Author(s)

E. Jach-Szakiel, J. Skrzypski & W. Kaminski

Abstract

The aim of this work was to optimize a model which gives a possibility of evaluating the localization of EMEP monitoring stations. To estimate if the location of air monitoring stations was correct, a single-layer artificial neural network (ANN) was used. In the procedure of unsupervised network training, a standard competitive training based on the k-means algorithm with a modification called the "neural gas" was used. The goal of modeling was to indicate the best locations of the hypothetical monitoring stations. The parameters, varying during model optimization, covered the value of subareas, the value of iterations and the value of hypothetical stations. As a result, an optimal model, useful for the EMEP monitoring network evaluation, was derived. Keywords: artificial neural network, air pollution, deposition fields. 1 Introduction Characterization of the air pollution deposition field in a given area is possible only when the number of monitoring stations is sufficient and its localization reflects precisely the real image of the spatial diversification of concentrations. The number of air monitoring stations and their location has a great impact on the applicability of mathematical procedures, e.g. interpolation or approximation in the estimation of spatial distribution of air pollutants. The description of an air pollution concentration field is required to estimate aerosanitary hazards and it is important for the environmental management system (by offering objective presumptions for making administrative decisions, e.g. concerning localization).

Keywords

artificial neural network, air pollution, deposition fields.