Volume 13 Issue 4 May - July 2018
Review Paper
Suitability of
Neural Network Techniques for Rainfall-Runoff Modelling Over a River Basin: A
Comprehensive Literature Review
Sanjeev Karmakar*, Pradeep Kumar Mishra **,
Shreerup Goswami ***
* Bhilai Institute
of Technology, Durg, Chhattisgarh, India
** Research Scholar, Bhilai Institute of Technology, Durg,
Chhattisgarh, India
*** Professor of Geology, Department of Earth Sciences,
Sambalput University Burla, Odisha, India..
Karmakar, S.,
Mishra, P.K., & Goswami, S. (2018). Suitability of Neural Network
Techniques for Rainfall-Runoff Modelling Over a River Basin: A Comprehensive
Literature Review. i-manager’s Journal
on Future Engineering and Technology, 13(4), 68-83.https://doi.org/10.26634/jfet.13.4.14434
Abstract
Rainfall-Runoff
(R-R) modelling over a river basin is a complex and challenging task for a
hydrological scientist. It is almost complicated due to chaos behaviour of R-R
data time series. However, since 1972 various techniques are suggested by the
world hydrological scientists and shown some extent of success. In this study,
a comprehensive review of the models of R-R modelling over river basin from
1972 to 2016 was performed, wherein models of various contribute is studied
year-wise. As an outcome, support vector machine and soft-computing, i.e.,
neural network, fuzzy logic, and genetic algorithm techniques based on
statistics have been found to be successfully applied. In numerical modelling,
two different architectures of neural network such as BPN and RBF were found
more suitable. The BPN was better evaluated over RBF architecture as far as
performance and complexity of implementation is concerned. Finally, it is
concluded that BPN is sufficient enough to resolve this complex problem since
it has shown 80% certainty in prediction of R-R over different river basins.
However, obtaining optimum architecture for better performance is mandatory.
These evidences are broadly discussed in this review article.
No comments:
Post a Comment