Environmental Engineering Reference
In-Depth Information
The growing computer capacity enabled model evaluation studies over longer
periods. Roemer et al. (2003) performed a study where ten different photochemical
CTM's participated in modelling the summer of 1997. The results gave a
correlation coefficient averaged over 30 selected stations between modelled and
observed daily maximum ozone of about 0.75. Hass et al. (2003) carried out a
study with six aerosol CTM's, focussing on the year 1995. It was shown that the
models calculate SO
4
, NO
3
and NH
4
within a range of a factor 2 from the
observations.
In a further study, van Loon et al. (2004) with seven CTM's, determined the
model performance over 1999 and 2001 for O
3
, NO, NO
2
and SO
2
, the aerosol
components SO
4
, NO
3
, NH
4
and total PM10 and wet deposition fluxes. In general,
similar conclusions were found as in the previous studies. The models are capable
of modelling O
3
well, and do show some skill in modelling the secondary inorganic
aerosol components. However, the models do severely underestimate the observed
PM10 concentrations.
In a recent study five European CTM's were tested for a period of about 80
days in February-April, 2003, with elevated PM10 and PM2.5 observations.,
Stern et al., 2008. It was again shown that the current modelling systems are
unable to simulate well higher PM10 and PM2.5 levels.
It should be noted that these European model evaluation studies are focussed
on operational models, and use mostly monitoring data for the comparison with
calculated concentrations.
3. Ensemble Approaches as Aspect of Model Evaluation
The concept of the ensemble approach is based on the situation that different CTM's
give different results, without being able to determine which model is better, or
even why a model would be better. Because models contain different para-
meterisations, which are all equally valid, models give different results.
The ensemble approach in weather forecasting has already a long tradition,
although in weather forecasts often only one model is used, but with uncertainties
in the parameterisations. The ensemble forecast of ECMWF is based on 50
members. The ensemble approach with CTM's is based on using different CTM's,
and different meteorological input data. In general, the emission data are identical
for the different models.
As stated, the first example of an ensemble approach can be found in Delle
Monache and Stull (2003).
McKeen et al. (2005) used seven CTM's for forecasting ozone. It was shown,
as in the study by Delle Monache and Stull, that the ensemble based on the mean
of the participating models has a significantly better temporal correlation to
the observed daily maximum 1-h average and maximum 8-h average than any
individual model. Similar results are found by van Loon et al. (2007). Vautard
et al. (2008) analysed the ensemble of seven models for O
3
, NO
2
and secondary
