6 Conclusions

The new three dimensional prognostic model Metphomod for the simulation of summer smog was developed and successfully applied to analyze a summer smog episode in Switzerland. Comparisons with measurements show that the model is capable to reproduce the real situation well.

While there are many other similar models used all over the world, Metphomod 's specific strengths are the direct coupling of chemical reactions with the meteorological computations, the capability of handling very rough topography without the requirement for smoothing, and the easiness of application. Direct coupling of the meteorological and chemical modules is supposed to improve the model accuracy in very complex terrain, where winds and turbulence change in a rapid and non-linear way. Although one might obtain similar results with a decoupled model, the storage area needed to save intermediate meteorological output at short time steps to be used as input to a chemical dispersion model would quickly use up available hard disk space on any computer facility, and therefore time slices in the order of 30 minutes to 1 hour are typically used to run decoupled photochemistry models.

Numeric models are never perfect and such is Metphomod . Many improvements could be made, mainly:

1.

A more accurate modeling of a thin layer above the local surface would allow more appropriate comparisons of model output to measurements which typically are made at a height of 2 m above ground. In the current state considerable uncertainty of model performance still exists, mainly due to the fact that measurements of chemical species concentrations made at about 2 m above ground can be completely unrepresentative of the lowest grid cell in the model which has a characteristic scale of 1000 m in the horizontal dimension and to a few tens of meters in the vertical. E.g. NO ₂ concentrations in reality are strongly affected even by minor traffic axes in the distance of 100-200 m from the measuring location, where at the spatial scale of a mesoscale model like Metphomod , such emissions are either absent or averaged over the surface area of a whole grid cell, which produces a significant dilution.

2.

Heterogeneous chemistry is completely absent in Metphomod . This is not specific to Metphomod , but the inclusion of heterogenous processes is a subject which attracts much attention at the moment. Modules for heterogenous and aerosol chemistry, will be intcluded in Metphomod in the near future.

3.

Eulerian prognostic models such as Metphomod use lots of input data. Synoptic meteorological data and chemical background concentrations are normally not readily available and therefore must be roughly estimated, which leads to high uncertainties in the predicted concentrations. Often the spatial and temporal resolution of the emission inventories is far from satisfactory, which leads to further errors and uncertainty. Important inconsistencies arise where emission inventories can only supply an average diurnal cycle of species like NO, while we attempt to simulate a non-average, specific day where we know the meteorology of that day, but do not know the real emissions for the same day. Our suggestion is that in a future version of Metphomod as many emission estimates as possible must be internalized, e.g. biogenic emissions of VOC computed using actual meteorological conditions at the canopy level.

Despite these suggested improvements we showed that Metphomod in its current state is able to model test cases that are generally used to validate a mesoscale model and two specific days with photosmog conditions observed during the Pollumet 1993 field campaign in Switzerland. Qualitative agreement between available surface measurements and model output is good. Therefore we suggest that Metphomod is a valid tool to compute scenarios to support the development of ozone abatement strategies which could be used by policy makers in their decision finding.