Atmospheric dispersion models play a critical role in the management of air quality. They are used to estimate the impact of sources of air pollutants on air quality. They are also used to infer emissions from sources from measurements of pollutant concentration. Regulatory agencies used air quality models to determine whether existing or proposed industrial facilities comply with regulatory requirements.

In my research, I developed a class of dispersion models to study four problems. I developed two Lagrangian dispersion models to estimate emissions of wildfires using data from PM monitoring networks and the High-Resolution Rapid Refresh (HRRR) meteorological model. By integrating data from ground-based monitors and NASA's Moderate Resolution Imaging Spectroradiometer (MODIS), I created a comprehensive model that enables improved spatial and temporal resolution for assessing PM2.5 concentrations during wildfires. This integrated technology, capable of evaluating concentrations at 1 km spatial resolution with 1 hour temporal resolution, has significant implications for health risk assessment, evacuation planning, and policy development related to wildfire impacts.

I next demonstrated the application of a dispersion model to estimate emissions of methane (CH4) using total column measurements from four field campaigns conducted in the San Joaquin Valley of California during four seasons from March 2019 until January 2020. The atmospheric column dry mixing ratios of CH4 were retrieved from multiple EM27/SUN solar spectrometers deployed upwind and downwind of a cluster of dairy farms. These measurements were complemented with satellite observations of column-averaged CH4 from the S5P/TROPOMI satellite instrument over the same area to extend the analysis to larger scales and periods.

The next study involved the development and application of a model to estimate the impact of vehicle tailpipe emissions on people waiting next to idling vehicles. We conducted a field study designed to collect CO2 concentration data at distances of a few meters from the tailpipe of a vehicle: the accelerator pedal was controlled to simulate idling and acceleration from a stop. Analysis of the data shows that the measurements are described within a factor of two with a dispersion model that uses micrometeorological variables as inputs and includes plume rise associated with the buoyancy of the exhaust plume. The data suggest that people situated a few meters from an idling vehicle are likely to be exposed to levels of NO2 that are above the Clean Air Act 1-hour standard of 100 ppb.

The final study examined the application of a dispersion model to estimate PM10 emissions from roads. This involved sampling silt loadings on roads using a mobile dust collection system that I helped to design and build. The sampling was conducted on two freeways and two city roads in Riverside, California. PurpleAir PM monitors and PICARRO CO2 monitors were deployed on the mobile platform to measure road dust concentrations, which was then used to infer emission factors with a line source dispersion model, and carbon mass balance approach. The results indicated that freeways have lower emission factors compared to the city roads. The data from the field studies was used to propose a new model for emission factors, which improves upon the currently used regulatory model.

This dissertation summarizes the results from a study to develop and evaluate models to estimate dispersion of pollutants in boundary layers whose properties change with downwind distance. Such boundary layers occur at the interface between rural and urban areas, and water and land bodies.

I first developed a method to estimate the meteorological inputs required to apply the current generation of dispersion models, such as AERMOD, to urban areas. This method uses measurements made at a single level on a tower located in an urban area, and is based on the assumption that similarity methods applicable to spatially homogeneous conditions are locally valid even in an inhomogeneous urban area. I show that under unstable conditions, measurements of temperature fluctuations improve upon commonly used energy balance methods to estimate heat flux. Also, any bias in heat flux estimates has a minor effect on the prediction of surface friction velocity and turbulent velocities.

I examined a method to estimate urban micrometeorology using measurements made on a tower located in an upwind suburban area. I applied an internal boundary layer model to trace the evolution of the boundary layer as it traveled from the suburban measurement location to the urban location of interest. Estimates from the model were observations made during a field study conducted in Riverside, CA. Estimates of friction velocity compare well with urban measurements only when the variation of friction velocity with height within the urban canopy was accounted for.

I examined the performance of two steady-state dispersion models in explaining concentrations measured during field studies designed to study low wind speed conditions typical of urban areas. One model is based on the numerical solution of the two-dimensional mass conservation equation and the other is AERMOD, which accounts for low wind speeds by including wind meandering. The numerical method performs better than AERMOD through a justifiable description of the interaction between dispersion and the gradient of the wind speed near the surface. Including wind meandering, which occurs under low wind speeds, improves the performance of the numerical model. As part of this study, I developed a method to improve estimates of surface friction velocity during low wind speeds.

The last model is applicable to elevated sources, such as power plants, situated close to shorelines. It is designed to be compatible with AERMOD (Cimorelli et al., 2005), the USEPA's regulatory model that is currently designed for spatially homogenous conditions. The semi-empirical shoreline dispersion model accounts for plume entrainment by the thermal internal boundary Layer (TIBL), whose height increases with distance from the shoreline. I show that AERMOD can be modified to account for two-dimensional shoreline effects, and this modified model performs as well in explaining observations as dispersion models specifically designed for shoreline sources. I also developed a method to generate meteorological inputs that are compatible with the current structure of AERMOD's inputs.

Exposure to elevated concentrations of vehicle emitted pollutants is associated with negative health effects. Elevated concentrations are typically found within several hundred meters of high traffic roads, where atmospheric dispersion has not sufficiently diluted pollutants. Tall buildings next to roads reduce dispersion, thereby creating pollutant hot spots and increasing exposure to vehicle emissions for city residents. Roadside barriers enhance dispersion of roadway emissions and thus can be used to mitigate elevated concentrations next to large roads. The work in this thesis develops semi-empirical dispersion models that are useful for estimating near road concentrations of vehicle emissions when there are buildings or barriers next to the road.

Dispersion models that account for the effect of near road barriers on concentrations are developed and evaluated with data from a wind tunnel and a field tracer study. The model evaluation shows that the primary effect of roadside barriers is enhancement of the vertical mixing by an amount proportional to the barrier height. Additionally, turbulence is enhanced in the barrier's wake, resulting in more rapid growth of the pollutant plume. The models perform well during neutral and stable atmospheric conditions. During unstable conditions the models overestimate concentrations. A model that accounts for reduction of the mean wind speed in the barrier wake is unbiased for all stabilities.

Models of the impact of tall buildings next to the road on near road concentrations of vehicle emissions are developed. The models are evaluated with data from field measurements conducted in Los Angeles and Riverside counties, CA, and with data from an urban area in Hannover, Germany. The study specifically investigates dispersion in cities with significant building height variability. Model evaluation shows that vertical turbulent transport dominates dispersion in cities. The primary variables governing near road concentrations of vehicle emissions in cities are the ratio of area weighted building height to street width and the vertical averaged standard deviation of vertical velocity fluctuations. The model informs design of transit oriented developments, dense residential areas located in close proximity to transportation infrastructure, which are used to reduce pollution and greenhouse gas emissions due to transportation.

Chronic exposure to high concentrations of pollutants such as NO2 and ultrafine particles is associated with negative health effects. Studies of exposure to these pollutants require estimates of concentrations at temporal and spatial scales relevant to exposure calculations. We have developed and applied methods to construct these concentration "maps" by using a combination of measurements and modeled results. To estimate concentration patterns at the urban scale of tens of kilometers we have formulated a Lagrangian model to estimate concentrations of NOx, NO2, and O3 over a domain extending over hundreds of kilometers. The model is evaluated with data collected at 21 regional monitoring stations in the San Joaquin Valley Air Basin during 2005. The model provides adequate descriptions of the spatial and temporal variation of concentrations of NO2, and NOx. We then use "residual" Kriging to combine the results from the dispersion model with observed concentrations to produce realistic concentration maps.

To estimate concentration patterns at scales of tens of meters in urban areas we developed a dispersion model that accounts for the effects of local building morphology on dispersion. The data used to evaluate the model was collected in field studies conducted in Los Angeles, California. The studies measured ultrafine particle concentrations and associated micrometeorology at several locations with different building morphologies. Surface concentrations in urban areas are primarily controlled by vertical dispersion, which depends on the street aspect ratio, defined as the ratio of the equivalent building height to the street width, and the vertical turbulent velocity ów. The presence of buildings increases the concentrations due to local traffic emissions relative to open areas. Since routine measurements of micrometeorological variables are usually not available in urban areas, we have developed models that allow us to estimate urban surface variables using values measured at an upwind rural location. Results from the urban street scale dispersion model can be combined with measurements from urban monitors to generate concentration maps with spatial resolution of meters and time resolution of minutes.

This dissertation summarizes the results of a five-year investigation of the impact of distributed generation (DG) of electricity on air quality in urban areas. I focused on the impact of power plants with capacities of less than 50 MW, which is typical of DG units in urban areas. These power plants are modeled as buoyant emissions from stacks less than 10 m situated in the midst of urban buildings. Because existing dispersion models are not designed for such sources, the first step of the study involved the evaluation of AERMOD, USEPA's state-of-the art dispersion model, with data collected in a tracer study conducted in the vicinity of a DG unit. The second step of the study consisted of using AERMOD to compare the impact of DG penetration in the South Coast Air Basin of Los Angeles with the impact of replacing DG generation with expansion of current central power plant capacity. The third topic of my investigation is the development and application of a model to examine the impact of non-power plant sources in a large urban area such as Los Angeles. This model can be used to estimate the air quality impact of DG relative to other sources in an urban area.

The first part of this dissertation describes a tracer study conducted in Palm Springs, CA. Concentrations observed during the nighttime experiments are generally higher than those measured during the daytime experiments. They fall off less rapidly with distance than during the daytime. AERMOD provides an adequate description of concentrations associated with the buoyant releases from the DG during the daytime when turbulence is controlled by convection induced by solar heating. However, AERMOD underestimates concentrations during the night when turbulence is generated by wind shear. Also, AERMOD predicts a decrease in concentrations with distance that is much more rapid than the relatively flat observed decrease. I have suggested modifications to AERMOD to improve the agreement between model estimates and observations during the night.

The second part of this dissertation examines the air quality impact of using DG to satisfy future growth in power demand in the South Coast Air Basin of Los Angeles (SoCAB), relative to the impact when the demand is met by expanding current central generation (CG) capacity. The air quality impacts of these two alternate scenarios are quantified in terms of hourly maximum ground-level and annually-averaged primary NOx concentrations, which are estimated using AERMOD. The shift to DGs has the potential for decreasing maximum hourly impacts of power generation in the vicinity of the DGs. The maximum hourly concentration is reduced from 25 ppb to 6 ppb if DGs rather than CGs are used to generate power. However, the annually-averaged concentrations are likely to be higher than for the scenario in which existing CGs are used to satisfy power demand growth. Future DG penetration will add an annual average of 0.1 ppb to the current basin average, 20 ppb, while expanding existing CGs will add 0.05 ppb.

The third part of my dissertation focused on formulating a model to estimate concentrations of NO2, NOx, and O3 averaged over a spatial scale of the order of a kilometer in a domain extending over tens of kilometers. The model can be used to estimate hourly concentrations of these species over time periods of years. It achieves the required computational efficiency by separating transport and chemistry using the concept of species age. Evaluation with data measured at 21 stations distributed over the Los Angeles air basin indicates that the model provides an adequate description of the spatial and temporal variation of the concentrations of NO2 and NOx. Estimates of maximum hourly O3 concentrations show little bias compared to observations, but the scatter is not small.

Near road air quality is a public concern because exposure to elevated concentrations of vehicular pollution is associated with adverse health effects. Roadway design is suggested as a potential strategy to mitigate near-road exposure. The first part of my dissertation describes the development and application of roadway dispersion models to examine the effectiveness of roadway configurations as pollutant mitigation strategies. These configurations include depressed roadways and at-grade roadways with the presence of solid/vegetative barriers.

Roadside solid barriers increase dispersion of pollutants by lofting emissions and inducing a recirculation zone on their leeward edge. I adapt a model, developed using data from a wind tunnel, to describe ultrafine particle measurements made in a field study. This requires modifying the model to account for uncertainties in emissions and meteorological parameters of real-world studies. Results suggest that 1) a model developed under controlled conditions is useful in the complex environment of urban areas, 2) the surface can be taken neutral in modeling dispersion in urban areas., and 3) the primary impact of the barrier is equivalent to shifting the road upwind by a distance of H(U/u*)cosθ.

I next analyze data from a wind tunnel that examined dispersion of emissions from depressed roadways using roadway dispersion models. I show that dispersion governed by the complex flow induced by depressed roads can be described using modified flat-terrain models. The modifications include 1) an initial vertical spread dependent on the geometry of the depressed roadway, and 2) increasing the friction velocity above its upwind value. Also, the vertical concentration profiles under neutral stability conditions are best explained with a vertical distribution function with an exponent of 1.3 rather than the 2 used in most currently used dispersion models.

Health risk assessment of PM2.5 on the community scale requires PM2.5 concentration estimations at the scale of tens of meters. The PM2.5 measurement from air quality stations and satellite-derived PM2.5 estimates cannot provide concentrations at this spatial resolution. In the last part of my dissertation, I describe a “downscaling” system that adapts roadway dispersion models to yield the concentration gradients that are not captured by satellite maps.

PM emissions from roadways can be classified into two primary categories: exhaust emissions and non-exhaust emissions (NEE). Non-exhaust emissions, often referred to as re-suspended dust or road dust, encompass a range of sources such as the mechanical wear of tires, brakes, vehicle components, road materials, and the re-suspension of particles into the atmosphere due to vehicle-induced and atmospheric turbulence. The currently used regulatory model to estimate the road dust is described in the “Compilation of Air Pollutant Emissions Factors” (AP-42).The current version of the AP-42 model is a semi-empirical equation based on two inputs: road surface silt loading sL, which refers to the fraction of surface dust with aerodynamic diameter below 75 μm, and average weight of vehicles on the road. The model has been criticized for yielding unreliable emission results and its lack of a mechanistic foundation. Furthermore, the procedures recommended by AP-42 for sampling silt loading on paved roads require manual collection of road surface materials using a vacuum cleaner. This collection is impractical for high traffic roads, which is the reason that there are few measurements of silt loading or emission factors for roads with average daily traffic (ADT) of more than 10000 vehicles/h.In this study, a mobile dust collection system has been designed to measure silt loading on California freeways with varying traffic volumes. A mobile dust collection system was first designed for a cargo van in spring 2023. Powered by a 3.5 kW UPS, the system utilizes a VacuMaid GV30, 740-watt vacuum cleaner equipped with a HEPA filter bag for dust collection. A GPS-enabled phone tracks the vehicle's location to compute the distance traveled. To ensure full contact between the brush and the ground, a telescopic hollow tube connects it to the vacuum cleaner, secured with straps and adjusted using bungee cords. The system also integrates a Picarro G2401-m CO/CO₂ analyzer, a PurpleAir sensor for measuring PM concentrations, and a 2D-sonic anemometer with a thermistor for monitoring meteorological conditions. In fall 2023, modifications were made to include a new brush fixture and an automated dust sieving system. Additionally, an automatic dust shaking system has been developed to minimize the error of the silt loading. To ensure data accuracy and reliability, measurements from the 2D anemometer and thermistor were calibrated against those from a 3D anemometer, correcting systematic biases in horizontal wind readings and improving emission factor and dispersion modeling precision. Additionally, PurpleAir sensors were calibrated against regulatory-grade systems, and measures were implemented to minimize dust mass loss during silt loading assessments. This mobile platform was deployed in field studies across six highways and two city roads: sections of CA-91, CA-60, CA-71, CA-55, I-15, and I-215 freeways, as well as Chicago Avenue and Iowa Avenue in three counties of California. Over the course of summer 2023 and summer 2024, a total of 109 dust samples were collected from road surfaces. In summer 2023, 64 samples were gathered from two highways (I-215 and CA-91) and two city roads (Iowa Avenue and Chicago Avenue). In spring and summer 2024, 45 samples were collected from CA-91, CA-60, CA-71, CA-55, I-15, I-215 freeways, and Chicago Avenue and Iowa Avenue in Riverside. Field experiments conducted from summer 2023 to summer 2024 provided datasets that facilitated the development of a mechanistic model for PM emission factors, serving as an alternative to the AP-42 model. The primary advantage of the mechanistic model over AP-42 lies in its foundation on physical principles, enhancing its applicability across a broader range of conditions. Additionally, the model's input parameters can be readily quantified without requiring measurements that would cause significant traffic disruptions, making it more practical for real-world applications.