Introduction

The coastal environment is regarded as a natural and ecotonic environment between the land and sea, which is affected by the ecological interactions of the two ecosystems and the functions of the two environments (Craig & Ruhl, 2010). Coastal areas are considered as the ground for major economic and social activity in the world (Ricketts & Harrison, 2007). However, Coastal areas are extremely vital and vulnerable ecologically to various stresses (Mani Murali et al., 2013). These areas are always undergoing extensive changes due to natural hazards, such as heavy rainfall, waves, climate change, coastal erosion, sea-level change, etc. (Andalecio, 2010; Tahri et al., 2017). Furthermore, human activities such as increasing population density, industrialization, and tourism in coastal areas are important factors that increase the pressure and intensity of exploitation in coastal areas (Nouri et al., 2008; Tang et al., 2018). Therefore, coastal environmental protection is considered as a management requirement to realize the principles of sustainable development considering the nature of the permanent threat to coastal ecosystems (Mani Murali et al., 2013).

Accordingly, ecological vulnerability (EV) identification is regarded as important to provide a sustainable management system with effective solutions for disaster management and prevention of further degradation (Doukakis, 2005; Sankari et al., 2015; Zhang et al., 2017). The main objective of EV research is to maintain the balance between conservation and exploitation, which is required for the sustainability of an ecosystem by identifying the factors affecting vulnerable areas. Thus, determining the boundaries of ecologically vulnerable areas is a necessary aspect of environmental protection management (He et al., 2018). EV is a function of pressure, sensitivity, and adaptive capacity, and its estimation is required for environmental protection modeling (Qiao et al., 2013). In general, a vulnerable area can be defined as an area that is both sensitive and exposed to human pressures, and the degree of vulnerability depends on the degree of sensitivity and the corresponding pressure (Bigot et al., 2000). Pressure and sensitivity factors in these areas are human activities and accelerated exploitation, especially destroying natural ecosystems and environmental pollution in recent years (Jafari, 2010; Jamshidi & Yousefi, 2013). Vulnerability mainly involves the three dimensions of exposure, sensitivity or susceptibility to loss, and the capacity of a system to adapt to changes and events (Adger, 2006; Füssel 2006; Field & Barros, 2014).

The ecological vulnerability index (EVI) has recently been developed for environmental assessment. This concept has been studied and used in different spatial scales and several regions (Fandohan et al., 2015; Jha & Gundimeda, 2019; Wang et al., 2019). Most of the studies conducted on coastal vulnerability have been based on the coastal vulnerability index (CVI) (Doukakis, 2005; Mani Murali et al., 2013; Sankari et al., 2015; Zhang et al., 2017). CVI is a relatively simple method of determining vulnerability including six variables related to physical processes about geological parameters, wave characteristic height, and changes in sea level, derived from the product of sequential variables divided by the total number of variables. This method only expresses the vulnerability of the coastline to physical and geological changes, where it is impossible to rank among the criteria based on expert opinions (Bagdanavičiūtė et al., 2015; Koroglu et al., 2019). Thus, this study aims to investigate the ecological and economic criteria and their different importance in the vulnerability assessment of coastal areas based on multi-criteria decision-making methods (MCDM) in the geographic information system (GIS) environment (Mani Murali et al., 2013; Tahri et al., 2017).

GIS-based MCDM is among the universally accepted approaches to help decision-makers to assess and map coastline vulnerabilities and threats based on the system of evaluating criteria and options (Andalecio, 2010; Lima et al., 2019; Tahri et al., 2017). So far, few studies have exploited GIS multi-criteria evaluation methods to assess the coastline vulnerability and risks. Cozannet et al. (2013) investigated the physical vulnerability of the Languedoc-Roussillon coastline in the northwest of the Mediterranean coast in France by the analytic hierarchy process (AHP) method. Tahri et al. (2017) used the FAHP method with seven ecological and economic factors to determine vulnerable areas on the coast of Morocco. They described the eastern coasts of the region as high-risk areas based on the vulnerability map in three levels as low, medium, and high. Further, Wu et al. (2018) prepared the EV map by FAHP and 21 indicators to determine such vulnerability in the Yellow River Delta in China.

Iran has approximately 5000 km of coastline along the Indian Ocean including the Oman Sea and the Persian Gulf, and more than 750 km of coastline adjacent to the Caspian Sea (Dibajnia et al., 2012; Pak & Farajzadeh, 2007). The Caspian Sea has an area of 371,000 km2 and is considered as the largest enclosed one in the world with huge energy resources, unique ecosystem features, and special geopolitical and social conditions (Fazli & Taghavi Jelodar, 2013; Mirzajani et al., 2015). Unfortunately, human activities and natural factors have posed widespread threats to the Caspian coastal areas in recent years. The most important of these threats are sea-level changes, biodiversity reduction, land use changes, improper exploitation of natural resources, overfishing, coastal erosion, and reduction of coastal water quality (Mirzajani et al., 2015). Each of these factors has a different impact on coastal vulnerability. Therefore, zoning coastal vulnerabilities is necessary to mitigate the effects of threats.

On the other hand, these zonings should provide appropriate solutions in line with the goals of sustainable development. Thus, recognizing and evaluating changes in the environment and the threatening factors of ecosystems is a process that leads to an appropriate understanding of how humans and the environment interact in environmental management studies. This is even more critical in biologically sensitive areas (Lambin & Geist, 2006). As a result, using environmental risk assessment methods is one of the significant tools in studying environmental management, as well as identifying and reducing the potential causes of environmental harm in wetlands to achieve sustainable development (Makvandi et al., 2016). Environmental risk assessment evaluates the environmental risks associated with human activities and natural disasters and the appropriate level of risk management measures to reduce the risks and their adverse effects to an acceptable level of risk. Environmental risk assessment is an essential component and one of the appropriate technical methods in environmental assessment. Thus, it can be used to investigate, prevent and reduce very adverse environmental effects (Wu & Zhang, 2014).

In this study, the conceptual framework of drivers, pressures, state, impact, and response (DPSIR) was used to organize cause-and-effect relationships and investigate the conditions of areas with severe vulnerability and express appropriate solutions. The DPSIR is an essential conceptual approach for organizing drivers, pressures, state changes, and impacts for an acceptable expression between human activities and environmental components, so that it can be useful in responding to policies (Gebremedhin et al., 2018). Therefore, this study attempted to use a hybrid approach based on MCDM methods and zone the EV in Anzali coasts based on the fuzzy gamma method and analytic network analysis (ANP) in the GIS environment. In addition, this study attempted to identify risk factors for an area with the severe vulnerability of this coast based on the DPSIR method, conduct risk assessment, and rank them based on vlse kriterijuska optamizacija I komoromisno resenje (VIKOR) method and express management strategies to achieve sustainable development. The results can pave the way for managers and planners to make significant environmental management decisions nationally and internationally for this ecosystem.

Material and Methods

Study Area

The coast of Bandar-e Anzali is located in an area of 891.475 km2 between 37° 20′–35° 37′ north latitude and 49° 55′–50° 51′ longitude in northern Iran and the vicinity of the Caspian Sea (Fig. 1). Bandar-e Anzali has a humid subtropical climate (Cfa) under the Köppen climate classification and a humid subtropical climate (Cf) under the Trewartha climate classification. Bandar-e Anzali is the rainiest city in Iran with a record of 1892 mm of total annual rainfall. It has the largest port in Iran on the southern coast of the Caspian Sea with a population of 118,564 inhabitants. The port of Anzali is regarded as the economic, fisheries, and tourism industry pole in Iran due to its geographical location. Part of the natural environment in the area includes the sea (40 km coastal strip) with a large number of islands and peninsula, and the other important part forms the highly valued Anzali wetland ecosystem. It is worth noting that the Anzali wetland was registered in the Ramsar Convention in 1975 (Mousazadeh et al., 2015). The Caspian coast is severely affected by the oscillating behavior of the Caspian Sea. Over the last thousand years, the Caspian Sea level varied between 19 and 28 m below the mean sea level, based on Caspian Sea level reconstruction from combined historical documents and geological records by Naderi Beni et al. (2013). Furthermore, the legal boundary of the Caspian Sea according to the definition provided in the Law of New and Coastal Lands of Iran is 60 m from the last point of seawater intrusion in 1963. Thus, the border of the studied coast was determined according to these issues.

Fig. 1
figure 1

Location of the study area

Full size image

Methodology

The Ecological Vulnerability Index (EVI)

EVI is an important tool to determine the ecological conditions in vulnerable areas (Hong et al., 2016). It is generally thought that EVI shows the degree of resistance of ecosystems in response to external interference, including natural and artificial factors, at a specific spatial scale (Beroya-Eitner, 2016; Hong et al., 2016). The EVI method in this study was performed with MCDM aggregation and GIS spatial analysis. In particular, the MCDM method can manage data from multiple sources and determine the weight of each criterion. GIS is used for spatial analysis and to generate an EV map (He et al., 2018). EVI is an indicator that integrates multiple multidimensional and multivariate criteria based on Eq. (1) (Wu et al., 2018):

$${\text{EVI}} = { }\mathop \sum \limits_{{{\text{i}} = 1}}^{{\text{N}}} {\text{W}}_{{\text{i }}} {\text{A}}_{{\text{i}}}$$
(1)

where EVI represents the degree of ecological vulnerability on Anzali coasts, \({W}_{i}\) is regarded as the weight of the i-th criterion, and \({A}_{i}\) indicates the fuzzy value of the i-th criterion obtained from the gamma membership function. In general, there are five steps in assessing EVI:

  • Selection of criteria based on the Delphi method;

  • Data collection and processing;

  • Normalization of ecological vulnerability criteria;

  • Weight determination by analytic network process (ANP);

  • Integrate criteria and produce an EV map in the GIS environment.

Selection of Criteria Based on the Delphi Method

In the current study, 19 primary sub-criteria were determined based on the comprehensive literature studies on coastal and marine issues, the report of the Department of Environmental Protection of Iran in coastal areas, experts’ opinions, local conditions, and access to data (Table 1). Then, the Delphi method was used to prioritize the identified sub-criteria using the opinions of 15 experts. Delphi is an appropriate method to analyze the opinions of experts in decision-making processes, which normally recognizes the need to structure a group communication process to achieve a useful result for management purposes (Foroozesh et al., 2002). The panel in this study included 15 experts including professional experts in government organizations and academic experts from various geographical locations. Table 2 shows the demographic information of the survey experts. Experts were asked to rank the significance of the sub-criteria based on their opinions in the rank of 1–9 (insignificant = 1, low significance = 3, significant = 5, very significant = 7, too much significant = 9, and intermediate values = 8, 6, 4, and 2). In this study, the Delphi technique continued in two rounds and the criteria with an arithmetic mean of less than five were removed. After summarizing the total opinions of experts, the sub-criteria were placed in each criterion category based on their common characteristics.

Table 1 Review of resources for effective sub-criteria in coastal ecological vulnerability
Full size table
Table 2 Information related to the survey experts
Full size table

Data Collection and Processing

The selected criteria along with the relevant sub-criteria and their information are shown in Table 3. In this study, satellite images, thematic layers, and statistical data were used to assess the EVI. Satellite images were accordingly selected so that (1) the cloud cover of the images was a maximum of 10% and (2) their dates were close enough to each other. In this study, geometric and atmospheric corrections have been carried out for all satellite images.

  • Groundwater status: the quantity and quality of groundwater are the significant factors in the change of surface vegetation coverage, soil salinity, and soil degradation (Ma et al., 2013; Song et al., 2021) that lead to EV. In this study, groundwater data were acquired from the monitoring data of 55 borehole wells along the coast of Anzali (July 2017–June 2018) that were conducted by Mazandaran Province Water and Wastewater Company (2018). In this report, 55 wells were randomly sampled and the electrical conductivity (EC) and level of the groundwater were measured. In this study, the mean of annual data was interpolated based on the kriging method and classified into 5 classes (Fig. 2a, b).

  • Land use: Land use is carried out to identify the various classes on this coast that are vulnerable to environmental degradation. In this study, MODIS/Terra products were used to produce a land use map. MODIS products were downloaded from https://ladsweb.modaps.eosdis.nasa.gov/. on 30 June 2018. Land use was classified into 6 categories based on the supervised classification method. The LULC features were categorized as built-up, farmlands, forest, grassland, water body, and wetland (Fig. 2c)

  • The normalized difference vegetation index (NDVI): NDVI was used to measure the level of vegetation density in the research area. The reduction of vegetation density is one of the main factors in soil erosion, biodiversity reduction, and other major ecological vulnerability (Zaitunah et al., 2019). In this study, NDVI is derived from the Sentinel-2A image acquired on 30 June 2018 based on Eq. (2). Then, the NDVI map was classified into 3 classes: none, moderate, and dense vegetation (Fig. 2d).

    $${\text{NDVI}} = \left( {\frac{{{\text{NIR}} - {\text{RED}}}}{{{\text{NIR}} + {\text{RED}}}}} \right)$$
    (2)

  • NIR and R represented the near-infrared band and red band of the image, respectively.

  • Topography: The topographical and morphological conditions of the coastal area are critical for assessing the vulnerability of groundwater resources to salinization (Michael et al., 2013) which can lead to EV. In this study, major geomorphological features were extracted from the Geological map of Anzali (Fig. 2e). Also, the SRTM DEM was retrieved from https://www.opendem.info/link_dem.html on June 2018 (Fig. 2f)

  • Soil: Soil is one of the important factors in biomass production, water storage capacity, and protection of flora and fauna (Meléndez-Pastor et al., 2017) and the reduction of its quality leads to ecological vulnerability. In this study, soil quality and texture data were derived from the report for soil and water resources in Iran's Ministry of Agriculture Jihad (2000). In the mentioned report, the soil quality map was prepared according to characteristics such as the percentage of soil mineral particles, pH, EC, and organic carbon, and the fuzzy membership function was used to standardize the sub-criteria in values ranging from 0 to 1. In this study, the soil quality map was classified into 4 classes based on the equal interval method: values less than 0.25 (bad), 0.25 to 0.5 (moderate), 0.5 to 0.75 (good), and more than 0.75 (best) (Fig. 3a). Soil texture was classified into 4 classes: silty loam, sand, sandy loam, and silty clay loam (Fig. 3b). Moreover, Landsat 5 TM and Landsat 8 OLI images (WRS_PATH = 166 and WRS_ROW = 34) were acquired on 5 May 2000, and 8 May 2018 to produce the soil salinity (Fig. 3c). These imageries are available on the https:// earthexplorer.usgs.gov/. The salinity index is derived based on Eq. (3).

    $${\text{Salinity Index}} = \left( {\frac{{{\text{Redband}}}}{{{\text{NIRband}}}}} \right){*}100$$
    (3)

where, NIR and R represented the near-infrared band and red band of the image, respectively.

Table 3 Selected criteria and sub-criteria and their information resources
Full size table
Fig. 2
figure 2

Map of sub-criteria a Groundwater level, b Electrical conductivity of groundwater, c Land use, d NDVI e Major geomorphological features, f Elevation

Full size image
Fig. 3
figure 3

Map of sub-criteria a Soil quality, b Soil texture, c Soil salinity, d Average annual precipitation, e Average annual temperature, f Drought

Full size image
Fig. 4
figure 4

Map of sub-criteria Coastal erosion

Full size image

  • Climate: In recent decades, coastal ecosystems and near-coast have been affected by climate change, which can lead to sea-level changes, loss of biodiversity, habitat destruction, and coastal erosion. The consequences of these changes potentially affect coastal livelihoods, the health of coastal and marine environments, ecosystem services, and coastal livelihoods (Halpern et al., 2012; Oloyede et al., 2021). In this study, precipitation and temperature data were obtained from 4 synoptic stations in Mazandaran province (July 2017–June 2018). Montazeri and Bayet (2012) stated that the change in climatic parameters in the northern regions of Iran which have led to the separation of climatic sub-regions is affected by factors such as distance from the sea and increasing elevation. Considering that the Anzali coast is a plain region (Fig. 2e), so the whole coastal area has a relatively homogeneous climate. Anzali city is the rainiest city in Iran with a total annual rainfall of 1892 mm and relative humidity fluctuates between 71 and 97% per year because there are no drastic changes in precipitation and temperature throughout the year. So the annual average has been used to study climatic parameters in this study. In this study, the mean of annual data was interpolated based on the kriging method (Figs. 3d and 3e). Moreover, the standardized precipitation index (SPI) is widely used in the research areas of drought analysis on a range of timescales (Wang et al., 2022). In this study, SPI was used to determine the drought for a year period and estimated based on Eq. (4). The SPI map is classified according to McKee et al. (1993) (Fig. 3f)

    $$SPI = \left( {{\raise0.7ex\hbox{${Pi - \overline{P}}$} \!\mathord{\left/ {\vphantom {{Pi - \overline{P}} {SD}}}\right.\kern-\nulldelimiterspace} \!\lower0.7ex\hbox{${SD}$}}} \right)$$
    (4)

    where, \(Pi\) is the amount of rainfall, \(\overline{P }\) is the average amount of rainfall and SD is the standard deviation.

  • Interaction with the sea: The tides of the sea and the influence of storm waves change the coastline (Wu et al., 2018). In this study, the subtract of NIR bands of Landsat 5 TM and Landsat 8 OLI was used to extract the boundary between water and soil. NIR band has the largest difference in reflectance values between water and non-water bodies (Mondejar & Tongco, 2019). Also, a high-pass filter was used to highlight the boundary between features. In this map, pixels with negative values were considered as areas that have changed from sea pixels to land pixels. These areas can be eroded by sea waves and wind (Fig. 4).

The data related to each sub-criterion should initially be transformed into a layer in the GIS software for mapping the EV in Anzali coasts based on the aforementioned sub-criteria. All of the layers were projected based on the UTM system and WGS 1984 datum and converted to grid layers with 30 m × 30 m cell size.

Normalization of Ecological Vulnerability Criteria

In the current study, the layers were standardized by the fuzzy logic technique. The theory of fuzzy logic as a knowledge-based approach is proposed by Zadeh (1965). The basis of this method is to convert input data into normalized outputs by calculating weights for fuzzy members based on fuzzy set theory (Zhang et al., 2017). In the theory of fuzzy logic, fuzzy members are weighed based on significant distances of different features, for each pixel or spatial position, a weight is assigned between zero and one. In this study, linear and small membership function was used. The linear membership function transforms the input values linearly on the 0 to 1 scale, with 0 being assigned to the lowest input value and 1 to the largest input value (Tabrizi et al., 2012). The small function is a nonlinear fuzzy membership. It is used when the low values in the membership map are close to one. This function is defined based on an intermediate point defined by the user. In this study, the minimum and maximum values for the sub-criteria were determined according to referring to the literature, the results of previous studies (Mani Murali et al., 2013; Rahmanipour et al., 2014; Wu et al., 2018) and professional books. The linear score allocation function was used for all sub-criteria, except the sub-criterion of coastal erosion. This function has two thresholds of minimum and maximum. In addition, other values were linearly fuzzy between these two values. In the case of coastal erosion sub-criteria, it is obvious that pixels with a negative value in the shoreline displacement map (indicating shoreline displacement and increased erodibility) have a higher fuzzy score. As a result, the small weighting method was used. Table 4 indicates the method of assigning fuzzy points to each sub-criterion.

Table 4 Fuzzy membership functions defined for the criteria and sub-criteria used in the study
Full size table

In this study, the gamma fuzzy operator expressed according to Eq. (5), was used.

$$\mu_{{{\text{Combination}}}} = \left( {{\text{FuzzySum}}} \right)^{\delta } *\left( {{\text{Fuzzyproduct}}} \right)^{1 - \delta }$$
(5)

where \(\mu_{{{\text{Combination}}}}\), Fuzzyproduct, and Fuzzysum indicate the fuzzy gamma layer, product of the fuzzified layers, and algebraic summation of the fuzzified layers, respectively. In the above equation, the gamma fuzzy operator \({\mu }_{\mathrm{Combination}}\) is a combination of the algebraic multiplicative fuzzy operator (PRODUCT) and the algebraic sum fuzzy operator (SUM) in which the gamma parameter (δ) is selected in the range of zero to one. Therefore, if (δ) is one, the composition will be the same product of algebraic sum (SUM) and if (δ) is zero, the composition will equal the product of algebraic multiplication (PRODUCT). The correct and conscious choice of gamma (δ) creates some values in the output which act as the increasing role of the fuzzy algebraic sum and the decreasing role of the fuzzy algebraic multiplication (Sadeghi & Khalajmasoumi, 2015). In this study, the fuzzy gamma operator was used to adjust the very high sensitivity of the algebraic multiplicative fuzzy operator and the very low sensitivity of the algebraic sum fuzzy operator. This operator works between two algebraic multiplication and sum operators. In this study δ = 0.7 was used to obtain the output map which has flexible compatibility between the increasing and decreasing tendencies of two fuzzy multiplication and sum operators.

Weight Determination by Analytic Network Process (ANP)

In the next step, the analytic network process (ANP) method was applied to weigh the criteria. ANP method is an extension of the analytic hierarchy process (AHP) method developed by Saaty (1996) to solve MCDM problems. ANP can model complex decision problems more realistically by eliminating the hierarchical structure (Moradpour & Long, 2019) and constructing a network with inter-level relationships. In general, the steps required to prioritize options based on the ANP method can be described in six steps as follows (Afzali et al., 2014; Eldrandaly, 2013):

Step I Determine criteria, options, and decision structure.

Step II Form the matrix for pairwise comparisons of decision structure.

Step III Examine the decision consistency ratio and calculate the consistency ratio.

Step IV Determine the relative weight of decision elements based on the eigenvector method.

Step V Form a supermatrix of decision structure based on computational weights.

Step VI Calculate the range of the even powers in the supermatrix and determine the supermatrix related to ultimate preference.

In this study, a pairwise comparison of the criteria with their sub-criteria was performed using the analytic network process (ANP) after determining and screening the criteria and sub-criteria. To determine the significance of one criterion, a table of scoring preferences was used and the criteria were based on the degree of significance from 1 to 9 (insignificant = 1, low significance = 3, significant = 5, very significant = 7, too much significant = 9, and intermediate values = 8, 6, 4 and 2). Then, six types of pairwise matrices were designed for pairwise comparisons and provided to 15 relevant experts in the form of the main questionnaire. In one of these matrices, the pairwise criteria were compared while the pairwise (pairwise matrices) sub-criteria related to each criterion and their preference over each other were compared in the other matrices. At this stage, Super Decision software was used to analyze the criteria. This software is reliable for evaluating multiple criteria with many capabilities such as calculating the incompatibility coefficient and the relative weight and final weight of options. In this study, all of the questionnaires which had an incompatibility coefficient of less than 0.1 were selected. The average compatibility rate of the questionnaires approved in this study was 0.063. In this study, the weight of the criteria was obtained by their averaging.

The Vulnerability of the Area Based on the DPSIR Method and Risk Assessment

In this study, the area which was placed in a very high vulnerability category based on the zoning map was used as a goal to assess the conditions and assess the risk. This section focuses on the environmental conditions of the region and all stressors causing the ecological balance, as well as endangering the survival of the ecosystem. The conceptual framework of this section is shown in Fig. 5. In general, this section includes three steps:

  • Identification of risks and vulnerable factors in the wetland based on the DPSIR method;

  • Risk assessment;

  • Rank of risks based on the VIKOR method.

Fig. 5
figure 5

DPSIR conceptual model framework and risk assessment

Full size image

Identification of Risks and Vulnerable Factors in the Wetland Based on the DPSIR Method

The first step of this study includes system description and analysis using data and library study in the framework of the DPSIR model. The DPSIR approach was proposed for the first time by the European Environment Agency and is widely used in the assessment and management of different environmental issues (Song & Frostell, 2012). This approach is based on a causal relationship that begins with human activities (driving force) and pressure on the environment and leads to responses with quantitative and qualitative changes in natural resources. Indeed, this model can be used with a practical approach as a tool for decision-making and management plans. This issue is highly necessary for planning and formulating a strategy to cope with and adapt to the phenomena. The components of this model include driver, pressure, state, impact, and response.

Drivers

Drivers (or driving forces) are various factors that lead to a change in the behavior of a system (Burkhard & Müller, 2008). These are any natural or human-induced factors that lead to major social, demographic, and economic changes (Schrevel & Kumar, 2010). In this study, the driving force refers to tourism, settlements and urban areas, population growth, and economic activities such as agriculture and aquaculture, which lead to environmental problems or changes in the lifestyle of coastal communities.

Pressure

This factor indicates the various consequences of human actions that are the result of driving forces. In most cases, all human activities affecting the environment can be classified as pressure (Zhou et al., 2013). In this study, various forms of human activities on the wetland that directly increase the risk of ecosystem degradation, such as land use patterns, water, and biological resources were been considered as factors of pressure.

State

The state index refers to the condition of the ecosystem resulting from drivers such as demographics, socio-economic natural factors, and pressures such as environmental factors (Kosamu et al., 2022). In this study, changes in the quantity and quality of various environmental elements in the wetland (soil, water, plants, animals, etc.) and subsequently their inability to provide ecosystem services were considered as factors of the wetland state.

Impact

The impact is defined as the socio-economic and environmental impacts that affect the local population’s livelihood, social welfare, and the environment due to the changing state of the ecosystem (Kosamu et al., 2022). The impact component is very important for management and decision-making because they directly explain the environmental and social consequences of human activities (Pinto et al., 2013). In this study, the effect of each of the components described in the previous steps on physical, chemical, biological, economic, and social circumstances in the receiving environment was determined in this step.

Responses

Responses are also activities performed by human society to reduce environmental pressures and the improvability of the environment. Responses are the best way to reduce the stresses and effects of dependent systems (Saadati et al. 2013). In this study, the responses to the problems related to the driving force and pressure on the wetland ecosystem were expressed using legal procedures, government guidelines, and public participation.

Risk assessment

In the second step of the study, three indicators of risk intensity, risk probability, and sensitivity of the acceptor environment were used for risk assessment. Then, the method of determining the scores of each indicator was presented. In this study, Tables 5 and 6 were used to determine risk intensity and environmental sensitivity. Based on Table 5, the score will be high when the risk of the wetland and the losses to the wetland resources are high. As shown in Table 6, a more sensitive risk acceptor environment leads to a higher risk score (Jozi & Shafiee, 2009; Makvandi et al., 2016). Based on the probability index in Table 7, the higher the probability of occurrence of the outcome due to the risk factor, the higher the risk score.

Table 5 Values related to the intensity of occurrence of risk aspects
Full size table
Table 6 Values related to the sensitivity of the recipient environment
Full size table
Table 7 Values related to the probability of the occurrence of risk
Full size table

The Rank of Risks Based on the VIKOR Method

In the third step, the VIKOR method was used to rank the risks. The VIKOR method has been developed for the multi-criteria optimization of complex systems by Opricovic and Tzeng (2004). This method focuses on rating and selecting a set of alternatives and determining the agreed solutions to an issue with conflicting criteria (Chen & Wan 2009). The multi-criteria measure for compromise ranking is developed from the Lp-metric used as an aggregating function in a compromise programming method (Yu, 1973; Zeleny 1982). The various J alternatives are denoted as a1, a2,..., aJ. For alternative aj, the rating of the ith aspect is denoted by fij, i.e. fij is the value of ith criterion function for the alternative aj; n is the number of criteria. Development of the VIKOR method started with the following form of Lp-metric in Eq. (6) (Opricovic & Tzeng, 2004):

$$L_{p.j} = \left\{ {\mathop \sum \limits_{i = 1}^{n} [w_{i} \left( {f_{i}^{*} - f_{ij} } \right)/\left( {f_{i}^{*} - f_{i}^{ - } } \right)]^{p} } \right\}^{1/p} .$$
(6)
$$1 \le p \le \infty ;j = 1. 2. \cdots . J.$$

In the VIKOR method L1,j (as Sj) and L∞,j (as Rj) are used to formulate ranking measures. The solution (Qj) obtained by minj Sj is with a maximum group utility, and the solution obtained by minj Rj is with a minimum individual regret of the opponent. The VIKOR algorithm is as follows:

Step 1 It is the determination of the alternatives (Am) and criteria (xm) and the formation of the decision matrix according to Eq. (7).

$$D = \begin{array}{*{20}c} {A_{1} } \\ {A_{2} } \\ {\begin{array}{*{20}c} \vdots \\ {A_{m} } \\ \end{array} } \\ \end{array} \left[ {\begin{array}{*{20}c} {\begin{array}{*{20}c} {x_{1} } & {x_{2} } \\ {x_{11} } & {x_{12} } \\ {x_{21} } & {x_{13} } \\ \end{array} } & \cdots & {\begin{array}{*{20}c} {x_{n} } \\ {x_{1n} } \\ {x_{2n} } \\ \end{array} } \\ \vdots & \vdots & \vdots \\ {\begin{array}{*{20}c} {x_{m1} } & {x_{m2} } \\ \end{array} } & \cdots & {x_{mm} } \\ \end{array} } \right]$$
(7)

Step 2 This step normalizes the decision matrix using Eq. (8).

$$r_{ij} = \frac{{{\text{x}}_{ij} }}{{\sqrt {\mathop \sum \nolimits_{{{\text{i}} = 1}}^{{\text{m}}} {\text{x}}_{ij}^{2} } }}$$
(8)

Step 3 Determine the best and the worst values of all criterion functions, for i = 1, 2,..,, n; based on Eqs. (9) and (10).

$$f^{*} = Max_{j} f_{ij} ;$$
(9)
$$f^{ - } = Min_{j} f_{ij}$$
(10)

Step 4 It determines the usefulness (Sj) namely the relative distance of alternative jth from the ideal point (Eq. 11), and regret (Rj), which shows the maximum discomfort of alternative j from the distance from the ideal point (Eq. 12). The wi is the weight of criteria that was calculated by the ANP method.

$$S_{j} = \mathop \sum \limits_{i = 1}^{n} w_{j} \left( {f_{i}^{*} - f_{ij} } \right)/\left( {f_{i}^{*} - f_{i}^{ - } } \right)$$
(11)
$$R_{j} = Max\left[ {w_{j} \left( {f_{i}^{*} - f_{ij} } \right)/(f_{i}^{*} - f_{i}^{ - } )} \right]$$
(12)

Step 5 It calculates the VIKOR index (Q) according to Eqs. (13), (14), and (15). V in relation to weights for the maximum strategy represents the desirability of group decision-making, which is normally a constant value of 0.5.

$$Q_{j} = \frac{{v\left( {S_{j} - S^{*} } \right)}}{{S^{ - } - S^{*} }} + \left( {1 - v} \right)\left( {R_{j} - R^{*} } \right)/\left( {R^{ - } - R^{*} } \right)$$
(13)

where,

$$S^{*} = Min_{j} S_{j} ; S^{ - } = Max_{j} S_{j} ;$$
(14)
$$R^{*} = Min_{j} R_{j} ; R^{ - } = Max_{j} R_{j}$$
(15)

and v is the weight of the strategy of the majority of criteria. Normally, v was assumed as v = 0.5. However, v can take any value from zero to one.

Step 6 It ranks the options based on the values of R, Q, and S in three groups from small to large.

Step 7 Suggest a compromise solution of the alternative (a') that is on the top of the ranked list of Q if satisfy two conditions:

Condition 1 Acceptable advantage: if Q (a'')–Q (a') \(\ge\) DQ, where a'' is the second best alternative based on Q ranking, and DQ = \(\frac{1}{J-1}\), while J is the number of alternatives.

Condition 2 Acceptable stability in decision-making. Alternative a' should be the best alternative based on S and R rankings. This compromise solution can be considered stable in a decision-making process, which could be ‘‘voting by majority rule’’ (when v > 0.5 is needed), ‘‘by consensus’’ v \(\approx\) 0.5, or ‘‘with veto’’ (v < 0.5), where v is the weight obtained for the strategy of decision-making ‘‘the majority of criteria’’ (or ‘‘the maximum group utility’’).

Results

Results of EVI on the Anzali Coast

As Table 8 represents, the weight of the criteria is obtained using the ANP method after entering the data layers related to the criteria into the GIS environment. Based on the results, the weight determined by the aforementioned method indicates that the soil criterion with a weight of 0.27 followed by the vegetation (0.24), groundwater (0.18), and land use (0.15) has the highest importance among the criteria involved in evaluating the vulnerability in the studied area, while the topography, interaction with the sea, and climate criteria account for the lowest weight (0.05).

Table 8 Weights obtained from the ANP method
Full size table

In the next step, the standardized maps related to the decision-making criteria were created in the GIS environment. Finally, the maps related to the criteria and weights obtained from the ANP method were combined based on the EVI formula, and the final map of the EV on the Anzali coast was created. Finally, the vulnerability map is divided into five categories based on the natural break method. This method aims to minimize the standard deviation between the data in each category and maximize the standard deviation between the categories. In other words, this method aims to reduce the variance in the category and maximize the variance between categories and groups. Accordingly, this method presents its classification (George, 1967) which is an appropriate method for classifying zones based on the desired goals. Figure 6 illustrates the ecological vulnerability zoning of Anzali coasts in the studied area, classified into five categories none, slight, moderate, severe, and extreme vulnerability. Based on Figs. 6 and 7, 19,179.9 (21.51%), 15,464.1 (17.35%), and 7153.04 (8.02%) hectares of the entire area have moderate, severe, and extreme degrees of vulnerability, respectively. However, 24,848.5 (27.87%) and 22,502 (25.24%) hectares of Anzali coasts have slight and none vulnerability potential, respectively.

Fig. 6
figure 6

Final map related to the EVI of the Anzali beaches

Full size image
Fig. 7
figure 7

Area and percentage of each EVI class on the Anzali coast

Full size image

Soil salinity is an important factor in reducing groundwater quality as well as the establishment of vegetation in natural habitats, which can reduce biodiversity, degrade natural habitats, increase soil erosion and accelerate the desertification process. In the present study, to evaluate and verify EV classes, field sampling was performed from soil salinity and vegetation in each of the EV classes (Table 9). As observed, two important indicators affecting the determination of ecological vulnerability, namely NDVI and soil salinity are measured in ds/m in the field samples. The results of the field validation revealed the accuracy of the designated classes so that the samples taken from the soil at the no risk ecological class had EC lower than 0.7 ds/m and vegetation was higher than 0.7, while the EC of soil salinity in soil with extreme vulnerability was more than 3.5 ds/m and lacked vegetation.

Table 9 Field validation matrix of ecological vulnerability classes based on soil salinity and NDVI
Full size table

Sensitivity Analysis Results

Sensitivity analysis determines the extent and effect of input data on output data (Malczweski, 2006). Based on the change in each of the studied criteria, the change in the ecological threshold of the constructed model was examined (Fig. 8). Based on the result, the ecological threshold changes by − 0.296, 0.3002, 0.5061, − 0.3187, − 0.2758, and − 0.2574 with a unit change in climate, soil, land use, groundwater, vegetation cover, topography, and interaction with coastline, respectively. The results of sensitivity analysis showed that the highest and lowest changes are related to land use and climate criteria, respectively.

Fig. 8
figure 8

Sensitivity analysis of the constructed model for changes in ecological threshold to coastline interaction change

Full size image

Results of Conditions in the Region and Risk Assessment

As indicated in Fig. 6, Anzali Wetland has been considered a significant and international ecosystem that is located in the high EV class. In the first part of the findings, the DPSIR analytical model for Anzali International Wetland was used to find the driving force and pressure factors on the conditions of the wetland and to express a solution to improve the current situation. A cause and effect relationship was determined using its conceptual framework as shown in Fig. 9.

Fig. 9
figure 9

Integrated DPSIR framework with vulnerability assessment

Full size image

The VIKOR method was used to rank the risk factors after determining 16 risk factors based on DPSIR results in this section. To conduct the VIKOR model, three criteria of effect intensity, probability of occurrence, and sensitivity of the recipient environment were estimated for each risk factor (Table 10). Tables 11, 12, and 13 show the results of ranking the hazards threatening the wetland in the three environments.

Table 10 Scoring the indicators of intensity and sensitivity of the recipient environment
Full size table
Table 11 Results of ranking the threatening risks in the physical–chemical environment of Anzali International Wetland based on VIKOR method
Full size table
Table 12 Results of ranking the threatening risks in the biological environment of the Anzali International Wetland based on VIKOR method
Full size table
Table 13 Results of ranking the threatening risks in the socio-economic environment of Anzali International Wetland based on the VIKOR method
Full size table

Based on Table 13, the results obtained from ranking the threatening risks of Anzali International Wetland using the VIKOR method indicated that the entry of wastewater from the city and surrounding villages into the wetland in the physical–chemical environment, irregular harvesting of plant wetland resources in the biological environment, and occupying and changing the land use in the socio-economic environment were the priorities of each environment. According to the risk ranking, control measures should be taken to reduce the environmental risks of Anzali International Wetland and some solutions in this regard are presented in Table 14.

Table 14 Control measures to reduce the environmental risks of Anzali International Wetland
Full size table

Discussion

The length of the coastline in the Caspian Sea is considered the focus of valuable ecosystems such as estuaries, wetlands, important wildlife habitats, and the like, requiring special management. Furthermore, preparing EV maps for such coasts can be a valuable strategy in managing biological resources to ensure sustainable productivity and management. There is a need to assess the sensitivity of the Anzali coast as it faces a variety of pollutants, erosion, and various physical obstacles along with the gradual growth of the population and increasing development (Doukakis, 2005; Sankari et al., 2015). The present study aims to zone the vulnerable spots on the Anzali coast based on the GIS-based MCDM. In this study, EVI mapping has been done based on the ANP method, which is one of the most efficient and widely used methods in vulnerability assessment (Andalecio, 2010; Mani Murali et al., 2013). ANP compares sub-criteria with the cluster of criteria as well as all criteria and their interrelationships with sub-criteria of other criteria. This analysis demonstrates the significance of each element on other elements as well as in the complete vulnerability assessment. Thus, the ANP was used to make optimized weights by considering the relative weights based on criteria and sub-criteria. Khodadoust et al. (2016) conducted a study on the prioritization and mapping of coastal areas to oil loss by comparing AHP and ANP methods in Nayband Marine National Park in Assaluyeh. They stated that the ANP method has more consistent than the AHP method to consider the interrelationship of different levels of decision with each other and also the internal relationship of decision criteria at one level. Mani Murali et al. (2013) and Tahri et al. (2017) introduced GIS-FAHP as an effective method in coastal vulnerability analysis by investigating the east coast of India and the Mohammedia coast in Morocco, respectively. In addition, Wu et al. (2018) conducted a study on the EV assessment of YRD based on the FAHP and fuzzy logic models. They demonstrated that the combination of the two models was useful for the assessment and the spatial distribution of all vulnerability grades was regular.

In the current study, the ANP results showed that the NDVI (0.24), landuse (0.15), and soil salinity (0.122) sub-criteria had the highest relative weights, respectively. It is shown that the mentioned criteria play the most significant role in assessing the EV of the coasts. In addition, the sub-criteria of major geomorphological features and climate such as drought, annual average temperature and precipitation due to the small size of the study area, low height difference, and presence of coastal plains were the least important in vulnerability assessment. Mani Murali et al. (2013) stated that the slope, geomorphology, elevation, and erosion rate are important criteria for physical vulnerability, and the remaining criteria, including sea-level change, wave height, and mean tidal range does not vary with respect to vulnerability. They stated that this is because the extent of the shoreline under consideration is small and so only one relevant value is obtained for the data. Tahri et al. (2017) declared that the shoreline evolution, elevation, geomorphology, wave height, sea-level rise, tidal range, and distance to an urban area are important criteria to analyze coastal vulnerability. Moreover, Wu et al. (2018) stated that soil salinity, soil quality, groundwater quality, and land use had the highest weight in EV assessment, respectively. Considering the different weights of the criteria in various studies, it can be concluded that the importance of the criteria will be a function of the environmental conditions and the location of the studied area.

Areas with slight or none vulnerability (53.11%) are mostly in the forest and grassland in the west and southwest, while the areas with severe and extreme vulnerability (25.37%) are more concentrated in the areas with the water bodies, Anzali Wetland, Siahkesham Protected Area, Sorkhangol Wildlife Refuge, and the Anzali City. The rest of the area is scattered in the northern part of the coastline, as well as east and southeast with the largest presence of agricultural land and scattered villages. In addition, the aforementioned areas correspond to those with medium to poor soil quality, high salinity of the soil, sandy to loamy sand, and the highest EC groundwater values at a glance. Further, GIS-based multi-criteria evaluation methods are considered an appropriate approach to decision-making given the nature of coastal vulnerability assessment, which involves various ecological, technical, and economic criteria (Cozannet et al., 2013; Ozyurt et al., 2011; Tahri et al., 2017; Wu et al., 2018). Furthermore, coastal vulnerability zoning based on fuzzy gamma is among the most efficient, widely used, and accepted methods (Andalecio, 2010; Mani Murali et al., 2013).

The evaluation of the zoning map in Fig. 6 indicated that Anzali International Wetland is located in a zone with severe vulnerability. The drivers, pressures, state, impacts, and responses were identified in the area of Anzali Wetland with the help of the conceptual model, as shown in Fig. 9. The pressures resulting from the drivers cause a change in the state of the ecosystem and affect the natural and man-made conditions of the environment. This impact forces society to offer some solutions in the form of responses to reduce the impact. Based on the DPSIR framework, the drivers identified for the changes are agriculture, aquaculture, settlements and urban areas, population growth, and tourism. The requirements of the growing population are considered a driver which puts pressure on the land's basic resources. Increasing human population density increases the demand for access to production and service services from the wetland. This increase in population increases the entry of urban and rural effluents into the lagoon and the Caspian Sea, the entry of pesticides and chemical fertilizers into the lagoon, the entry of nutrients into water such as nitrogen, and phosphate, and the growth of wetland plants and eutrophication.

On the other hand, population increase has caused land use changes in the margins of Anzali Wetland. Based on the study of Mousazadeh et al. (2015) about 6500 hectares (about 69%) of pastures in the watershed of Anzali Wetland were destroyed and urban and agricultural areas expanded to 2982 hectares (~ 74%) and 2228 hectares (~ 6%) during 1975–2013. This change of use can increase the inflow of effluents, flood potential, and water utilization but reduce groundwater quality in the Anzali wetland, leaving numerous negative effects on the appearance and aesthetics of the wetland and reducing the habitat of aquatic species, birds, and the continuation of regeneration. In addition, another driver of this wetland is unstable tourism activities. Pristine nature, diversity of plant and animal communities, and temperate climate annually attract many tourists who use the welfare facilities along the lagoon created regardless of the ecological status of the lagoon in the region. This activity can have adverse effects on the environment such as soil erosion, destruction of wetland habitats, noise pollution caused by motor boat traffic, groundwater pollution, and endangering the health of inadequate waste and sewage networks in the wetland environment. Jahanishakib and Malekmohammadi (2017), Pirrone et al. (2005), and Zacharias et al. (2008) considered agriculture, population growth, tourism, and industry as drivers. In addition, the results of ranking the threatening risks of Anzali International Wetland using the VIKOR method indicated that the entry of wastewater from the city and surrounding villages into the wetland, uncontrolled harvesting of plant wetland resources, as well as occupying, and changing the land use cover significant factors in the vulnerability of Anzali wetland can be studied in all three dimensions.

Conclusion

Coastal ecosystems play a significant role in human welfare by providing essential services to human communities. However, the health of such valuable ecosystems is highly threatened by environmental pressures from natural and man-made activities throughout the world. Therefore, EV mapping on the Anzali coast is essential to implement integrated protection management. In this study, a quantitative analysis of the ecological vulnerability of the Anzali coast was conducted using ecological criteria and based on the MCDM approach in the GIS environment. Quantitative assessment of vulnerability can help decision-makers make accurate decisions regarding the management of vulnerable areas and investigate the factors affecting the vulnerability of the area. The vulnerability zoning map shows that the spatial distribution of vulnerability classes in the Anzali coasts has been irregular and scattered. In this study, areas with none and slight vulnerability are mostly located in the forest and grassland in the west and southwest, while most areas with severe to extreme vulnerability are mostly located near Anzali wetland, adjacent protected areas, and Anzali city. However, the comparison of the vulnerable areas with criteria maps showed that the vulnerable areas are located in areas with medium to poor soil quality, high soil salinity, sandy to loamy sand, and the highest amounts of EC groundwater. These results showed that GIS-MCDM is an appropriate approach for vulnerability zoning of coastal areas, and its results can provide better decision support for planning in the sustainable management of coastal areas.

In this study, a systematic analysis and identification of risks and vulnerable factors were conducted based on the DPSIR method and risk assessment of Anzali wetland as one of the high vulnerability areas. Based on the results, the entry of wastewater from the city and surrounding villages into the wetland, uncontrolled harvesting of wetland plant resources, and land use changes are the most significant factors threatening the environment of the Anzali wetland. Such factors are the main causes of adverse effects on the unique features of the wetland. Since Anzali International Wetland is a valuable wetland in the Ramsar Convention and Black Wetlands of Montreux, preserving and preventing its ecological changes as functional ecosystems are necessary for sustainable development. The results of this process can be highly essential for decision-makers to obtain a framework of the situation and the effects of their actions on wetlands based on factors such as driver and pressure for taking management measures to reduce the pressures on this international wetland. Of course, this study has its limitations. As this study only investigated the spatial distribution of EV in 2018, the analyses were homogenous and limited. As a result, it is suggested that EV be conducted in different seasons and time series for a more in-depth analysis of environmental features due to the spatial dynamic changes of the studied data. In this study, the ANP method was used to nonlinear structure and considered the interrelationship of different decision levels with each other as well as the internal relationship of decision criteria at the same level. So, the authors suggest the use of other weighting methods such as BWM, DEMATEL-ANP, SWARA, or fuzzy methods such as fuzzy ANP to eliminate the uncertainty and ambiguity of verbal variables. In this study, ecological vulnerability assessment has been done with an emphasis on ecological criteria. Therefore, it is suggested that socio-economic criteria be used in future studies to increase the accuracy of analyses to consider all the principles of sustainable development to provide a comprehensive and complete view of the region's condition for decision-makers.