PM10 particles are one of the main air pollutants with many negative impacts
on human health, the environment and the climate. As part of the thesis, we
have searched for several different chemistry-transport models that are used
to predict PM10 in different locations across Europe. We have evaluated how
the predicted values compare with measured values in Slovenia. We wanted
to evaluate which model predicts the values closest to the measured values,
or which model predicts exceedances of the limit value most accurately. We
were also interested in which model best predicts sudden large increases or
decreases in concentrations. We used a variety of metrics. We were also
interested in whether certain meteorological conditions affect the accuracy of
the predictions, so we modelled model errors using meteorological parameters
as attributes for machine learning models.
|
