Ground-level ozone concentration is a key indicator of air quality, which varies in space and time. A statistical method for modeling ground-level ozone concentration in a given region is undertaken in this paper. In addition, an environmental network design problem, is also explored. The ozone concentrations in the Pittsburgh region of Pennsylvania, United States are considered for demonstration purpose. In this region, there are 25 stations to collect the hourly ozone concentration data. All but one station have missing observations. This region is covered by grid boxes with a spatial resolution of latitude 0.2° × longitude 0.2°. By applying hierarchical Bayesian spatio-temporal modeling (see Le and Zidek (2006) for details), a conditional predictive distribution over these grid points can be obtained. In terms of an entropy criterion, the environmental network design problem is solved using the obtained predictive distributions. Model evaluation is also provided.