Current practice in the control of urban air pollution concentrates on reducing emissions from both fixed and mobile sources directly, without consideration of the fact that (i) zonal restrictions and

relocalization of activities -both directly polluting activities such as industrial sources and transport generating activities- affect emissions, and (ii) that the direct control of emissions impacts the costs of emitting sources, and this in turn affects location decisions. Thus the application of regulations that do not consider these effects can result in unexpected changes in emissions in different parts of the city, in response to underlying economic forces. Modeling this relationship is a difficult task because it involves relating the spatial distribution of sources that are responsive to economic conditions, with their emission and abatement cost characteristics. In response to this challenge, this research project presents a methodological proposition to determine an urban location/emission pattern which can be used to predict the net cost and emission result of controls on location of activities and unit emissions.

Specifically, implementing a self organizing neural network is proposed in order to model the relationship between local urban configuration parameters with the local characteristics of emissions and abatement costs. This model has been implemented for the 34 districts of Santiago and allows estimating changes in emissions of two pollutants (PM-10 and NOx) resulting from relocation of activities, zonal restrictions and/or emission control regulations. Results show that even minor housing developments can change the emission pattern of the whole city. Thus, for long term environmental planning it is necessary to evaluate the impacts on emission of policies affecting location decisions.