A predictive model of the impact of urbanization on bacterial loads in watersheds

Bacterial concentration is one of the most important aspects of water quality. Many regions in the world are affected by increasing urbanization and a potential increase in bacterial concentrations in waters. We used long-term data from 68 stations in eight watersheds in Eastern Thailand to quantify the temporal and geographical variation in total and fecal coliform bacteria. Descriptive statistics showed considerable seasonal, inter-annual, and geographical variation. In order to quantify this multi-level variation, we built a predictive model of bacterial loads. Using fixed- and mixed-effects regression models, we built a model including the effects of urbanization and other significant variables. The best model, fitted by restricted maximum likelihood, included the effects of season, year, urbanization as fixed effects, and of watershed and station as nested, random effects. Temporal variation was related to seasonal and annual variations. Spatial variation had a very significant impact on the bacterial concentrations. Urbanization was an important factor controlling concentrations of bacteria in rivers: we found that the proportion of urban area around a station had a statistically significant effect on log-transformed total coliform bacterial concentration with a slope equal to 1.3 (SE = 0.3), and on log-transformed fecal coliform bacterial concentration with a slope equal to 1.4 (SE = 0.3). Our model predicts that bacterial concentrations would be multiplied by 20 if land is transformed from non-urban to fully urban.