WETLAND ECOLOGICAL RISK ASSESSMENT AND MANAGEMENT: TAKING WENZHOU SANYANG WETLAND AS A CASE STUDY by Ge Xu B.Sc., Wenzhou University, China, 2016 THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF SCIENCE IN NATURAL RESOURCES AND ENVIRONMENTAL STUDIES (ENVIRONMENTAL SCIENCE) UNIVERSITY OF NORTHERN BRITISH COLUMBIA December 2018 © Ge Xu, 2018 ABSTRACT Based on the traditional framework of wetland ecological risk assessment, this thesis proposed a new method for more comprehensive assessment by considering two major pollution types faced by wetlands, including heavy metal pollution and water eutrophication. Artificial neural network (ANN) method was applied to evaluate the eutrophication risk level, while an improved potential ecological risk index was used to estimate the risk of heavy metals in surface sediments. Fuzzy set theory was then used to combine the two risk levels to obtain a general risk level, which could be used for recommending appropriate risk management actions. The Sanyang Wetland in Wenzhou, China was used as a case study to demonstrate the proposed wetland ecological risk assessment method. This thesis indicated that the new framework of wetland ecological assessment could provide a more objective risk level and then give more appropriate suggestions to decision making. Key words: wetland; risk assessment; eutrophication; heavy metal; fuzzy set; Sanyang wetland; scenario analysis ii TABLE OF CONTENTS ABSTRACT ............................................................................................................................. ii Table of Contents ..................................................................................................................... iii List of Tables ............................................................................................................................vi List of Figures ........................................................................................................................ viii ACKNOWLEDGEMENT ......................................................................................................... x CHAPTER 1: INTRODUCTION .............................................................................................. 1 1.1 Background of wetland .................................................................................................... 1 1.2 Wetland Eutrophication ................................................................................................... 1 1.3 Heavy metal pollution in sediments and its hazards ........................................................ 4 1.4 Research objectives .......................................................................................................... 5 1.5 Organization of the thesis ................................................................................................ 5 CHAPTER 2: LITERATURE REVIEW ................................................................................... 6 2.1 Water quality evaluation .................................................................................................. 6 2.2 Eutrophication evaluation ................................................................................................ 7 2.2.1 Conventional evaluation method ............................................................................... 7 2.2.2 Artificial neural network method .............................................................................. 9 2.3 Heavy metal risk evaluation........................................................................................... 11 2.4 Fuzzy Risk Assessment.................................................................................................. 12 2.5 Scenario analysis ............................................................................................................ 13 2.6 Wetland risk assessment ................................................................................................ 15 2.7 Summary ........................................................................................................................ 16 CHAPTER 3: METHODOLOGY ........................................................................................... 18 iii 3.1 Step one: Problem statement .......................................................................................... 18 3.2 Step two: Water quality assessment .............................................................................. 18 3.3 Step three: Risk assessment model ............................................................................... 21 3.3.1 Eutrophication risk assessment ............................................................................... 21 (1) Artificial Neural Network ....................................................................................... 21 (2) Comprehensive nutritional status index .................................................................. 24 3.3.2 Modified potential ecological risk index ................................................................. 26 (1) Risk assessment code (RAC) .................................................................................. 26 (2) Potential ecological risk index ................................................................................ 27 (3) Modified potential ecological risk index ................................................................ 28 3.4 Step four: Overall Risk ................................................................................................. 29 3.5 Step five: Risk Management and Scenario analysis ..................................................... 35 CHAPTER 4: CASE STUDY ................................................................................................. 36 4.1 Overview of the study area ............................................................................................ 36 4.2 Eutrophication risk assessment results........................................................................... 37 4.2.1 Environmental data collection ................................................................................. 37 4.2.2 Water quality status results...................................................................................... 39 4.2.3 Eutrophication risk assessment results .................................................................... 41 (1) The result of artificial neural network (ANN) ........................................................ 41 (2) The result of comprehensive nutritional status index method ................................ 43 4.3 Heavy metal risk assessment results .............................................................................. 44 4.3.1 Sampling and analysis ............................................................................................. 44 4.3.2 Modified ecological risk assessment ....................................................................... 46 (1) Results of total contamination and RAC of heavy metal ........................................ 46 iv (2) Potential ecological risk index assessment ............................................................. 51 4.4 General Risk Levels in Sanyang Wetland...................................................................... 54 4.5 Scenario Analysis ........................................................................................................... 59 4.5.1 Scenario 1: sediment dredging method.....................................................................60 4.5.2 Scenario 2: ecological restoration............................................................................ 66 CHAPTER 5: CONCLUSION ................................................................................................ 73 REFERENCES ........................................................................................................................ 76 v List of Tables Table 3.1: Water quality parameters and their weights ........................................................... 20 Table 3.2: Water Quality Rating for NSF Water Quality Index methods ............................... 21 Table 3.3: Control Standard for Surface Water Eutrophication .............................................. 22 Table 3.4: Correlation between Chl-a and other parameters as well as the weight value ....... 25 Table 3.5: Trophic levels using comprehensive nutritional status index................................. 26 Table 3.6: classification of RAC and values of δ(toxic index)................................................ 27 Table 3.7: Indices and grades of potential ecological risk ....................................................... 28 Table 3.8: Survey results on fuzzy rules .................................................................................. 32 Table 3.9: Recommended risk management actions ............................................................... 35 Table 4.1: Average values of water quality parameters at the sampling locations.................. 40 Table 4.2: WQI Assessment results for the two sampling locations ....................................... 41 Table 4.3: ANN model calculated score of trophic state for Sanyang Wetland ...................... 42 Table 4.4: Assessment results of the comprehensive nutritional status index method in Sanyang wetland ...................................................................................................................... 44 Table 4.5: Concentrations of heavy metal in sediment of Sanyang wetland ........................... 47 Table 4.6: Contamination coefficient and CPI of heavy metals .............................................. 48 Table 4.7: RAC classification of heavy metal in sediments under different land uses ........... 51 Table 4.8: Modified index (Ω) of heavy metal ........................................................................ 51 ~ Table 4.9: Eir, Eir and risk level ............................................................................................... 53 Table 4.10: The result of heavy metal risk assessment under Scenario 1 ............................... 61 Table 4.11: The result of eutrophication risk assessment under Scenario 1 ............................ 61 Table 4.12: The result of heavy metal risk assessment under Scenario 2 ............................... 67 vi Table 4.13: The result of eutrophication risk assessment under Scenario 2 ............................ 67 vii List of Figures Figure 2.1: Schematic illustration of a three-layered feed-forward neural network................ 10 Figure 2.2: The framework of wetland risk assessment(Jose, 1999)....................................... 16 Figure 3.1: The framework of wetland risk assessment .......................................................... 19 Figure 3.2: The framework of general risk. ............................................................................. 31 Figure 3.3: Membership functions of fuzzy general risk levels .............................................. 33 Figure 3.4: Membership functions of fuzzy eutrophication risk level events ......................... 34 Figure 3.5: Membership functions of fuzzy heavy metal risk level events ............................. 34 Figure 4.1: Sampling points in around Sanyang wetland in Wen Zhou .................................. 38 Figure 4.2: Land use distribution around two sampling locations (A6 and S4) ...................... 39 Figure 4.3: Sampling locations and distribution of different land uses (OP, OPRI and RAI) in Sanyang wetland ...................................................................................................................... 45 Figure 4.4: Distribution of chemical species in metal under different land use types, (a) OP land use, (b) RAI land use, (c) OPRI land use......................................................................... 50 Figure 4.5: Comparison of total risk (MRI and RI) of heavy metals under different land uses ................................................................................................................................................. 53 Figure 4.6: Fuzzy inference process under S4; (a) and (d): heavy metal risks, (b) and (e): eutrophication risks, (c), (f) and (g): general risk levels.......................................................... 56 Figure 4.7: Fuzzy inference process under A6; (a) and (d): heavy metal risks, (b) and (e): eutrophication risks, (c), (f) and (g): general risk levels.......................................................... 58 viii Figure 4.8: Fuzzy inference process under A6 of scenario 1; (a) and (d): heavy metal risks, (b) and (e): eutrophication risks, (c), (f) and (g): general risk levels ............................................ 62 Figure 4.9: Fuzzy inference process under S4 of scenario 1; (a) and (d): heavy metal risks, (b) and (e): eutrophication risks, (c), (f) and (g): general risk levels ............................................ 63 Figure 4.10: Fuzzy inference process under A6 of scenario 2; (a) and (d): heavy metal risks, (b) and (e): eutrophication risks, (c), (f) and (g): general risk levels....................................... 69 Figure 4.11: Fuzzy inference process under S4 of scenario 2; (a) and (d): heavy metal risks, (b) and (e): eutrophication risks, (c), (f) and (g): general risk levels....................................... 71 ix ACKNOWLEDGEMENT Each person's life will face numerous choices, and every choice determines the direction of our life. I've been wondering what made me meet UNBC, my teachers, and my friends. When my MSc graduate study life is coming to the end, I know that I will always love and miss UNBC. First and foremost, I would like to express my appreciation to all of my teachers, especially my supervisor Professor Jianbing Li and my co-supervisor Professor Min Zhao. They gave me meticulous care in my study and life. The instructors' profound knowledge, rigorous academic attitude, active and insightful academic thinking, and the working attitude will benefit me for the rest of my life. This thesis was completed under the careful guidance of theses teachers. From the beginning of research topic selection, paper writing and revision, to the final draft, there are a lot of efforts and energy from the instructors. Secondly, I am grateful to my family and friends. Thank my parents for their understanding and tolerance. Thank my friends for helping me to solve the difficult problems and providing their valuable opinions. Their consistent support and care are the driving force and motivation of my progress. Thirdly, I am grateful to UNBC for the rigorous but open academic atmosphere that allows me to think and comprehend better. Every day at UNBC is unforgettable. It embraces all my joys and frustrations. It's like a mother accepting my everything. I've made progress and grown up in UNBC. This is the best gift that this experience has bestowed on me. When I am about to walk out of school, I might feel nervous and hesitant, but I know that with courage and determination, I can challenge everything. Thanks again for all the things here. x CHAPTER 1: INTRODUCTION 1.1 Background of Wetland Wetland provides an irreplaceable ecological service to human ecosystems through the interaction of land and water systems (Qu et al., 2011). It has important ecological service function and ecological value and is an important feature of the global landscape. It was estimated that about 6% of the Earth's land surface is covered by wetlands, while wetland ecosystem provides 15% of the global ecosystem services. As the “kidneys” of the Earth and the multifunctional and biodiversity-rich ecosystem, wetlands provide comprehensive ecological services for flood control and mitigation, climate control, pollution prevention, soil erosion reduction, and biodiversity conservation (Shao et al., 2012). However, wetlands have been suffering from serious degradation and loss due to intense anthropogenic disturbances such as wetland pollution, eutrophication, land use change, and global change (e.g., seawater intrusion) (Shao et al., 2012). Especially, the problem of wetland eutrophication has received great concerns all over the world. Because of its unique hydrological conditions, wetlands are more likely to experience eutrophication and cause serious consequences such as "water bloom". 1.2 Wetland Eutrophication Eutrophication refers to situation that the excessive discharge of nutrients to natural water causes the abnormal propagation and growth of plants. It is often accompanied by many algae breeding which can be divided into naturally caused eutrophication and human 1 caused eutrophication, while the major concern has been on human-induced eutrophication (Cheng, 2012). The OECD (The Organisation for Economic Co-operation and Development) defined eutrophication as a series of changes in the production of algae and aquatic plants to cause water quality decline due to increased nutrient content (Ferreira et al., 2011). China's "Ocean Dictionary" stated eutrophication as "excessive body of nutrients in the water, resulting in large breeding algae and deterioration of water quality" (Yang et al., 2010). Eutrophication process is associated with a series of biological, chemical, and physical changes that can be affected by wetland morphology and many other factors (Cheng, 2012). In general, the eutrophication of wetland is due to the input and output imbalance of nutrients, which results in the breakage of species distribution and thus the infiltration of single species (such as algae), further destroying the energy and material flow in the whole ecosystem. Another reason is due to the longer wetting time in the wetland which results in excessive nutrients in the sediment, thus forming a nutrient exchange between the sediment and the upper water interface (Carrillo et al., 2011). When eutrophication reaches a certain level, it is likely to cause the concentration of algae outbreak and the formation of bloom. Therefore, bloom is one of the typical characteristics of eutrophication. Wetland eutrophication can not only cause significant economic losses, but also seriously endanger human health. According to Smith (2003), eutrophication occurs in nearly half of the damaged lakes and 60% of the blackened rivers in the United States, and eutrophication-related water quality impairment can have a significantly negative economic impact. First, eutrophication will reduce the wetland function and economic benefits. Eutrophication in the water will occur over the breeding of algae, so that the water will produce a musty odour (Cheng, 2012). During the high-temperature period, a large amount of algae proliferation would greatly consume the dissolved oxygen in water, leading to the death 2 and decomposition of many species in the water, and thus producing a strong smell. At the same time, many algae float in the water, causing water to become turbid and thus reducing water transparency (Smith, 2003). Changes of water quality can also cause a negative effect on soil around the wetlands and the surface water sediments, thereby affecting the surrounding terrestrial ecosystems. Depending on different land uses, eutrophication will cause different effects. In addition, when the water experiences eutrophication, many algae can secrete and release toxic and harmful substances, which can not only harm aquatic organisms and damage aquatic ecosystems, but also affect human health (Qu et al., 2011). For example, many cyanobacteria can produce a neurotoxic compound with high degree of hepatotoxicity and cytotoxicity against human and domestic animals (Smith, 2003). Moreover, in the wetland ecosystem, the large number of algae breeding will lead to the remaining species with rapidly reduced living space. For example, local native aquatic plants will be deprived of the algae growth space as well as the oxygen and nutrients in water, which will eventually lead to an aquatic plant ecosystem with single species. Local animals will also be affected by eutrophication through the impacts on food chain (Walden et al., 2004). Therefore, the study of eutrophication has become one of the hot environmental research issues. Many scientists have studied the eutrophic status, hazards, models, control, and governance, and have achieved promising results (Wang et al., 2008). However, the mechanism of water eutrophication formation and the constraint factors are still controversial. It is of fundamental importance to establish the water eutrophication prediction and evaluation model due to the influence of various environmental factors on water eutrophication (Malekmohammadi & Blouchi, 2014). 3 1.3 Heavy metal pollution in sediments and its hazards Heavy metal pollution is an important part of sediment risk assessment, and has attracted much attention due to its toxicity, extensive sources, non-degradability, and cumulative behavior (Yu et al., 2008). Due to the slow flow of water in most wetlands, sediments are more likely to accumulate heavy metals which will seriously affect the aquatic ecosystems (Charkhabi et al. 2005). In addition to harm the organisms in water, heavy metals may also migrate through water flow, causing the spread of pollution. The release of heavy metals from sediments can be hazardous to the water environment, although the toxicity of heavy metals depends on their concentration and exposure pathways (Liu et al. 2009). Heavy metals are inert in the sedimentary environment and are often considered conservative pollutants (Wilcock, 1999). In sediments, heavy metals can exist in a variety of chemical forms that exhibit different physical and chemical behaviors in relation to chemical interactions, mobility, bioavailability, and potential toxicity (Liu et al., 2009). The toxicity of heavy metals in sediment not only depends on their total concentration, but also on their specific chemical form, such as the water solubility, exchangeability, carbonate-related forms, oxide association, organic associations, and residual forms (Liu et al., 2009). Water soluble and exchangeable components are considered bioavailable, while the combined oxide, carbonate, and organic components may be bioavailable (He et al., 2005). The determination of heavy metal contaminants in sediments is the basis for solving many environmental problems, including the control and management of water environment. 4 1.4 Research objectives The purpose of this thesis research is to propose a new wetland ecological risk assessment method so that it can provide a more comprehensive scientific basis for decision making. It will focus on both eutrophication and heavy metal pollution issues faced by wetlands. The specific objectives include: • Identification of a suitable water eutrophication risk assessment model for wetland ecological risk assessment • Investigation of a more comprehensive heavy metal risk assessment model for wetland ecological risk assessment • Development of methods to combine the results of eutrophication and heavy metal risk assessments to obtain a general risk level for wetland ecological risk assessment • Application of the wetland ecological risk assessment method to a case study 1.5 Organization of the thesis The organization of the thesis is as follows: Chapter 1 introduces the research, including research background, research objectives and thesis structure; Chapter 2 reviews available methods and applications in the related areas; Chapter 3 describes the eutrophication and heavy metal risk assessment methods as well as the proposed wetland ecological risk assessment method; Chapter 4 presents a case study to demonstrate the applicability of the proposed method; and Chapter 5 summarizes the research results, limitations, and future research directions. 5 CHAPTER 2: LITERATURE REVIEW 2.1 Water Quality Evaluation Various water quality evaluation methods have been developed by scientists during the past years. In the past, single-factor evaluation methods were widely used for investigating water quality in rivers, reservoirs, and lakes (Xin et al. 2006). One typical model is the Vollenweider model to study lake eutrophication with phosphorus as a limiting nutrient (Cloren, 2000). However, the evaluation of water quality is a complex process and the single-factor method cannot reflect the overall situation of the waterbody. The National Sanitation Foundation (NSF) (USA) developed a comprehensive index method such as the WQI (Water Quality Index) and has then developed a variety of water quality indicators (Xin et al. 2006). Some researchers developed and applied water quality evolutionary trend analysis methods, including the rank correlation method (El-Shaarawi et al., 1983), time series analysis method (Long et al., 2009), and parametric test method (Kundzewicz and Robson, 2004; Yenilmez et al., 2011; Ocampo-Duque et al., 2006). For example, Chang et al. (2001) applied the fuzzy mathematics theory to water quality evaluation by developing a fuzzy comprehensive evaluation method. The goal of such methods is to reduce the uncertainty and inaccuracy of the evaluation standard used in decision making (OcampoDuque et al., 2006). The water quality of a waterbody can be assessed using various physical, chemical, and biological parameters. These parameters can be integrated into a water quality index (WQI) to provide a very simple and efficient method for environmental monitoring and decision-making. The National Sanitation Foundation (USA) established a general water 6 quality evaluation framework in terms of selecting parameters, formulating universal scale, and distributing weights, to calculate the water qualities of various waterbodies (Tyagi et al., 2013). The National Sanitation Foundation Water Quality Index (NSFWQI) is based on nine water quality parameters (temperature, pH, turbidity, fecal coliforms, dissolved oxygen, BOD, total phosphate, nitrate, and total solids) (Tyagi et al., 2013). As a representative of traditional water quality assessment methods, WQI is often used for basic water quality evaluation and the comparison of new methods. Some recent applications of using improved models to calculate water quality index can be found in Ranisavljevic & Zerajic (2018). Dharmendra et al. (2018) also applied the improved Water Quality Index (WQI) to assess the time variation of water quality in the port of Paradip, India, while many other parameters were included in addition to the physical and chemical properties of water, such as petroleum hydrocarbons, heavy metals, total viable bacterial (TVB) count, total coliform bacteria (TCB) count, total E. coli bacteria (TEB) count, total phytoplankton count (TPC) and chlorophyll-a concentration. 2.2 Eutrophication Evaluation 2.2.1 Conventional evaluation method The water eutrophication evaluation is a quantitative description of the nutritional status of a waterbody. It investigates and analyzes the indicators relative to the eutrophication status and process and then forecasts the development trend to provide the basis for eutrophication control and management (Cheng, 2012). A variety of evaluation methods have been applied, such as the water quality index method, biological index method, nutrition status index method, fuzzy mathematics evaluation method, and artificial neural networks (ANN). 7 Many scientists have studied a variety of issues related to eutrophication (e.g., eutrophic status, hazards, models, control, and governance) and have achieved promising results (Wang et al., 2008). Tang & Li (2016) applied the water quality index method to assess whether the waterbody is in eutrophication based on the concentration of nitrogen and phosphorus in water (i.e., total nitrogen, total phosphorus) and chlorophyll as well as other parameters associated with water morphology and hydrology. The biological parameter method has been used to evaluate the degree of eutrophication of algae through a number of measurements such as existing stock and diversity. It can be divided into algae species pollution indicator method, comprehensive index method, diversity index method, and biological indicator method. This method is suitable for situations with comprehensive monitoring data (Ignatiades et al., 1992). Sun & Chen (1994) applied the fuzzy mathematics to develop a fuzzy evaluation method for the comprehensive assessment of water quality. Gong et al. (2005) made further optimization of the above model and applied fuzzy mathematics and analytic hierarchy theory for water eutrophication evaluation. They also applied geographic information system and geo-statistics techniques to obtain water eutrophication map by taking Taihu Lake as an example. The goal of such method is to reduce the uncertainty and inaccuracy of the standard used in the decision-making process (Ocampo-Duque et al., 2006). However, the conventional evaluation models are associated with complex structure and require a lot of parameters. Other alternative evaluation methods are thus needed. With the rapid development and maturity of artificial neural network technology, researchers have successfully applied this technique to water quality evaluation. 8 2.2.2 Artificial neural network method Artificial neural network (ANN) is a computational form inspired by the brain and the functioning of the nervous system. Traditionally, ANNs have been used to perform tasks that cognitive brains do naturally, such as face recognition, learning to speak and understand natural language, handwritten text recognition, and defining a goal to analyze different aspects of the same subject. The use of ANNs has been rapidly increasing in the past years, and it has been widely applied to a variety of areas related to forecasting, including water resources (Maier & Dandy, 1996). Artificial neural network is a “black box” which can map the nonlinear relationship among the variables in the ecosystem (Kuo et al., 2006). Several typical models of ANN have been studied, such as Back-propagation (BP) neural network, Hopfield network model, self-organizing neural network, associative memory neural network, and the genetic algorithms. Among them, BP network is one of the most effective and widely used methods, such as for using in the prediction of water quality. Back-propagation is a commonly used learning algorithm in neural network applications. It applies a gradient descent algorithm to determine the weights in the network. Figure 2.1 presents the illustration of a three-layered feed-forward neural network, including an input layer, a hidden layer, and an output layer. The input layer contains the input nodes (neurons), and the output layer contains the expected system output. The hidden layer usually contains a series of nodes associated with the transfer function (Huo et al., 2013). Each layer of the network is linked by weights that can be determined by the learning algorithm. Sigmoid function is one of commonly used transfer functions (Huo et al., 2013). The development of ANN model requires a series of input and output data to establish their relationship, and part of the data is used for training, while part of the data is used for testing. 9 Figure 2.1: Schematic illustration of a three-layered feed-forward neural network (Lek & Gue´gan, 1999) With the technology advancement and maturity, artificial neural network has been applied to a variety of water environmental research studies. Karul et al. (2000) utilized a three-layer Levenberg-Marquardt feedforward learning algorithm to model the eutrophication process in three waterbodies in Turkey (Keban Dam Reservoir, Mogan and Eymir Lakes), and the model achieved relatively good correlation although the hydrological conditions at the selected sites were complicated. Joseph (2003) used artificial neural network as a data simulation method to examine the kinetics of algae bloom in Hong Kong coast water. Kuo et al. (2006) combined BP neural network with key factors affecting water quality in a reservoir in central Taiwan, such as dissolved oxygen (DO), total phosphorus (TP), chlorophyll-a, and transparency (SD), to capture reservoir variables, and their results showed that the model can 10 serve as an important tool in reservoir management. Lin et al. (2009) applied the BP artificial neural network method for the eutrophication risk assessment of reservoirs in Shenzhen, China. Huo et al. (2013) used the BP neural network to develop the relationship between the water quality factors and the eutrophication indicators, and the model was successful for prediction, suggesting that the neural network is a valuable tool for lake water quality management. Cui et al. (2014) established a BP artificial neural network prediction model of TN, TP and Chl-a changes in Taihu Lake by using the Minitab software, and the results indicated that the established ANN model can effectively predict the change of eutrophication in Taihu Lake, China. Zhang et al. (2018) established a BP neural network model to classify mangrove water quality and establish the relationship between water quality and pests’ diseases. Salari et al. (2018) also used the mathematical method and artificial neural network to obtain a water quality parameter optimization artificial neural network model. It was found that the ANN model can play a positive role in the management of lakes and reservoirs. It can also be applied to the prediction and management of wetland water quality, and it is thus feasible to use artificial neural network method in eutrophication evaluation. 2.3 Heavy Metal Risk Evaluation Methods for assessing the ecological risk of heavy metals in sediments include the calculation of geo-accumulation index (Porstner, 1989), potential ecological risk index (Ha kanson, 1980), and excess regression analysis (ERA) (Hilton et al., 1985), where the first two index are the most commonly used (Yi et al., 2011). Most studies involving heavy metal polluted sediments only used the total metal contents as the standard to assess their potential 11 contamination effects. However, the total metal concentration provides insufficient information to assess the bioavailability or toxicity of metals (Sundaray et al., 2011). RI (potential ecological risk index) is one method to evaluate the potential risk of heavy metals. However, its disadvantage is that it does not consider the chemical forms of metals. Heavy metals have significant differences in different chemical forms (Li et al., 2007). The Risk Assessment Code (RAC) is another method for risk assessment of heavy metals. The risk level was classified according to the chemical form of heavy metals (Singh et al., 2005). Liu et al. (2009) proposed a modified potential ecological risk index (MRI) model by multiplying RI by the toxicity index of different chemical forms of heavy metals associated with RAC. Rahman et al. (2013) adopted the modified potential ecological risk index to evaluate heavy metal contamination in sediment and waterbody around Dhaka export processing zone, Bangladesh. Liu et al. (2018) used the modified potential ecological risk index (MRI) model to assess metal contamination in estuarine surface sediments from Dongying City, with the study metals including Cd, Hg, As, Cr, Cu, Pb and Zn. Liu & Ni (2018) analyzed the contamination level, chemical fraction and ecological risk of heavy metals in sediments from Daya Bay, South China Sea according to the modified potential ecological risk index. 2.4 Fuzzy Risk Assessment In order to combine the information of both eutrophication and heavy metal pollution for better evaluating the waterbody of wetland, some artificial intelligence methods can be used. The fuzzy logic and fuzzy sets methods have been tested for addressing practical environmental issues. Their goal is to reduce the information uncertainty and inaccuracies 12 related to decision making (Ocampo-Duque et al., 2006; McKone & Deshpande, 2005). In the past decades, several fuzzy-set based approaches have been proposed to estimate environmental quality and achieved promising results. Smith (1995) developed a fuzzy aggregation approach for environmental quality evaluation. Veiga et al. (1995) proposed a risk assessment model of mercury pollution in the Amazon basin based on fuzzy set theory. Li et al. (2003) and Liu et al. (2004) proposed a hybrid method combining probability and possibility to represent the uncertainty of modeling parameters involved in risk assessment. Ocampo-Duque et al. (2006) developed a water quality index that was calculated with fuzzy reasoning, and the application of the fuzzy index was tested with a case study. Li et al. (2007) developed an integrated fuzzy stochastic modeling approach (IFSM) to assess the impact of air pollution on asthma susceptibility. Caniani et al. (2011) used a fuzzy model as a fast and economic method for the assessment of groundwater pollution risk of aquifers underlying some uncontrolled landfills in the Basilicata Region territory. It was found from the literature that most previous risk assessment studies were based on a single fuzzy set approach. Few previous studies incorporated a variety of information in a risk assessment framework (Li et al., 2007). 2.5 Scenario Analysis Scenarios and scenario analysis have become popular methods of organizational planning and participatory activities for sustainable development (Duinker & Greig, 2006). They can be developed in contexts related to stakeholders involved in their applications as the assessment of scenario outcomes and impacts can enhance decision making activities 13 (Mahmoud et al., 2009). A scenario can enable policies and decisions to be tested against possible futures and inspire new ideas (Lang, 2001). A variety of methods can be used to develop scenarios. In the process of creating a scenario, it is necessary to use various collected information. For example, it should consider the problems faced by the environment, the main pollutants, and some mainstream governance methods (Cornish, 2004). Scenario-based work is the most powerful when creating and analyzing several alternative scenarios, and each scenario should provide significant contrast to other scenarios. While each scenario describes an alternative future in terms of qualitative and/or quantitative, each scenario must be credible (Duinker & Greig, 2006). Although the method of scenario analysis originated in other fields, it has been successfully applied to environmental studies. Baker et al. (2004) used the inputs from local stakeholders to create three alternative future landscape scenarios for the year 2050 when studying the Willamette River Basin in western Oregon, and the results indicated that scenario analysis can be applied to long-term prediction. Jia et al. (2011) developed a hydrodynamic model of river water pollution control by taking the Nansha River in Beijing, China, as a case study, and three water pollution control scenarios were proposed by using the developed model to identify the management objectives of the point source pollution control plan. Zhu & Song (2013) established a one-dimensional river network hydrodynamic and water quality numerical model and applied it to the analysis of water environment management in the inland river system of Fuzhou, while three different interception scenarios were proposed to calculate the water quality results and identify the optimal scenario. 14 2.6 Wetland Risk Assessment Wetland risk assessment is based on ecological risk assessment framework and targets the change of wetland ecological characteristics. The US Environmental Protection Agency (EPA) proposed a quantitative and qualitative wetland ecological risk assessment on wetland ecosystem pressures (Dam et al., 1999). Pasco (1993) explored the concept of wetland risk assessment by using two case studies to demonstrate. The US EPA (1998) then developed a similar watershed eco-risk assessment framework, and proposed the guidelines for ecological risk assessment, including detailed information on physical and biological stress and chemical stress prediction and assessment (Dam et al., 1999). Figure 2.2 presents a basic modeling framework of wetland risk assessment (van Leeuwen, 1995; Dam et al., 1999). This framework is an integral part of the wetland management planning process (Jose, 1999). The purpose of the framework is to outline how wetland risk assessment can be used as a "vehicle" in the process of predicting and assessing changes in ecological characteristics, with emphasis on the application of early warning techniques (Jose, 1999). The framework has been applied to the relevant wetland projects in many countries for effective wetland management decision-making (Zhang, 2004). 15 Identification of the problem Identification of the effects Identification of the extent (field assessment) (e.g., chemical concentrations) Identification of the risk Risk management Monitoring Figure 2.2 The framework of wetland risk assessment (Jose, 1999) 2.7 Summary A variety of methods have been used for eutrophication evaluation, with each having its own advantages and disadvantages. Among them, artificial neural network (ANN) can be successfully applied to evaluate the risk of eutrophication with the advantage of nonlinear mapping and more objective parameter selection. In terms of the risk assessment of heavy metals in surface sediments, there are also many evaluation methods available, but most of them only consider the total amount of heavy metals and neglect the metal speciation forms that determine the toxicity and bioavailability of metals. An improved potential ecological risk assessment method is thus desired to make a more reasonable evaluation. In terms of the assessment of wetland ecological risk, the traditional risk assessment framework can only be 16 used to assess the changes in wetland ecological characteristics under a specific pollution source (e.g., eutrophication or heavy metal pollution) instead of multiple types of pollution sources. This thesis research will then use fuzzy set theory to combine the risk resulted from different types of pollution (e.g., eutrophication and heavy metal pollution in wetlands) in order to obtain a more comprehensive ecological risk assessment result and reduce the associated uncertainty and inaccuracy, while scenario analysis will be conducted to recommend appropriate wetland management solution (Jia et al., 2011). 17 CHAPTER 3: METHODOLOGY In this thesis research, the proposed wetland risk assessment framework was established by combining the eutrophication evaluation with heavy metal evaluation, and fuzzy set approach was then applied to obtain the general risk level of wetland. Figure 3.1 presents a flow chart of the framework, and each step is briefly described below. 3.1 Step One: Problem Statement Problem statement is a process of identifying research area background, including the identification of the pressure and the receptor. It defines the goals and scope of the study and provides the basis for the overall risk assessment (Jose, 1999). It also describes the source of data for subsequent evaluation. In this thesis, the eutrophication and heavy metal pollution of wetlands were considered as the sources of stress. The sources and properties of eutrophication and heavy metal pollution as well as the possible adverse effects of water quality changes were analyzed (Jose, 1999). 3.2 Step Two: Water Quality Assessment This step is a data analysis process. Deteriorating water quality would lead to changes in the physical and chemical properties of wetlands (Dam et al., 1999). In this thesis study, field survey and monitoring data will be used to analyze the water quality. Using WQI could get a general idea of the overall situation of the local waterbody. 18 Problem statement (e.g., background of study area, defining objectives, data sources) Input: water quality indicator concentrations (e.g. TN, TP, turbidity) Input: the concentrations of heavy metals and their fractions Water quality index Output: water quality level Risk assessment model Modified potential ecological risk assessment of heavy metal Eutrophication risk assessment (artificial neural network) Output: the degree of heavy metal potential risk (HR) Output: the degree of eutrophication risk (ER) Overall risk assessment (use fuzzy rules to combine HR and ER) Scenario analysis Figure 3.1 The framework of wetland risk assessment 19 Input: water quality indicator concentration (e.g., TN, TP, COD, Chl-a) The National Sanitation Foundation (USA) established a common water quality indicator method to calculate the water quality of various waterbodies in terms of selecting parameters, formulating universal scale, and distributing weights (Tyagi et al., 2013). The National Sanitation Foundation Water Quality Index (NSFWQI) is based on nine water quality parameters (temperature, pH, turbidity, fecal coliforms, dissolved oxygen, BOD, total phosphate, nitrate, and total solids) (Tyagi et al., 2013). If there are n types of water quality parameters, each would be assigned a weight through a decision process. The water quality index can then be calculated by the following equations (Chang et al., 2001): Qi = Ci/Cn (3.1) ∑𝑛 𝑄 𝑊 (3.2) 𝑖 𝑖 𝑊𝑄𝐼 = ∑𝑖 𝑛 𝑊 × 100 𝑖 𝑖 Where Qi = sub-index for ith water quality parameter; Ci = measured value of the ith water quality parameter; Cn = Chinese Surface Water Quality Standard for the ith water quality parameter; Wi = weight associated with the ith water quality parameter; n = number of water quality parameters. The water parameters and their weights in calculating WQI are listed in Table 3.1. Table 3.2 presents the water quality categories based on the WQI results. Table 3.1 Water quality parameters and their weights (Tania et al., 2013) Parameter Revised Weight Turbidity 0.18 DO 0.38 Total phosphorus 0.22 Nitrates 0.22 20 Table 3.2 Water quality category for NSF water quality index methods (Li et al., 2010). WQI Value Water Quality Category 0-50 Excellent water quality 50-100 Good water quality 100-150 Medium water quality 150-200 Bad water quality >200 Very bad water quality 3.3 Step Three: Risk Assessment Model This step is to assess the extent of water eutrophication and heavy metal pollution in wetlands. A BP Artificial Neural Network (ANN) eutrophication risk evaluation method was used to evaluate the eutrophication level in the wetland. In terms of heavy metal risk level, the modified potential ecological risk index was used. These two risk levels would then serve as the basis for the subsequent evaluation of the overall risk of wetland. 3.3.1 Eutrophication risk assessment (1) Artificial neural network The key to build a model using neural networks is the adequate training samples. In this study, the evaluation standard of eutrophication status as stipulated in the "Control Standard for Surface Water Eutrophication" of the Ministry of Water Resources of China was used to obtain enough training samples through interpolation (Lin et al., 2009). Table 3.3 presents the standard values of the five water quality parameters corresponding to 10 trophic 21 status score values (Deng et al., 2006). The trophic state was used as an indicator, and 100 data samples were uniformly interpolated within each classification interval of trophic state score (from 0 to 100), thus leading to a total of 1000 samples to be normalized (Lin et al., 2009). The normalization was conducted using x−Min Max−Min , where “x” is the value of parameter, “Max” is the upper bound of this parameter in Table 3.3 and “Min” is the lower bound of this parameter. For example, a parameter “TP” with a value of 0.001 means that “x” is 0.001, while its “Max” and “Min” value shown in Table 3.3 is 1.3 and 0, respectively. The corresponding value of normalization is then 0.001−0 1.3−0 . Among the 1000 data samples, 800 samples were taken as training samples, 100 as test samples and 100 as validation samples. 22 Table 3.3 Control standard for surface water eutrophication (Deng et al., 2006). Levels of trophic state Score of Chl-a TP TN CODMN SD trophic state (mg/m3) (mg/L) (mg/L) (mg/L) (m) 0 0 0 0 0 37 10 0.5 0.001 0.02 0.15 10 20 1 0.004 0.05 0.4 5 30 2 0.01 0.1 1 3 40 4 0.025 0.3 2 1 50 10 0.05 0.5 4 1 Light eutrophic 60 26 0.1 1 8 0.5 Moderate eutrophic 70 64 0.2 2 10 0.4 Heavy eutrophic 80 160 0.6 6 25 0.3 90 400 0.9 9 40 0.2 100 1000 1.3 16 60 0.12 Oligotrophic Mesotrophic A 3-layer BP neural network was used in this study, including an input layer, a hidden layer, and an output layer. As can be seen from Table 3.3, there are five input parameters (Chl-a, TP, TN, CODMn, and SD) (“SD” is transparency), and one output variable which is the trophic state with a value between 0 and 100. Therefore, the input layer had five nodes, and the output layer had one node. The number of nodes of the hidden layer would change with the training process (Ren et al., 2004). The neural network was trained using the 800 training samples, and tested using 100 test samples, while the initial learning rate was set to 0.01 in the parameter selection. The remaining 100 sets of data were then used to validate the 23 model so that the error between the expected output value and predicted value was low. Once the results of the ANN prediction are relatively satisfactory, the model can be applied to evaluate the eutrophication of wetland. In order to use the developed ANN model, the values of the five water quality parameters obtained in the wetland area were put into the network, thus obtaining an output score that can be used for assessing the degree of eutrophication. (2) Comprehensive nutritional status index method In order to evaluate the accuracy of the ANN results, another method using comprehensive nutrition index was used. Based on the classification and technical requirements for “Lakes and Reservoir Eutrophication Assessment Methods” of the China Environmental Monitoring Center, a comprehensive nutritional status index can be used to evaluate lake eutrophication. It was reported that this index method provides a more comprehensive evaluation range than other nutritional index methods. It comprehensively considers a variety of water quality parameters such as TN, TP, SD, Chl-a and CODMn (Li et al., 2017). It was calculated using the following equations (Zhang et al., 2006). (3.3) 𝑇𝐿𝐼(∑) = ∑𝑚 𝑗=1 𝑊𝑗 × 𝑇𝐿𝐼(𝑗) Where TLI (∑) is the comprehensive nutritional status index; Wj is a normalized weighted value of parameter j; and TLI (j) is the universal index of parameter j (Li et al., 2017). Using chl-a as the reference parameter, the weight calculation is as follows: 24 𝑟 2 (3.4) 𝑊𝑗 = ∑𝑚 𝑖𝑗𝑟 2 𝑗=1 𝑖𝑗 Where rij is the correlation coefficient between parameter j and the reference parameter Chl-a; m is the number of evaluation parameters (Li et al., 2017). Table 3.4 shows the values of Wj of the five water quality parameters. Table 3.4 Correlation between Chl-a and other parameters as well as the weight value (Zhang et al., 2006). Parameter Chl-a TP TN CODMn SD rij 1 0.84 0.82 0.83 -0.83 rij2 1 0.7056 0.6724 0.6889 0.6889 wj 0.266 0.188 0.179 0.183 0.183 The trophic level index (TLI) is calculated by the following equations (Liu et al., 2011): TLI (TN) = 10(5.453+1.694 ln TN) (3.5) TLI (TP) = 10(9.436 + 1.624 ln TP) (3.6) TLI (COD MN) = 10(0.109 + 2.661 ln COD MN) (3.7) TLI (Chl-a) = 10(2.5 + 1.086 ln Chl-a) (3.8) TLI (SD) = 10(5.118-1.94ln SD) (3.9) Consequently, the comprehensive nutritional status index can be calculated as: TLI() = 0.179TLI(TN) + 0.188TLI(TP) +0.183TLI(CODMn) +0.266 TLI(Chl-a) +0.183TLI(SD) (3.10) 25 After obtaining this comprehensive eutrophication index, the reservoir eutrophication is divided into five different trophic levels as listed in Table 3.5. Under the same trophic level, the higher the index value, the more serious the eutrophication (Li et al., 2017). Table 3.5 Trophic levels using comprehensive nutritional status index (Li et al., 2017). Trophic levels Comprehensive nutritional status index values Oligotrophic TLI (∑) < 30 Mesotrophic 30 ≤ TLI (∑) ≤50 Light eutrophic 50 < TLI (∑) ≤ 60 Moderate eutrophic 60 < TLI (∑) ≤70 Hyper eutrophic TLI (∑) > 70 3.3.2 Modified potential ecological risk index (1) Risk assessment code (RAC) To assess heavy metal risk in sediment, risk assessment code (RAC) represents one of the guidelines based on the percentage of heavy metal in the carbonate and exchangeable fractions according to the three- step BCR method (Liu et al., 2009). The RAC evaluation is based on the facts that the concentrations and exposure time of different metals would affect their toxicity, and thus the risk of heavy metals in sediments can be assessed based on the information of the percentage of exchangeable and carbonate fractions (Zhu et al., 2011). According to the RAC guideline, when the total ratio of the exchangeable and carbonate fractions (F1) in metal is less than 1%, there is no risk; when the total ratio is in the range of 26 1-10%, there is a low risk; a medium risk corresponds to the ratio range of 11-30%, and a high risk corresponds to a ratio range of 31-50%; when this ratio is more than 50%, the risk is very high, which means the heavy metals can easily enter the food chain (Liu et al., 2009; Zhu et al., 2011). Table 3.6 shows the RAC classification of risk. Table 3.6 Classification of RAC and values of δ (toxic index) (Zhu et al., 2011). Risk level Carbonate and exchangeable fraction (F1) δ in metal (%) No risk <1 1.00 Low risk 1-10 1.00 Medium risk 11-30 1.20 High risk 31-50 1.20 Very high risk >50 1.40 (2) Potential ecological risk index To assess ecological risks for aquatic system, Hakanson (1980) developed a methodology to calculate a risk index (RI) based on the synergy among various heavy metal elements, toxicity level, pollution concentration, and the sensitivity of the ecological environment to heavy metals (Yi et al., 2011). This method has been widely used by scientists in many countries. The value of RI can be calculated by the following formulas (Zhu et al., 2011): Cif=CiD/CiR (3.11) Eir =Tir×Cif (3.12) 27 Where Eir is the potential ecological risk factor; Tir is the toxic-response factor for a given heavy metal i; Cif is the contamination coefficient; CiD is the present concentration of heavy metal i in sediments; CiR is reference concentration of heavy metal I in sediments. Based on Hakason (1980), the toxic-response factor Tir is described as Cd (30)> Cu=Pb (5) > Cr (2) > Zn (1), respectively. The individual potential ecological risk factors is then used to obtain the heavy metal risk index for the study area (Liu et al., 2009): n RI=  Eir (3.13) i Table 3.7 presents the evaluation criteria of this method. Table 3.7 Indices and levels of potential ecological risk (Zhu et al., 2011). Eir Level of ecological risk of RI single metal Eir<40 Level of potential ecological risk Low risk RI<150 Low risk 40 ≤ Eir<80 Moderate risk 150≤ RI<300 Moderate risk 80≤ Eir<160 Considerable risk 300≤ RI>600 Considerable risk 160≤ Eir<320 High risk RI≥600 Very high risk Eir ≥320 Very high risk (3) Modified potential ecological risk index Many previous studies on risk assessment for heavy metals in sediment were relying on the total concentrations of heavy metals. However, the risk not only depends on the total metal content, but also their chemical speciation (Zhu et al., 2011). The metal speciation 28 greatly affects the bioavailability which could further affect the risk of the overall aquatic ecosystem. Zhu et al. (2011) developed a modified potential ecological risk index using different toxicity indices for different proportions of exchangeable and carbonate fractions (F1): Ω= Aδ +B (3.14) ~ CiD= CiD Ω ~ (3.15) ~ Cif = CiD /CiR (3.16) ~ Eir =Tir×Cif n (3.17) MRI=  Eir ~ (3.18) i ~ ~ ~ Where CiD , Cif , Eir and MRI are the modified Cif , CiD, Eir and RI; Ω is the modified index of heavy metal concentration; A is the percentage of exchangeable and carbonate fractions in metal (F1); B is equal to 1-A; δ is the toxic index as shown in Table 3.6. The evaluation criteria of this method can also be referenced in Table 3.7. 3.4 Step Four: Overall Risk In this thesis, an overall risk of wetland needs to be determined based on the eutrophication risk level and heavy metal risk level. A fuzzy set approach was then used to combine the information from both risk levels since the general risk quantification can only be based on subjective opinions rather than on probabilistic analysis (Li et al., 2007). The general risk was acquired through a questionnaire survey of experts based on the fuzzy rules 29 which could effectively capture the subjective opinions. Rules usually include conditional parts (such as antecedents) and conclusion parts (such as consequences). An antecedent can be a simple clause, or a combination of several clauses joined by fuzzy logic operators. For example, “if eutrophication risk is “High” and heavy metal potential risk is “High”, then the general risk is “Very High”. This is a form of fuzzy rules, where the “eutrophication risk” and” heavy metal potential risk” are inputs and the “general risk” is an output (Mohamed and Cote, 1999). In this thesis, two fuzzy operators (“AND” and “OR”) were used to illustrate the association. If A and B were used to represent two fuzzy events, and µA and µB represent their membership degree respectively, then AND operation would generate a membership equal to Min [µA (x), µB(x)], and OR operation would produce a membership of Max [µA(x), µB(x)] (Wang et al., 2008). The framework of obtaining the general risk level is shown in Figure 3.2. 30 Eutrophication risk (ER) Very high risk Considerable risk Moderate risk Low risk High eutrophic Moderate eutrophic Light eutrophic Mesotrophic Oligotrophic Fuzzy rule base Heavy metal potential risk (HR) General risk levels (GRL) Very high risk High risk Medium-High risk Medium risk Low-medium risk Low risk Figure 3.2 The framework of obtaining general risk In this study, based on the questionnaire survey of experts, the general risk levels were classified into six categories of fuzzy set, including “low’’, ‘‘low-to-medium’’, ‘‘medium’’, ‘‘medium-to-high”, “high”, and ‘‘very-high’’ (Wang et al., 2008). The fuzzy logical operator "AND" was used. Because ER includes five types of eutrophication risk level events, and HR includes four potential ecological risk level events, there are a total of 20 rules (Li et al., 2007). If a rule gets the highest response frequency in the survey, it will be saved in the rule base to determine the general risk level. Table 3.8 presents the 20 fuzzy rules. 31 Table 3.8 Survey results on fuzzy rules Antecedent Consequence If eutrophication risk If heavy metal risk Then the general risk Survey response (ER) is (HR) is level (GRL) is frequency (%) Oligotrophic Low Low 100 Oligotrophic Moderate Low-Medium 73 Oligotrophic Considerable Medium 45 Oligotrophic Very High Medium-High 55 Mesotrophic Low Low 55 Mesotrophic Moderate Medium 55 Mesotrophic Considerable Medium-High 55 Mesotrophic Very High High 73 Light eutrophic Low Low 64 Light eutrophic Moderate Medium 64 Light eutrophic Considerable Medium-High 64 Light eutrophic Very High High 55 Moderate eutrophic Low Low-Medium 36 Moderate eutrophic Moderate Medium 55 Moderate eutrophic Considerable High 55 Moderate eutrophic Very High Very High 64 Heavy eutrophic Low Medium 55 Heavy eutrophic Moderate Medium-High 55 Heavy eutrophic Considerable Very High 55 Heavy eutrophic Very High Very High 73 32 Figure 3.3 shows the membership functions of the fuzzy general risk levels of “low’’, ‘‘low-to-medium’’, ‘‘medium’’, ‘‘medium-to-high”, “high” and ‘‘very-high’’ (Li et al. (2007). The purpose of using the fuzzy membership functions of these risk level events is to calculate a risk score for risk management. This score is obtained by applying an “AND” or “OR” fuzzy operator (Wang et al., 2008). 1 L L-M M M-H H V-H 0.8 µGR 0.6 0.4 0.2 0 0 10 20 30 40 50 60 70 80 90 100 GRL Figure 3.3 Membership functions of fuzzy general risk levels In order to obtain the membership functions of the general risk level events, the membership functions of both eutrophication risk level (ER) and heavy metal potential ecological risk level (HR) events need to be established. Since ER includes five fuzzy events, and HR includes four fuzzy events, their membership functions can be presented in Figure 3.4 and Figure 3.5 (Li et al., 2007). The membership function of the general risk level is then obtained through fuzzy “AND” and “OR” operations. 33 M L L-E M-E H-E 1 0.8 µER 0.6 0.4 0.2 0 0 10 20 30 40 50 60 70 80 90 100 ER Figure 3.4 Membership functions of fuzzy eutrophication risk level events L M H V-H 1 0.8 µHR 0.6 0.4 0.2 0 0 75 150 225 300 375 450 525 600 675 750 HR Figure 3.5 Membership functions of fuzzy heavy metal risk level events 34 825 900 3.5 Step Five: Risk Management and Scenario Analysis Risk management is the ultimate decision-making process, using the information obtained from risk assessment and attempting to minimize the risk. The general risk level score of the waterbody can provide the basis for wetland risk management decision-making. Table 3.9 lists the relationship between the general risk level score and the proposed management action (Mohamed and Cote, 1999). In the context of the International Convention on Wetlands, risk management must also consider the concept of the potential role of sensible use and management of decision-making (Jose et al., 1999). After comprehensive analysis of major wetland environmental issues, major pollutants, and appropriate pollution control techniques, two scenarios were designed and analyzed in the thesis, including conservative and positive action (Jia et al., 2011). Table 3.9 Recommended risk management actions Risk score Risk management action 90-100 Take full treatment of the wetland region 70-90 Take target treatment on the most serious pollution 50-70 Warning of wetland pollution and take emergency measures 30-50 Reduce discharge of pollutants around the wetland region 10-30 The wetland region should strengthen monitoring program 0-10 No actions are required 35 CHAPTER 4: CASE STUDY 4.1 Overview of Study Area Wenzhou Sanyang Wetland is within the Wenzhou Ecological Park on the southern bank of the Oujiang River estuary. It is in the southeast corner of Wenzhou City in Zhejiang Province, China, and is close to Daluo Mountain. It is a river network wetland that evolved from the ancient sea-coast and consists of 161 "island" shaped fields with a total area of 11.81 km2. The water area is 3.31 km2, accounting for 28% of the entire wetland area (Wu et al., 2012). As a typical suburban wetland, Sanyang Wetland has a long history of development with a land area of approximately 8.5 km2, including residential buildings and industrial factories (1.1 km2), citrus-based farms (6.6 km2), and other crops and pastures (0.8 km2) (Chen, 2008). According to various environmental monitoring results and analysis, the regional environmental conditions in Sanyang Wetland especially the water environment quality has been severely deteriorated. The water quality parameters such as nitrogen, phosphorus, and heavy metals were far exceeding their environmental quality standard. The Sanyang Wetland is in a subtropical monsoon climate zone with four distinct seasons, which is warm and humid year-round. The annual average temperature is 17.9 °C, with the extreme low and high temperature of -4.5 °C and 39.03 °C, respectively. The annual temperature difference is about 20 °C. Appropriate temperature is conducive to the growth and reproduction of algae which would further lead to deteriorating water quality. The main sources of pollution in the wetland area include the heavy use of pesticides caused by orange plantation, and the significant increase in the number of manufacturing facilities (for example, small leather 36 tanneries and textile mills owned by households), animal farms (e.g. pigs and ducks), and residential buildings on wetlands (Li et al., 2017). In the past years, the majority of research work in the Sanyang Wetland and its nearby areas focused on water eutrophication (Liu & Jia, 2007). However, Gao et al. (2014) found that the heavy metal pollution in wetland surface sediments cannot be neglected. Heavy metal pollution in surface sediments can also affect the quality of overlying water and aggravate the risk of pollution in wetland waters. 4.2 Eutrophication Risk Assessment Results 4.2.1 Environmental data collection Chen et al. (2016) conducted an environmental quality study in the Wen-Rui Tang River which is located in Sanyang Wetland area. The data collected from two sampling locations in their study were used for this thesis research (Figure 4.1). Figure 4.2 shows the land use type around these two sampling locations. Water samples were collected monthly from May 2008 to December 2012. The monitored parameters include turbidity, transparency (SD), chemical oxygen demand (CODMn), total nitrogen (TN), Nitrates, total phosphorus (TP), dissolved oxygen (DO), and chlorophyll a (Chl-a). The measurement method followed the Surface Water Monitoring Standard of the State Environmental Protection Administration (China) (Chen et al., 2016). 37 Sanyang street S4 Wu Ti Street A6 Ouhai Avenue Wen Ruitang River Sanyang Avenue Figure 4.1 Sampling locations in Sanyang Wetland area (Chen et al., 2016) 38 100% 90% 80% 70% 60% urban 50% Industrial and mining 40% 30% Water 20% Vegetated 10% Agriculture 0% A6 S4 Figure 4.2 Land use distribution around two sampling locations (A6 and S4) (Chen et al., 2016) 4.2.2 Water quality status results Table 4.1 lists the average values of water quality parameters obtained at two sampling locations during different seasons. It can be found that the four selected water quality parameters were generally exceeding the Chinese environmental quality standard values, especially the turbidity and dissolved oxygen (DO), which exceeded the standard by 5-10 times. Such worse water quality situation may be caused by the excessive growth of plankton which causes the water to be turbid. The decomposition of organic matter also consumes oxygen in the water, resulting in insufficient dissolved oxygen. 39 Table 4.1 Average values of water quality parameters at the sampling locations (Chen et al., 2016) DO TP Nitrates Turbidity (mg/L) (mg/L) (mg/L) （m） Sampling location Season A6 Winter 1.33 0.33 13.13 30.23 Spring 1.34 0.39 16.65 15.12 Summer 1.66 0.32 14.49 19.65 Autumn 1.19 0.41 10.83 15.71 Winter 3.98 0.23 12.36 21.62 Spring 5.73 0.21 11.30 10.59 Summer 5.03 0.22 12.70 19.20 Autumn 3.86 0.18 10.14 25.93 ≥5 ≤ 0.2 ≤ 10 ≤5 S4 Surface water environmental quality standard of China Table 4.2 lists the results of WQI for the two sampling locations by using Eqs. 3.1 and 3.2. It can be seen that the water quality at location S4 was significantly better than that at location A6. The water qualities at A6 in four seasons were rated as “very bad”, while the water qualities at S4 were rated as “bad” except for the spring when they were rated as “medium” water quality. These results are related to the land use around the two sampling locations, indicating that human life, industrial and mining activities had a greater impact on the surrounding waters than agricultural activities. Moreover, it was observed that the water 40 quality in spring and summer was generally better than that in autumn and winter. The reason may be that the summer and spring are the rainy seasons, and the soil erosion would influence water quality. Overall, the water quality assessment of Sanyang Wetland indicated that the waterbody was extremely polluted. Table 4.2 WQI Assessment results for the two sampling locations Sampling Season location A6 S4 QiWi WQI Water quality rating DO TP Nitrates Turbidity Winter 142.66 36.49 28.89 108.84 316.88 Very Bad Spring 142.06 42.36 36.63 54.44 275.49 Very Bad Summer 114.36 35.46 31.87 70.75 252.44 Very Bad Autumn 159.91 44.86 23.84 56.57 285.18 Very Bad Winter 47.72 25.58 27.20 77.82 178.31 Bad Spring 33.13 22.66 24.85 38.12 118.76 Medium Summer 37.80 24.55 27.94 69.12 159.41 Bad Autumn 49.27 19.93 22.30 93.33 184.84 Bad 4.2.3 Eutrophication risk assessment results (1) The results of artificial neural network (ANN) model Taking the monitoring data of the five parameters from sampling locations A6 and S4 as the ANN modeling inputs, the artificial neural network (ANN) model was implemented to calculate the score of the trophic state, and the results are in Table 4.3. 41 Table 4.3 ANN model calculated score of trophic state for Sanyang Wetland Sampling Season location Normalized concentration of Score of Trophic Level parameter trophic TN TP SD state Winter 11.47 35.23 9.11 0.33 2.11 80.00 Heavy eutrophic Spring 13.36 35.81 9.89 0.39 3.06 79.76 Moderate eutrophic Summer 12.83 25.90 8.41 0.32 3.85 76.20 Moderate eutrophic Autumn 12.91 31.43 8.80 0.41 3.03 78.28 Moderate eutrophic Annual 12.67 31.91 9.04 0.36 3.05 78.48 Moderate eutrophic Winter 20.31 13.89 8.82 0.23 2.66 71.02 Moderate eutrophic Spring 28.57 15.32 8.41 0.21 3.96 70.00 Moderate eutrophic Summer 25.84 14.32 6.75 0.22 2.84 70.24 Moderate eutrophic Autumn 23.57 14.93 7.62 0.18 2.91 70.07 Moderate eutrophic Annual 27.41 14.61 7.87 0.21 3.08 70.81 Moderate eutrophic Chl-a A6 S4 COD The ANN modeling results indicated that the eutrophication degree around both sampling locations (A6 and S4) was rated as “moderate eutrophic”, except for the “heavy eutrophic” state in winter around sampling location A6. It can also be found that the eutrophication around location A6 was more serious than location S4. The reason may be that most of the land around location S4 was used for agricultural purposes. The large amount of chemical fertilizers and pesticides used in modern agricultural production would lead to continuous leaching of nutrients into the surrounding environment, especially into the waterbody. However, the vicinity of sampling location A6 was mostly associated with residential land use and industrial land use. The content of nitrogen and phosphorus in industrial (such as chemical, printing and dyeing) wastewater was quite high. In addition, the 42 domestic wastewater from residential land uses and the animal husbandry in rural areas would lead to excessive nutrients in the water which could cause severe eutrophication of waterbodies. Overall, the waterbody in the Sanyang Wetland was in a “moderate eutrophication” state, which agrees with the WQI results. (2) The results of comprehensive nutritional status index Five parameters (TN, TP, COD, Chl-a, and SD) were used for calculating the comprehensive nutritional index (TLI). Table 4.4 presents the results for the two sampling locations. It was found from this method that the Sanyang Wetland around these two sampling locations were rated as “moderate eutrophic”, which is generally consistent with the ANN modeling results, indicating that the ANN model is feasible for the evaluation of water eutrophication. 43 Table 4.4 Assessment results of the comprehensive nutritional status index in Sanyang Wetland Normalized concentrations of parameters Sampling Season Chl-a COD TN TP SD TLI Level (mg/m3) (mg/L) (mg/L) (mg/L) (m) Winter 11.47 35.23 9.11 0.33 2.11 68.78 Moderate eutrophic Spring 13.36 35.81 9.89 0.39 3.06 68.69 Moderate eutrophic Summer 12.83 25.90 8.41 0.32 3.85 65.15 Moderate eutrophic Autumn 12.91 31.43 8.80 0.41 3.03 67.81 Moderate eutrophic Annual 12.67 31.91 9.04 0.36 3.05 67.53 Moderate eutrophic Winter 20.31 13.89 8.82 0.23 2.66 63.90 Moderate eutrophic Spring 28.57 15.32 8.41 0.21 3.96 63.44 Moderate eutrophic Summer 25.84 14.32 6.75 0.22 2.84 63.58 Moderate eutrophic Autumn 23.57 14.93 7.62 0.18 2.91 63.16 Moderate eutrophic Annual 27.41 14.61 7.87 0.21 3.08 63.84 Moderate eutrophic location A6 S4 4.3 Heavy Metal Risk Assessment Results 4.3.1 Sampling and analysis The data obtained from Li et al. (2017) were used for metal risk assessment in this thesis. The study area (3.2 km2) was located on the northwest part of Sanyang Wetland (Figure 4.3). 44 Figure 4.3 Sampling locations and distribution of different land uses (OP, OPRI and RAI) in Sanyang wetland (Li et al., 2017) In the selected study area in Sanyang Wetland, the sampling locations represent the influence of three dominant land uses, including orange plantation (OP), mixed orange plantation, residential and industrial land (OPRI), as well as the mixed residential and industrial land (RAI) (Figure 4.3) (Li et al., 2017). The surface sediment samples were 45 collected in December 2012, September 2013, and May 2014, respectively. The samples collected in 2012 were used to estimate total carbon, nitrogen, phosphorus, pH, and bulk density. The samples collected in 2013 were used to analyze the metal speciation (chemical form), and the samples collected in 2014 were used to determine the total metal contents (Li et al., 2017). The determination of the total concentrations of heavy metals (Zn, Pb, Cu, Cr, Cd) and their chemical speciation can be found in Li et al. (2017). The heavy metal chemical forms were divided into the exchangeable fraction (F1), the bound to oxides fraction (F2), the organic bound fraction (F3), and the residual fraction (F4). 4.3.2 Modified ecological risk assessment (1) Results of total contamination and RAC of heavy metal Table 4.5 presents the total concentrations of heavy metals (Zn, Pb, Cu, Cr, Cd) in Sanyang wetland corresponding to three land use types.It is obvious that the contents of heavy metals in the surrounding sediments of the three land use types in Sanyang wetland exceeded the environmental background values. Among them, Zn and Cd are the two metals that presented the most exceedance of the background values, with the exceedance reaching 51.20 and 416.62 times, respectively. It was found that the heavy metals contained in sediments near the OP land use had lower concentrations than those near the other two land use types, indicating that relatively simple agricultural activities led less pollution to the surrounding environments than residential and industrial land use activities. Since the OPRI and RAI land use types had a significantly higher impact on all types of heavy metals in sediments, wastewater from electroplating and galvanizing facilities may be the main culprit 46 (Li et al., 2017). In summary, the sediments of Sanyang Wetland were contaminated seriously by heavy metals. Table 4.5 Concentrations of heavy metals in the sediments of Sanyang wetland (Li et al., 2017) Total metal concentration (mg/kg) Land use type Zn Pb Cu Cr Cd OP 3208.63 67.22 117.36 156.67 37.1 OPRI 5580.48 94.2 173.3 177.66 69.93 RAI 4794 95.2 247.8 256.36 63.73 Average value in Sanyang Wetland 1049.88 149.11 348.67 324.5 50.19 National soil background value 74.2 26 22.6 61 26.9 Wenzhou soil background value 108.99 38.38 32.7 88.11 0.168 According to Zhu et al. (2011), a composite pollution index (CPI) was used to evaluate the level of heavy metal pollution as calculated below: Cif = Ci /Cin (4.1) n CPI=  Cif /n (4.2) i Where Cif is the contamination coefficient; Ci is the concentration of heavy metal i; Cin is the background value of metal i; n is the number of heavy metals; CPI is the composite pollution index. Wenzhou soil background values are presented in Table 4.5. Sediment is rated as “unpolluted” for CPI < 1, and “contaminated” for CPI ≥ 1. The degree of heavy metal pollution 47 increases with the increase in CPI (Zhou et al., 2007). Table 4.6 lists the results of contamination coefficient and CPI of Sanyang Wetland under three land use types. Table 4.6 Contamination coefficient and CPI of heavy metals Metal contamination coefficient Land use type Zn Pb Cu Cr Cd CPI OP 29.44 1.75 3.59 1.78 221.03 51.52 OPRI 51.20 2.45 5.30 2.02 416.6 95.52 RAI 43.99 2.48 7.58 2.91 379.68 87.33 The contamination coefficients of all the metals exceeded the critical value of “1.0”, indicating the existence of heavy metal pollution in Sanyang Wetland. The degree of heavy metal pollution in this area was ranked in the order of Cd > Zn > Cu > Pb ≈ Cr. In addition, the CPI value was the highest for areas around the OPRI land use type, followed by RAI and OP land use types, indicating that heavy metal pollution was the most serious for sediments impacted by mixed land uses (e.g., mixed agriculture, residential and industrial land use). The toxicity of heavy metals in sediment can be evaluated more accurately from the chemical form of the metal as compared with the total concentration of metal. Figure 4.4 presents the distribution of chemical species (fractions) of the five selected heavy metals (Cu, Cr, Cd, Pb, Zn) in surface sediments impacted by different land use types. 48 (a) OP 100% 90% 80% Fraction% 70% 60% 50% 40% 30% 20% 10% 0% Zn Pb Cd Cr Cu (b) RAI 100% 90% 80% Fraction% 70% 60% 50% 40% 30% 20% 10% 0% Zn Pb Cd 49 Cr Cu (c) OPRI 100% Fraction% 80% 60% exchangeable 40% reduced 20% oxdized residual 0% Zn Pb Cd Cr Cu Figure 4.4 Distribution of chemical species in metal under different land use types, (a) OP land use, (b) RAI land use, (c) OPRI land use (Li et al., 2017) Based on the RAC evaluation standard, it can be seen from Figure 4.4 that the risk of Zn and Cd was higher than that of other heavy metals under the three different land use conditions. The F1 fraction (i.e. exchangeable fraction) of Zn in sediments impacted by OP, OPRI and RAI land uses were 57.64%, 61.97% and 59.12%, respectively, indicating “very high risk” level. Furthermore, F1 fraction of Cd in in sediments impacted these three land use types were 68.03%, 71.65% and 62.66%, respectively, indicating that Cd could easily enter the food chain and pose a “very high risk” to the wetland environment. Table 4.7 presents the risk levels of all the five metals in sediments under different land use type conditions. The metal Cr posed no risk under the three land use types, while the risks posed by Cu and Pb 50 were also relatively low. Overall, Sanyang Wetland was facing a serious ecological risk of heavy metal pollution, especially for Zn and Cd. Table 4.7 RAC classification of heavy metal in sediments under different land uses RAC risk classification Land use type Zn Pb Cu Cr Cd OP Very high Low Low No risk Very high OPRI Very high Low No risk No risk Very high RAI Very high Low Low No risk Very high (2) Potential ecological risk index assessment Table 4.8 presents the modified indexes (Ω) of heavy metal concentration. Such index can be seen as the toxicity index of heavy metals, and it is related to the chemical form of heavy metals. Compared with other metals, the index of Cd and Zn was higher because of their higher proportion of exchangeable fraction (Zhu et al., 2011). Table 4.8 Modified index (Ω) of heavy metals Ω index Land use type Zn Pb Cu Cr Cd OP 1.35 1.00 1.00 1.00 1.41 OPRI 1.37 1.00 1.00 1.00 1.43 RAI 1.35 1.00 1.00 1.00 1.38 51 Table 4.9 presents the potential ecological risk factor (Eir) and the modified potential ~ ecological risk factor (Eir) of heavy metals under three land use types in Sanyang Wetland. The difference between the conventional and modified potential ecological risk factor is that the modified method incorporated the chemical speciation of heavy metals. The results show ~ that for both Eir and Eir, the risks of heavy metals were ranked in order of Cr < Pb < Cu < Zn < Cd. The modified potential ecological risk factor of Cd in sediments impacted by OP, OPRI, and RAI land uses was 9337.53, 17871.83, and 15672.91, respectively, far exceeding the standard, which indicates that Cd posed a very high risk to the wetland ecosystem. The risk factor values of Pb, Cu and Cr in sediments under the impacts of three land use types were all below 40, indicating their “low risk” level. The risk factor of Zn in sediments ~ impacted by OP land use was also relatively low, with Eir and Eir value of 39.62 and 29.44, respectively, indicating its low risk under this land use type impacts. However, in sediments impacted by OPRI and RAI land uses, the risk factor value of Zn was in the range of 40 ~80, indicating its “moderate risk” level. 52 ~ Table 4.9 Results of Eir and Eir and risk level ~ Modified and original risk factor Eir/ Eir Land use OP OPRI RAI Zn Pb Cu Cr Cd 39.62/29.44 8.76/8.76 17.94/17.94 3.56/3.56 9337.53/6630.92 LR/ LR LR/ LR LR/ LR LR/ LR VHR/VHR 70.24/51.20 12.27/12.27 26.50/26.50 4.03/4.03 17871.83/12498.66 MR /MR LR/ LR LR/ LR LR/ LR VHR/VHR 59.59/43.99 12.40/12.40 37.89/37.89 5.82/5.82 15672.91/11390.53 MR /MR LR/ LR LR/ LR LR/ LR VHR/VHR Note: LR - Low risk; MR - Moderate risk; VHR - Very high risk 20000.00 18000.00 16000.00 14000.00 Risk 12000.00 10000.00 8000.00 RI 6000.00 4000.00 MRI 2000.00 0.00 OP OPRI RAI Figure 4.5 Comparison of total risk (MRI and RI) of heavy metals under different land uses 53 To quantify the overall potential ecological risk of heavy metals in the sediments of Sanyang Wetland, the values of RI and MRI were calculated and shown in Figure 4.5. All of MRI and RI values for sediments under the three land use types were higher than 600, indicating a very high potential ecological risk posed by heavy metals in Sanyang Wetland. In sediments impacted by the OPRI land use, the heavy metal risk index was the highest (i.e. 17984.9). The order of overall risks (for both MRI and RI) of heavy metals impacted by different land uses was OP < RAI < OPRI. This may indicate that a complex source of pollution would be one of the possible factors that increase the risk of heavy metal pollution. The results also showed that the overall risk calculated by MRI was higher than that by RI. For example, the modified risk index (MRI) of heavy metals in sediments impacted by OPRI land use was approximately 1.43 times that of RI. Due to the consideration of chemical speciation of metals, MRI may provide a more reasonable estimation of heavy metal risks. 4.4 General Risk Levels in Sanyang Wetland The land use type around sampling location A6 and S4 was corresponding to RAI and OP, respectively. The eutrophication risk information was then combined with the heavy metal risk information at these two sampling locations to obtain a general risk level of the wetland. At sampling location S4, the potential ecological risk value of heavy metals (HR) calculated by MRI (modified risk level) was 9407.41, far exceeding the evaluation criteria, so the heavy metal risk level would be “Very high”, with a membership grade of 1.0 shown in Figures 4.6 (a) according to Figure 3.5. Meanwhile, the eutrophication risk score was calculated as 70.81 using the BP Artificial neural network (ANN) model. It could be found from Figure 3.4 that the corresponding eutrophication risk (ER) would be partly “Low54 eutrophication” (with a membership grade of 0.419) and partly “Medium-eutrophication” (with a membership grade of 0.581). The membership functions of the fuzzy eutrophication risk events can then be shown in Figures. 4.6 (b) and (c). Therefore, there are two combinations of antecedents, including (a) if HR (heavy metal risk) is “Very high”, and ER (eutrophication risk) is “Low- eutrophication”, and (b) if HR (heavy metal risk) is “Very high”, and ER (eutrophication risk) is “Medium- eutrophication”. Figure 4.6 demonstrates the related fuzzy reasoning process. According to Table 3.8, the result of the fuzzy "AND" operation can be determined first, for example, µ GR= Min [µER, µHR]. It is means that the minimum degree of membership grade of the two input factors (ER and HR) was given to the output factor (GRL) (Li et al., 2007). The fuzzy rule showed that “if HR is Very high, and ER is Low-eutrophication, then the general risk is High”. Based on Fig.3.3, the “high” level of general risk is “represented by a triangular membership function” with a score between 60 and 100 (Li et al., 2007). Besides, the corresponding membership grade of this risk level was µGR= Min [1.0, 0.419] = 0.419 as shown in Figure 4.6 (c). In the meantime, the second antecedent “if HR is Very high, and ER is Medium-eutrophication” would generate a conclusion “then the general risk is Very high”. This “very high” general risk level event had a membership grade of µGR = Min [1.0, 0.581] = 0.581 as shown in Figure 4.6 (f). The fuzzy “OR” operation was then applied to combine the two fuzzy GRL events and as shown in Figure 4.6 (g). The score of the overall GRL was calculated as the centroid of 89. Based on this score, the recommended risk management actions would be “Take the full treatment of the region” (Table 3.9). 55 ER (a) (d) (b) (e) (c) GRL (f) (g) Note: fuzzy “AND” operation of (a) and (b) to obtain (c); fuzzy “AND” operation of (d) and (e) to obtain (f); fuzzy “OR” operation of (c) and (f) to obtain (g). Figure 4.6 Fuzzy inference process for sampling location S4, (a) and (d): heavy metal risks; (b) and (e): eutrophication risks; (c), (f) and (g): general risk levels. 56 In terms of sampling location A6, the potential ecological risk value of heavy metals (HR) calculated by MRI (modified risk level) is 15788.6, also far exceeding the evaluation criteria, so the heavy metal risk level would be “Very high” with a membership grade of 1.0 show in Figures 4.7(a) according to Figure 3.5. Meanwhile, the eutrophication risk score was calculated as 78.48 using the BP ANN model. It could thus be found from Figure 3.4 that the corresponding eutrophication risk (ER) would be partly “Medium-eutrophication” (with a membership grade of 0.652) and partly “High-eutrophication” (with a membership grade of 0.348), as shown in Figures 4.7 (b) and (c), respectively. Therefore, two combinations of antecedents were involved, including (a) if HR (heavy metal risk) is “Very high”, and ER (eutrophication risk) is “Medium - eutrophication”, and (b) if HR (heavy metal risk) is “Very high”, and ER (eutrophication risk) is “High - eutrophication”. The first antecedent of “if HR is Very high, and ER is Medium-eutrophication” would produce a conclusion of “then the general risk is Very high”. The corresponding membership grade of this risk level was µGRv= Min [1.0, 0.652] =b0.652 as shown in Figure 4.7(f). The second antecedent of “if HR is Very high, and ER is High-eutrophication” would lead to a conclusion of “then the general risk is Very high”, with a membership grade of µGR = Min [1.0, 0.348] = 0.348 as shown in Figure 4.7 (c). The fuzzy “OR” operation was then applied to the two obtained fuzzy GRL events. As seen from Figure 4.7 (g), both GRLs are "very high risk” with a score ranging from 80-100. The membership grade of the overall GRL was then µGR = Max [0.652, 0.348] = 0.652 as shown in Figure 4.7 (g). Based on the patterning of Figure 4.7 (g), the score of the overall GRL event was calculated as the centroid of 93.3. Based on this risk score, the recommended risk management actions would still be “Take the full treatment of the region” (Table 3.9). 57 GRL (a) (d) (b) (e) (c) (f) (g) 1 V-H 0.8 0.652 µGR Note: fuzzy “AND” operation of (a) and (b) to obtain (c); fuzzy “AND” operation of (d) and (e) to obtain (f); fuzzy “OR” operation of (c) and (f) to obtain (g). 0.6 Figure 4.7 Fuzzy inference process for sampling location A6, (a) and (d): heavy metal risks; (b) and (e): eutrophication risks; (c), (f) and (g): general risk levels. 0.4 0.2 0 70 90 110 130 58 In summary, in sampling S4, the general risk level is partly “very high” and partly “high” by considering two types of risk (eutrophication risk and heavy metal risk), and in sampling A6, the general risk level is “very high”. The reason of lower overall risk value at S4 than A6 may be caused by the type of land use around them, with the pollution caused by industrial activities and human daily life having a greater impact on wetlands. Such high overall risks would not only affect the wetland ecosystem, but also threaten human health. Compared with the conventional method of considering individual risk information, the fuzzy risk assessment method would provide more realistic risk perception for decision making. 4.5 Scenario Analysis The above analysis indicated that the Sanyang Wetland had a serious general risk level by comprehensively taking into account the eutrophication risk and heavy metal risk, and thus it may need appropriate risk management actions such as “taking the full treatment of the region”. Two management scenarios were designed by considering the source of pollution as well as the characteristics and current status of pollution control methods. One is a more conservative approach and the other is a more active approach. 59 4.5.1 Scenario 1: sediment dredging method The sediment dredging method is one of the main measures in water pollution control. It can not only permanently remove the pollutants in the sediment, but also reduce the influence of pollutants in the sediment on the overlying waterbody. However, it is also easy to cause secondary pollution during the implementation of the dredging method, thus causing a certain impact on the effect of this management action. Scenario 1 was then designed to implement a sediment dredging program to manage the wetland risk. According to previous studies, after dredging, the content of heavy metals in sediments could generally decrease, and the decrease could be approximately Pb (50%) > Cu (45%) > Cd (40%) = Zn (40%) > Cr (30%). Table 4.10 lists the heavy metal risk assessment results of scenario 1 based on the modified potential ecological risk assessment method. Meanwhile, sediment dredging may also significantly affect the physical and chemical parameters and nutrient contents of water. For example, the dissolved oxygen content in water could be increased by about 60%, the transparency could be enhanced by approximately 30%, the total nitrogen and total phosphorus could be decreased by 20% and 25%, respectively, and the chemical oxygen demand could drop significantly by about 30% (Tong et al., 2015). Based on such approximation and using the ANN eutrophication model, the water eutrophication risk assessment results of scenario 1 can be obtained (Table 4.11) 60 Table 4.10 Heavy metal risk assessment results under scenario 1 Land use type Concentration (mg/Kg) MRI Zn Pb Cu Cr Cd OP (S4) 1925.18 33.61 64.55 109.67 22.26 5643.03 RAI (A6) 2876.4 47.6 136.29 179.45 38.24 9470.61 Table 4.11 Eutrophication risk assessment results under scenario 1 Land use type Normalized concentration of parameter Score of trophic Chl-a COD TN TP SD state calculated (mg/m3) (mg/L) (mg/L) (mg/L) (m) by ANN model RAI (A6) 9.50 22.34 7.23 0.27 3.96 65.39 OP (S4) 20.56 10.23 6.30 0.16 4.01 59.43 The general risk level of the wetland after implementing scenario 1 was then evaluated according to the above general risk assessment method. At sampling location A6, the potential ecological risk value of heavy metals (HR) calculated by MRI (modified risk level) was 9470.61. Although it is only 60% of the MRI before implementing scenario 1, it is still far exceeding the evaluation criteria, so the risk level would be “Very high”, with a membership grade of 1.0 according to Figure 3.5. Meanwhile, the eutrophication risk score after implementing scenario 1 was calculated as 65.39 using the ANN model. It could also be found from Figure 3.4 that the corresponding eutrophication risk (ER) would be partly “Medium-eutrophication” (with a membership grade of 0.039) and partly “Loweutrophication” (with a membership grade of 0.961). Therefore, two combinations of antecedents were involved, including (a) if HR (heavy metal risk) is “Very high”, and ER 61 (eutrophication risk) is “Medium - Eutrophication”, and (b) if HR (heavy metal risk) is “Very high”, and ER (eutrophication risk) is “Low - eutrophication”. The first antecedent “if HR is Very high, and ER is Medium-eutrophication” would generate a conclusion part of “then the general risk is Very high”, with the corresponding membership grade as µGR = Min [1.0, 0.039] = 0.039. The second antecedent “if HR is Very high, and ER is Low- eutrophication” would lead to the conclusion part of “then the general risk is high”, with a membership grade of µGR = Min [1.0, 0.961] = 0.961. The fuzzy “OR” operation was then applied to combine the two fuzzy GRL events and the final overall fuzzy general risk level event (“high risk”) is shown in Figure 4.8 (g). The risk score of the final overall GRL was calculated as the centroid of 84, and thus further risk management actions may need to be taken such as “Taking target treatment on the most serious pollution” (Table 3.9). 62 (a) (d) (b) (e) (c) H (f) Note: fuzzy “AND” operation of (a) and (b) to obtain (c); fuzzy “AND” operation of (d) and (e) to obtain (f); fuzzy “OR” operation of (c) and (f) to obtain (g). (g) H Figure 4.8 Fuzzy inference process for sampling location A6 under scenario 1, (a) and (d): heavy metal risks; (b) and (e): eutrophication risks; (c), (f) and (g): general risk levels. 63 In terms of sampling location S4, the potential ecological risk value of heavy metals (HR) calculated by MRI (modified risk level) was 5643.03 after implementing scenario 1, still far exceeding the evaluation criteria, so the risk level would be “Very high” with a membership grade of 1.0 according to Figure 3.5. Meanwhile, the eutrophication risk score after implementing scenario 1 was calculated as 59.43 using the ANN model. It could also be found from Figure 3.4 that the corresponding eutrophication risk (ER) would be partly “Mesotrophic” (with a membership grade of 0.557) and partly “Low-eutrophication” (with a membership grade of 0.443). The first antecedent of “if HR is Very high, and ER is Loweutrophication” would lead to the conclusion of “then the general risk is high”, with a corresponding membership grade of µGR = Min [1.0, 0.443] = 0.443. The second antecedent of “if HR is Very high, and ER is Mesotrophic” led to “then the general risk is high”, with a membership grade of µGR = Min [1.0, 0.557] = 0.557. The fuzzy “OR” operation was then applied to combine these two fuzzy GRL events to obtain a “high risk” GRL event, with a membership grade of µGR = Max [0.443, 0.557] = 0.557 as shown in Figure 4.9(c). The risk score of the final overall GRL was then calculated as the centroid of 80, indicating that further risk management actions would be needed such as “taking target treatment on the most serious pollution” (Table 3.9). 64 (b) (a) (e) (b) H (c) H (g) (f) Note: fuzzy “AND” operation of (a) and (b) to obtain (c); fuzzy “AND” operation of (d) and (e) to obtain (f); fuzzy “OR” operation of (c) and (f) to obtain (g). Figure 4.9 Fuzzy inference process for sampling location S4 under scenario 1, (a) and (d): heavy metal risks; (b) and (e): eutrophication risks; (c), (f) and (g): general risk levels. 65 In summary, the effect of implementing sediment dredging on location S4 was more obvious than sampling location A6. The main difference is the impact on reducing the risk of eutrophication, but not on the risk of heavy metals in sediments. Even though the removal rate of heavy metals in sediments under scenario 1 was high, the risk of heavy metals remained unchanged because the Cd content as the main pollutant was still far exceeding the standard. The general risk level at both S4 and A6 locations was decreased, indicating that sediment dredging had an obvious effect on managing the risk of wetland, especially on the risk of eutrophication. However, through sediment dredging, the overall risk of wetlands was still relatively high, and more effective methods could be sought. The sediment dredging also requires the subsequent treatment of the heavy metal-rich sediments that can commonly be dealt with dehydration and drying, followed by landfill. This method consumes a lot of energy and may cause pollution risks of land near the landfill. Therefore, considering the removal effect of sediment dredging and the post-processing problems, it is necessary to find another way to solve the problem. 4.5.2 Scenario 2: ecological restoration The ecological restoration measures mainly include the following: (1) microbial remediation; (2) phytoremediation; (3) reaeration. In general, this risk management measure is more complicated and has a longer duration, but at the same time, it also has the advantages such as not causing secondary pollution. Scenario 2 was then designed in this thesis to examine the effects of using ecological restoration measures to reduce wetland risk. According to previous studies, after a series of ecological restoration measures, the heavy metal content in the sediment could be greatly reduced, especially for Cr and Zn, which can be removed by about 90%. However, the removal rate of Cd could be about 40% which is 66 similar as that under scenario 1 (He et al., 2007). Table 4.12 presents the heavy metal risk assessment results of scenario 2 based on the modified potential ecological risk assessment method. As for the water quality parameters, after ecological restoration measures, the removal rate of COD could reach about 80%, and the decrease of ammonia nitrogen and total phosphorus could also achieve about 50% and 75%, respectively (Tan et al., 2006). Based on such information, the water eutrophication risk results of scenario 2 can be obtained using the ANN model (Table 4.13). It can be seen that scenario 2 had a better effect on both heavy metal pollution and water eutrophication. Table 4.12 Heavy metal risk assessment results of scenario 2 Land use Concentration (mg/kg) MRI type Zn Pb Cu Cr Cd OP (S4) 320.86 53.78 29.34 15.67 22.26 5618.33 RAI (A6) 479.4 76.16 61.95 25.636 38.24 9429.68 Table 4.13 Eutrophication risk assessment results of scenario 2 Land use type Normalized concentration of parameter Score of trophic Chl-a COD TN TP SD state calculated (mg/m3) (mg/L) (mg/L) (mg/L) (m) by ANN model RAI (A6) 3.17 6.38 4.52 0.09 4.57 50.66 OP (S4) 6.85 2.92 3.94 0.05 4.62 46.05 The general risk level in the wetland after implementing scenario 2 can then be obtained using similar method introduced above. In terms of sampling location A6, the 67 potential ecological risk value of heavy metals (HR) calculated by MRI (modified risk level) was 9429.68, also far exceeding the evaluation criteria, so the risk level would be “Very high” with a membership grade of 1.0 according to Figure 3.5. Meanwhile, the eutrophication risk score after implementing scenario 2 was 50.66 as calculated by the ANN model. It could be found from Figure 3.4 that the corresponding eutrophication risk (ER) would be partly “Mesotrophic” (with a membership grade of 0.9132) and partly “Oligotrophic” (with a membership grade of 0.0868). Therefore, the two combinations of antecedents were involved, including (a) if HR (heavy metal risk) is “Very high”, and ER (eutrophication risk) is “Mesotrophic”, and (b) if HR (heavy metal risk) is “Very high”, and ER (eutrophication risk) is “Oligotrophic”. The first antecedent of “if HR is Very high, and ER is Mesotrophic” led to a conclusion part of “then the general risk is high”, with a corresponding membership grade of µGR = Min [1.0, 0.9132] = 0.9132. In the meantime, the second antecedent of “if HR is Very high, and ER is Oligotrophic” would lead to “then the general risk is Medium-High”, with a membership grade of µGR = Min [1.0, 0.0868] = 0.0868. Again, the fuzzy “OR” operation was then applied to combine the two fuzzy GRL events as shown in Figure 4.10 (g). The risk score of the final overall GRL was then calculated as the centroid of 68, indicating that further risk management actions might be needed such as “warning of water pollution and take some emergency measures” (Table 3.9). 68 (a) (d) ER (b) (c) (e) (f) (g) Note: fuzzy “AND” operation of (a) and (b) to obtain (c); fuzzy “AND” operation of (d) and (e) to obtain (f); fuzzy “OR” operation of (c) and (f) to obtain (g). H Figure 4.10 Fuzzy inference process for sampling location A6 under scenario 2, (a) and (d): heavy metal risks; (b) and (e): eutrophication risks; (c), (f) and (g): general risk levels. 69 In terms of sampling location S4, the potential ecological risk value of heavy metals (HR) calculated by MRI (modified risk level) was 5618.33 after implementing scenario 2, still far exceeding the evaluation criteria, so the risk level would be “Very high” with a membership grade of 1.0 according to Figure 3.5. Meanwhile, the eutrophication risk score was calculated as 46.05 using the ANN model, and the corresponding eutrophication risk (ER) would be partly “Mesotrophic” (with a membership grade of 0.831) and partly “Oligotrophic” (with a membership grade of 0.169). The first antecedent of “if HR is Very high, and ER is Low” would lead to “then the general risk is Medium-high”, with a membership grade of µGR = Min [1.0, 0.169] = 0.169. The second antecedent “if HR is Very high, and ER is Mesotrophic” led to “then the general risk is high”, with a membership grade of µ GR = Min [1.0, 0.831] = 0.831. The fuzzy “OR” operation was then applied to combine the two fuzzy GRL events as shown in Figure 4.11 (g). The risk score of the final overall GRL was then calculated 68, indicating that other risk management actions are needed after implementing scenario 2 such as “warning of water pollution and take some emergency measures” (Table 3.9). 70 (a) (d) ER (b) (e) (f) (c) Note: fuzzy “AND” operation of (a) and (b) to obtain (c); fuzzy “AND” operation of (d) and (e) to obtain (f); fuzzy “OR” operation of (c) and (f) to obtain (g). (g) H Figure 4.11 Fuzzy inference process for sampling location S4 under scenario 2, (a) and (d): heavy metal risks; (b) and (e): eutrophication risks; (c), (f) and (g): general risk levels. 71 Similar to the implementation of scenario 1, despite the high removal rate of heavy metals by ecological restoration, the risk of heavy metals remained high after implementing scenario 2 due to the extremely high concentration of heavy metals in Sanyang Wetland sediments and the insufficiency to improve the removal rate of Cd. However, the ecological restoration measure had more obvious effects on reducing eutrophication risk. It can be seen from the above results that the effects of scenario 1 and scenario 2 on the risk reduction of heavy metal pollution were generally similar (i.e. reducing risk by about 40%), mainly because the Cd content in the Sanyang Wetland region was excessively high. Moreover, the removal rates of Cd by the two management scenario measures were similar, leading to similar levels of heavy metal pollution risk. However, scenario 2 could reduce the risk of water eutrophication much higher than scenario 1. Overall, the implementation of scenario 1 and scenario 2 could both reduce the general risk of wetland, but the risk could decrease to a healthier level when implementing scenario 2. 72 CHAPTER 5: CONCLUSION With the continuous development of the society and economy as well as the continuous advancement of industrialization, the landscape of wetlands faces some major environmental problems. Among them, eutrophication of waterbodies and heavy metal pollution of surface sediments are the most common and serious problems. How to conduct a more comprehensive risk assessment of wetlands and obtain a more reasonable risk estimation for risk management decision making have become a hot research topic recently. In this thesis research, the existing risk evaluation methods for eutrophication and heavy metal pollution were analyzed. The artificial neural network (ANN) and the improved potential ecological risk index method were then proposed to evaluate the eutrophication risk and heavy metal risk, respectively. The heavy metal risk was proposed to be assessed using the metal speciation information instead of just using the total concentration of heavy metals. A fuzzy set approach was proposed to combine the information from two types of risk (eutrophication and heavy metal risk) to obtain a general risk level which can then be used to assess the overall risk of wetlands. A fuzzy rule base was established for facilitating the combination of different risk types. Moreover, the proposed general risk assessment framework was then demonstrated by taking Sanyang Wetland in China as a case study using the water quality and sediment sampling data obtained in the past years. The water eutrophication risk and heavy metal risk under different land use types in the study wetland were calculated and then used for estimating the general wetland risk level through a fuzzy inference process. Two different risk management scenarios were examined for their effects on reducing risk levels. From the obtained results, the following conclusions can be drawn: 73 (1) The artificial neural network (ANN) is effective in obtaining the score of trophic state of waterbodies by inputting a series of water quality parameters, and it can be reasonably applied to assess the water quality eutrophication risk. (2) The Sanyang Wetland was associated with eutrophication risk all year round, especially in areas close to residential and industrial activities where the water quality was even worse. The risk of eutrophication was mainly due to the frequent anthropogenic activities around the waterbody. The large amount of industrial, agricultural and domestic sewage discharges led to higher nutrients in water. In addition, Wenzhou's warmer climate is more suitable for algae growth. (3) The modified potential ecological risk index (MRI) method incorporated the chemical speciation and bioavailability of heavy metals into the risk assessment process. It was used to analyze the risk of heavy metal pollution in Sanyang Wetland. The results showed that all the MRI values of sediments impacted by three land use types were higher than 600, indicating that the study area had a very high potential ecological risk. Among the five selected heavy metals, Cd posed the highest ecological risk, with the risk ranking order of Cr < Pb < Cu < Zn < Cd. Overall, Sanyang Wetland faced the serious ecological risk of heavy metal pollution, and Cd should be regarded as a priority pollutant. (4) Different types of land use had a certain impact on the risk of heavy metal pollution in Sanyang Wetland. It was observed that residential and industrial activities had a greater impact on Zn pollution, but the agricultural activities didn’t cause serious accumulation of Zn. (5) A fuzzy set approach was applied to assess the general risk level of wetland by using the fuzzy logic method to combine different risk information. Compared with conventional 74 risk assessment methods, the effective reflection of information uncertainties into the general risk assessment framework enhanced the robustness of the proposed method. The general risk level of Sanyang Wetland obtained from the proposed fuzzy assessment method was rated as “very high”, indicating that Sanyang wetland needs immediate risk management measures. (6) Between the two proposed risk management scenarios for Sanyang Wetland, scenario 2 (i.e. ecological restoration measures) would take longer duration and more investment, but it could lead to much more significant effects on reducing risk than scenario 1 (i.e. sediment dredging measure). The proposed fuzzy risk assessment method also had some limitations. The membership functions of the relevant fuzzy events and the fuzzy rule base for generating general risk levels were established based on questionnaire survey, and they represent the cognition of the experts involved. The boundaries and shapes of membership functions are also subjective. The uncertainty in the fuzzy membership function and the rule base itself was not considered in this thesis research, and their impacts on the risk assessment results need further investigation. The future direction of this research could focus on reducing the subjectivity of evaluation. 75 REFERENCES Aguilera, P. A., Fernández, A., Fernández, R., Rumí, R., & Salmerón, A. (2011). Bayesian networks in environmental modelling. Environmental Modelling & Software, 26(12), 1376-1388. Andersen, J. H., & Conley, D. J. (2009). Eutrophication in coastal marine ecosystems: towards better understanding and management strategies. Hydrobiologia, 629(1), 1-4. Arhonditsis, G. B., & Brett, M. T. (2005). Eutrophication model for Lake Washington (USA): Part I. Model description and sensitivity analysis. Ecological modelling, 187(2), 140178. Arhonditsis, G., Eleftheriadou, M., Karydis, M., & Tsirtsis, G. (2003). Eutrophication risk assessment in coastal embayments using simple statistical models. Marine Pollution Bulletin, 46(9), 1174-1178. Arhonditsis, G. B., Qian, S. S., Stow, C. A., Lamon, E. C., & Reckhow, K. H. (2007). Eutrophication risk assessment using Bayesian calibration of process-based models: application to a mesotrophic lake. ecological modelling, 208(2), 215-229. Baker, J.P., Hulse, D.W., Gregory, S.V., White, D., Van Sickle, J., Berger, P.A., Dole, D., Schumaker, N.H., 2004. Alternative futures for the Willamette River Basin, Oregon. Ecol. Appl. 14 (2), 313–324. Bai, J., Cui, B., Chen, B., Zhang, K., Deng, W., Gao, H., & Xiao, R. (2011). Spatial distribution and ecological risk assessment of heavy metals in surface sediments from a typical plateau lake wetland, China. Ecological Modelling, 222(2), 301-306. Bao, C., Li, J., & Wu, D. (2018). A fuzzy mapping framework for risk aggregation based on risk matrices. Journal of Risk Research, 21(5), 539-561. 76 Benson, N. U., Adedapo, A. E., Fred-Ahmadu, O. H., Williams, A. B., Udosen, E. D., Ayejuyo, O. O., & Olajire, A. A. (2018). New ecological risk indices for evaluating heavy metals contamination in aquatic sediment: a case study of the Gulf of Guinea. Regional Studies in Marine Science, 18, 44-56. Bo, L., Wang, D., Li, T., Li, Y., Zhang, G., Wang, C., & Zhang, S. (2015). Accumulation and risk assessment of heavy metals in water, sediments, and aquatic organisms in rural rivers in the Taihu Lake region, China. Environmental Science and Pollution Research, 22(9), 6721-6731. Borsuk, M. E., Stow, C. A., & Reckhow, K. H. (2002). Integrative environmental prediction using Bayesian networks: A synthesis of models describing estuarine eutrophication. Burns, N. M., Rutherford, J. C., & Clayton, J. S. (1999). A monitoring and classification system for New Zealand lakes and reservoirs. Lake and Reservoir Management, 15(4), 255-271. Caniani, D., Lioi, D. S., Mancini, I. M., Masi, S., & Sdao, F. (2011). Fuzzy logic model development for groundwater pollution risk assessment. Eur Water, 35, 13-22. Carlson, R. E. (1977). A trophic state index for lakes. Limnology and oceanography, 22(2), 361-369. Chang, N. B., Chen, H. W., & Ning, S. K. (2001). Identification of river water quality using the fuzzy synthetic evaluation approach. Journal of environmental management, 63(3), 293-305. Cloern, J. E. (2001). Our evolving conceptual model of the coastal eutrophication problem. Marine ecology progress series, 210, 223-253. 77 Chen, Q., Mei, K., Dahlgren, R. A., Wang, T., Gong, J., & Zhang, M. (2016). Impacts of land use and population density on seasonal surface water quality using a modified geographically weighted regression. Science of the Total Environment, 572, 450-466. Chen (2012). Risk Assessment Method and Case Analysis of Water Eutrophication - Taking Xiangxi River as an Example. Master 's Degree Thesis of Central University for Nationalities. Chen, S., Chen, B., & Fath, B. D. (2013). Ecological risk assessment on the system scale: a review of state-of-the-art models and future perspectives. Ecological Modelling, 250, 25-33. Cornish, E. (2004). Futuring: The exploration of the future. World Future Society. Correll, D. L. (1998). The role of phosphorus in the eutrophication of receiving waters: A review. Journal of Environmental Quality, 27(2), 261-266. Dam, R. A., Camilleri, C., Bayliss, P., & Markich, S. J. (2004). Ecological risk assessment of tebuthiuron following application on tropical Australian wetlands. Human and Ecological Risk Assessment, 10(6), 1069-1097. Deng, D.P., Liu, G.,Li, X.D., Tang, F.,& Hua, R. M.(2006). Lake eutrophication comprehensive evaluation slope weighted scoring method. Journal of Environmental Science, 26(8),1386-1392. Dharmendra, S., Kumar, M. R., Chinmayee, A., Ranjan, S. D., & Ranjan, P. C. (2018). Appraisal of Temporal Variation of Water Quality Status in terms of Water Quality Index (WQI) at Paradip Port, India. Research Journal of Chemistry and Environment. Vol, 22, 12. Gudimov, A., Stremilov, S., Ramin, M., & Arhonditsis, G. B. (2010). Eutrophication risk assessment in Hamilton Harbour: system analysis and evaluation of nutrient loading scenarios. Journal of Great Lakes Research, 36(3), 520-539. 78 Fernandes, H. M. (1997). Heavy metal distribution in sediments and ecological risk assessment: the role of diagenetic processes in reducing metal toxicity in bottom sediments. Environmental Pollution, 97(3), 317-325. Ferreira, J. G., Andersen, J. H., Borja, A., Bricker, S. B., Camp, J., Da Silva, M. C., ... & Lancelot, C. (2011). Overview of eutrophication indicators to assess environmental status within the European Marine Strategy Framework Directive. Estuarine, Coastal and Shelf Science, 93(2), 117-131. Finlayson, C. M., van Dam, R. A., & Humphrey, C. L. (1998). Wetland risk assessment: A framework and methods for predicting and assessing change in ecological character. Technical report submitted to the Ramsar STRP. Gudimov, A., Stremilov, S., Ramin, M., & Arhonditsis, G. B. (2010). Eutrophication risk assessment in Hamilton Harbour: system analysis and evaluation of nutrient loading scenarios. Journal of Great Lakes Research, 36(3), 520-539. Gao, Y., Zheng, X, M., & Zhou, L, M. (2014). Spatial Distribution and Pollution Assessment of Heavy Metals of Sediments in Upper Reach OF the Wenruitang River. Urban Environment &Urban Ecology, 27(4):1-5. Hakanson, L. (1980). An ecological risk index for aquatic pollution control. A sedimentological approach. Water research, 14(8), 975-1001. Hakanson, L., Jasson, M. (Eds.), 1983. Principles of Lake Sedimentology. Springer-Verlag, Berlin. Hongyi, N. I. U., Wenjing, D. E. N. G., Qunhe, W. U., & Xingeng, C. H. E. N. (2009). Potential toxic risk of heavy metals from sediment of the Pearl River in South China. Journal of Environmental Sciences, 21(8), 1053-1058. 79 Hou, D., He, J., Lü, C., Ren, L., Fan, Q., Wang, J., & Xie, Z. (2013). Distribution characteristics and potential ecological risk assessment of heavy metals (Cu, Pb, Zn, Cd) in water and sediments from Lake Dalinouer, China. Ecotoxicology and environmental safety, 93, 135-144Liu, J., Li, Y., Zhang, B., Cao, J., Cao, Z., & Domagalski, J. (2009). Ecological risk of heavy metals in sediments of the Luan River source water. Ecotoxicology, 18(6), 748-758. Huo, S., He, Z., Su, J., Xi, B., & Zhu, C. (2013). Using artificial neural network models for eutrophication prediction. Procedia Environmental Sciences, 18, 310-316. Jain, C. K. (2004). Metal fractionation study on bed sediments of River Yamuna, India. Water research, 38(3), 569-578. Jian-Min, ZHOU., Zhi, D., Mei-Fang, C. A. I., & Cong-Qiang, L. I. U. (2007). Soil heavy metal pollution around the Dabaoshan mine, Guangdong province, China. Pedosphere, 17(5), 588-594. Jiang, Y., Xu, Z., & Yin, H. (2006). Study on improved BP artificial neural networks in eutrophication assessment of China eastern lakes. Journal of Hydrodynamics, 18(1), 517-521. Jones, J. R., & Bachmann, R. W. (1976). Prediction of phosphorus and chlorophyll levels in lakes. Journal (Water Pollution Control Federation), 2176-2182. Jørgensen, S. E. (2010). A review of recent developments in lake modelling. Ecological modelling, 221(4), 689-692. Karydis, M., & University of the Aegean, Lesvos I.(Greece). Department of Environmental Studies; University of the Aegean, Lesvos I.(Greece). Department of Environmental Studies. (2009, December). Eutrophication assessment of coastal waters based on indicators: a literature review. In Proceedings of the International Conference on 80 Environmental Science and Technology (Vol. 1). University of the Aegean, Chania(Greece). Karydis, M., & University of the Aegean, Lesvos I. (Greece). Department of Environmental Studies; University of the Aegean, Lesvos I. (Greece). Department of Environmental Studies. (2009, December). Eutrophication assessment of coastal waters based on indicators: a literature review. In Proceedings of the International Conference on Environmental Science and Technology (Vol. 1). University of the Aegean, Chania (Greece). Karul, C., Soyupak, S., Çilesiz, A. F., Akbay, N., & Germen, E. (2000). Case studies on the use of neural networks in eutrophication modeling. Ecological modelling, 134(2-3), 145-152. Kuo, J. T., Hsieh, M. H., Lung, W. S., & She, N. (2007). Using artificial neural network for reservoir eutrophication prediction. Ecological modelling, 200(1-2), 171-177. King, R. S., & Richardson, C. J. (2003). Integrating bioassessment and ecological risk assessment: an approach to developing numerical water-quality criteria. Environmental management, 31(6), 795-809. Koelmans, A. A., Van der Heijde, A., Knijff, L. M., & Aalderink, R. H. (2001). Integrated modelling of eutrophication and organic contaminant fate & effects in aquatic ecosystems. A review. Water Research, 35(15), 3517-3536. Lackey, R. T. (1994). Ecological risk assessment. Fisheries, 19(9), 14-19. Latour, J. B., Reiling, R., & Slooff, W. (1994). Ecological standards for eutrophication and desiccation: perspectives for a risk assessment. Water, Air, & Soil Pollution, 78(3), 265-277. 81 Lamon, E. C., & Stow, C. A. (2004). Bayesian methods for regional-scale eutrophication models. Water Research, 38(11), 2764-2774. Lang T. Scenario planning: your key to the future. Curtin University of Technology; 2001. Lemly, A. D. (1997). PROFILE: risk assessment as an environmental management tool: considerations for freshwater wetlands. Environmental Management, 21(3), 343-358. Lek, S., & Guégan, J. F. (1999). Artificial neural networks as a tool in ecological modelling, an introduction. Ecological modelling, 120(2-3), 65-73. Li, T., Chu, C., Zhang, Y., Ju, M., & Wang, Y. (2017). Contrasting Eutrophication Risks and Countermeasures in Different Water Bodies: Assessments to Support Targeted Watershed Management. International journal of environmental research and public health, 14(7), 695. Li, Y., Arocena, J. M., Zhang, Q., Thring, R. W., & Li, J. (2017). Heavy metals and nutrients (carbon, nitrogen, and phosphorus) in sediments: relationships to land uses, environmental risks, and management. Environmental Science and Pollution Research, 24(8), 7403-7412. Lin, G.S., Huang, X.Y., & Li, J. (2010). Application of Artificial Neural Network in Evaluation of Eutrophication of Reservoirs in Shenzhen City. Environmental Monitoring Management and Technology, 22(1), 59-63. Liou, S. M., Lo, S. L., & Wang, S. H. (2004). A generalized water quality index for Taiwan. Environmental Monitoring and Assessment, 96(1-3), 35-52. Liu, J., Jia, H, Y. (2007). Pollution and prevention of river water environment in Wenzhou. Zhejiang hydro-technics, 152(4), 73-77. 82 Liu, J. J., Ni, Z. X., Diao, Z. H., Hu, Y. X., & Xu, X. R. (2018). Contamination level, chemical fraction and ecological risk of heavy metals in sediments from Daya Bay, South China Sea. Marine pollution bulletin, 128, 132-139. Liu, Q., Wang, F., Meng, F., Jiang, L., Li, G., & Zhou, R. (2018). Assessment of metal contamination in estuarine surface sediments from Dongying City, China: Use of a modified ecological risk index. Marine pollution bulletin, 126, 293-303. Liu, W., Li, S., Bu, H., Zhang, Q., & Liu, G. (2012). Eutrophication in the Yunnan Plateau lakes: the influence of lake morphology, watershed land use, and socioeconomic factors. Environmental Science and Pollution Research, 19(3), 858-870. Malmaeus, J. M., & Håkanson, L. (2004). Development of a lake eutrophication model. Ecological Modelling, 171(1), 35-63. Malekmohammadi, B., & Blouchi, L. R. (2014). Ecological risk assessment of wetland ecosystems using multi criteria decision making and geographic information system. Ecological Indicators, 41, 133-144. Mladenović-Ranisavljević, I. I., & Žerajić, S. A. (2018). Comparison of different models of water quality index in the assessment of surface water quality. International Journal of Environmental Science and Technology, 15(3), 665-674. Morrison, G., Fatoki, O. S., Persson, L., & Ekberg, A. (2001). Assessment of the impact of point source pollution from the Keiskammahoek Sewage Treatment Plant on the Keiskamma River-pH, electrical conductivity, oxygen-demanding substance (COD) and nutrients. Water Sa, 27(4), 475-480. Norton, S. B., Rodier, D. J., van der Schalie, W. H., Wood, W. P., Slimak, M. W., & Gentile, J. H. (1992). A framework for ecological risk assessment at the EPA. Environmental toxicology and chemistry, 11(12), 1663-1672. 83 Perin, G., Craboledda, L., Lucchese, M., Cirillo, R., Dotta, L., Zanette, M. L., & Orio, A. A. (1985). Heavy metal speciation in the sediments of northern Adriatic Sea. A new approach for environmental toxicity determination. Heavy metals in the environment, 2(1), 454-456. Picart, S. S., Allen, J. I., Butenschön, M., Artioli, Y., de Mora, L., Wakelin, S., & Holt, J. (2015). What can ecosystem models tell us about the risk of eutrophication in the North Sea? Climatic change, 132(1), 111-125. Pollino, C. A., Woodberry, O., Nicholson, A., Korb, K., & Hart, B. T. (2007). Parameterisation and evaluation of a Bayesian network for use in an ecological risk assessment. Environmental Modelling & Software, 22(8), 1140-1152. Qu, C. S., Chen, W., Bi, J., Huang, L., & Li, F. Y. (2011). Ecological risk assessment of pesticide residues in Taihu Lake wetland, China. Ecological Modelling, 222(2), 287292. Qian, S. J., & Richardson, C. J. (1997). Estimating the long‐term phosphorus accretion rate in the Everglades: A Bayesian Approach with risk assessment. Water Resources Research, 33(7), 1681-1688. Rahman, M. S., Saha, N., & Molla, A. H. (2014). Potential ecological risk assessment of heavy metal contamination in sediment and water body around Dhaka export processing zone, Bangladesh. Environmental earth sciences, 71(5), 2293-2308. Rajankar, P. N., Gulhane, S. R., Tambekar, D. H., Ramteke, D. S., & Wate, S. R. (2009). Water quality assessment of groundwater resources in Nagpur Region (India) based on WQI. Journal of Chemistry, 6(3), 905-908. 84 Reckhow, K. H. (1994). Water quality simulation modeling and uncertainty analysis for risk assessment and decision making. Ecological Modelling, 72(1-2), 1-20. Ren, L., Dong, Z.C., & Li, S.H. (2004). Application of artificial neural network in eutrophication evaluation of Taihu Lake. Journal of Hohai University,32(2),147-150. Ren-Ying, L. I., Hao, Y. A. N. G., Zhi-Gao, Z. H. O. U., Jun-Jie, L. Ü., Xiao-Hua, S. H. A. O., & Feng, J. I. N. (2007). Fractionation of Heavy Metals in Sediments from Dianchi Lake, China1. Pedosphere, 17(2), 265-272. Ruilian, Y. U., Xing, Y., Yuanhui, Z., Gongren, H. U., & Xianglin, T. U. (2008). Heavy metal pollution in intertidal sediments from Quanzhou Bay, China. Journal of Environmental Sciences, 20(6), 664-669. Salari, M., Shahid, E. S., Afzali, S. H., Ehteshami, M., Conti, G. O., Derakhshan, Z., & Sheibani, S. N. (2018). Quality assessment and artificial neural networks modeling for characterization of chemical and physical parameters of potable water. Food and Chemical Toxicology, 118, 212-219. Sánchez-Carrillo, S., Angeler, D. G., Álvarez-Cobelas, M., & Sánchez-Andrés, R. (2010). Freshwater wetland eutrophication. In Eutrophication: causes, consequences, and control (pp. 195-210). Springer Netherlands. Salihoglu, G., & Karaer, F. (2004). Ecological risk assessment and problem formulation for Lake Uluabat, a Ramsar State in Turkey. Environmental Management, 33(6), 899-910. Serveiss, V. B., Dow, D., & Valiela, I. (2004). Using ecological risk assessment to identify the major anthropogenic stressor in the Waquoit Bay Watershed, Cape Cod, Massachusetts. Environmental management, 33(5), 730-740. 85 Singh, K. P., Mohan, D., Singh, V. K., & Malik, A. (2005). Studies on distribution and fractionation of heavy metals in Gomti river sediments—a tributary of the Ganges, India. Journal of hydrology, 312(1-4), 14-27. Somlyódy, L., & van Straten, G. (Eds.). (2012). Modeling and managing shallow lake eutrophication: with application to Lake Balaton. Springer Science & Business Media. Smith, V. H. (2003). Eutrophication of freshwater and coastal marine ecosystems a global problem. Environmental Science and Pollution Research, 10(2), 126-139. Sundaray, S. K., Nayak, B. B., Lin, S., & Bhatta, D. (2011). Geochemical speciation and risk assessment of heavy metals in the river estuarine sediments—a case study: Mahanadi basin, India. Journal of hazardous materials, 186(2-3), 1837-1846. Suter II, G. W. (2016). Ecological risk assessment. CRC press. Tang, X, J., & Li, X, M. (2016). Application of Comprehensive Nutrition Level Index Method in Eutrophication Evaluation of Water. Stochastic Environmental Research and Risk Assessment, 22(4), 128-129. Tania, M., Radu, M., Dan, V., Rodica, V., & Mihnea, M. (2013). Water quality assessment of the nadas river (romania) in terms of NSF water quality index. Analele Univ. din Oradea, 21, 649-654. Tyagi, S., Sharma, B., Singh, P., & Dobhal, R. (2013). Water quality assessment in terms of water quality index. American Journal of Water Resources, 1(3), 34-38. Uusitalo, L. (2007). Advantages and challenges of Bayesian networks in environmental modelling. Ecological modelling, 203(3), 312-318. 86 Van Dam, R. A., Finlayson, C. M., & Humphrey, C. L. (1999). Wetland risk assessment. Techniques for enhanced wetland inventory and monitoring, edited by Finlayson, CM and Spiers, AG, Supervising Scientist Report, 147, 83-118. Van Dam, R. A., Camilleri, C., Turley, C. J., & Markvich, S. J. (2001). Ecological risk assessment of tebuthiuron following application on northern Australian wetlands. Environment Research Institute of the Supervising Scientist (eriss). Varol, M. (2011). Assessment of heavy metal contamination in sediments of the Tigris River (Turkey) using pollution indices and multivariate statistical techniques. Journal of Hazardous Materials, 195, 355-364. Walden, D., van Dam, R., Finlayson, M., Storrs, M., Lowry, J., & Kriticos, D. (2004). A risk assessment of the tropical wetland weed Mimosa pigra in northern Australia. Supervising Scientist Report, 177. Wang. (2012). Ecological Risk Assessment and Research Progress. AMBIO: A Journal of the Human Environment, 36(4), 335-342. Wagh, V., Panaskar, D., Muley, A., Mukate, S., & Gaikwad, S. (2018). Neural network modelling for nitrate concentration in groundwater of Kadava River basin, Nashik, Maharashtra, India. Groundwater for Sustainable Development. Wang, X., Bai, S., Lu, X., Li, Q., Zhang, X., & Yu, L. (2008). Ecological risk assessment of eutrophication in Songhua Lake, China. Stochastic Environmental Research and Risk Assessment, 22(4), 477-486. Wilcock, D. N. (1999). River and inland water environments. In Environmental Management in Practice: Vol 3 (pp. 61-84). Routledge. 87 Xiao, R., Bai, J., Huang, L., Zhang, H., Cui, B., & Liu, X. (2013). Distribution and pollution, toxicity, and risk assessment of heavy metals in sediments from urban and rural rivers of the Pearl River delta in southern China. Ecotoxicology, 22(10), 1564-1575. Xu, X., Lin, H., & Fu, Z. (2004). Probe into the method of regional ecological risk assessment—a case study of wetland in the Yellow River Delta in China. Journal of Environmental Management, 70(3), 253-262. Xu, F. L., Tao, S., Dawson, R. W., & Li, B. G. (2001). A GIS-based method of lake eutrophication assessment. Ecological Modelling, 144(2), 231-244. Xu, Y., Peng, H., Yang, Y., Zhang, W., & Wang, S. (2014). A cumulative eutrophication risk evaluation method based on a bioaccumulation model. Ecological modelling, 289, 7785. Yang, Z., Wang, Y., Shen, Z., Niu, J., & Tang, Z. (2009). Distribution and speciation of heavy metals in sediments from the mainstream, tributaries, and lakes of the Yangtze River catchment of Wuhan, China. Journal of Hazardous Materials, 166(2-3), 11861194. Yongming, H., Peixuan, D., Junji, C., & Posmentier, E. S. (2006). Multivariate analysis of heavy metal contamination in urban dusts of Xi'an, Central China. Science of the total environment, 355(1-3), 176-186. Yousefi, H., Zahedi, S., & Niksokhan, M. H. (2018). Modifying the analysis made by water quality index using multi-criteria decision-making methods. Journal of African Earth Sciences, 138, 309-318. Yu, F. C., Fang, G. H., & Ru, X. W. (2010). Eutrophication, health risk assessment and spatial analysis of water quality in Gucheng Lake, China. Environmental Earth Sciences, 59(8), 1741-1748. 88 Yu, F. C., Fang, G. H., & Ru, X. W. (2010). Eutrophication, health risk assessment and spatial analysis of water quality in Gucheng Lake, China. Environmental Earth Sciences, 59(8), 1741-1748. Zhang (2004). Wetland Risk Assessment Method and Approach. Science China Earth Sciences, 53(7), 1036-1042. Zhang, R., Liang, S., Ou, M., & Xiong, Q. (2018, August). Evaluation of Water Quality for Mangrove Ecosystem Using Artificial Neural Networks. In 2018 International Conference on Advanced Mechatronic Systems (ICAMechS) (pp. 257-261). ZHU, H. N., YUAN, X. Z., ZENG, G. M., JIANG, M., LIANG, J., ZHANG, C., & JIANG, H. W. (2012). Ecological risk assessment of heavy metals in sediments of Xiawan Port based on modified potential ecological risk index. Transactions of Nonferrous Metals Society of China, 22(6), 1470-1477. 89