Title: Regenerated inflow forecast in a data-sparse catchment

Authors: Oseni Taiwo Amoo; Ikudayisi Akinola

Addresses: Risk and Vulnerability Science Centre, Walter Sisulu University, Mthatha Campus, Eastern Cape, South Africa ' Civil Engineering Department, Walter Sisulu University, Buffalo Campus, Eastern Cape, EC, South Africa

Abstract: The lack of hydrological data because of a progressive decline in the ground-based observation network and pervasive ungauged basins has made an adequate assessment of water resources more challenging. Hence, this study evolved best-fitted rainfall time series model with ARIMA, moving average, simple, and double exponential with stochastic models of log Gumbel, Normal, and Pearson fitted on the available streamflow data for best monthly flow. Regionalised flow duration curves (FDCs) was developed to model future years' inflow forecast into the Hazelmere Dam was tested using back propagation neural network, general regression neural network, and multiple linear inflow regression models for the best fitted trained algorithm to forecast future inflows into the Dam. The results of the double exponential model show a more accurate past rainfall prediction compared to others while stochastic fitted FDCs and the validated ANNs model can be used as a viable environmental data transformation stratagem in a data-sparse catchment.

Keywords: inflow-forecast; data-sparse catchment; ungauged basins; neural networks; flow duration curves; sustainable water management.

DOI: 10.1504/IJHST.2025.144256

International Journal of Hydrology Science and Technology, 2025 Vol.19 No.2, pp.187 - 211

Received: 05 Nov 2022
Accepted: 04 Oct 2023

Published online: 03 Feb 2025 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article