MULTIPLE-SITE RECEPTOR MODELING WITH A MINIMAL SPANNING TREE COMBINEDWITH A NEURAL-NETWORK

A combination of two pattern recognition methods has been developed th at allows the generation of geographical emission maps from multivaria te environmental data. In such a projection into a visually interpreta ble subspace by a Kohonen self-organizing feature map, the topology of the higher dimensional variables space can be preserved, but parts of the information about the correct neighborhood among the sample vecto rs will be lost. This loss can partly be compensated for by an additio nal projection of Prim's minimal spanning tree into the trained neural network. This new environmental receptor modeling technique has been adapted for multiple sampling sites. The behavior of the method has be en studied using simulated data. Subsequently, the method has been app lied to mapping data sets from the Southern California Air Quality Stu dy (SCAQS). The projection of 17 chemical variables measured at up to eight sampling sites provided a two-dimensional, visually interpretabl e, geographically reasonable arrangement of air pollution sources in t he South Coast Air Basin.