Abstract:Accurate surface runoff prediction is vital for water resources engineers for various applications. Advances in the artificial intelligence techniques can act as robust tools for modelling hydrological processes. The present study focuses on testing the reliability of different data sources and choosing the correct source to model the rainfall-runoff process under data scarce situations using AI techniques. In this study, an absolute homogeneity test was performed for TRMM, gridded and observed precipitation data and found that the observed precipitation dataset is homogeneous and best suitable for modelling rainfall-runoff process in Kallada river basin, Kerala. Emotional artificial neural network (EANN) is a novel hybrid neural network and it is suggested in the present study for accurate monthly surface runoff prediction. This study was also conceived to address and investigate the efficiency of EANN for forecasting monthly surface runoff and compare the performances with conventional feed forward neural network (FFNN) and multivariate adaptive regression spline (MARS) models. Suitable goodness-of-fit criteria such as Nash–Sutcliffe efficiency (NSE), root mean square error (RMSE) and coefficient of determination (R2) and graphical indicators are used for assessing the efficacy of the developed models. The results showed that the EANN model performs better with R2 = 0.80 for the training phase and R2 = 0.77 for validation phase compared to other models. The improvement in the performance of EANN model over FFNN model is 12% and 5.8% for coefficient of determination in the training and validation phase, respectively. Further, the Taylor diagram indicates that there is a close match between the observed and EANN model predicted values in terms of statistical parameters. Overall, this study demonstrated the effectiveness of EANN in modelling the rainfall-runoff process and also could be a useful technique in other fields of water resources engineering.<h3 class="a-plus-plus">Highlights</h3><ul class="a-plus-plus"><li class="a-plus-plus">The present study focuses on testing the reliability of different data sources such as gridded, observed and TRMM precipitation datasets and choosing the correct source to model the rainfall-runoff process using AI techniques.</li><li class="a-plus-plus">From the selected homogeneous dataset and the observed runoff data, potential predictors were identified based on correlation analysis and partial autocorrelation function (PACF).</li><li class="a-plus-plus">The monthly runoff prediction models were developed using three AI techniques namely FFNN, MARS and EANN in a tropical river basin (Kallada) of Kerala with scarce amount of data.</li><li class="a-plus-plus">The performance of the developed models were assessed using statistical indicators (NSE, RMSE, and R2) and graphical indicators (Taylor diagram, REC plots and Random walk test).</li></ul>

Application of neural network and adaptive neuro-fuzzy inference systems for river flow prediction

Flood Forecasting Using Artificial Neural Network

Simulation of river stage using artificial neural network and MIKE 11 hydrodynamic model

Streamflow forecasting by modeling the rainfall–streamflow relationship using artificial neural networks

Performance evaluation of artificial intelligence paradigms—artificial neural networks, fuzzy logic, and adaptive neuro-fuzzy inference system for flood prediction

Performance assessment of artificial neural networks and support vector regression models for stream flow predictions

Comprehensive performance analysis of training functions in flow prediction model using artificial neural network

Discharge Forecasting By Applying Artificial Neural Networks At The Jinsha River Basin, China

Rainfall-runoff modeling for the Hoshangabad Basin of Narmada River using artificial neural network

Development of Machine Learning Flood Model Using Artificial Neural Network (ANN) at Var River

Potential of hybrid wavelet-coupled data-driven-based algorithms for daily runoff prediction in complex river basins

An approach for flood flow prediction utilizing new hybrids of ANFIS with several optimization techniques: a case study

Hybrid artificial neural network and cooperation search algorithm for nonlinear river flow time series forecasting in humid and semi-humid regions

Application of a novel artificial neural network model in flood forecasting

A comparative study of artificial neural networks, Bayesian neural networks and adaptive neuro-fuzzy inference system in groundwater level prediction

Reservoir Inflow Forecasting Using Ensemble Models Based on Neural Networks, Wavelet Analysis and Bootstrap Method

Daily Streamflow Forecasting Based on Flow Pattern Recognition

Flood discharge prediction using improved ANFIS model combined with hybrid particle swarm optimisation and slime mould algorithm

Application of augmented bat algorithm with artificial neural network in forecasting river inflow in Malaysia

Superiority of artificial neural networks over conventional hydrological models in simulating urban catchment runoff

Monthly surface runoff prediction using artificial intelligence: A study from a tropical climate river basin