A three-way principal factor analysis for assessing the time variability of freshwaters related to a municipal water supply

Abstract

Chemical analyses (total hardness, HARD; dissolved oxygen, DO; chlorides; sulfates; nitrates; nitrites; ammonia; orthophosphates; and UV-absorbing organic constituents, UV-ORG), physical data (turbidity, TURB; temperature, TEMP; conductivity, COND), and biological monitors (total and faecal coliforms, FAEC; faecal streptococci, STREPTO) constitute the 15 parameters, monitored with monthly frequency in the space of 4 years on freshwaters sampled at seven sites in a karstic area of northeastern Italy. The data set was used for a three-way principal factor analysis aimed at exploring the pattern of information about the environmental quality of the monitored freshwaters, since four wells are feeding the municipal water supply of the Province of Trieste, and the other water courses can influence them. The selected three-way (3,3,2) model uses three components for describing the analytical parameters, three for temporal variations and two for spatial variations. The method optimising the ‘variance of squares’ of the core elements has permitted a simple and meaningful interpretation of the Tucker-3 solution. The procedure succeeded in decomposing the overall temporal variation in three parts, thus highlighting nonperiodic critical events, a periodic seasonal component and a constant term. The seasonality has been confirmed by the examination of the autocorrelation function of the second temporal component. An environmental interpretation and an estimate of the relative relevance of phenomena conditioning the considered water body, detected by the multiway analysis, have been proposed.