Application of experimental design for the optimization of artificial neural network-based water quality model: a case study of dissolved oxygen prediction

A comprehensive review of water quality indices for lotic and lentic ecosystems

Freshwater resources play a pivotal role in sustaining life and meeting various domestic, agricultural, economic, and industrial demands. As such, there is a significant need to monitor the water quality of these resources. Water quality index (WQI) models have gradually gained popularity since their maiden introduction in the 1960s for evaluating and classifying the water quality of aquatic ecosystems. WQIs transform complex water quality data into a single dimensionless number to enable accessible communication of the water quality status of water resource ecosystems. To screen relevant articles, the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) method was employed to include or exclude articles. A total of 17 peer-reviewed articles were used in the final paper synthesis. Among the reviewed WQIs, only the Canadian Council for Ministers of the Environment (CCME) index, Irish water quality index (IEWQI) and Hahn index were used to assess both lotic and lentic ecosystems. Furthermore, the CCME index is the only exception from rigidity because it does not specify parameters to select. Except for the West-Java WQI and the IEWQI, none of the reviewed WQI performed sensitivity and uncertainty analysis to improve the acceptability and reliability of the WQI. It has been proven that all stages of WQI development have a level of uncertainty which can be determined using statistical and machine learning tools. Extreme gradient boosting (XGB) has been reported as an effective machine learning tool to deal with uncertainties during parameter selection, the establishment of parameter weights, and determining accurate classification schemes. Considering the IEWQI model architecture and its effectiveness in coastal and transitional waters, this review recommends that future research in lotic or lentic ecosystems focus on addressing the underlying uncertainty issues associated with the WQI model in addition to the use of machine learning techniques to improve the predictive accuracy and robustness and increase the domain of application.

Environmental Assessment of Potentially Toxic Elements Using Pollution Indices and Data-Driven Modeling in Surface Sediment of the Littoral Shelf of the Mediterranean Sea Coast and Gamasa Estuary, Egypt

Coastal environmental assessment techniques have evolved into one of the most important fields for the long-term development and management of coastal zones. So, the overall aim of the present investigation was to provide effective approaches for making informed decisions about the Gamasa coast sediment quality. Over a two-year investigation, sediment samples were meticulously collected from the Gamasa estuary and littoral shelf. The inductively coupled plasma mass spectra (ICP-MS) was used to the total concentrations of Al, Fe, Ti, Mg, Mn, Cu, P, V, Ba, Cr, Sr, Co, Ni, Zn, Pb, Zr, and Ce. Single elements environmental pollution indices including the geoaccumulation index (Igeo), contamination factor (CF), and enrichment factor (EF), as well as multi-elements pollution indices comprising the potential ecological risk index (RI), degree of contamination (Dc), and pollution load index (PLI) were used to assess the sediment and the various geo-environmental variables affecting the Mediterranean coastal system. Furthermore, the Dc, PLI, and RI were estimated using the random forest (RF) and Back-Propagation Neural Network (BPNN) depending on the selected elements. According to the Dc results, all the investigated sediment samples categories were considerably contaminated. Cr, Co, Ni, Cu, Zr, V, Zn, P, and Mn showed remarkable enrichment in sediment samples and were originated from anthropogenic sources based on the CF, EF, and Igeo data. Moreover, the RI findings revealed that all the samples tested pose a low ecologically risk. Meanwhile, based on PLI, 70% of the Gamasa estuary samples were polluted, while 93.75% of littoral shelf sediment was unpolluted. The BPNNs -PCs-CD-17 model performed the best and demonstrated a better association between exceptional qualities and CD. With R2 values of 1.00 for calibration (Cal.) and 1.00 for validation (Val.). The BPNNs -PCs-PLI-17 models performed the best in terms of measuring PLI with respective R2 values of 1.00 and 0.98 for the Cal. and Val. datasets. The findings showed that the RF and BPNN models may be used to precisely quantify the pollution indices (Dc, PLI, and RI) in calibration (Cal.) and validation (Val.) datasets utilizing potentially toxic elements of surface sediment.

Utilization of Pollution Indices, Hyperspectral Reflectance Indices, and Data-Driven Multivariate Modelling to Assess the Bottom Sediment Quality of Lake Qaroun, Egypt

Assessing the environmental hazard of potentially toxic elements in bottom sediments has always been based entirely on ground samples and laboratory tests. This approach is remarkably accurate, but it is slow, expensive, damaging, and spatially constrained, making it unsuitable for monitoring these parameters effectively. The main goal of the present study was to assess the quality of sediment samples collected from Lake Qaroun by using different groups of spectral reflectance indices (SRIs), integrating data-driven (Artificial Neural Networks; ANN) and multivariate analysis such as multiple linear regression (MLR) and partial least square regression (PLSR). Jetty cruises were carried out to collect sediment samples at 22 distinct sites over the entire Lake Qaroun, and subsequently 21 metals were analysed. Potential ecological risk index (RI), organic matter (OM), and pollution load index (PLI) of lake’s bottom sediments were subjected to evaluation. The results demonstrated that PLI showed that roughly 59% of lake sediments are polluted (PLI > 1), especially samples of eastern and southern sides of the lake’s central section, while 41% were unpolluted (PLI < 1), which composed samples of the western and western northern regions. The RI’s findings were that all the examined sediments pose a very high ecological risk (RI > 600). It is obvious that the three band spectral indices are more efficient in quantifying different investigated parameters. The results showed the efficiency of the three tested models to predict OM, PLI, and RI, revealing that the ANN is the best model to predict these parameters. For instance, the determination coefficient values of the ANN model of calibration datasets for predicting OM, PLI, and RI were 0.999, 0.999, and 0.999, while they were 0.960, 0.897, and 0.853, respectively, for the validation dataset. The validation dataset of the PLSR produced R2 values higher than with MLR for predicting PLI and RI. Finally, the study’s main conclusion is that combining ANN, PLSR, and MLR with proximal remote sensing could be a very effective tool for the detection of OM and pollution indices. Based on our findings, we suggest the created models are easy tools for forecasting these measured parameters.

Assessment of Water Quality in Lake Qaroun Using Ground-Based Remote Sensing Data and Artificial Neural Networks

Monitoring and managing water quality parameters (WQPs) in water bodies (e.g., lakes) on a large scale using sampling-point techniques is tedious, laborious, and not highly representative. Hyperspectral and data-driven technology have provided a potentially valuable tool for the precise measurement of WQPs. Therefore, the objective of this work was to integrate WQPs, derived spectral reflectance indices (published spectral reflectance indices (PSRIs)), newly two-band spectral reflectance indices (NSRIs-2b) and newly three-band spectral indices (NSRIs-3b), and artificial neural networks (ANNs) for estimating WQPs in Lake Qaroun. Shipboard cruises were conducted to collect surface water samples at 16 different sites throughout Lake Qaroun throughout a two-year study (2018 and 2019). Different WQPs, such as total nitrogen (TN), ammonium (NH4+), orthophosphate (PO43−), and chemical oxygen demand (COD), were evaluated for aquatic use. The results showed that the highest determination coefficients were recorded with the NSRIs-3b, followed by the NSRIs-2b, and then followed by the PSRIs, which produced lower R2 with all tested WQPs. The majority of NSRIs-3bs demonstrated strong significant relationships with three WQPs (TN, NH4+, and PO43−) with (R2 = 0.70 to 0.77), and a moderate relationship with COD (R2 = 0.52 to 0.64). The SRIs integrated with ANNs would be an efficient tool for estimating the investigated four WQPs in both calibration and validation datasets with acceptable accuracy. For examples, the five features of the SRIs involved in this model are of great significance for predicting TN. Its outputs showed high R2 values of 0.92 and 0.84 for calibration and validation, respectively. The ANN-PO43−VI-17 was the highest accuracy model for predicting PO43− with R2 = 0.98 and 0.89 for calibration and validation, respectively. In conclusion, this research study demonstrated that NSRIs-3b, alongside a combined approach of ANNs models and SRIs, would be an effective tool for assessing WQPs of Lake Qaroun.

A Review of the Artificial Neural Network Models for Water Quality Prediction

Water quality prediction plays an important role in environmental monitoring, ecosystem sustainability, and aquaculture. Traditional prediction methods cannot capture the nonlinear and non-stationarity of water quality well. In recent years, the rapid development of artificial neural networks (ANNs) has made them a hotspot in water quality prediction. We have conducted extensive investigation and analysis on ANN-based water quality prediction from three aspects, namely feedforward, recurrent, and hybrid architectures. Based on 151 papers published from 2008 to 2019, 23 types of water quality variables were highlighted. The variables were primarily collected by the sensor, followed by specialist experimental equipment, such as a UV-visible photometer, as there is no mature sensor for measurement at present. Five different output strategies, namely Univariate-Input-Itself-Output, Univariate-Input-Other-Output, Multivariate-Input-Other(multi), Multivariate-Input-Itself-Other-Output, and Multivariate-Input-Itself-Other (multi)-Output, are summarized. From results of the review, it can be concluded that the ANN models are capable of dealing with different modeling problems in rivers, lakes, reservoirs, wastewater treatment plants (WWTPs), groundwater, ponds, and streams. The results of many of the review articles are useful to researchers in prediction and similar fields. Several new architectures presented in the study, such as recurrent and hybrid structures, are able to improve the modeling quality of future development.

A water quality prediction method based on the multi-time scale bidirectional long short-term memory network

As an important factor affecting the mangrove wetland ecosystem, water quality has become the focus of attention in recent years. Therefore, many studies have focused on the prediction of water quality to help establish a regulatory framework for the assessment and management of water pollution and ecosystem health. To make a more accurate and comprehensive forecast analysis of water quality, we propose a method for water quality prediction based on the multi-time scale bidirectional LSTM network. In the method, we improve data integrity and data volume through data preprocessing. And the network processes input data forward and backward and considers the dependencies at multiple time scales. Besides, we use the Box-Behnken experimental design method to adjust hyper-parameters in the process of modeling. In this study, we apply this method to the water quality prediction research of Beilun Estuary, and the performance of our proposed model is evaluated and compared with other models. The experiment results show that this model has better performance in water quality prediction than that of using LSTM or bidirectional LSTM alone. Graphical Abstract Schematic of research work.

Optimizing a coastal monitoring network using a water-quality response grid (WRG)-based sampling design for improved reliability and efficiency

Marine monitoring in Bohai Sea is delivered within three networks by lacking appropriate sampling and assessment methodologies. Water-quality response grid (WRG)-based sampling design using optimization and multi-factors assessment can reliably detect a variety of environmental impacts. Which includes 5 steps: selects environmental reference factors, divides the sampling grid, sets the initial stations, optimizes the sampling stations, and assesses the proposed network's reproducibility and efficiency. We applied this method to the Bohai Sea, the networks proposed here have 225 stations for optimized special surveys (OSS) and 181 stations for optimized operational monitoring (OOM), accounting for 46.5% and 37.4% of the original station totals, respectively. Besides, the reproducibility and efficiency index (REI) of OSS and OOM stations approximately 15.4% and 13.3% higher than three current monitoring networks on average among multi-factors in 4 seasons. Thus, the method can improve the reproducibility, efficiency and land-sea spatial matching of monitoring network.