Mobile and short-term monitoring campaigns are increasingly used to develop land-use regression (LUR) models for ultrafine particles (UFP) and black carbon (BC). It is not yet established whether LUR models based on mobile or short-term stationary measurements result in comparable models and concentration predictions. The goal of this paper is to compare LUR models based on stationary (30 min) and mobile UFP and BC measurements from a single campaign. An electric car collected both repeated stationary and mobile measurements in Amsterdam and Rotterdam, The Netherlands. A total of 2964 road segments and 161 stationary sites were sampled over two seasons. Our main comparison was based on predicted concentrations of the mobile and stationary monitoring LUR models at 12 682 residential addresses in Amsterdam. Predictor variables in the mobile and stationary LUR model were comparable, resulting in highly correlated predictions at external residential addresses (R of 0.89 for UFP and 0.88 for BC). Mobile model predictions were, on average, 1.41 and 1.91 times higher than stationary model predictions for UFP and BC, respectively. LUR models based upon mobile and stationary monitoring predicted highly correlated UFP and BC concentration surfaces, but predicted concentrations based on mobile measurements were systematically higher.

Background: There is growing evidence that exposure to ultrafine particles (UFP; particles smaller than ) may play an underexplored role in the etiology of several illnesses, including cardiovascular disease (CVD). Objectives: We aimed o investigate the relationship between long-term exposure to ambient UFP and incident cardiovascular and cerebrovascular disease (CVA). As a secondary objective, we sought to compare effect estimates for UFP with those derived for other air pollutants, including estimates from two-pollutant models. Methods: Using a prospective cohort of 33,831 Dutch residents, we studied the association between long-term exposure to UFP (predicted via land use regression) and incident disease using Cox proportional hazard models. Hazard ratios (HR) for UFP were compared to HRs for more routinely monitored air pollutants, including particulate matter with aerodynamic diameter ( ), PM with aerodynamic diameter ( ), and . Results: Long-term UFP exposure was associated with an increased risk for all incident CVD [ per ; 95% confidence interval (CI): 1.03, 1.34], myocardial infarction (MI) ( ; 95% CI: 1.00, 1.79), and heart failure ( ; 95% CI: 1.17, 2.66). Positive associations were also estimated for ( ; 95% CI: 1.01, 1.48 per ) and coarse PM ( ; HR for all ; 95% CI: 1.01, 1.45 per ). CVD was not positively associated with (HR for all ; 95% CI: 0.75, 1.28 per ). HRs for UFP and CVAs were positive, but not significant. In two-pollutant models ( and ), positive associations tended to remain for UFP, while HRs for and generally attenuated towards the null. Conclusions: These findings strengthen the evidence that UFP exposure plays an important role in cardiovascular health and that risks of ambient air pollution may have been underestimated based on conventional air pollution metrics. https://doi.org/10.1289/EHP3047

Air pollution measurements collected through systematic mobile monitoring campaigns can provide outdoor concentration data at high spatial resolution. We explore approaches to minimize data requirements for mapping a city's air quality using mobile monitors with "data-only" versus predictive modeling approaches. We equipped two Google Street View cars with 1-Hz instruments to collect nitric oxide (NO) and black carbon (BC) measurements in Oakland, CA. We explore two strategies for efficiently mapping spatial air quality patterns through Monte Carlo analyses. First, we explore a "data-only" approach where we attempt to minimize the number of repeated visits needed to reliably estimate concentrations for all roads. Second, we combine our data with a land use regression-kriging (LUR-K) model to predict at unobserved locations; here, measurements from only a subset of roads or repeat visits are considered. Although LUR-K models did not capture the full variability of onroad concentrations, models trained with minimal data consistently captured important covariates and general spatial air pollution trends, with cross-validation R 2 for log-transformed NO and BC of 0.65 and 0.43. Data-only mapping performed poorly with few (1−2) repeated drives but obtained better cross-validation R 2 than the LUR-K approach within 4 to 8 repeated drive days per road segment.

Land use regression (LUR) models for air pollutants are often developed using multiple linear regression techniques. However, in the past decade linear (stepwise) regression methods have been criticized for their lack of flexibility, their ignorance of potential interaction between predictors, and their limited ability to incorporate highly correlated predictors. We used two training sets of ultrafine particles (UFP) data (mobile measurements (8200 segments, 25 s monitoring per segment), and short-term stationary measurements (368 sites, 3 × 30 min per site)) to evaluate different modeling approaches to estimate long-term UFP concentrations by estimating precision and bias based on an independent external data set (42 sites, average of three 24-h measurements). Higher training data R 2 did not equate to higher test R 2 for the external long-term average exposure estimates, making the argument that external validation data are critical to compare model performance. Machine learning algorithms trained on mobile measurements explained only 38−47% of external UFP concentrations, whereas multivariable methods like stepwise regression and elastic net explained 56−62%. Some machine learning algorithms (bagging, random forest) trained on short-term measurements explained modestly more variability of external UFP concentrations compared to multiple linear regression and regularized regression techniques. In conclusion, differences in predictive ability of algorithms depend on the type of training data and are generally modest.

No abstract