Re-evaluation of the Lagrangian particle modelling system on an experimental campaign in complex terrain

Slovenian legislation for industrial air pollution control requires efficient modelling systems for small domains over complex topography. To determine the performance and efficiency of the Lagrangian particle modelling system used for this purpose a study was made where a general purpose modelling system designed for local-scale areas was used. The main goal of the study was to evaluate a modelling system of this kind using an operational configuration of bot h input data and model parameters, choosing a testing period with very complex dispersion conditions. This severe check could help to better understand the general quality that a model call achieve in these conditions giving some idea on flow to better evaluate and use some results that seem to be very negative simply looking at some statistical parameter. Data from a three-week experimental campaign performed around the Sostanj thermal power plant during the spring of 1991 was used (analyzed) for evaluation. The database covers very high ambient concentrations (due to the absence of desulphurisation plants) over complex terrain. The simulation was performed for the full duration of the campaign and a particular situation during the 1st and the 2nd of April 1991 was used as an example to outline the model behaviour in complex conditions. During this selected sub-period measurements revealed that; (measured) wind speeds were very low, wind changed course in all directions rapidly and consequently the plume spread in all directions. A comparison between measured and reconstructed SO2 concentrations was made at the positions of several automatic air quality measuring stations located around the thermal power plant. Standard statistical indexes to evaluate model performances are instead computed at the same positions for the entire period of the experimental campaign. Overall the reconstructed SO2 concentrations were underestimated relative to the measured ones, but all direct air pollution events were reconstructed. Some weaknesses of the model in the correct reconstruction of peak events are explained and a way to better describe them and to enhance statistical indexes is proposed.