A method has been developed to obtain a joint solution to the problem of optimum number and configuration of ambient air quality monitors, on the principles of spatial correlation analysis and the minimum spanning tree. The interest in this case is to represent the patterns of regional air quality, at a minimum of an overlap of information. This methodology is extended to account for the uncertainties in air quality simulations and also to incorporate the probabilities of occurrence. As an illustration to these methodologies, an example of Taipei City, Taiwan has been considered.