Chapter 9 Cuckoo Search Optimisation Neural Networks for Runoff Simulation in a Equatorial Rural Watershed

Rainfall–runoff relationship is one of the most complex hydrological phenomena. A conventional neural network (NN) with backpropagation algorithm has successfully modelled various non-linear hydrological processes in recent years. However, the convergence rate of the backpropagation NN is relatively slow, and solutions may trap at local minima. Therefore, a new metaheuristic algorithm named as cuckoo search optimisation was proposed to combine with the NN to model the daily rainfall–runoff relationship at Sungai Bedup Basin, Sarawak, Malaysia. Two-year rainfall–runoff data from 1997 to 1998 had been used for model training, while one-year data in 1999 was used for model validation. Input data used are current rainfall, antecedent rainfall and antecedent runoff, while the targeted output is current runoff. This novel NN model is evaluated with the coefficient of correlation ( R ) and the Nash–Sutcliffe coefficient ( E 2 ). Results show that cuckoo search optimisation neural network (CSONN) is able to yield R and E 2 to 0.99 and 0.94, respectively, for model validation with the optimal configuration of number of nests ( n ) = 20, initial discovery rate of alien eggs ( <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink" altimg="E9-001.gif" display="inline"> <mml:mrow> <mml:msub> <mml:mi>p</mml:mi> <mml:mrow> <mml:msub> <mml:mi mathvariant="normal">a</mml:mi> <mml:mrow> <mml:mtext>initial</mml:mtext> </mml:mrow> </mml:msub> </mml:mrow> </mml:msub> </mml:mrow> </mml:math>) = 0.6, hidden neuron (HN) = 100, iteration number (IN) = 1,000 and learning rate (LR) = 1 for CSONND4 model. The results revealed that the newly developed CSONN is able to simulate runoff accurately using only precipitation and runoff data.