Daily Suspended Sediment Load Estimation by Meta-heuristic Optimization Approaches and Fuzzy-C-Means Clustering Method

Mahmoodreza Tabatabaei

doi:10.17485/ijst/2014/v7i10.16

Abstract

Accurate Suspended Sediment Load (SSL) estimation is very important for water resources quantity and quality studies. In this regard, Sediment Rating Curve (SRC) is a common regression model in predicting SSL of discharge. Studies in this field have shown that the data log-transformation in SRC model causing a bias which underestimates SSL prediction. In this study, using data from the daily flow discharge and suspended sediment discharge of Karaj Dam watershed at Siera station, a 30–year period (1981 to 2011), SRC equation derived, and then, using meta-heuristic algorithms (genetic algorithm and particle swarm optimization algorithm) it was calibrated again. Before modeling to increase the generalization power of the models, using fuzzy clustering method, the data were clustered and then by doing data sampling, they were classified into two homogeneous groups (calibration and test data set). The results show that meta-heuristic algorithms are appropriate methods for optimizing coefficients of SRC model and their results are much more favorable than those of the conventional SRC models or SRC models corrected by correction factors. In this relation, the sediment rating curve models calibrated with meta-heuristic algorithms, by reducing the RMSE of the test data set of 3718.87 ton/day (in the initial SRC model) to 2615/119 ton/day (in the calibrated models by meta-heuristic algorithms) increased the accuracy of suspended sediment load estimation at a rate of 1103.68 ton/day. However, the SRC model corrected by FAO factor decreased the efficiency of initial SRC model by increasing the RMSE of the test data set to 4128/73 ton/day. Using meta-heuristic algorithms in calibrating SRC models also prevents data log-transformation and use of correction factors and increases the accuracy of results. Keywords: Fuzzy-C-Means Clustering, Genetic and PSO Algorithms, Karaj Dam, Sediment Rating Curve, Suspended Sediment Load

Highlights

It is necessary to have adequate up-to-date information about the Suspended Sediment Load (SSL) of rivers and monitor them continually in order to be aware of the watershed sediment yield condition, the amount of erosion and changes in the river bed and river bank, the quality of water, and optimum design and favorable performance of water resource structures[1,2,3,4,5]
To correct the bias resulting from the logarithmic transformation, different correction factors have been introduced so far (FAO, QMLE (Quasi-Maximum Likelihood Estimator), MVUE (Minimum Variance Unbiased Estimator) etc.), and all of them aim at increasing the values calculated through Sediment Rating Curve (SRC) model
The comparative analysis of the results showed that the Adaptive Neuro-Fuzzy Inference System (ANFIS) model has superiority over the other models for estimating daily suspended sediment concentration

Summary

Introduction

It is necessary to have adequate up-to-date information about the Suspended Sediment Load (SSL) of rivers and monitor them continually in order to be aware of the watershed sediment yield condition, the amount of erosion and changes in the river bed and river bank, the quality of water, and optimum design and favorable performance of water resource structures[1,2,3,4,5]. In other words, when calculating a and b coefficients, a kind of bias appears in SRC regression model and makes the estimated values of SSL lower than its corresponding observed values[8] This problem is more obvious in flood discharges and causes more errors. To correct the bias resulting from the logarithmic transformation, different correction factors have been introduced so far (FAO, QMLE (Quasi-Maximum Likelihood Estimator), MVUE (Minimum Variance Unbiased Estimator) etc.), and all of them aim at increasing the values calculated through SRC model. These factors sometimes cause another bias in the form of an overestimation besides making the results with the same data different[7]

Methods

Results

Conclusion