Development and application of an aerosol screening model for size-resolved urban aerosols

Charles O Stanier, Sang-Rin Lee
Research Report (Res Rep Health Eff Inst) 2014, (179): 3-79
Predictive models of vehicular ultrafine particles less than 0.1 microm in diameter (UFPs*) and other urban pollutants with high spatial and temporal variation are useful and important in applications such as (1) decision support for infrastructure projects, emissions controls, and transportation-mode shifts; (2) the interpretation and enhancement of observations (e.g., source apportionment, extrapolation, interpolation, and gap-filling in space and time); and (3) the generation of spatially and temporally resolved exposure estimates where monitoring is unfeasible. The objective of the current study was to develop, test, and apply the Aerosol Screening Model (ASM), a new physically based vehicular UFP model for use in near-road environments. The ASM simulates hourly average outdoor concentrations of roadway-derived aerosols and gases. Its distinguishing features include user-specified spatial resolution; use of the Weather Research and Forecasting (WRF) meteorologic model for winds estimates; use of a database of more than 100,000 road segments in the Los Angeles, California, region, including freeway ramps and local streets; and extensive testing against more than 9000 hours of observed particle concentrations at 11 sites. After initialization of air parcels at an upwind boundary, the model solves for vehicle emissions, dispersion, coagulation, and deposition using a Lagrangian modeling framework. The Lagrangian parcel of air is subdivided vertically (into 11 levels) and in the crosswind direction (into 3 parcels). It has overall dimensions of 10 m (downwind), 300 m (vertically), and 2.1 km (crosswind). The simulation is typically started 4 km upwind from the receptor, that is, the location at which the exposure is to be estimated. As parcels approach the receptor, depending on the user-specified resolution, step size is decreased, and crosswind resolution is enhanced through subdivision of parcels in the crosswind direction. Hourly concentrations and size distributions of aerosols were simulated for 11 sites in the Los Angeles area with large variations in proximal traffic and particle number concentrations (ranging from 6000 to 41,000/cm3). Observed data were from the 2005-2007 Harbor Community Monitoring Study (HCMS; Moore et al. 2009), in Long Beach, California, and the Coronary Health and Air Pollution Study (CHAPS; Delfino et al. 2008), in the Los Angeles area. Meteorologic fields were extracted from 1-km-resolution meteorologic simulations, and observed wind direction and speed were incorporated. Using on-road and tunnel measurements, size-resolved emission factors ranging from 1.4 x 10(15) to 16 x 10(15) particles/kg fuel were developed specifically for the ASM. Four separate size-resolved emissions were used. Traffic and emission factors were separately estimated for heavy-duty diesel and light-duty vehicles (LDV), and both cruise and acceleration emission factors were used. The light-duty cruise size-resolved number emission factor had a single prominent mode at 12 nm. The diesel cruise size-resolved number emission factor was bimodal, with a large mode at 16 nm and a secondary mode at around 100 nm. Emitted particles were assumed to be nonvolatile. Data on traffic activity came from a 2008 travel-demand model, supplemented by data on diurnal patterns. Simulated ambient number size distributions and number concentrations were compared to observations taking into account estimated losses from particle transmission efficiency in instrument inlet tubing. The skill of the model in predicting number concentrations and size distributions was mixed, with some promising prediction features and some other areas in need of substantial improvement. For long-term (-15-day) average concentrations, the variability from site to site could be modeled with a coefficient of determination (r2) of 0.76. Model underprediction was more common than overprediction. The average of the absolute normalized bias was 0.30; in other words, long-term mean particle concentrations at each site were on average predicted to within 30% of the measured values. Observed 24-hour number concentrations were simulated to within a factor of 1.6 on 48% of days at HCMS sites and 81% at CHAPS sites, lower than the original design goal of 90%. Extensive evaluation of hourly concentrations, diurnal patterns, sizedistributions, and directional patterns was performed. At two sites with heavy freeway and heavy-duty-vehicle (HDV) influences and extensive size-resolved measurements, the ASM made significant errors in the diurnal pattern, concentration, and mode position of the aerosol size distribution. Observations indicated a shift in concentrations and size distributions corresponding to the afternoon development of offshore wind at the HCMS sites. The model did not reproduce the changes in particles associated with this wind shift and suffered from overprediction for particles of less than 15 nm and underprediction for particles of between 15 and 500 nm, raising doubt about the applicability of the HDV emission factors and the model's assumptions that particles were nonvolatile. The model's temporal prediction skill at individual monitoring sites was variable; the index of agreement (IOA) for hourly values at single sites ranged from 0.30 to 0.56. The model's ability to reproduce diurnal patterns in aerosol concentrations was site dependent; midday underprediction as well as underprediction for particle sizes greater than 15 nm were typical errors. Despite some problems in model skill, the number of time periods and locations evaluated as well as the extent of our qualitative and quantitative evaluations versus physical measurements well exceeded other published size-resolved modeling efforts. As a trial of a typical application, the sensitivity of the concentrations at each receptor site to LDV traffic, HDV traffic, and various road classes was evaluated. The sensitivity of overall particle numbers to all types of traffic ranged from 0.87 at the site with the heaviest traffic to 0.28 at the site with the lightest traffic, meaning that a 1% reduction in traffic could yield a reduction in particle number of 0.87% to 0.28%. Key conclusions and implications of the study are the following: 1. That variable-resolution (down to 10 m) modeling in a relatively simple framework is feasible and can support most of the applications mentioned above; 2. That model improvements will be required for some applications, especially in the areas of the HDV emission factor and the parameterization of meteorologic dispersion; 3. That particle loss from instrument transmission efficiency can be significant for particles smaller than 50 nm, and especially significant for particles smaller than 20 nm. In cases where loss corrections are not accounted for, or are inaccurate, this loss can cause disagreements in observation-model and observation-observation comparisons. 4. That LDV traffic exposures likely exceed HDV traffic exposures in some locations; 5. That variable step size and adaptive parcel width are critical to balancing computational efficiency and resolution; and 6. That the effects of roadways on air quality depend on both traffic volume and distance--in other words, low traffic volumes at close proximity need to be considered in health and planning studies just as much as do high traffic volumes at distances up to several kilometers. Future improvements to the model have been identified. They include improved emission factors; integration with the U.S. Environmental Protection Agency (EPA) Motor Vehicle Emission Simulator (MOVES) model; nesting with three-dimensional (3D) Eulerian models such as the Community Multi-scale Air Quality (CMAQ) model; increased emission dependence on acceleration, load, grade, and speed as well as evaporation and condensation of semivolatile aerosol species; and modeling of carbon dioxide (CO2) as an on-road and near-road dilution tracer. In addition, comparison with other statistically and physically based models would be highly beneficial.

Full Text Links

Find Full Text Links for this Article


You are not logged in. Sign Up or Log In to join the discussion.

Related Papers

Remove bar
Read by QxMD icon Read

Save your favorite articles in one place with a free QxMD account.


Search Tips

Use Boolean operators: AND/OR

diabetic AND foot
diabetes OR diabetic

Exclude a word using the 'minus' sign

Virchow -triad

Use Parentheses

water AND (cup OR glass)

Add an asterisk (*) at end of a word to include word stems

Neuro* will search for Neurology, Neuroscientist, Neurological, and so on

Use quotes to search for an exact phrase

"primary prevention of cancer"
(heart or cardiac or cardio*) AND arrest -"American Heart Association"