Browsing Teagasc funded research by Subject "Neural network model"

Now showing items 1-1 of 1

Assessment of factors affecting flood forecasting accuracy and reliability. Carpe Diem Centre for Water Resources Research : Deliverable 10.3

Bruen, Michael; Nasr, Ahmed Elssidig; Yang, Jianqing; Parmentier, Benoit (University College Dublin. Departmetn of Civil Engineering, 2012-07-02)

In Deliverable 10.1, a optimal methodology for combining precipitation information from raingauges, radar and NWP models (in this case HIRLAM) was described. It was based on an artificial neural network combination model, fitted to historic data, and operating on one-dimensional time-series of discharges. In this report, this new methodology is tested by applying it to (i) a rural catchment (Dargle)and (ii) a small urban catchment (CityWest). The results are compared with measured discharge series in both cases. Various measures of performance, applied to both the entire discharge series and also to the peaks-only are reported for various combinations of lead-time, spatial resolution and numbers of neurons in the hidden layer of the ANN model.