Machine learning predictions of chlorophyll-a in the Han river basin, Korea

Developing a real-time water quality simulation toolbox using machine learning and application programming interface

Rivers are vital for sustaining human life as they foster social development, provide drinking water, maintain aquatic ecosystems, and offer recreational spaces. However, most rivers are being increasingly contaminated by pollutants from non-point sources, urbanization, and other sources. Consequently, real-time river water quality modeling is essential for managing and protecting rivers from contamination, and its significance is growing across various sectors, including public health, agriculture, and water treatment systems. Therefore, a real-time river water quality simulation toolbox was developed using machine learning (ML) and an application program interface (API). To create the toolbox, models that simulated water quality parameters such as chlorophyll a (Chl-a), dissolved oxygen (DO), total nitrogen (TN), total organic carbon (TOC), and total phosphorus (TP) at each point in the Nakdong River were constructed. The models were constructed using Artificial neural network (ANN), Random Forest (RF), support vector machines (SVM), and data from API. Subsequently, hyperparameter optimization was conducted to enhance the model's performance. During training, the models' performances were evaluated and compared based on the data sampling method and ML algorithms. Models trained with random sampling data outperformed those trained with time-series data. Among the algorithm models that used random sampling data, the RF exhibited the best performance. The average coefficient of determination (R2) values for each water quality simulation with randomly sampled data using RF for DO, TN, TP, Chl-a, and TOC were 0.79, 0.65, 0.74, 0.45, and 0.48, respectively. For ANN, they were 0.7, 0.51, 0.64, 0.35, and 0.35, respectively, and for SVM, they were 0.73, 0.51, 0.59, 0.21, and 0.3, respectively. The Chl-a and TOC models exhibited relatively poor performance, whereas the DO, TN, and TP models demonstrated superior performance. Diversifying the input data variables is necessary to improve the performance of the Chl-a and TOC models. Sensitivity and uncertainty analyses were conducted to evaluate and enhance the models' understanding. Furthermore, using a graphic user interface (GUI) toolbox, user convenience was maximized.

Algal classification and Chlorophyll-a concentration determination using convolutional neural networks and three-dimensional fluorescence data matrices

In recent years, the frequency of harmful algal blooms has increased, leading to the release of large quantities of toxins and compounds that cause unpleasant odors and tastes, significantly compromising drinking water quality. Chlorophyll-a (Chl-a) is commonly used as a proxy for algal biomass. However, current methods for measuring Chl-a concentration face challenges in accurately quantifying algae by categories and effectively adapting to natural aquatic environments. This study combined convolutional neural networks (CNNs) and three-dimensional fluorescence data matrices to address these challenges. The algal classification model achieved over 99.5% accuracy in identifying thirteen types of algal samples, with class activation maps showing that the model primarily focused on algal pigment regions. In determining Chl-a concentrations of each algal species in mixed algae solutions (Microcystis aeruginosa, Cyclotella, and Chlorella), the Chl-a models demonstrated Mean Absolute Percentage Errors (MAPEs) ranging from 6.55% to 10.56% in the ultrapure water background, 11.57%-14.12% in the Qingcaosha Reservoir raw water background, and 21.46%-123.37% in the Lake Taihu raw water background. After calibration, the models were significantly improved, achieving MAPEs ranging from 11.86% to 14.18% in the Lake Taihu raw water background. Discrepancies in determination performance indicated that the intensity and locations of characteristic algal pigment fluorescence peaks greatly influenced the Chl-a models' accuracy. This research introduces a novel approach for algal classification and Chl-a concentration determination in water bodies, with significant potential for practical applications.

Source apportionment and influencing factors of surface water pollution through a combination of multiple receptor models and geodetector

In line with sustainable development goals (SDGs), precise quantification of water pollution and analysis of environmental interactions are crucial for effectively safeguarding water resources. In this study, Nemerow's pollution index was used to evaluate water quality, three receptor models were used to identify pollution sources, and Geodetector analysis was applied to explore environmental interactions in the North Shangyu Plain, Southeast China. Using 5207 surface water samples from September 2023 with 11 physicochemical parameters, the results showed that surface rivers in the North Shangyu Plain exhibited varying degrees of pollution: slight pollution upstream, moderate pollution in midstream and downstream, and concentrated high pollution in certain areas, with TN, CODCr, and TP as the primary pollutants. Multimethod source apportionment significantly improved the accuracy of pollution source attribution and identified five main sources: domestic sewage (1.42%-3.54%) characterized by NO3-N, phytoplankton source (38.43%-50.05%) indicated by chl and PC, agricultural cultivation (16.1%-17.63%) marked by TP and CODMn, industrial wastewater (17.64%-25.1%) primarily associated with TN, and natural source (10.32%-13.26%) characterized by DO, NH3-N, and CODCr. Influencing factor analysis validated the source identification. Natural factors had minor impacts on water parameters, while pollution control from agricultural activities was suggested to diversify fertilizer types rather than merely reduce quantities. The combined effects of industrial and aquaculture activities intensified pollution from TN, chl, and PC, underscoring the need for targeted management practices. This study showed the objectivity and reliability of using a combined approach of multiple receptor models and Geodetector to evaluate the river water quality status, which helps assist decision-makers in formulating more effective water resource protection strategies.

Seasonal responses of microbial communities to water quality variations and interaction of eutrophication risk in Gehu Lake

Gehu Lake, as a key upstream reservoir of Taihu Lake, China, plays a crucial role in improving the water quality, and eutrophication control of the Taihu Lake Basin. Although the microbial communities are significantly important in maintaining the ecological health of lake, the microbial response to water quality, especially for eutrophication has been rarely reported in Gehu Lake. In this study, the water quality parameters and the corresponding effects on the structure and function of microbial communities were determined seasonally. It was found that the poorest water quality in summer (Water Quality Index = 116.52) with severe eutrophication (Trophic Level Index >70), was primarily driven by agricultural non-point sources (33.4%) and seasonal pollution (23.8%). The chemical oxygen demand (COD) was the most important indicator of water quality that affected the concentration of Chlorophyll-a (Chla) according to Pearson correlation analysis (p < 0.001), random forest modeling (p < 0.01), and structural equation modeling (path coefficient = 0.926). Redundancy analysis revealed that total nitrogen, total phosphorus, Chla, and COD significantly influenced the microbial community (p < 0.05). Microbial co-occurrence networks demonstrated significantly seasonal variations, and winter exhibited a more complex structure under lower temperature and limited nutrients compared to the other seasons. In addition, the Chla-sensitive microbial species that involved in nitrogen and phosphorus metabolism were identified as the biological indicators of eutrophication in response to the changes of seasonal water quality. These findings have taken insights into the interactions between water quality and microbial communities, and might provide the basis for improvement of the ecological and environmental management of Gehu Lake, as well as the control of eutrophication in Taihu Lake.

Enhancing the content of phycoerythrin through the application of microplastics from Porphyridium cruentum produced in wastewater using machine learning methods

Microalgae can produce secondary metabolites like phycoerythrin (Phy). The effects of some microplastics (MPs), wastewater (WW), and light intensity (LI) parameters, including complex data sets, on Phy concentration from Porphyridium cruentum were investigated using machine learning methods in this study. Also, the deep learning (DL) model was developed to get the maximum phy concentration from the dataset. The dataset (232 data groups), including a feature set, polyethylene (PE), polypropylene (PP), polystyrene (PS), polyvinyl chloride (PVC), WW, LI, and an output variable, Phy, were randomly divided into training and test sets to create and evaluate the models. The highest experimental and predicted Phy concentrations were 52.3 mg/g and 58.32 mg/g in a scenario with 15% WW, 80 mg/L PE, PP, PS, and PVC, and a LI of 175 μmolm-2 s-1, respectively. The Pearson correlation coefficient (r) indicates a positive correlation between Phy and the variables PE (r = 0.35), PVC (r = 0.69), PP (r = 0.27), PS (r = 0.29), and LI (r = 0.22). However, variables such as WW (r = -0.05) have a weak correlation, and while PVC and PE showed the most significant effect on Phy concentration, WW had the lowest effect. Furthermore, LIME (local interpretable model-agnostic explanations) and SHAP (shapley additive explanations) provided us with important results for interpreting the random forest regression (RF) and DL models' predictions, respectively. The LIME and SHAP analyses suggest that the system with more PVC has a higher predicted Phy value. For WW, the reverse is true; higher WW values result in lower Phy predictions. Researchers were given the model explainability decision tree (DT) structure to study reactants' effects on output (Phy). In conclusion, the dye industry can use microalgae to treat WW contaminated with MPs while also producing high amounts of Phy using a DL model.

Smartphone as an alternative to measure chlorophyll-a concentration in small waterbodies

Monitoring chlorophyll-a concentrations (Chl-a, μg·L-1) in aquatic ecosystems has attracted much attention due to its direct link to harmful algal blooms. However, there has been a lack of a cost-effective method for measuring Chl-a in small waterbodies. Inspired by the increase of smartphone photography, a Smartphone-based convolutional neural networks (CNN) framework (SCCA) was developed to estimate Chl-a in Aquatic ecosystem. To evaluate the performance of SCCA, 238 paired records (a smartphone image with a 12-color background and a measured Chl-a value) were collected from diverse aquatic ecosystems (e.g., rivers, lakes and ponds) across China in 2023. Our performance-evaluation results revealed a NS and R2 value of 0.90 and 0.94 in Chl-a estimation, demonstrating a satisfactory (NS = 0.84, R2 = 0.86) model fit in lower Chl-a (<30 μg L-1) conditions. SCCA had involved a realtime-update method with hyperparameter optimization technology. In comparison with the existing methods of measuring Chl-a, SCCA provides a useful screening tool for cost-effective measurement of Chl-a and has the potential for being an algal bloom screening means in small waterbodies, using Huajin River as a case study, especially under limited resources for water measurement. Overall, we highlight that the SCCA can be potentially integrated into a smartphone application in the future to diverse waterbodies in environmental management.

Predicting the governing factors for the release of colloidal phosphorus using machine learning

Predicting the parameters that influence colloidal phosphorus (CP) release from soils under different land uses is critical for managing the impact on water quality. Traditional modeling approaches, such as linear regression, may fail to represent the intricate relationships that exist between soil qualities and environmental influences. Therefore, in this study, we investigated the major determinants of CP release from different land use/types such as farmland, desert, forest soils, and rivers. The study utilizes the structural equation model (SEM), multiple linear regression (MLR), and three machine learning (ML) models (Random Forest (RF), Support Vector Regression (SVR), and eXtreme Gradient Boosting (XGBoost)) to predict the release of CP from different soils by using soil iron (Fe), aluminum (Al), calcium (Ca), pH, total organic carbon (TOC) and precipitation as independent variables. Results show that colloidal-cations (Fe, Al, Ca) and colloidal-TOC strongly influence CP release, while bioclimatic variables (precipitation) and pH have weaker effects. XGBoost outperforms the other models with an R2 of 0.94 and RMSE of 0.09. SHapley Additive Explanations described the outcomes since XGBoost is accurate. The relative relevance ranking indicated that colloidal TOC had the highest ranking in predicting CP. This was supported by the analysis of partial dependence plots, which showed that an increase in colloidal TOC increased soil CP release. According to our research, the SHAP XGBoost model provides significant information that can help determine the variables that considerably influence CP contents as compared to RF, SVM, and MLR.

Prediction of microbial activity and abundance using interpretable machine learning models in the hyporheic zone of effluent-dominated receiving rivers

Serving as a vital linkage between surface water and groundwater, the hyporheic zone (HZ) plays a fundamental role in improving water quality and maintaining ecological security. In arid or semi-arid areas, effluent discharge from wastewater treatment facilities could occupy a predominant proportion of the total base flow of receiving rivers. Nonetheless the relationship between microbial activity, abundance and environmental factors in the HZ of effluent-receiving rivers appear to be rarely addressed. In this study, a spatiotemporal field study was performed in two representative effluent-dominated receiving rivers in Xi'an, China. Land use data, physical and chemical water quality parameters of surface and subsurface water were used as predictive variables, while the microbial respiratory electron transport system activity (ETSA), the Chao1 and Shannon index of total microbial community, as well as the Chao1 and Shannon index of denitrifying bacteria community were used as response variables, while ETSA was used as response variables indicating ecological processes and Shannon and Chao1 were utilized as parameters indicating microbial diversity. Two machine learning models were utilized to provide evidence-based information on how environmental factors interact and drive microbial activity and abundance in the HZ at variable depths. The models with Chao1 and Shannon as response variables exhibited excellent predictive performances (R2: 0.754-0.81 and 0.783-0.839). Dissolved organic nitrogen (DON) was the most important factor affecting the microbial functions, and an obvious threshold value of ∼2 mg/L was observed. Credible predictions of models with Chao1 and Shannon index of denitrifying bacteria community as response variables were detected (R2: 0.484-0.624 and 0.567-0.638), with soluble reactive phosphorus (SRP) being the key influencing factor. Fe (Ⅱ) was favorable in predicting denitrifying bacteria community. The ESTA model highlighted the importance of total nitrogen in the ecological health monitoring in HZ. These findings provide novel insights in predicting microbial activity and abundance in highly-impacted areas such as the HZ of effluent-dominated receiving rivers.

Algae development in rivers with artificially constructed weirs: Dominant influence of discharge over temperature

Algal blooms contribute to water quality degradation, unpleasant odors, taste issues, and the presence of harmful substances in artificially constructed weirs. Mitigating these adverse effects through effective algal bloom management requires identifying the contributing factors and predicting algal concentrations. This study focused on the upstream region of the Seungchon Weir in Korea, which is characterized by elevated levels of total nitrogen and phosphorus due to a significant influx of water from a sewage treatment plant. We employed four distinct machine learning models to predict chlorophyll-a (Chl-a) concentrations and identified the influential variables linked to local algal bloom events. The gradient boosting model enabled an in-depth exploration of the intricate relationships between algal occurrence and water quality parameters, enabling accurate identification of the causal factors. The models identified the discharge flow rate (D-Flow) and water temperature as the primary determinants of Chl-a levels, with feature importance values of 0.236 and 0.212, respectively. Enhanced model precision was achieved by utilizing daily average D-Flow values, with model accuracy and significance of the D-Flow amplifying as the temporal span of daily averaging increased. Elevated Chl-a concentrations correlated with diminished D-Flow and temperature, highlighting the pivotal role of D-Flow in regulating Chl-a concentration. This trend can be attributed to the constrained discharge of the Seungchon Weir during winter. Calculating the requisite D-Flow to maintain a desirable Chl-a concentration of up to 20 mg/m3 across varying temperatures revealed an escalating demand for D-Flow with rising temperatures. Specific D-Flow ranges, corresponding to each season and temperature condition, were identified as particularly influential on Chl-a concentration. Thus, optimizing Chl-a reduction can be achieved by strategically increasing D-Flow within these specified ranges for each season and temperature variation. This study highlights the importance of maintaining sufficient D-Flow levels to mitigate algal proliferation within river systems featuring weirs.

Data-driven modelling for assessing trophic status in marine ecosystems using machine learning approaches

Assessing eutrophication in coastal and transitional waters is of utmost importance, yet existing Trophic Status Index (TSI) models face challenges like multicollinearity, data redundancy, inappropriate aggregation methods, and complex classification schemes. To tackle these issues, we developed a novel tool that harnesses machine learning (ML) and artificial intelligence (AI), enhancing the reliability and accuracy of trophic status assessments. Our research introduces an improved data-driven methodology specifically tailored for transitional and coastal (TrC) waters, with a focus on Cork Harbour, Ireland, as a case study. Our innovative approach, named the Assessment Trophic Status Index (ATSI) model, comprises three main components: the selection of pertinent water quality indicators, the computation of ATSI scores, and the implementation of a new classification scheme. To optimize input data and minimize redundancy, we employed ML techniques, including advanced deep learning methods. Specifically, we developed a CHL prediction model utilizing ten algorithms, among which XGBoost demonstrated exceptional performance, showcasing minimal errors during both training (RMSE = 0.0, MSE = 0.0, MAE = 0.01) and testing (RMSE = 0.0, MSE = 0.0, MAE = 0.01) phases. Utilizing a novel linear rescaling interpolation function, we calculated ATSI scores and evaluated the model's sensitivity and efficiency across diverse application domains, employing metrics such as R2, the Nash-Sutcliffe efficiency (NSE), and the model efficiency factor (MEF). The results consistently revealed heightened sensitivity and efficiency across all application domains. Additionally, we introduced a brand new classification scheme for ranking the trophic status of transitional and coastal waters. To assess spatial sensitivity, we applied the ATSI model to four distinct waterbodies in Ireland, comparing trophic assessment outcomes with the Assessment of Trophic Status of Estuaries and Bays in Ireland (ATSEBI) System. Remarkably, significant disparities between the ATSI and ATSEBI System were evident in all domains, except for Mulroy Bay. Overall, our research significantly enhances the accuracy of trophic status assessments in marine ecosystems. The ATSI model, combined with cutting-edge ML techniques and our new classification scheme, represents a promising avenue for evaluating and monitoring trophic conditions in TrC waters. The study also demonstrated the effectiveness of ATSI in assessing trophic status across various waterbodies, including lakes, rivers, and more. These findings make substantial contributions to the field of marine ecosystem management and conservation.

Identifying the drivers of chlorophyll-a dynamics in a landscape lake recharged by reclaimed water using interpretable machine learning

The water quality of lakes recharged by reclaimed water is affected by both the fluctuation of reclaimed water quality and the biochemical processes in the lakes, and therefore the main controlling factors of algal blooms are difficult to identify. Taking a typical landscape lake recharged by reclaimed water as an example and using the spatiotemporal distribution characteristics and correlation analysis of water quality indexes, we propose an interpretable machine learning framework based on random forest to predict chlorophyll-a (Chl-a). The model considered nutrient difference indexes between reclaimed water and lake water, and further used feature importance ranking and partial dependence plot to identify nutrient drivers. Results show that the NO3--N input from reclaimed water is the dominant nutrient driver for algal bloom especially at high temperatures, and the negative correlation between NO3--N and Chl-a in the lake water is the consequence of algal bloom rather than the cause. Our study provides new insights into the identification of eutrophication factors for lakes recharged by reclaimed water.

Deep learning based soft-sensor for continuous chlorophyll estimation on decentralized data

Monitoring the concentration of pigments like chlorophyll (Chl) in water-bodies is a key task to contribute to their conservation. However, with the existing sensor technology, measurement in real-time and with enough frequency to ensure proper risk management is not completely feasible. In this work, with the concept of data-driven soft-sensing, three hydrophysical features are used together with three meteorological ones to estimate the concentration of Chl in two tributaries of the River Thames. Data driven models, specifically neural networks, are used with three learning approaches: individual, centralized and federated. Data reduction scenarios are proposed in order to analyze the performance of each approach when less data is available. The best results in the training are usually obtained with the individual approach. However, the federated learning provides better generalization ability. It was also observed that in most of the cases the results of the federated learning approach improve those of the centralized one.

Water quality pollution assessment and source apportionment of lake wetlands: A case study of Xianghai Lake in the Northeast China Plain

Surface water pollution has always posed a serious challenge to water quality management. Improving water quality management requires figuring out how to comprehend water quality conditions scientifically and effectively as well as quantitatively identify regional pollution sources. In this study, Xianghai Lake, a typical lake-type wetland on the Northeast China Plain, was taken as the research area. Based on a geographic information system (GIS) method and 11 water quality parameters, the single-factor evaluation and comprehensive water quality index (WQI) methods were used to comprehensively evaluate the water quality of the lake-type wetland in the level period. Four key water quality parameters were determined by the principal component analysis (PCA) method, and more convenient comprehensive water quality evaluation models, the minimum WQI considering weights (WQImin-w) and the minimum WQI without considering weights (WQImin-nw) were established. The multiple statistical method and the absolute principal component score-multiple liner regression (APCS-MLR) model were combined to analyse the lake pollution sources based on the spatial changes in pollutants. The findings demonstrated that the WQImin-nw model's water quality evaluation outcome was more accurate when weights were not taken into account. The WQImin-nw model can be used as a simple and convenient way to comprehend the variations in water quality in wetlands of lakes and reservoirs. It was concluded that the comprehensive water quality in the study area was at a "medium" level, and CODMn was the main limiting factor. Nonpoint source pollution (such as agricultural planting and livestock breeding) was the most important factor affecting the water quality of Xianghai Lake (with a comprehensive contribution rate of 31.65%). The comprehensive contribution rates of sediment endogenous and geological sources, phytoplankton and other plants, and water diversion and other hydrodynamic impacts accounted for 25.12%, 19.65%, and 23.58% of the total impact, respectively. This study can provide a scientific method for water quality assessment and management of lake wetlands, and an effective support for migration of migratory birds, habitat protection and grain production security.

Chlorophyll-a prediction in tropical reservoirs as a function of hydroclimatic variability and water quality

The study goal was to determine spatiotemporal variations in chlorophyll-a (Chl-a) concentration using models that combine hydroclimatic and nutrient variables in 150 tropical reservoirs in Brazil. The investigation of seasonal variability indicated that Chl-a varied in response to changes in total nitrogen (TN), total phosphorus (TP), volume (V), and daily precipitation (P). Therefore, an empirical model for Chl-a prediction based on the product of TN, TP, and normalized functions of V and P was proposed, but their individual exponents as well as a general multiplicative factor were adjusted by linear regression for each reservoir. The fitted relationships were capable of representing algal temporal dynamics and blooms, with an average coefficient of determination of R2 = 0.70. The results revealed that nutrients yielded better predictability of Chl-a than hydroclimatic variables. Chl-a blooms presented seasonal and interannual variability, being more frequent in periods of high precipitation and low volume. The equations demonstrate different Chl-a responses to the parameters. In general, Chl-a was positively related to TN and/or TP. However, in some cases (22%), high nutrient concentrations reduced Chl-a, which was attributed to limited phytoplankton growth driven by light deficiency due to increased turbidity. In 49% of the models, precipitation intensified Chl-a levels, which was related to increases in the nutrient concentration from external sources in rural watersheds. Contrastingly, 51% of the reservoirs faced a decrease in Chl-a with precipitation, which can be explained by the opposite effect of dilution of nutrient concentration at the reservoir inlet in urban watersheds. In terms of volume, in 67% of the reservoirs, water level reduction promoted an increase in Chl-a as a response to higher nutrient concentration. In the other cases, Chl-a decreased with lower water levels due to wind-induced destratification of the water column, which potentially decreased the internal nutrient release from bottom sediment. Finally, applying the model to the two largest studied reservoirs showed greater sensitivity of Chl-a to changes in water use classes regarding variations in TN, followed by TP, V, and P.

Assessing the algal population dynamics using multiple machine learning approaches: Application to Macao reservoirs

The quality of reservoir water is important to the health and wellbeing of human and animals. Eutrophication is one of the most serious problems threatening the safety of reservoir water resource. Machine learning (ML) approaches are effective tools to understand and evaluate various environmental processes of concern, such as eutrophication. However, limited studies have compared the performances of different ML models to reveal algal dynamics using time-series data of redundant variables. In this study, the water quality data from two reservoirs in Macao were analyzed by adopting various ML approaches, including stepwise multiple linear regression (LR), principal component (PC)-LR, PC-artificial neuron network (ANN) and genetic algorithm (GA)-ANN-connective weight (CW) models. The influence of water quality parameters on algal growth and proliferation in two reservoirs was systematically investigated. The GA-ANN-CW model demonstrated the best performance in reducing the size of data and interpreting the algal population dynamics data, which displayed higher R-squared, lower mean absolute percentage error and lower root mean squared error values. Moreover, the variable contribution based on ML approaches suggest that water quality parameters, such as silica, phosphorus, nitrogen, and suspended solid have a direct impact on algal metabolisms in two reservoirs' water systems. This study can expand our capacity in adopting ML models in predicting algal population dynamics based on time-series data of redundant variables.

Does water temperature influence in microcystin production? A case study of Billings Reservoir, São Paulo, Brazil

We investigated the relationship between some water quality parameters and microcystin, chlorophyll-a, and cyanobacteria in different conditions of water temperature. We also proposed to predict chlorophyll-a concentration in the Billings Reservoir using three machine learning techniques. Our results indicate that in the condition of higher water temperatures with high density of cyanobacteria, microcystin concentration can increase severely (>10² μg/L). Besides the magnitude observed in higher concentrations, in water temperatures above 25.3 °C (classified as high extreme event), higher frequencies of inadequate values of microcystin (87.5%), chlorophyll-a (70%), and cyanobacteria (82.5%) compared to cooler temperatures (<19.6 °C) were observed. The prediction of chlorophyll-a in Billings Reservoir presented good results (0.76 ≤ R² ≤ 0.82; 0.17 ≤ RMSE≤0.20) using water temperature, total phosphorus, and cyanobacteria as predictors, with the best result using Support Vector Machine.

Predictive modeling of pharmaceutical product removal by a managed aquifer recharge system: Comparison and optimization of models using ensemble learners

Pharmaceutical products (PPs) are emerging water pollutants with adverse environmental and health-related impacts, owing to their toxic, persistent, and undetectable microscopic nature. Globally, increasing scientific knowledge and advanced technologies have allowed researchers to study PP-associated problems and their removal for water reuse. Experimental modeling methods require laborious, lengthy, expensive, and environmentally hazardous lab-work to optimize the process. On the other hand, predictive machine learning (ML) models can trace the complex input-output relationship of a process using available datasets. In this study, ensemble ML techniques, including decision tree (DT), random forest (RF), and Xtreme gradient boost (XGB), were used to explore PP (diclofenac, iopromide, propranolol, and trimethoprim) removal by a managed aquifer recharge (MAR) system. The model input parameters included characteristics of reclaimed water and soil used in the columns, pH, dissolved organic carbon, operating time, nitrogen dioxide, sulfate, nitrate, electrical conductivity, manganese, and iron. The selected PP removal was the model output. Datasets were collected through a one-year experimental study of continuous MAR system operation to predict the removal of PPs. DT, RF, and XGB models were then developed for one of the selected compounds and tested for the others to check the reliability of the ML model results. The developed models were assessed using statistical performance matrices. The experimental results showed >80% removal of propranolol and trimethoprim; however, removal of diclofenac and iopromide was only ≈50% by the MAR system. The proposed DT and RF models presented higher coefficients of determination (R² ≥ 0.92) for diclofenac, propranolol, and trimethoprim than for iopromide (R² ≤ 0.63). In contrast, the XGB model showed better results for diclofenac, iopromide, propranolol, and trimethoprim, with R² values of 0.92, 0.72, 0.96, and 0.97, respectively. Therefore, XGB could be the best predictive model to provide insight into the adaptation of ML models to predict PP removal by the MAR system, thereby minimizing experimental work.