Evaluating and Updating the Sample Design in Repeated Environmental Surveys: Monitoring Air Quality in Padua

In many repeated environmental surveys, data are collected in space without a rigorous statistical design, based only on practical circumstances such as social importance of the site, availability of space, or nearness to main roads. Furthermore, the sample size is usually fixed because of financial constraints. Given these conditions, two issues are of interest: evaluating the performance of the current design and building a decision support system to assist any future modification of it. The case examined here is a survey on air pollutants run by the Padua City Council from 1991-1994. We propose a methodology to identify a network of monitoring stations that are optimally distributed in the study area so as to minimize the prediction variance of the population mean. The performance of the current design can then be compared with the optimal design. An optimal location criterion is used to improve the precision of sample estimates where one (or more) monitoring stations are added, deleted, or relocated.