A stochastic-dynamic model to predict fecal coliforms at the mouth of the Anasco River

This research effort attempts to predict one year ahead the concentration of fecal coliforms at the mouth of the A (n) over tilde asco River, located in Puerto Rico. One of the most efficient techniques to represent stochastic processes is time series modeling. These models decompose the process into three major components: trend, seasonality and stochastic components. Unfortunately, time series models require observations at equal time intervals. Since the water quality data are not given at regular intervals, an adaptive estimation technique is proposed to estimate the missing values and, therefore, to generate an approximated time series at equal time intervals, to be able to study trend, seasonality, and stochastic components of fecal coliforms. Water quality data were collected from three water quality stations, located on the A (n) over tilde asco River. Historical data from 1973 to 2000 were used to model fecal coliforms at each of the water quality stations of the A (n) over tilde asco River. Time series models were identified at each station and were used to predict for one year the concentration of fecal coliforms at each station. A spatial interpolation algorithm was used to estimate the fecal coliforms at the mouth of the river.