Search

Due to excessive streamflow (SF), Peninsular Malaysia has historically experienced floods and droughts. Forecasting streamflow to mitigate municipal and environmental damage is therefore crucial. Streamflow prediction has been extensively demonstrated in the literature to estimate the continuous values of streamflow level. Prediction of continuous values of streamflow is not necessary in several applications and at the same time it is very challenging task because of uncertainty. A streamflow category prediction is more advantageous for addressing the uncertainty in numerical point forecasting, considering that its predictions are linked to a propensity to belong to the pre-defined classes. Here, we formulate streamflow prediction as a time series classification with discrete ranges of values, each representing a class to classify streamflow into five or ten, respectively, using machine learning approaches in various rivers in Malaysia. The findings reveal that several models, specifically LSTM, outperform others in predicting the following n-time steps of streamflow because LSTM is able to learn the mapping between streamflow time series of 2 or 3 days ahead more than support vector machine (SVM) and gradient boosting (GB). LSTM produces higher F1 score in various rivers (by 5% in Johor, 2% in Kelantan and Melaka and Selangor, 4% in Perlis) in 2 days ahead scenario. Furthermore, the ensemble stacking of the SVM and GB achieves high performance in terms of F1 score and quadratic weighted kappa. Ensemble stacking gives 3% higher F1 score in Perak river compared to SVM and gradient boosting. [ABSTRACT FROM AUTHOR]

Streamflow forecasting process exhibited highly nonstationary and stochastic pattern, thus not easy to be done with simple models. There is a need to develop an efficient and precise streamflow forecasting system which is vital for water management at hydrological infrastructures like Aswan High Dam (AHD). As the decision makers will be able to decide on water allocation for different purposes such as irrigation, domestic and industrial uses. This study explores the potential of AI model: nonlinear autoregressive neural network (NAR) in performing inflow forecasting to AHD. The dataset of past 130 years of Nile River discharge rate was used for the network development as well as evaluation of models' performance. This study also proposes an integration process of NAR with Holt-Winters exponential smoothing to improve the accuracy of the model. To determine the models' performance, different indicators were employed and calculated (MAE, MAPE, RMSE, R2). The results were compared to identify the optimal network architecture. The results show that the NAR models are capable of predicting the future values of AHD inflow in monthly time steps accurately. For standard NAR model, the root mean squared error (RMSE) was 2.0072, and the coefficient of determination (R2) between recorded and forecasted values was 0.9152. Values of RMSE = 1.5421 and R2 = 0.9760 and RMSE = 1.0843 and R2 = 0.9823 were obtained by NAR-SES and NAR-HW models respectively. The results reveal that combination of Holt-Winters exponential smoothing with NAR significantly improved the precision beyond the standard model. This study proved that NAR neural networks can be useful to address streamflow forecasting problems. [ABSTRACT FROM AUTHOR]

One of the largest hydropower facilities currently in operation in Malaysia is the Terengganu hydroelectric facility. As a result, for hydropower generation to be sustainable, future water availability in hydropower plants must be known. Therefore, it is necessary to precisely estimate how the river flow will alter as a result of changing rainfall patterns. Finding the best value for the hyper-parameters is one of the problems with machine learning algorithms, which have lately been adopted by many academics. In this research, Artificial Neural Network (ANN) is integrated with a nature-inspired optimizer, namely Cuckoo search algorithm (CS-ANN). The performance of the proposed algorithm then will be examined based on statistical indices namely Root-Mean-Square Error (RSME) and Determination Coefficient (R2). Then, the accuracy of the proposed model will be then examined with the stand-alone Artificial Neural Network (ANN). The statistical indices results indicate that the proposed Hybrid CS-ANN model showed an improvement based on R2 value as compared to ANN model with R2 of 0.900 at training stage and R2 of 0.935 at testing stage. RMSE value, for ANN model, is 127.79 m3/s for training stage and 12.7 m3/s at testing stage. While for the proposed Hybrid CS-ANN model, RMSE value is equal to 121.7 m3/s for training stage and 10.95 m3/s for testing stage. The results revealed that the proposed model outperformed the stand-alone model in predicting the river flow with high level of accuracy. Although the proposed model could be applied in different case study, there is a need to tune the model internal parameters when applied in different case study. [ABSTRACT FROM AUTHOR]

Streamflow forecasting has always been important in water resources management, particularly the peak flow, which often determines the seriousness of the impending flood. However, the highly imbalanced flow distribution often hinders the machine learning algorithm's performance. In this paper, streamflow forecasting was approached through the formulation of two distinct machine learning problems: categorical streamflow forecast and regression streamflow forecast. Due to the distinctive characteristics of these two adopted forms, selecting the correct algorithm for the machine learning problem along with their hyperparameter tuning process is critical to the realization of the desired results. For the distinct streamflow formulated scenarios, three neural network algorithms and their hyperparameter tuning strategy were investigated. The comparative empirical studies had revealed that formulated categorical-based streamflow forecast is a better choice than a regression-based streamflow forecast, regardless of the algorithms used; for instance, the f1-score of 0.7 (categorical based) is obtained compared to the 0.53 (regression based) for the LSTM in scenario 1 (binary). Furthermore, forest-based algorithms were investigated and shown to be superior at forecasting high streamflow fluctuations in situations featuring low-dimensional streamflow input. Besides, encoding the streamflow time series as images (input) for forecasting purposes would require a thorough analysis as there is a discrepancy in the results, revealing that not all approaches are suitable for streamflow image transformation. The functional ANOVA analysis provided evidence to substantiate the Bayesian optimization results, implying that the hyperparameters were effectively optimized. [ABSTRACT FROM AUTHOR]

The development of a river inflow prediction is a prerequisite for dam reservoir management. Precise forecasting leads to better irrigation water management, reservoir operation refinement, enhanced hydropower output and mitigation of risk of natural hazards such as flooding. Dam created reservoirs prove to be an essential source of water in arid and semi-arid regions. Over the years, Artificial Intelligence (AI) has been used for development of models for prediction of various natural variables in different engineering fields. Also, several AI models have been proved to be beneficial over the conventional models in efficient prediction of various natural variables. In this study, four AI models, namely, Artificial Neural Network (ANN), Support Vector Machine (SVM), Random Forest (RF) and Boosted Tree Regression (BTR) were developed and trained over 130-years of monthly historical rainfall data to forecast streamflow at Aswan High Dam, Egypt. The input parameters were selected according to the Autocorrelation Function (ACF) plot. The findings revealed that RF model outperformed other techniques and could provide precise monthly streamflow prediction with the lowest RMSE (2.2395) and maximum WI (0.998462), R2 (0.9012). The input combination for the optimum RF model was Qt-1, Qt-11, and Qt-12 (i.e., one-, eleven- and twelve-months delay inputs). The optimum RF model provides a reliable source of data for inflow predictions, which allows improved utilization of water resources and long-term water resource planning and management. [ABSTRACT FROM AUTHOR]

Due to the need to reduce the flooding disaster, river streamflow prediction is required to be enhanced by the aid of deep learning algorithms. To achieve accurate model of streamflow prediction, it is important to provide suitable data sets to train the predictive models. Thus, this research has investigated two sampling approaches by using deep learning algorithms. These sampling approaches are linear and stratified selection in deep learning algorithms. This investigation has been performed on the Tigris River data set in terms of 2 scenarios. The first scenario: implementation of 12 different linear and stratified sampling selection in deep learning models. This scenario is trained and tested as much as a number of months per year—12 months. The second scenario: the complete time series is taken into consideration while performing the two approaches that are utilized in this research. Furthermore, the optimal input combination is identified via correlation analysis. To evaluate the performance of the algorithms utilized in this research, a number of metrics have been used which are Root Mean Square Error RMSE, Absolute Error AE, Relative Error RE, Relative Error Lenient REL, Relative Error Strict RES, Root Relative Squared Error RRSE, Coefficient of determination R2, Spearman rho and Kendall tau. The results have indicated that in both scenarios, stratified-deep learning (SDL) improves the accuracy by about 7.96–94.6 with respect to several assessment criteria. Thus, finally, it is worth mentioning that SDL outperforms Linear-deep learning (LDL) in monthly streamflow modelling. [ABSTRACT FROM AUTHOR]

Floods and droughts are environmental phenomena that occur in Peninsular Malaysia due to extreme values of streamflow (SF). Due to this, the study of SF prediction is highly significant for the purpose of municipal and environmental damage mitigation. In the present study, machine learning (ML) models based on the support vector machine (SVM), artificial neural network (ANN), and long short-term memory (LSTM), are tested and developed to predict SF for 11 different rivers throughout Peninsular Malaysia. SF data sets for the rivers were collected from the Malaysian Department of Irrigation and Drainage. The main objective of the present study is to propose a universal model that is most capable of predicting SFs for rivers within Peninsular Malaysia. Based on the findings, the ANN3 model which was developed using the ANN algorithm and input scenario 3 (inputs consisting of previous 3 days SF) is deduced as the best overall ML model for SF prediction as it outperformed all the other models in 4 out of 11 of the tested data sets; and obtained among the highest average RMs with a score of 3.27, hence indicating that the model is very adaptable and reliable in accurately predicting SF based on different data sets and river case studies. Therefore, the ANN3 model is proposed as a universal model for SF prediction within Peninsular Malaysia. [ABSTRACT FROM AUTHOR]

This paper focuses on the development of a robust accurate streamflow prediction model by balancing the abilities of exploitation and exploration to find the best parameters of a machine learning model. To do so, the simulated annealing (SA) algorithm is integrated with the mayfly optimization algorithm (MOA) as SAMOA to determine the optimal hyper-parameters of support vector regression (SVR) to overcome the exploration weakness of the MOA method. The proposed method is compared with the classical SVR and hybrid SVR-MOA. To examine the accuracy of the selected methods, monthly hydroclimatic data from Jhelum River Basin is used to predict the monthly streamflow on the basis of RMSE, MAE, NSE, and R2 indices. Test results show that the SVR-SAMOA outperformed the SVR-MOA and SVR models. SVR-SAMOA reduced the prediction errors of the SVR-MOA and SVR models by decreasing the RMSE and the MSE from 21.4% to 14.7% and from 21.7% to 15.1%, respectively, in the test stage. [ABSTRACT FROM AUTHOR]

Accurate runoff forecasting plays a key role in catchment water management and water resources system planning. To improve the prediction accuracy, one needs to strive to develop a reliable and accurate forecasting model for streamflow. In this study, the novel combination of the adaptive neuro-fuzzy inference system (ANFIS) model with the shuffled frog-leaping algorithm (SFLA) is proposed. Historical streamflow data of two different rivers were collected to examine the performance of the proposed model. To evaluate the performance of the proposed ANFIS-SFLA model, six different scenarios for the model input–output architecture were investigated. The results show that the proposed ANFIS-SFLA model (R2 = 0.88; NS = 0.88; RMSE = 142.30 (m3/s); MAE = 88.94 (m3/s); MAPE = 35.19%) significantly improved the forecasting accuracy and outperformed the classic ANFIS model (R2 = 0.83; NS = 0.83; RMSE = 167.81; MAE = 115.83 (m3/s); MAPE = 45.97%). The proposed model could be generalized and applied in different rivers worldwide. [ABSTRACT FROM AUTHOR]

In nature, streamflow pattern is characterized with high non-linearity and non-stationarity. Developing an accurate forecasting model for a streamflow is highly essential for several applications in the field of water resources engineering. One of the main contributors for the modeling reliability is the optimization of the input variables to achieve an accurate forecasting model. The main step of modeling is the selection of the proper input combinations. Hence, developing an algorithm that can determine the optimal input combinations is crucial. This study introduces the Genetic algorithm (GA) for better input combination selection. Radial basis function neural network (RBFNN) is used for monthly streamflow time series forecasting due to its simplicity and effectiveness of integration with the selection algorithm. In this paper, the RBFNN was integrated with the Genetic algorithm (GA) for streamflow forecasting. The RBFNN-GA was applied to forecast streamflow at the High Aswan Dam on the Nile River. The results showed that the proposed model provided high accuracy. The GA algorithm can successfully determine effective input parameters in streamflow time series forecasting. [ABSTRACT FROM AUTHOR]

Hydrological modeling is one of the important subjects in managing water resources and the processes of predicting stochastic behavior. Developing Data-Driven Models (DDMs) to apply to hydrological modeling is a very complex issue because of the stochastic nature of the observed data, like seasonality, periodicities, anomalies, and lack of data. As streamflow is one of the most important components in the hydrological cycle, modeling and estimating streamflow is a crucial aspect. In this study, two models, namely, Optimally Pruned Extreme Learning Machine (OPELM) and Chi-Square Automatic Interaction Detector (CHAID) methods were used to model the deterministic parts of monthly streamflow equations, while Autoregressive Conditional Heteroskedasticity (ARCH) was used in modeling the stochastic parts of monthly streamflow equations. The state of art and innovation of this study is the integration of these models in order to create new hybrid models, ARCH-OPELM and ARCH-CHAID, and increasing the accuracy of models. The study draws on the monthly streamflow data of two different river stations, located in north-western Iran, including Dizaj and Tapik, which are on Nazluchai and Baranduzchai, gathered over 31 years from 1986 to 2016. To ascertain the conclusive accuracy, five evaluation metrics including Correlation Coefficient (R), Root Mean Square Error (RMSE), Nash-Sutcliffe Efficiency (NSE), Mean Absolute Error (MAE), the ratio of RMSE to the Standard Deviation (RSD), scatter plots, time-series plots, and Taylor diagrams were used. Standalone CHAID models have better results than OPELM methods considering sole models. In the case of hybrid models, ARCH-CHAID models in the validation stage performed better than ARCH-OPELM for Dizaj station (R = 0.96, RMSE = 1.289 m³/s, NSE = 0.92, MAE = 0.719 m³/s and RSD = 0.301) and for Tapik station (R = 0.94, RMSE = 2.662 m³/s, NSE = 0.86, MAE = 1.467 m³/s and RSD = 0.419). The results remarkably reveal that ARCH-CHAID models in both stations outperformed all other models. Finally, it is worth mentioning that the new hybrid "ARCH-DDM" models outperformed standalone models in predicting monthly streamflow. [ABSTRACT FROM AUTHOR]

Hydrological models play a crucial role in water planning and decision making. Machine Learning-based models showed several drawbacks for frequent high and a wide range of streamflow records. These models also experience problems during the training process such as over-fitting or trapping in searching for global optima To overcome these limitations, the current study attempts to hybridize the recently developed physics-inspired metaheuristic algorithms (MHAs) such as Equilibrium Optimization (EO), Henry Gases Solubility Optimization (HGSO), and Nuclear Reaction Optimization(NRO) with Multi-layer Perceptron (MLP). These models' accuracy will be inspected to solve the streamflow forecasting problem where the streamflow dataset was collected through 130 years from a station located on the High Aswan Dam (HAD). The performance of proposed models then will be compared with two traditional neural network models(MLP and RNN), and nine well-known hybrid MLP-based models belong to the different branches of the metaheuristic field (evolutionary group, swarm group, and physics group). The internal parameters of the proposed models will be initialized and optimized. Different performance metrics will be used to examine the performance of the proposed models. The stability of the proposed models and the convergence speed will be evaluated. Finally, ranking these models based on different performance evaluations will be carried out. The results show that the models in the group of Physics-MLP are more reliable in capturing the streamflow patterns, followed by the Swarm-MLP group and then by the Evolutionary-MLP group. Finally, among the all employed methods, the NRO has the best accuracy with the lowest RMSE(2.35), MAE(1.356), MAPE(16.747), and the highest WI(0.957), R(0.924), and confidence in forecasting the streamflow of Aswan High Dam. It can be concluded that augmenting the NRO algorithm with MLP can be a reliable tool in forecasting the monthly streamflow with a high level of precision, speed convergence, and high constancy level. • Develop an improved hybrid Machine Learning model for streamflow forecasting. • Augment MLP model with three physics-inspired metaheuristics algorithms. • Hybridized (Hybrid-MLP) models more reliable than standalone (MLP and RNN) models. • Physics-based hybrid models outperformed evolutionary-based and swarm-based models. • NRO-MLP is reliable in terms of prediction accuracy, convergence, and stability. [ABSTRACT FROM AUTHOR]

The current study investigates the effect of a large climate index, such as NINO3, NINO3.4, NINO4 and PDO, on the monthly stream flow in the Aydoughmoush basin (Iran) based on an improved Adaptive Neuro Fuzzy Inference System (ANFIS) during 1987–2007. The bat algorithm (BA), particle swarm optimization (PSO) and genetic algorithm (GA) were used to obtain the ANFIS parameter for the best ANFIS structure. Principal component analysis (PCA) and Varex rotation were used to decrease the number of effective components needed for the streamflow simulation. The results showed that the large climate index with six-month lag times had the best performance, and three components (PCA1, PCA2 and PCA3) were used to simulate the monthly streamflow. The results indicated that the ANFIS-BA had better results than the ANFIS-PSO and ANFIS-GA, with a root mean square error (RMSE) 25% and 30% less than the ANFIS-PSO and ANFIS-GA, respectively. In addition, the linear error in probability space (LEPS) score for the ANFIS-BA, based on the average values for the different months, was less than the ANFIS-PSO and ANFIS-GA. Furthermore, the uncertainty values for the different ANFIS models were used and the results indicated that the monthly simulated streamflow by the ANFIS was computed well at the 95% confidence level. It can be seen that the average streamflow for the summer season is 75 m3/s, so that the stream flow for summer, based on climate indexes, is more than that in other seasons. [ABSTRACT FROM AUTHOR]

• Proposed the novel PROFINN method for daily streamflow imputation. • Integrates probabilistic fusion principles with deep neural networks. • Proposed method maintains precision and allows quantification of uncertainty. • PROFINN performs remarkably well on comparing with RF, K-NN and PMM. • Performs consistently well under different types of data gaps and flow states. The present-day accessibility of streamflow data, particularly in the developing countries, is often marked by a multitude of data shortfalls or distortions. This study investigates the estimate of missing streamflow data using machine learning approaches, including K-nearest neighbour (KNN), Predictive Mean Matching (PMM), Random Forest (RF) and a novel technique of Probabilistic Fusion Imputer with Neural Networks (PROFINN). This study tackles the issue of data insufficiency in such time series considering the inclusion of numerous hydrological parameters by means of RF feature selector, to determine the most significant ones among them. The study explores the efficacy of selected models on various data gaps under different hydrological scenarios, including diverse flow characteristics (mean, high and low flows) and gap lengths (long and short gaps, continuous and discontinuous gaps) and presents a pattern of ranking system that assesses the level of suitability of each technique for various data gaps. This study underscore PROFINN's remarkable performance across all scenarios, yielding an average Root Mean Square Error (RMSE) of 0.91 and an average Nash-Sutcliffe Efficiency (NSE) value of 0.93 when applied for an intermittent river system of Pamba in Southern Kerala, India. RF follows PROFINN in the imputation of extreme flows as well as long and short gap scenarios. KNN closely follows PROFINN for the imputation of continuous and discontinuous gap scenarios. This study augments the significance of tailored machine learning techniques in enhancing the integrity of hydrological datasets, offering valuable insights for effective decision-making in water resource management and related fields. Furthermore, the success of PROFINN in streamflow data imputation suggests its potential extension to other hydrological research domains, like historical data reconstruction assisting climate change studies and future water resources planning, emphasizing the broader relevance and applicability of this study's findings. [ABSTRACT FROM AUTHOR]

• Several hybrid multilayer Perceptrons and Coupla Bayesian model averaging are proposed for streamflow prediction. • Multilayer Perceptrons had been trained using several evolutionary optimization methods. • The hybrid models using either regular or Copula-based-Bayesian model averaging outperformed other models. • Accuracy of the streamflow prediction had been significantly improved using combined MLP and Copula-based-Bayesian model averaging. Streamflow prediction help the modelers to manage water resources in watersheds. It gives essential information for flood control and reservoir operation. This study uses the copula-based -Bayesian model averaging (CBMA) as an improved version of the BMA model for predicting streamflow in the Golok River, the Kelantan River, the Lanas River, and the Nenggiri River of Malaysia. The CBMA corrected the assumption of the utilization of Gaussian distortion in the BMA. While the BMA used normal distribution for the variables, the CBMA uses different distribution and copula functions for the variables. This study works on the Archimedes optimization algorithm (AOA) to train the mutlilayer perceptron (MLP) model. The ability of the MLP-AOA model was benchmarked against the MLP-bat algorithm (BA), MLP-particle swarm optimization (MLP-PSO), and the MLP-firefly algorithm (MLP-FFA). The models used significant climate signals, namely, the southern oscillation index (SOI), El NiÑo–Southern Oscillation (ENSO), North Atlantic oscillation (NAO), and the pacific decadal oscillation (PDO) as the inputs to the models. The Gamma test (GT) was coupled with the AOA to provide the hybrid GT for choosing the best inputs. The gamma test was used to determine the suitable lag times of the Nino 3.4, PDO, NAO, and SOI as the inputs. The novelty of the current paper includes introducing new hybrid MLP models, new gamma test for choosing the best input combination, the comprehensive uncertainty analysis of outputs, and the use of an advanced ensemble CBMA model for predicting streamflow. First, the outputs of the MLP-AOA, MLP-BA, MLP-FFA, MLP-PSO, and MLP were obtained, following which, the CBMA as an ensemble framework based on outputs of the hybrid and standalone MLP models was used to predict monthly streamflow. The CBMA at the training level, decreased the root mean square error (RMSE) of BMA, MLP-AOA, MLP-BA, MLP-FFA, MLP-PSO, and MLP models by 28%, 32%, 52%, 53 53%, and 55%, respectively. The CBMA at the training level of another station decreased the mean absolute error (MAE) of BMA, MLP-AOA, MLP-BA, MLP-FFA, MLP-PSO, and MLP models by 6.04, 29%,42%, 49%, 52%, and 52%, respectively. The Nash Sutcliff efficiency (NSE) of the CBMA at the training level was 0.94 while it was 0.92, 0.90, 0.85, 0.84, 0.82, and 0.80 for the BMA, MLP-AOA, MLP-BA, MLP-FFA, MLP-PSO, and MLP models. The RMSE of the MLP-AOA was reported 4.3%, 12%, 14%, and 17% lower than those of the MLP-BA, MLP-FFA, MLP-PSO, and MLP models, respectively. The current research showed the CBMA and the BMA models had high abilities for predicting monthly streamflow. The results of this current study indicated that the CBMA and BMA provided lower uncertainty the standalone MLP models. The general results indicated that the streamflows in the hotter months decreased and flood control is of higher priority during the other months. [ABSTRACT FROM AUTHOR]

• Investigate the potential of Fast-Orthogonal Search (FOS) for streamflow forecasting. • Utilize FOS identification technique to provide a wider selection of candidate. • Examine several training approaches for developing robust forecasting model. • Compare the FOS based model to existing AI based models. • Demonstrate the performance using realistic data from the Nile river in Egypt. Data-driven models for streamflow forecasting have attracted considerable attention, as they are independent of physical system features. The physical features of the river basin are extremely hard to collect, especially for large rivers. Empirical data-driven models, such as stochastic and regression models, have been widely used in the field of streamflow forecasting. However, they suffered limited accuracy in predicting extreme streamflow. They also required raw data pre-processing prior to the modeling process, especially for lengthy data records and for large time-scale increments (e.g. monthly resolution). To overcome these challenges, data-driven forecasting models based on Artificial Intelligence (AI) have been widely used and resulted in enhancing the forecasting accuracy. Nevertheless, AI-based models required augmentation with proper optimization schemes to adjust the model parameters for optimal accuracy. Furthermore, in some cases, due to unsuitability of the optimization model, there is high possibility for overfitting of the AI model, which might cause poor prediction of input patterns that were not adequately mimicked. This study introduces a new approach to streamflow forecasting based on nonlinear system identification. The proposed technique employs Fast Orthogonal Search (FOS) to develop a nonlinear model of stream flow. The main advantage of using FOS is eliminating the requirement of raw data pre-processing and the need for an optimization scheme for model parameter adjustment since the FOS algorithm takes this into account while building the model. In addition, the FOS algorithm includes a pole-zero cancellation procedure that can detect and avoid the over-fitted models. The FOS-based nonlinear modeling approach was adopted in this research for the development of a streamflow forecasting model at Aswan High Dam using monthly basis natural streamflow records for 130 years. The results indicated that the proposed FOS algorithm outperformed the previously developed AI models of streamflow forecasting for large data records and for large time-scale increment (monthly resolution). [ABSTRACT FROM AUTHOR]