May a highly touristic tropical island sustain water resources? Navigating troubled waters to berth on a safe shore

Abstract

While being a major economic development force for tropical islands, tourism may aggravate disturbances on local water resources (scarcity, pollution) and the surrounding ecosystem. In this work, we used a discrete events modelling approach to characterize trajectories of Cat Ba Island (Vietnam) and its ability to reach specific trajectories. We wanted to assess if a healthy social ecosystem could be compatible with tourism. We found that co-existence between water resources and tourism may be reached; moreover, a systemic state where ecosystem and sociosystem can sustainably coexist may be reached. The implementation of urban ecological infrastructure is a key process to shift from the present state towards more favourable states. Another significant result is that the departure from the present state to reach more favourable ones may lead to a transitory degraded situation; only by keeping in mind the systemic objective significant transformations can be carried out.

Supplementary Information

The online version contains supplementary material available at 10.1007/s13280-024-02079-4.

Introduction

Tourism has been a key economic sector in many Southeast Asian countries, accounting for 12% of the region’s gross domestic product (GDP) in 2019. Around 13% of total jobs in 2019 has been created by this industry, which was equivalent to 42 million jobs (OECD 2023). Nevertheless, over-tourism, a consequence of dramatic growth of tourism, is generating negative impacts on local environment, mainly via waste increase and greenhouse gases emission (Ahmad et al. 2019). Southeast Asian islands, among all destinations, are particularly suffering from these consequences. Another disadvantage of islands compared to mainland is its remoteness from various sources of food, water, utilities like gas and electricity, or public services (Partelow and Nelson 2020). Therefore, problems associated with over-tourism are critical on islands, especially regarding waste management, carbon dioxide emission, crowding and environment and water degradation with evidences in Malaysia (Ghaderi et al. 2012), Philippine (Fernandez-Abila et al. 2024), and Indonesia (Hayati et al. 2020; Adrianto et al. 2021).

Vietnamese coastal region is also under pressure from a hot touristic development (Veettil et al. 2020). Serious concerns have been raised in terms of solid waste management (Tsai et al. 2021), plastic pollution (Kerber and Kramm 2022), environment and ecological conservation (Hayward and Tran 2014; Tien and Nguyen 2019), or freshwater shortage (Nguyen et al. 2021). Cat Ba Island is the biggest limestone island in tropical Southeast Asian region, located in the northwest margin of the Gulf of Tonkin, Northern Vietnam. The island has unique tropical limestone landscape and ecosystem (Van et al. 2010; Phuong et al. 2013) which has been renowned internationally as an UNESCO-recognized biosphere reserve since 2004. As a consequence, the island became a famous tourist destination in Vietnam, receiving around 2.5 million tourists comparing to the island population of 31.3 thousand in 2018 (Cat Hai Division of Statistics 2021). The increase in anthropogenic activities, particularly tourism, has generated a heavy burden on Cat Ba Island (Cao and Nguyen 2018; Mai and Smith 2018). Among other disturbances, those activities resulted in water shortage status in the island, especially during the high touristic season (Phan et al. 2019). Climate change and over-tourism have been continuously surpassing water supply capacity of Cat Ba Island. Serious water shortage occasions were recorded in July 2010 and June 2018 which were the results of drought combined to overcrowding (Hung 2010; Thinh 2018). Wastewater management is another challenge with high nitrate levels detected in the public beaches of the island (Trang et al. 2022). In addition, the current policies aiming to promote dramatic growth of tourism in Cat Ba Island have been warned that it could lead to unsustainable outcomes including over-exploitation of local resources (land and freshwater), uncontrolled temporary immigrants resulting to local unemployment, limited access to school and environmental degradation (Mai and Smith 2015).

Water resources and ecosystem statuses, as driven by social behaviours (tourists and managers actions) and biophysical processes (environmental degradation and changes of the hydrological cycle), are the main issues that were targeted in this work. We thus aimed at modelling Cat Ba social-ecological system using variables related to water resources and tourism, potential threats generated by human action and the stakeholders involved in the management of both these resources and degradations (Cao and Nguyen 2018; Nguyen et al. 2021). Based on the pressure exerted by tourism on water resources on tropical touristic islands, we would like to question the co-existence of tourism and water resources on Cat Ba Island, as follows: (Q1) "Is Cat Ba system able to reach a state to simultaneously achieve sustainable tourism and water resources?" and (Q2) "Does the trajectory towards sustainable system states correspond to a gradual shift from detrimental to beneficial system status?". The associated hypotheses are: (H1) The presence of tourists may generate depletion of (both surface and groundwater) water resources, additional degradation of soil physicochemical and biological status, additional disturbance of fauna and flora and additional discharge of wastewater and littering of solid waste (Mai and Smith 2015, 2018; Cao and Nguyen 2018); (H2) The activity of managers may help limiting artificialization and may help reclaiming water resources through limitation of solid waste littering and wastewater discharge (Falkenmark and Rockström, 2006); (H3) The use of alternative management strategies (grey water reuse, urban ecological infrastructure (Childers et al. 2019) may help reclaiming water resources, through wastewater discharge reduction, and subsequent soil physicochemical and biological statuses improvement. The implementation of the model with relevant ecological, social and social-ecological processes should allow to answer these questions, while testing these hypotheses.

We chose to use a discrete event model developed into the EDEN (Ecological Discrete Event Network) modelling framework. This framework is possibilistic, i.e. non-deterministic and non-probabilistic, and allows free exploration of all the possible system trajectories, once the model is defined and validated (Gaucherel and Pommereau 2019; Cosme et al. 2022). This modelling framework has been chosen, as compared to more traditional equation-based and statistical models, precisely for exploring all possible outcomes of a system dynamics. We will thus be able to search in the computed dynamics some reachable states, specific trajectories (i.e. state successions or pathways) and (un)stable behaviours of the Cat Ba system. In addition, the model is qualitative and rule-based, thus allowing to integrate a large number of processes, then rigorously formalized by specific rules. The definition of the model processes through expert knowledge and its subsequent validation relies on system historical knowledge (Gaucherel et al. 2024).

We will use the EDEN modelling framework to explore ways to reach the most favourable states, where tourism and water resources co-exist, along with other vital features of the system. The trajectories towards these states will be explored, identifying successive states and processes responsible for the transitions towards the target. We also wish to identify the trajectories that would lead to dead-ends, starting from the "present-day situation", precisely to avoid them. Eventually, we will try and identify the type of rules that either lead to the most favourable states or contrarily astray from it, to provide recommendations on where to lay efforts on.

Materials and methods

Study site

Cat Ba Island is the biggest island (285 km²) in the Cat Ba Archipelago, a combination of 367 limestone islands and islets located in the Northwest margin of the Gulf of Tonkin, Northern Vietnam. The island is under the authority of Cat Hai District, Hai Phong City and is known for its unique biodiversity with various rare species (Abramov and Kruskop 2012). In 1986, Cat Ba National Park was established (President of the Council of Minister decision 79-CT, 1986–03-31). In 2004, Cat Ba Archipelago was recognized as World Biosphere Reserve area by UNESCO with a total area of 264.2 km² of land and sea. The Cat Ba Island has six communes and one town with a total island population of 32,515 people in 2021 (Cat Hai Division of Statistics 2021). Human activities on Cat Ba Island dated back thousands of years. In recent years, the main activities focus on tourism, fishing, aquaculture (ponds and cages), fishery services and water transportation (Van et al. 2010). Since the recognition of UNESCO, the number of tourists visiting Cat Ba Island has continuously increased from 309,000 tourists in 2004 to a peak of 2,718,000 tourists in 2019 and 2,360,000 after the COVID outbreak and recovery (Cat Hai Division of Statistics 2022). During the period of 2017–2021, tourism accounted for roughly 70% of total production value of the island. Present political orientations of the People's Committee of Hai Phong City plan on the expansion of the tourism sector in Cat Ba Island, with 17,600 jobs creation in 2025 (decision 2732/QĐ-UBND, 2014–12-05).

Chronological features of Cat Ba Island trajectory

We considered the area of Cat Ba Island, including the near-shore area. We decided to explore the time scale between the beginning of archives on the island, to present days and ahead. Prior to the exploration of the model, we searched for documented Cat Ba system states using historical archives (decision no. 57-CP, 1977–03-11; decision 79-CT, 1986–03-31), grey literature (Ministry of Transport 2017; UNESCO 2018; Hai Phong People’s Committee 2020; Thuy 2023) and expert knowledge (Phan et al. 2019; Quan 2012) for validation. Data from various sources were collected to comprehensively determine the historical and present status of water resources management, storm-/wastewater management and identify the related stakeholders. Moreover, the data were collected from various locations on the island accounting for different socioeconomic activities, population densities and geographical characteristics. Socioeconomic statistical data of Cat Ba Island during the period of 2003–2021 were collected from Cat Hai Division of Statistics and literature (Cat Hai Division of Statistics 2021, 2022). The water supply data of Cat Ba Island in the year 2022 were collected from Cat Hai Water Supply Plant, Cat Hai District, Hai Phong (CHWSP 2022).

Features of the discrete event model

For this purpose, we explored the trajectories of the Cat Ba system as openly as possible. To make the study as unbiased as possible, we chose to implement decreasing water demand and tourism as well as their increase in the model, widening perspective compared to previous studies (Nguyen et al. 2021) and coherent with the finiteness of planetary resources (Rockström et al. 2009). To study the site dynamics, we used a discrete event modelling framework, which is qualitative (with Boolean variables, symbolized by Var + and Var−), non-deterministic (i.e. successive states are possibly numerous but are systematically the same) and possibilistic (i.e. non-probabilistic, with all possible futures computed according to the predefined rules and chosen initial state) (Gaucherel and Pommereau 2019; Cosme et al. 2022; Pommereau et al. 2022). It uses variables that are deemed to adequately describe the system given the research question asked, to further explore the possible states of the modelled system. These variables can be either active (labelled "+") or inactive (labelled "−") following the crossing of a given threshold (specified below in the ‘Construction of the Cat Ba Model’ section). A given configuration of variables is called a state of the system, or system state. The rules that change the variable states (from active to inactive, or vice versa) are defined to describe as adequately as possible the processes known or suspected to occur within the system. These rules conjugate initial conditions on variables (left part of the rule) that lead to realizations (right part of the rule), i.e. a discrete change of state, onto the same or other predefined variables. Constraints are a specific kind of priority rules explained below. The model starts with an initial system state provided by the modeller, either for its known existence or for its value regarding the research question. From this initial state, a round of application entails the following steps:

1. Rules (and constraints) that are not applicable because of unsuitable conditions (i.e. some variables are not in the appropriate state) are discarded by the algorithm;

2. Constraints that can apply are mandatorily executed by the algorithm, as a priority to all other rules;

3. For each rule with suitable conditions, two outcomes are computed: either the rule is executed, possibly leading to a new system state, or it is not executed and the system remains unchanged.

The algorithm performs then the next step by applying the same protocol as above to all the system states resulting from the previous application step. The algorithm stops when no more rule can be executed. (And all reachable system states have thus been encountered.)

Each EDEN model is based on a symbolic mathematical engine (Pommereau et al. 2022), written in Python language (Van Rossum and Drake Jr 1995) and edited with Jupyter notebook (Kluyver et al. 2016). Any modeling work is providing two main outputs: the so-called hyper-network (HN) (Gaucherel et al. 2024) of the modeled system and the state-transition graph (STG), which is a space gathering all computed states and transitions. The HN gathers the system components (variables, represented by nodes) and processes that link them (rules, represented by oriented arrows) resulting from the model definition. This graph is a simplified representation of the modeled system and it indicates whether a variable is interacting with another, is influencing or is influenced by another (Gaucherel et al. 2024). The STG is not displayed by default, as it may gather a great number (typically millions or billions) of states and transitions. For this reason, it requires to explore the computed dynamics through specific queries: CTL queries exploring reachable states of any STG (Thomas et al. 2022). These queries returned graphs in which the searched states are reached or/and not. Any EDEN model is fully tractable (i.e. analytical, in that it provides all computed states and only them) (Gaucherel and Pommereau 2019; Pommereau et al. 2022). It is possible and convenient to summarize the STG computed dynamics by automatically identifying some structural (qualitative) stabilities of the system. For example, a structural stability in the STG is defined as a set of mutually reachable states.

A simplistic model involving tourists and water resources is used for illustration purpose (Fig. 1). At the beginning, both tourists and water resources may be present (two active variables, S₀). When tourists over-consume water resources, the latter may disappear (R₁). Tourists are then still present, but there may be no more water resources (S₁). As there is no more water, tourists may leave (R₂), which may result in the absence of both water resources and tourists (S₂). As tourists have left, water resources may start restoring (R₃), tourists being still out (S₃). Eventually, as water resources are restored, tourists may come back (R₄) and the situation may return to the initial state (S₀). From the initial system state, queries could be for instance: does the state "absence of tourists and water resources" exist? (i.e. is the state "Tou-, Wat-" reachable?) Is the transition from "presence of tourists, absence of water resources" towards "presence of tourists and of water resources" possible (i.e. is the transition from "Tou + , Wat-" to "Tou + , Wat + " retrieved in the dynamics)?

Fig. 1.

2-variables, 4-rules (R₁—> R₄) possibilistic model. Tou: tourists; Wat: water resource. An active variable is represented as a white, black-circled disc. An inactive variable is represented as a grey disc. S₀—> S₃: system states

Definition of the target states

The exploration of the model will be carried out searching for trajectories and specific states within computed system states (Thomas et al. 2022). We chose to first explore the existence (presence of states) and transition (between two states) between the combination of three independent situations: (1) Preserved water resources: groundwater and surface water is present; (2) a preserved ecosystem: soil is functional, plants and animals are present in the soil and display ecological health, and plants and animals are present on the island and display ecological health; and (3) presence of tourists.

Then, among the myriad of states potentially available in the model, we decided to search for four specific states of the system, characterized by well-contrasted scenarios, to broaden the analysis and avoid its pointless systematic exploration of the model. As they are hypothetical situations (except for the present state), they will all be explicited using conditional processes, and formalized by the verb "may". We labelled them as "present", "perilous", "anthropo-centred" and "systemic" states, respectively. These four situations are unique combinations of defined variable states. We provide some context in parentheses to explicit these states and then define them in terms of variable states. We are looking for sequential links between these states, starting from the "present" one:

• Present state (Cao and Nguyen 2018; Tsai et al. 2021; Trang et al. 2022): what is presently observed on the field. Today, tourists visit the island, artificialization is significant, and wastewater discharge and solid waste littering are common. Soil physicochemical and biological statuses are degraded.

• Perilous state (El-Fadel et al. 1997; Scalenghe and Marsan 2009; Englert et al. 2013): (conjugated pollution and over-consumption may generate resources scarcity and biotope|biocenosis non-functional statuses; tourists may leave because of a degraded envionment). Every part of the environment may be severely degraded: fauna, flora, soil micro-flora/fauna and soils may be no longer functioning because of significant artificialization and massive discharge of wastewater and solid waste littering. Water resources may be scarce. Managers may be no longer active; water reuse and nature-based solutions may be not implemented. Tourists may have gone.

• Anthropo-centred state (Crini and Lichtfouse 2019): (water resources management and wastewater management may be targeted as priority issues to bring back tourists, independently of fauna, flora and soil statuses). Determined action of water resources and wastewater managers may lead to water resource presence, through the use of both energy-/matter-consuming technology and nature-based solutions. This action, synergized with wastewater discharge cut and blackwater reuse, may make tourists present again. As land use and solid waste managers may have been inactive, artificialization may still be significant and solid waste littering may still be on. Fauna, flora and soil may still suffer from severe degradation;

• Systemic state (Negrutiu 2022; Rohr et al. 2023): (water resources may be dwindling; unhealthy animals and plants populating severely degraded soils may give a poor image of the island and may bring fear for the sustainability of this system. The growing awareness that conditions for well-living for all (humans, plants, animals) require global environment restoration may create a paradigm shift, including soil unsealing and regeneration, and involvement of tourists into environmental preservation activities). Land use and solid waste managers may join forces with water resource and wastewater managers, to cut solid waste littering and raw wastewater discharge (through reuse and nature-based solutions implementation). Artificialization may be unsignificant. Fauna, flora and soil organisms may be in a functional state.

Construction of the Cat Ba Model (CBM)

The model we used did not contain any time duration (i.e. it is not temporized, no quantification such as duration or probability of transitions), making the STG only chronological (time-oriented, but not quantified in terms of duration) and remained causal (i.e. precedent and subsequent states are preserved in the dynamics). In this study, we defined 15 variables (Table 1), 71 rules (10 rules of ecological functioning–R1–R10–, 29 of social functioning–R11–R39–and 32 of social-ecological functioning–R40–R71) (Filani 1975; Butler 2000; Sophocleous 2002; Gössling 2001, 2006; Landry and Pu 2010; Simões et al. 2010; Becken 2014; Keesstra et al. 2018; Tobias et al. 2018; Zeng et al. 2018; Condon et al. 2020; Gonzalez-Flo et al. 2023; Van de Walle et al. 2023; Yu et al. 2013) and two constraints (Table S1), which appeared as a good trade-off between simplistic and over-complicated models. In the absence of explicit reference, we defined two reasons for defining a rule: general truth (well-known and acknowledged process) and hypothesis on fundamental role (action one would expect from the agent). We defined variables that express both the ecological and the social status of the system, thereafter called "ecological variables" and "social variables", respectively. The ecological variables were soil micro-fauna and flora (Mic), physical soil (Soi), macro-fauna and flora (Bio), surface water (Wra) and groundwater (Wgr) resources. The social variables were land artificialization (Art), four kinds of managers (solid waste–Msw, wastewater–Mww, water resources–Mwr-, land use–Mar), tourists (Tou), solid waste (Was), wastewater (Wwa), greywater reuse (Reu) and urban ecological infrastructure (Uei). For the ease of understanding, we classified the rules by their intricate (either ecological, social or social-ecological component of Cat Ba Island, stressing their intertwined natures) and developed their meaning (Table S1). Historical and present states of the system (Table 2) were searched for through the STG to assess its ability to faithfully model the observed (past) trajectory of Cat Ba Island. Once the Cat Ba Model (CBM) was validated, we checked its robustness with the following protocol: ideally, all EDEN model rules should be slightly altered one by one and the resulting dynamics should be checked for subsequent modifications. Given all the rules in the CBM, this combinatory task was impossible; we thus focused on altering the few remarkable rules that would appear in the system dynamics after CBM computation.

Table 1.

Model variables with their initial state and explicited meaning. Without additional information, we assumed the very existence of a threshold

Variable	Initial state	Meaning	Threshold
Bio	+	Plants and animals present on the island display ecological health	Abundance, Biodiversity index, functional traits
Mic	+	Plants and animals present in the soil display ecological health	Abundance, Biodiversity index, functional traits
Soi	+	Soil is functional (permeability, chemical composition)	Hydraulic conductivity value, organic matter content
Wgr	+	Groundwater is present	Available volume
Wra	+	Surface water is present	Available volume
Art	−	Land is artificialized below the recommended threshold	Ratio of artificialized vs natural soil
Mar	−	Managers responsible for urban planning are not active	Regulation proposition and enforcement
Msw	−	Managers responsible for solid waste are not active	Regulation proposition and enforcement
Mwr	+	Managers responsible for water resources are active	Regulation proposition and enforcement
Mww	−	Managers responsible for wastewater are not active	Regulation proposition and enforcement
Tou	+	Tourists come to visit Cat Ba island and surroundings	Number of tourists per year
Was	+	Solid waste (packing, plastic bags, organic waste) is discharged in the environment	Mass of littered solid waste per day per surface unit
Wwa	+	Wastewater is discharged in the environment	Volume of discharged wastewater per day
Reu	−	Greywater is not reused	Volume of reused greywater
Uei	−	Urban ecological infrastructures are not implemented	Number of implemented infrastructures

Table 2.

Documented CBM states (sources, see ‘Chronological features of Cat Ba Island trajectory’ paragraph). ‘*’ indicates a wild card

	1900	1957	1962	1986	1988	1994	1998	2004	2010
Bio	+	−	−	−	−	−	−	+	+
Mic	+	+	−	−	−	−	−	−	−
Soi	+	−	−	−	−	−	−	−	−
Wgr	+	+	+	+	+	+	+	+	+
Wra	+	+	+	+	+	+	+	+	−
Art	+	+	−	−	−	−	−	−	−
Mar	−	−	−	−	+	+	+	+	+
Msw	−	−	−	−	−	+	+	+	+
Mwr	−	−	−	−  −	−	−	+	+	+
Mww	−	−	−	−	−	+	+	+	+
Tou	−	−	−	+	+	+	+	+	+
Was	+	−	−	−	−	-	−	−	−
Wwa	+	−	−	−	−	−	−	−	−
Reu	+	+	−	*	*	*	*	+	+
Uei	−	−	−	−	−	−	−	−	−

Exploration of CBM dynamics

Once CBM was computed, it was purposely explored in two specific directions, to avoid everlasting and irrelevant search. We targeted two important kinds of states, within the multitude that was computed: specific configurations and specific states, that will be detailed below.

First, we chose to assemble variables in specific configurations, where variables can be defined with AND logical operator. Three configurations were defined (Fig. 4): (1) wat(er) = "Wra + AND Wgr + " indicated preserved water resources; (2) bio(diversity) = "Bio + AND Mic + AND Soi + " indicated a preserved ecosystem; and (3) tou(rist) = "Tou + " indicated the presence of tourists. States that would include a combination of these configurations were searched for, with a total of 8 (= 2³) theoretical presence/absence combinations. The states where all three configurations are absent are grouped as the NULL configuration. As the search synthesizes all CBM dynamics, each eligible configuration gathers a great number of states (tou/wat/bio + all other variables in unconstrained states).

Fig. 4.

Transitions from one configuration to another. Icons indicate either the presence of tourists, preserved water resources and preserved ecosystem. Arrows indicate transitions

Second, the CBM dynamics was explored with four specific states, the so-called present, perilous, human-centred and systemic ones described in the introduction. These situations were translated into variable states (Table 3). Using the specific exploration engine (temporal logic) provided by EDEN, we searched for their presence and transition sequences potentially leading from one situation to other ones. We then discussed them from a qualitative point of view. States were qualitatively characterized as situations; transitions will be qualitatively characterized as labels. Any transition or trajectory towards the "perilous" situation was thus labelled as "unfavourable", while any transition or trajectory from this situation was labelled as "favourable". Similarly, for any situation, a transition or trajectory that contributed to switching the variables to the state finally activated in the "systemic" situation (e.g. " + " for Tou, " + " for Mic) was labelled as "favourable". The resilience of the "present" situation was understood as the capacity for this situation, once left, to re-establish.

Table 3.

Variable states associated with the explored states

	Bio	Mic	Soi	Wgr	Wra	Art	Mar	Msw	Mwr	Mww	Tou	Reu	Uei	Was	Wwa
Present state	+	−	−	+	+	−	+	+	+	+	+	+	−	−	−
Perilous state	−	−	−	−	−	+	−	−	−	−	−	−	−	+	+
Anthropo-centred state	−	−	−	+	+	+	−	−	+	+	+	+	+	+	−
Systemic state	+	+	+	+	+	−	+	+	+	+	+	+	+	−	−

In brief, we will look after the model dynamics from four facets: (i) situations are specific states we would like to focus on, (ii) configurations are sets of states we would like to focus on, (iii) labels are transitions and/or sets of transitions (trajectories) we would like to focus on, and (iv) intricates are sets of rules we would like to focus on.

Results

The interactions (processes) between the chosen variables in the CBM model can be visualized with the HN (Fig. 2). Spatially close variables on the graph are variables who are in closer interaction. Soil, micro-fauna/flora and plants/animals (Soi, Mic and Bio) are central variables and are close to one another, as well as water resources and wastewater managers (Wgr, Wra and Mww). Urban eological infrastructures, greywater reuse and water resource managers (Uei, Reu and Mwr) are the most peripheral ones in this model. A looser group of variables appears with solid waste, urban planning managers, tourists, solid waste managers and water resource managers (Was, Mar, Tou, Msw and Mwr). Micro-fauna/flora, soil, plants/animals, groundwater, surface water, urban planning managers, wastewater managers, wastewater (Wwa), solid waste and tourists are the most connected variables with more than seven interactions each; greywater reuse, with four interactions, is the less connected variable.

Fig. 2.

CBM hyper-network. Variables are circled, links indicate relationships between variables, and arrows indicate directional influence. Green: ecological variables (Bio, Mic, Soi); blue: water resources (Wgr, Wra); turquoise: NBS techniques (Uei, Reu); red: disturbances (Art, Was, Wwa); orange: stakeholders (Mar, Msw, Mwr, Mww, Tou)

All documented states (Table 2) are retrieved within the CBM computed dynamics and are reachable (Fig. 3). More precisely, all observed states are connected, through a central cluster; system states in 1962, 1986, 1988, 1998, 2004 and 2010 are even directly (and reversely) connected. As a first and central result, the CBM is thus validated. Intermediate states between the targeted dates do not explicitly appear in the search; they are included with all the other computed states in the large green central cluster (Fig. 3). A fine-grain description of the latter was not carried out, as it was not the main point to identify all possible trajectories between past observed states.

Fig. 3.

Result of the search for historical configurations of Cat Ba SES (model validation). Arrows indicate transitions between situations. All model states other than the nine exhibited ones are gathered in the green central cluster

As the model is validated, it can be reasonably safely explored. The CBM yields 12 352 accessible states from the initial one (Table 1), strongly partitioned between a 6 164-state stability without implemented UEI (Uei-) and a 6 188-state stability with implemented UEI (Uei +), while 32 768 (= 2¹⁵) states are theoretically possible. The structural stability without UEI irreversibly leads towards the one with UEI through a single rule (R18, Table S1). The robustness of the model was subsequently examined upon modification of this rule. Changing it from "only wastewater managers are needed to implement Uei" to "wastewater AND water resource managers are needed to implement Uei" (i.e. R18 becomes "Mww + , Mwr + » Uei + ") did not change CBM dynamics (i.e. computed states are the same). Hence, the CBM is robust regarding this change of rule.

The first exploration was carried out on the states characterized by configurations involving "tourists" (tou), "water resources" (wat) and "biodiversity" (bio) (Fig. 4), i.e. where variables Wra, Wgr, Bio, Mic, Soi and Tou are in ‘ + ’ state (see “Exploration of CBM dynamics” section). All are retrieved in the STG and partially connected, as expected. The three-configuration ("tou + wat + bio") states only connect with the single-configuration states where tourists are present ("tou"; upper left downward black arrow). Exclusive states are not connected ("wat + bio" and "tou", "tou + wat" and "bio", "tou + bio" and "wat", "tou + wat + bio" and NULL). The three-configuration states bi-directionally connect to the two-configuration states, except from "tou + wat + bio" to "wat + bio" (upper right upward black arrow). Two-configuration states only connect with the NULL configuration from "wat + bio" states (lower right downward black arrow).

The second exploration was carried out on the four situations defined above (Table 3), which are all found within the STG (Fig. 5). "Present" and "perilous" situations are located in the set of states without UEI and are indirectly connected within it through a set of rules. "Anthropo-centred" and "systemic" situations are located in the other set of states, with UEI, and are also indirectly connected within it (brown, respectively green, circles in lower part on Fig. 5). A single rule (R18: Mww +  >> Uei +) leads to the irreversible transition between these two sets of states. All the rules that can lead to a shift from or towards a situation are classified in terms of social, social-ecological or ecological intricate (Table 4 and Tables S2 and S3). For instance, the "present" situation can be left by the application of R11; the "systemic" situation can be reached by the application of R10.

Fig. 5.

Search for system configurations through CBM. Black circle: "present" situation; red circle: "perilous" situation; brown circle: "anthropo-centred" situation; green circle: "systemic" situation. The downward arrow indicates the irreversible transition between the two stable state complexes, represented as grey fuzzy blocks. Upper block: without UEI; lower block: with UEI. The downward vertical arrow represents R18

Table 4.

Rules related to the system dynamics around the systemic situation. Normal font: inward—favourable—rules; bold font: outward—unfavourable—rules

Intricate	Rule	Process
Social	R24: Tou +  >  > Mar +	Environmentally committed tourists action
	R25: Tou +  >  > Msw +
	R26: Tou +  >  > Mwr +
	R27: Tou +  >  > Mww +
	R15: Msw +  >  > Was-	Managers action
	R16: Mww +  >  > Wwa-
	R18: Mww +  >  > Uei +
	R19: Mww +  >  > Reu +
	R30: Mar +  >  > Msw + , Mwr + , Mww +
	R31: Msw +  >  > Mar + , Mwr + , Mww +
	R32: Mwr +  >  > Msw + , Mar + , Mww +
	R33: Mww +  >  > Msw + , Mwr + , Mar +
	R57: Mwr + , Mww + , Msw +  >  > Wgr + , Wra +
	R58: Mww + , Msw + , Mar +  >  > Mic + , Soi + , Bio +
	R59: Tou + , Mwr + , Msw + , Mww +  >  > Wra +
	R60: Tou + , Mwr + , Msw + , Mww +  >  > Wgr +
	R43: Art-, Was-, Wwa- >  > Mic + , Soi +	Best practices
	R44: Art-, Was-, Wwa- >  > Bio +
	R45: Uei +  >  > Wgr + , Wra +
	R49: Uei + , Was- >  > Bio + , Mic + , Soi +
Social-ecological	R22: Reu +  >  > Wwa-	Best practices
	R55: Wwa-, Was- >  > Wgr + , Wra +
	R62: Reu + , Soi + , Mic +  >  > Wgr +
	R70: Bio + , Wra +  >  > Tou +	Biosphere and water resources recovery
	R71: Bio + , Wgr +  >  > Tou +
Ecological	R01: Wra + , Soi + , Mic +  >  > Wgr +	Natural recovery of water resources
	R10: Wra- >  > Wra +
	R53: Wgr +  >  > Tou +
	R54: Wra +  >  > Tou +
Social	R11: Tou +  >  > Was +	Tourists action
	R12: Tou +  >  > Wwa +
	R23: Tou +  >  > Mar-
	R28: Tou +  >  > Mwr-
Ecological	R04: Wra + , Bio +  >  > Wra-	Water over-consumption
	R05: Wgr + , Bio + , Soi +  >  > Wgr-

Discussion

Can the Cat Ba system reach a safe shore?

The answer to the initial question appears clearly: Cat Ba social-ecological system can indeed reach a situation where sustainable tourism and water resources are simultaneously achieved, in the so-called anthropo-centered situation for instance. In this situation, however, tourism may be seen as the main driver of local economy (Mejjad et al. 2022), without regard to the ecosystem status; typically, this was once achieved on Cat Ba Island when drinking water was brought from the continent to cope with a water scarcity episode (MONRE 2018) and maintain water availability for both inhabitants and tourists. This way of handling the lack of resource leaves untackled the issue of the ecosystem functionality, with great concern about the overall sustainability of ecosystem water production, not to mention surrounding soil and fauna/flora functionality (Singh et al. 2004). Navigating further, the system can reach an even healthier situation that accounts for these other needs as well, qualified as "systemic". Here, tourists and water resources not only co-exist within Cat Ba system, but also a healthy ecosystem (soil, micro-fauna and micro-flora, fauna and flora), while urban ecological infrastructures and active managers can take place. This state and contribution to the management of the SES will help sustain its water supply function while maintaining other crucial ones (biodiversity support, soil food production) for local viability (Rohr et al. 2023). The irreversible transition between the upper and lower state sets appears to be driven by the implementation of Urban Ecological Infrastructures following wastewater managers action; thus, the hypothesis of UEI implementation as agent of favourable change (Hypothesis H3) is confirmed (Ferreira et al. 2023). One of the main interests of UEI lies in the achievement of ecological functions, regardless of their various physical configurations (aquatic, terrestrial or semi-aquatic landscapes, with or without vegetation for instance (Childers et al. 2019)). Thus, the locally relevant configuration should be searched keeping in mind the objectives of ecological functionality. (For instance, if water quantity regulation is the aim in the case of this tropical island, it could be achieved with a constructed wetland for inland or coastal zones, riparian forest along a stream, lake in impervious areas or de-sealed soil.) With this in mind, a careful implementation design must be set to avoid other pitfalls, such as the governance-related ones (Cohen-Shacham et al. 2019). Finally, it is worth noting that the singular role of this rule was not detectable at the model construction phase and required the model computation to reveal.

From troubled waters to a safe shore

The rules that generate transitions towards the "present" state, once it has been left, may be interpreted as markers of the state resilience (Elmqvist et al. 2019), in our case a back and forth process that keeps Cat Ba Island in the present state; they all lead to a status improvement, though (Table S2). As the above state is not a systemic one that would lead to a perennial system functioning, the transition from the "present" to the "systemic" state requires a temporary degradation. May it be towards the "perilous" state (Table S4), or specifically related to a process leading to less "favourable" status (Table S3 and Fig. 5), it would nevertheless lead to a sustainable improvement in the long run, locally preserving ecosystem, society and people health (Negrutiu 2022). Moreover, in the CBM, the "present" state resilience lies mainly in behaviours, as the related rules refer mainly to action of managers or presence of tourists; this would strongly promote immobility, as behavioural momentum and perceived risks of change have been shown to be drivers of inaction (Gifford 2011). Only by keeping in mind the "systemic" state as an objective, temporary inconvenience can be endured. To allow seeing the more perennial trajectory behind the transient less favourable state, workshops with Cat Ba citizens and stakeholders may help to lay the scientific foundation about the issue and collectively discuss why change would be needed. For instance, initial effort to keep littering at bay could be rewarded by anticipating nicer landscapes and reduced soil pollution as well as associated health problems.

Building the ship, recruiting the crew

The close connection between ecological variables (soil, micro-fauna and flora, fauna and flora) in the CBM model reflects the importance given to the ecological functioning in its construction (Fig. 2, Tables 1 and S1). Water resources (both surface and groundwater) are located in the centre of the HN, which mirrors the question asked about the water issue in Cat Ba Island. The influence of biotope and biocenosis (outside Humans) on water resources is emphasized by the proximity of the corresponding variables (Bio, Mic, Soi, Wgr and Wra) and stresses on the many contributors of ecosystem and citizens health (Rohr et al. 2023). Likewise, the influence of wastewater on water resources, as is the case in Cat Ba (Trang et al. 2022), is highlighted by the proximity of water resources and wastewater manager (Wgr, Wra and Mww). To summarize, CBM as it was built lays great emphasis on the processes within the biotope and biocenosis, yet includes the influence of managers action and pollution types, which echoes well the questions asked in this study. The crew may be further completed, with economic and common governance variables that could be recruited. To complete the ship and for the model to become truly systemic (with the drawback of much increased complexity), chosen economic variables should account for dynamics between resources and needs, rather than offer and demand (Georgescu-Roegen 1971). This is crucial for water issue in Cat Ba Island as over-consumption was responsible for scarcity episodes (Hung 2010; Thinh 2018). Eventually, chosen governance variables should acknowledge the multi-stakeholder status of water issue in Cat Ba and the dynamic—adaptive—nature of governance regarding the multiplicity of actors and the retroactive effect of economic (among others) incentives (Folke et al. 2005).

Navigating waters from beacon to beacon

While searching for system configurations within CBM dynamics, it appears impossible for such a complex system, even simplified for the purpose of this work, to lose all of its resources in one transition (Fig. 4) (Gaucherel et al. 2020). The corresponding degradation would gradually proceed, which cannot happen in one single transition (i.e. process); this, regardless to the followed trajectory towards the state where degraded ecosystem and water resources lead to tourist absence (bottom configuration, Fig. 4). Conversely, it is impossible for a severely degraded Cat Ba Island (i.e. without water resources, tourists and with a degraded ecosystem) to recover all of the targeted resources in one transition. Changes, such as losing tourists while recovering water resources and ecosystem or losing tourists and ecosystem while recovering water resource, are not possible in a single transition, likely due to the complexity (i.e. intertwined processes) of the system. According to the model, wastewater discharge would be responsible for the simultaneous loss of water resources and ecosystem good status (downward red arrows, Fig. 4), in agreement with both general (Singh et al. 2004) and Cat Ba specific (Trang et al. 2022) literature. As expected, good statuses of ecosystem and water resources are key in the maintenance of tourist presence, as it is not possible for Cat Ba SES to lose tourists, while these two features are maintained. And while on an improvement trajectory, water resources and ecosystem health are the conditions to recover tourists (upward red arrow, Fig. 4). To summarize, while a complete degradation of Cat Ba system status requires here several steps, a recovery from a severely degraded situation reciprocally could not be achieved all at once. As this would mean stress for many organisms (including citizens) within the SES while navigating through all the states buoying this recovery trajectory, simultaneous degradation of water resources, ecosystem and tourist attendance should be avoided.

Who is the captain on the ship?

Globally, reaching or leaving the "systemic" situation seems to be a strong social(-ecological) issue, beyond anything else, as previously proposed (Ostrom 1990). Socially triggered improvement includes the autonomous and/or collective intervention of managers (also allowing to leave the perilous situation, Table S4), the adoption of best practices, e.g. reducing pollutions and artificialization, and UEI implementation leading to recovery of biosphere status and/or tourist presence (Table 4) (Van der Brugge et al. 2005). Social-ecologically triggered improvement includes best practices regarding the urban water cycle (Keesstra et al. 2018) and return of tourists following biosphere and water resources recovery. Finally, ecologically triggered improvement involves the recovery of water resources through natural evolution of the water cycle (also responsible for departure from the perilous situation). Here, hypothesis about tourism-induced degradation (H1) would not be confirmed, in the case of environmentally aware tourists, instead of hypotheses H2 and H3 that are confirmed. Conversely, the system may enter the perilous situation for various (social-)ecological reasons (Table S4): (i) biosphere degraded by artificialization and water reuse impossibility due to water resource scarcity; (ii) degraded water resources, both surface- and groundwater, by artificialization, solid waste littering or wastewater. Given the situation, a large panel of rules may need to apply for improving the situation(Rohr et al. 2023), as many of the system intricates are implied (Rohr et al. 2023). Among them, the actions of managers may be generated by a variety of reasons, either resource scarcity, pollution or degraded biocenosis (R67, R20/R21 and R63-R66, respectively; hypothesis H2). On the ground, this would mean steps are taken and enforced to thwart the noxious behaviours that cascade into socioecological degradation (soil and degradation, fauna/flora impairment, tourist departure). The last source of improvement, we cannot directly act on though, is the hydrological cycle evolution. Yet, as the indirect impact of Human on hydrological cycle is clearly established, an obvious improvement trail is the evolution of our practices (Masson-Delmotte et al. 2007).

A social-ecosystem cannot be qualified as healthy if water resources and/or biosphere are degraded; neither can it be called healthy if tourists massively depart, as we acknowledged that tourism (in moderate amount) can be a cause for SES economic thriving, especially on Cat Ba Island, where tourists are presently counted by the million every year. Reducing the dependency of the island on this type of tourism and encouraging SES-caring types of tourism (Ateljevic 2020) may help conserving tourism economic benefits while alleviating some of its environmental burdens (direct ecosystem degradation or impairing managers action, H1 hypothesis).

Eventually, this multiplicity of processes and stakeholders, related to the system complexity and the water-tourism issue (actually any systemic issue), means that while the trajectory from the Cat Ba Island “present” state towards the “systemic” scenario is clearly outlined by the model, the system management remains everything but straightforward. Multiple stakeholders are involved, with sometimes diverging immediate objectives. Yet what could be sometimes seen as the “tragedy of the commons” (Hardin 1968) is not mandatory, and many other examples show that these commons can be altruistically managed by local stakeholders (Ostrom 1999) supported by appropriate local regulation.

Need for a compass

To precise CBM recommendations beyond the broad orientations provided here (Figs. 4, 5) that provide a first compass (Negrutiu 2022; Rohr et al. 2023), estimates for reduction of water consumption, reduction of artificialization, increase of UEI and wastewater reuse are needed for Cat Ba Island. Adding this operational layer to the work would require gathering or carrying out a significant amount of field work, e.g. bringing all stakeholders around the table and explore the model consequences for local actions. Some of this work has already been undertaken, in the form of water sampling and analyses, island inhabitants’ surveys, meteorological and land use data (Pham et al., unpublished results). Still the orientation (preceding quantification) is crucially needed to know where we want to berth, before navigating randomly towards an ill-defined destination (keeping in mind Goodhart’s Law that warns against the counterproductive effect of measures mistaken for targets (Strathern 1997)) or generating frustration or reluctance on board the ship. Similarly, qualitative extrapolation for other touristic cities on tropical islands could be relevant, but numbered recommendations would require local data and knowledge (Mai and Smith 2015, 2018).

Conclusion: navigating troubled waters to berth on a safe shore

In this study, we showed it is possible, starting from the "present" situation, to reach some system states combining sustainable water resources and tourism. UEI implementation appears to be a main agent of change for this evolution. Moreover, Cat Ba system can reach a "systemic" situation where a healthy ecosystem co-exists with sustainable water resources and tourism, thanks to various stakeholder (i.e. managers, environmentally committed tourists) actions. Yet, on the way towards this favourable situation, it may be possible to temporarily lose ecosystem and water resources, which would be due to environmental discharge of wastewater. As it is impossible to regain both water resources and ecosystem favourable status in a single transition, this type of environmental degradation should be especially scrutinized. More generally, leaving the "present" situation may lead to temporarily uncomfortable situation, but then the ultimate "systemic" objective should be kept in mind to navigate troubled waters towards a safe shore.

Supplementary Information

Below is the link to the electronic supplementary material.

Biographies

Paul Bois

is an Associate Professor at the ENGEES/ ICube. His research interests include urban ecology and river thermal modelling.

Le Anh Pham

is a lecturer at the University of Science and Technology of Hanoi (USTH). His research interests include natural based water treatment systems and water treatment modeling.

Cédric Gaucherel

is a Senior Scientist at INRAe. His research interests include theoretical ecology.