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

D. K. Joe, Hongliang Zhang, S. P. DeNero, Hsiang‐He Lee, Shu‐Hua Chen, Brian McDonald, Robert A. Harley, Michael J. KleemanAtmospheric 2013 · DOI

TL;DR

This paper introduces a high-resolution (250m) version of the Source-Oriented WRF/Chem (SOWC-HR) model to better capture localized air pollution hotspots. By implementing Large Eddy Simulation (LES) at this fine scale, the model reveals that higher resolution significantly increases predicted population-weighted exposure to pollutants like elemental carbon (EC), particularly during extreme events, highlighting the importance of micro-scale atmospheric processes.

What Problem Does This Paper Address?

Accurately predicting air pollution exposure, especially for reactive mixtures and in densely populated areas with complex terrain, is a significant challenge. Traditional regional air quality models often operate at resolutions (e.g., 1-4 km) that can smooth out fine spatial features and local micro-climates, leading to an underestimation of peak concentrations and localized population exposure. This is particularly problematic for primary pollutants like elemental carbon (EC), which exhibit sharp spatial gradients. The authors aim to bridge this gap by developing and implementing a model that can resolve these neighborhood-scale phenomena.

Methods

The core of this work is the adaptation of the Source-Oriented WRF/Chem (SOWC) model to achieve high spatial resolution (250m) using Large Eddy Simulation (LES) within a nested domain framework. The SOWC model tracks pollutants from their sources, allowing for more realistic simulations of particle size distributions and optical properties. Key methodological advancements include:

  • High-Resolution Nesting: A multiscale domain configuration was used, with the innermost domain reaching 250m resolution.
  • Large Eddy Simulation (LES): LES was employed in the 250m domain to explicitly resolve turbulent mixing at scales where traditional parameterizations are insufficient.
  • Adaptive Time Step: To handle the computational demands of the fine grid, an adaptive time step method was implemented.
  • Bilinear Interpolation Scheme: A simplified bilinear interpolation scheme was developed to address numerical issues that arise with sharp spatial gradients in source-oriented pollutant fields when nesting.

Datasets

  • Meteorological Inputs: NARR (North American Regional Reanalysis) data was used for meteorological inputs, processed via the WRF preprocessing system (WPS).
  • Emissions Data:
    • 4 km resolution emissions data from the California Air Resources Board (CARB) for larger domains.
  • Higher resolution on-road diesel vehicle emissions data for the finer domains, incorporating Caltrans truck counts, BAAQMD surveys, and updated emission factors.
  • Shipping, train, and "other" source emissions were downscaled from CARB data, with modifications based on local analyses and previously reported studies.
  • Speciated PM Measurements: Data from the BAAQMD West Oakland monitoring site was used for model evaluation.
  • Source Apportionment Data: Positive Matrix Factorization (PMF) results from the BAAQMD West Oakland site and the Port of Oakland were used to compare with model predictions.
  • Population Data: Tract level 2007 census data for Oakland was used for population-weighted exposure calculations.

Key Insights and Ideas

  • LES is Crucial for Extreme Events: The study strongly suggests that LES at fine scales is essential for accurately predicting extreme pollutant concentrations, which are critical for assessing health risks in densely populated areas.
  • Resolution Matters for Exposure: While total EC concentrations at sampling sites might not differ drastically between high-resolution and lower-resolution models, the population-weighted exposure can increase significantly (up to 32% for maximum hourly EC) when using the higher resolution, especially in areas with high population density near emission sources.
  • Dilution Effects are Real: The difference between HR (250m) and non-HR (1km) results, even when HR is averaged back to 1km, indicates that LES's explicit treatment of turbulence fundamentally alters predicted concentration fields compared to parameterized boundary layer schemes.
  • Source Dominance Confirmed: The study reinforces that on-road diesel traffic is the dominant source of EC in Oakland, followed by trains and ships, aligning with previous receptor-based analyses.

Key Results

  • Increased Exposure: Population-weighted 8-day average EC exposure increased by 1-17% with the HR model, while maximum hourly EC exposure increased by 9-32%.
  • Maximum EC Concentration: The HR model predicted a nearly twofold increase in maximum hourly EC from on-road diesel (3.74 µg m⁻³ to 6.69 µg m⁻³) when increasing resolution from 1 km to 250 m, with the 250m HR model predicting up to 9.94 µg m⁻³.
  • Model Performance: The SOWC-HR model showed reasonable agreement with measurements and PMF results at the West Oakland site, though it tended to underpredict EC at the Port of Oakland site due to its proximity to a busy road.
  • Spatial Gradients: The HR case resolved higher EC concentrations around sources and identified localized areas with extremely high concentrations not seen in the non-HR simulations.

Limitations and Future Work

  • Measurement Site Proximity: The Port of Oakland sampling site's proximity to a major road led to underprediction of on-road diesel EC, making direct model-measurement comparisons challenging for this source at this location.
  • PMF Limitations: The PMF analysis struggled to distinguish between on-road diesel and train emissions, and the "train" source was not resolved by PMF at all.
  • Future Work: The authors suggest further analysis to determine the associated changes in public health impacts and risks due to the identified differences in exposure. They also propose more spatial analysis to investigate policy implications. The study highlights the need for further research into how these high-resolution predictions can inform policy and mitigation strategies.

Summary updated

BibTeX
@inproceedings{joe2013implementation,
  author = {D. K. Joe and Hongliang Zhang and S. P. DeNero and Hsiang‐He Lee and Shu‐Hua Chen and Brian McDonald and Robert A. Harley and Michael J. Kleeman},
  title = {{Implementation of a high-resolution Source-Oriented WRF/Chem model at the Port of Oakland}},
  booktitle = {Atmospheric},
  year = {2013},
  doi = {10.1016/j.atmosenv.2013.09.055},
}
Figure