Search

In agricultural regions prone to dust storms, heavy metal contamination of soil and crops from airborne particulates poses significant risks to food safety and public health. This study has assessed the potential of machine learning models for predicting concentrations of toxic heavy metals like arsenic, chromium, and lead in dust from the agricultural Sistan region of southeastern Iran. This region experiences frequent dust storms mobilizing particulates from local dried lakes onto agricultural lands. The metals including nickel, copper, magnesium, cobalt, zinc, chromium, arsenic, and lead were measured in summer dust samples during 2012-2018 across 15 stations. Two hybrid models were developed combining group method of data handling (GMDH) and support vector regression (SVR) machine learning with harmony search optimization (H) so as to predict toxic metals arsenic, chromium, and lead using nickel, copper, magnesium, cobalt, and zinc inputs. Standard error maps were uncertainty higher in southern and western areas, and they are most impacted by dust storms. Results demonstrated that the hybrid GMDH + H and SVR + H models improved the accuracy of individual GMDH and SVR models in predicting heavy metals. The GMDH + H model performed the best for the lead with an agreement index (d-index) of 0.98, root mean square error (RMSE) of 2.87 ppm, normalized RMSE (NRMSE) of 0.12, and coefficient of determination (RR) of 0.96. The SVR + H model showed the highest accuracy for arsenic and chromium, obtaining d-index 0.96, RMSE 0.47 ppm, NRMSE 0.09, and RR 0.92 for arsenic, and d-index 0.96, RMSE 11.24 ppm, NRMSE 0.16, and RR 0.93 for chromium. Taylor's diagram and heatmap analysis confirmed the superiority of the hybrid techniques. This work demonstrates the utility of state-of-the-art computing for addressing complex environmental health challenges., (© 2024. The Author(s), under exclusive licence to Springer-Verlag GmbH Germany, part of Springer Nature.)

Biochemical oxygen demand (BOD) is one of the most important parameters used for water quality assessment. Alternative methods are essential for accurately prediction of this parameter because the traditional method in predicting the BOD is time-consuming and it is inaccurate due to inconstancies in microbial multiplicity. In this study, the applicability of four hybrid neuro-fuzzy (ANFIS) methods, ANFIS with genetic algorithm (GA), ANFIS with particle swarm optimization (PSO), ANFIS with sine cosine algorithm (SCA), and ANFIS with marine predators algorithm (MPA), was investigated in predicting BOD using distinct input combinations such as potential of hydrogen (pH), dissolved oxygen (DO), electrical conductivity (EC), water temperature (WT), suspended solids (SS), chemical oxygen demand (COD), total nitrogen (TN), and total phosphorus (T-P) acquired from two river stations, Gongreung and Gyeongan, South Korea. The applicability of multi-variate adaptive regression spline (MARS) in determination of the best input combination was examined. The ANFIS-MPA was found to be the best model with the lowest root mean square error and mean absolute error and the highest determination coefficient. It improved the root mean square error of ANFIS-PSO, ANFIS-GA, and ANFIS-SCA models by 13.8%, 12.1%, and 6.3% for Gongreung Station and by 33%, 25%, and 6.3% for Gyeongan Station in the test stage, respectively., (© 2023. The Author(s).)

Due to its heterogeneous and complex nature, groundwater modeling needs great effort to quantify the aquifer, a crucial tool for policymakers and hydrogeologists to understand the variations in groundwater levels (GWL). This study proposed a set of supervised machine learning (ML) models to delineate the GWL changes in the Zarand-Saveh complex aquifer in Iran using 15-year (2005-2020) monthly dataset. The wavelet transform (WT) procedure was also used to improve the GWL prediction ability of ML models for 3-month horizons using input datasets of precipitation, evapotranspiration, temperature, and GWL. The four well-accepted standalone ML methods, i.e., artificial neural network (ANN), adaptive neuro-fuzzy inference system (ANFIS), group method of data handling (GMDH), and least square support vector machine (LSSVM), were implemented and compared with the hybrid wavelet conjunction models. The methods were compared based on root mean square error (RMSE), mean absolute error (MAE), correlation coefficient (R), and Nash-Sutcliffe efficiency (NSE). Comparison outcomes showed that the hybrid wavelet-ML considerably improved the standalone model results. The wavelet transform-least square support vector machine (WT-LSSVM) model was superior to other standalone and hybrid wavelet-ML methods to predict GWL. The best GWL predictions were acquired from the WT-LSSVM model with input scenario 5 involving all influential variables, and this model produced RMSE, MAE, R, and NSE as 0.05, 0.04, 0.99, and 0.99 for 1 month ahead of GWL prediction, while the corresponding values were obtained as 0.18, 0.14, 0.95, and 0.90 for 3 months ahead of GWL prediction, respectively., (© 2022. The Author(s), under exclusive licence to Springer-Verlag GmbH Germany, part of Springer Nature.)

Accurate and precise values of hydrodynamic parameters are needed for groundwater modeling and management. Pumping test in the aquifer is the standard method to estimate the transmissivity, hydraulic conductivity, and storage coefficient as the key hydrodynamic parameters. Analytical solutions with curve matching and numerical modeling are two methods to estimate these parameters in the aquifer. Graphical analyses are commonly applied to time-drawdown/water table data which are time-consuming and approximate. Graphical type-curve methods as promising tools are used extensively in water resources studies, while applying these methods is still new in pumping test analysis. In the current study, the first effort based on our knowledge, we have reviewed the literature type-curve graphical methods in pumping test analysis. To achieve this goal, we reviewed and compared the journal articles regarding the characteristics and capabilities of the modeling process from 2000 to 2022. We have clustered the reviewed papers into graphical, modeling, and hybrid categories. Then, a comprehensive review of the selected papers was presented to delineate the highlight of every paper. This review could guide researchers in pumping test analysis. Also, we have presented various recommendations for future research to improve the quality of hydrodynamic parameter estimation., (© 2022. The Author(s), under exclusive licence to Springer-Verlag GmbH Germany, part of Springer Nature.)

Machines learning models have recently been proposed for predicting rivers water temperature (T w ) using only air temperature (T a ). The proposed models relied on a nonlinear relationship between the T w and T a and they have proven to be robust modelling tools. The main motivation for this study was to evaluate how the variational mode decomposition (VMD) contributed to the improvement of machines learning performances for river T w modelling. Measured data collected at five stations located in Poland from 1987 to 2014 were acquired and used for the analysis. Six machines learning models were used and compared namely, K-nearest neighbor's regression (KNNR), least square support vector machine (LSSVM), generalized regression neural network (GRNN), cascade correlation artificial neural networks (CCNN), relevance vector machine (RVM), and locally weighted polynomials regression (LWPR). The six models were developed according to three scenarios. First, the models were calibrated using only the T a as input and obtained results show that the models were able to predict consistently water temperature, showing a high determination coefficient (R 2 ) and Nash-Sutcliffe efficiency (NSE) with values near or above 0.910 and 0.915, respectively, and in overall the six models worked equally without clear superiority of one above another. Second, the air temperature was combined with the periodicity (i.e., day, month and year number) as input variable and a significant improvement was achieved. Both models show their ability to accurately predict river T w with an overall accuracy of 0.956 for R 2 and 0.955 for NSE values, but the LSSVM2 have some advantages such as a small errors metrics, and high fitting capabilities and it slightly surpasses the others models. Thirdly, air temperature was decomposed into several intrinsic mode functions (IMF) using the VMD method and the performances of the models were evaluated. The VMD parameters appeared to cause much influence on the prediction accuracy, exhibiting an improvement of about 40.50% and 39.12% in terms of RMSE and MAE between the first and the third scenarios, however, some models, i.e., GRNN and KNNR have not benefited from the VMD. This research has demonstrated the high capability of the VMD algorithm as a preprocessing approach in improving the accuracies of the machine learning models for river water temperature prediction., (© 2022. The Author(s), under exclusive licence to Springer-Verlag GmbH Germany, part of Springer Nature.)

The nonlinear groundwater level fluctuations depend on the interaction of many factors such as evapotranspiration, precipitation, groundwater abstraction, and hydrogeological characteristics, making groundwater level prediction a complex task. Groundwater level changes are among the most critical issues in water resource management, which can be predicted to effectively provide management solutions to conserve renewable water resources. Understanding the aquifer status using numerical models is time-consuming and also is associated with inherent uncertainty; therefore, in recent decades, the application of artificial intelligence methods to predict water table fluctuations has significantly gained momentum. In this study, artificial neural network (ANN), fuzzy logic (FL), adaptive neuro-fuzzy inference system (ANFIS), and least square support vector machine (SVM) methods were utilized to predict groundwater level (GWL) with 1-, 2-, and 3-month lead time in Tehran-Karaj plain. Several input scenarios were developed considering groundwater levels, average temperature, total precipitation, total evapotranspiration, and average river flow on a monthly interval. The four error criteria, the correlation coefficient (R), root mean squared error (RMSE), Nash-Sutcliffe efficiency (NSE), and mean absolute error (MAE), were the basis to evaluate the models. Results showed that all the applied methods could provide acceptable GWL prediction, but the ANFIS was the most accurate. However, the ANFIS model showed slightly better performance by yielding R = 0.98 for the training stage and R = 0.98 for the testing stage in the P84 observation well and the second combination of inputs and 1-month lead time. The outcomes also revealed that all the approaches mentioned above could appropriately predict GWL for the leading time of 1 and 2 months, but the models provided unsatisfactory results for a 3-month leading time., (© 2022. The Author(s), under exclusive licence to Springer Nature Switzerland AG.)

Total organic carbon (TOC) has vital significance for measuring water quality in river streamflow. The detection of TOC can be considered as an important evaluation because of issues on human health and environmental indicators. This research utilized the novel hybrid models to improve the predictive accuracy of TOC at Andong and Changnyeong stations in the Nakdong River, South Korea. A data pre-processing approach (i.e., complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN)) and evolutionary optimization algorithm (i.e., crow search algorithm (CSA)) were implemented for enhancing the accuracy and robustness of standalone models (i.e., multivariate adaptive regression spline (MARS) and M5Tree). Various water quality indicators (i.e., TOC, potential of Hydrogen (pH), electrical conductivity (EC), dissolved oxygen (DO), water temperature (WT), chemical oxygen demand (COD), and suspended solids (SS)) were utilized for developing the standalone and hybrid models based on three input combinations (i.e., categories 1~3). The developed models were evaluated utilizing the correlation coefficient (CC), root-mean-square error (RMSE), and Nash-Sutcliffe efficiency (NSE). The CEEMDAN-MARS-CSA based on category 2 (C-M-CSA2) model (CC = 0.762, RMSE = 0.570 mg/L, and NSE = 0.520) was the most accurate for predicting TOC at Andong station, whereas the CEEMDAN-MARS-CSA based on category 3 (C-M-CSA3) model (CC = 0.900, RMSE = 0.675 mg/L, and NSE = 0.680) was the best at Changnyeong station.

Adopting methodologies utilizing exogenous data from ancillary stations for determining crop water requirement is a suitable approach to exempt local shortcomings due to the lack of meteorological data/stations. Meanwhile, soft computing techniques might be suitable tools to be used with such data management scenarios. The present paper aimed at evaluating the generalizability of the gene expression programming (GEP) technique for estimating reference evapotranspiration (ET 0 ) through cross-station assessment and exogenous data supply, using data from Turkey and Iran. The GEP-based models were established and learnt using data from 10 stations in Turkey, and then the developed models were tested (validated) in 18 stations of Iran with considerable latitude differences. Different time periods (beginning and the end of time series) were selected for the training and testing stations so that there was no overlap among the dates of the events in both the groups. A comparison was also performed between the GEP models and the corresponding commonly used empirical equations. The obtained results revealed that the generalized GEP models presented promising outcomes in simulating daily ET 0 values when they were trained and tested in quite distant stations with different chronological periods of the applied parameters. The performance accuracy of the empirical equations calibrated using exogenous data was reduced in comparison with their original (non-calibrated) versions. Further, although the generalization ability of the GEP models was reduced when the climatic context of the training-testing stations was different, the overall performance accuracy of those models was higher than those of the commonly used classic empirical equations.

The high cost and time for determining water quality parameters justify the importance of application of mathematical models in discovering connection among them. This paper presents a data mining technique and its improved version in estimating water quality parameters. For this purpose, the surface and ground water quality data from Hamedan (Iran) between 2006 and 2015 were analyzed using M5 model tree and its modified version optimized with Excel Solver Platform (ESP). The values of electrical conductivity (EC), total dissolved solids (TDS), sodium adsorption ratio (SAR), and total hardness (TH) were considered as target variables, whereas pH, concentrations of sodium (Na), chlorine (Cl), bicarbonate (HCO 3 ), sulfate (SO 4 ), magnesium (Mg), calcium (Ca), and potassium (K) were as inputs. The results showed that in both the sources, pH was the least influential parameter on EC, TDS, SAR, and TH. It was found that among the objective parameters, the accuracy of models in estimating TH was higher than the other parameters, whereas SAR was a complex variable. The comparison of performances of the M5 and the M5-ESP models illustrated that the application of the ESP significantly decreased the normal root mean error (NRMSE) of the M5 model; the mean NRMSEs were decreased by 18.95% and 20.29% in estimating groundwater and surface water quality parameters, respectively. Moreover, ability of both the M5 and the M5-ESP models in computing objective parameters of the groundwater was found to be better than the surface water.

There is a need to develop an accurate and reliable model for predicting suspended sediment load (SSL) because of its complexity and difficulty in practice. This is due to the fact that sediment transportation is extremely nonlinear and is directed by numerous parameters such as rainfall, sediment supply, and strength of flow. Thus, this study examined two scenarios to investigate the effectiveness of the artificial neural network (ANN) models and determine the sensitivity of the predictive accuracy of the model to specific input parameters. The first scenario proposed three advanced optimisers-whale algorithm (WA), particle swarm optimization (PSO), and bat algorithm (BA)-for the optimisation of the performance of artificial neural network (ANN) in accurately predicting the suspended sediment load rate at the Goorganrood basin, Iran. In total, 5 different input combinations were examined in various lag days of up to 5 days to make a 1-day-ahead SSL prediction. Scenario 2 introduced a multi-objective (MO) optimisation algorithm that utilises the same inputs from scenario 1 as a way of determining the best combination of inputs. Results from scenario 1 revealed that high accuracy levels were achieved upon utilisation of a hybrid ANN-WA model over the ANN-BA with an RMSE value ranging from 1 to 6%. Furthermore, the ANN-WA model performed better than the ANN-PSO with an accuracy improvement value of 5-20%. Scenario 2 achieved the highest R 2 when ANN-MOWA was introduced which shows that hybridisation of the multi-objective algorithm with WA and ANN model significantly improves the accuracy of ANN in predicting the daily suspended sediment load.

For effective planning of irrigation scheduling, water budgeting, crop simulation, and water resources management, the accurate estimation of reference evapotranspiration (ET o ) is essential. In the current study, the hybrid support vector regression (SVR) coupled with Whale Optimization Algorithm (SVR-WOA) was employed to estimate the monthly ET o at Algiers and Tlemcen meteorological stations positioned in the north of Algeria under three different optimal input scenarios. Monthly climatic parameters, i.e., solar radiation (R s ), wind speed (U s ), relative humidity (RH), and maximum and minimum air temperatures (T max and T min ) of 14 years (2000-2013), were obtained from both stations. The accuracy of the hybrid SVR-WOA model was appraised against hybrid SVR-MVO (Multi-Verse Optimizer), and SVR-ALO (Ant Lion Optimizer) models through performance measures, i.e., mean absolute error (MAE), root-mean-square error (RMSE), index of scattering (IOS), index of agreement (IOA), Pearson correlation coefficient (PCC), Nash-Sutcliffe efficiency (NSE), and graphical interpretation (time-variation and scatter plots, radar chart, and Taylor diagram). The results showed that the SVR-WOA model performed superior to the SVR-MVO and SVR-ALO models at both stations in all scenarios. The SVR-WOA-1 model with five inputs (i.e., T min, T max, RH, U s , R s : scenario-1) had the lowest value of MAE = 0.0658/0.0489 mm/month, RMSE = 0.0808/0.0617 mm/month, IOS = 0.0259/0.0165, and the highest value of NSE = 0.9949/0.9989, PCC = 0.9975/0.9995, and IOA = 0.9987/0.9997 for testing period at both stations, respectively. The proposed hybrid SVR-WOA model was found to be more appropriate and efficient in comparison to SVR-MVO and SVR-ALO models for estimating monthly ET o in the study region.

Accurate estimation of reference evapotranspiration (ET o ) is profoundly crucial in crop modeling, sustainable management, hydrological water simulation, and irrigation scheduling, since it accounts for more than two-thirds of global precipitation losses. Therefore, ET o -based estimation is a major concern in the hydrological cycle. The estimation of ET o can be determined using various methods, including field measurement (the scale of the lysimeter), experimental methods, and mathematical equations. The Food and Agriculture Organization recommended the Penman-Monteith (FAO-56 PM) method which was identified as the standard method of ET o estimation. However, this equation requires a large number of measured climatic data (maximum and minimum air temperature, relative humidity, solar radiation, and wind speed) that are not always available on meteorological stations. Over the decade, the artificial intelligence (AI) models have received more attention for estimating ET o on multi-time scales. This research explores the potential of new hybrid AI model, i.e., support vector regression (SVR) integrated with grey wolf optimizer (SVR-GWO) for estimating monthly ET o at Algiers, Tlemcen, and Annaba stations located in the north of Algeria. Five climatic variables namely relative humidity (RH), maximum and minimum air temperatures (T max and T min ), solar radiation (R s ), and wind speed (U s ) were used for model construction and evaluation. The proposed hybrid SVR-GWO model was compared against hybrid SVR-genetic algorithm (SVR-GA), SVR-particle swarm optimizer (SVR-PSO), conventional artificial neural network (ANN), and empirical (Turc, Ritchie, Thornthwaite, and three versions of Valiantzas methods) models by using root mean squared error (RMSE), Nash-Sutcliffe efficiency (NSE), Pearson correlation coefficient (PCC), and Willmott index (WI), and through graphical interpretation. Through the results obtained, the performance of the SVR-GWO provides very promising and occasionally competitive results compared to other data-driven and empirical methods at study stations. Thus, the proposed SVR-GWO model with five climatic input variables outperformed the other models (RMSE = 0.0776/0.0613/0.0374 mm, NSE = 0.9953/ 0.9990/0.9995, PCC = 0.9978/0.9995/0.9998 and WI = 0.9988/0.9997/0.9999) for estimating ET o at Algiers, Tlemcen, and Annaba stations, respectively. In conclusion, the results of this research indicate the suitability of the proposed hybrid artificial intelligence model (SVR-GWO) at the study stations. Besides, promising results encourage researchers to transfer and test these models in other locations in the world in future works.

Prediction of dissolved oxygen which is an important water quality (WQ) parameter is crucial for aquatic managers who have responsibility for the ecosystem health's maintenance and for the management of reservoirs related to WQ. This study proposes a new ensemble method, Bayesian model averaging (BMA), for estimating hourly dissolved oxygen. The potential of the BMA was investigated and compared with five data-driven methods, extreme leaning machine (ELM), artificial neural networks (ANNs), adaptive neuro-fuzzy inference system (ANFIS), classification and regression tree (CART), and multilinear regression (MLR), by considering hourly temperature, pH, and specific conductivity data as inputs. The methods were compared with respect to three statistics, root mean square errors (RMSE), Nash-Sutcliffe efficiency, and determination coefficient. Results based on two stations' data indicated that the proposed method performed superior to the ELM, ANN, ANFIS, CART, and MLR in estimation of hourly dissolved oxygen; corresponding improvements obtained by BMA are about 5-8%, 13-12%, 7-9%, and 18-27% with respect to RMSE. The ELM also outperformed the other four methods (ANN, ANFIS, CART, and MLR), and the CART and MLR indicated the lowest estimation accuracy in both stations. Examination of various input combinations revealed that the most effective variable is water temperature while the specific conductivity has negligible effect on hourly dissolved oxygen.

Accurate prediction of suspended sediment concentration (SSC) carried by a river or watershed basin is essential for understanding the hydrology of basin in terms of water quality, river bed sustainability and aquatic habitats. In this study, four heuristic methods, namely, radial basis neural network (RBNN), self-organizing map neural network (SOMNN), least square support vector regression (LSSVR), and multivariate adaptive regression spline (MARS) were employed for daily SSC modeling at Ashti, Bamini, and Tekra stations located in Godavari River basin, Andhra Pradesh, India. The Gamma test (GT) was utilized for identifying the most significant input variables for the applied heuristic approaches. The results obtained by RBNN, SOMNN, LSSVR, and MARS models were compared with those of the traditional sediment rating curve (SRC). The performance of the models was evaluated based on the root mean square error (RMSE), coefficient of efficiency (COE), Pearson correlation coefficient (PCC), Willmott index (WI), and pooled average relative error (PARE) indices, as well as the visual inspection using line diagram, scatter diagram, and Taylor diagram (TD). The results of comparison revealed that the four heuristic methods gave higher accuracy than the SRC model. Among the heuristic models, the RBNN-3 (RMSE = 0.045, 0.062, 0.131 g/l; COE = 0.884, 0.883, 0.914; PCC = 0.955, 0.961, 0.958; and WI = 0.970, 0.963, 0.976) outperformed the other models in simulating daily SSC records in the studied stations.

Chlorination, the basic treatment utilized for drinking water sources, is widely used for water disinfection and pathogen elimination in water distribution networks. Thereafter, the proper prediction of chlorine consumption is of great importance in water distribution network performance. In this respect, data mining techniques-which have the ability to discover the relationship between dependent variable(s) and independent variables-can be considered as alternative approaches in comparison to conventional methods (e.g., numerical methods). This study examines the applicability of three key methods, based on the data mining approach, for predicting chlorine levels in four water distribution networks. ANNs (artificial neural networks, including the multi-layer perceptron neural network, MLPNN, and radial basis function neural network, RBFNN), SVM (support vector machine), and CART (classification and regression tree) methods were used to estimate the concentration of residual chlorine in distribution networks for three villages in Kerman Province, Iran. Produced water (flow), chlorine consumption, and residual chlorine were collected daily for 3 years. An assessment of the studied models using several statistical criteria (NSC, RMSE, R 2 , and SEP) indicated that, in general, MLPNN has the greatest capability for predicting chlorine levels followed by CART, SVM, and RBF-ANN. Weaker performance of the data-driven methods in the water distribution networks, in some cases, could be attributed to improper chlorination management rather than the methods' capability.

In this paper, several extreme learning machine (ELM) models, including standard extreme learning machine with sigmoid activation function (S-ELM), extreme learning machine with radial basis activation function (R-ELM), online sequential extreme learning machine (OS-ELM), and optimally pruned extreme learning machine (OP-ELM), are newly applied for predicting dissolved oxygen concentration with and without water quality variables as predictors. Firstly, using data from eight United States Geological Survey (USGS) stations located in different rivers basins, USA, the S-ELM, R-ELM, OS-ELM, and OP-ELM were compared against the measured dissolved oxygen (DO) using four water quality variables, water temperature, specific conductance, turbidity, and pH, as predictors. For each station, we used data measured at an hourly time step for a period of 4 years. The dataset was divided into a training set (70%) and a validation set (30%). We selected several combinations of the water quality variables as inputs for each ELM model and six different scenarios were compared. Secondly, an attempt was made to predict DO concentration without water quality variables. To achieve this goal, we used the year numbers, 2008, 2009, etc., month numbers from (1) to (12), day numbers from (1) to (31) and hour numbers from (00:00) to (24:00) as predictors. Thirdly, the best ELM models were trained using validation dataset and tested with the training dataset. The performances of the four ELM models were evaluated using four statistical indices: the coefficient of correlation (R), the Nash-Sutcliffe efficiency (NSE), the root mean squared error (RMSE), and the mean absolute error (MAE). Results obtained from the eight stations indicated that: (i) the best results were obtained by the S-ELM, R-ELM, OS-ELM, and OP-ELM models having four water quality variables as predictors; (ii) out of eight stations, the OP-ELM performed better than the other three ELM models at seven stations while the R-ELM performed the best at one station. The OS-ELM models performed the worst and provided the lowest accuracy; (iii) for predicting DO without water quality variables, the R-ELM performed the best at seven stations followed by the S-ELM in the second place and the OP-ELM performed the worst with low accuracy; (iv) for the final application where training ELM models with validation dataset and testing with training dataset, the OP-ELM provided the best accuracy using water quality variables and the R-ELM performed the best at all eight stations without water quality variables. Fourthly, and finally, we compared the results obtained from different ELM models with those obtained using multiple linear regression (MLR) and multilayer perceptron neural network (MLPNN). Results obtained using MLPNN and MLR models reveal that: (i) using water quality variables as predictors, the MLR performed the worst and provided the lowest accuracy in all stations; (ii) MLPNN was ranked in the second place at two stations, in the third place at four stations, and finally, in the fourth place at two stations, (iii) for predicting DO without water quality variables, MLPNN is ranked in the second place at five stations, and ranked in the third, fourth, and fifth places in the remaining three stations, while MLR was ranked in the last place with very low accuracy at all stations. Overall, the results suggest that the ELM is more effective than the MLPNN and MLR for modelling DO concentration in river ecosystems.