Ozgur Kisi

Multi-layer perceptrons with Levenberg-Marquardt training algorithm for suspended sediment concentration prediction and estimation / Prévision et estimation de la concentration en matières en suspension avec des perceptrons multi-couches et l’algorithme d’apprentissage de Levenberg-Marquardt

Abstract Abstract The prediction and estimation of suspended sediment concentration are investigated by using multi-layer perceptrons (MLP). The fastest MLP training algorithm, that is the Levenberg-Marquardt algorithm, is used for optimization of the network weights for data from two stations on the Tongue River in Montana, USA. The first part of the study deals with prediction and estimation of upstream and down-stream station sediment data, separately, and the second part focuses on the estimation of downstream suspended sediment data by using data from both stations. In each case, the MLP test results are compared to those of generalized regression neural networks (GRNN), radial basis function (RBF) and multi-linear regression (MLR) for the best-input combinations. Based on the comparisons, it was found that the MLP generally gives better suspended sediment concentration estimates than the other neural network techniques and the conventional statistical method (MLR). However, for the estimation of maximum sediment peak, the RBF was mostly found to be better than the MLP and the other techniques. The results also indicate that the RBF and GRNN may provide better performance than the MLP in the estimation of the total sediment load.

Suspended sediment estimation using neuro-fuzzy and neural network approaches/Estimation des matières en suspension par des approches neurofloues et à base de réseau de neurones

Abstract The abilities of neuro-fuzzy (NF) and neural network (NN) approaches to model the streamflow–suspended sediment relationship are investigated. The NF and NN models are established for estimating current suspended sediment values using the streamflow and antecedent sediment data. The sediment rating curve and multi-linear regression are also applied to the same data. Statistic measures were used to evaluate the performance of the models. The daily streamflow and suspended sediment data for two stations—Quebrada Blanca station and Rio Valenciano station—operated by the US Geological Survey were used as case studies. Based on comparison of the results, it is found that the NF model gives better estimates than the other techniques.

Neural Networks and Wavelet Conjunction Model for Intermittent Streamflow Forecasting

Intermittent streamflow estimates are important for water quality management, planning water supplies, hydropower, and irrigation systems. This paper proposes the application of a conjunction model (neurowavelet) for forecasting daily intermittent streamflow. The neurowavelet conjunction model is improved by combining two methods, discrete wavelet transform and artificial neural networks (ANN), for 1 day ahead streamflow forecasting and results are compared with those of the single ANN model. Intermittent streamflow data from two stations in the Thrace Region, the European part of Turkey, in the northwest part of the country are used in the study. The comparison results revealed that the suggested model could significantly increase the forecast accuracy of single ANN in forecasting daily intermittent streamflows. The neurowavelet conjunction model reduced the prediction root mean square errors and mean absolute errors with respect to the single ANN model by 74–65% and 43–12%, and increased the determinati...

Modeling rainfall-runoff process using soft computing techniques

Rainfall-runoff process was modeled for a small catchment in Turkey, using 4 years (1987-1991) of measurements of independent variables of rainfall and runoff values. The models used in the study were Artificial Neural Networks (ANNs), Adaptive Neuro-Fuzzy Inference System (ANFIS) and Gene Expression Programming (GEP) which are Artificial Intelligence (AI) approaches. The applied models were trained and tested using various combinations of the independent variables. The goodness of fit for the model was evaluated in terms of the coefficient of determination (R^2), root mean square error (RMSE), mean absolute error (MAE), coefficient of efficiency (CE) and scatter index (SI). A comparison was also made between these models and traditional Multi Linear Regression (MLR) model. The study provides evidence that GEP (with RMSE=17.82l/s, MAE=6.61l/s, CE=0.72 and R^2=0.978) is capable of modeling rainfall-runoff process and is a viable alternative to other applied artificial intelligence and MLR time-series methods.

Comparison of different ANN techniques in river flow prediction

Forecasts of future events are required in many of the activities associated with the planning and operation of the components of a water resource system. For the hydrologic component, there is a need for both short- and long-term forecasts of river flow events in order to optimize the system or to plan for future expansion or reduction. This paper presents the comparison of different artificial neural network (ANN) techniques in short- and long-term continuous and intermittent daily streamflow forecasting. The studies in modelling the intermittent series are quite limited because of the complexity of fitting models in to these series. The available conventional models necessitate the adjustment of numerous parameters for calibration. Three different ANN techniques, namely, feed-forward back propagation (FFBP), generalized regression neural networks, and radial basis function-based neural networks (RBF) are applied to continuous and intermittent river flow data of two Turkish rivers for short-range and long-range forecasting studies. The k-fold partitioning method is employed for preparing the ANN training data successfully. In general, the forecasting performance of RBF is found to be superior to the other two ANN techniques and a time series model in terms of the selected performance criteria. It was observed that the FFBP method had some drawbacks such as a local minima problem and negative flow generation.

Comparison of genetic programming with neuro-fuzzy systems for predicting short-term water table depth fluctuations

This paper investigates the ability of genetic programming (GP) and adaptive neuro-fuzzy inference system (ANFIS) techniques for groundwater depth forecasting. Five different GP and ANFIS models comprising various combinations of water table depth values from two stations, Bondville and Perry, are developed to forecast one-, two- and three-day ahead water table depths. The root mean square errors (RMSE), scatter index (SI), Variance account for (VAF) and coefficient of determination (R^2) statistics are used for evaluating the accuracy of models. Based on the comparisons, it was found that the GP and ANFIS models could be employed successfully in forecasting water table depth fluctuations. However, GP is superior to ANFIS in giving explicit expressions for the problem.

Forecasting daily lake levels using artificial intelligence approaches

Accurate prediction of lake-level variations is important for planning, design, construction, and operation of lakeshore structures and also in the management of freshwater lakes for water supply purposes. In the present paper, three artificial intelligence approaches, namely artificial neural networks (ANNs), adaptive-neuro-fuzzy inference system (ANFIS), and gene expression programming (GEP), were applied to forecast daily lake-level variations up to 3-day ahead time intervals. The measurements at the Lake Iznik in Western Turkey, for the period of January 1961-December 1982, were used for training, testing, and validating the employed models. The results obtained by the GEP approach indicated that it performs better than ANFIS and ANNs in predicting lake-level variations. A comparison was also made between these artificial intelligence approaches and convenient autoregressive moving average (ARMA) models, which demonstrated the superiority of GEP, ANFIS, and ANN models over ARMA models.

Precipitation forecasting by using wavelet-support vector machine conjunction model

A new wavelet-support vector machine conjunction model for daily precipitation forecast is proposed in this study. The conjunction method combining two methods, discrete wavelet transform and support vector machine, is compared with the single support vector machine for one-day-ahead precipitation forecasting. Daily precipitation data from Izmir and Afyon stations in Turkey are used in the study. The root mean square errors (RMSE), mean absolute errors (MAE), and correlation coefficient (R) statistics are used for the comparing criteria. The comparison results indicate that the conjunction method could increase the forecast accuracy and perform better than the single support vector machine. For the Izmir and Afyon stations, it is found that the conjunction models with RMSE=46.5mm, MAE=13.6mm, R=0.782 and RMSE=21.4mm, MAE=9.0mm, R=0.815 in test period is superior in forecasting daily precipitations than the best accurate support vector regression models with RMSE=71.6mm, MAE=19.6mm, R=0.276 and RMSE=38.7mm, MAE=14.2mm, R=0.103, respectively. The ANN method was also employed for the same data set and found that there is a slight difference between ANN and SVR methods.

River Suspended Sediment Load Prediction: Application of ANN and Wavelet Conjunction Model

Accurate suspended sediment prediction is an integral component of sustainable water resources and environmental systems. This study considered artificial neural network (ANN), wavelet analysis and ANN combination (WANN), multilinear regression (MLR), and sediment rating curve (SRC) models for daily suspended sediment load (S) modeling in the Iowa River gauging station in the United States. In the proposed WANN model, discrete wavelet transform was linked to the ANN method. For this purpose, observed time series of river discharge (Q) and S were decomposed into several subtime series at different scales by discrete wavelet transform. Then these subtime series were imposed as inputs to the ANN method to predict one-day-ahead S. The results showed that the WANN model was in good agreement with the observed S values and that it performed better than the other models. The coefficient of efficiency was 0.81 for the WANN model and 0.67, 0.6, and 0.39 for the ANN, MLR, and SRC models, respectively. In addition, ...

River Flow Estimation and Forecasting by Using Two Different Adaptive Neuro-Fuzzy Approaches

This paper demonstrates the application of two different adaptive neuro-fuzzy (ANFIS) techniques for the estimation of monthly streamflows. In the first part of the study, two different ANFIS models, namely ANFIS with grid partition (ANFIS-GP) and ANFIS with sub clustering (ANFIS-SC), were used in one-month ahead streamflow forecasting and the results were evaluated. Monthly flow data from two stations, the Besiri Station on the Garzan Stream and the Baykan Station on the Bitlis Stream in the Firat-Dicle Basin of Turkey were used in the study. The effect of periodicity on the model’s forecasting performance was also investigated. In the second part of the study, the performance of the ANFIS techniques was tested for streamflow estimation using data from the nearby river. The results indicated that the performance of the ANFIS-SC model was slightly better than the ANFIS-GP model in streamflow forecasting.