Six sampling points located at different distances, two by two (near-far), from the three main supply reservoirs, were selected in the distribution system of the drinking water network of the city of Salamanca in order to follow the evolution of trihalomethanes (THMs). Data obtained were correlated statistically with the chlorination dosages in the treatment plants and the distribution system, the distances run by the water, residual free chlorine and total chlorine, the total organic carbon (TOC), pH and temperature. The prechlorination dosage did not correlate strongly (P = 0.1268) with chloroform content at the first point of the distribution system, while that of postchlorination and rechlorination did not correlate at all. The organic matter, measured as the TOC, showed a positive correlation (P = 0.0322), with a consumption of free chlorine (R = -0.327, P = 0.0190) and total chlorine (R = -0.290, P = 0.0393) that was proportional to the content in THMs. However, the distance among the six points of the distribution system-an approximate means of estimating the reaction time-yielded conflicting results. Temperature and pH proved to be the parameters with the strongest influence. Temperature alone did not show linear dependence, while pH did (R = 0.665, P = 0.0001), showing a dramatic increase in THM levels on two sampling dates with exceptionally high pH levels. Functions of these two parameters permitted the elaboration of a predictive mathematical model (R = 0.995, P < 0.0001) in which increasing values of pH and temperature are seen to increase the level of chloroform up to a given temperature value (T-c = 17.30 degrees C) after which a sharp decrease in the chloroform content occurs. This value was comparable with that obtained with a similar model in the finished water treatment plants. A global modeling of both systems (treatment plants and distribution system) was also possible (R = 0.827, P = 0.0001), with a T-c value of 18.74 degrees C for each pH. (C) 1997 Elsevier Science Ltd.