Increasing active travel is a commonly cited strategy to improve public health and reduce the environmental impact of transportation. Active travel has a number of benefits to both the individual (e.g., physical activity) and to society (e.g., reduced emissions, shifting trips off the motor-vehicle network); however, how cyclists and pedestrians are exposed to hazards during active travel is understudied. One such risk factor is exposure to on-road particulate air pollution. My dissertation explores how people are exposed to particulate air pollution during active travel and how spatial patterns of cycling and walking traffic may impact exposure. I measured real-time particulate air pollution (particle number, black carbon, PM2.5, and particle size) while cycling in Minneapolis, MN. I then use those measurements to explore exposure to particulate air pollution during active travel including: (1) assessing whether measures of particulates are correlated with each other, with motor-vehicle traffic counts, and with traffic mix (i.e., truck, bus, or passenger vehicle), (2) developing land use regression (LUR) models to determine correlates of particulate air pollution and to extrapolate estimates to locations without measurements, and (3) comparing the spatial patterns of modeled particulate air pollution with the spatial patterns of bicycle and pedestrian traffic. I collected 129 hrs of measurements (886 miles) on streets and off-street trails during 42 sampling runs in year-2012. Mobile bicycle-based sampling was completed using a bicycle trailer, modified to safely carry air pollution equipment and to allow for an elevated air intake that samples at heights close to the breathing zone of cyclists and pedestrians. I carried air pollution instrumentation that measures four aspects of particulate air pollution: (1) TSI CPC 3007 (particle number concentration), (2) AethLabs AE-51 micro-aethalometer (black carbon mass concentration), (3) TSI DustTrak 8530 (PM2.5 mass concentration), and (4) TSI NanoScan (particle size distribution). A GPS and temperature and relative humidity sensor were included on the sampling platform to geo-locate and adjust air pollution measurements for meteorological variables. Two video cameras were included to assess the impact of traffic volume and mix on exposure concentrations. In Chapter 3, I explore trends in measured on-road concentrations associated with traffic dynamics (i.e., volume and mix), between-pollutant correlations, and how concentrations vary by street characteristics. I developed a set of underwrite functions to control for day-to-day variability in background concentrations, and I used the background-adjusted dataset for all spatial comparisons of concentrations. (Analyses here that are not focused on spatial patterns do not use this adjustment). I found that black carbon and particle number (two traffic-related pollutants) were the most correlated among the pollutants measured (R2: 0.39 and 0.61, respectively, for 1- and 30-minute averages). Correlations among all pollutant pairs increased with averaging time. Particle size was modestly correlated with particle number (negative correlation) and with black carbon (positive correlation). In regression models that control for meteorological parameters, median particle size on-road is more correlated with background concentration of black carbon and particle number than on-road concentration (ratio of background to on-road model β for black carbon [particle number]: 5.3 [8.3]). Traffic counts (assessed from the video footage) were correlated with particulate concentrations; the presence of a truck in the video frame (from a camera mounted on the handlebars) was correlated with a 30,600 pt/cc [1.6 µg/m3] increase in particle number [black carbon] concentration compared to a 2,300 pt/cc [0.26 µg/m3] increase for buses and a 240 pt/cc [0.02 µg/m3] increase for passenger-vehicles. Concentrations were associated with street functional class and decreased with distance from a major road; presence of a bike facility had little impact on concentrations. In Chapter 4, I describe an approach for developing land use regression (LUR) models from the mobile measurements to estimate concentrations at locations where measurements are not available. I aggregated measurements at various spatial resolutions and tested model performance using a number of different averaging times. Models were developed using 215 candidate independent variables and a stepwise regression approach that selects the variables most correlated with concentrations from those candidate variables. In total 1,224 LUR models were generated; 8 models were chosen to extrapolate concentration estimates to the entire City of Minneapolis. Variables that were most commonly selected in the models either described separation from emission sources (e.g., distance from a major road, area of open space, or length of local [low-traffic] road) or proximity to or density of emissions (e.g., length of major road or freeway, industrial area, or traffic intensity). In general model goodness-of-fit statistics were modest; morning [afternoon] adjusted R2 were 0.50 [0.48] for particle number, 0.28 [0.42] for black carbon, 0.30 [0.49] for PM2.5, and 0.29 [0.20] for particle size. Models generally underestimated the highest concentration locations; analysis of spatial autocorrelation of model residuals suggests this may be owing to measurements at highly congested street segments during rush-hour. Since data were not available on real-time congestion my models do not capture this relationship. In Chapter 5, I present a sample application of the LUR models by comparing spatial patterns of particulate air pollution to those of bicycle and pedestrian traffic. I estimated bicycle and pedestrian traffic volumes for every street segment in Minneapolis using regression models based on peak-hour bicycle and pedestrian traffic counts. I then used the LUR models developed in Chapter 4 to estimate concentrations for the same street segments. I overlaid the estimates of these two factors to assess spatial patterns of population exposure during active travel. Few neighborhoods in Minneapolis were classified as "healthy" (i.e., high rates of active travel and low particulate concentrations). Bicycle and pedestrian traffic is highest on arterial and collector streets where estimated concentrations are also highest; moving one block away from major roads reduces morning [afternoon] exposure concentrations by 21% [12%] for particle number, 15% [20%] for black carbon, and 7% [3%] for PM2.5. Non-motorized traffic increased for areas with greater land use mix while air pollution concentrations were unchanged; non-motorized traffic and particulate concentrations were not well correlated with population density. Finally, the burden of exposure during active travel appears to occur in neighborhoods that are low-income and have the highest proportion of non-whites (i.e., low-income, non-white neighborhoods were associated with high rates of active travel and increased particulate concentrations). People exposed in those locations may be residents or visitors (e.g., in downtown Minneapolis where job density is high). This dissertation presents work to describe the spatial patterns of exposure to particulate air pollution during active travel in Minneapolis. The analyses herein provide evidence that mobile measurements of air pollution concentrations can yield insight into how and where cyclists and pedestrians are exposed. The results presented here may be of interest to planners and policy-makers interested in designing clean, healthy cities.