Skip to main content

Water level prediction by artificial neural network in a flashy transboundary river of Bangladesh

  • Authors (legacy)
    Corresponding: ROBIN K. BISWAS
    Co-authors: Biswas R.K. and Jayawardena A.W.
    Download PDF
  • gnest_01226_published.pdf
  • Paper ID
    gnest_01226
  • Paper status
    Published
  • Date paper accepted
  • Date paper online
Abstract

This paper presents the sensitivity analysis results of feed forward multilayer perceptron based Artificial Neural Network model for water level prediction in a data constraint transnational Surma River of Bangladesh. Catchment characteristics, hydro-geomorphological, meteorological and headwater information of the upper catchment area are not available to the authors. As such past daily total rainfall and water levels data available within the country are utilized in this study. Logistic sigmoid activation function with unit steepness parameter is exercised for non-linear transformations in both hidden and output layers. Synaptic weights are adjusted using modified delta rule through error back propagation algorithm. Batch mode of training is adopted for global error minimization. Finally, statistical indicators are used to evaluate the prediction performance of the neural network. The model is then applied to predict water levels with twenty four and forty eight hours lead time.

It is found that a single hidden layer with two hidden neurons are adequate to train the network. A higher number of hidden neurons is speeding up the training procedure, but with an unacceptable generalization for the application. The authors have successfully created a model that recognizes the intricate pattern of water levels, without having the spatially distributed geomorphic characteristics of the watershed and the time-series of the climatic factors.

 

Copy to clipboard
Cite this article
BISWAS, R.K., Biswas, R. and Jayawardena, A. (2014) “Water level prediction by artificial neural network in a flashy transboundary river of Bangladesh”, Global NEST Journal, 16(2). Available at: https://doi.org/10.30955/gnj.001226.