Environmental Engineering Reference
In-Depth Information
the models tend to underpredict peak ozone concentrations both at the surface and
aloft and model performance does not meet traditional goals for ozone. Prior
modeling efforts in the San Joaquin Valley (SJV) have also shown this tendency.
For example, during the SJV Air Quality Study (SJVAQS) - Atmospheric Utility
Signatures, Predictions, and Experiments (AUSPEX) Regional Modeling Adap-
tation Project (SARMAP), ozone concentrations at about 500 m above the surface
were underestimated by as much as 60 ppb (Thuillier and Ranzieri, 1995; DaMassa
et al., 1996). These performance issues undermine confidence in the models as
planning tools and need to be addressed.
While some limited evaluations of model performance aloft have been made,
model evaluation studies (e.g., Tesche et al., 2004; San Joaquin Valley Air
Pollution Control District, 2004) have focused on the ability to replicate surface-
based measurements. Greater attention to assessing aloft model performance using
data collected on tall towers, balloon soundings, and instrumented aircraft is
needed. This study focused on understanding and improving model performance
for ozone and ozone precursors aloft. Understanding model performance both at
the surface and aloft allows us to identify critical performance problems, diagnose
their causes, and implement changes in the modeling system that will improve its
ability to predict the spatial and temporal patterns of observed ozone and thus,
increase credibility.
2. Methodology
Two ozone episodes and their associated simulations with CAMx were investigated
in this study: (1) July 29-August 2, 2000 and (2) September 16-20, 2000. Two
simulations were evaluated for the first episode. The first simulation used
hybrid meteorology from the Pennsylvania State University/National Center for
Atmospheric Research Mesoscale Model version 5 (MM5) and the California
Meteorological Model (CALMET) while the second simulation used meteorology
from MM5 alone. Only one simulation, based on MM5 meteorology, was evaluated
for the second episode.
Extensive analyses of the available air quality and meteorological data were
performed to clarify the chemical and physical processes leading to elevated con-
centrations of ozone aloft and provide the basis for diagnosis of aloft model
performance deficiencies. The analyses included descriptive statistics; three-
dimensional (3-D) visualization; chemical composition; indicator species ratios;
biogenic hydrocarbons; spatial variations in chemical composition; pollutant
fluxes; multivariate time series; and transport and mixing height statistics.
Model results were evaluated by comparing measured aloft air quality using
traditional model performance statistical and graphic products. These products
include peak performance statistics by level, location, and day; bias and error
statistics by level, location, and day; scatter plots of observations versus predic-
tions by level and location; quantile-quantile plots; vertical profiles of observed
