visual range data have been developed elsewhere .9 We
developed a similar approach to estimate daily PM10 concentrations
with the available data for Bangkok. The estimation
algorithm was developed by estimating the relationship
between PM10 measured at the MOSTE monitoring
locations and visual range on days when measured
PM10 data were available (n = 581). Dewpoint temperature
and visual range are the variables included in the
algorithm, which was then used to estimate PM10 concentrations
on all days for which visual range and dewpoint
temperature were available (n = 1367).
To assess the reasonableness of the estimated PM10
concentrations, we first compare them to the measured
PM10 concentrations on the days for which monitor data
are available. The mean values are very close (i.e., 65.1
mg/m3 for the measured PM10 versus 64.4 mg/m3 for the
estimated PM10), but the estimated PM10 measure somewhat
overpredicts the low concentrations and
underpredicts the high concentrations. This suggests that
the predictive algorithm does better at the middle visual
range levels than on very clean or very polluted days. The
simple correlation between measured and estimated PM10
is 0.67 (p < 0.0001). Examination of the plots of estimated
daily PM10 concentrations overlaid on the measured PM10
concentrations suggests that the estimation technique was
able to capture a pattern of day-to-day and seasonal variations
in daily PM10 concentrations.