Optimising statistical models to predict faecal pollution in coastal areas based on geographic and meteorological parameters

Abstract

This article describes a methodology for optimising predictive models for concentrations of faecal indicator organisms (FIOs) in coastal areas based on geographic and meteorological characteristics of upstream catchments. Concentrations of FIOs in mussels and water sampled from 50 sites in the south of Brazil from 2012 to 2013 were used to develop models to separately predict the spatial and temporal variations of FIOs. The geographical parameters used in predictive models for the spatial variation of FIOs were human population, urban area, percentage of impervious cover and total catchment area. The R² of models representing catchments located within 3.1 km from the monitoring points was up to 150% higher than that for the nearest catchment. The temporal variation of FIOs was modelled considering the combined effect of meteorological parameters and different time windows. The explained variance in models based on rainfall and solar radiation increased up to 155% and 160%, respectively.

Keywords: Coliforms; Escherichia coli; Faecal pollution; Florianópolis; Perna perna; Regression models.