Savvidou Pinelopi, Dotro Gabriela, Campo Pablo, Coulon Frederic, Lyu Tao
School of Water, Energy and Environment, Cranfield University, College Road, Cranfield, Bedfordshire MK43 0AL, United Kingdom.
School of Water, Energy and Environment, Cranfield University, College Road, Cranfield, Bedfordshire MK43 0AL, United Kingdom.
Sci Total Environ. 2024 Jul 15;934:173237. doi: 10.1016/j.scitotenv.2024.173237. Epub 2024 May 17.
Per- and poly-fluoroalkyl substances (PFAS) have emerged as newly regulated micropollutants, characterised by extreme recalcitrance and environmental toxicity. Constructed wetlands (CWs), as a nature-based solution, have gained widespread application in sustainable water and wastewater treatment and offer multiple environmental and societal benefits. Despite CWs potential, knowledge gaps persist in their PFAS removal capacities, associated mechanisms, and modelling of PFAS fate. This study carried out a systematic literature review, supplemented by unpublished experimental data, demonstrating the promise of CWs for PFAS removal from the influents of varying sources and characteristics. Median removal performances of 64, 46, and 0 % were observed in five free water surface (FWS), four horizontal subsurface flow (HF), and 18 vertical flow (VF) wetlands, respectively. PFAS adsorption by the substrate or plant root/rhizosphere was deemed as a key removal mechanism. Nevertheless, the available dataset resulted unsuitable for a quantitative analysis. Data-driven models, including multiple regression models and machine learning-based Artificial Neural Networks (ANN), were employed to predict PFAS removal. These models showed better predictive performance compared to various mechanistic models, which include two adsorption isotherms. The results affirmed that artificial intelligence is an efficient tool for modelling the removal of emerging contaminants with limited knowledge of chemical properties. In summary, this study consolidated evidence supporting the use of CWs for mitigating new legacy PFAS contaminants. Further research, especially long-term monitoring of full-scale CWs treating real wastewater, is crucial to obtain additional data for model development and validation.
全氟和多氟烷基物质（PFAS）已成为新受监管的微污染物，其特点是具有极强的难降解性和环境毒性。人工湿地（CWs）作为一种基于自然的解决方案，已在可持续水和废水处理中得到广泛应用，并带来多种环境和社会效益。尽管人工湿地具有潜力，但在其去除PFAS的能力、相关机制以及PFAS归宿的建模方面仍存在知识空白。本研究进行了系统的文献综述，并辅以未发表的实验数据，证明了人工湿地从不同来源和特征的进水去除PFAS的前景。在五个自由水面（FWS）湿地、四个水平潜流（HF）湿地和18个垂直流（VF）湿地中，分别观察到PFAS的去除性能中位数为64%、46%和0%。认为基质或植物根/根际对PFAS的吸附是关键的去除机制。然而，现有的数据集不适合进行定量分析。采用包括多元回归模型和基于机器学习的人工神经网络（ANN）在内的数据驱动模型来预测PFAS的去除。与包括两种吸附等温线在内的各种机理模型相比，这些模型显示出更好的预测性能。结果证实，在对化学性质了解有限的情况下，人工智能是模拟新兴污染物去除的有效工具。总之，本研究汇总了支持使用人工湿地减轻新的遗留PFAS污染物的证据。进一步的研究，特别是对处理实际废水的全尺寸人工湿地进行长期监测，对于获取模型开发和验证的额外数据至关重要。
