Environmental Risk Assessment of Wetland Ecosystems Using Bayesian Belief Networks

: Wetlands are valuable natural capital and sensitive ecosystems facing signiﬁcant risks from anthropogenic and climatic stressors. An assessment of the environmental risk levels for wetlands’ dynamic ecosystems can provide a better understanding of their current ecosystem health and functions. Different levels of environmental risk are deﬁned by considering the categories of risk and the probability and severity of each in the environment. Determining environmental risk levels provides a general overview of ecosystem function. This mechanism increases the visibility of risk levels and their values in three distinct states (i.e., low, moderate, and high) associated with ecosystem function. The Bayesian belief network (BBN) is a novel tool for determining environmental risk levels and monitoring the effectiveness of environmental planning and management measures in reducing the levels of risk. This study develops a robust methodological framework for determining the overall level of risks based on a combination of varied environmental risk factors using the BBN model. The proposed model is adopted for a case study of Shadegan International Wetlands (SIWs), which consist of a series of Ramsar wetlands in the southwest of Iran with international ecological signiﬁcance. A comprehensive list of parameters and variables contributing to the environmental risk for the wetlands and their relationships were identiﬁed through a review of literature and expert judgment to develop an inﬂuence diagram. The BBN model is adopted for the case study location by determining the states of variables in the network and ﬁlling the probability distribution tables. The environmental risk levels for the SIWs are determined based on the results obtained at the output node of the BBN. A sensitivity analysis is performed for the BBN model. We proposed model-informed management strategies for wetland risk control. According to the BBN model results, the SIWs ecosystems are under threat from a high level of environmental risk. Prolonged drought has been identiﬁed as the primary contributor to the SIWs’ environmental risk levels.


Introduction
Wetlands provide key environmental benefits including water purification and pollution removal, flood protection, shoreline stabilization, groundwater recharge, stream flow maintenance, and ecological enhancement [1,2]. Vegetation, soils, and the microbial communities in wetlands facilitate the treatment of water and wastewater run-off. The hydraulics and hydrological processes in wetlands offer a low-emission and sustainable alternative to traditional wastewater treatment plants and can accelerate the natural biodegradation of pollutants [3]. However, anthropogenic activities, urbanization, and This study aims to develop a probabilistic decision support model using BBNs to identify, quantify, and analyze the environmental risk levels of wetlands. The proposed model is adopted for the case study of the Shadegan International Wetlands (SIWs), located in southwestern Iran, which is a Ramsar site of significant ecological value. Key environmental variables and the relationships amongst them are identified through the expert judgment of five environmental specialists, and an influence diagram (ID) is developed. The BBNs are developed through determining the states and probability distribution tables, and environmental management strategies are proposed to reduce the risk levels for the SIWs.

Study Area
The Shadegan International Wetlands (SIWs) are located in Khuzestan Province, southwest of Iran, between 48 • 20 and 49 • 20 E longitude and 30 • 50 and 31 • 00 N latitude. Figure 1 shows the geographical location and characteristics of the study area and the wetland catchment. SIWs are located in the Jarahi River Delta, with very flat land and a low-gradient plains' topography. The Jarahi basin has an area of approximately 24,310 km 2 and is in the southern parts of the Zagros Mountain Range. The SIWs cover about 5377 km 2 , of which almost 61% is protected as a wildlife refuge [35].
Hydrology 2023, 9, x FOR PEER REVIEW 3 of 15 insufficient and uncertain data. The applications of BN's framework in participatory modeling have been successfully tested by Borsuk et al. [33] and Jensen [34]. This study aims to develop a probabilistic decision support model using BBNs to identify, quantify, and analyze the environmental risk levels of wetlands. The proposed model is adopted for the case study of the Shadegan International Wetlands (SIWs), located in southwestern Iran, which is a Ramsar site of significant ecological value. Key environmental variables and the relationships amongst them are identified through the expert judgment of five environmental specialists, and an influence diagram (ID) is developed. The BBNs are developed through determining the states and probability distribution tables, and environmental management strategies are proposed to reduce the risk levels for the SIWs.

Study Area
The Shadegan International Wetlands (SIWs) are located in Khuzestan Province, southwest of Iran, between 48°20′ and 49°20′ E longitude and 30°50′ and 31°00′ N latitude. Figure 1 shows the geographical location and characteristics of the study area and the wetland catchment. SIWs are located in the Jarahi River Delta, with very flat land and a low-gradient plains' topography. The Jarahi basin has an area of approximately 24,310 km 2 and is in the southern parts of the Zagros Mountain Range. The SIWs cover about 5377 km 2 , of which almost 61% is protected as a wildlife refuge [35]. The SIWs have important hydrological, biological, and ecological significance as a Ramsar site and in terms of maintaining normal ecosystem functions in the surrounding river basin and coastal system. The unique diversity of this wetland includes plant and animal species specific to freshwater, brackish, and saltwater environments. SIWs are listed in the Montreux Record [36], and currently, they are under significant existential threat from several environmental risks. The recent surveys of SIWs show that despite unique and outstanding ecological values and ecosystem function, the wetlands are much degraded and removed from their natural ecological condition due to anthropogenic The SIWs have important hydrological, biological, and ecological significance as a Ramsar site and in terms of maintaining normal ecosystem functions in the surrounding river basin and coastal system. The unique diversity of this wetland includes plant and animal species specific to freshwater, brackish, and saltwater environments. SIWs are listed in the Montreux Record [36], and currently, they are under significant existential threat from several environmental risks. The recent surveys of SIWs show that despite unique and outstanding ecological values and ecosystem function, the wetlands are much degraded and removed from their natural ecological condition due to anthropogenic activities [20]. The expert team consulted in this research identified the most important risk factors that SIWs have faced in recent years, including land use change, water pollution, uncontrolled exploitation, sedimentation, loss of biodiversity, drought, and changes in water flow regimes. The following are brief discussions of the underlying mechanisms that introduced these risks: activities [20]. The expert team consulted in this research identified the most important risk factors that SIWs have faced in recent years, including land use change, water pollution, uncontrolled exploitation, sedimentation, loss of biodiversity, drought, and changes in water flow regimes. The following are brief discussions of the underlying mechanisms that introduced these risks: Reduction of discharge and changes in water flow regime: development of surface water and flood control plans (i.e., storage dams) in the upstream area and extensive irrigation and drainage plans for the upstream lands have led to a reduction of water entering from the main artery of the wetland (the Jarhari River). The flow reduction of the Karun River and the inflow of drainage from the sugarcane development projects, with their high salinity and large volume, have changed the nature of the ecosystem across the freshwater part of the wetland and facilitated salinization and drying [37,38]. activities [20]. The expert team consulted in this research identified the most important risk factors that SIWs have faced in recent years, including land use change, water pollution, uncontrolled exploitation, sedimentation, loss of biodiversity, drought, and changes in water flow regimes. The following are brief discussions of the underlying mechanisms that introduced these risks: ▪ Sedimentation: The sedimentation and filling of the wetland by the sediments that supply the wetland's water; the construction of dams upstream, which has increased sedimentation and sediment accumulation [35].
▪ Loss of biodiversity: Fish breeding activities in the upstream part of the catchment area have led to the introduction of non-native species that compete with the native species of the wetland and have caused biodiversity issues. Road construction inside the wetland, hunting, and the development of fishing ports are also threatening biodiversity at the wetland [17,37]. ▪ Sedimentation: The sedimentation and filling of the wetland by the sediments that supply the wetland's water; the construction of dams upstream, which has increased sedimentation and sediment accumulation [35].
▪ Loss of biodiversity: Fish breeding activities in the upstream part of the catchment area have led to the introduction of non-native species that compete with the native species of the wetland and have caused biodiversity issues. Road construction inside the wetland, hunting, and the development of fishing ports are also threatening biodiversity at the wetland [17,37]. Reduction of discharge and changes in water flow regime: development of surface water and flood control plans (i.e., storage dams) in the upstream area and extensive irrigation and drainage plans for the upstream lands have led to a reduction of water entering from the main artery of the wetland (the Jarhari River). The flow reduction of the Karun River and the inflow of drainage from the sugarcane development projects, with their high salinity and large volume, have changed the nature of the ecosystem across the freshwater part of the wetland and facilitated salinization and drying [37,38].

Modeling and Analysis
Bayesian Belief Networks (BBNs) are a class of probabilistic graphical models capable of describing conditional dependencies between a set of variables. BBNs use directed acyclic graphs to describe the modeling outcomes. A BBN is a graph consisting of nodes (variables) and arcs. Nodes represent discrete random variables, and arcs depict conditional relationships between variables. Variables that depend on other variables are referred to as "child nodes," while nodes directly preceding them are called "parent nodes." Two variables are independent if they are not directly linked by an arc. The quantitative component is a set of probability distributions that quantifies the strength of the conditional dependencies between variables represented in the directed acyclic graph [30,38]. In general, BBNs are useful for situations in which the current state of the system depends on its previous state. Using BBNs, the posterior probabilities of the output variables can be determined [39,40]. Hence, BBNs, as probabilistic networks, can be utilized for decisionmaking in environmental management problems where uncertainty in variables can be challenging [41].
This study adopts the BBNs to determine the risk levels of wetland ecosystems. At the beginning of the modeling process, an influence diagram (ID) was developed by selecting variables and determining the causal relationships between them. Selection variables, variables' states, and relations between nodes were identified through expert judgments (see §3 Results for further discussions).
Three types of nodes are introduced in the network. Input nodes are variables related to wetland conditions and risks. Intermediate nodes are related to variables describing the probability of hazards and consequences, which are the main indicators in determining risk levels. The output node specifies the environmental risk levels of wetlands. Figure 2 shows the schematic of the methodological framework adopted in this study to determine wetland risk levels. The BBNs model proposed in this study is developed using Netica open-source code. variables can strongly influence the behavior of the system and which variables are not very sensitive to changes in the system. By selecting each state of a selected node, the model assumes 100% probability for the state and estimates the consequent probabilities for other nodes in the network. This feature was used in two forms: "predictive model" and "diagnostic model" [46,47]. In the "diagnostic model," by selecting a state of one or more "child nodes" at 100%, changes are observed in the "parent nodes" of the probability distribution. In the "predictive model," the state of one or more "parent nodes" is altered, and changes in probability distributions in "child nodes" are examined [46].

Start
Step 1 Identifying the environment and current status of the wetland and choosing risk factors Step 2

Proposing influence diagram in BBN
Step 3 Developing primary structure of BBN Step 4 Completing CPTs and developing final structure of BBN Step 5 Determining risk levels Step 6 Step 7 End Sensitivity analysis Offering control measures in order reduce the risk levels Figure 2. A methodological framework adopted for determining the risk levels of wetland ecosystems.

Results and Discussion
To determine the levels of environmental risk at the SIWs, initially, the ID of causal relationships was formed in the model. The nodes and arcs between the parameters were selected based on a literature review, the judgments of experts, and the characteristics of the study area. Land use change, water pollution, uncontrolled exploitation, sedimentation, loss of biodiversity, drought, and changing water regimes are identified as the key risk factors for SIWs conservation. Table 1 describes a comprehensive list of the Following the identification of the list of variables influencing risk levels at SIWs, the primary structure of BBNs is developed by determining the causal relationships amongst the influential parameters and the setting states of the variables in the network. To com-plete the network, the combined possibilities of states are entered for each node in the form of conditional probability tables (CPTs). Interdependencies between the nodes are qualitatively modeled by the links in the network and quantitatively by the CPTs associated with each node. The joint probability distribution captured by the network's structure and CPTs encodes the domain expert's knowledge of interdependencies among variables [42]. The associated uncertainty of the system can be examined using CPTs [43].
The development of BBNs, including the selection of nodes, states, relationships between variables, and values of CPTs, is completed with a combination of expert judgments, results from simulation models, literature reviews, and analysis of real data obtained from the case study area. In Bayesian modeling, the existing scientific information in different formats as well as the experts' judgments are valuable sources of information that can be combined to enhance the output results [44,45]. The ability of BBNs to integrate multiple data types into a model is particularly useful for wetlands conservation and the development of robust decision support tools for environmental management. The environmental risk levels are determined in the form of a probability distribution at the output node of the BBNs. The results of the output node (i.e., risk levels) are used to devise management strategies to control and mitigate the risk factors according to the characteristics of the study area (i.e., SIWs) and based on the existing literature.
Sensitivity analysis is carried out to examine the model's response to variations in the network inputs and quantify the effects of changes in one variable on other variables. In BNs, sensitivity means that changes in the probabilities of the considered nodes are a function of changes in the values of the conditional probability tables of the network. The sensitivity analysis of the proposed Bayesian model provides information on which variables can strongly influence the behavior of the system and which variables are not very sensitive to changes in the system. By selecting each state of a selected node, the model assumes 100% probability for the state and estimates the consequent probabilities for other nodes in the network. This feature was used in two forms: "predictive model" and "diagnostic model" [46,47]. In the "diagnostic model", by selecting a state of one or more "child nodes" at 100%, changes are observed in the "parent nodes" of the probability distribution. In the "predictive mode", the state of one or more "parent nodes" is altered, and changes in probability distributions in "child nodes" are examined [46].

Results and Discussion
To determine the levels of environmental risk at the SIWs, initially, the ID of causal relationships was formed in the model. The nodes and arcs between the parameters were selected based on a literature review, the judgments of experts, and the characteristics of the study area. Land use change, water pollution, uncontrolled exploitation, sedimentation, loss of biodiversity, drought, and changing water regimes are identified as the key risk factors for SIWs conservation. Table 1 describes a comprehensive list of the characteristics of the risk factors considered in this study. The probability of hazard and consequence were selected as the main indicators in determining the environmental risk levels. The ID used to determine the levels of environmental risk for the SIWs is shown in Figure 3. Table 1. Characteristics of the risk factors in SIWs.

Risk Factor Adverse Environmental Effects Receivers
Land use change -Reduces groundwater recharge -Increases evapotranspiration -Increases concentration of inorganic compounds All organisms in the soil and aquatic life in wetland Table 1. Cont.

Risk Factor Adverse Environmental Effects Receivers
Water pollution -Impact on the interaction between Biotop and Biocenose - Creating an unfavorable view - The loss or migration of species with low ecological valence All organisms associated with wetland

Risk Factor Adverse Environmental Effects Receivers
Land use change -Reduces groundwater recharge -Increases evapotranspiration -Increases concentration of inorganic compounds All organisms in the soil and aquatic life in wetland Following the completion of the ID setup, the states of each node were determined to form the initial BBN model. A range of states were considered for each node based on the data from expert judgments and the literature. Figure 4 illustrates the initial BBN developed to determine the levels of environmental risks at the SIWs. Once the network structure was completed, the probabilities of risk factors were entered into the network in the form of CPT codes for each node in the network. The quantitative relations between variables were modeled by the CPT associated with each node. The child node was assigned a probability for each possible combination of each parent state. The probability percentage values considered in the model were estimated based on expert judgments and background field study data from the SIWs. The BBN model developed for this study considers the states at each node. For instance, in the risk level node, three states (i.e., low, moderate, and high) are specified. The CPT of the risk level node is shown in Figure 5. Following the completion of the ID setup, the states of each node were determined to form the initial BBN model. A range of states were considered for each node based on the data from expert judgments and the literature. Figure 4 illustrates the initial BBN developed to determine the levels of environmental risks at the SIWs. Once the network structure was completed, the probabilities of risk factors were entered into the network in the form of CPT codes for each node in the network. The quantitative relations between variables were modeled by the CPT associated with each node. The child node was assigned a probability for each possible combination of each parent state. The probability percentage values considered in the model were estimated based on expert judgments and background field study data from the SIWs. The BBN model developed for this study considers the states at each node. For instance, in the risk level node, three states (i.e., low, moderate, and high) are specified. The CPT of the risk level node is shown in Figure 5.  The risk level node is influenced by two nodes, including the probability of hazards and consequences. By combining the states of the nodes in BBN, several scenarios for the risk level node are considered. The probability distribution obtained from the output node (risk level) indicates that the risk level is very high for the SIWs (Figure 6). The results show that the risk levels for the SIWs are about 10%, 17%, and 73%, respectively, for the Following the completion of the ID setup, the states of each node were determined to form the initial BBN model. A range of states were considered for each node based on the data from expert judgments and the literature. Figure 4 illustrates the initial BBN developed to determine the levels of environmental risks at the SIWs. Once the network structure was completed, the probabilities of risk factors were entered into the network in the form of CPT codes for each node in the network. The quantitative relations between variables were modeled by the CPT associated with each node. The child node was assigned a probability for each possible combination of each parent state. The probability percentage values considered in the model were estimated based on expert judgments and background field study data from the SIWs. The BBN model developed for this study considers the states at each node. For instance, in the risk level node, three states (i.e., low, moderate, and high) are specified. The CPT of the risk level node is shown in Figure 5.  The risk level node is influenced by two nodes, including the probability of hazards and consequences. By combining the states of the nodes in BBN, several scenarios for the risk level node are considered. The probability distribution obtained from the output node (risk level) indicates that the risk level is very high for the SIWs (Figure 6). The results show that the risk levels for the SIWs are about 10%, 17%, and 73%, respectively, for the The risk level node is influenced by two nodes, including the probability of hazards and consequences. By combining the states of the nodes in BBN, several scenarios for the risk level node are considered. The probability distribution obtained from the output node (risk level) indicates that the risk level is very high for the SIWs (Figure 6). The results show that the risk levels for the SIWs are about 10%, 17%, and 73%, respectively, for the low, moderate, and high states ( Figure 6). The results shown in Figure 6 highlight the drought as the highest (or fastest) state of risk for the SIWs, with 75%, followed by sedimentation (60%), loss of biodiversity (60%), changing water regimes (55%), uncontrolled exploitation (38%), land use change (37.5%), and water pollution (17.8%).
Hydrology 2023, 9, x FOR PEER REVIEW 9 of 15 low, moderate, and high states ( Figure 6). The results shown in Figure 6 highlight the drought as the highest (or fastest) state of risk for the SIWs, with 75%, followed by sedimentation (60%), loss of biodiversity (60%), changing water regimes (55%), uncontrolled exploitation (38%), land use change (37.5%), and water pollution (17.8%). To understand the sensitivity of the BBN in the "predictive model," each risk factor was assumed to be 100% at the low (i.e., slow) state; it was expected that the low states' probability of hazards and consequences, as the child nodes, would be significantly reduced. Figure 7 shows the predictive model determined for the sensitivity analysis of the environmental risk levels at the SIWs. The results show that when the low state values in risk factors were changed to 100%, the two low state values of the parent nodes (probability of hazard and consequence) were altered to 93% and 97%, respectively. Also, with a significant reduction in the probability of hazards and consequences, the risk level in the low state changed to 90.4%.  To understand the sensitivity of the BBN in the "predictive model," each risk factor was assumed to be 100% at the low (i.e., slow) state; it was expected that the low states' probability of hazards and consequences, as the child nodes, would be significantly reduced. Figure 7 shows the predictive model determined for the sensitivity analysis of the environmental risk levels at the SIWs. The results show that when the low state values in risk factors were changed to 100%, the two low state values of the parent nodes (probability of hazard and consequence) were altered to 93% and 97%, respectively. Also, with a significant reduction in the probability of hazards and consequences, the risk level in the low state changed to 90.4%.
Hydrology 2023, 9, x FOR PEER REVIEW 9 of 15 low, moderate, and high states ( Figure 6). The results shown in Figure 6 highlight the drought as the highest (or fastest) state of risk for the SIWs, with 75%, followed by sedimentation (60%), loss of biodiversity (60%), changing water regimes (55%), uncontrolled exploitation (38%), land use change (37.5%), and water pollution (17.8%). To understand the sensitivity of the BBN in the "predictive model," each risk factor was assumed to be 100% at the low (i.e., slow) state; it was expected that the low states' probability of hazards and consequences, as the child nodes, would be significantly reduced. Figure 7 shows the predictive model determined for the sensitivity analysis of the environmental risk levels at the SIWs. The results show that when the low state values in risk factors were changed to 100%, the two low state values of the parent nodes (probability of hazard and consequence) were altered to 93% and 97%, respectively. Also, with a significant reduction in the probability of hazards and consequences, the risk level in the low state changed to 90.4%.   For this model, in the risk level node, the high state was assumed to be 100%, so the probability of hazards and consequences in the high state were increased to 93.6% and 77.4%, respectively. According to the results of the sensitivity analysis shown in Table 2, the probability of hazard has a greater impact on the level of risk. The results highlight that, among the risk factors, drought has a greater effect on the environmental risk levels of the SIWs. Figure 8 presents the diagnostic model for determining the environmental risk levels of the SIWs. For this model, in the risk level node, the high state was assumed to be 100%, so the probability of hazards and consequences in the high state were increased to 93.6% and 77.4%, respectively. According to the results of the sensitivity analysis shown in Table  2, the probability of hazard has a greater impact on the level of risk. The results highlight that, among the risk factors, drought has a greater effect on the environmental risk levels of the SIWs.  We propose management strategies to reduce risk levels and manage the environmental risks of the SIWs based on the results gathered from BBNs, the characteristics of the study area, existing data, and expert opinions (Table 3). This research highlights the significant role of prolonged droughts as a climatic stressor that adversely influences the environmental preservation of SIWs. As such, it is proposed that harvesting water from the SIWs should be banned, specifically during the drought forecast period. To limit the environmental damage caused by drought, changes in flow regimes, such as construction of hydraulic structures (e.g., dams, control structures), must be avoided across the SIWs. A robust hydrological and drought monitoring network within Jarahi Basin is necessary to improve management and planning for the SIWs' preservation.  We propose management strategies to reduce risk levels and manage the environmental risks of the SIWs based on the results gathered from BBNs, the characteristics of the study area, existing data, and expert opinions (Table 3). This research highlights the significant role of prolonged droughts as a climatic stressor that adversely influences the environmental preservation of SIWs. As such, it is proposed that harvesting water from the SIWs should be banned, specifically during the drought forecast period. To limit the environmental damage caused by drought, changes in flow regimes, such as construction of hydraulic structures (e.g., dams, control structures), must be avoided across the SIWs. A robust hydrological and drought monitoring network within Jarahi Basin is necessary to improve management and planning for the SIWs' preservation. Furthermore, water transfer into wetlands from areas with high (or higher) water levels during drought is recommended to reduce the environmental risk levels imposed by drought. BBNs are a robust and effective probabilistic modeling technique for determining the levels of environmental risks in different ecosystems and assessing the potential outcomes of alternative management actions. The capabilities of BBNs in providing the possibility of combining different types of data, evaluating modeling uncertainty, and updating the model when new knowledge and data become available make BBNs an ideal tool for examining risk factors across complex environmental domains. As such, BBNs can represent multiple environmental risk factors in complicated wetland systems as a whole without the need to fully capture and illustrate all the underlying processes in the system. The associated uncertainty in the environmental risk factors of a wetland ecosystem can be indicated and quantified using CPTs.
To protect and manage wetlands in a sustainable manner, it is necessary to quantify and reduce levels of environmental risk that impact the wetlands' health and function. In this study, a BBN model was developed to determine the relative levels of environmental risk, which were defined by considering the categories of probability and severity. To standardize the risk levels and their value, three distinct states, including low, moderate, and high, were adopted for BBN. This study utilized BBN to model environmental risk levels for a case study of SIWs, which is a Ramsar site of high environmental and conservation significance. The BBN results show a high level of risk (73% at high state) for SIWs, emphasizing the critical need for SIW risk management strategies. The analysis of risk levels and the available survey data for SIWs show that without robust interventions, the environmental condition and ecological health of the wetlands will worsen. Changes in land use and the increased development plans across the SIWs area are key parameters affecting the wetland's health and function.
The BBN model developed in this study determined drought as the main risk factor influencing the SIWs ecosystem. Prolonged drought in Jarahi's Basin has been a chronic problem in recent decades, which will be exacerbated by the long-term effects of a changing climate. Previously, Malekmohammadi and Rahimi Blouchi [17] conducted an environ-mental risk assessment for Shadgan wetland and concluded that the most important risk was the construction of a dam and irrigation network. Given that Iran and especially the southern regions of the country have faced prolonged drought and low rainfall in recent years, which have led to the drying up of a significant part of the Shadegan wetland, this study identified drought as the most important risk. Preventing the harvest of water from the SIWs, designing a hydrological and drought monitoring network in the Jarahi Basin, and preventing deforestation and erosion upstream of the SIWs are proposed as the key control measures for reducing the level of drought risk.
Considering the close relationship between the biotope and the biogenesis in wetland ecosystems and the interconnectivity between the upstream and downstream interacting processes of wetlands, a catchment-scale environmental management approach is the most appropriate way to sustainably manage wetlands' dynamic ecosystems. Monitoring risk levels in wetland ecosystems with BBNs aids in the implementation of ecosystem-based environmental management strategies. Implementing risk management strategies can help improve the current conditions at wetlands, reduce risk levels, and prevent further degradation of these valuable natural capitals.
In developing BBNs for assessing the environmental risk levels of wetland ecosystems, consideration must be given to identifying the most influential variables and their interdependencies for designing and refining the network and constructing CPTs in the model. The developed BBNs can be considered a decision-support tool that helps devise robust environmental management protocols. New variables (i.e., new risk factors) and variables that affect the existing risk factors can be incorporated into BBNs' model, and the temporal and spatial changes in risk levels can be monitored. Also, the rate of decrease in risk factors and related risk levels can be monitored and evaluated by BBNs when specific management strategies are implemented. Integration of BBNs with conventional methods for environmental risk assessment (e.g., multi-criteria decision-making method) can provide a more comprehensive decision-support tool for monitoring and managing environmental risk factors in wetland ecosystems.
In this study, risk factors are considered independent of each other. However, the proposed model can also consider inter-relationships among the risk factors if a comprehensive dataset on the influential risk parameters exists for the case study location. The proposed BBN model can be adopted as a robust auxiliary tool in determining the status of each risk factor affecting the wetland and determining the final level of wetland risk. The prioritization of different risk factors for environmental management and decision-making will vary as a function of time, budget, and study area characteristics.
The lack of suitable quantitative data for risk factors, and especially for determining the relationships between the variables, is one of the key challenges that environmental risk assessment studies face. As such, it is necessary to provide a set of standardized protocols and rules that can serve as a step-by-step guide for the environmental risk assessments and to specify the influential variables and the relationship between them in the design and refinement of the BBN model. Also, the addition of other advanced statistical models that can be combined with Bayesian networks can help to provide more reliable outputs (e.g., ANN [48,49], GP [50]). For example, using a full probability density function (PDF) for characterizing the risk factors can enhance the quality and reliability of the modeling results. Further research is needed to develop a more holistic environmental risk assessment framework for wetlands that includes both qualitative and quantitative risk assessment approaches.

Conclusions
This study proposed a robust methodological approach to determine the overall environmental and conservation risk level for wetlands using BBNs. The risk factors that influence a wetland's health and function and the relationships between these variables can vary in the proposed model depending on the characteristics of the study area, data from field-based observations, or expert opinions. The complexity in setting up relationships between risk variables is a function of the availability of data, the purpose of the assessment, and the resources allocated for the study. In this study, following the literature review and surveys and examining the study area, the expert team identified the most important risk factors for the SIWs, including land use change, water pollution, uncontrolled exploitation, sedimentation, loss of biodiversity, drought, and changing water regimes. The results show prolonged drought in the Jarahi basin as a key risk factor threatening the ecological function and health of SIWs. Catchment scale management strategies are proposed to reduce environmental risk levels for SIWs.