Application of newly developed ensemble machine learning models for daily suspended sediment load prediction and related uncertainty analysis

ABSTRACT

Ensemble machine learning models have been widely used in hydro-systems modeling as robust prediction tools that combine multiple decision trees. In this study, three newly developed ensemble machine learning models, namely gradient boost regression (GBR), AdaBoost regression (ABR) and random forest regression (RFR) are proposed for prediction of suspended sediment load (SSL), and their prediction performance and related uncertainty are assessed. The SSL of the Mississippi River, which is one of the major world rivers and is significantly affected by sedimentation, is predicted based on daily values of river discharge (Q) and suspended sediment concentration (SSC). Based on performance metrics and visualization, the RFR model shows a slight lead in prediction performance. The uncertainty analysis also indicates that the input variable combination has more impact on the obtained predictions than the model structure selection.     

Monthly suspended sediment load prediction using artificial intelligence: testing of a new random subspace method

ABSTRACT

The predictive capability of a new artificial intelligence method, random subspace (RS), for the prediction of suspended sediment load in rivers was compared with commonly used methods: random forest (RF) and two support vector machine (SVM) models using a radial basis function kernel (SVM-RBF) and a normalized polynomial kernel (SVM-NPK). Using river discharge, rainfall and river stage data from the Haraz River, Iran, the results revealed: (a) the RS model provided a superior predictive accuracy (NSE = 0.83) to SVM-RBF (NSE = 0.80), SVM-NPK (NSE = 0.78) and RF (NSE = 0.68), corresponding to very good, good, satisfactory and unsatisfactory accuracies in load prediction; (b) the RBF kernel outperformed the NPK kernel; (c) the predictive capability was most sensitive to gamma and epsilon in SVM models, maximum depth of a tree and the number of features in RF models, classifier type, number of trees and subspace size in RS models; and (d) suspended sediment loads were most closely correlated with river discharge (PCC = 0.76). Overall, the results show that RS models have great potential in data poor watersheds, such as that studied here, to produce strong predictions of suspended load based on monthly records of river discharge, rainfall depth and river stage alone.     

Modelling reservoir sedimentation and estimating historical deposition rates using a data-based mechanistic (DBM) approach

Abstract

The data-based mechanistic (DBM) modelling methodology is applied to the study of reservoir sedimentation. A lumped-parameter, discrete-time model has been developed which directly relates rainfall to suspended sediment load (SSL) at the reservoir outflow from the two years of measured data at Wyresdale Park Reservoir (Lancashire, UK). This nonlinear DBM model comprises two components: a rainfall to SSL model and a second model, relating the SSL at the reservoir inflow to the SSL at the reservoir spillway. Using a daily measured rainfall series as the input, this model is used to reconstruct daily deposition rates between 1911 and 1996. This synthetic sediment accretion sequence is compared with the variations in sand content within sediment cores collected from the reservoir floor. These profiles show good general agreement, reflecting the importance of low reoccurrence, high magnitude events. This preliminary study highlights the potential of this DBM approach, which could be readily applied to other sites.     

Neuro-fuzzy models employing wavelet analysis for suspended sediment concentration prediction in rivers

Abstract

The study of sediment load is important for its implications to the environment and water resources engineering. Four models were considered in the study of suspended sediment concentration prediction: artificial neural networks (ANNs), neuro-fuzzy model (NF), conjunction of wavelet analysis and neuro-fuzzy (WNF) model, and the conventional sediment rating curve (SRC) method. Using data from a US Geological Survey gauging station, the suspended sediment concentration predicted by the WNF model was in satisfactory agreement with the measured data. Also the proposed WNF model generated reasonable predictions for the extreme values. The cumulative suspended sediment load estimated by this model was much higher than that predicted by the other models, and is close to the observed data. However, in the current modelling, the ANN, NF and SRC models underestimated sediment load. The WNF model was successful in reproducing the hysteresis phenomenon, but the SRC method was not able to model this behaviour. In general, the results showed that the NF model performed better than the ANN and SRC models.

Citation Mirbagheri, S. A., Nourani, V., Rajaee, T. & Alikhani, A. (2010) Neuro-fuzzy models employing wavelet analysis for suspended sediment concentration prediction in rivers. Hydrol. Sci. J. 55(7), 1175–1189.     

Estimation of suspended sediment concentration in the Saint John River using rating curves and a machine learning approach

ABSTRACT

Sedimentation in navigable waterways and harbours is of concern for many water and port managers. One potential source of variability in sedimentation is the annual sediment load of the river that empties in the harbour. The main objective of this study was to use some of the regularly monitored hydro-meteorological variables to compare estimates of hourly suspended sediment concentration in the Saint John River using a sediment rating curve and a model tree (M5?) with different combinations of predictors. Estimated suspended sediment concentrations were multiplied by measured flows to estimate suspended sediment loads. Best results were obtained using M5? with four predictors, returning an R² of 0.72 on calibration data and an R² of 0.46 on validation data. Total load was underestimated by 1.41% for the calibration period and overestimated by 2.38% for the validation period. Overall, the model tree approach is suggested for its relative ease of implementation and constant performance.

EDITOR M.C. Acreman; ASSOCIATE EDITOR B. Touaibia     

Assessing the performance and uncertainty analysis of the SWAT and RBNN models for simulation of sediment yield in the Nagwa watershed,India

Abstract

The process-based Soil and Water Assessment Tool (SWAT) model and the data-driven radial basis neural network (RBNN) model were evaluated for simulating sediment load for the Nagwa watershed in Jharkhand, India, where soil erosion is a severe problem. The SWAT model calibration and uncertainty analysis were performed with the Sequential Uncertainty Fitting algorithm version 2 and the bootstrap technique was applied on the RBNN model to analyse uncertainty in model output. The percentage of data bracketed by the 95% prediction uncertainty (95PPU) and the r factor were the two measures used to assess the goodness of calibration. Comparison of the results of the two models shows that the value of r factor (r = 0.41) in the RBNN model is less than that of SWAT model (r = 0.79), which means there is a wider prediction interval for the SWAT model results. More values of observed sediment yield were bracketed by the 95PPU in the RBNN model. Thus, the RBNN model estimates the sediment yield values more accurately and with less uncertainty.

Editor D. Koutsoyiannis; Associate editor H. Aksoy

Citation Singh, A., Imtiyaz, M., Isaac, R.K., and Denis, D.M., 2014. Assessing the performance and uncertainty analysis of the SWAT and RBNN models for simulation of sediment yield in the Nagwa watershed, India. Hydrological Sciences Journal, 59 (2), 351–364.     

Daily suspended sediment modelling using a fuzzy differential evolution approach / Modélisation journalière des matières en suspension par une approche d’évolution différentielle floue

Abstract

Abstract Correct estimation of the sediment volume carried by a river is important with respect to pollution, channel navigability, reservoir filling, hydroelectric equipment longevity, fish habitat, river aesthetics and scientific interests. However, conventional sediment rating curves are not able to provide sufficiently accurate results. In this study, models incorporating fuzzy logic are developed as a superior alternative to the sediment rating curve technique for determining the daily suspended sediment concentration for a given river cross-section. This study provides forecasting benchmarks for sediment concentration prediction in the form of a numerical and graphical comparison between fuzzy and rating curve models. Benchmarking was based on a five-year period of continuous streamflow and sediment concentration data from the Quebrada Blanca Station operated by the United States Geological Survey (USGS). Nine different fuzzy models were developed to estimate sediment concentration from streamflow. Each fuzzy model has a different number of membership functions. The parameters of the membership functions were found using a differential evolution algorithm. The benchmark results showed that the fuzzy models were able to produce much better results than rating curve models for the same data inputs.     

Adaptive neural-based fuzzy inference system (ANFIS) approach for modelling hydrological time series

Abstract

The main aim of this study is to develop a flow prediction method, based on the adaptive neural-based fuzzy inference system (ANFIS) coupled with stochastic hydrological models. An ANFIS methodology is applied to river flow prediction in Dim Stream in the southern part of Turkey. Application is given for hydrological time series modelling. Synthetic series, generated through autoregressinve moving-average (ARMA) models, are then used for training data sets of the ANFIS. It is seen that the extension of input and output data sets in the training stage improves the accuracy of forecasting by using ANFIS.     

Predictability of river flow and suspended sediment transport in the Mississippi River basin: a non‐linear deterministic approach

As the Mississippi River plays a major role in fulfilling various water demands in North America, accurate prediction of river flow and sediment transport in the basin is crucial for undertaking both short‐term emergency measures and long‐term management efforts. To this effect, the present study investigates the predictability of river flow and suspended sediment transport in the basin. As most of the existing approaches that link water discharge, suspended sediment concentration and suspended sediment load possess certain limitations (absence of consensus on linkages), this study employs an approach that presents predictions of a variable based on history of the variable alone. The approach, based on non‐linear determinism, involves: (1) reconstruction of single‐dimensional series in multi‐dimensional phase‐space for representing the underlying dynamics; and (2) use of the local approximation technique for prediction. For implementation, river flow and suspended sediment transport variables observed at the St. Louis (Missouri) station are studied. Specifically, daily water discharge, suspended sediment concentration and suspended sediment load data are analysed for their predictability and range, by making predictions from one day to ten days ahead. The results lead to the following conclusions: (1) extremely good one‐day ahead predictions are possible for all the series; (2) prediction accuracy decreases with increasing lead time for all the series, but the decrease is much more significant for suspended sediment concentration and suspended sediment load; and (3) the number of mechanisms dominantly governing the dynamics is three for each of the series. Copyright © 2005 John Wiley & Sons, Ltd.     

Periodicity of sediment load and runoff in the Yangtze River basin and possible impacts of climatic changes and human activities / Périodicité de la charge sédimentaire et de l'écoulement dans le bassin du Fleuve Yangtze et impacts possibles des changements climatiques et des activités humaines

Abstract

Periodicity of the runoff and the sediment load, and possible impacts from human activities and climatic changes, in the Yangtze River basin during 1963–2004 are discussed based on the monthly sediment and runoff data, and using the wavelet approach. Research results indicated that: (a) Sediment load changes are severely impacted by the different types of human activity (e.g. construction of water reservoirs, deforestation/afforestation); and the runoff variability is the direct result of climatic changes, e.g. the precipitation changes. (b) The impacts of human activity and climatic changes on the sediment load and runoff changes are greater in smaller river basins (e.g. the Jialingjiang River basin) than in larger river basins. The response of sediment load and runoff changes to the impacts of human activities and climatic changes are prompt and prominent in the Jialingjiang River basin relative to those in the mainstem of the Yangtze River basin. (c) Construction of the Three Gorges Dam has already had obvious impacts on the sediment transport process in the middle and lower Yangtze River basin, but shows no obvious influence on the runoff changes. Construction of the Three Gorges Dam will result in further re-adjustment of the scouring/filling process within the river channel in the middle and lower Yangtze River basin, and have corresponding effects on the altered sediment load because of the Dam's operation for the river channel, ecology, sustainable social economy and even the development of the Yangtze Delta. This will be of concern to local governments and policy makers.     

Sediment dynamics and temporal variation of runoff in the Yom River,Thailand

The Yom River is one of the four major sediment sources to the Chao Phraya River in Thailand. Human activities and changes in climate over the past six decades may have affected the discharge and sediment load to some extent. In the current study, the river discharge and sediment characteristics in the mainstream of the Yom River were investigated using the field observation data from 2011 to 2013 and the historical river flow and sediment data from 1954 to 2014 at six hydrological stations operated by the Royal Irrigation Department of Thailand (RID). The non-parametric Mann-Kendall test and double mass curve were used to analyze the sediment dynamics and temporal changes in the discharge of the Yom River. The results revealed that the sediment was mainly transported in suspension, and the bed-to-suspended sediment loads ratio varied between 0 and 0.05. The daily suspended sediment load (SSL) in the upper and middle basins had a strong correlation with the daily discharge and could be represented by power equations with coefficients of determination higher than 0.8. The daily suspended sediment load in the lower basin did not directly depend on the corresponding discharge because of the reduction in river slope and water diversion by irrigation projects. It also appeared that the river discharges and sediment loads were mainly influenced by climate variation (floods and droughts). Moreover, the average sediment transport of the upper, middle, and lower reaches were 0.57, 0.71, and 0.35 million t/y, respectively. The sediment load in the lower basin decreased more than 50% as a result of changes in the river gradient (from mountainous to floodplain areas). The results from sediment analysis also indicated that the construction of the Mae Yom Barrage, the longest diversion dam in Thailand, and land-use changes did not significantly affect the sediment load along the Yom River.     

A field investigation of the sediment transport characteristics of a high sediment load intermittent river in Taiwan

Most rivers in Taiwan are intermittent rivers with relatively steep slopes and carry rapid sediment‐laden flows during typhoon or monsoon seasons. A series of field experiments was conducted to collect suspended load data at the Tzu‐Chiang Bridge hydrological station of the lower Cho‐Shui River, which is a major river with the highest sediment yield in Taiwan. The river reach was aggrading with a high aspect ratio during the 1980s. Because of sand mining and extreme floods, it was incised and has had a relatively narrow main channel in recent years. The experimental results indicated that typical sediment transport equations can correctly predict the bed material load for low or medium sediment transport rates (e.g. less than about 1000 tons/day‐m). However, these equations far underestimate the bed material load for high sediment transport rates. The effects of cross‐sectional geometry change (i.e. river incision) and earthquakes on the sediment load were investigated in this study. An empirical sediment transport equation with consideration of the aspect ratio was also derived using the field data collected before and after river incision. Copyright © 2012 John Wiley & Sons, Ltd.     

Evaluation of temporal concentration profiles for ungauged rivers following pollution incidents

Abstract

In order to predict the impact of pollution incidents on rivers, it is necessary to predict the dispersion coefficient and the flow velocity corresponding to the discharge in the river of interest. This paper explores methods for doing this, particularly with a view to applications on ungauged rivers, i.e. those for which little hydraulic or morphometric data are available. An approach based on neural networks, trained on a wide-ranging database of optimized parameter values from tracer experiments and corresponding physical variables assembled for American and European rivers, is proposed. Tests using independent cases showed that the neural networks generally gave more reliable parameter estimates than a second-order polynomial regression approach. The quality of predictions of temporal concentration profiles was heavily influenced by the accuracy of the velocity prediction.

Citation Piotrowski, A. P., Napiorkowski, J. J., Rowinski, P. M. & Wallis, S. G. (2011) Evaluation of temporal concentration profiles for ungauged rivers following pollution incidents. Hydrol. Sci. J. 56(5), 883–894.     

Effects of placer mining on suspended sediment budget: case study of north of Russia’s Kamchatka Peninsula

Abstract

Placer mines are located in river valleys, along river benches, or along the pathways of ancient channels. Open-pit mining alters the stream hydrology and enhances sediment transport. The present study focuses on sediment transport in the area of the platinum placer mining located at the north of Russia’s Kamchatka Peninsula (Seynav-Galmoenan placer deposits). Based on hydrological field investigations, a conceptual model was derived to assess anthropogenic effects on the total sediment budget of rivers. The model illustrates key processes controlling sediment dynamics in the Vyvenka River basin. Field work included water-discharge and sediment-load measurements, assessment of annual channel change in rivers in mining site areas, and evaluation of the relative importance of sediment sources and transport processes. In this study, we estimated total sediment delivery from opencast placer mining of 60 t year^-1; the annual mass wasting rate ranges from 2 to 5.5 kg m^-2 year^-1, which is three orders of magnitude higher than from non-mined streams. Mass wasting dominates surface erosion on the hillslopes and produces significant wastewater effluents; however, erosion of the artificially stratified channel reaches is the primary contributor to the annual sediment yield of the mined rivers (21.4%).

Editor D. Koutsoyiannis

Citation Chalov, S.R., 2014. Effects of placer mining on suspended sediment budget: case study of north of Russia’s Kamchatka Peninsula. Hydrological Sciences Journal, 59 (5), 1081–1094.     

Reach-scale suspended sediment balance downstream from dams in a large Mediterranean river

Abstract

This paper presents a reach-scale sediment balance of a large impounded Mediterranean river (the lower Ebro, 1998–2008). Multi-temporal sediment storage and the influence of floods and tributaries on the sediment load were examined using continuous discharge and turbidity records. The mean annual suspended sediment load at the reach outlet (Xerta) is 0.12?×?10⁶ t, corroborating previous results. Suspended sediment concentrations were low (SSC_mean?=?13 mg L^-1), attaining a maximum of 274 mg L^-1. Erosion processes (channel-scour, bank erosion) are dominant, and net export of sediment occurs over the long term. Unexpectedly, ephemeral tributaries were found to contribute significantly: sediment delivered during torrential events attained 5% of the Ebro annual load, and was even larger than that in flushing flows. Overall, most of the suspended sediment load is transported by floods (up to 65% in some years). The results constitute basic information to underpin current management actions aiming to achieve the sustainability of the riverine and deltaic system.

Editor D. Koutsoyiannis; Associate editor D. Hughes

Citation Tena, A., Batalla, R.J. and Vericat, D., 2012. Reach-scale suspended sediment balance downstream from dams in a large Mediterranean river. Hydrological Sciences Journal, 57 (5), 831–849.     

Modelling daily suspended sediment of rivers in Turkey using several data-driven techniques / Modélisation de la charge journalière en matières en suspension dans des rivières turques à l'aide de plusieurs techniques empiriques

Abstract

The transport of sediment load in rivers is important with respect to pollution, channel navigability, reservoir filling, longevity of hydroelectric equipment, fish habitat, river aesthetics and scientific interest. However, conventional sediment rating curves cannot estimate sediment load accurately. An adaptive neuro-fuzzy technique is investigated for its ability to improve the accuracy of the streamflow—suspended sediment rating curve for daily suspended sediment estimation. The daily streamflow and suspended sediment data for four stations in the Black Sea region of Turkey are used as case studies. A comparison is made between the estimates provided by the neuro-fuzzy model and those of the following models: radial basis neural network (RBNN), feed-forward neural network (FFNN), generalized regression neural network (GRNN), multi-linear regression (MLR) and sediment rating curve (SRC). Comparison of results reveals that the neuro-fuzzy model, in general, gives better estimates than the other techniques. Among the neural network techniques, the RBNN is found to perform better than the FFNN and GRNN.     

Wavelet and Neuro‐fuzzy Conjunction Approach for Suspended Sediment Prediction

Simulation approaches employed in suspended sediment processes are important in the areas of water resources and environmental engineering. In the current study, neuro‐fuzzy (NF), a combination of wavelet transform and neuro‐fuzzy (WNF), multi linear regression (MLR), and the conventional sediment rating curve (SRC) models were considered for suspended sediment load (S) modeling in a gauging station in the USA. In the proposed WNF model, the discrete wavelet analysis was linked to a NF approach. To achieve this aim, the observed time series of river flow discharge (Q) and S were decomposed to sub time series at different scales by discrete wavelet transform. Afterwards, the effective sub time series were added together to obtain a useful Q and S time series for prediction. Eventually, the obtained total time series were imposed as inputs to the NF method for daily S prediction. The results illustrated that the predicted values by the proposed WNF model were in good agreement with the observed S values and gave better results than other models. Furthermore, the WNF model satisfactorily estimated the cumulative suspended sediment load and produced relatively reasonable predictions for extreme values of S, while NF, MLR, and SRC models provided unacceptable predictions.     

Design and evaluation of SVR,MARS and M5Tree models for 1, 2 and 3-day lead time forecasting of river flow data in a semiarid mountainous catchment

Accurate and reliable river flow forecasts attained with data-intelligent models can provide significant information about future water resources management. In this study we employed a 50-model ensemble of three data-driven predictive models, namely the support vector regression (SVR), multivariate adaptive regression spline (MARS) and M5 model tree (M5Tree) to forecast river flow data in a semiarid and ecologically significant mountainous region of Pailugou catchment in northwestern China. To attain stable and accurate forecast results, 50 different models were trained by randomly sampling the entire river flow data into 80% for training and 20% for testing subsets. To attain a complete evaluation of the ensemble-model based results, the global mean of six quantitative statistical performance evaluation measures: the coefficient of correlation (R), mean absolute relative error (MAE), root mean squared error (RMSE), Nash–Sutcliffe efficiency coefficient (NS), relative RMSE, and the Willmott’s Index (WI), and Taylor diagrams, including skill scores relative to a persistence model, were selected to assess the performances of the developed predictive models. The results indicated that all of the averaged R value attained was higher than 0.900 and all of the averaged NS values were higher than 0.800, representing good performance of the SVR, MARS and M5Tree models applied in the 1-, 2- and 3-day ahead modeling horizon, and this also accorded with the deductions made through an assessment of the Willmott’s Index. However, the M5Tree model outperformed both the SVR and MARS models (with NS?=?0.917 vs. 0.904 and 0.901 for 1-day, 0.893 vs. 0.854 and 0.845 for 2-day, and 0.850 vs. 0.828 and 0.810 for 3-day forecasting horizons, respectively), which was in concurrence with the high value of WI. Therefore, based on the ensemble of 50 models, the performance of the M5Tree can be considered as superior to the SVR and MARS models when applied in a problem of river flow forecasting at multiple forecast horizon. A detailed comparison of the overall performance of all three models evaluated through Taylor diagrams and boxplots indicated that the 1-day ahead forecasting results were more accurate for all of the predictive models compared to the 2- and 3-day ahead forecasting horizons. Data-intelligent models designed in this study indicate that the M5Tree method could successfully be explored for short-term river flow forecasting in semiarid mountainous regions, which may have useful implications in water resources management, ecological sustainability and assessment of river systems.     

Using ANNs to analyse effects of the Three Gorges Dam on sedimentation in Dongting Lake,China

Since a river system is a unidirectional network, the upstream influencing factors often interfere with those downstream. We quantify the effects of the TGD on the sediment exchange processes between the Yangtze River and Dongting Lake. Based on yearly sediment load data from 1981 to 2008, two multi-layer perceptron (MLP) models were constructed to predict the potential sedimentation in Dongting Lake without implementation of the TGD. The sediment discharge at Yichang station decreased from 622.5 to 61.1 Mt/year between 1981–1985 and 2003–2008, while the sedimentation in Dongting Lake reduced from 127.3 to 6.6 Mt/year. The MLP models indicate that only 27.9% of the decreased sediment load at Yichang station and 16.9% of that in Dongting Lake is caused by the TGD, while the rest is caused mostly by other upstream climate variations and anthropogenic impacts.

EDITOR A. Castellarin; ASSOCIATE EDITORN. Ilich