Environmental Engineering Reference
In-Depth Information
sta-tistical interpolation methods are applied to annual monitoring (AirBase) and
modeling (EMEP) data to produce individual yearly maps of AOT40 and annual
mean SO
2
for the years 1996-2005. These maps are then assessed to determine
both the spatial distribution and the uncertainty of the 10 year trend. A number of
sensitivity tests are carried out. This paper provides a summary of a more extensive
report from Denby at al. (2008a).
2. Methodology
To spatially map trends in air quality a number of steps are needed. For each year,
from 1996 to 2005, maps of AOT40 and annual mean SO
2
are produced at a
spatial resolution of 25 km using statistical interpolation methods that combine
both monitoring (AirBase, 2008) and modeling (EMEP) data (Solberg et al., 2009).
Residual log-normal kriging, after multiple linear regression, is the statistical inter-
polation method employed (Denby et al., 2008b; Horálek et al., 2007).
For each grid point in the maps trends are calculated, producing maps of air
quality trends. Both linear regression and Sen's method (Sen, 1968) are applied.
For each year and at each grid point an uncertainty is determined using the
residual kriging variance. Monte Carlo methods are used to determine the uncertainty
in the trend by randomly sampling the kriging variance. For each random sample
set a trend is determined and the standard deviation of the ensemble of resulting
trends is used to indicate the uncertainty in the mean trend.
A range of sensitivity studies are carried out concerning the influence of station
selection, trend assessment methodology and interpolation methodology. Details
concerning these sensitivity studies can be found in Denby et al. (2008a).
3. Modeled and Observed Trends
In
Fig. 1
the modeled (EMEP) and observed trends in the station averaged AOT40
and annual mean SO
2
concentrations are shown. From this it can be seen that the
modeled and observed trends for AOT40 are opposite in sign, whilst the modeled
and observed trends for SO
2
are of the same sign but the modeled con-centrations,
and trends, are around 40% less in magnitude than those observed. These
differences make the use of only models, to study trends, inappropriate. By
applying the statistical interpolation method for each year of data the general
observed trends will be followed and the spatial distribution will be represented.
