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.     

Magnitude and variation in the contribution of bank erosion to the suspended sediment load of the River Severn,UK

The wide range of studies describing the role of bank erosion in fluvial sediment supply have mostly lumped amounts of bank erosion into coarse temporal units, such as years. This paper investigates sediment yields from individual bank erosion events within the upper River Severn, UK (basin area 380 km²). Manual erosion pins and photo-electronic erosion pins were used to estimate bank erosion, and turbidity meters were used to determine suspended sediment transport. At the annual time-scale, the silt-clay fraction of bank-derived sediment accounted for an equivalent of 17 per cent of the suspended load, increasing to an average of 38 per cent at the monthly timescale, and then to an average of 64 per cent at the event timescale. This research highlighted that for an upland catchment, bank erosion was an important supply of suspended sediment, and that for some flood events bank erosion can supply more sediment than is transported. © 1997 John Wiley & Sons, Ltd.     

Temporal variability of suspended sediment sources in an alpine catchment combining river/rainfall monitoring and sediment fingerprinting

Influence of the rainfall regime on erosion and transfer of suspended sediment in a 905‐km² mountainous catchment of the southern French Alps was investigated by combining sediment monitoring, rainfall data, and sediment fingerprinting (based on geochemistry and radionuclide concentrations). Suspended sediment yields were monitored between October 2007 and December 2009 in four subcatchments (22–713 km²). Automatic sediment sampling was triggered during floods to trace the sediment origin in the catchment. Sediment exports at the river catchment outlet (330 ± 100 t km^‐2 yr^‐1) were mainly driven (80%) by widespread rainfall events (long duration, low intensities). In contrast, heavy, local and short duration storms, generated high peak discharges and suspended sediment concentrations in small upstream torrents. However, these upstream floods had generally not the capacity to transfer the sediment down to the catchment outlet and the bulk of this fine sediment deposited along downstream sections of the river. This study also confirmed the important contribution of black marls (up to 70%) to sediment transported in rivers, although this substrate only occupies c. 10% of the total catchment surface. Sediment exports generated by local convective storms varied significantly at both intra‐ and inter‐flood scales, because of spatial heterogeneity of rainfall. However, black marls/marly limestones contribution remained systematically high. In contrast, widespread flood events that generate the bulk of annual sediment supply at the outlet were characterized by a more stable lithologic composition and by a larger contribution of limestones/marls, Quaternary deposits and conglomerates, which corroborates the results of a previous sediment fingerprinting study conducted on riverbed sediment. Copyright © 2012 John Wiley & Sons, Ltd.     

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.     

Magnitude‐frequency characteristics of effective discharge for suspended sediment transport,Fraser River,British Columbia,Canada

This study analyses archival discharge and sediment concentration data (1965–1988), monitored by Water Survey of Canada, to examine suspended sediment transport rates and their relationship to effective discharge (Q_eff) based on daily discharge duration curves. Effective discharge was determined as the mid‐point of the discharge class transporting the greatest portion of the suspended sediment load (hence class‐based Q_eff). Results showed that the concept of effective discharge was applicable to the Fraser River basin where the average class‐based Q_eff occurred during 8·4% of the study period with individual values ranging from 0·03% to 16·1%. The durations of effective discharge classes ranged from 0·02% to 19·6% while the transport of 50% of total sediment loads ranged from 3% to 22% with an average of 14% of the time. Equations for predicting the class‐based Q_eff in the Fraser River basin from bankfull discharge and drainage area are presented. The observed variations among stations in sediment‐discharge regimes based on subjectively selected 20 discharge classes, seem to reflect the influence of sediment controlling factors such as geology, physiography, catchment size and land use practice in the basin. Future directions of research on applications of the effective discharge concept are explored. As a solution to the problem of lack of an objective method for determining the effective discharge, the effective discharge should be determined from event based assessments of sediment transport (event‐based Q_eff), avoiding any subjectivity in the selection of number of discharge classes used for its determination. In conclusion, it is proposed that continued use of the conventional method of determining Q_eff should cease. Copyright © 1999 John Wiley & Sons, Ltd.     

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.     

Predictability performance enhancement for suspended sediment in rivers:Inspection of newly developed hybrid adaptive neuro-fuzzy system model

Reliable modeling of river sediments transport is important as it is a defining factor of the economic viability of dams, the durability of hydroelectric-equipment, river susceptibility to pollution, suitability for navigation, and potential for aesthetics and fish habitat. The capability of a new machine learning model, fuzzy c-means based neuro-fuzzy system calibrated using the hybrid particle swarm optimization-gravitational search algorithm（ANFIS-FCM-PSOGSA） in improving the estimation accur...     

Suspended sediment and dissolved load budgets of two Amazonian rivers from the Guiana Shield: Maroni River at Langa Tabiki and Oyapock River at Saut Maripa (French Guiana)

This study presents the data collected within the framework of an Observatory of Research in Environment on the Amazonian Basin—the ORE HYBAM. It relates to the dissolved and solid loads of the two main rivers of French Guiana, the Maroni and Oyapock rivers, running on the Guiana Shield and draining respectively 64 230 and 24 630 km². The low coefficient of variation of the average annual flows of the two rivers indicates an inter‐annual hydrological stability probably related to the immediate vicinity of the Atlantic Ocean. The sedimentary load is mainly composed of quartz and kaolinite. The total suspended solid (TSS) concentrations are among the world's lowest values; they range from 1 to 130 mg l⁻¹ during the hydrological cycle, with averages of 22 and 12 mg l⁻¹, for the Maroni and Oyapock rivers, respectively. The seasonal variability of these values is significantly higher than that of hydrologic flows, but without simple relationship with the discharge. Water chemical composition of the two rivers indicates a very weak mineralization, very similar to that found in the Amazonian rivers running on the Brazilian and Guianese shields, and in the Congo River and its tributaries in the Central African Shield. Seasonal variations are observed in both basins; they correspond to higher concentrations during low water stage (from October to February) and to more diluted water during the flood, from April to July. A signature enriched in Cl⁻ is present at the Saut Maripa station on the Oyapock River indicating a more marked influence of the trade winds in this basin. The computation of atmospheric contributions to ions budget indicated a weak contribution for Ca²⁺ and Mg²⁺, which originates mainly from water‐rock interactions in both stations, while more than half of Na⁺ is derived from atmospheric inputs. Copyright © 2010 John Wiley & Sons, Ltd.     

Suspended sediment load estimation and the problem of inadequate data sampling: a fractal view

Suspended sediment load estimation at high resolutions is an extremely difficult task, because: (1) it depends on the availability of high‐resolution water discharge and suspended sediment concentration measurements, which are often not available; (2) any errors in the measurements of these two components could significantly influence the accuracy of suspended sediment load estimation; and (3) direct measurements are very expensive. The purpose of this study is to approach this sampling problem from a new perspective of fractals (or scaling), which could provide important information on the transformation of suspended sediment load data from one scale to another. This is done by investigating the possible presence of fractal behaviour in the daily suspended sediment load data for the Mississippi River basin (at St. Louis, Missouri). The presence of fractal behaviour is investigated using five different methods, ranging from general to specific and from mono‐fractal to multi‐fractal: (1) autocorrelation function; (2) power spectrum; (3) probability distribution function; (4) box dimension; and (5) statistical moment scaling function. The results indicate the presence of multi‐fractal behaviour in the suspended sediment load data, suggesting the possibility of transformation of data from one scale to another using a multi‐dimensional model. Copyright © 2005 John Wiley & Sons, Ltd.     

Effect of reservoir construction on suspended sediment load in a large river system: thresholds and complex response

Dam construction greatly alters the channel boundary of rivers, making the dammed river system a human‐controlled system. Based on hydrometric data in the upper Changjiang River basin, the change in behaviour of sediment transport of some dammed rivers was studied. As a result, some phenomena of threshold and complex response were found. When the coefficient (C_r,a) of actual runoff regulation by reservoirs, defined as the ratio of total capacity of reservoirs to annual runoff input, is smaller than 10%, suspended sediment load at Yichang station, the control station of the Changjiang River, shows a mild decreasing trend. When this coefficient becomes larger than 10%, suspended sediment load decreases sharply. The coefficient of 10% can be regarded as a threshold. The C_r,a of 10% is also a threshold, when the variation of suspended sediment concentration (SSC) with C_r,a at Yichang station is considered. The impacts of reservoir construction can be divided into several stages, including road construction, dam building and closure, water storage and sediment trapping. During these stages, some complex response was identified. At the station below the dam, SSC increases and reaches a maximum, and then declines sharply. This phenomenon was found on the main‐stem and several major tributaries of the upper Changjiang River. In the Minjiang River, where a series of dams were built successively, the response of SSC is more complicated. Copyright © 2010 John Wiley & Sons, Ltd.