Model assessment and verification - Air Pollution Modeling and Its Application - page 357

The PM 10 dispersion modeling used CALPUFF (version 6.263) forced with
CALMET fields; both domains were 120 by 120 km on a 1.2 km grid centered on
the city. Area sources were used to represent home heating and vehicle emissions,
and industrial stacks were included as point sources. The 'slug' rather than 'puff'
dispersion option was used since output concentrations were required from sites
within area emission sources. The input hourly diurnal pattern for sources was
typical for a cold winters' day [4], therefore, whilst different zones of the city had
different diurnal patterns and individual stacks had hourly varying patterns, an
identical pattern was repeated every day of the CALPUFF run. The model run was
for May to the end of August 2006.
The main air quality reporting site for the city, and one that typically experiences
the greatest number of days in excess of 50 μg/m 3 , is in the residential suburb of
St Albans (Fig. 1) . A second site is located at Woolston, near to an industrial area.
PM 10 is measured continuously with a TEOM FDMS; St Albans has temporal data
coverage of greater than 98% and Woolston of over 90%. These two sites provided
the verification data for the CALPUFF modeling results.
3. Results and Conclusions
The CALPUFF modeled concentrations have good qualitative agreement with
observations on both an hourly and 24-hourly basis . Fig. 2. shows a period of 2
weeks in June 2006, the model successfully simulating the diurnal variability, with
concentrations peaking strongly in the late evening. The period encompasses a
range of dispersion scenarios, including low wind speeds and dawn frosts for the
first 6 days, followed by snow during the early hours of the 12 June, a frost on the
14 June and milder, breezier conditions in the end of the period. The false modeled
peak in concentrations around midnight on 12 June coincides with what had been
a mild evening, the mean daily temperature on the 11 June was 13°C and potentially
the emissions from home heating would be low. The use of a fixed diurnal emission
pattern representing a typical cold winters' day is likely to lead to a bias in
overestimating concentrations.
The relationship between modeled and measured 24-h averaged concentrations
was investigated for all days of 8°C or lower May to August 2005 and 2006 ( Fig. 3.
These lower temperature days were chosen to reflect that the emission inventory
focuses on a cold winters' day. The Spearman rank correlation coefficient was
0.80 for St Albans and 0.63 for Woolston. The model tended to predict higher
concentrations than were measured. The long term average measured concentration at
St Albans for June and July 2005 and 2006 was 49 μg/m 3 and the modeled 53
μg/m 3 , at Woolston the measured average was 41 μg/m 3 and the modeled 37 μg/m 3 .