The urban background dispersion model, BUM, used in the SIMAIR-system, is a simple trajectory model for evaluation of Air
Quality in urban areas on 1 x 1 km spatial resolution. The urban contribution to concentrations in a receptor point is calculated from
the emission sources in an upstream influence area whose size is dependent on the wind speed.
This simple and attractive concept enables fast model calculations and the model is applied for more than 100 Swedish towns within
SIMAIR. However, comparison with measured concentrations has shown that BUM underestimates levels of NO2 and NOX,
especially for towns in northern Sweden. The reason for this is probably meteorological, i.e. it exemplifies the difficulties in
describing the dispersion of air pollutants during strong stable atmospheric conditions.
This problem has previously been solved by a statistical method (regression analysis), to adjust the calculations against
measurements for towns in northern Sweden. The result of this method has varied widely; for some urban areas the result has been
good while the correlation between measured and calculated concentrations has been lower for others.
The aim of this study is, through a sensitivity analysis, to examine the parameters of the model that most significantly affect the
levels of NO2, and subsequently improve the parametrization of these during stable atmospheric conditions. Furthermore, the results
are validated against measurements from 13 urban areas in Sweden.
According to the sensitivity analysis, it is the parametrization of the vertical dispersion parameter σz that most affects the levels of
NO2. A new parametrization, which takes into account the stability, is introduced for urban areas outside major cities. This generally
raises the concentrations with several μgm-3 on annual basis and 10’s μgm-3 for 98-percentile daily mean concentration. Furthermore,
a correction of the meteorology (from Mesan) is introduced used in the calculations of BUM, for the meteorology to represent more
urban (rough) conditions.
The improvements of BUM lead to a better consistency between the model and the measurements. Generally, the correlation between
the calculated and the measured concentrations of NO2 increases, and the time variation of concentrations is better captured in the
model. Annual averages, and especially 98-percentile daily- and hourly mean value, are better reproduced in the improved version of
BUM; when compared to measured concentrations, 37 % of data points are within ± 50 % for the original BUM while the
corresponding results for the new BUM is 95 %. However, the new BUM model still doesn’t succeed, for all towns in northern
Sweden, to fully reproduce the highest daily and hourly peaks of concentrations.
In comparison with the original BUM climate corrected concentrations (in northern Sweden), the correlation between calculated and
measured concentrations is higher for the new BUM, especially in terms of annual average, correlation coefficient and coefficient of
variation.