Halocarbons in a drinking water supply system: Forecasting contamination values and population exposure

The network, which supplies drinking water to the city of Milan, has a number of sampling points, where the concentrations of several contaminants are periodically measured. From the time series of total halocarbon (HC) concentration values (in the ppb range) autoregressive models can be obtained for each sampling point. To this end a suitable figure of merit is minimized, which depends on the backcast error and on the model order. The best model is then applied to predicting concentration one time step (one month or one year) forward. The experimental data pertain to the years 1981 to 1992. Predicted values are compared with the measured ones. Some features are consistent with independent information about the water supply system. The models are applied to estimate the population exposure to HC and its short term trend. All computer codes which implement these procedures are inserted into an IBM AIX - X11 R3 graphics interface. Results are shown by means of bar charts and contour maps, where the average predicted values and their confidence intervals are plotted. The method and the procedures may be used by public authorities to manage drinking water resources, whenever more detailed models give rise to non observable systems.