Artificial neural network modeling of the river water quality—A case study

The paper describes the training, validation and application of artificial neural network (ANN) models for computing the dissolved oxygen (DO) and biochemical oxygen demand (BOD) levels in the Gomti river (India). Two ANN models were identified, validated and tested for the computation of DO and BOD concentrations in the Gomti river water. Both the models employed eleven input water quality variables measured in river water over a period of 10 years each month at eight different sites. The performance of the ANN models was assessed through the coefficient of determination (R²) (square of the correlation coefficient), root mean square error (RMSE) and bias computed from the measured and model computed values of the dependent variables. Goodness of the model fit to the data was also evaluated through the relationship between the residuals and model computed values of DO and BOD. The model computed values of DO and BOD by both the ANN models were in close agreement with their respective measured values in the river water. Relative importance and contribution of the input variables to the model output was evaluated through the partitioning approach. The identified ANN models can be used as tools for the computation of water quality parameters.     

Comparing discriminant analysis, neural networks and logistic regression for predicting species distributions: a case study with a Himalayan river bird

We assessed the occurrence of a common river bird, the Plumbeous Redstart Rhyacornis fuliginosus, along 180 independent streams in the Indian and Nepali Himalaya. We then compared the performance of multiple discrimant analysis (MDA), logistic regression (LR) and artificial neural networks (ANN) in predicting this species’ presence or absence from 32 variables describing stream altitude, slope, habitat structure, chemistry and invertebrate abundance. Using the entire data (=training set) and a threshold for accepting presence in ANN and LR set to P≥0.5, ANN correctly classified marginally more cases (88%) than either LR (83%) or MDA (84%). Model performance was assessed from two methods of data partitioning. In a ‘leave-one-out’ approach, LR correctly predicted more cases (82%) than MDA (73%) or ANN (69%). However, in a holdout procedure, all the methods performed similarly (73–75%). All methods predicted true absence (i.e. specificity in holdout: 81–85%) better than true presence (i.e. sensitivity: 57–60%). These effects reflect species’ prevalence (=frequency of occurrence), but are seldom considered in distribution modelling. Despite occurring at only 36% of the sites, Plumbeous Redstarts are one of the most common Himalayan river birds, and problems will be greater with less common species. Both LR and ANN require an arbitrary threshold probability (often P=0.5) at which to accept species presence from model prediction. Simulations involving varied prevalence revealed that LR was particularly sensitive to threshold effects. ROC plots (received operating characteristic) were therefore used to compare model performance on test data at a range of thresholds; LR always outperformed ANN. This case study supports the need to test species’ distribution models with independent data, and to use a range of criteria in assessing model performance. ANN do not yet have major advantages over conventional multivariate methods for assessing bird distributions. LR and MDA were both more efficient in the use of computer time than ANN, and also more straightforward in providing testable hypotheses about environmental effects on occurrence. However, LR was apparently subject to chance significant effects from explanatory variables, emphasising the well-known risks of models based purely on correlative data.     

One-dimensional numerical modelling of dam-break waves over movable beds: application to experimental and field cases

This paper reports a numerical study on dam-break waves over movable beds. A one-dimensional (1-D) model is built upon the Saint-Venant equations for shallow water waves, the Exner equation of sediment mass conservation and a spatial lag equation for non-equilibrium sediment transport. The set of governing equations is solved using an explicit finite difference scheme. The model is tested in various idealized experimental cases, with fairly good agreement between the numerical predictions and measurements. Discrepancies are observed at the earlier stage of the dam-break wave and around the dam location due to no vertical velocity component being taken into account. Sensitivity tests confirm that the friction coefficient is an important parameter for the evaluation of sediment transport processes operating during a dam-break wave. The influence of the non-equilibrium adaptation length (or the lag distance) is negligible on the wavefront celerity and weak on the free surface and bed profiles, which indicates that one may ignore the spatial lag effect in dam-break wave studies. Finally, the simulation of the Lake Ha!Ha! dyke-break flood event shows that the model can provide relevant results if a convenient formula for computing the sediment transport capacity and an appropriate median grain diameter of riverbed material are selected.     

河流水体中重金属形态模型研究

本文提出了一种重金属形态模型,用于模拟水体中的水溶态、颗粒态和底泥、悬浮物的活动态和非活动态。以乐安江水体中铜污染为背景,计算结果与现场调查分析的数据比较表明,两者基本吻台,说明模型基本上是合理的,概括了重金属污染的重要过程,参数的选择反映了河流的实际情况。     

The application of artificial neural networks to flow and phosphorus dynamics in small streams on the Boreal Plain,with emphasis on the role of wetlands

Historically, management strategies in Canada's boreal forest have focused on forest polygons and terrestrial biodiversity to address ecological considerations in forest management. The Forest Watershed and Riparian Disturbance (FORWARD) project examines the problem from a watershed perspective rather than a forest polygon viewpoint. The main objective of this study was to devise an artificial neural network (ANN) modeling tool that can predict flow and total phosphorus (TP) concentration for ungauged watersheds (where daily flow is not monitored). This dictates that all inputs should be easily accessed via a public domain database, like the Environment Canada weather database, without the need to install flow gauges in each modeled watershed. Daily flow and TP concentration for two of the project watersheds were modeled using ANNs. The two watersheds (1A Creek, 5.1 km² and Willow Creek, 15.6 km²) were chosen to reflect variations in wetland area and composition in the study area. Flow was modeled with a feed-forward multilayer perceptron ANN trained with the error back-propagation algorithm. Simulated values for flow were then used, as inputs, to model TP concentration using the same neural networks algorithm. One hidden layer with three slabs; each operating with a different activation function was utilized to simulate the conceptual differences between base flow, snowmelt, and storm events. Time domain analysis was conducted to identify possible model time-lagged inputs reflecting the time dependency of the modeled variables. Spectral analysis was used to address data hystereses. Our results highlight the capabilities of ANN in modeling complex ecosystems and highly correlated variables. Results also indicated that more research towards the phosphorus dynamics in wetlands is required to better represent the impact of wetland area and composition on the water-phase phosphorus in ANN modeling.     

Simulation-based optimization of in-stream structures design: rock vanes

We employ a three-dimensional coupled hydro-morphodynamic model, the Virtual Flow Simulator (VFS-Geophysics) in its Unsteady Reynolds Averaged Navier–Stokes mode closed with \(k-\omega\) model, to simulate the turbulent flow and sediment transport in large-scale sand and gravel bed waterways under prototype and live-bed conditions. The simulation results are used to carry out systematic numerical experiments to develop design guidelines for rock vane structures. The numerical model is based on the Curvilinear Immersed Boundary approach to simulate flow and sediment transport processes in arbitrarily complex rivers with embedded rock structures. Three validation test cases are conducted to examine the capability of the model in capturing turbulent flow and sediment transport in channels with mobile-bed. Transport of sediment materials is handled using the Exner equation coupled with a transport equation for suspended load. Two representative meandering rivers, with gravel and sand beds, respectively, are selected to serve as the virtual test-bed for developing design guidelines for rock vane structures. The characteristics of these rivers are selected based on available field data. Initially guided by existing design guidelines, we consider numerous arrangements of rock vane structures computationally to identify optimal structure design and placement characteristics for a given river system.     

Hierarchical modeling of the dilute transport of suspended sediment in open channels

We propose, discuss and validate a theoretical and numerical framework for sediment-laden, open-channel flows which is based on the two-fluid-model (TFM) equations of motion. The framework models involve mass and momentum equations for both phases (sediment and water) including the interactive forces of drag, lift, virtual mass and turbulent dispersion. The developed framework is composed by the complete two-fluid model (CTFM), a partial two-fluid model (PTFM), and a standard sediment-transport model (SSTM). Within the umbrella of the Reynolds-Averaged Navier-Stokes (RANS) equations, we apply K–ε type closures (standard and extended) to account for the turbulence in the carrier phase (water). We present the results of numerical computations undertaken by integrating the differential equations over control volumes. We address several issues of the theoretical models, especially those related to coupling between the two phases, interaction forces, turbulence closure and turbulent diffusivities. We compare simulation results with various recent experimental datasets for mean flow variables of the carrier as well as, for the first time, mean flow of the disperse phase and turbulence statistics. We show that most models analyzed in this paper predict the velocity of the carrier phase and that of the disperse phase within 10% of error. We also show that the PTFM provides better predictions of the distribution of sediment in the wall-normal direction as opposed to the standard Rousean profile, and that the CTFM is by no means superior to the PTFM for dilute mixtures. We additionally report and discuss the values of the Schmidt number found to improve the agreement between predictions of the distribution of suspended sediment and the experimental data.     

Spatial variability in characteristics and trace metal levels in sediments from watercourses

Assessing sediment quality requires sampling designs which address variability. Often however, only one sample or composited sample is collected at a sample site. Existing studies on sediment variability or sampling strategies primarily concern lake or marine sediments. Two rivers and one canal were sampled to assess sediment variability. Further, it was determined if sediment contamination in running water could be predicted by using visual criteria and/or knowledge of the presence of depositional and erosion areas. Metal concentrations and sediment characteristics were measured in different visually distinct areas of the river. At all sample sites the coefficient of variance was relatively high for most sediment characteristics (1.4‐ > 100%) and metal levels (15.3–63.6%). In one river the majority of sediment characteristics variation was between two sediment types and detected by visually distinct differences. Significant differences in cadmium and zinc concentrations were also detected. Contrary to what was expected, cadmium, and zinc levels were highest in the coarse fraction. No differences were found in the second river. In the canal a greater concentration of fine grained sediments and metals were found in the deep areas. It was not possible to predict sediment areas with the highest levels of metal contamination using visual criteria or knowledge of the erosion and sedimentation pattern of the river.     

Idealized study on cohesive sediment flux by tidal asymmetry

The flux of cohesive sediment in an estuary is determined by many factors, including tidal asymmetry, wave effect, fluvial influence, phase difference between tidal velocity and tidal level fluctuations, sediment properties, flocculation, bed erodibility, bathymetry effect and other nonlocal effects. Our capability in predicting sediment fluxes in tide-dominant environments is critical to the morphodynamics and water quality of estuaries. Due to the difficulties in carrying out detailed measurement of sediment flux with high spatial and temporal resolutions, an one-dimensional-vertical (1DV) numerical model for cohesive sediment transport, previously verified and calibrated with field measured cohesive sediment concentration data, is utilized here to study some of the aforementioned factors in affecting tidal-driven sediment fluxes in idealized condition. Tidal-averaged sediment flux is shown to be correlated with tidal velocity skewness with a linear relationship. This linear relationship is different from that of non-cohesive sediment and it is demonstrated here to be mainly due to variable critical shear stress implemented for the mud bed in order to parameterize consolidation. The reason that tidal velocity skewness causes tidal-averaged residual sediment transport is shown to be due to nonlinear intra-tidal interactions between flow velocity and sediment concentration. Moreover, the effects of nonlinear intra-tidal interaction between tidal velocity and tidal level fluctuations is shown to mainly cause seaward transport, which is the most significant under progressive wave system (phase difference 90°) and almost negligible for standing wave system (phase difference 0°).     

A model diagnostic study of age of river-borne sediment transport in the tidal York River Estuary

As nutrients and organic matters are transported preferentially in an adsorbed state and tend to bind to the sediments, sediment transport plays an important role on eutrophication processes in the estuaries. The timescale of sediment transport is of significance for studying the retention of pollutants and eutrophication processes in the estuaries. Unlike transport of dissolved substances that is mainly controlled by advection and diffusion processes, the sediment transport is significantly affected by the intermittent settling and resuspension processes. A three-dimensional model with suspended sediment transport was utilized to investigate the transport timescale of river-borne sediment in the tidal York River Estuary. The results indicate that river discharge dominantly determines the age of river-borne sediment in the estuary. High river discharge results in a low sediment age compared to that under mean flow. The intermittent effects of settling and resuspension events greatly affect the river-borne sediment age. Both settling velocity and critical shear stress are shown to be key parameters in determining the sediment transport timescale. The sediment age decreases as settling velocity and/or critical shear stress decrease, while it increases with the increase of settling velocity that prevents the sediment to be transported out of the estuary.     

Neural network modelling of coastal algal blooms

An artificial neural network (ANN), a data driven modelling approach, is proposed to predict the algal bloom dynamics of the coastal waters of Hong Kong. The commonly used back-propagation learning algorithm is employed for training the ANN. The modeling is based on (a) comprehensive biweekly water quality data at Tolo Harbour (1982–2000); and (b) 4-year set of weekly phytoplankton abundance data at Lamma Island (1996–2000). Algal biomass is represented as chlorophyll-a and cell concentration of Skeletonema at the two locations, respectively. Analysis of a large number of scenarios shows that the best agreement with observations is obtained by using merely the time-lagged algal dynamics as the network input. In contrast to previous findings with more complicated neural networks of algal blooms in freshwater systems, the present work suggests the algal concentration in the eutrophic sub-tropical coastal water is mainly dependent on the antecedent algal concentrations in the previous 1–2 weeks. This finding is also supported by an interpretation of the neural networks’ weights. Through a systematic analysis of network performance, it is shown that previous reports of predictability of algal dynamics by ANN are erroneous in that ‘future data’ have been used to drive the network prediction. In addition, a novel real time forecast of coastal algal blooms based on weekly data at Lamma is presented. Our study shows that an ANN model with a small number of input variables is able to capture trends of algal dynamics, but data with a minimum sampling interval of 1 week is necessary. However, the sufficiency of the weekly sampling for real time predictions using ANN models needs to be further evaluated against longer weekly data sets as they become available.     

3D numerical modelling of turbidity currents

During floods, the density of river water usually increases due to a subsequent increase in the concentration of the suspended sediment that the river carries, causing the river to plunge underneath the free surface of a receiving water basin and form a turbidity current that continues to flow along the bottom. The study and understanding of such complex phenomena is of great importance, as they constitute one of the major mechanisms for suspended sediment transport from rivers into oceans, lakes or reservoirs. Unlike most of the previous numerical investigations on turbidity currents, in this paper, a 3D numerical model that simulates the dynamics and flow structure of turbidity currents, through a multiphase flow approach is proposed, using the commercial CFD code FLUENT. A series of numerical simulations that reproduce particular published laboratory flows are presented. The detailed qualitative and quantitative comparison of numerical with laboratory results indicates that apart from the global flow structure, the proposed numerical approach efficiently predicts various important aspects of turbidity current flows, such as the effect of suspended sediment mixture composition in the temporal and spatial evolution of the simulated currents, the interaction of turbidity currents with loose sediment bottom layers and the formation of internal hydraulic jumps. Furthermore, various extreme cases among the numerical runs considered are further analyzed, in order to identify the importance of various controlling flow parameters.     

Use of genetic algorithms to select input variables in decision tree models for the prediction of benthic macroinvertebrates

Predicting freshwater organisms based on machine learning is becoming more and more reliable due to the availability of appropriate datasets, advanced modelling techniques and the continuously increasing capacity of computers. A database consisting of measurements collected at 360 sampling sites in non-navigable watercourses in Flanders was applied to predict the absence/presence of benthic macroinvertebrate taxa by means of decision trees. The measured variables were a combination of physical–chemical (temperature, pH, dissolved oxygen concentration, conductivity, total organic carbon, Kjeldahl nitrogen and total phosphorus), structural (granulometric analysis of the sediment, width, depth and flow velocity of the river) and two ecotoxicological variables. The predictive power of decision trees was assessed on the basis of the number of Correctly Classified Instances (CCI). A genetic algorithm was introduced to compare the predictive power of different sets of input variables for the decision trees. The number of input variables was reduced from 15 to 2–8 variables without affecting the predictive power of the decision trees significantly. Furthermore, reducing the number of input variables allowed to ease the identification of general data trends.     

Advent of sheet flow in suction affected alluvial channels

The presence of suction (flow of water from channel to ground water) affects the channel hydrodynamics and increases the bed shear stress. At high bed shear stress in alluvial channels made of the non-cohesive material, sediment transport occurs as sheet flow layer of high sediment concentration. The sediment transport in the form of sheet flow has been observed in the present study when suction was applied to the non-transporting channels designed on incipient motion condition. The erosion of the channel banks contributed to the sheet flow because of the increased channel bed shear stress. An empirical relation for the thickness of sheet flow layer has been developed which includes suction as independent parameter along with others.     

Effect of streambed bacteria release on E. coli concentrations: Monitoring and modeling with the modified SWAT

Streambed sediment has been attracting attention as a reservoir for bacteria, including pathogenic strains. Soil and Water Assessment Tool (SWAT) has been augmented with a bacteria transport subroutine in SWAT2005 in which bacteria die-off is the only in-stream process. The purpose of this study was to develop the partial model of sediment-associated bacteria transport in stream and to evaluate the potential significance of streambed Escherichia coli (E. coli) release and deposition within the SWAT microbial water quality simulations. Streambed E. coli release and deposition were simulated based on the sediment resuspension and deposition modules in SWAT. The modified SWAT was applied to the Little Cove Creek watershed, Pennsylvania, which has forestry and dairy pasture landuses. Temporal changes in sediment E. coli concentrations were derived from monitoring data rather than from a streambed bacteria population model. Sensitivity analyses and calibrations were separately conducted for both hydrologic and bacteria parameters. Hydrologic calibration characterized soils in the watershed as pervious and thus the surface runoff was only moderately contributing to the streamflow. However, the surface runoff carried large numbers of E. coli to the stream, and sediment resuspension contributed to the persistent concentration of E. coli in stream water. Although the uncertainty of E. coli concentrations in streambed sediments and from wildlife probably affected the performance of the modified SWAT model, this study qualitatively confirmed the significance of modeling E. coli release from streambed and deposition for the SWAT microbial water quality simulations. Further developments should include modeling dynamics of bacteria populations within streambeds.     

Secondary flow within a river bed and contaminant transport

We have developed a numerical method to simulate the transport of non-sorbing contaminants within the sediment layer of a stream and the leaching of these contaminants in the steam. Typical stream bottom surfaces are uneven with triangularly shaped undulation forms. The flow of the water above such triangular surfaces causes external pressure changes that result in a “pumping effect” and a secondary flow within the sediment. The latter causes a significant contaminant advection within the sediment layer. The flow field in the porous sediment layer is obtained by solving numerically Darcy’s equations. The unsteady mass transfer equation is solved by using a finite-difference method with an up-wind scheme. The effects of parameters, such as channel slope, hydraulic head and dispersion, are studied by quantitatively comparing the numerical results of the total mass flow rate from the contaminant source, the concentration front propagation, and the contaminant mass flow rate into the water column. The “pumping effect,” increases the flow in the vertical direction and, thus, enhances the vertical advective mass transport of the contaminant. This bedform-shape induced flow is largely responsible for the mass transfer of contaminants into the water column. The numerical results also show that the mechanical dispersion inside the sediment bed will significantly increase the contaminant mass flow rate from the source.     

Suspended particulate matter dynamics in a particle framework

Suspended particulate matter (SPM) dynamics in ocean models are usually treated with an advection–diffusion equation for one or more sediment size classes coupled to the hydrodynamical part of the model. Numerical solution of these additional partial differential equations unavoidably introduces numerical diffusion, i.e. in the case of sharp gradients the possible occurrence of artificial oscillations and non-positivity. A Lagrangian particle-tracking model has been developed to simulate short-term SPM dynamics. Modelling individual sediment particles allows a straightforward physical interpretation of the processes. The tracking of large numbers of individual and independent particles (up to 25 million in total in a single sediment class) can be achieved on high performance computer clusters, due to efficient parallelisation of particle tracking. The movement of the particles is described by a stochastic differential equation, which is consistent with the advection–diffusion equation. Here, the concentration profile is represented by a set of independent moving particles, which are advected according to the 3D velocity field, while the diffusive displacements of the particles are sampled from a random distribution, which is related to the eddy diffusivity field. To account for erosion a new parameterisation is proposed. Three numerical particle tracking schemes (EULER, MILSTEIN and HEUN) are presented and validated in idealised test cases. Finally, the particle tracking algorithms are applied to a realistic scenario, a severe winter storm in the East Frisian Wadden Sea (southern North Sea). The comparison with observations and an Eulerian SPM transport model seems to indicate a somewhat better fidelity of the Lagrangian approach.     

Empirical equation for transverse dispersion coefficient based on theoretical background in river bends

There are different approaches to estimating the transverse dispersion coefficient in river mixing. Theoretical approaches have derived the dispersion coefficient from the concept of shear flow, which has dominant effects on the transverse mixing. Empirical approaches have developed an equation using the hydraulic and geometric data of rivers through dimensional analysis and regression techniques. These two equations interact closely with each other. For example, the complicated theoretical equation can be simplified by empirical approaches, and the functional relationships of the empirical equation can be derived from theoretical bases. In this study, a new empirical equation for the transverse dispersion coefficient has been developed based on the theoretical background in river bends. As a regression method, the least-square iterative method was used because the equation was a nonlinear model. The estimated dispersion coefficients derived by the new equation were compared with observed transverse dispersion coefficients acquired from natural rivers and coefficients calculated by the other existing empirical equations. From a comparison of the existing transverse dispersion equations and the proposed equation, it appears that the behavior of the existing formula in a relative sense is very much dependent on the flow condition and the river geometry. Moreover, the proposed equation does not vary widely according to variation of flow conditions. Also, it was revealed that the equation proposed in this study becomes an asymptotic curve as the curvature effect increases.     

Occurrence and distribution of micro- and mesoplastics in the high-latitude nature reserve,northern China

• The first study on micro(meso)plastics (MMPs) in the Liaohe River Reserve is reported. • Diverse MMP were detected in surface water and sediment at all 32 sites. • The abundance of MMPs decreased in the course of the river. • The MMPs abundance in water is significant association with the county population. Microplastics pollution has received growing attention worldwide in recent years. However, data on microplastics in the freshwater environment are still limited, especially in high-latitude nature reserves in Northern China. The first study on microplastic pollution in the Liaohe River Reserve in Northern China is reported here, and mesoplastics were also incorporated. Surface water and sediment samples were collected from 32 sites along the nature reserve. The abundance, type, shape, color, and size of micro- and mesoplastics were measured using density extraction, optical microscopy, and FTIR spectroscopy. The data showed that diverse micro- and mesoplastics were found widespread in the 32 sites, and the average abundance of these plastics was 0.11±0.04 10⁻² items/L in surface water and 62.29±54.30 items/kg in sediment. Moreover, 70% and 66% were smaller than 2000 μm in surface water and sediment, respectively. Fiber accounted for 91.86% in surface water and 43.48% in sediment, indicating that the major source of micro- and mesoplastics in the Liaohe River Reserve may be domestic sewage and aquaculture. A total of 16 and 27 polymers were identified in surface water and sediment, respectively, and mostly consisted of rayon, polyester, polystyrene, and poly(ethylene terephthalate). Moreover, both the risk index and the pollution load index demonstrated a low risk of micro- and mesoplastics in surface water and sediment in the Liaohe River Reserve.