Prediction of the removal efficiency of emerging organic contaminants based on design and operational parameters of constructed wetlands

https://doi.org/10.1016/j.jece.2020.104592Get rights and content

Highlights

  • Design and operational parameters of CWs can predict removal efficiency of EOCs.

  • OLR, HLR, and HRT are the most significant predictors compared with depth and area.

  • A combination of two-four predictors resulted in the acceptable models.

  • A novel decision support tool (REOCW-DOP) readily estimates removal efficacy of EOCs.

Abstract

This study investigates the prediction of removal efficiency of pharmaceuticals (PhCs), personal care products (PCPs), and steroidal hormones (SHs) based on design and operational parameters (depth, area, hydraulic loading rate-HLR, organic loading rate-OLR, and hydraulic retention time-HRT) of constructed wetlands (CWs). A comprehensive statistical analysis was performed by applying principle component analysis, correlation and multiple linear regression analyses. The data used in this analysis was compiled from peer reviewed publications. The CWs design and operational parameters are good predictors of the removal efficiency of these emerging organic contaminants. Operational parameters (HLR, OLR, and HRT) are the most significant predicators and combination with design parameters (depth and area) often improved reliability of the predictions. The best predictive models for PhCs and PCPs were composed of depth, OLR, and HRT (root mean square errors-RMSEs: training set: 7–14 %; test set: 22–27 %). A combination of area, HLR, and OLR formed a credible model for predicting the removal efficiency of SHs (RMSEs: training set: 6 %; test set: 11 %). Similarly, generic models by combining data of PhCs and PCPs, PhCs and SHs, PCPs and SHs, and PhCs, PCPs, and SHs showed acceptable performance. The best performing combined model for the prediction of PhCs, PCPs, and SHs was based on area, HLR, OLR, and HRT (RMSEs: training set: 13 %; test set: 22 %). The information obtained by the use of these models may guide researchers, practitioners, policy makers, and citizens in enhancing knowledge and understanding for the design and operation of CWs in the field conditions.

Graphical abstract

  1. Download : Download high-res image (228KB)
  2. Download : Download full-size image

Prediction of the removal efficiency of emerging organic contaminants based on design and operational parameters of constructed wetlands.

Introduction

Constructed wetlands (CWs) are nature-based treatment technologies that have been extensively investigated for the removal of emerging organic contaminants (EOCs) from wastewater [[1], [2], [3], [4], [5]]. To date, a large number of individual case studies have been published in peer reviewed journals related to the removal of EOCs by CWs. Several studies indicated that design and operational parameters such as depth, area, hydraulic loading rate (HLR), organic loading rate (OLR), and hydraulic retention time (HRT) are among the major governing factors in the removal of EOCs by CWs because these parameters considerably influence the possible removal mechanisms such as biodegradation and adsorption/sorption (e.g., [[6], [7], [8]]). The design and operational factors play an important role in the treatment process in CWs for the removal of pharmaceuticals (PhCs), personal care products (PCPs), and steroidal hormones (SHs) [[9], [10], [11]]. For instance, a significant correlation was found between the removal efficiency of some of the studied PhCs and design as well as operational parameters (depth, area, HLR, OLR, and HRT) [9].

However, the previous research did not investigate the possibility of predicting the removal efficiency by using design and operational parameters. For example, multiple linear regression analysis to develop predictive models for removal efficiency of PhCs, PCPs, and SHs with design and operational parameters was not conducted by the above-mentioned studies. Similarly, a principal component analysis (PCA) was not conducted to understand the role of different design and operational parameters of CWs in removal processes. Thus, a comprehensive statistical analysis is lacking to develop predictive models for the removal efficiency of EOCs (PhCs, PCPs, and SHs) based on design and operational parameters of CWs.

Moreover, research is needed on all aspects of modelling of EOCs removal in CWs [12]. The CWs models are broadly classified as process-based and black-box [[12], [13], [14]]. The widely studied process-based models of CWs are RTD/GPS-X [15], Diph-M [16], FITOVERT [17], HYDRUS-CW2D [18], HYDRUS-CWM1 [19], CWM1-RETRASO [20], CFD Model [21], and BIO-PORE [22]. These models mainly focus on the removal of conventional parameters such as chemical oxygen demand (COD), nitrogen, phosphorus, and ammonia. The process-based models focus on the hydraulic, chemical, and biological mechanisms occurring in CWs. While these models attempt to cover in detail the processes happening in CWs, most of them only include to model one or few processes (e.g., hydraulic, reactive-transport, biochemical, plants, and clogging), thus, partly representing several processes happening at the same times [[12], [13], [14]]. Including more processes in available CW models is still a big challenge because it increases complexity as well as the number of parameters to quantify, and lack comprehensive experimental data for calibration and validation. On the other hand, black-box models, which are data driven models, such as regression models, first-order models, time-dependent retardation model, tank-in-series model, Monod models, neural networks, and statistical approaches, mainly focus on input and output rather than processes [13]. Although these models are formed using experimental data, the application possibilities are limited to the range of data used in their development. Meyer et al. [14] suggested that all types of models (both process-based and black-box) are valuable, although some give more understanding of the scientific processes and some are more useful for engineers to design the CW systems. In general, there is a need to include EOCs in further development of CW models.

Therefore, the main objective of this study is to comprehensively analyze the possibility of developing reliable predictive models for removal efficiency of EOCs based on design and operational parameters of CWs. The specific objectives are: (1) to develop predictive models in the form of multiple linear regression equations between removal efficiency of EOCs (PhCs, PCPs, and SHs), and design and operational parameters of CWs; (2) to develop a generic predictive model for the removal efficiencies of PhCs, PCPs, and SHs based on design and operational parameters of CWs; (3) to examine the uncertainties in the prediction process; and (4) to discuss potential applications of the developed predictive models.

Section snippets

Data

In this study, the predictive models were formulated using data of widely studied EOCs (31 PhCs, 13 PCPs, and eight SHs) in CWs. The data were collected from the peer reviewed published sources. All the sources of data are acknowledged in the database compiled in our previous work [10,11,23], which formed the basis of this study. The removal efficiency is used as an independent variable to be predicted. The dependent variables were design and operational parameters: depth, area, HLR, OLR, and

Predictive models for PhCs

The first two principal components (PCs) with eigenvalue > 1.0 explained 70 % variance in the data of 31 widely studied PhCs (Fig. 1 and Table 1). The first PC could explain 49 % variance in the data. In PC1, the high positive loadings were observed for OLR, and high negative loadings for depth, area, and HLR (Fig. 2). In PC2, area indicated high positive loadings; however, HRT showed high negative loading (Fig. 2 and Table 1). The removal efficiency indicated positive loadings in these two

Further development and applications

The predictive models developed in this study could provide useful information in the decision-making process. For example, initial estimates of removal efficiencies could be generated using design and operational parameters of a CW system. Thus, the developed models could serve as useful screening tools to provide first-hand information on the expected removal of PhCs, PCPs, and SHs in CWs. Similarly, preliminary values of design and operational parameters can be generated from the proposed

Conclusions

The following conclusions are drawn from this study:

  • 1

    The design and operational parameters of CWs are good predictors of the removal efficiency of EOCs in CWs. Operational parameters, HLR, OLR, and HRT are the most significant predicators followed by design parameters (depth and area). Thus, plausible predictive models were formulated by combining these parameters, and acceptable models contained two-four of these parameters.

  • 2

    The removal efficiency of PhCs, PCPs, and SHs in CWs could be predicted

Declaration of Competing Interest

The authors report no declarations of interest.

References (60)

  • A. Petitjean et al.

    Modelling aerobic biodegradation in vertical flow sand filters: impact of operational considerations on oxygen transfer and bacterial activity

    Water Res.

    (2012)
  • D. Giraldi et al.

    FITOVERT: a dynamic numerical model of subsurface vertical flow constructed wetlands

    Environ. Model. Softw.

    (2010)
  • T.G. Pálfy et al.

    The verification of the Constructed Wetland Model No. 1 implementation in HYDRUS using column experiment data

    Ecol. Eng.

    (2014)
  • E. Llorens et al.

    CWM1 implementation in RetrasoCodeBright: first results using horizontal subsurface flow constructed wetland data

    Chem. Eng. J.

    (2011)
  • A.R. Rajabzadeh et al.

    Multiphysics modelling of flow dynamics, biofilm development and wastewater treatment in a subsurface vertical flow constructed wetland mesocosm

    Ecol. Eng.

    (2015)
  • R. Samsó et al.

    BIO_PORE, a mathematical model to simulate biofilm growth and water quality improvement in porous media: application and calibration for constructed wetlands

    Ecol. Eng.

    (2013)
  • S. Rühmland et al.

    Fate of pharmaceuticals in a subsurface flow constructed wetland and two ponds

    Ecol. Eng.

    (2015)
  • Y.-J. Choi et al.

    Removal characteristics and mechanism of antibiotics using constructed wetlands

    Ecol. Eng.

    (2016)
  • J. Chen et al.

    Removal of steroid hormones and biocides from rural wastewater by an integrated constructed wetland

    Sci. Total Environ.

    (2019)
  • J. Nivala et al.

    Dynamics of emerging organic contaminant removal in conventional and intensified subsurface flow treatment wetlands

    Sci. Total Environ.

    (2019)
  • O. Carranza-Diaz et al.

    Removal of selected organic micropollutants in planted and unplanted pilot-scale horizontal flow constructed wetlands under conditions of high organic load

    Ecol. Eng.

    (2014)
  • V. Matamoros et al.

    A comparative assessment of intensive and extensive wastewater treatment technologies for removing emerging contaminants in small communities

    Water Res.

    (2016)
  • J. Vymazal et al.

    Occurrence and removal of pharmaceuticals in four full-scale constructed wetlands in the Czech Republic – the first year of monitoring

    Ecol. Eng.

    (2017)
  • Y. Wang et al.

    Bioaccumulation behaviour of pharmaceuticals and personal care products in a constructed wetland

    Chemosphere

    (2019)
  • A. Dordio et al.

    Constructed wetlands with light expanded clay aggregates for agricultural wastewater treatment

    Sci. Total Environ.

    (2013)
  • T.-C. Chen et al.

    Estrogen degradation and sorption onto colloids in a constructed wetland with different hydraulic retention times

    J. Hazard. Mater.

    (2014)
  • B. Petrie et al.

    Biotic phase micropollutant distribution in horizontal sub-surface flow constructed wetlands

    Sci. Total Environ.

    (2018)
  • D.Q. Zhang et al.

    Removal of pharmaceutical compounds in tropical constructed wetlands

    Ecol. Eng.

    (2011)
  • D.Q. Zhang et al.

    Pharmaceutical removal in tropical subsurface flow constructed wetlands at varying hydraulic loading rates

    Chemosphere

    (2012)
  • D.Q. Zhang et al.

    Batch versus continuous feeding strategies for pharmaceutical removal by subsurface flow constructed wetland

    Environ. Pollut.

    (2012)
  • Cited by (12)

    • Nature-based solutions for nutrient pollution control in European agricultural regions: A literature review

      2023, Ecological Engineering
      Citation Excerpt :

      This information alone is important in order to evaluate the applicability and expected performance of the NBS. Furthermore, we investigated the relationship between climate, landscape and design of NBS and their performance, following a procedure similar to the one recently applied by Ilyas et al., 2021. As a preliminary step, we applied Multiple Correspondence Analysis (MCA) in order to identify the most and least significant variables, thus reducing the problem complexity, while also checking whether the data spontaneously collected among meaningful clusters.

    • Machine learning exhibited excellent advantages in the performance simulation and prediction of free water surface constructed wetlands

      2022, Journal of Environmental Management
      Citation Excerpt :

      Although numerical simulations are not limited by site and weather conditions, their construction and parameter calibration processes require considerable effort (May and Sivakumar, 2009). Different results obtained from various test treatments may lead to considerable challenges in optimizing model parameters (Ilyas et al., 2021; Tomenko et al., 2007). More importantly, there are a number of deficiencies in batch processing, weighing multiple test data, and accompanying errors.

    View all citing articles on Scopus
    View full text