiliary electric loads (air conditioning, entertainment systems, seat warmers, etc.), among others [106]. In addition, other real-world impacts, such as aging and road gradients, were simply not included in the compliance tests [103]. The growing issues with the NEDC led to the development and adoption of the new and more realistic Worldwide Harmonized Light-duty Test Procedure (WLTP) in the EU in 2017, as well as the introduction of the RDE on-road emissions testing. In addition, the EU has introduced the compulsory application of On-Board Fuel Consumption Monitoring systems (OBFCM) since 2021 with the aim of monitoring the real-world fuel consumption of the fleet.

The emission “gap” also existed in the US when the two-cycle FTP (i.e., Federal Test Procedure) was the norm [107]. The problem was recognized and addressed with the development of the new five-cycle test, which is a better estimate of real-world fuel use and emissions. For instance, the five-cycle test appears to provide a reasonable approximation of real-world fuel efficiency and CO₂ emission rates and may even be slightly conservative [108,109]. The five-cycle test benefited from the European experience and was created explicitly to fend off the potential for a growing “gap” in emission performance.

The gap appears to have created several policy issues in the EU, including recognition of substantially lower emission reduction benefits than intended, overestimation of the emission benefits of diesel vehicles—in particular that the image of clean and climate-friendly diesel vehicles is not in line with reality—local air quality issues, and inaccurate emissions and fuel consumption information for consumers [35,92,105,110,111,112].

But for the purposes of this review, the relevant question is what portion of the gap can be attributed to cheating by vehicle manufacturers and what portion can be attributed to natural differences between on-road driving and official test procedures.

AES exemptions appear to have been misused to the extent of switching off emission controls in various driving conditions that commonly occur, including low and high ambient temperatures, high vehicle speeds, high altitudes, and high engine loads [95]. However, another reason for excess emissions is simply that legislative test protocols do not capture all aspects that affect real-world operation as well. One example is SCR systems in Euro IV/V heavy-duty engines that were not working in low-load, low-exhaust temperature urban conditions, which was rectified after changes in the Euro VI legislation [38]. It is thus challenging to tease out the presence and use of defeat devices simply from the results obtained from regulatory emission tests.

The latest research suggests that the current mean gap between the WLTP and real-world on-road fuel consumption is 15–20% for light-duty ICE vehicles and close to zero percent for battery electric vehicles [104]. This implies that in the EU, the potential impact of cheating and the use of illegal defeat devices by LDV manufacturers in terms of CO₂ emissions from current-production new vehicles may be significantly less than 20%, on average, once the on-road fleet begins to turn over in the next five to 10 years. However, the “gap” impacts from other vehicle types, such as trucks and buses, are less clear.

Cheating appears to result in substantially larger conventional pollutant emissions than fuel consumption and CO₂ emissions. For instance, previous research has established that average on-road emissions for Euro 5 and 6 passenger cars are, on average, about 4 to 5 times the NO_x emission limit and that no engine family had real-world emission levels below the regulatory limits [94]. This is even though Euro 5 and 6 standards were intended to reduce NO_x emissions to alleviate persistent urban ozone pollution (e.g., [113]). This gap may be the result of the use of defeat devices and the limitations of the official test protocols. For reference, in the resolution of the cheating scandal by the HDDE manufacturers in the late 1990s, in-use measurement allowances of approximately 1.5 to nearly 2 times the applicable NO_X limits were approved [114]. However, several additional exemptions or “carve outs” rendered most of the real-world on-road truck emissions unusable for the purposes of the NTE regulation. Meyer et al. [94] state that comprehensive testing in the EU shows that the application of a defeat device is likely present in results from most tests, which is further supported by an analysis of statements made by vehicle manufacturers as to the technological reasons for the observed high emission levels.

3.3. In-Service Vehicles

Vehicle emission performance deteriorates naturally; hence, exhaust emissions increase with age (e.g., [115,116,117]). Beyond that, vehicles can become high or super emitters due to, for example, malfunctioning emission control systems or engines, incorrect repairs, a lack of proper servicing and maintenance, poor emission control system retrofits, the use of improper or low-quality fuels, or tampering. In fact, a large portion of total fleet emissions originates from a disproportionately small number of vehicles, and a large portion of total emissions from a vehicle originates during a disproportionately small fraction of its operating envelop, as has been confirmed repeatedly by, for instance, remote sensing and other empirical studies around the world [21,118,119,120,121,122,123,124,125,126]. Thus, although average per-vehicle emission rates (g/km) have generally decreased over time in absolute terms as automotive technology improves, total emissions from the on-road fleet have become more skewed: i.e., a small fraction of vehicles contributes to the largest share of emissions, leading to the “high-emitter conundrum”.

In addition, the skewness of the high-emitter problem has increased over time [122,127]. In the US, about 1% of the on-road fleet contributed to about 10% of total emissions in the late 1980s, increasing to about 30% in the 2000s [128]. It is this small but high-emissions portion of the on-road fleet that has been assumed to offer the greatest potential for emissions reductions through repair or removal from use [129]. The distribution of on-road emissions reflects two trends [21]: (1) newer and cleaner vehicles making it into the fleet and replacing older, higher-emitting technology. The newer vehicles are an advancement in emission performance driven primarily by stricter emission standards (including durability requirements and improved emission controls). They operate primarily in a low-emissions mode, interspersed by brief and large emission excursions during commanded high-load, transient operation, and (2) the presence of discrete vehicles with elevated or excessive emissions.

As more stringent emission standards continue to reduce fleet-average emissions, identifying vehicles with high in-use emissions, with or without emissions defeat devices, is becoming increasingly important [55]. There is not a universally accepted or exact definition of a high-emitter, but they are typically defined as either exceeding a predefined multiple of the applicable standards or the relative position of their emission rate in the distribution of an on-road fleet [50]. Traditionally, a high and excessive emitter was defined in the EU and US as exceeding the standard by more than 1.5 to 2 times, respectively (e.g., [51,130,131]). The poor emissions performance of modern diesel vehicles seems to have stretched this threshold to 2 to 5 times the standard for real-world emissions [94]. The measurement allowances in the US heavy-duty engine emission regulations of 2007 and 2010, which are on the order of 1.5 to 2.0, may be facilitating an expanded view of a “high-emitter”.

As the odometer mileage increases, the likelihood of a failing vehicle or an excessive polluter generally increases [129,132]. For instance, Ando et al. [133] reported that failure rates are low for new vehicles and reach about 20% and 40% for vehicles 10 years and 15 years old, respectively. Zyl et al. [134] reported the following aging factors for 10-year-old Euro 1 and Euro 2 petrol vehicles: CO—1.79, THC—1.48, NO_x—1.97, and PM₁₀—0.

Deterioration used to be primarily determined by wear and tear and maintenance [115], but the importance of tampering has increased over time. Previous research has found that certain vehicle brands were more likely to fail or have excessive emissions [129,132]. Information about non-compliant vehicles allows authorities to focus investigations on specific vehicle brands and models with consistently high emissions and to activate a manufacturer recall for corrective action. Similarly, vehicle attributes such as age, fuel economy, mileage, engine characteristics, weight, and time of year (meteorology and variation in fuel quality) have been found to be strong determinants of emissions and failure rates [129].

3.4. Tampering by Vehicle Owners

Whereas the Volkswagen scandal highlighted cheating behavior by vehicle manufacturers, tampering is undertaken by regular vehicle owners. Owners resort to tampering to (1) improve performance and/or fuel economy, (2) avoid costs of running the emission control system, (3) avoid costs of necessary repairs to the emission control system, (4) sell the vehicle for a higher price (i.e., odometer tampering), or (5) simply to show off and rebel against norms, a practice by some US diesel pickup truck owners called “rolling coal”.

Again, this is not a new phenomenon; it happens globally and applies to all vehicle types [100] and continues to this day. Back in 1974, the US EPA reported that motorists allegedly requested service stations to adjust or disconnect the emission controls to improve fuel economy. Interestingly, a survey showed that most garages declined to perform the work as they did not want to contribute to a deterioration of air quality, but about 25% of the service stations would accept the tampering request [135]. Tampering was reportedly widespread in the US when emission regulations first required the use of oxidation catalysts (OC). The OCs would be removed because they were thought to impair performance, and there was widespread use of cheaper leaded petrol, which poisons catalysts [136]. Lawson [137] analyzed data from roadside surveys in California and reported that about 10% of vehicles showed tampering in a visual test. Unfortunately, this practice may still be commonplace in emerging economies that are home to an increasing number of older, second-, or even third-hand vehicles, where conformity checks are not sufficiently robust to prevent or deter tampering [138].

Improving performance generally involves reprogramming the engine control unit (ECU) to a different engine map [139], also referred to as re-chipping or re-flashing. Chip tuning is popular, usually easy to install or remove (e.g., to pass in-service testing), and a cheap way to modify or override the factory settings and improve the performance of lower-powered versions of some makes and models. It has been estimated that about 25% of new diesel cars sold in the Netherlands are equipped with chip tuning [140]. The costs vary from about EUR 70 to EUR 1000, showing different levels of sophistication. For instance, chip tuning varies from simply overriding fuel temperature sensors (increasing fuel injection) to intercepting and modifying actual ECU signals with an external “power box” that can reprogram via, for example, the OBD connection if this option is not disabled by the manufacturer. Tampering with the emission control system is generally more involved and can include the physical removal or disabling of emission control components, in addition to changes to the engine management software, as discussed later.

Tampering and the use of defeat devices in the freight sector adds the specific issue of undercutting responsible and compliant operators [141], giving operators with tampered vehicles an unfair economic advantage while causing harm to the environment (refer to Section 3.5.2 and Section 3.5.3). In the US, a transport company was recently sentenced in Federal court for knowingly and intentionally tampering with emissions monitoring devices and had to pay a fine [142].

One solution is to ensure manufacturers produce tampering-resistant vehicles. An example is SCR-equipped heavy-duty vehicles in California that progressively go through driver inducements to prevent bad actions—an alert, engine de-rating, and eventual vehicle immobilization if the DEF tank is empty and not refilled, or if DEF is replaced with water [143]. The same study in California of in-use trucks pulled over for inspection found that empty or water-filled DEF tanks were not a common practice.

EU legislation also includes anti-tampering protections. Tampering with emission control systems is explicitly prohibited (EU Regulation 595/2009), and vehicle manufacturers are required to install systems that deter owners from modifying emission control systems and odometer readings (EU Regulation 2017/1151). The European Commission [100] notes that relevant authorities should test the robustness of anti-tampering technology installed by vehicle manufacturers by manipulating the emission control system using various known tampering approaches [100].

Finally, there are complex legal considerations when it comes to tampering by vehicle owners. It is not necessarily illegal for businesses to sell tampering devices, but it is illegal to install and use them. An interesting case in this respect was a settlement in 2018 between a company that sold aftermarket defeat devices and the US EPA. The company (Derive Systems) violated the Clean Air Act with its sale of approximately 363,000 defeat devices involving engine tuning software. The software enabled the user to remove emission control components, including catalysts, DPFs, EGRs, and elements of the OBD system. Derive also sold parts that bypass, defeat, or render inoperative elements of design that were installed by the vehicle or engine manufacturer to comply with the emission standards. These handheld products—commonly known as “tuners”—allowed the user to easily turn off emission controls originally installed and certified by vehicle manufacturers for regulatory compliance. Under the terms of the settlement, Derive had to stop selling non-compliant tuners into commerce and retrofit existing tuners for compliance. The terms of the settlement squarely focused on functions that limited vehicle emission controls. However, other types of “tuning” are still permitted. The settlement was intended to demonstrate to other manufacturers that products designed to unlawfully thwart vehicle emissions control systems will not be tolerated. In a statement, ironically, Derive noted that this is a very positive agreement for the business as they are the only federally compliant player in the automotive aftermarket industry: “every company in the aftermarket will now be held to the same standards” [107,144]. No such cases are known outside of the US. Effective policies to address tampering will need to consider changes to the legal framework to make selling defeat devices to owners explicitly illegal.

3.5. Emission Impacts of Cheating, Tampering, or Malfunctioning Vehicle Emission Controls

Although emission control in modern vehicles involves a broad range of technologies as discussed before, we focus here on diesel emission control technologies that are probably most susceptible to cheating, tampering, and detection by in-service maintenance. Tampering with SI engine petrol/LPG vehicles is certainly technically possible, but it requires a substantial level of effort and is relatively easy to detect in inspection programs. Importantly, emission control systems in SI engine vehicles appear to be relatively robust, as was discussed previously, and therefore likely less prone to tampering. Nevertheless, three-way catalysts have sometimes been removed and replaced with a de-cat pipe to increase the power output of the engine, but the number of vehicles affected in the on-road fleet is small [139]. In China, vehicle owners allegedly rented a new three-way catalyst before an emissions test, returning it after the test was completed, with significant impacts to the effectiveness of test and repair programs [145].

The type, quality, and durability of an emission control system depend, to a large extent, on the make and model and the applicable regulations. For instance, about ten years ago, diesel cars in the US required either SCR or LNT in combination with EGR to meet US Tier 2 NO_x emission standards, whereas EGR alone was sufficient to meet the EU Euro 5 standards [60,63]. Diesel automakers were reluctant to tackle the stringency of the US NO_X emission standards, leading to very few diesel cars sold there.

3.5.1. Exhaust Gas Recirculation for NO_x Control

Exhaust gas recirculation has been around for years, and it is a proven system to effectively control NO_x emissions in diesel exhaust. EGR introduces cooler exhaust gas back into the combustion chamber to absorb some of the combustion enthalpy and, hence, decrease flame temperature and oxygen content, the main drivers for NO_x production [32]. In addition to optimization of the combustion process (e.g., reconfiguration of engine maps and higher injection pressures), EGR was sufficient to meet Euro 3 and 4 emission standards for diesel cars and generally sufficient to meet the Euro I, II, and III limits for diesel trucks [146]. EGR is commonly used in diesel vehicles but can also be fitted to petrol vehicles. EGR can provide a reduction in NO_x emissions of 20% to 70%, depending on engine speed and load. However, EGR leads to reduced engine efficiency, higher fuel consumption on the order of 1% to 4%, and increased PM emissions due to incomplete combustion (e.g., [13]). Hence, EGR is commonly used in combination with a DPF to meet current PM emission standards.

There is limited information regarding EGR tampering. EGR can suffer from specific problems over time, such as a sticking valve, blocked manifolds, or contaminated combustion chambers due to carbon buildup (i.e., a maintenance issue). EGR maintenance can be costly because, from time to time, engines need to be cleaned due to blockage and internal contamination. The EGR system can also be intentionally disabled by either removing it completely or by simply removing the control to it (i.e., a vacuum pipe or the electrical connection) to overcome a fault, to achieve more engine power, and/or improve fuel economy [143]. If the EGR is disconnected or missing, this may or may not be detected by the OBD system, depending on the sensors fitted to the vehicle [139].

Lee et al. [147] conducted a study in an emissions laboratory and reported that NO_x emissions increased by a factor of almost three in motorway driving conditions when the EGR valve failed (i.e., stuck due to rust and soot particles). On-road tests showed a smaller increase in emissions, 14% to 27%, depending on the driving conditions. CARB [143] reported that disabling the EGR led to 9.3 times the NO_x emission limit. Assuming an average real-world NO_x reduction efficiency of the EGR of 40%, NO_x emissions would increase by about 70% due to EGR removal or EGR deactivation.

3.5.2. Selective Catalytic Reduction for NO_x Control

SCR has mainly been used on Euro IV (2005+) and US EPA 2010 (2010+) heavy-duty vehicles [50,148]. The most common tampering in HDVs relates to the SCR system [149]. SCR requires ammonia (NH₃) to catalytically reduce NO_x into nitrogen (N₂) and water. A urea/water solution or diesel exhaust fluid (DEF), marketed as AdBlue, is commonly used. It is stored in a separate tank that needs refilling just like diesel fuel. A minimum exhaust temperature of typically 200 °C is required for the SCR to promote the hydrolysis of urea into ammonia. Excess urea injection can result in ammonia slip, but this is often controlled with an ammonia slip catalyst (ASC). SCR efficiency depends on the catalyst materials used, the urea dosing strategy, temperature, physical system layout, and driving conditions [32]. SCR can provide a reduction in NO_x emissions of approximately 60% to 95% over a wide operating range. SCR, like other control systems, continues to evolve and improve.

In the US, technology has now been developed, demonstrated, and extensively tested by industry and regulators that can result in an order of magnitude reduction in NO_X emissions from 2010 levels (i.e., taking the current 0.2 g/bhp-hr limit to 0.02 g/bhp-hr). This improved SCR/DPF technology is expected in future vehicles subject to the most recent US HD emission regulations. SCR is commonly used in trucks and more recently in some late-model larger diesel cars and SUVs (usually >3-L engines [13]). In Europe, all Euro 6d diesel vehicles are SCR-equipped.

SCR systems can be either defective or intentionally manipulated. Using DEF increases the operational cost to truck operators, creating an incentive to find ways to circumvent the SCR system and reduce costs by minimizing DEF usage [143]. In China, Ding et al. [150] reported that the use of water in the SCR system in HDDV significantly undermined the effectiveness of NOx control.

SCR emulators effectively switch off the SCR systems to save on DEF consumption and also avoid expensive maintenance of the SCR system [151,152]. There are two types of SCR emulators. Hardware emulators are small electronic devices that are installed in the truck. They can be located at positions that are difficult to reach (e.g., below the trailer) and thus difficult to find in a roadside inspection. Software emulators are uploaded to manipulate engine software without any physical modification of the truck. It is currently not possible to detect these emulators. Visual inspection will therefore only detect a fraction of SCR emulators. Whereas DEF costs are substantial for trucks, up to EUR 2000/year [151], DEF costs for cars are insignificant. It is therefore unlikely that car owners would go through the trouble to purchase and install SCR emulators. For trucks, on the other hand, there is a clear financial incentive as SCR emulators are available, starting at the price of about EUR 30, and are easy to install [151].

California has now introduced a new in-use inspection program for HD diesel trucks to proactively ensure adequate in-use emission performance. The program involves inspection testing of modern diesel vehicles equipped with current emission controls. The new testing goes beyond opacity to include OBD interrogation, visual inspection, and new periodic testing requirements in a referee testing network. Roadside emission monitoring devices, which may include remote sensing devices [126], CARB’s portable emissions acquisition systems, and automated license plate recognition cameras, will assist with enforcement of the new regulation [153]. Other US states are expected to follow suit.

EU and US legislations require DEF monitoring. If the DEF tank is empty, the engine cannot start [149]. If the OBD system detects an issue with the SCR system, it will likely de-rate the engine [148]. SCR emulators override this functionality and drastically increase NO_x emissions, but the impact depends on driving conditions. For instance, Rexeis [154] reports that a Euro V truck and trailer combination with a non-functional SCR system would exceed Euro III NO_x emissions, with an increase that can be as high as a factor of four under motorway operation. CARB [143] also reported an increase in NO_x emissions of four times when DEF supply or SCR are tampered with. Assuming an average real-world NO_x conversion efficiency of the SCR of 70%, NO_x emissions would increase by a factor of more than three due to an SCR emulator. But this may be significantly higher as SCR, vehicles are calibrated to produce high NO_x engine-out emissions [139].

3.5.3. Diesel Particulate Filters for PM Control

DPFs are usually highly effective in reducing PM emissions (both mass and particle number). To avoid blockage, a regeneration phase burns off the collected soot, a step that must be repeated approximately every 500 km and lasts 10 to 15 min [13,19]. During regeneration, particle mass and number emissions increase by two or three orders of magnitude [5].

Regeneration can be achieved actively or passively. Active DPFs rely on continuous and periodic filter regeneration to burn off the accumulated soot. Passive DPFs use a catalyst and regenerate continuously using oxygen, nitrogen dioxide, or a combination of both. A DPF imposes on performance, has a fuel penalty, and can become a hassle for owners who drive only short distances [13]. DPFs are used in all Euro 5s, six diesel LDVs, Euro VI (2013+), and US EPA 2007 diesel HDVs [43,50]. The DPF, similar to the TWC in gasoline vehicles, is now standard equipment in all new diesel vehicles in the EU (LDV and HDV), although these primarily appear in heavy-duty applications like commercial freight in the US.

DPFs can become damaged (e.g., breaking, cracking, or melting of substrate) or blocked, which means the filter needs to be either cleaned or replaced [92]. The cause is not always just a restricted DPF or a failed regeneration. For example, a clogged air filter, excessive engine oil consumption, a faulty injector, a faulty turbocharger, or faulty sensors can disrupt the DPF. Excessive engine oil consumption can be particularly detrimental. The heavy metals (sulfated ash) and heavier hydrocarbons found in lubricants lead to blockage and may lead to excessive temperatures during filter regeneration and possibly catastrophic failure, including fire. If these problems are not resolved, a cleaned or new DPF will probably become restricted more quickly than usual [155]. As diesel vehicles age, the DPF will require cleaning of deposits that are mainly metals from engine wear and the lubricant.

Replacing a DPF is expensive and estimated at EUR 1000–5000 for a passenger car. A truck DPF replacement is substantially more [54,139,155]. The cost of DPF cleaning is EUR 300–400 [139,155], which can be prohibitively expensive to some, especially because it must be repeated every 100,000 to 200,000 km of driving.

There is, therefore, an incentive to look for cheaper options. Mechanics commonly offer physical or software (i.e., delete kits) removals of DPF for a price. These methods include replacing the DPF with a piece of exhaust pipe or with a dummy DPF, removing the filter substrate and welding the can back together, or drilling out the DPF [139,156].

The removal of a DPF is also offered as part of “chip tuning” and serves to achieve better engine performance and improved fuel economy. But it is noted that a well-designed engine-DPF system in a well-maintained vehicle can comfortably meet the expected performance and emission control. Occasionally, along with the elimination of the DPF, the EGR system is also deactivated or removed as part of chip tuning [155]. Tests on a diesel car with various chip tuning options suggest that emission impacts are highly variable (depending on the chip tuning device and pollutant) and vary from a reduction or stabilization to an increase in emissions [140].

The OBD system should detect DPF failures [157], but it appears this is often not the case [33]. Emission tests have shown that only one out of three high-emission vehicles equipped with a DPF produced an emission-related OBD fault code. This indicates that OBD systems are not yet designed to detect DPF failures (e.g., [43]). In any case, in addition to DPF removal, the OBD system can also be modified to trick it into believing that the DPF is still there and working. OBD tampering can be performed in different ways [139]. An OBD simulator uses the OBD scanner port and controls what is reported (e.g., DTCs) and how the OBD functions (e.g., MIL turned off). OBD sensors (e.g., the sensor measuring the pressure difference across the DPF) can be modified directly to feed the OBD incorrect data. The ECU can be reprogrammed (ECU reflashing) to disable emission control equipment and the OBD MIL and fault codes [50]. The DPF software routines are eliminated from the vehicle’s engine management system to prevent “undesirable” warning lights on the dashboard and ensure the eliminated DPF goes unnoticed during the technical inspection [155].

DPFs significantly reduce the real-world PM emissions of diesel cars and SUVs to a typical value of 1–2 mg/km, including the emissions associated with regeneration events [54]. Wall-flow DPFs can provide an average reduction in PM emissions of 85% to 99.9%. Assuming an average real-world PM removal efficiency of the DPF of 98%, PM emissions would increase by a factor of more than 50 from a vehicle due to DPF removal. Similar increases of a factor of 25 to 100 have been suggested by other research [54]. At idle, a car without a DPF can emit 1000 times more particles than a car with a functioning DPF [92], although absolute emission levels at idle are relatively low [158].

3.6. Fleet-Wide Impacts of Cheating and Tampering

Due to the significant impact of excess emissions on air quality, cheating by vehicle manufacturers and tampering by vehicle owners are receiving increased attention in the EU and the US. It is, however, difficult to precisely determine the fleet-wide impacts of cheating or tampering. Although we have a reasonable idea of what intentional interference with the emission control system does to the emissions from individual vehicles (Section 3.5), assessing fleet-wide impacts requires reliable information on the extent of cheating and tampering and how widespread the practices are. This type of information is scarce, and the available information suggests significant geographic variability.

Table 1. Overview and summary of key points by impact category retrieved from the literature.

Impact Category	Key Point Summary
Health	About 30 hospital admissions and 10 to 350 premature deaths in the US in the period 2009–2015 with an associated excess damage of USD 430 million have been directly related to excess emissions from the Volkswagen vehicles involved in the dieselgate scandal [8,9,44,159]. The impact in the EU would have been substantially larger due to the larger number of affected vehicles and differences in compliance actions [10]. Approximately 482,000 cars were implicated in the VW dieselgate scandal in the US, compared with more than 8.5 million in the EU [60,62]. Alexander and Schwandt [160] reported that an additional VW cheating diesel car per 1000 cars in the US increases PM2.5, PM10, and ozone by 2, 2.2, and 1.3%, respectively, while the low birth weight rate and infant mortality rate increase by 1.9 and 1.7%, respectively. Similar impacts are found for acute asthma attacks in children. These health impacts occur at all pollution levels and across the socioeconomic spectrum.
Economy/health	In Ireland, the economic costs of dieselgate (cheating as well as poor performance of on-road diesel LDVs) were estimated to be EUR 540 million, with 70 additional deaths and a total burden of more than 600 disability-adjusted life years (DALYs). Oldenkamp et al. [62] estimated the emissions and health impacts due to the VW scandal to be 526 ktons of excess NOx emissions in the EU and US in 2009–2015 and associated burdens of 45,000 DALYs and USD 39 billion in health costs, with the bulk of these impacts (98–99%) in the EU. They note this is more than five times larger than the USD 7.3 billion that VW had set aside to cover the worldwide costs of the diesel emissions fraud.
Use of tampering devices (SCR)	Visual inspections in the EU suggest 1–3% of trucks have installed (hardware) SCR emulators. However, vehicle emission measurements (plume chasing) in Germany and Austria suggest substantially higher violations up to 35% (Euro V) and 25% (Euro VI) [151]. A more recent plume-chasing study [152] showed regional differences, with the percentage of suspicious high-emission trucks being less than 10% in Denmark and 40% in the Czech Republic. Most of these trucks were found to have defects or software issues (about 60–65%), and about 25–30% of the vehicles were tampered with, with the remaining emissions being due to cold starts. A vehicle measurement campaign in Austria (remote sensing) suggested SCR tampering devices in 15% of the trucks inspected [149]. A comparison of Adblue sales data and expected Adblue consumption estimated that about 15–20% of HDVs operated with ineffective SCR systems in the EU [154]. A substantially higher proportion was found for Euro V trucks in Spain, where 47% of the trucks inspected may have been illegally tampered with [149]. A remote sensing campaign in Denmark combined with roadside inspections or RSI [161] measured 874 trucks and concluded that approximately 25% are cheating with SCR catalysts. DEPA reported that out of nine vehicles inspected, two had a cheating device, and four had malfunctioning SCR catalysts (no DEF, engine problem, or sensor issue). The latter could indicate software tampering rather than malfunctioning. The fact that only about 30% (3 out of 9) of trucks in Denmark are operating with fully compliant emission controls is concerning. Li et al. [162] investigated 66 Chinese heavy-duty vehicles and found that 15 vehicles had excessive emissions and that 2 vehicles were tampered with. Tampered vehicles accounted for 3% of the sample, but emitted 1.4 times more NOx emissions when compared with the other vehicles.
Tampering of diesel particulate filters (DPFs)	Research in Europe suggests DPFs are frequently disabled or removed if there are problems with DPF regeneration [54,156]. An investigation of diesel vehicles in 2016 in the Netherlands deployed a stationary PN emission test and found elevated particle number emissions for 5% to 7% of tested vehicles [54]. A higher figure was found in neighboring Belgium, where about 15% of vehicles had elevated PN emissions [92]. Elevated PN levels are attributed to malfunctioning filters (e.g., cracked) and potential filter removal [54]. A further survey in the Netherlands suggests that about 1.5% of diesel vehicles have their DPFs removed, amounting to about 20,000 vehicles. However, DPF removals may increase with vehicle age. On the other hand, DPF technology may mature, reducing the malfunctions and maintenance issues. It has been suggested that the reliability of DPFs has improved over time [92], which means high-emission vehicles are increasingly likely to be tampered with. This would also depend on the region. For instance, there is some evidence [21] that new light-duty diesel vehicles sold in Australia may have low-quality DPFs, possibly due to lagging emission standards and a lack of a PTR program. In addition, regulations could be improved to require car and engine makers to design and deploy emission control systems like DPFs and SCRs with anti-tampering in mind. DPF problems in the UK are reportedly widespread, and DPF removal services are widely advertised across the UK [163]. Roadside checks in the UK in 2017 found that 8–20% of trucks had cheat devices fitted [141]. The cheat devices found were illegal engine modifications, removed DPFs, EGR removal, or bypassing and SCR emulators. The drivers and operators were given 10 days for a fix or face a GBP 300 fine and having the vehicle impounded. Visual checks conducted by the (UK) Department for Transport caught almost 1200 vehicles with a removed DPF. However, repair centers offer removal in a way that is not visible through visual checks. Clean Air London estimates that tens to hundreds of thousands of vehicles have been tampered with this way. It is expected that the problem will grow as vehicle warranties expire and DPFs age. A UK remote sensing database with 50,000 emission measurements was interrogated for evidence of significant removal of DPF filters in the diesel passenger car fleet and at what levels DPF removal could be detected [164]. Euro 5 and 6 vehicles were found to have low emissions, suggesting effective DPF functioning and no evidence of significant levels of DPF removal. The analysis suggests that DPF removal or DPF failures can be detected in remote sensing data at rates of approximately 1%. A similar study in Australia found that a few percent of new diesel passenger vehicles appear to have DPF issues, although the sample size was small [21]. Quiros et al. [165] reported that 8% of heavy-duty vehicles in California had damaged or malfunctioning DPFs and that these vehicles contributed disproportionately (approximately 70%) to fleet-wide PM emissions. The study also showed that some engines and emission control systems are not sufficiently durable to effectively operate and control emissions over the lifetime of the vehicle. In contrast, roadside inspections in California of SCR-equipped trucks showed that all of the 243 trucks were using proper DEF solution, no tampering was present, and no DEF warning lights or audible alerts had been initiated [143]. However, US EPA [15] reported a high tampering rate of 30% for diesel pick-up trucks in the USA

shows a summary of key points retrieved from the international literature. There is large variability and uncertainty in the reported occurrence of tampering by vehicle owners, as well as the associated impacts. In comparison, the impacts of cheating on fleet emissions may be easier to quantify since the number of vehicles affected is generally known. In contrast, it is not yet feasible to quantify the fleet-wide impacts of tampering on emissions with a sufficient level of confidence. What is clear is that, given the effectiveness of modern (diesel) emission control systems, the available evidence suggests that impacts are undeniable, likely substantial, and cannot be ignored.

Clearly, understanding of real-world emissions and the impacts of cheating and tampering is an integral part of an effective clean vehicle standard-setting process. Compliance with these emission standards directly affects the ability to meet ambient air quality goals. The process is iterative. It relies on quantification of emissions from all sources, an inventory of the sum of all emissions for each source, and a robust understanding of the pollution carrying capacity of each geographic region. And it employs sophisticated chemical models of the atmosphere and the dynamic interactions of the mixture of pollutants present in ambient air. NO_x emissions are well-established precursors of ozone formation and one of the most pervasive urban air pollution problems around the world. When unplanned and additional NO_x pollution is added to an airshed, be that accidentally or intentionally, additional ozone formation will occur and lead to additional adverse health impacts on those exposed.

4. In-Service Emission Test and Repair Programs

4.1. The Three Common Types of In-Service Emission Programs

The rationale for an in-service emissions program is that modern vehicles are dependent on properly functioning and well-maintained systems to keep exhaust pollution levels in check [166]. Without effective in-service emission programs, compliance is significantly weakened [17].

The importance of properly functioning vehicle emission control systems is obvious. Horowitz [167] stated fifty years ago that “It is generally recognized that mechanical malfunctions of automobiles can lead to abnormally high exhaust emissions of air pollutants. The detection and control of these emissions requires programs of inspection and maintenance for in-use automobiles”. In addition, vehicles that fail and are repaired under these programs are expected to return to lower-emission performance. The threat of failing an inspection is thought to deter tampering, misfuelling, and encourage better maintenance [168]. In-service emission programs are designed to identify large faults rather than gradual deterioration of the emission controls. These programs are needed to ensure that the benefits of new-vehicle control technologies are not lost through poor maintenance and tampering.

In-service emission programs may involve national and local authorities, inspection stations, and car owners. Therefore, for these programs to be effective, they must include a simple test that is technically sound, socially acceptable, and not too demanding in terms of time and costs (testing and repairs) to the vehicle owner. [16,131,169]. Programs exist that provide financial assistance for repairs for income-eligible consumers. In-service emission programs can take different forms, and the main types include the following (e.g., [50,137,166,168]):

• Periodic inspection—registered vehicles are inspected at regular intervals at a dedicated inspection facility and must pass a test as a condition for (re)registration;
• Self-certification—fleet owners are allowed to conduct periodic tests in their own maintenance facility and report the results to the relevant government authority;
• Roadside inspections—a sample of on-road vehicles that are randomly pulled over and tested at specific locations.

Periodic inspections are the most common form of in-service programs used around the world. They are different from in-use compliance (in-use conformity, conformity of production, and surveillance testing) and enforcement programs, which are primarily aimed at early detection of excess emissions caused by manufacturer defects [50].

There is little published information about self-inspection programs, but it is clear that effective self-inspection and self-certification require robust and strict oversight with frequent auditing [50,168].

Roadside testing in the US has been primarily used in California [170]. It is conducted throughout Europe to check emissions over time (Directive 2000/30/EC) and has been considered favorably for reducing test costs and including all vehicles, even unregistered vehicles. But there is little empirical evidence to support these observations [168]. Although roadside pullover data can be useful, historically, it has only been possible to consistently obtain such data in California (e.g., [137]). Aside from the logistical and public acceptance difficulties of roadside testing, it is expensive and voluntary. Therefore, some argue that a roadside pullover sample is not an accurate representation of the fleet [170,171].

4.2. Periodic Test and Repair (PTR) Programs

The purpose of periodic in-service emission testing is to allow authorities to check that vehicles are well-maintained and conform as far as possible to their expected emission levels [16,115]. The terminology regarding periodic in-service testing and repair varies between jurisdictions. In Europe, they are commonly referred to as periodic technical inspections (PTIs). In the US, they are known as inspection and maintenance (I/M) programs. Modern in-service testing programs are not really inspection and maintenance programs, but rather inspection and repair programs. If a vehicle has excessive emissions, it is likely the result of either a system component failure or tampering. A faulty component needs to be replaced or fixed, rather than undergo maintenance, as was traditionally the case—for example, by adjusting the carburetor or other simple devices [127]. Indeed, mandatory tune-ups where vehicles were adjusted to a manufacturer’s specifications by private mechanics [167] are a thing of the past, given the complexity of modern vehicles and software-related tampering practices. Here, we refer to PTI and I/M programs as periodic (emission) testing and repair (PTR) programs.

When looking at global vehicle inspection programs, most information comes from the US and the EU, where most experience with light-duty vehicle programs has been in the US, dating back to the early 1970s. The first introduction of such a program was in the state of New Jersey in 1974. Since then, many other US states have followed suit [136]. The program was implemented because of the requirements in the federal Clean Air Act (1970 amendments) to develop a plan to bring ambient air quality levels into compliance in areas that exceed the National Ambient Air Quality Standards (NAAQS) [172]. PTR programs were regarded as more cost-effective and less disruptive than alternative strategies [172], although the actual cost-effectiveness of PTR programs has been challenged (e.g., [130,169,173,174]).

In the European Union, periodic inspections were first introduced in the early 1980s in some member states [50]. In China, in-service programs were rolled out later in 1999, preceding China’s new vehicle standards [145,175]. Today, programs have been implemented in various countries, including Bangladesh, Brazil, Canada, Hong Kong, Colombia, India, Indonesia, Japan, Malaysia, Nepal, New Zealand, Pakistan, Philippines, Singapore, South Korea, Sri Lanka, Taiwan, Thailand, Turkey, and Vietnam (e.g., [33,175,176]). But interestingly, they have not been implemented in Australia, despite the fact that Australia has developed its own inspection test cycle. A substantial amount of work was performed in Australia in the 1990s in preparation for the diesel vehicle National Environment Protection Measure (DNEPM). This included the development of a short I/M test, called the DT80, which correlated well with real-world Australian drive cycles (e.g., [115,177]). Relatively effective programs have been reported for Chile and Mexico City, but many are allegedly ineffective programs because they tend to be ignored by motorists or are allegedly fraudulent (e.g., India, Kazakhstan, Colombia, Philippines, Egypt, and Indonesia) [166].

An effective PTR program requires the following fundamental elements [17,178]:

• Comprehensive fleet coverage;
• A suitable test procedure;
• Vehicle compliance and effective enforcement;
• Technician training;
• Routine quality control;
• Periodic program evaluation;
• Minimization of waivers and exemptions (e.g., new or older vehicles).

This means that in-service programs need to be properly designed, well-funded, politically supported, and staffed with technically competent personnel [17]. The test and repair program consists of the following steps:

• Vehicle identification via license number, Vehicle Identification Number (VIN), and registration of the odometer reading.
• Visual inspection of the emission-control equipment to determine its presence or if it is absent, disconnected, modified, or defective, and to identify any leaks that would affect the emission measurements [33]. The outcome of this step would be “passed”, “failed”, or “tampered” [132].
• A functional pass or fail inspection that includes examination of the OBD system and other checks, such as fuel cap integrity [132]. The idea behind OBD inspections is that it will fail vehicles if either the emission control components are or have been malfunctioning or if the sensor monitoring emissions control components are malfunctioning [127]. In the US, an exhaust emission test while the vehicle is operated on a chassis dynamometer is required after a failed OBD-based I/M check. An exhaust emission test is used in the EU for assessment of Euro 5/6 vehicles if they fail the OBD test, and the requirement to test older-technology vehicles depends on specific policies in each member state.
• Performance of an emissions test (if required). There is a range of options. The outcomes of this step could be “pass”, “fail”, “excessive polluter”, or “aborted” [132]. The test is primarily a measurement of exhaust emissions with an approved instrument, but it may also include examination of the evaporative emission control systems [169]. After a failed inspection, further diagnosis will be necessary to determine the cause of the technical problem. Given the complexity of modern emission control systems, diagnostic skills and technical knowledge are essential but may be lacking in some PTR programs.

At the end of an inspection, a vehicle owner will receive a vehicle inspection report from a certified repair technician. Owners of vehicles that fail the test generally have a certain amount of time (e.g., one month) to seek the necessary repairs before returning for a re-inspection. A “pass” test is required for vehicle registration renewal. The repair technician is required to enter the repair data, such as components replaced, emissions results, and repair costs. Naturally, PTR programs require strong enforcement to prevent unscrupulous activities of corrupt technicians or station owners (“clean piping”).

4.3. PTR Program Regulations

Established in 1970 and amended several times, including last in 1990, the US Clean Air Act (CAA) is the federal legislation for controlling air pollution. The 1977 CAA amendments mandated PTR programs for states persistently in nonattainment of the ozone NAAQS, although the general nature of the guidance resulted in programs that varied significantly in design and implementation [127,170]. When the 1977 Amendments failed to produce clean air, the CAA was amended again in 1990. Among its many requirements, the new legislation mandated ‘‘enhanced” motor vehicle PTR programs for light-duty vehicles (petrol) < 8500 lbs (3.8 t) in the most seriously polluted areas—those designated severe or extreme non-attainment of the NAAQS [179,180]—in addition to the “basic” program (idle tests) in areas designated moderate or marginal nonattainment [170,181]. The Clean Air Act Amendments of 1990 mandated that areas with the worst air quality adopt an enhanced PTR program and vehicle emissions testing at centralized, test-only facilities [127]. The US EPA revised its 1992 inspection mandates several times and, by the end of 2001, in effect abandoned emissions testing entirely as a requirement for 1996 and newer OBD-II-equipped vehicles [180].

As will be discussed in Section 4.5.1, collectively, US states used just about every conceivable test on cars, ranging from component inspection (with no actual emission testing) to no-load tests (idle and two-speed idle), to a panoply of loaded tests (ASM2, ASM2525, ASM5015, IM240, MA31, RI2000, etc.), to OBD testing [166]. The acceleration simulation mode (ASM) was developed by California as a less costly and complicated but equivalent test to the federal IM240 test.

The first consolidated EU Directive 96/96/EC (and 2001/9/EC) regulated in-service test programs and safety inspections. The directive effectively sets a floor on in-service testing and allows member states to take stronger steps at their discretion. The European Union created four categories for cars referencing both age (pre- and post-1986) and technology (catalytic converters, OBD). The current legislation is contained in 2009/40/EC and 2010/48/EC. In addition, Directive 2000/30/EC addresses emission measurement during roadside roadworthiness tests [50]. In 2014, the European Union adopted a package of three directives aiming to improve road safety through minimum common requirements for annual roadworthiness tests and roadside inspections of all vehicles within the EU. This included Directive 2014/45/EU on periodic roadworthiness tests, referred to as periodic technical inspections (PTR), which replaced the previous Directive 2009/40/EC. Directive 2014/45/EU sets out the maximum time intervals for inspections of vehicles by authorized testing centers. The directive requires all passenger vehicles and light-commercial vehicles (M1 and N1 vehicles) to be inspected at a minimum of four years after the vehicles are registered and then every two years subsequently. Member states may choose to implement requirements for more frequent inspections. For example, Malta and the United Kingdom require inspections of M1 and N1 vehicles three years after registration and then annually thereafter. For HDVs, the directive requires all M2/3 and N2/3 vehicles to be inspected (at a minimum) annually starting one year after the vehicles are registered. Each member state has an emission testing scheme, which takes the EU legislation as the minimum requirement, but with adaptations to suit the local situation [33].

In Europe, EU legislation (Directives 2009/40/EC and 2010/48/EC) essentially requires for vehicles with positive ignition (petrol) engines, the measurement of the CO concentrations in the exhaust with the engine at idle and high idle, as well as the calculated lambda value (i.e., the normalized air/fuel ratio) by measuring the CO₂, O₂, CO, and THC volumetric concentrations at idle for lambda-controlled vehicles. For diesel vehicles, the measurement is performed for exhaust opacity or smoke using the free acceleration test. The measurement of NO, NO₂, or PM is not currently a requirement of the EU legislation. However, the particle number test for DPF-equipped diesel vehicles was introduced in Belgium, the Netherlands, and Switzerland in 2022 and 2023. In March 2023, the PTI particle number test was introduced by the European Commission as a Recommendation [182]. In July 2023, Germany started with the PTI-PN test and the European Commission has just proposed a comprehensive package to address the issues discussed in this paper.

4.4. On-Board Diagnostics (OBD) and PTR

While some in-service deterioration can be identified relatively easily (e.g., excessive smoke), others, like reduced fuel efficiency or performance (e.g., faulty fuel cap and reduced catalyst efficiency), are more difficult for the vehicle user to spot [136]. This changed with the introduction of OBD, which is designed to detect malfunctions in electronically controlled systems and essentially encompasses the entire engine and emission control system [183]. OBD testing is now an integral part of test and repair programs around the world, where technicians use an OBD scanning tool to retrieve fault codes for malfunctions. OBD identifies where repairs are needed [50,145]. The most common problems detected by the OBD reportedly relate to oxygen sensors, engine misfires, and EGR [51].

The potential benefits of OBD are comprehensive fault detection, preventative maintenance, the use of more durable components by vehicle manufacturers, reduced costs for vehicle owners by addressing (minor) issues early on and within the warranty period, and targeted and faster repairs [51,183]. It is therefore not surprising that OBD was initially considered to be a feasible replacement for emission measurements at periodic inspection facilities, thus reducing costs and inconvenience for vehicle owners [47,51]. For instance, in the early 2000s, the US EPA anticipated that OBD-II-equipped vehicles could avoid traditional emissions tests and rely solely on OBD-based inspections.

However, OBD was never designed for the specific purpose of periodic inspections. OBD is a diagnostic tool that identifies malfunctioning components operating outside their design specifications, rather than specific emission thresholds [51]. This also means that operation outside design specifications may not necessarily lead to elevated emissions, although components may still require repair (e.g., post-catalyst oxygen sensors). The component fault may also indicate an intermittent emission problem or problems under driving conditions outside of the regulatory emissions test. For instance, Durbin and Norbeck [51] tested 77 OBD II vehicles with the MIL illuminated and reported that 62% of the vehicles had emissions below the certification standard. Thus, problems that might not be identified by an emissions test may be revealed by the OBD. For instance, low emissions were found for vehicles that had passed steady-state or IM240 tests but had the MIL on [51].

On the other hand, OBD thresholds for fault detection are set high to avoid unnecessary malfunction warnings for vehicle owners. Consequently, a vehicle with, for example, a defective DPF or high NO_x emissions will often not be detected by the OBD because those faults might remain below the MIL threshold and no diagnostic trouble codes will result (e.g., [54,148]). It has also been reported that OBD misses emission issues when the battery is disconnected or someone has reset the OBD monitor, a practice referred to as pre-test code clearing [180]. OBD systems in China IV (~ Euro IV) HDVs even allowed for the removal of parts of the emission control system (e.g., DPF) without being detected [50]. The error was corrected in the Euro VI standards.

For the purposes of effective periodic inspection, OBD appears to result in too many cases where emission-related defects in the system are not detected. For instance, CITA [148] reported that 50% of (EU) vehicles with induced failures showed increased emissions beyond the MIL threshold, yet the MIL was not illuminated. Nevertheless, the use of OBD for a functional check of emission-control systems is permitted in the EU Directive 2009/40/EC as an alternative to emission measurements at engine idle (SI engines). In addition, for vehicles up to Euro 5/V, an exhaust emission test is used for assessment, whereas OBD can be used instead of an emission test for Euro 6/VI [148].

US states also moved to abandon emissions tests and implement OBD-only tests in the early 2000s. It was found that OBD failed to find high-emission vehicles, and OBD and dynamometer emissions tests identified different problems, the so-called “overlap issue” [127]. California, for example, reported that OBD tests missed more than 75% of dynamometer-based (ASM) emission test failures, with similar results reported for other states. Colorado found that the emissions tests identified high-emitters 91% of the time, while OBD identified high-emitters 53% of the time. Research indicates that only a small fraction (about 10%) of vehicles failing an OBD test also failed the IM240 test.

In relation to potential tampering with emission controls, historic per-vehicle OBD data can be examined. Readiness codes (RCs) provide information regarding the state of the OBD system since the computer memory was last cleared. If RCs are set, then an emissions evaluation based on OBD is valid [148]. A readiness profile that has changed from one inspection to the next and which is inconsistent with similar vehicles may indicate tampering has taken place [183]. In that case, emissions testing is required to detect faults or tampering [148].

A potentially more efficient system is to add telemetry to OBD systems, sometimes referred to as remote OBD or OBD III [145,168,183]. This approach uses radio transponders (e.g., currently used in e-toll collection systems) to report an emissions problem directly to authorities, either continuously or on a regular basis (e.g., the VIN number and fault codes). If this works well, there could be significant emission benefits with shorter time periods between detection and repair, and the need for periodic inspections could disappear because only vehicles with problems would have to be tested and fixed [183,184]. Recent examples of this approach are the Californian Real Emissions Assessment Logging (“REAL”) program [185] and a Chinese pilot program with 30 urban buses [145]. We are moving toward a world where vehicles could send notifications when there is a problem. Otherwise, it is free to keep on running.

4.5. Exhaust Emission Test for a PTR Test

While emission regulations for new vehicles apply to the original auto manufacturer, are relatively detailed, and generally require specialists and expensive laboratories or on-road testing equipment to demonstrate compliance, by necessity, a lower level of sophistication and more practicality (i.e., time and costs) are needed for an in-service emission test under the responsibility of the vehicle owner [167]. In-service tests target the vehicle owner, are currently based on shorter, simplified vehicle operation, involve the measurement of fewer pollutants (typically CO, HC, and diesel “smoke”, although a NO_x test has been included in some), and make use of equipment that is less precise and less expensive than that used for a regulatory test. Thus, it becomes a screening test.

4.5.1. Tailpipe Exhaust Emission Tests

Several short in-service exhaust emissions tests have been developed in the US, Europe, and other countries like Australia, and they can be classified into three categories [16,33,130]: (i) no-load idle, (ii) steady-state, and (iii) transient.

• In a no-load idle test, there is no load exerted on the engine other than simply the inertial weight of the moving engine components (i.e., piston and camshaft) while the car operates with the transmission in the neutral position. The idle test (normally idle at 400–1250 rpm and high idle at 2500–3000 rpm) is considered the basic emissions test for SI (petrol, gaseous fuel) engine vehicles, measuring only CO and HC [50,127,131,132]. These basic test programs are still widely used in the US and elsewhere because they are fast, cheap, and easy to perform with the minimum possible testing equipment [181]. Moreover, if a vehicle cannot be tested on a dynamometer because, for instance, it is too large, heavy, or is a 4WD, the idle test may be the only viable option [132]. The effectiveness of the test is, however, limited as it excludes loaded engine operation (transient or steady state) when the bulk of emissions are generated. A recent study [23] confirmed that the current four-gas test is not suitable for the detection of elevated NO_x emissions. Further development of this test procedure is expected to make it effective for better detection of NO_x emissions from petrol vehicles. An additional drawback of the basic test is that it is not adequate for measuring the all-important ozone precursor—NO_x emissions. From 2012 to 2020, a new idle test was developed in the Netherlands for modern diesel vehicles. The application of wall-flow DPFs reduced the particle emissions from diesel vehicles to nearly zero. Thus, DPF failures can only be detected with more sensitive test equipment. Low-cost particle counters were defined and developed for measuring the particle number concentration in emissions at a low idle speed, which is referred to as the low-idle PN test [23,43,186].
• The Free Acceleration Smoke (FAS) test (also referred to as the snap-idle test in the US) is another example of a no-load test, and it is commonly used in PTR programs around the world for (HD) CI (diesel) engine vehicles without a DPF. In the diesel smoke opacity test, the vehicle is operated through a sequence of so-called “free accelerations” where the inertia of the engine components provides the only load on the engine. The test results in a measurable amount of smoke for the opacimeter. Smoke is an undesirable pollutant in its own right, and reducing opacity levels may also tend to reduce particulate emissions. The presence of smoke in diesel exhaust is suggestive of poor combustion resulting from a malfunction, maladjustment, or use of improper fuel [33]. The FAS test has been shown to successfully identify vehicles with fuel pump and fuel injector issues in the past [50,167]. However, the correlation of smoke opacity with PM is poor at best since the test was developed for higher-emitting, mechanically controlled engines and no emission after-treatment systems [33,177]. It has been suggested that other pollutants, such as CO and HC, may be better predictors for PM emissions [187]. And in DPF-equipped diesel or gasoline engines, a test of particle number emissions is arguably a better option, as has been suggested and applied in Europe and Mexico [188,189].
• During the 1970s, several loaded tests were developed in the US, but they saw limited application due to costs. The simplest loaded test is steady-state operation of the vehicle at constant speed and acceleration. But any loaded test requires a dynamometer, correspondingly either a dynamometer with steady-state power absorption or an eddy current transient dynamometer. The ASM (Acceleration Simulation Mode) test is an example of this type of test to measure CO, HC, and potentially NO_x emissions from SI engine vehicles [50,190].
• In transient load tests, vehicles are driven on a dynamometer following a specified driving schedule (i.e., a drive cycle) and require a constant volume sampling (CVS) dilution tunnel and laboratory-grade emissions analyzers. In the US, the IM240 test is a well-known example of this kind of test for LDVs measuring CO, HC, and NO_x emissions [50,116]. The test was a part of “enhanced PTR programs”, which were regarded by some as the “gold” standard [169]. Another example is the Australian DT80 test for diesel HDVs measuring PM and NO_x emissions [50]. While the measured emissions on loaded tests may not be representative of all in-use emissions, they have been reasonably accurate in identifying vehicles with high emissions [117,178]. A downside of testing on a chassis dynamometer is that the vehicle may be able to “detect” and cheat on the test by shifting to cleaner engine operation [168], as was discussed earlier.

4.5.2. Cut-Points

The amount of emissions measured in the exhaust determines whether a vehicle passes or fails a PTR test. The emission limit is referred to as the cut-point [127,133,169]. Cut-points are typically set high relative to the applicable emission standard. For instance, fail opacity in the FAS test is typically higher than 40%, which means that trained smoke inspectors are generally able to visually screen vehicles that would pass the test [168]. In general, PTR programs do not fail vehicles unless they have exhaust emissions that are typically 2–7 times higher than the certification standards [17,127].

Importantly, cut-points are determined by local fleet emissions performance and practical and political considerations. Typically, the test procedure and cut-points are tailored to the local conditions and specified to achieve a 10% to 25% failure rate [50,127,132,178]. For instance, PTR test results in California indicate a failure rate of 12.4%, of which 4.6% would be quantified as excessive emitters or “gross polluters” [132]. A lower failure rate is considered to be ineffective at reducing pollution, whereas a higher failure rate could be politically untenable, too costly, and may overwhelm the vehicle repair industry [17,191].

Figure 1 shows the basic principles of an in-service emissions program. The emissions test data are for illustrative purposes only. The dots represent test results for individual vehicles over a benchmark emission test and a short and simplified in-service emission test (any of the previously discussed tests in Section 4.5.1). The associated vehicle emission standard for the target pollutant is shown as the lower dashed horizontal line. The higher horizontal line is set as a percentage above the emission standard (φ) and distinguishes elevated from high/excessive emitters (φ is set to 75% in this case), as tested over a benchmark test cycle. The benchmark test could involve either a legislative dynamometer cycle or a real-world in-use test cycle [131]. An effective in-service emissions program needs an acceptable correlation between the benchmark and in-service emissions test and an optimized combination of cut-points and φ. These features will minimize the number of vehicles falling into the undesirable Groups 1, 4, and 6. In Figure 1, the linear Pearson correlation coefficient is 0.93, which is acceptable for a short in-service test cycle (e.g., [131]). Three colored emission regimes are shown. The vertical line represents a hypothetical in-service test cut-point, which, in this case, results in a 19% failure rate.

Six different emission regimes are shown in Figure 1 (G1–G6) to illustrate how to assess program effectiveness and undesirable outcomes. Vehicles in Groups 1 and 2 have excessive emissions, are in need of repair, and represent a failure rate of 45% of all vehicles tested. In an effective program, the bulk of the tested vehicles fall within G2 as they are correctly identified by the in-service test as excessive emitters. G1 vehicles are “errors of omission”—they have excessive emissions yet fall below the cut-point and are, therefore, not detected for repair by the in-service test. In this example, the identification rate (G2) is only 19% and is missing 58% of the excessive emitters that should be flagged. To improve the outcome, the cut-point could be reduced to, for example, 4 g/km, such that 91% of high-emitters are identified, but the result is a 41% failure rate, which is probably unacceptably high [132,178].

Any scenario requires the examination of the impact of the combined cut-point and φ value on other emission groups. Groups 4 and 6 influence the program’s cost-effectiveness. Ideally, no tested vehicles fall in these groups. Group 6 vehicles are “errors of commission”—these are vehicles incorrectly identified by the in-service test as excessive emitters despite having emissions below the applicable standard. Ideally, no vehicle falls in this group as there is an unnecessary cost to the vehicle owner and test program with no improvement in emissions. In the example, this concerns one vehicle, which means 0.5% of all vehicles tested, and may be acceptable. Samaras et al. [131] recommend G6 to be less than 5%. One vehicle with elevated emissions falls within Group 4 and is targeted for repair. Again, there is an unnecessary cost, but this time with a small improvement in emissions.

When multiple pollutants are considered, a vehicle may be placed in different groups for different pollutant emissions. Samaras and Kitsopanidis [178] suggest that a test vehicle is in need of repair when at least one pollutant falls into Group 2. The effectiveness of the program can be estimated by assuming that any vehicle selected for repairs (Groups 2, 4, and 6) will have the same average emission levels as vehicles in Group 5, which, in this case, is 1.1 g/km. The potential reduction—also referred to as Emission Reduction Rate Potential or ERRP—for this pollutant is 29% and 54%, with a cut-point of 6 g/km and 4 g/km, respectively. In reality, emission reductions are expected to be less than the estimated ERRP for several reasons, as discussed next.

4.6. Assessing the Effectiveness of PTR Programs

PTR programs have been cited as one of the most cost-effective methods of reducing in-use vehicle emissions [180]. In the US, it is common practice in federally required regional air pollution control plans to credit I/M programs for specific air quality improvements [136]. Admittedly, assessing the benefits of I/M programs was hotly debated and a topic of intense research activity in the 2000s. And while it may be intuitively plausible to consider that reducing emissions from the transport sector, which can be the largest source of air and climate pollution—will result in ambient air quality improvements, accurately quantifying the benefits of PTR programs is complex and requires an understanding of emissions, atmospheric transport processes of pollutants, and the various chemical production and loss mechanisms for the air pollution mixture when present in the ambient air shed. But these analyses are standard practice in air quality management and, over the last three decades, many related methodologies and tools have been developed, including sophisticated multiscale air quality modeling systems, source apportionment techniques, and accurate emission inventories to support the claims [192]. For instance, air quality monitoring programs in Switzerland (2008–2018) showed a major reduction in black carbon concentrations in ambient air, coinciding with the deployment of DPF-equipped diesel vehicles into the European market [193].

Initial studies in the 1970s suggested that engine tune-ups and other mechanical adjustments could reduce CO and HC emissions by up to 50% in some cars [194]. Given the efficiency of current engine and emission control systems—when working properly—it is not surprising that fixing a vehicle’s related malfunctions or tampering will still generally result in substantial emission reductions (Section 3.5). For instance, Durbin and Norbeck [51] reported large average reductions in emissions from high-emission vehicles (>1.5 standard) of 87%, 84%, and 60% for NMHC, CO, and NO_x, respectively, after catalyst replacements and fuel injection repairs.

However, the overall effect of PTR on fleet-level emissions is more relevant than impacts on individual vehicles. It is useful to distinguish between estimated (modeled) and measured performance of test and repair programs either using or deriving from empirical data.

Modeled (theoretical) reductions in fleet CO and HC emissions on the order of a few percent to 50% have been reported, with generally lower reductions for NO_x of about 5 to 10% (e.g., [17,116,127,130,167,169,178,180,195,196]). This large spread reflects differences in modeling methods and the many necessary underlying assumptions [196,197,198]. Relatively little information is available for PM, but McCormick et al. [187] reported an estimated improvement of about 40% for truck PM emissions after repairs, with an associated increase of about 20% in NO_x emissions, likely due to advanced injection timing on older non-computer-controlled engines [179]. Due to this PM-NO_x trade-off, it has been suggested that an effective diesel PTR program must include the measurement of both pollutants, clearly an action impacting test simplicity and costs [179]. Li and Crawford-Brown [175] proposed feasible PTR reduction targets for PM of about 20% to 30%. Motor vehicle repair may improve fuel economy and reduce greenhouse gas emissions, but little information is published. Modeled estimates in the US suggest a benefit of 13%, but further analysis of PTR data suggested a smaller benefit of about 3% to 4% [133,169].

Empirical data to support modeled program benefits are limited. Moreover, it is challenging to disaggregate the benefits of more stringent emission standards (including durability requirements) from the impacts of “natural” degradation of emission performance and the effects of PTR programs (e.g., [175]). The empirical methods include analysis of PTR data, random roadside testing, remote sensing, and analysis of roadside ambient concentration measurements. Like modeling results, the analysis of empirical data with different methods has reported variable effectiveness of test and repair programs, providing a vague picture of actual outcomes. But, in general, these analyses suggest smaller emission reduction benefits than modeled results. Possible reasons are further explored in Section 4.7.

• The requirement for vehicles to be inspected periodically by an authorized testing center offers the opportunity for empirical data to be systematically collected for analysis [199]. PTR test records are cost-effectively collected because they are routinely generated and easily accessible. They can, however, be biased by factors such as fraudulent testing behavior, omission of pre-inspection maintenance, and limited representativeness of real-world driving conditions [170]. Harrington et al. [169] analyzed test and repair program data from an enhanced PTR program and found observed emission improvements of 8% (NO_x), 13% (HC), and 13% (CO). Similarly, Singer and Wenzel [200] analyzed test and repair program data and found observed emission improvements of approximately 15% (NO_x), 25% (HC), and 35% (CO). In contrast, Easley [201] analyzed CO and VOC emissions data from a PTR program in the US, but they did not find a statistically significant benefit of the program.
• A benefit of roadside surveys is that they could exclude any influence of fraudulent test behavior. However, they introduce a new bias—self-selection—because the test is voluntary and there is approximately a 10% refusal rate (e.g., perhaps from those trying to hide fraudulent behavior). These surveys also suffer from the limited representativeness of real-world driving conditions [170]. Lawson [137] analyzed data from California’s random roadside surveys and could not detect a difference in tampering and failure rates between PTR and non-PTR locations. It was concluded that PTR programs show little effect on in-use fleet emissions mainly due to vehicle owner and technician behavior. A roadside emission study in 1999 estimated PTR program effectiveness of 13% for CO, 14% for HC, and 6% for NO_x [170]. In California, a pilot roadside program reportedly resulted in a drop of smoky HDVs from about 45 to 19% in two years’ time [168].
• Proponents of remote sensing device studies argue that motorists do not prepare for an RSD test, so fraud is not an issue [171]. Remote sensing measurements are made on the road, reflecting actual driving conditions rather than an artificial PTR test [181]. On the other hand, remote sensing provides only a snapshot (less than a second) measurement of emissions in tailpipe exhaust that may not reflect a vehicle’s typical lifetime pollution potential, but the ability to sample a large number of vehicles in a short time is a strength of RSD [21]. Remote sensing can be used in different ways in the evaluation of PTR programs, and three methods have been distinguished in the US [170,181]: (1) a comprehensive method involving sampling before and after PTR implementation in the same geographical area presumably capturing the same test fleet, (2) a step method of two sub-fleets of tested and non-tested vehicles in the same area, and (3) a reference method involving PTR versus non-PTR fleets traveling in different areas. A remote sensing study by Zhang et al. [195] concluded that the performance of PTR programs in the US was less effective than that predicted by the US EPA. Stedman et al. [202,203] observed a small improvement (6%) for CO, but no observed improvements for HC and NO. Bishop [204] could not find evidence that cities with and without PTR programs had different malfunction and repair rates, suggesting that the programs do not work as expected. Corley et al. [171] reported program benefits varying from 5% to 12%, depending on the city and area considered. It is noted that these studies quantified the emission benefits from an incremental change of the PTR program rather than the full benefits as compared to a non-PTR case [171]. Wenzel [181] reported a small reduction of 3.3% in CO emissions from the PTR program.
• Tiao et al. [205] analyzed roadside concentration data and concluded that the PTR program reduces CO emissions by 8% to 15%. In a later study [206], evidence for the positive impact of PTR programs on CO concentrations was confirmed but not quantified in terms of percent reductions. One of the issues with this approach is the confounding effect of increasingly stringent emission standards.

Figure 2 visualizes the reported impacts of PTR programs using empirical data collected with different techniques. The intention was to conduct a statistical random-effects meta-analysis [207], but this was not possible as relevant studies generally do not provide the required information, such as the mean effect, sample size, and uncertainty (standard error). Therefore, Figure 2 shows the distribution of the reported mean effects from different studies and the grand mean effect and associated asymmetric 95% confidence interval of the grand mean determined with bootstrap resampling.

Figure 2 illustrates the substantial variability in the reported impact of PTR programs. The grand mean effect and associated 95% confidence intervals are −11% (−14% to −8%) for CO, −11% (−15% to −7%) for HC and −4% (−10% to +1%) for NO_x. This analysis suggests that the general impacts of PTR programs on fleet emissions of CO and HC are statistically significant, with a reduction of about 10%, whereas the impact on NO_x is not statistically significant and cannot be distinguished from a zero effect.

4.7. Factors Affecting PTR Program Outcomes

The previous section has shown that there is a large variability in reported effectiveness of PTR programs, which has led to controversy and a dose of healthy skepticism in the US. PTR effectiveness depends on a range of factors that can significantly reduce the effectiveness of in-service emissions programs, and they include [50,169,178,198,209,210,211] the aspects outlined below and discussed in Section 4.7.1, Section 4.7.2, Section 4.7.3, Section 4.7.4, Section 4.7.5 and Section 4.7.6.

• The test configuration:

  ○

  The test procedures and associated pass/fail criteria (e.g., cut-off points);

  ○

  The execution of the program (calibration and maintenance of equipment).

• The fraction of the on-road fleet included in the program:

  ○

  Finding and including high emitters;

  ○

  Exemptions;

  ○

  The compliance rate (the percentage of eligible vehicles that have complied with the requirement to pass the periodic inspection).

• The effectiveness of repairs and fraud:

  ○

  The quality and durability of repairs;

  ○

  Fraud by vehicle owners of test facilities.

4.7.1. PTR Test Configuration

The test procedures used in a PTR program impact program effectiveness. Samaras and Kitsopanidis [178] found that a steady-state test identified about 15% of the high-emitters, whereas transient tests identified about 70% of high-emitters. A subsequent study [131] confirmed that a short transient load test performs better for TWC-equipped petrol and diesel vehicles while also showing the low effectiveness of the FAS test for diesel vehicles. Choo et al. [132] also found that an idle test is less likely to result in a failed vehicle or identification of a gross polluter than an ASM test but noted the limited emission measurements (CO and HC only) as a potential reason for underperformance. Other studies did not find a significant difference between emission tests. Harrington and McDonnell [197] used a stochastic model to evaluate program performance and only found a small difference between the IM240 and a high-idle test. An analysis of program data in China did not find a difference in effectiveness between steady-state and transient tests [145]. In addition, failure rates may be sensitive to other test aspects, such as vehicle preconditioning and calibration procedures [168]. The FAS test has also been shown to have poor correlation with results obtained over the NEDC and a strong influence on pre-conditioning [131].

4.7.2. Fraction of On-Road Fleet in the PTR Program

PTR program effectiveness is also a function of the portion of the on-road fleet that is tested, which depends on the types of vehicles targeted; the frequency of the PTR tests; and any exemptions, test avoidance, or compliance rate. PTR programs have traditionally focused on light-duty vehicles (cars and light commercial vehicles), but these programs also exist for two-wheelers [176] and heavy-duty vehicles. The different types of on-road vehicles are usually handled using similar test procedures [33,50]. In the US, in-service testing for diesel trucks has received less attention than in the rest of the world. A reduction in ozone precursor emissions has been the main driver for PTR programs in the US. The US EPA has not mandated or regulated HDV PTR programs yet [168]. Only a handful of states have a diesel PTR program. California is now implementing the first of such programs for HDVs [153], and other states are expected to follow. In contrast, the EU requires PTR testing for all vehicles regardless of fuel and weight, so it includes diesel-powered cars and trucks [166,168]. The EU emissions tests are also integrated into a comprehensive roadworthiness inspection approach for on-road vehicles, which includes safety [168]. The wide coverage of the on-road fleet in the EU is expected to result in better PTR outcomes than in the US. The frequency of in-service tests can vary between six months and two years [50], but tests are generally conducted every one or two years.

Vehicles may be exempted from testing, for instance, under a first four-year new vehicle exemption [132]. Unlike in the EU, a waiver may be given if a vehicle owner in the US has spent more than a specified amount on emissions-related repairs, even if the vehicle is still exceeding the allowable limits [169,191]. Although exemptions for testing have been considered to undermine the effectiveness of PTR programs [17,178], it has been suggested that PTR cost-effectiveness would be improved significantly by, for example, increasing the model years of vehicles subject to testing [174]. Importantly, these exemptions rely on the recognition that modern vehicles generally have lower emissions than older ones and that today’s high-emitters are cleaner than yesterday’s.

Test avoidance is real, such that not all vehicles will undergo a required test. ERG [198] noted that the US EPA assumes a compliance rate of 96% and a waiver rate of 3%. However, analysis of PTR data suggests these rates are lower, leading to an overestimation of program benefits. Program non-compliance, either by avoiding the test altogether or by never returning after initially failing a PTR test, lowers the emissions benefits of the program. Studies have indicated that about 10–25% of all vehicles that failed an emissions test never received a passing mark. Many of these vehicles were observed to be still operating on the road. Another 5–10% of vehicles on the road have been found to be eligible but have never participated in testing [127].

4.7.3. Effectiveness of Repairs and Fraud

Modeling often assumes that repairs will be completely effective and durable, whereas in reality, this may not be the case [196,197,201]. Unscrupulous vehicle owners and mechanics may resort to a “clean for a day” situation in which a vehicle would fail the test but is passed due to a temporary fix or fraud [171]. Lawson [130] reported that random roadside surveys showed that California’s PTR, the Smog Check program, had done little to reduce tampering and emissions from the in-use fleet. This lack of success appeared to be caused mostly by motorist or technician behavior. Because the costs of properly maintaining a car can be high, motorists appeared to have found ways to “pass the test” and therefore avoid the costs of repairs [130]. In addition, owners might have purchased cheap parts that provide only a temporary fix [212]. Repairs performed in these programs might not be robust and long-lasting. A desire to pass the test at the minimal possible cost affects the type of repairs motorists obtain [127].

In addition, PTR programs may not always have a beneficial effect in all cases, as shown in Figure 2. Previous studies have shown that a substantial portion of repaired vehicles can have higher emissions for specific air pollutants after repairs (e.g., [127]). This reflects well-known trade-offs between CO and HC versus NO_x for petrol vehicles and PM versus NO_x for diesel vehicles [132,187,197].

Some studies suggest that highly variable emissions, and not test fraud, were the cause of a large number of vehicles failing a PTR test soon after completing a previous PTR test. Analysis suggests that as much as 74% of the vehicles that failed their initial PTR test passed a retest because of emissions variability, and not because any repairs were performed. Vehicles with highly variable emissions may not only fail and pass a subsequent retest without repair but are also likely to fail again if tested shortly thereafter [117].

PTR programs are either centralized or decentralized [50,168]. In centralized programs, all inspections are performed in dedicated high-volume test facilities. In a decentralized program, both emissions testing and repairs can be conducted in private garages. There is a broad consensus that decentralized programs are generally less effective because of possible fraud and improper inspections [50]. Studies in the US have found test fraud exceeding 50% in decentralized programs [117]. A small percentage of repair shops in the US even specialized in providing fraudulent certificates or stop-gap repairs to pass the test [196]. Undercover investigations suggest that less than 25% of vehicles inspected in decentralized programs receive thorough and accurate inspections [17]. The underlying issue could relate to perverse incentives, such as failing clean vehicles to charge for unnecessary repairs or passing dirty vehicles to assure repeat business and remain competitive [169,196]. Eventually, separating the test and repair aspects of the program was implemented to address those conflicts. The concept of a test-only facility is expected to take away the potential incentive for corruption in repairs.

But not all studies agree. A remote sensing study by Zhang et al. [195] found no evidence that centralized programs are more effective than decentralized programs. Moreover, any program can suffer from fraudulent behavior. For instance, Stedman et al. [203] reported that vehicles were found to avoid repairs after failing the test through sales to individuals who register them outside the program area but still drive the vehicles in the area. Corley et al. [171] similarly referred to the halo effect, where vehicles that fail a test in areas with a long-standing PTR program are sold off to non-PTR areas, causing a potential build-up of failing vehicles in these areas. Vehicle owners can also pass an initial test but avoid subsequent tests, never obtain a passing test after failing an initial test, or simply avoid testing altogether [181]. Ando et al. [133] reported that about 22% of vehicles fail their initial test and do not show a passing test record (e.g., sold outside the region, scrapped, or still in the process of passing the test).

4.7.4. Selective Sampling

Selective sampling can be used to increase the likelihood of high emitters appearing in the PTR program [213]. Indeed, the Smog Check Program in California uses a high-emitter profile to identify the fraction of the on-road fleet (i.e., model years of interest) that is likely to fail the emissions test for inspection at test-only facilities. This is achieved in addition to the required biennial inspection for all vehicles and testing when there is a change of ownership [132]. These types of approaches are based on PTR data of emission test results and relevant vehicle characteristics, such as vehicle age, engine size, and odometer reading [129,132].

4.7.5. Diesel Vehicles in a PTR Program

As discussed in Section 4.6, PM and NO_x emissions are arguably the most important pollutants from the on-road transport sector (e.g., [192]), but historically, they have not been included in in-service test programs. Traditionally, smoke (i.e., soot) opacity has been the in-use measurement of choice for diesel vehicles as a proxy for PM emissions. Opacity is an inexpensive and practical measurement, in contrast to PM [50,177]. However, opacity measurements only work well for vintage diesel engines without post-combustion emission controls, where PM emissions are primarily black carbon. In contrast, there is no simple way to determine NO_x emissions, as they require the use of a “loaded” test on a chassis dynamometer, which increases the complexity and costs of the program [50]. Now that both PM (e.g., DPF) and NO_x (e.g., SCR) emission controls are common in all modern diesel vehicles, new measurement approaches for assessing in-use performance are needed. Studies searching for alternatives are already underway in some jurisdictions in Europe, the US, and a few other places. But the identification of a suitable PM measurement approach for an in-service test program for modern diesel vehicles has been particularly challenging. The size of the emitted particles (whose composition is mostly organic carbon, sulfate, and metals) from approximately all MY 2000 and newer vehicles in the EU and US falls below 400 nm in diameter. Therefore, these particles are largely invisible to an opacimeter, resulting in a lack of correlation between the current smoke test and the measurement of PM or PN emissions. For these reasons, an opacity measurement is inadequate for evaluating the proper operation of DPFs or catalysts [50,92,131,168]. In fact, current Euro 5 and 6 limit values for diesel smoke can be met without a DPF [92,155]. Consequently, DPF malfunctions (e.g., a cracked substrate) cannot be identified using smoke detection [50]. PEMS technology, including particle counters, is promising. It has been amply demonstrated and can be readily deployed to conduct in-use emission measurements. Other measurement technologies are emerging in the form of portable, convenient, smaller, and less costly on-board sensing and reporting options (e.g., four gas testing, including particle counters). These technologies will help authorities improve PTR program implementation [214].

Increased attention to diesel HDVs is warranted, as they generally represent less than 5 percent of the global on-road vehicle population but account for 40% to 70% of on-road NO_x and PM emissions [50].

Nearly all HDV programs around the world use and are limited to exhaust smoke opacity test to identify the need for repair because of simplicity and low costs [168,179]. Due to a general lack of guidance on diesel emission testing and allowed flexibility, PTR programs in the US have developed a wide range of emissions tests (variations of idle, snap-idle, and loaded tests), although most of the programs use the snap-idle or FAS test [179]. OEIT [166] reported that there were 16 HDV diesel PTR programs in operation in the US and 2 in Canada, all of which measured smoke opacity only as a surrogate for diesel PM emissions, although program structures and test methods varied across states. Lower opacity cut points have been implemented in California via regulation, as these have been shown to be effective at identifying drastic system failures like cracked DPFs [153]. Loaded testing of diesel trucks is still relatively rare, with Australia and Chile being a prominent exception, along with a few US states [168]. Despite many years of advocacy by environmental advocates, EU countries have still not adopted load testing for diesel trucks [166]. Some possible improvements for diesel PTR programs have been proposed, including the following:

• The use of laser light scattering photometry (LLSP) in conjunction with effective vehicle preconditioning procedures [50,168].
• The use of remote sensing and RSD smoke factor readings above 0.15 (i.e., 1.5 g of soot/kg fuel) to flag a malfunctioning or tampered DPF with high confidence and target for repairs [21].
• A new PN low-idle test that has shown a good correlation with real-world PM emissions and may be a good option to detect vehicles with removed or malfunctioning DPFs [43,54,92,186].

Measurement of both PM and NO_x is critical for an effective diesel PTR program [179]. However, a sufficiently accurate and simple NO_x test is still missing. The detection of SCR emulators in diesel HDVs is not yet possible, whereas the use of these emulators in the light-duty fleet may not be as big a concern.

4.7.6. Catching Tampering Behavior in PTR

As discussed in the previous section, modern diesel vehicles with advanced emission control systems, such as SCR and DPF systems, have been increasingly difficult to evaluate with emission tests used in legacy test and repair programs. Alternative approaches are needed [50]. The absence of in-service inspection programs or the use of ineffective high-emitter detection methods may encourage diesel vehicle owners to tamper with emission control (e.g., [43]) or to use low-quality fuels [215,216], in both cases to save money.

Since actual software is modified during the removal of a DPF when using a “delete kit”, the current PTR cannot determine whether a DPF has been removed. A smoke opacity test is also similarly constrained. Checking for the absence of a DPF during the PTR is difficult because no parts may be removed during the inspection. At best, an inspector may inspect the DPF housing for signs of repair welds or knock on the housing to assess if it is empty. If the housing seems filled, however, there is no way to determine with certainty the presence of a working DPF or a dummy [155]. This issue has been addressed with the implementation of the new low-idle PN screening test for DPF failures and removals [43,186]. Specifically, researchers in Switzerland and later the Particle Measurement Programme (PMP), supported by the European Commission, have made significant advances using particle counters and a PN test in a measurement methodology that is now integrated into a PTR program [217].

Supplementary programs to PTR will be important to effectively address tampering. New practical methods are being developed and tested, such as simplified PEMS/monitors [19,218,219,220] emissions plume-based methods [151,152,221,222,223], remote sensing [21,125,126,128,149,218,224], on-road emissions monitoring [32,225], remote access to on-board (emissions related) sensors [226], and improved in-service inspection methods [43,54,92].

Given the growing pollution burden of cheating and tampering, a combination of low-cost on-road emissions screening and targeted vehicle repair programs could be particularly cost-effective [17]. For instance, a combination of remote sensing or roadside inspections and a call-back procedure to assess the worst emitters by means of laboratory or PEMS testing has been suggested as a good way forward [35]. However, the downside of these approaches also needs to be considered. Additional study of in-use performance using various PEMS could also aid in determining the next screening methods.

It has been suggested that remote sensing and similar techniques, such as on-road emission measurement, can be used as the backbone for a test and repair program for clean screening (i.e., a test waiver for a certain vehicle model year) and high-emitter identification, where high emitters are directed for further testing, including OBD checks and dynamometer tests [50]. RSD techniques have been considered to identify high-emitters, for instance, as a means of tracing the results of the in-service testing programs [16], or even to replace these programs [227]. RSD is useful to identify “gross polluters” and to rank vehicles in order of increasing emissions so vehicles belonging to the most polluting subset can be targeted for further inspection and maintenance [166]. Some US states had hoped that RSD could supplement or replace routine emissions inspections. However, there were several issues, including identifying only a small portion of high-emission vehicles in “dirty screening” (e.g., a high detection threshold, limited number of locations, and cost-effectiveness). Similarly, there were issues with identifying clean vehicles or “clean screening” to exempt low emitters from routine inspections (waiver). The inability to check OBD data to reduce evaporative emissions and to intercept potential problems that could result, later, in higher emissions proved detrimental [180]. There is also the need for multiple readings to ensure that including or excluding a vehicle for subsequent in-service testing is based on acceptably robust data [127]. It is noted that innovative approaches with RSD can have value for high-emitter detection. For instance, Bishop et al. [225] successfully demonstrated a method that combined RSD with plume capture using a tent for fast emission testing of heavy-duty diesel trucks serving in US ports. Moreover, remote emission sensing has undergone continuous improvements, and the latest research [126] indicates that remote sensing shows good agreement with other methods, such as PEMS, suggesting that remote sensing should be an important component of programs aimed at catching excessive polluters.

RSIs can be conducted randomly by authorities to focus on illegal activities such as DPF removal and SCR tampering, tachograph tampering, weight exceedance, road worthiness, and others. Roadside inspections can be an efficient means of identifying high-emission vehicles and forcing them to be repaired. In Chongqing, China, inspection stations failed about 10 percent of vehicles brought in by drivers. In contrast, roadside inspectors failed about 40 percent of vehicles flagged [166].

Work continues to explore alternative ways for vehicle inspection. Remote fleet monitoring with on-board sensors (Simplified Emissions Measurement System, or SEMS) and remote sensing as essential components of market surveillance and assessment of on-road emissions performance have been assessed in EU research projects such as CARES and NEMO [100,126]. A more convenient and low-cost wireless vehicle inspection system has been proposed in which engine parameter data are automatically collected and wirelessly transferred to the road authority. Owners are then notified if repairs are required [212]. EU market surveillance has been conducted by inspection authorities from 2020 onwards. OSAR, on-board sensing and reporting, is also another concept being explored by the University of California, Riverside researchers. This goes beyond PEMS and has potential for leading to a better and more straightforward approach for assessing and monitoring in-use vehicle performance [214]. Thus, we may not have seen the last of I/M. And the lessons of the past provide valuable information for the future of in-use real-world vehicle performance assessments by industry, the research community, and regulators.

4.8. Outlook and PTR Improvements

To minimize real-world emissions of ICE vehicles over their entire life cycle, a set of measures for specific issues is needed. An overview of these measures is provided in

Table 2. Overview of key aspects, stakeholders, goals, impacts, and potential measures for controlling real-world emissions from fossil-fuelled ICE vehicles during the vehicle lifecycle.

Aspect	Stakeholder	Goal or Interest of Stakeholder	Impacts	Potential Measure
1. Type approval	National Government	Sustainably low emissions on the road	Poor legislation leads to higher emissions in the real world	Improved type approval tests; investments in real world emission research; structural enforcement/durability test programs
2. Defeat devices, cheating	Vehicle manufacturer	Lower product price/higher profits; improved fuel economy; less engine failure	Significantly increased emissions in the real world for a range of pollutants with associated public health, environmental, and economic impacts	Improved type approval tests; on-road surveillance emission tests; electrification of the on-road fleet; information campaigns for vehicle owners (including registration of known defeat devices by make/model)
3. Defeat devices, tampering	Vehicle owner	Lower operational costs; improvement in driveability and engine power	Significantly increased emissions in the real-world for a range of pollutants with associated public health, environmental, and economic impacts	Roadside inspections, including remote emission sensing and plume chasing; modernized, improved, and fit-for-purpose PTR-programs; electrification of the on-road fleet; information campaigns for vehicle owners
4. PTR programs	National, regional, and local government	Maintaining acceptable emission levels for the on-road fleet, preventing high emitters on the road to maximum extent possible	Reduction in real-world emissions with associated benefits but funding required	Legal basis for PTR-programs; research on efficiency of PTR programs; modernization and improvement of traditional PTR programs; dedicated and secured funding; information campaigns for vehicle owners
5. Lack of vehicle maintenance	Vehicle owner and service/repair shops	Reduced operational costs on short term, but potentially increased costs in the longer term	Reduced driveability (e.g., less engine power); more technical failures; increased repairs; elevated air pollutant and greenhouse gas emissions	Improved OBD; power restriction/release via software; electrification of the on-road fleet; information campaigns for vehicle owners

. A general increase in vehicle durability and associated vehicle manufacturer warranties have accompanied improvements in vehicle emission performance [127,169] and perhaps, at first sight, reduced the relevance of in-service programs. For instance, OEIT [166] stated that the benefits of PTR programs were decreasing in the US due to improvements in the performance and longevity of emission control technologies. Interestingly, it is the very test and repair programs that may have played a significant role in forcing auto manufacturers to improve durability [169]. In addition, more stringent and technology-forcing regulations in key markets like the US have also pushed car makers to bring better performing products to market. It is noted that the European Commission has just proposed a comprehensive package to address the issues discussed in this paper [228].

On the other hand, the relative impact of a small proportion of high-emitters in the on-road fleet on total emissions has also increased. In addition, the relevance of tampering and malfunctioning emission controls has increased due to the effectiveness of modern systems when working appropriately. Thus, it could be argued that it is more important and cost-effective than ever to find and repair high-emission vehicles now. Monitoring of emissions from on-road vehicles is therefore essential. Less effective but practical are RSD test programs with additional measures for high-emission vehicles. More effective for total fleets are PTR programs with improved emission test procedures and checks and balances to ensure compliance. There are several ways to measure vehicle emissions, but commonly used laboratory methods and the use of PEMS are simply too complicated, expensive, and time-consuming to be used for in-service emission monitoring of on-road fleets. It is noted that PEMS technology is improving and becoming more accessible due to lower costs.

Figure 3 visualizes the relationship of expected fleet emission reductions for different test methods. Several methods are available. But an important policy consideration is the relationship between the expected decrease in the average emissions of a vehicle fleet and the costs and efforts of the measures to achieve those reductions. Broadly, three approaches exist: (1) periodic inspections under a PTR program, (2) roadside inspection (RSI), or (3) remote emission measurement (RES). In practice, the following options can be considered:

• PTR OBD: periodic technical inspections restricted to the assessment of OBD codes.
• PTR E-test: periodic technical inspections with simple and fast emissions tests.
• RSI E-test: roadside inspections with simple and fast emissions tests.
• PTR IE-test: periodic technical inspections with improved emissions tests at additional costs and complexity.
• PTR CD-test: periodic technical inspections with a simple chassis dynamometer test.
• RS Recall: remote sensing or plume-based high-emitter detection and recall for further inspection.

The choice is a political and technical decision. Ultimately, the effectiveness of any of these options depends, to a large extent, on an implementation strategy with clear requirements for compliance and enforcement.

5. Conclusions

Electrification of the road transport sector is the most appropriate and cost-effective approach to reduce greenhouse gas (GHG) emissions, improve energy efficiency and air quality, and promote energy security. However, the penetration of EVs into the global on-road fleet will likely be slow. The relatively long useful life of ICE vehicles (typically 15 to 20 years for a first buyer) suggests that they will probably still be present to some extent in many markets (particularly emerging economies) in 2050. Thus, good performance and effective emission control of conventional and aging ICE vehicles will continue to be of paramount importance for the foreseeable future.

In this study, we provided a broad outlook by (i) offering a comprehensive review of the published literature, (ii) take historical stock of the issues, (iii) articulate the current state of play in the world of vehicle emission control, and (iv) formulate questions for potential future research. We have identified and discussed two main problems and their potential solutions.

5.1. Problem 1—Cheating (New Vehicles) and Tampering (In-Service Vehicles)

With progressively more stringent vehicle emission and fuel quality standards adopted in all major markets, engine design and emission control systems have become increasingly sophisticated, often combining different technology and fuel approaches. Today, these controls are complex, computer-controlled systems generally optimized to balance performance and efficient air and climate pollution abatement. Modern vehicles today are highly optimized machines, achieving a high efficiency of 90% or more for air pollutant removal. But when these systems malfunction due to natural deterioration, intentional cheating, or tampering, it can have a large negative effect on vehicle performance and emissions.

Nearly all aspects of vehicle and engine operation are controlled with precision by ECUs (electronic control units), which allows manufacturers to drive significant improvements in performance, reliability, and fuel economy. But, at the same time, the complexity and sophistication create opportunities for cheating and tampering. The type, quality, and durability of an emission control system employed by a vehicle manufacturer depend, to a large extent, on the make and model and the applicable regulations.

In this study, we distinguished between new vehicles and in-service vehicles. For new vehicles, we considered (1) the impacts of legally approved modifications granted to a vehicle manufacturer that temporarily increase emissions and (2) a defeat device or intentional cheating by a vehicle manufacturer. For in-use vehicles, we specifically considered intentional and unlawful tampering by vehicle owners. Cheating by vehicle manufacturers and tampering by vehicle owners refer to intentional and unlawful modifications of a vehicle’s emission control system, resulting in environmentally adverse effects from sub-optimal emission control performance. Cheating involves a defeat device.

The review suggests that there are important differences between the main types of ICE technology and associated fuels—SI engines (petrol, CNG, or LPG) and CI engines (diesel). Experience with SI engines in the last three decades suggests that sophisticated control systems have generally led to a major reduction in emissions. Previous studies have confirmed that real-world emission control for SI engines is generally effective and, importantly, robust and may also be less affected by cheating or tampering. However, gasoline direct injection technology is a recent and rapidly growing entrant into the light-duty vehicle market. Thus, PN emissions and the need for properly functioning GPFs are of growing interest and concern.

Emission controls for CI engines have become increasingly sophisticated and require the use of relatively new technologies. Previous studies suggest that real-world emission control for diesel CI engines is more expensive, less effective, less robust, and more susceptible to cheating or tampering. Defeat devices seem to primarily be a concern with diesel vehicles (both light and heavy), where emission control deactivation in real-world driving can lower manufacturing costs (i.e., inferior and cheap systems and engine calibration), improve fuel economy, reduce engine noise, improve vehicle performance, and extend refill intervals for diesel exhaust fluid, if present.

This review examined the emission impacts of cheating and tampering with a focus on modern and relevant emission control systems, namely, EGR, SCR, and DPFs. It is clear that the removal or deactivation of these systems will lead to large increases in emissions of specific air pollutants—for instance, 70% (EGR), a factor of 3 or more (SCR), and a factor of 25–100 (DPF). Although we have a reasonable idea of what intentional interference with the emission control system does to the emissions from individual vehicles, assessing fleet-wide impacts requires reliable information on how widespread the practice is. Our analysis shows a large geographic variability and uncertainty regarding the reported occurrence of tampering by vehicle owners in particular. In comparison, the impacts of cheating by manufacturers on fleet emissions may be easier to quantify since the number of vehicles affected is generally well known. In any case, the available evidence suggests that fleet-wide impacts of cheating and tampering on emissions are undeniable, substantial, and cannot be ignored.

Cheating with emission control systems by vehicle manufacturers is not a new phenomenon (the first cases were reported in the 1970s), is still happening (the latest case was reported in 2024), and does not seem to be restricted to specific manufacturers or vehicle types.

Cheating refers to corporate malfeasance and involves the vehicle or engine manufacturer because modifications are complex, often made deep in the vehicle’s inner workings. Cheating involves a defeat device, which is generally easy to design and difficult to detect. Emission control adds cost to vehicle production and generally reduces overall fuel efficiency and increases CO₂ emissions. In addition, technological limitations can put compliance with emission standards in conflict with other factors, such as performance, durability, and efficiency, creating an incentive for vehicle manufacturers to evade regulatory requirements, sometimes by exploiting the peculiarities and flexibilities of official tests.

The notorious Volkswagen “dieselgate” scandal (including its subsidiaries Audi and Porsche) in 2015 received a lot of attention around the world and raised awareness about cheating. But many vehicle manufacturers seem to have been deploying increasingly sophisticated and illegal strategies to disable or reduce the effectiveness of emission controls on the road. In fact, almost as soon as vehicle emission testing began in the 1970s, a wide range of vehicle manufacturers have actively found ways to cheat, leading to fines ranging from millions to over a billion dollars.

Suspicious observed real-world emissions behavior suggests that the use of defeat devices by vehicle manufacturers may currently be widespread. Differences in vehicle emissions between laboratory tests and real-world on-road operation (i.e., “the gap”) have been understood for a long time. The relevant question is what portion of the gap can be attributed to cheating or tampering versus justifiable differences, including natural variability in on-road driving and test procedures. For CO₂ emissions from current production new vehicles, the impact of cheating appears to be, on average, less than 20%, but substantially larger for conventional air pollutants. In some cases, vehicle emissions can increase by up to two orders of magnitude.

Although a review of the available evidence suggests large economic (up to a billion dollars or more) and health costs (several thousands of disability-adjusted life years), it also suggests a substantial level of geographical variability. Further research is, therefore, required to enable an accurate assessment of the overall impacts of cheating and tampering on global emissions, air quality, and public health.

This review documents a long history of enforcement action against cheating in the US, but, importantly, not in other parts of the world. Enforcement action in the US is supported by specific US EPA regulations and guidance, which provide clarity and reduce the possibility of manufacturers exploiting loopholes. The differences between the EU and US may be disappearing now as recent EC guidance documents and rulings by the Court of Justice of the European Union provide a strict and narrow definition of what is considered an allowable use of an AES.

The assessment of emissions strategies and the search for defeat devices remains a complex, resource-intensive, and time-consuming endeavor. Typically, a defeat device detects if a vehicle is running on a dynamometer for an official emissions test, and if so, it then switches to a special low-emission mode. In the real world, engine performance, fuel economy, and drivability are then optimized by the defeat device at the expense of excess tailpipe emissions. They are typically embedded in proprietary computer code that is difficult to access and interpret by a third party. A clearly established and agreed set of (international) criteria is not yet available, and a possible test program for defeat device detection will require a clear definition of the test protocol. It is clear that a combination of repeated and iterative laboratory dynamometer and on-road PEMS measurements is the “gold standard” for the detection of defeat devices, and—if possible—further enhanced with a detailed analysis of the vehicle software code.

Despite the financial penalties, undesired global attention automakers get when caught cheating, associated damage to brand reputation, temporary drop in vehicle sales and stock value, and forced recalls, cheating may continue. At least some vehicle manufacturers can appear to treat the use of illegal defeat devices in new vehicles as a manageable risk.

Vehicle owners resort to tampering to (1) improve performance and/or fuel economy, (2) avoid costs of running the emission control system, (3) avoid costs of necessary repairs to the emission control system, (4) sell the vehicle for a higher price (i.e., odometer tampering), or (5) simply to rebel against norms.

Again, this is not a new phenomenon; it has happened globally since the 1970s, and it applies to all vehicle types. Although there appears to be significant geographical variability where tampering is occurring and a larger range of estimates of the number of malfunctioning systems (from undetectable, a few percent, to almost 50%), the reported proportions of on-road vehicles that are tampered generally relate to the fraction of diesel vehicles in the on-road fleet, including both light-duty and heavy-duty vehicles. Tampering and the use of defeat devices in the freight sector also means undercutting responsible and law-abiding operators, giving those who tamper an unfair economic advantage while causing harm to the environment. There are complex legal considerations when it comes to tampering by vehicle owners. It is not necessarily illegal for businesses to sell tampering devices, but it is illegal to install and use them.

5.2. Problem 2—The Effectiveness of Current In-Service Emission Programs

Without effective in-service emission programs, compliance with new vehicle emission standards is significantly weakened. In-service emission programs are designed to identify and fix large faults, rather than a gradual deterioration in the control of emissions. These programs are needed to ensure that the benefits of new-vehicle control technologies are not lost through malfunctioning emission control systems, engine issues, incorrect repairs, a lack of servicing and maintenance, poorly retrofitted fuel and emission control systems, the use of improper or low-quality fuels, or tampering. Moreover, these programs may have played a significant role in creating incentives for the manufacturers to improve durability.

In-use vehicle emission performance deteriorates naturally over time; hence, exhaust emissions increase with age. Beyond that, vehicles can become excessive emitters due to, for example, malfunctioning emission control systems or engines and the other reasons mentioned before. Consequently, a large portion of total fleet emissions originates from a disproportionately small number of high-emission vehicles, and this is increasingly the case. It is this small and high-emissions portion of the on-road fleet that has been assumed to offer the greatest potential benefit through a repair or removal program. There is not a universally accepted or exact definition of a high emitter, but they are typically defined as vehicles exceeding either a predefined multiple of the applicable emission standard (i.e., 1.5 to 5.0 times) or the relative position on the emission rate distribution of a related on-road fleet.

In-service emission programs impact a range of stakeholders. An effective program involves a simple test that is technically sound, socially acceptable, and not too onerous in terms of test and repair costs. In-service emission tests are not meant to meet the rigor of a regulatory compliance test. They target the vehicle owner, are based on a brief and simplified operation of the test vehicle, involve the measurement of fewer pollutants (typically CO, HC, and diesel “smoke” or PN), and involve the use of affordable equipment that is less precise. Several short in-service exhaust emissions tests have been developed in the US and Europe. These include (1) the no-load test, (2) the steady-state loaded test, and (3) the transient loaded test. A vehicle fails a test when its emissions are determined to be typically above two to seven times the applicable standard. These thresholds are pollutant-specific.

On-board diagnostics testing is now an integral part of test and repair programs around the world because it offers several potential benefits, such as comprehensive fault detection, preventative maintenance, the use of more durable components by vehicle manufacturers, and reduced costs for vehicle owners by addressing (minor) issues early on and within the warranty period and targeted and faster repairs. While some in-service deterioration, such as excessive smoke or higher fuel consumption, can be identified relatively easily, a technician uses an OBD scanning tool to retrieve fault codes to detect any malfunctions for potential repair. However, OBD alone is not sufficient. Although controversial, a potential improvement is adding telemetry to the OBD system for reporting an emissions problem directly to authorities either continuously or on a schedule, similar to on-board fuel monitoring systems (OBFCM) now in place in the EU.

Periodic Test and Repair programs are the most common form of in-service programs used around the world. Accurately quantifying the benefits of past PTR programs is complex, and the overall effect of PTR on fleet-level emissions is more relevant than impacts on individual vehicles. This study collected and analyzed modeled and empirical data on PTR program effectiveness, showing substantial variability in the reported impact of PTR programs. The grand mean effect and associated 95% confidence intervals are −11% (−14% to −8%) for CO, −11% (−15% to −7%) for HC, and −4% (−10% to +1%) for NO_x. This analysis suggests that the general impacts of PTR programs on fleet emissions of CO and HC are statistically significant, with a reduction of about 10%, whereas the impact on NO_x is not statistically significant and cannot be distinguished from a zero effect.

However, there is now a need to modernize PTR programs since the effectiveness depends on a range of design factors that can significantly reduce the effectiveness of in-service emissions programs. They have the following fundamental elements: (1) comprehensive fleet coverage, (2) a suitable test procedure, (3) vehicle compliance and effective enforcement, (4) proper technician training, (5) quality control and quality assurance, (6) periodic program evaluation, and (7) minimization of waivers and exemptions. PM and NO_x emissions are now arguably the most important pollutants in the on-road transport sector, but historically, they have not been included in in-service test programs. Now that both PM and NO_x emission controls are common in all modern new diesel vehicles, and commonly the focus of cheating and tampering practices, new measurement approaches for assessing in-use emissions performance are needed. As an example, the new PTI-PN test for DPF-equipped vehicles is a newcomer in this field.

5.3. The Solutions—Identification and Repair

Potential solutions are discussed separately for new vehicles and in-service vehicles.

For new vehicles, the following apply:

• The vehicle emission certification and approval processes could be modified (1) to reflect as closely as possible the variability in real-world conditions and (2) by introducing random aspects in the test protocol wherever possible to close loopholes that may currently be exploited by manufacturers. Moreover, the situations where AECDs or AESs may be temporarily deployed should be clearly defined, limited, and narrowly applied to specific operating conditions for engine protection. These exemptions should not lead to extended high emissions in real-world driving conditions.
• Sufficient funding and scope for independent in-use compliance and enforcement programs (in-use conformity, conformity of production, and surveillance testing) will assist regulators in the early detection of excess emissions caused by defeat devices and manufacturer defects.
• The act of cheating should become more unattractive for manufacturers. This can be achieved through an active, strong and consistent enforcement regime, which could impose large penalties and force vehicle recalls. Related and critical is the need for robust legal frameworks and governments sufficiently supportive of the agencies entrusted with the enforcement of the standards.
• An international and agreed-upon test protocol specifically for the detection of defeat devices would put further pressure on manufacturers to cease cheating. This should include different options, such as combined on-road (PEMS) and laboratory emissions testing, as well as the analysis of vehicle software code.
• The implementation of national or state-level on-road emission test programs in which ICE vehicles are tested when they reach specific accumulated mileages (e.g., 50,000, 100,000, 150,000, and 200,000 km) and releasing these data in the public domain will create more clarity and transparency.

For in-service vehicles:

• Compulsory PTR programs play a key role. They are arguably the first line of defence in the detection and repair of high emitters but require careful consideration of a range of program design parameters that will determine the program’s overall effectiveness. For instance, selective sampling based on relevant vehicle characteristics can increase the likelihood of including and fixing high emitters.
• Include PM and NO_x emission tests in the PTR program, as they are currently either ineffective or missing. There are several options for excessive PM emissions, including the use of laser light scattering photometry (LLSP) in conjunction with effective vehicle preconditioning procedures (instead of the currently used ineffective smoke opacity) and a new PN low-idle test that has shown a good correlation with real-world PM emissions. A sufficiently accurate and simple NO_x test is still lacking. For NO_x, a combination of remote sensing or roadside inspections and a call-back procedure to assess the worst emitters by means of laboratory or PEMS testing could be a good way forward.
• OBD will also play a critical role. The first examination of OBD data forms an essential part of an effective PTR program, and specific checks (e.g., changes to readiness codes) should be explicitly included in the PTR program. Second, adding reasonably accurate but low-cost emission sensors to the OBD system will improve OBD’s effectiveness in detecting emission issues. Third, remote OBD, where telemetry is added to OBD systems, will enable fast, wireless, and direct reporting of emission problems to regulatory agencies (detection) and trigger further action (confirmation and repair).
• A combination of (1) low-cost on-road emissions’ clean or dirty screening, such as remote sensing (fixed location and plume chasing) or on-road emissions monitoring, and (2) targeted vehicle repair programs could be a particularly cost-effective way forward to address tampering, malfunctioning, and cheating.
• Randomized roadside inspections can focus on illegal activities, such as DPF removal and SCR tampering, tachograph tampering, weight exceedance, roadworthiness, and others. This will strengthen the identification and repair of high-emission vehicles.
• Address current legal barriers by making changes or introducing new regulations that penalize the sale to and/or use of defeat devices by vehicle owners. An additional solution could be mandatory registration of owners who purchase a defeat device. These changes should include robust enforcement action, in particular outside the US.
• Ensure that vehicle manufacturers produce tampering-resistant vehicles.

6. Outlook and Recommendations

This review has shown that as long as ICEVs provide motive power for vehicles, controlling fleetwide emissions and maintaining a minimum level of environmental performance will require attention at various stages of a vehicle’s service life. Both dynamometer and on-road emissions tests will continue to be an important element of the type-approval or emission certification process. But emission performance in real-world operation must also be tracked. To combat potential fraud by vehicle manufacturers or vehicle owners and to detect and repair malfunctions that degrade emission controls, adequate in-service inspection programs are necessary. While generally increasing vehicle durability and warranties at all market segments are positive developments, the presence of a relatively small fraction of high-emitters and cheating and tampering will contribute a higher and disproportionate amount of excess pollution that must be tackled by environmental authorities around the world, and in particular in emerging economies where millions of used ICE vehicles from the US and EU ended up.

This review has identified diesel vehicles as likely the most problematic technology in terms of cheating by vehicle manufacturers, tampering by vehicle owners, and the efficiency of conventional PTR programs. Modern diesel vehicles with advanced emission control systems, such as SCR and DPF, have been increasingly difficult to inspect and repair with legacy test and repair programs. Moreover, the absence of effective PTR programs for these vehicles, or the use of ineffective high-emitter detection methods, may also further encourage vehicle owners to tamper. Since software defeat devices are very difficult to detect, new methods are required to find them.

This leads to the conclusion that a modernized and practical approach to PTR is required to address cheating and tampering of ICE vehicles, in particular diesel vehicles. We recommend an enhanced and fit-for-purpose PTR program that includes the following:

• On-road, real-world emission monitoring programs for the detection of high-emitters using screening methods such as remote (emission) sensing, plume-based and on-road methods, and remote OBD.
• For those identified as high-emitters, a targeted test and repair program using improved in-service inspection methods, including (remote) OBD scans, a low-idle PN emission test, and a loaded dynamometer test for NO_x emissions.
• The development of high-emitter profiles based on fleet-specific attributes, such as vehicle age, fuel economy, mileage, engine characteristics, weight, seasonal fuel quality, and meteorology.
• Active audit analysis and testing under real-world conditions for emission regulation compliance and enforcement by the relevant authorities.
• Appropriate funding, equipment, and training for PTR inspectors and repair and diagnostic technicians.
• Further development of low-cost methods for inclusion in PTR programs to detect tampering impacts on NO_x emissions by SCR emulators and EGR deactivation.
• Transparency and (public) availability of PTR data on failed vehicles and the underlying technical flaws.
• The continuation of research and development into improved PTR methodologies and technologies.
• Strengthening of enforcement action for tampering and cheating, in particular outside the US.
• Modifying existing legal frameworks to address issues such as the sale of tampering devices.

Finally, this study was wide in scope and intended to provide an up-to-date and in-depth review of the current state of play regarding cheating, tampering, malfunctioning emission control systems, and possible solutions, including modernized PTR programs. There are various aspects, however, that could be examined in further detail in future follow-up studies. Some suggestions include the following:

• A comparative analysis of legislative frameworks, enforcement actions, and effectiveness across different jurisdictions and how manufacturers responded to enforcement actions (i.e., specific case studies).
• A cost–benefit analysis of emissions regulations and in-service test and repair programs (case-studies).
• A more in-depth comparative analysis (advantages and limitations) of different policy tools and cheating device detection methods, such as OBD, Real Driving Emissions (RDEs), remote sensing, plume chasing, software analysis, etc.
• Further analysis of the long-term effects of testing schemes in different regions based on relevant empirical studies or data.