ANNs to forecast waterflow, soluble reactive phosphorus and total phosphorus at the lower end of a river stretch

These ANN models are instruments capable to forecast flows and concentrations with an anticipation of up to 15h, at resolutions up to an hour, at the downstream end of a river stretch, based on the values measured at the upstream end of the stretch (together with data from tributaries for the concentrations). These instruments are based on artificial neural networks (ANN) and comprises separate prediction modules (1) for flows and (2) for concentrations; and (3) for flows and concentrations. The ANNs have been tested employing field data collected on the river Swale, UK, during 1993 to 2000, in two different type of occasions (a) long term usual water flow monitoring at 15 minutes resolution and (b) intensive monitoring campaigns for water flow and phosphorus species concentrations at up to three hours resolution.

The ANNs development methodology and results are described in the paper:
Horia Hangan, Elisabeta Cristina Timis, Vasile Mircea Cristea, Michael George Hutchins, 2023. Improved artificial neural network models of river stretches extending the forecasting horizon across wide ranges of water flow and phosphorus concentration, submitted to Journal of Hydrology.