Developing a new approach for design support of subsurface constructed wetland using machine learning algorithms

Meta-analysis review for pilot and large-scale constructed wetlands: Design parameters, treatment performance, and influencing factors

Despite their longstanding use in environmental remediation, constructed wetlands (CWs) are still topical due to their sustainable and nature-based approach. While research and review publications have grown annually by 7.5 % and 37.6 %, respectively, from 2018 to 2022, a quantitative meta-analysis employing advanced statistics and machine learning to assess CWs has not yet been conducted. Further, traditional statistics of mean ± standard deviation could not convey the extent of confidence or uncertainty in results from CW studies. This study employed a 95 % bootstrap-based confidence interval and out-of-bag Random Forest-based driver analysis on data from 55 studies, totaling 163 cases of pilot and full-scale CWs. The study recommends, with 95 % confidence, median surface hydraulic loading rates (HLR) of 0.14 [0.11, 0.17] m/d for vertical flow-CWs (VF) and 0.13 [0.07, 0.22] m/d for horizontal flow-CWs (HF), and hydraulic retention time (HRT) of 125.14 [48.0, 189.6] h for VF, 72.00 [42.00, 86.28] h for HF, as practical for new CW design. Permutation importance results indicate influent COD impacted primarily on COD removal rate at 21.58 %, followed by HLR (16.03 %), HRT (12.12 %), and substrate height (H) (10.90 %). For TN treatment, influent TN and COD were the most significant contributors at 12.89 % and 10.01 %, respectively, while H (9.76 %), HRT (9.72 %), and HLR (5.87 %) had lower impacts. Surprisingly, while HRT and H had a limited effect on COD removal, they substantially influenced TN. This study sheds light on CWs' performance, design, and control factors, guiding their operation and optimization.

Machine learning assisted combined systems of wastewater treatment plants with constructed wetlands optimal decision-making

This study proposed an efficient framework for optimizing the design and operation of combined systems of wastewater treatment plants (WWTP) and constructed wetlands (CW). The framework coupled a WWTP model with a CW model and used a multi-objective evolutionary algorithm to identify trade-offs between energy consumption, effluent quality, and construction cost. Compared to traditional design and management approaches, the framework achieved a 27 % reduction in WWTP energy consumption or a 44 % reduction in CW cost while meeting strict effluent discharge limits for Chinese WWTP. The framework also identified feasible decision variable ranges and demonstrated the impact of different optimization strategies on system performance. Furthermore, the contributions of WWTP and CW in pollutant degradation were analyzed. Overall, the proposed framework offers a highly efficient and cost-effective solution for optimizing the design and operation of a combined WWTP and CW system.

Estimating ammonium changes in pilot and full-scale constructed wetlands using kinetic model, linear regression, and machine learning

Constructed wetlands (CWs) are a widely utilized nature-based wastewater treatment method for various effluents. However, their application has been more focused on pilot and full-scale CWs with substantial surface areas and extended operation times, which hold greater relevance in practical scenarios. This study used kinetics, linear regression (LR), and machine learning (ML) models to estimate effluent ammonium in pilot and full-scale CWs. From screening 1476 papers, 24 pilot and full-scale CW studies were selected to extract data containing 15 features and 975 data points. Nine models were fit to this data, revealing that linear models were less effective in capturing CW effluent compared to nonlinear ML algorithms. For training data, the Monod kinetic model predicted the poorest performance with an RMSE of 41.84 mg/L and R2 of 0.34, followed by simple LR (RMSE 24.29 mg/L and R2 0.77) and multiple LR (RMSE 22.63 mg/L and R2 0.80). In contrast, Cubist and Random Forest achieved high performances, with an average RMSE of 12.01 ± 5.38 and an average R2 of 0.93 ± 0.07 for Cubist, and an average RMSE of 15.94 ± 10.69 and an average R2 of 0.91 ± 0.08 for RF. The trained Random Forest performed the best for new data, with an R2 of 0.93 and RMSE of 13.48 mg/L. This ML-based model is a valuable tool for efficiently estimating effluent ammonium concentration in pilot and full-scale CWs, thereby facilitating the design of systems.

Machine learning-guided determination of Acinetobacter density in waterbodies receiving municipal and hospital wastewater effluents

A smart artificial intelligent system (SAIS) for Acinetobacter density (AD) enumeration in waterbodies represents an invaluable strategy for avoidance of repetitive, laborious, and time-consuming routines associated with its determination. This study aimed to predict AD in waterbodies using machine learning (ML). AD and physicochemical variables (PVs) data from three rivers monitored via standard protocols in a year-long study were fitted to 18 ML algorithms. The models' performance was assayed using regression metrics. The average pH, EC, TDS, salinity, temperature, TSS, TBS, DO, BOD, and AD was 7.76 ± 0.02, 218.66 ± 4.76 µS/cm, 110.53 ± 2.36 mg/L, 0.10 ± 0.00 PSU, 17.29 ± 0.21 °C, 80.17 ± 5.09 mg/L, 87.51 ± 5.41 NTU, 8.82 ± 0.04 mg/L, 4.00 ± 0.10 mg/L, and 3.19 ± 0.03 log CFU/100 mL respectively. While the contributions of PVs differed in values, AD predicted value by XGB [3.1792 (1.1040-4.5828)] and Cubist [3.1736 (1.1012-4.5300)] outshined other algorithms. Also, XGB (MSE = 0.0059, RMSE = 0.0770; R² = 0.9912; MAD = 0.0440) and Cubist (MSE = 0.0117, RMSE = 0.1081, R² = 0.9827; MAD = 0.0437) ranked first and second respectively, in predicting AD. Temperature was the most important feature in predicting AD and ranked first by 10/18 ML-algorithms accounting for 43.00-83.30% mean dropout RMSE loss after 1000 permutations. The two models' partial dependence and residual diagnostics sensitivity revealed their efficient AD prognosticating accuracies in waterbodies. In conclusion, a fully developed XGB/Cubist/XGB-Cubist ensemble/web SAIS app for AD monitoring in waterbodies could be deployed to shorten turnaround time in deciding microbiological quality of waterbodies for irrigation and other purposes.

Effects of two plant species combined with slag-sponges on the treatment performance of contaminated saline water in constructed wetland

Constructed wetland (CW), an ecological water treatment system, can purify and repair the damaged saline water body in an open watershed, but its repairing function is limited at low temperature under salt stress. In this study, two different plant species with slag-sponge layer were operated to enhance the purification effect of CW on the damaged saline water body. The results showed that the combination of Scirpus mariqueter and slag-sponges in CW had a better purification effect especially under the condition of salinity of 10‰ (S = 10) with a respective removal efficiency of 91.04% of total nitrogen, 80.07% of total phosphorus, and 93.02% of COD in high temperature (25 ~ 35 °C). Furthermore, ecological traits (enzyme activity and amino acids) of plants, the abundance and distribution of functional microorganisms on the surface of slag-sponges, and the microbial state on the substrate surface of the denitrifying zone of CW were analyzed to explain how exactly the combinations worked. It was found that the enrichment of functional microorganisms in slag-sponge and the anaerobic zone of plants have improved the nitrogen and phosphorus removal. Plants maintained high enzyme activities and the ability to synthesize key amino acids under salt stress to ensure the growth and reproduction of plants and achieve the assimilation function. Scirpus mariqueter combined with slag-sponges in CW effectively improved the purification effect of damaged saline water, indicating that it is an ecological and green saline water treatment way.

Removal of bacterial indicators in on-site two-stage multi-soil-layering plant under arid climate (Morocco): prediction of total coliform content using K-nearest neighbor algorithm

This study aims to evaluate and monitor the efficacy of a full-scale two-stage multi-soil-layering (TS-MSL) plant in removing fecal contamination from domestic wastewater. The TS-MSL plant under investigation consisted of two units in series, one with a vertical flow regime (VF-MSL) and the other with a horizontal flow regime (HF-MSL). Furthermore, this study attempts to see whether linear model (LM) and K-nearest neighbor (KNN) model can be used to predict total coliform (TC) removal in the TS-MSL system. For 24 months, the TS-MSL system was monitored, with bimonthly measurements recorded at the inlet and outlet of each compartment. Obtained results show removal of 85% of COD, 67% of TP, 27% of TN, and 3 log units of coliforms with good system stability. Thus, the effluent meets the Moroccan water quality code for reuse in the irrigation of green spaces. In addition, as compared to LM, the KNN model (R² = 0.988) may be considered as an effective method for predicting TC removal in the TS-MSL system. Finally, sensitivity analysis has shown that TC and dissolved oxygen level in the influent were the most influential parameters for predicting TC removal in the TS-MSL system.

Modelling hydrological strength and alteration in moribund deltaic India

The Ganga-Brahmaputra moribund deltaic floodplain region hosted many socio-ecologically precious freshwater wetland ecosystems experiencing hydrological alteration. The present study aimed to model hydrological strength (HS) to show the spatial difference and account for the degree and direction of hydrological alteration of Indian moribund deltaic wetland in three phases e.g. (1) phase I (1988-1997), (2) phase II (1998-2007) and phase III (2008-2017). Three key hydrological parameters, such as Water Presence Frequency (WPF), water depth, and hydro-period were considered for hydrological strength modelling using two ensemble Machine Learning (ML) techniques (Random Forest (RF) and XGBoost). Image algebra was employed for phasal change detection. Hydrological strength models show that around 75% of the wetland area was lost in-between phases I to III and the loss was found more intensive in moderate and weak HS zones. Existing wetland shows a clear spatial difference of HS between wetland core and periphery and river linked and delinked or not linked wetlands. Regarding the suitability of the ML models, both are acceptable, however, the XGBoost outperformed in reference to applied 15 statistical validation techniques and field evidence. HS models based on change detection clarified that more than 22% and 55% of the weak HS zone in phases II and III respectively were turned into non-wetland. The degree of alteration revealed that about 40% of wetland areas experienced a negative alteration during phases I to II, and this proportion increased to 63% in between phases II to III. Since the study figured out the spatial nature of HS, degree and direction of alteration at a spatial scale, these findings would be instrumental for adopting rational planning towards wetland conservation and restoration.

Generating 3D Geothermal Maps in Catalonia, Spain Using a Hybrid Adaptive Multitask Deep Learning Procedure

Mapping the subsurface temperatures can efficiently lead to identifying the geothermal distribution heat flow and potential hot spots at different depths. In this paper, an advanced adaptive multitask deep learning procedure for 3D spatial mapping of the subsurface temperature was proposed. As a result, predictive 3D spatial subsurface temperatures at different depths were successfully generated using geolocation of 494 exploratory boreholes data in Catalonia (Spain). To increase the accuracy of the achieved results, hybridization with a new modified firefly algorithm was carried out. Subsequently, uncertainty analysis using a novel automated ensemble deep learning approach for the predicted temperatures and generated spatial 3D maps were executed. Comparing the accuracy performances in terms of correct classification rate (CCR) and the area under the precision–recall curves for validation and whole datasets with at least 4.93% and 2.76% improvement indicated for superiority of the hybridized model. According to the results, the efficiency of the proposed hybrid multitask deep learning in 3D geothermal characterization to enhance the understanding and predictability of subsurface spatial distribution of temperatures is inferred. This implies that the applicability and cost effectiveness of the adaptive procedure in producing 3D high resolution depth dependent temperatures can lead to locate prospective geothermally hotspot active regions.

Analysis of changes in air pollution quality and impact of COVID-19 on environmental health in Iran: application of interpolation models and spatial autocorrelation

In the global COVID-19 epidemic, humans are faced with a new challenge. The concept of quarantine as a preventive measure has changed human activities in all aspects of life. This challenge has led to changes in the environment as well. The air quality index is one of the immediate concrete parameters. In this study, the actual potential of quarantine effects on the air quality index and related variables in Tehran, the capital of Iran, is assessed, where, first, the data on the pollutant reference concentration for all measuring stations in Tehran, from February 19 to April 19, from 2017 to 2020, are monitored and evaluated. This study investigated the hourly concentrations of six particulate matters (PM), including PM2.5, PM10, and air contaminants such as nitrogen dioxide (NO₂), sulfur dioxide (SO2), ozone (O3), and carbon monoxide (CO). Changes in pollution rate during the study period can be due to reduced urban traffic, small industrial activities, and dust mites of urban and industrial origins. Although pollution has declined in most regions during the COVID-19 quarantine period, the PM2.5 rate has not decreased significantly, which might be of natural origins such as dust. Next, the air quality index for the stations is calculated, and then, the interpolation is made by evaluating the root mean square (RMS) of different models. The local and global Moran index indicates that the changes and the air quality index in the study area are clustered and have a high spatial autocorrelation. The results indicate that although the bad air quality is reduced due to quarantine, major changes are needed in urban management to provide favorable conditions. Contaminants can play a role in transmitting COVID-19 as a carrier of the virus. It is suggested that due to the rise in COVID-19 and temperature in Iran, in future studies, the effect of increased temperature on COVID-19 can be assessed.