Performance analysis of the water quality index model for predicting water state using machine learning techniques

Process Safety and Environmental Protection, Год журнала: 2022, Номер 169, С. 808 - 828

Опубликована: Ноя. 28, 2022

Existing water quality index (WQI) models assess using a range of classification schemes. Consequently, different methods provide number interpretations for the same properties that contribute to considerable amount uncertainty in correct quality. The aims this study were evaluate performance model order classify coastal correctly completely new scheme. Cork Harbour data was used study, which collected by Ireland's environmental protection agency (EPA). In present four machine-learning classifier algorithms, including support vector machines (SVM), Naïve Bayes (NB), random forest (RF), k-nearest neighbour (KNN), and gradient boosting (XGBoost), utilized identify best predicting classes widely seven WQI models, whereas three are recently proposed authors. KNN (100% 0% wrong) XGBoost (99.9% 0.1% algorithms outperformed accurately models. validation results indicate outperformed, accuracy (1.0), precision (0.99), sensitivity specificity F1 (0.99) score, predict Moreover, compared higher prediction accuracy, precision, sensitivity, specificity, score found weighted quadratic mean (WQM) unweighted root square (RMS) respectively, each class. findings showed WQM RMS could be effective reliable assessing terms classification. Therefore, helpful providing accurate information researchers, policymakers, research personnel monitoring more effectively.

Язык: Английский

---

Optimization of water quality index models using machine learning approaches

Water Research, Год журнала: 2023, Номер 243, С. 120337 - 120337

Опубликована: Июль 11, 2023

Язык: Английский

---

A sophisticated model for rating water quality

The Science of The Total Environment, Год журнала: 2023, Номер 868, С. 161614 - 161614

Опубликована: Янв. 18, 2023

Here, we present the Irish Water Quality Index (IEWQI) model for assessing transitional and coastal water quality in an effort to improve method develop a tool that can be used by environmental regulators abate pollution Ireland. The developed has been associated with adoption of standards formulated waterbodies according framework directive legislation regulator water. consists five identical components, including (i) indicator selection technique is select crucial indicator; (ii) sub-index (SI) function rescaling various indicators' information into uniform scale; (iii) weight estimating values based on relative significance real-time quality; aggregation computing index (WQI) score; (v) score interpretation scheme state quality. IEWQI was Cork Harbour, applied four Ireland, using 2021 data summer winter seasons order evaluate sensitivity terms spatio-temporal resolution waterbodies. efficiency uncertainty were also analysed this research. In different magnitudes domains, shows higher application domains during winter. addition, results reveal architecture may effective reducing avoid eclipsing ambiguity problems. findings study could efficient reliable assessment more accurately any geospatial domain.

Язык: Английский

---

Assessing and forecasting water quality in the Danube River by using neural network approaches

The Science of The Total Environment, Год журнала: 2023, Номер 879, С. 162998 - 162998

Опубликована: Март 24, 2023

The health and quality of the Danube River ecosystems is strongly affected by nutrients loads (N P), degree contamination with hazardous substances or oxygen depleting substances, microbiological changes in river flow patterns sediment transport regimes. Water index (WQI) an important dynamic attribute characterization quality. WQ scores do not reflect actual condition water We proposed a new forecast scheme for based on following qualitative classes very good (0-25), (26-50), poor (51-75), (76-100) extremely polluted/non-potable (>100). forecasting using Artificial Intelligence (AI) meaningful method protecting public because its possibility to provide early warning regarding harmful pollutants. main objective present study WQI time series data physical, chemical status parameters associated scores. Cascade-forward network (CFN) models, along Radial Basis Function Network (RBF) as benchmark model, were developed from 2011 2017 forecasts produced period 2018-2019 at all sites. nineteen input features represent initial dataset. Moreover, Random Forest (RF) algorithm refines dataset selecting eight considered most relevant. Both datasets are employed constructing predictive models. According results appraisal, CFN models better outcomes (MSE = 0.083/0,319 R-value 0.940/0.911 quarter I/quarter IV) than RBF In addition, show that both could be effective predicting when relevant used variables. Also, CFNs accurate short-term curves which reproduce first fourth quarters (the cold season). second third presented slightly lower accuracy. reported clearly demonstrate successfully they may learn historic determine nonlinear relationships between output

Язык: Английский

---

Assessing the impact of land use and land cover on river water quality using water quality index and remote sensing techniques

Environmental Monitoring and Assessment, Год журнала: 2023, Номер 195(4)

Опубликована: Март 8, 2023

Язык: Английский

---

Marine waters assessment using improved water quality model incorporating machine learning approaches

Journal of Environmental Management, Год журнала: 2023, Номер 344, С. 118368 - 118368

Опубликована: Июнь 24, 2023

In marine ecosystems, both living and non-living organisms depend on "good" water quality. It depends a number of factors, one the most important is quality water. The index (WQI) model widely used to assess quality, but existing models have uncertainty issues. To address this, authors introduced two new WQI models: weight based weighted quadratic mean (WQM) unweighted root squared (RMS) models. These were in Bay Bengal, using seven indicators including salinity (SAL), temperature (TEMP), pH, transparency (TRAN), dissolved oxygen (DOX), total oxidized nitrogen (TON), molybdate reactive phosphorus (MRP). Both ranked between "fair" categories, with no significant difference models' results. showed considerable variation computed scores, ranging from 68 88 an average 75 for WQM 70 76 72 RMS. did not any issues sub-index or aggregation functions, had high level sensitivity (R2 = 1) terms spatio-temporal resolution waterbodies. study demonstrated that approaches effectively assessed waters, reducing improving accuracy score.

Язык: Английский

---

Developing a novel tool for assessing the groundwater incorporating water quality index and machine learning approach

Groundwater for Sustainable Development, Год журнала: 2023, Номер 23, С. 101049 - 101049

Опубликована: Ноя. 1, 2023

Groundwater plays a pivotal role as global source of drinking water. To meet sustainable development goals, it is crucial to consistently monitor and manage groundwater quality. Despite its significance, there are currently no specific tools available for assessing trace/heavy metal contamination in groundwater. Addressing this gap, our research introduces an innovative approach: the Quality Index (GWQI) model, developed tested Savar sub-district Bangladesh. The GWQI model integrates ten water quality indicators, including six heavy metals, collected from 38 sampling sites study area. enhance precision assessment, employed established machine learning (ML) techniques, evaluating model's performance based on factors such uncertainty, sensitivity, reliability. A major advancement incorporation metals into framework index model. best authors knowledge, marks first initiative develop encompassing heavy/trace elements. Findings assessment revealed that area ranged 'good' 'fair,' indicating most indicators met standard limits set by Bangladesh government World Health Organization. In predicting scores, artificial neural networks (ANN) outperformed other ML models. Performance metrics, root mean square error (RMSE), (MSE), absolute (MAE) training (RMSE = 0.361; MSE 0.131; MAE 0.262), testing 0.001; 0.00; 0.001), prediction evaluation statistics (PBIAS 0.000), demonstrated superior effectiveness ANN. Moreover, exhibited high sensitivity (R2 1.0) low uncertainty (less than 2%) rating These results affirm reliability novel monitoring management, especially regarding metals.

Язык: Английский

---

Predicting the impacts of urban development on urban thermal environment using machine learning algorithms in Nanjing, China

Journal of Environmental Management, Год журнала: 2024, Номер 356, С. 120560 - 120560

Опубликована: Март 27, 2024

Язык: Английский

---

Data-driven evolution of water quality models: An in-depth investigation of innovative outlier detection approaches-A case study of Irish Water Quality Index (IEWQI) model

Water Research, Год журнала: 2024, Номер 255, С. 121499 - 121499

Опубликована: Март 20, 2024

Recently, there has been a significant advancement in the water quality index (WQI) models utilizing data-driven approaches, especially those integrating machine learning and artificial intelligence (ML/AI) technology. Although, several recent studies have revealed that model produced inconsistent results due to data outliers, which significantly impact reliability accuracy. The present study was carried out assess of outliers on recently developed Irish Water Quality Index (IEWQI) model, relies techniques. To author's best knowledge, no systematic framework for evaluating influence such models. For purposes assessing outlier (WQ) this first initiative research introduce comprehensive approach combines with advanced statistical proposed implemented Cork Harbour, Ireland, evaluate IEWQI model's sensitivity input indicators quality. In order detect outlier, utilized two widely used ML techniques, including Isolation Forest (IF) Kernel Density Estimation (KDE) within dataset, predicting WQ without these outliers. validating results, five commonly measures. performance metric (R2) indicates improved slightly (R2 increased from 0.92 0.95) after removing input. But scores were statistically differences among actual values, predictions 95% confidence interval at p < 0.05. uncertainty also contributed <1% final assessment using both datasets (with outliers). addition, all measures indicated techniques provided reliable can be detecting their impacts model. findings reveal although had architecture, they moderate rating schemes' This finding could improve accuracy as well helpful mitigating eclipsing problem. provide evidence how influenced reliability, particularly since confirmed effective accurately despite presence It occur spatio-temporal variability inherent indicators. However, assesses underscores important areas future investigation. These include expanding temporal analysis multi-year data, examining spatial patterns, detection methods. Moreover, it is essential explore real-world revised categories, involve stakeholders management, fine-tune parameters. Analysing across varying resolutions incorporating additional environmental enhance assessment. Consequently, offers valuable insights strengthen robustness provides avenues enhancing its utility broader applications. successfully adopted affect current Harbour only single year data. should tested various domains response terms resolution domain. Nevertheless, recommended conducted adjust or revise schemes investigate practical effects updated categories. potential recommendations adaptability reveals effectiveness applicability more general scenarios.

Язык: Английский

---

Assessing water quality of an ecologically critical urban canal incorporating machine learning approaches

Ecological Informatics, Год журнала: 2024, Номер 80, С. 102514 - 102514

Опубликована: Фев. 13, 2024

This study assessed water quality (WQ) in Tongi Canal, an ecologically critical and economically important urban canal Bangladesh. The researchers employed the Root Mean Square Water Quality Index (RMS-WQI) model, utilizing seven WQ indicators, including temperature, dissolve oxygen, electrical conductivity, lead, cadmium, iron to calculate index (WQI) score. results showed that most of sampling locations poor WQ, with many indicators violating Bangladesh's environmental conservation regulations. eight machine learning algorithms, where Gaussian process regression (GPR) model demonstrated superior performance (training RMSE = 1.77, testing 0.0006) predicting WQI scores. To validate GPR model's performance, several measures, coefficient determination (R2), Nash-Sutcliffe efficiency (NSE), factor (MEF), Z statistics, Taylor diagram analysis, were employed. exhibited higher sensitivity (R2 1.0) (NSE 1.0, MEF 0.0) WQ. analysis uncertainty (standard 7.08 ± 0.9025; expanded 1.846) indicates RMS-WQI holds potential for assessing inland waterbodies. These findings indicate could be effective approach waters across study's did not meet recommended guidelines, indicating Canal is unsafe unsuitable various purposes. implications extend beyond contribute management initiatives

Язык: Английский

---