A platform for research: civil engineering, architecture and urbanism
A platform for research: civil engineering, architecture and urbanism
Flood Forecasting Using Simple and Ensemble Artificial Neural Networks
With the increased flood havoc in many river basins worldwide, flood forecasting has been recognized as one of the feasible nonstructural measures of flood management. For accurate and reliable extreme flood forecasts, different artificial neural networks (ANNs)-based modelling approaches are developed in this study. For daily streamflow forecasting, a non-clustered feedforward backpropagation ANN (NCANN) model is developed using the daily observed streamflow training dataset. Further, in order to forecast high flow with ANN, two different types of models are developed, namely pre-classified ANN and post-processed ANN model. The pre-classified ANN models are basically the cluster-based ANN (CANN), and the post-processed ANN models are the ensemble ANNs (EANN). The high flow forecasting efficacy of the ANN models is compared for a common set of high flow regimes at one-, two- and three-day lead times in the flood-prone Mahanadi River basin in Eastern India. The results reveal that consideration of the training dataset consisting of various flow stratifications does not ensure a better forecasting of the high flow regime by NCANN model; rather the use of homogeneous set of training dataset gives improved forecasting during testing. Moreover, the high flow forecasting error is further reduced when the model ensembles are used. Seasonal flow assessment can be improved by using these developed models equally effectively.
Flood Forecasting Using Simple and Ensemble Artificial Neural Networks
With the increased flood havoc in many river basins worldwide, flood forecasting has been recognized as one of the feasible nonstructural measures of flood management. For accurate and reliable extreme flood forecasts, different artificial neural networks (ANNs)-based modelling approaches are developed in this study. For daily streamflow forecasting, a non-clustered feedforward backpropagation ANN (NCANN) model is developed using the daily observed streamflow training dataset. Further, in order to forecast high flow with ANN, two different types of models are developed, namely pre-classified ANN and post-processed ANN model. The pre-classified ANN models are basically the cluster-based ANN (CANN), and the post-processed ANN models are the ensemble ANNs (EANN). The high flow forecasting efficacy of the ANN models is compared for a common set of high flow regimes at one-, two- and three-day lead times in the flood-prone Mahanadi River basin in Eastern India. The results reveal that consideration of the training dataset consisting of various flow stratifications does not ensure a better forecasting of the high flow regime by NCANN model; rather the use of homogeneous set of training dataset gives improved forecasting during testing. Moreover, the high flow forecasting error is further reduced when the model ensembles are used. Seasonal flow assessment can be improved by using these developed models equally effectively.
Flood Forecasting Using Simple and Ensemble Artificial Neural Networks
Water Sci.,Technol.Library
Pandey, Ashish (editor) / Chowdary, V. M. (editor) / Behera, Mukunda Dev (editor) / Singh, V. P. (editor) / Sahoo, Bhabagrahi (author) / Nanda, Trushnamayee (author) / Chatterjee, Chandranath (author)
Geospatial Technologies for Land and Water Resources Management ; Chapter: 24 ; 429-456
2021-12-07
28 pages
Article/Chapter (Book)
Electronic Resource
English
Real-Time Flood Forecasting Using Neural Networks
| British Library Online Contents | 1998
Real-Time Flood Forecasting Using Neural Networks
| Online Contents | 1998
Real‐Time Flood Forecasting Using Neural Networks
| Wiley | 1998
Improving flood forecasting in Bangladesh using an artificial neural network
| British Library Online Contents | 2010