Groundwater level forecasting with machine learning models: A review.

Groundwater, the world's most abundant source of freshwater, is rapidly depleting in many regions due to a variety of factors. Accurate forecasting of groundwater level (GWL) is essential for effective management of this vital resource, but it remains a complex and challenging task. In recent years, there has been a notable increase in the use of machine learning (ML) techniques to model GWL, with many studies reporting exceptional results. In this paper, we present a comprehensive review of 142 relevant articles indexed by the Web of Science from 2017 to 2023, focusing on key ML models, including artificial neural networks (ANN), adaptive neuro-fuzzy inference systems (ANFIS), support vector regression (SVR), evolutionary computing (EC), deep learning (DL), ensemble learning (EN), and hybrid-modeling (HM). We also discussed key modeling concepts such as dataset size, data splitting, input variable selection, forecasting time-step, performance metrics (PM), study zones, and aquifers, highlighting best practices for optimal GWL forecasting with ML. This review provides valuable insights and recommendations for researchers and water management agencies working in the field of groundwater management and hydrology.

Assessment of Evaporative Demand Drought Index for drought analysis in Peninsular Malaysia.

An effective drought monitoring tool is essential for the development of timely drought early warning system. This study evaluates Evaporative Demand Drought Index (EDDI) as a drought indicator in measuring spatiotemporal evolution of droughts over Peninsular Malaysia during 1989-2018. The modified Mann-Kendall and Sen's slope tests were performed to detect the presence of monotonic trends in EDDI, Standardized Precipitation Index (SPI), Standardized Precipitation Evapotranspiration Index (SPEI) and their related climate variables. The performance of EDDI in capturing the drought onset, evolutions and demise of historical severe droughts was also compared with SPI and SPEI at multiple timescales. EDDI demonstrates strong spatiotemporal correlations with SPI and SPEI and comparable performance in historical drought events identification. At sub-monthly timescale, 2-week EDDI displays equivalent drought severities and durations for all historical severe droughts corresponding to the monthly EDDI. In the case when rainfall deficits are normalized in an otherwise warm and dry month, EDDI may serve as a great alternative to SPI and SPEI due to it being sensitive to the changes in prevalent atmospheric conditions. Collectively, the results fill in the knowledge gaps on drought evolutions from the evaporative perspective and highlight the efficacy of EDDI as a valuable drought early warning tool for Peninsular Malaysia. Future study should explore the physical mechanisms behind the development of flash drought and the role of evaporation in the drought propagation processes.

Groundwater level forecasting using ensemble coactive neuro-fuzzy inference system.

A modeling framework utilizing the coactive neuro-fuzzy inference system (CANFIS) has been developed for multi-lead time groundwater level (GWL) forecasting in four different wells located in Texas and Florida, USA. Various model input combinations, including GWL, precipitation, temperature, and surface water level variables, have been derived based on proposed correlation analysis using singular spectrum analysis (SSA) remainders. The models have been trained on data subsets of varying lengths to identify the optimal training data duration. Additionally, we have introduced the bagging ensemble learning method to enhance the performance of the CANFIS model. As part of a comprehensive model evaluation process, the best-performing CANFIS model for each forecasting scenario has undergone uncertainty analysis using bootstrap sampling. Our results reveal that the CANFIS model performs satisfactorily for daily forecasting but leaves room for improvement in monthly forecasting, particularly for two-month and three-month ahead forecasts. Moreover, we have identified several optimal input combinations, highlighting the significance of the temperature variable in monthly forecasting. Furthermore, our findings indicate that additional training data does not necessarily lead to improved performance. The ensemble CANFIS model has demonstrated significant performance enhancement, particularly for monthly forecasting. Finally, the CANFIS model uncertainty analysis has shown satisfactory results for daily forecasting scenarios, while monthly forecasting models exhibit higher uncertainties, particularly during periods with distinctly different GWL fluctuation patterns.

Optimized fuzzy inference system to enhance prediction accuracy for influent characteristics of a sewage treatment plant.

Sewage treatment plants (STPs) keep sewage contamination within safe levels and minimize the risk of environmental disasters. To achieve optimum operation of an STP, it is necessary for influent parameters to be measured or estimated precisely. In this research, six well-known influent chemical and biological characteristics, i.e., biochemical oxygen demand (BOD), chemical oxygen demand (COD), Ammoniacal Nitrogen (NH3-N), pH, oil and grease (OG) and suspended solids (SS), were modeled and predicted using the Sugeno fuzzy logic model. The membership function range of the fuzzy model was optimized by ANFIS, the integrated Genetic algorithms (GA), and the integrated particle swarm optimization (PSO) algorithms. The results were evaluated by different indices to find the accuracy of each algorithm. To ensure prediction accuracy, outliers in the predicted data were found and replaced with reasonable values. The results showed that both integrated GA-FIS and PSO-FIS algorithms performed at almost the same level and both had fewer errors than ANFIS. As the GA-FIS algorithm predicts BOD with fewer errors than PSO-FIS and the aim of this study is to provide an accurate prediction of missing data, GA-FIS was only used to predict the BOD parameter; the other parameters were predicted by PSO-FIS algorithm. As a result, the model successfully could provide outstanding performance for predicting the BOD, COD, NH3-N, OG, pH and SS with MAE equal to 3.79, 5.14, 0.4, 0.27, 0.02, and 3.16, respectively.

Leachate generation rate modeling using artificial intelligence algorithms aided by input optimization method for an MSW landfill.

Leachate is one of the main surface water pollution sources in Selangor State (SS), Malaysia. The prediction of leachate amounts is elementary in sustainable waste management and leachate treatment processes, before discharging to surrounding environment. In developing countries, the accurate evaluation of leachate generation rates has often considered a challenge due to the lack of reliable data and high measurement costs. Leachate generation is related to several factors, including meteorological data, waste generation rates, and landfill design conditions. The high variations in these factors lead to complicating leachate modeling processes. This study aims at identifying the key elements contributing to leachate production and developing various AI-based models to predict leachate generation rates. These models included Artificial Neural Network (ANN)-Multi-linear perceptron (MLP) with single and double hidden layers, and support vector machine (SVM) regression time series algorithms. Various performance measures were applied to evaluate the developed model's accuracy. In this study, input optimization process showed that three inputs were acceptable for modeling the leachate generation rates, namely dumped waste quantity, rainfall level, and emanated gases. The initial performance analysis showed that ANN-MLP2 model-which applies two hidden layers-achieved the best performance, then followed by ANN-MLP1 model-which applies one hidden layer and three inputs-while SVM model gave the lowest performance. Ranges and frequency of relative error (RE%) also demonstrate that ANN-MLP models outperformed SVM models. Furthermore, low and peak flow criterion (LFC and PFC) assessment of leachate inflow values in ANN-MLP model with two hidden layers made more accurate values than other models. Since minimizing data collection and processing efforts as well as minimizing modeling complexity are critical in the hydrological modeling process, the applied input optimization process and the developed models in this study were able to provide a good performance in the modeling of leachate generation efficiently.

Predicting sanitary landfill leachate generation in humid regions using ANFIS modeling.

Landfill leachate is one of the sources of surface water pollution in Selangor State (SS), Malaysia. Leachate volume prediction is essential for sustainable waste management and leachate treatment processes. The accurate estimation of leachate generation rates is often considered a challenge, especially in developing countries, due to the lack of reliable data and high measurement costs. Leachate generation is related to several variable factors, including meteorological data, waste generation rates, and landfill design conditions. Large variations in these factors lead to complicated leachate modeling processes. The aims of this study are to determine the key elements contributing to leachate production and then develop an adaptive neural fuzzy inference system (ANFIS) model to predict leachate generation rates. Accuracy of the final model performance was tested and evaluated using the root mean square error (RMSE), the mean absolute error (MAE), and the correlation coefficient (R). The study results defined dumped waste quantity, rainfall level, and emanated gases as the most significant contributing factors in leachate generation. The best model structure consisted of two triangular fuzzy membership functions and a hybrid training algorithm with eight fuzzy rules. The proposed ANFIS model showed a good performance with an overall correlation coefficient of 0.952.

Analysing the accuracy of machine learning techniques to develop an integrated influent time series model: case study of a sewage treatment plant, Malaysia.

The function of a sewage treatment plant is to treat the sewage to acceptable standards before being discharged into the receiving waters. To design and operate such plants, it is necessary to measure and predict the influent flow rate. In this research, the influent flow rate of a sewage treatment plant (STP) was modelled and predicted by autoregressive integrated moving average (ARIMA), nonlinear autoregressive network (NAR) and support vector machine (SVM) regression time series algorithms. To evaluate the models' accuracy, the root mean square error (RMSE) and coefficient of determination (R2) were calculated as initial assessment measures, while relative error (RE), peak flow criterion (PFC) and low flow criterion (LFC) were calculated as final evaluation measures to demonstrate the detailed accuracy of the selected models. An integrated model was developed based on the individual models' prediction ability for low, average and peak flow. An initial assessment of the results showed that the ARIMA model was the least accurate and the NAR model was the most accurate. The RE results also prove that the SVM model's frequency of errors above 10% or below - 10% was greater than the NAR model's. The influent was also forecasted up to 44 weeks ahead by both models. The graphical results indicate that the NAR model made better predictions than the SVM model. The final evaluation of NAR and SVM demonstrated that SVM made better predictions at peak flow and NAR fit well for low and average inflow ranges. The integrated model developed includes the NAR model for low and average influent and the SVM model for peak inflow.

Optimization of Multiple and Multipurpose Reservoir System Operations by Using Matrix Structure (Case Study: Karun and Dez Reservoir Dams).

Optimal operation of water resources in multiple and multipurpose reservoirs is very complicated. This is because of the number of dams, each dam's location (Series and parallel), conflict in objectives and the stochastic nature of the inflow of water in the system. In this paper, performance optimization of the system of Karun and Dez reservoir dams have been studied and investigated with the purposes of hydroelectric energy generation and providing water demand in 6 dams. On the Karun River, 5 dams have been built in the series arrangements, and the Dez dam has been built parallel to those 5 dams. One of the main achievements in this research is the implementation of the structure of production of hydroelectric energy as a function of matrix in MATLAB software. The results show that the role of objective function structure for generating hydroelectric energy in weighting method algorithm is more important than water supply. Nonetheless by implementing ε- constraint method algorithm, we can both increase hydroelectric power generation and supply around 85% of agricultural and industrial demands.

Assessment of water quality parameters using multivariate analysis for Klang River basin, Malaysia.

This case study uses several univariate and multivariate statistical techniques to evaluate and interpret a water quality data set obtained from the Klang River basin located within the state of Selangor and the Federal Territory of Kuala Lumpur, Malaysia. The river drains an area of 1,288 km(2), from the steep mountain rainforests of the main Central Range along Peninsular Malaysia to the river mouth in Port Klang, into the Straits of Malacca. Water quality was monitored at 20 stations, nine of which are situated along the main river and 11 along six tributaries. Data was collected from 1997 to 2007 for seven parameters used to evaluate the status of the water quality, namely dissolved oxygen, biochemical oxygen demand, chemical oxygen demand, suspended solids, ammoniacal nitrogen, pH, and temperature. The data were first investigated using descriptive statistical tools, followed by two practical multivariate analyses that reduced the data dimensions for better interpretation. The analyses employed were factor analysis and principal component analysis, which explain 60 and 81.6% of the total variation in the data, respectively. We found that the resulting latent variables from the factor analysis are interpretable and beneficial for describing the water quality in the Klang River. This study presents the usefulness of several statistical methods in evaluating and interpreting water quality data for the purpose of monitoring the effectiveness of water resource management. The results should provide more straightforward data interpretation as well as valuable insight for managers to conceive optimum action plans for controlling pollution in river water.

Arsenic, zinc, and aluminium removal from gold mine wastewater effluents and accumulation by submerged aquatic plants (Cabomba piauhyensis, Egeria densa, and Hydrilla verticillata).

The potential of three submerged aquatic plant species (Cabomba piauhyensis, Egeria densa, and Hydrilla verticillata) to be used for As, Al, and Zn phytoremediation was tested. The plants were exposed for 14 days under hydroponic conditions to mine waste water effluents in order to assess the suitability of the aquatic plants to remediate elevated multi-metals concentrations in mine waste water. The results show that the E. densa and H. verticillata are able to accumulate high amount of arsenic (95.2%) and zinc (93.7%) and resulted in a decrease of arsenic and zinc in the ambient water. On the other hand, C. piauhyensis shows remarkable aluminium accumulation in plant biomass (83.8%) compared to the other tested plants. The ability of these plants to accumulate the studied metals and survive throughout the experiment demonstrates the potential of these plants to remediate metal enriched water especially for mine drainage effluent. Among the three tested aquatic plants, H. verticillata was found to be the most applicable (84.5%) and suitable plant species to phytoremediate elevated metals and metalloid in mine related waste water.

Trend analysis of a tropical urban river water quality in Malaysia.

Rivers play a significant role in providing water resources for human and ecosystem survival and health. Hence, river water quality is an important parameter that must be preserved and monitored. As the state of Selangor and the city of Kuala Lumpur, Malaysia, are undergoing tremendous development, the river is subjected to pollution from point and non-point sources. The water quality of the Klang River basin, one of the most densely populated areas within the region, is significantly degraded due to human activities as well as urbanization. Evaluation of the overall river water quality status is normally represented by a water quality index (WQI), which consists of six parameters, namely dissolved oxygen, biochemical oxygen demand, chemical oxygen demand, suspended solids, ammoniacal nitrogen and pH. The objectives of this study are to assess the water quality status for this tropical, urban river and to establish the WQI trend. Using monthly WQI data from 1997 to 2007, time series were plotted and trend analysis was performed by employing the first-order autocorrelated trend model on the moving average values for every station. The initial and final values of either the moving average or the trend model were used as the estimates of the initial and final WQI at the stations. It was found that Klang River water quality has shown some improvement between 1997 and 2007. Water quality remains good in the upper stream area, which provides vital water sources for water treatment plants in the Klang valley. Meanwhile, the water quality has also improved in other stations. Results of the current study suggest that the present policy on managing river quality in the Klang River has produced encouraging results; the policy should, however, be further improved alongside more vigorous monitoring of pollution discharge from various point sources such as industrial wastewater, municipal sewers, wet markets, sand mining and landfills, as well as non-point sources such as agricultural or urban runoff and commercial activity.