A system of hydrodynamic, water quality and neural network models for predicting water quality in the rio de la plata estuary
Diego Norberto Bottelli, Sebastian H. Martijena, Sebastian Santisi
Friday 3 july 2015
8:45 - 9:00h
at Central America (level 0)
Themes: (T) Water resources and hydro informatics (WRHI), (ST) Management support systems and serious gaming
Parallel session: 14H. Water resources - Serious gaming
The Río de la Plata (RDLP) is a wide estuary formed by the confluence of the Paraná River delta and the Uruguay River (total average flow of 22,000 m3/s). This estuary is the main source of drinking water for 11 million residents of Buenos Aires city and its metropolitan. The RDLP coastal sector holds densely populated areas with tributaries running across them. This is associated with the generation of pollution plumes on the estuary which can originate, under certain specific hydrometeorological conditions, water quality deteriorating events that affect raw water at water intakes (5,000,000_m3/day). In order to provide management support, the implementation of a raw water quality forecasting system was decided. This paper describes the development and validation of that system, based on the integration of different models with real-time sensors and weather forecast. This work also includes performance analysis of different kinds of models (deterministic, stochastic and Artificial Neural Networks - ANN) to work as a system constituting a prediction tool, and description of the finally adopted models, which are i) 1D Hydrodynamic / Water Quality Model of the Paraná River and its Delta for predicting raw water turbidity, ii) a Back Propagation ANN model for the prediction of RDLP tides at water intakes and iii) Kohonen ANN model for the prediction of ammonium in raw water. The validation of this system of models is presented by the comparison of calculated and continuously measured Hydrodynamic and quality parameters for at least a two-year period. This raw water quality forecasting system has already been implemented and is at present running daily as a management support tool.
