Abstract. We evaluate four high-resolution model simulations of pollutant
emissions, chemical transformation, and downwind transport for the Athabasca
oil sands using the Global Environmental Multiscale – Modelling Air-quality
and Chemistry (GEM-MACH) model, and compare model results with surface
monitoring network and aircraft observations of multiple pollutants, for
simulations spanning a time period corresponding to an aircraft measurement
campaign in the summer of 2013. We have focussed here on the impact of
different representations of the model's aerosol size distribution and
plume-rise parameterization on model results. The use of a more finely resolved representation of the aerosol size
distribution was found to have a significant impact on model performance,
reducing the magnitude of the original surface PM2.5 negative biases
32 %, from −2.62 to −1.72 µg m−3. We compared model predictions of SO2, NO2, and speciated
particulate matter concentrations from simulations employing the commonly
used Briggs (1984) plume-rise algorithms to redistribute emissions from large
stacks, with stack plume observations. As in our companion paper (Gordon et
al., 2017), we found that Briggs algorithms
based on estimates of atmospheric stability at the stack height resulted in
under-predictions of plume rise, with 116 out of 176 test cases falling below
the model : observation 1 : 2 line, 59 cases falling within a factor of 2
of the observed plume heights, and an average model plume height of 289 m
compared to an average observed plume height of 822 m. We used a
high-resolution meteorological model to confirm the presence of significant
horizontal heterogeneity in the local meteorological conditions driving plume
rise. Using these simulated meteorological conditions at the stack locations,
we found that a layered buoyancy approach for estimating plume rise in stable
to neutral atmospheres, coupled with the assumption of free rise in
convectively unstable atmospheres, resulted in much better model performance
relative to observations (124 out of 176 cases falling within a factor of 2
of the observed plume height, with 69 of these cases above and 55 of these
cases below the 1 : 1 line and within a factor of 2 of observed values).
This is in contrast to our companion paper, wherein this layered approach
(driven by meteorological observations not co-located with the stacks) showed
a relatively modest impact on predicted plume heights. Persistent issues with
over-fumigation of plumes in the model were linked to a more rapid decrease
in simulated temperature with increasing height than was observed. This in
turn may have led to overestimates of near-surface diffusivity, resulting in
excessive fumigation.
The Dependence of Atmospheric Extinction on Meteorological Conditions and Aerosol Size Distribution
