On-road emissions sources degrade air quality, and these sources have been highly regulated. Epidemiological and environmental justice studies often use road proximity as a proxy for traffic-related air pollution (TRAP) exposure, and other studies employ air quality models or satellite observations. To assess these metrics’ abilities to reproduce observed near-road concentration gradients and changes over time, we apply a hierarchical linear regression to ground-based observations, long-term air quality model simulations using Community Multiscale Air Quality (CMAQ), and satellite products. Across 1980−2019, observed TRAP concentrations decreased, and road proximity was positively correlated with TRAP. For all pollutants, concentrations decreased fastest at locations with higher road proximity, resulting in “flatter” concentration fields in recent years. This flattening unfolded at a relatively constant rate for NO x , whereas the flattening of CO concentration fields has slowed. CMAQ largely captures observed spatial−temporal NO 2 trends across 2002−2010 but overstates the relationships between CO and elemental carbon fine particulate matter (EC) road proximity. Satellite NO x measures overstate concentration reductions near roads. We show how this perspective provides evidence that California’s on-road vehicle regulations led to substantial decreases in NO 2 , NO x , and EC in California, with other states that adopted California’s light-duty automobile standards showing mixed benefits over states that did not adopt these standards. ■ INTRODUCTION On-road mobile source emissions have been shown to have large impacts on air pollution, human exposure, and environmental health. 1−3 Documenting spatial and temporal variability in pollution from on-road sources (often called traffic-related air pollution, TRAP) informs epidemiological and environmental justice (EJ) studies of automobile contributions to adverse health outcomes and inequality. 4−7 In addition, identifying emissions and exposure trends over long periods is important for estimating benefits of regulatory programs and projecting impacts of future air pollution policies. 8,9 Tracking source-specific spatial−temporal trends has the potential to improve evaluations of broad regulations implemented on different sources congruently over long periods. 10,11 Research has shown that the emissions mix from the United States vehicle fleet has evolved over recent decades, 12 and studies have revealed multidecadal reductions in air pollution mixture differences between near-and far-road environments. 13−15 These differences will likely continue to shrink with increased electrification and as the current fleet turns over to newer vehicles certified to the EPA’s Tier 3 program beginning with 2017 model year vehicles. 16,17 Even with reduced impact of road travel broadly, however, mobile air quality monitoring continues to show elevated air pollution concentrations in the near-road environment 18 (some of which is due to growing relative importance of stationary sources in urban areas, e.g., restaurants 19), and recent research has identified continued ecological and health impacts imparted by mobile sources. 20−22 The ability of exposure assessment tools such as air quality models, satellite measurements, and distance-based metrics to capture observed air pollution trends in space and time are important for these tools’ expanded viability in health studies and policy evaluations. 23 There is evidence derived from such tools that United States mobile source nitrogen oxide (NO x) emissions estimates may be biased high by as much as a factor of two. 2,24−28 Juxtaposed with these findings of bias, however, multiple studies have concluded that modeled emissions capture air pollutant trends accurately enough for air quality models to provide adequate estimates of air quality and changes over time. 4,29−35 For some model applications (such as regulatory accountability studies and/or epidemiological studies), capturing spatial and/or temporal trends in emissions changes is more important than accurately capturing absolute emissions. 10