Implementation of a high-resolution Source-Oriented WRF/Chem model at the Port of Oakland

Highlights

• We develop a high resolution Source-Oriented WRF/Chem model (SOWC-HR).

• We predict PM_2.5 EC source contributions at 250 m resolution over Oakland CA.

• Higher resolution increased population-weighted EC exposure by 17% for some sources.

• Traffic, rail, and ships are dominant sources.

Abstract

A version of the Source-Oriented WRF/Chem (SOWC) model with 250 m spatial resolution (SOWC-HR) was developed and implemented to perform high resolution simulations over the community of Oakland, California, during March 2010. A multiscale set of nested domains was used to predict contributions to airborne particulate elemental carbon (EC) concentrations from ships, trains, and on-road diesel trucks. The final domain at 250 m resolution used Large Eddy Simulation (LES) to predict turbulent mixing at scales where traditional first order closure models are not valid. Results of the high resolution simulation with the nested LES (HR case) and without the nested LES (non-HR case) were compared to speciated particulate matter (PM) measurements and source contributions calculated using Positive Matrix Factorization (PMF). The PMF results showed that on-road diesel traffic was a major EC contributor, a result consistent with previous studies for Oakland. The average EC concentration predicted at the site by the SOWC-HR model was 0.42 μg m⁻³, with source contributions of 0.22 μg m⁻³ from on-road diesel, 0.05 μg m⁻³ from ship fuel combustion, 0.08 μg m⁻³ from trains, and 0.09 μg m⁻³ from other sources. Both simulation cases predicted similar total EC concentrations and source contributions at the sampling sites, but more substantial differences were predicted at other locations in the study region. The HR case predicted higher average and maximum hourly EC contributions from all sources compared the non-HR case. The greatest relative increase of maximum hourly EC was seen in the on-road diesel source, which increased by nearly a factor of 2 (3.74 μg m⁻³ to 6.69 μg m⁻³) when spatial resolution was increased from 1 km to 250 m. The SOWC-HR model predicted greater population-weighted EC concentrations from all sources when compared to the SOWC model without HR. The increase in period-averaged EC exposure from each source ranged from +1% to +17%, while the increase in maximum hourly EC exposure from each source ranged from +9% to +32%. This evaluation shows that resolving neighborhood scales through the representation of local mixing phenomena can significantly impact pollutant concentration predictions, especially when examining extreme exposures in a densely populated area with many sources and complex terrain.

Introduction

The San Francisco Bay Area in California is a densely populated metropolitan region with a variety of air pollution sources and complex topography. The West Oakland community within the Bay Area has a population of 22,200 in a relatively small area of 7.7 km², which lies adjacent to the Port of Oakland and the Union Pacific Rail yard, and is bounded by three major freeways (Di, 2008) (see Fig. 1). The terrain surrounding Oakland has elevation ranging from sea level to approximately 500 m in the hills 15 km to the east. Mesoscale circulation driven by differential heating over inland areas versus over the ocean produces a land–sea breeze wind system that interacts with the terrain to produce complex wind patterns and regions of micro-climates. High spatial resolution and sophisticated modeling treatments are needed to accurately predict population exposure to air pollution mixtures given these conditions.

The air pollutant of greatest concern in the Oakland region is airborne particles with diameter less than 2.5 μm (PM_2.5). PM_2.5 is composed of numerous solid and liquid chemical components in size fractions as small as a few nanometers (nm). The chemical components in PM_2.5 may be emitted directly to the atmosphere in the condensed form or they can be produced from atmospheric chemical reactions. The majority of the PM_2.5 in Oakland is thought to originate from various types of fuel combustion (Tanrikulu et al., 2011b), but the dominant sources are difficult to identify given the complex formation pathways and number of different sources. A lack of clear relationships between emissions sources and air pollution exposure makes it difficult to design control programs to protect public health.

Previous modeling studies have examined the sources of PM_2.5 and associated health risks in the Bay Area using a variety of multiscale regional air quality models, including CAMx, CMAQ, and WRF (Deng and Stauffer, 2011, Tanrikulu et al., 2009a, Tanrikulu et al., 2009b). These simulations were performed at high spatial resolutions (4 km–1 km) and identified sharp spatial gradients of PM concentration around major sources. These sharp gradients lead to complex patterns of population exposure at the neighborhood scale, which can have a significant impact on health risks in these densely populated areas (Tanrikulu et al., 2011b). Higher spatial resolutions (250 m) have been used in receptor-based models to simulate annual average air pollution in Oakland (Di, 2008), but these receptor models use simplified treatments of meteorology, particle size distributions, and chemical reactions, and are not typically suited to predict population exposure over an entire city. A need exists to predict exposure to reactive air pollution mixtures at neighborhood scales in communities like Oakland across the US.

The use of high spatial resolution avoids numerical artifacts that can smooth fine spatial features in predicted concentration fields, but previous studies show that the accuracy of the overall model prediction is still influenced by the accuracy of the input data. Primary pollutants such as PM_2.5 EC have sharper spatial gradients than secondary pollutants such as ozone, but a study by Valari and Menut (2008) suggests that high resolution emissions input data is needed to capture these features. A study by Thompson and Selin (2012) suggests that increased model spatial resolution may not reduce uncertainties enough to recognize significant differences in health impact predictions for ozone exposure.

The objective of this study is to develop a method to predict source contributions to chemically reacting air pollution mixtures with sufficient spatial resolution to accurately calculate population exposure in the presence of sharp spatial concentration gradients. This method is applied to predict the spatial distribution of a primary pollutant (PM_2.5 EC) in a region where secondary transformations (condensation of nitrate, SOA, and other semi-volatile material) could influence the dry deposition rate and therefore the concentration field. A version of the Source-Oriented WRF/Chem (SOWC) model (Zhang et al., 2013) was modified to work at high resolution (HR) for this purpose. The model uses Large Eddy Simulation (LES) to predict source contributions to the size and composition distribution of airborne particulate matter at neighborhood scales of 250 m. The SOWC-HR model was implemented to simulate pollutant concentrations over the city of Oakland during the month of March 2010. Predictions of source-resolved elemental carbon (EC) concentrations were compared to receptor-based source apportionment results calculated using Positive Matrix Factorization at the Port of Oakland and at the West Oakland community monitoring location. Population-weighted EC exposure was also calculated to evaluate the differences caused by spatial variation ranging from 1 km down to 250 m.