Enhancing Environmental Cognition Through Kayaking in Aquavoltaic Systems in a Lagoon Aquaculture Area: The Mediating Role of Perceived Value and Facility Management

Abstract

Tainan’s Cigu, located on Taiwan’s southwestern coast, is a prominent aquaculture hub known for its extensive ponds, tidal flats, and lagoons. This study explored the novel integration of kayaking within aquavoltaic (APV) aquaculture ponds, creating a unique hybrid tourism landscape that merges industrial land use (aquaculture and energy production) with nature-based recreation. We investigated the relationships among facility maintenance and safety professionalism (FM), the perceived value of kayaking training (PV), and green energy and sustainable development recognition (GS) within these APV systems in Cigu, Taiwan. While integrating recreation with renewable energy and aquaculture is an emerging approach to multifunctional land use, the mechanisms influencing visitors’ sustainability perceptions remain underexplored. Using data from 613 kayaking participants and structural equation modeling, we tested a theoretical framework encompassing direct, mediated, and moderated relationships. Our findings reveal that FM significantly influences both PV (β = 0.68, p < 0.001) and GS (β = 0.29, p < 0.001). Furthermore, PV strongly affects GS (β = 0.56, p < 0.001). Importantly, PV partially mediates the relationship between FM and GS, with the indirect effect (0.38) accounting for 57% of the total effect. We also identified significant moderating effects of APV coverage, guide expertise, and operational visibility. Complementary observational data obtained with underwater cameras confirm that non-motorized kayaking causes minimal ecological disturbance to cultured species, exhibiting significantly lower behavioral impacts than motorized alternatives. These findings advance the theoretical understanding of experiential learning in novel technological landscapes and provide evidence-based guidelines for optimizing recreational integration within production environments.

1. Introduction

The global push toward renewable energy development, coupled with increasing competition for land use and the urgency of climate change mitigation, has catalyzed innovative approaches to space utilization [1]. Aquavoltaic (APV) systems, integrating solar photovoltaic installations with aquaculture production, have emerged as a promising strategy to address these challenges [2]. APV installations, often described as “top-level electricity generation, bottom-level aquaculture,” represent a paradigm shift in the conceptualization and management of productive water bodies [3]. Taiwan, with its ambitious energy-transition goals targeting a “nuclear-free homeland” by 2030, has actively promoted renewable energy development, particularly solar power [4]. As of September 2023, Taiwan’s solar installation capacity reached 11.7 gigawatts, with APV systems forming a key component of this renewable energy portfolio [5].

Coastal areas with extensive aquaculture infrastructure, such as Cigu District in Tainan, have become focal points for this transformation. Cigu District, renowned for its extensive aquaculture ponds, tidal flats, and lagoons, is a significant aquaculture center experiencing substantial APV development [6]. Projects such as the Tainan Cigu Solar Power Plant and Star Aquaculture’s “Sun Operation Project” exemplify this transition [7]. The rationale for promoting APV in Cigu includes revitalizing traditional aquaculture, enhancing land-utilization efficiency, and creating diversified income streams for local communities [8]. However, the APV transformation in Cigu represents more than a technological shift; it constitutes a complex socio-ecological experiment. Driven by national energy policies [9], APV development is viewed as a solution to land-use efficiency challenges and economic revitalization of aquaculture areas [5]. Yet, Cigu’s unique social structure, characterized by a high proportion of tenant farmers (approximately 80%), and its ecological sensitivity create implementation challenges that exceed mere technological considerations [10]. Concerns regarding tenant farmers’ rights, ecological impacts on species such as the black-faced spoonbill, and aquaculture productivity maintenance have generated significant local apprehension and resistance [11]. While the “70% yield” claim is crucial for the economic and social acceptance of APV, it faces scrutiny due to the lack of peer-reviewed, long-term, multi-species data from commercial-scale operations in Cigu [9], exacerbating local concerns [11]. The FRI itself acknowledges that on-site management issues remain to be resolved and that establishing a universal cost–benefit model is challenging [12]. Additionally, while the temperature-regulation function of APV panels generally benefits summer heat stress mitigation [5], it may have subtle effects on optimal temperature requirements for specific species and phytoplankton dynamics, ultimately affecting productivity and overall pond ecosystem health [13].

Kayaking, as a popular non-motorized water recreation activity, aligns with low-carbon tourism principles. Taiwan offers numerous kayaking experiences across diverse aquatic environments, including coastal areas, lakes, and even rice paddies [14]. The Taiwanese government has encouraged fishermen to develop low-carbon green tourism activities, such as kayaking, as potential supplementary income sources—a particularly relevant consideration for areas like Cigu undergoing APV transition [15]. The introduction of kayaking activities in APV aquaculture ponds represents a novel intersection of industrial land use (aquaculture and energy production) and nature-based recreation, creating a unique hybrid tourism landscape. While APV sites are essentially modified industrial–agricultural landscapes [16], kayaking is typically associated with natural or semi-natural water bodies [17]. The integration of kayaking into APV ponds merges these distinct uses, potentially offering a “novel” experience [18] while raising questions about authenticity and ecological compatibility. The core issue addressed in this research concerns whether kayaking activities may cause ecological disturbance and stress to the main cultured species in Cigu’s APV ponds—hard clam (Meretrix lyrata) and whiteleg shrimp (Penaeus vannamei). Potential disturbance mechanisms include the physical presence of kayaks and paddlers, water column turbulence and sediment resuspension from paddle strokes, casting of moving shadows, and noise and vibration. There is a critical knowledge gap regarding such impacts in APV-modified aquaculture environments, as existing studies have not specifically addressed this context.

The unique environment of APV ponds (partial shading and presence of structures) may alter the baseline stress levels of cultured species, potentially making them more or less sensitive to additional stressors from kayaking activities compared to traditional open ponds [5]. APV systems modify light conditions, water temperature, and potentially water flow [5]. These altered conditions likely already affect the physiology and behavior of the clams and shrimp [13]. Therefore, the introduction of kayaking as an additional novel stimulus (shadows, movement, and turbulence) needs to be assessed against this already modified baseline, meaning that findings from traditional ponds may not be directly applicable. A frequently cited claim by the Fisheries Research Institute, Mod [19], states that APV systems with 40% shading can maintain at least 70% of the original aquaculture productivity [19]. Specific FRI trials have shown that hard clams (Meretrix sp.) performed well under 40% shading during summer months [19], though growth was slower in autumn and winter due to reduced algal production [12]. The validity of this claim for Meretrix lyrata and Penaeus vannamei under actual operating conditions in Cigu APV ponds (rather than merely simulated environments) requires verification. The shading provided by solar panels helps regulate water temperature, mitigating summer heat stress and potentially reducing evaporation [1]. A study in southeastern China found that floating solar panels reduced water temperature by an average of 1.5 °C [20].

Given the rapid development of APV landscapes and their complex interactions, particularly the insufficiently studied ecological impacts of introducing recreational activities like kayaking, a deeper understanding is urgently needed. This requires rigorously reviewing existing literature on APV systems, ecological impacts of non-motorized boat recreation, and the sensitivity of Meretrix lyrata and Penaeus vannamei to relevant disturbances; integrating contextual information on APV development in Cigu, including local concerns and policy drivers; identifying potential ecological impact pathways of kayaking activities on target cultured species in APV ponds; exploring the potential and challenges of integrating kayaking tourism as a sustainable livelihood diversification strategy in Cigu’s APV systems; assessing the recreational benefits and visitor experience aspects of kayaking in these novel, modified environments. This research aims to inform sustainable management practices for regions like Cigu in balancing energy production, food security, local livelihoods, and conservation. Recreational kayaking has gained popularity as an outdoor activity that connects participants with nature while providing physical and psychological benefits. The quality of facilities, safety measures, and professional instruction (FM) are critical factors that may impact participants’ perceived value of the training experience (PV). Additionally, as environmental awareness grows, understanding participants’ recognition of green energy and sustainable development concepts (GS) becomes increasingly important in outdoor recreational contexts.

Previous literature suggests interconnections between these constructs, but empirical research examining their relationships, particularly in kayaking recreation, remains limited. This study aims to address this gap by investigating how facility maintenance and safety professionalism influence participants’ perceived value of kayaking training, and how both factors contribute to participants’ recognition of green energy and sustainable development concepts.

2. Literature Review

2.1. Aquavoltaics (APV): Global and Local Perspectives

Aquavoltaics (APV) integrates solar photovoltaic (PV) electricity generation with aquaculture within the same water body [1]. This “top-level electricity generation, bottom-level aquaculture” model aims to achieve dual utilization efficiency of land/water resources [16] and represents a branch of agrivoltaics (agriculture and solar coexistence) [2].

The benefits of APV systems include enhanced land-utilization efficiency, enhancing land-utilization efficiency by maximizing the productivity per plot [2]; green energy production, reducing dependence on fossil fuels and lowering carbon emissions [1] (for example, the TW fishery–complementary PV project in mainland China is projected to significantly reduce carbon dioxide emissions over its lifecycle) [16]; improved aquaculture environments, as shading from solar panels helps regulate water temperature, reduce evaporation, and stabilize water quality [5], particularly beneficial for shade-preferring aquatic species (e.g., white shrimp and crabs) [16]; and economic advantages, offering potential for increased/diversified income for farmers/landowners through electricity sales and continued aquaculture [6] (in a case study from mainland China, aquaculture accounted for 25% of project income [16], with smart aquavoltaics enhancing aquaculture value and profitability) [5].

However, APV systems also face significant challenges: high initial investment costs, requiring substantial upfront capital for solar panel installation and infrastructure [2]; technical expertise requirements, necessitating specialized knowledge in both solar technology and aquaculture [5]; potential reduced crop/aquaculture yields, as shading reduces light for photosynthesis, potentially affecting productivity if improperly managed or with unsuitable species [2] (some agrivoltaic studies have reported yield reductions of 3–62% in over 80% of tested crops) [21]; ecological concerns, including impacts on biodiversity, habitat alteration, and landscape aesthetics (this can include changes to water temperature regulation, leading to potential shifts in optimal conditions for specific aquatic species) [5]; and social acceptance and conflict issues, involving land tenure, community acceptance, and impacts on existing livelihoods, as evidenced in Cigu [5]. The physical presence of solar panels and their supporting infrastructure modifies the natural habitat. This can include changes to water temperature regulation, leading to potential shifts in optimal conditions for specific aquatic species. Additionally, the construction and presence of these systems can alter benthic communities and sediment structures. APV system types include floating, pond embankment, indoor facility rooftop, and smart aquavoltaic greenhouse models [5].

Discourse on APV benefits often emphasizes technical aspects and broad environmental gains (carbon reduction, land efficiency, etc.), but the realization of these benefits at the local scale heavily depends on addressing socioeconomic challenges and ensuring ecological compatibility—factors often underestimated in initial planning. While APV offers clear macro-benefits such as green energy and land efficiency [16], it also presents significant challenges, including high costs, technical barriers, and potential yield reductions [2]. The Cigu case particularly highlights how social issues (such as tenant displacement) [10] and ecological concerns (such as the black-faced spoonbill’s habitat) [9] can overshadow or offset anticipated benefits without active management. This suggests a gap between the idealized model of APV and the complexity of its on-the-ground implementation. Additionally, “smart” APV integrating Internet of Things, artificial intelligence, and real-time monitoring technologies [5] offers pathways to mitigate some technical and productivity challenges but may introduce new issues such as data security, higher costs, and increased technical skill requirements, potentially widening the gap between large corporate adopters and small-scale aquaculturists.

2.2. Environmental Co-Benefits: Temperature Regulation, Water Quality, and Shading Dynamics

Solar panels provide partial shading, helping to regulate water temperature by reducing direct sunlight exposure [5], preventing excessive summer water temperatures [19], and reducing evaporation [16]. A study in southeastern China found that solar panels reduced the average water body temperature by 1.5 °C [20]. Shading simultaneously affects water quality, potentially limiting algal blooms by restricting light [16], which can be beneficial in eutrophic waters. The aforementioned chinese study [20] found that solar power generation led to significant reductions in chlorophyll-a (72–94%) and increases in dissolved oxygen (DO) concentrations (8–24%), along with decreased concentrations of certain nutrients. However, excessive shading (e.g., coverage exceeding 75%) significantly reduces plankton, which serves as a food source for some cultured species [16]. The impact on dissolved oxygen may be more complex; while reduced photosynthesis may lower DO, lower water temperatures can hold more oxygen, and reduced algal presence may lower organic loading and oxygen demand.

The claim of “40% shading maintaining 70% yield” represents a core issue [19]. FRI studies indicate that for hard clams, 40% shading is beneficial in summer but may slow growth in cooler months when algae (their food source) are less abundant [19]. Different crops/species respond differently to shading; some shade-tolerant species (e.g., lettuce, chiltepin peppers, and certain tomato varieties) [2] benefit, while those requiring ample sunlight (e.g., grains and corn) [2] may show yield reductions. Species tested by FRI include milkfish, hard clams, tilapia, golden sea bream, and spotted sea bass [12]. In Cigu, hard clams (Meretrix sp.) are a primary cultured species, depending on sunlight-generated algae for food [10]. The impact of shading on algae (clam food) and consequently on clam growth is a key consideration [12]. Research indicates that at 40% shading, water quality begins to change, with more pronounced changes at 70% shading, affecting benthic community composition [12].

The optimal shading percentage for APV is highly dependent on specific species and seasonal factors, likely requiring dynamic management rather than fixed coverage to simultaneously maximize aquaculture and energy yields. The “40% shading maintaining 70% yield” claim is likely an oversimplification. FRI research has shown that hard clams benefit from 40% shading in summer but not necessarily in cooler months [19]. Different species have different light and temperature requirements [2]. Algal production, critical for filter-feeding animals like hard clams, directly correlates with light and temperature [12]. A fixed shading regime may not be optimal across seasons or for multi-trophic aquaculture systems with diverse species requirements (such as the Sun Operation Project cultivating white shrimp, mullet, and milkfish) [5]. Furthermore, APV’s impact on water quality is a double-edged sword: Reducing algal blooms may be beneficial, but substantially reducing primary productivity could negatively affect filter-feeding aquaculture species and overall pond ecosystem health. Solar panel shading reduces solar radiation, causing chlorophyll-a reduction and potentially decreasing harmful algal bloom occurrence [20]. However, phytoplankton form the foundation of many aquaculture pond food webs, particularly important for species like hard clams [12]. A dramatic reduction in phytoplankton (e.g., when solar panel coverage exceeds 75%) [3] may cause food shortages for cultured species. This creates a delicate balance between managing excessive algal growth and ensuring a sufficient food supply.

2.3. Ecological Footprint of Kayaking and Non-Motorized Water Recreation

Recreational activities such as boating, fishing, and swimming can negatively impact biodiversity and ecosystem function [22]. Boat navigation and shoreline use behaviors show consistent negative ecological impacts across different levels of biological organization—individual, population, and community [22]. Shoreline activities may lead to trampling, reducing vegetation cover and compacting soil [22], while boat navigation can also reduce vegetation cover [23]. Leisure boat anchoring causes major impacts on seagrass beds (e.g., Posidonia oceanica) [24]. Although kayaks typically do not anchor in shallow aquaculture ponds, the principles of physical disturbance to benthic habitats still apply. Negative effects from general recreational activities are strongest on invertebrates and plants [25].

Boat activities, including kayak paddling, can cause sediment resuspension, increasing water turbidity [22]. This affects light penetration, covers benthic organisms, and releases nutrients or pollutants from sediments [23]. Experiments have shown that sediment resuspension alters water column nutrient concentrations (e.g., ammonium, nitrite, and phosphate) [26]. The increased turbidity from sediment resuspension may also inhibit phytoplankton growth [27]. Recreational disturbances can negatively affect wildlife, causing bird flushing [22] or stress responses in aquatic animals. While many studies have focused on motorized boats, some impacts are common. In certain studies, non-motorized boats such as kayaks constitute a smaller proportion of recreational fleets (e.g., approximately 2% in Mediterranean studies) [24], but their localized impact in sensitive or enclosed environments like aquaculture ponds could be significant.

Though kayaking is often considered a “low-impact” activity, when conducted in shallow, enclosed aquaculture ponds, it may cause chronic low-intensity disturbances (e.g., sediment resuspension and visual cues) with cumulative effects on sensitive benthic species like clams and shrimp that remain unclear and potentially underestimated. In shallow ponds, paddle strokes near the bottom may disturb fine-grained sediments [22]. Hard clams and whiteleg shrimp are benthic or near-benthic species sensitive to water and substrate conditions [13]. Repeated, albeit low-intensity disturbances over time may lead to chronic stress, reduced feeding efficiency (due to turbidity or avoidance behavior), or habitat degradation. Meta-analyses [22] indicate that invertebrates are particularly vulnerable to the impacts of recreational activity. Additionally, regular kayaking may alter APV pond nutrient dynamics and primary productivity through sediment resuspension, potentially offsetting or complicating APV shading effects. APV shading aims to control algal growth and water temperature [5]. Kayak-induced sediment resuspension releases nutrients (nitrogen and phosphorus) from sediments into the water column [23]. This nutrient release may stimulate algal growth, or, conversely, increased turbidity may limit the light required by algae [27]. This creates a complex interaction between APV structural effects and recreational activity impacts on pond ecology (Table 1).

Kayaking activities may disturb aquatic environments through multiple mechanisms. For physical disturbance, direct contact of paddles or hulls with benthic organisms or substrates, while potentially minimal with careful operation, may cause trampling if entering/exiting kayaks from within ponds. Human wading/trampling affects benthic communities (a general inference lacking specific data for kayaking in aquaculture ponds, though [28] mentions direct disturbance from harvesting equipment). Sediment resuspension represents a key potential impact pathway for kayaking activities, as paddling in shallow water disturbs bottom sediments, increasing turbidity and potentially releasing adsorbed nutrients or pollutants from sediments [22].

Regarding sensory cues, kayaks passing overhead create moving shadows and visual disturbances. Certain aquatic organisms, including mollusks and crustaceans, are known to respond to light changes and visual stimuli [29]. Bivalves may close their valves in response to sudden stimuli [30]. Nudibranchs (gastropod mollusks) show preferences for dark/shadowed areas and respond to distant visual targets [31]. Cleaner shrimps respond to visual signals from symbiotic fish (such as color changes) [32]. Although quieter than motorized boats, kayaks still produce underwater sounds and vibrations from paddling and hull movement. Bivalves can detect and respond to sounds and vibrations [29]; mussels close their valves at the onset of pulsed sound waves, coquina clams (Donax variabilis) respond to wave sounds [29], and anthropogenic noise causes stress responses in bivalves [29]. Kayak-generated boat wakes are minimal compared to motorized vessels [22], but water displacement still occurs.

Table 1. Documented ecological impacts of non-motorized recreational vessels (e.g., kayaks) on aquatic organisms and habitats.

Impact Type	Specific Manifestations	Affected Organisms/Habitats	Potential Mechanisms	Key References
Sediment Resuspension and Increased Turbidity	Water turbidity, reduced light penetration, increased suspended solids	Benthic invertebrates, filter feeders, aquatic plants, phytoplankton, water quality	Paddle disturbance, hull movement disturbing bottom sediments	[22]
Physical Disturbance to Benthic Organisms	Direct contact causing injury or displacement, habitat structure alteration, trampling (entry points)	Benthic invertebrates (e.g., mollusks, crustaceans), aquatic plants, sediment structure	Paddle or hull contact with benthos, trampling during entry/exit	[22]
Wildlife Disturbance/Stress	Behavioral changes (e.g., avoidance, startle responses, feeding interruption), physiological stress responses, altered habitat use	Fish, birds, aquatic mammals, large invertebrates	Visual stimuli (moving objects, shadows), noise/vibration, human proximity	[33]
Water Quality Alterations	Nutrient release (from resuspended sediments), resuspension of contaminants, dissolved oxygen changes (indirect effects)	Water chemistry, organisms dependent on specific water quality conditions	Sediment disturbance releasing pore water and adsorbed substances	[34]
Effects on Aquatic Vegetation	Physical damage (leaf breakage, root disturbance), coverage reduction (due to boat travel paths or entry point activities)	Submerged plants, emergent vegetation, attached algae	Direct contact with paddles or hull, flow disturbance, human trampling	[22]

Table 1. Documented ecological impacts of non-motorized recreational vessels (e.g., kayaks) on aquatic organisms and habitats.

Impact Type	Specific Manifestations	Affected Organisms/Habitats	Potential Mechanisms	Key References
Sediment Resuspension and Increased Turbidity	Water turbidity, reduced light penetration, increased suspended solids	Benthic invertebrates, filter feeders, aquatic plants, phytoplankton, water quality	Paddle disturbance, hull movement disturbing bottom sediments	[22]
Physical Disturbance to Benthic Organisms	Direct contact causing injury or displacement, habitat structure alteration, trampling (entry points)	Benthic invertebrates (e.g., mollusks, crustaceans), aquatic plants, sediment structure	Paddle or hull contact with benthos, trampling during entry/exit	[22]
Wildlife Disturbance/Stress	Behavioral changes (e.g., avoidance, startle responses, feeding interruption), physiological stress responses, altered habitat use	Fish, birds, aquatic mammals, large invertebrates	Visual stimuli (moving objects, shadows), noise/vibration, human proximity	[33]
Water Quality Alterations	Nutrient release (from resuspended sediments), resuspension of contaminants, dissolved oxygen changes (indirect effects)	Water chemistry, organisms dependent on specific water quality conditions	Sediment disturbance releasing pore water and adsorbed substances	[34]
Effects on Aquatic Vegetation	Physical damage (leaf breakage, root disturbance), coverage reduction (due to boat travel paths or entry point activities)	Submerged plants, emergent vegetation, attached algae	Direct contact with paddles or hull, flow disturbance, human trampling	[22]

2.4. Hypothesis Development

The integration of recreational activities within aquavoltaic (APV) systems represents a novel approach to multifunctional land use that may enhance the sustainability transition of coastal aquaculture regions. Based on an extensive review of the literature encompassing aquavoltaic development, recreational ecology, visitor experience, and environmental perception, we propose a conceptual framework examining the relationships between facility maintenance and safety professionalism (FM), perceived value of kayaking training (PV), and green energy and sustainable development recognition (GS).

2.4.1. Facility Quality and Safety Considerations in Aquavoltaic Recreational Environments

The quality and maintenance of recreational facilities significantly influence visitor satisfaction and perceived value [18,35]. In the context of APV systems, where infrastructure combines energy production with aquaculture, facility maintenance takes on heightened importance due to both safety considerations and operational complexities [5]. Prior research on agri-tourism and aqua-tourism has demonstrated that well-maintained facilities and professional safety protocols enhance visitor perceptions of experience quality [36].

Furthermore, kayaking within modified environments such as APV ponds requires specialized infrastructure adaptations and safety considerations beyond those encountered in natural waterways [6]. The visible professionalism of staff, clarity of safety protocols, and physical quality of facilities serve as tangible cues that influence visitors’ pre-experience expectations and post-experience evaluations [37]. Therefore, we hypothesize that higher levels of facility maintenance and safety professionalism will positively influence participants’ perceived value of kayaking training experiences in APV environments. The following hypothesis guides our investigation:

H1. 

Facility maintenance and safety professionalism (FM) has a significant positive effect on the perceived value of kayaking training (PV).

2.4.2. Experiential Learning and Environmental Knowledge Acquisition in Sustainable Recreation Contexts

Experiential learning theory suggests that direct engagement with sustainable technologies and systems can enhance environmental knowledge acquisition and value recognition [38]. Kayaking within APV systems provides visitors with firsthand exposure to renewable energy infrastructure combined with food production systems, potentially creating powerful contextualized learning opportunities [18].

Studies in environmental education have demonstrated that the perceived value of educational experiences acts as a key determinant of knowledge internalization and attitude formation [36]. When visitors perceive high value in their kayaking training experience, they are more likely to actively engage with the educational components regarding green energy and sustainable development principles embedded within the experience [37]. Research on agricultural heritage tourism has shown that experiences perceived as valuable significantly enhance visitors’ recognition and appreciation of sustainability practices [38]. Accordingly, we hypothesize that a higher perceived value of kayaking training will positively influence participants’ recognition of green energy and sustainable development principles. Thus, we propose the following hypothesis:

H2. 

Perceived value of kayaking training (PV) has a significant positive effect on green energy and sustainable development recognition (GS).

2.4.3. Environmental Perception and Infrastructure Quality in Technological Landscapes

The physical manifestation of sustainability principles through well-maintained facilities can directly influence environmental attitudes and recognition [39]. In APV systems, the visible integration of renewable energy infrastructure with aquaculture represents a tangible demonstration of sustainable development principles in action [2]. Research on landscape perception indicates that visitors’ interpretations of technological landscapes are significantly influenced by maintenance standards and professional presentation [40].

Additionally, safety professionalism in recreational settings creates cognitive space for visitors to engage with educational content rather than focusing on concerns about personal safety [41]. By providing a secure environment for exploring novel technological systems, high-quality facility maintenance and safety protocols may enhance visitors’ receptivity to sustainability messaging [35]. Several studies have documented the importance of physical infrastructure quality in shaping visitors’ environmental perceptions at renewable energy sites [2]. Therefore, we hypothesize that facility maintenance and safety professionalism will directly and positively influence green energy and sustainable development recognition. Accordingly, we propose the following hypothesis:

H3. 

Facility maintenance and safety professionalism (FM) has a significant positive effect on green energy and sustainable development recognition (GS).

2.4.4. Mediating Role of Perceived Value in Experiential Environmental Education

While facility quality may directly influence sustainability perceptions, theoretical frameworks in experiential education suggest that this relationship is likely mediated by the perceived value of the educational experience itself [42]. The stimulus–organism–response (S–O–R) model indicates that environmental stimuli (facility maintenance) influence internal states (perceived value), which then determine behavioral and cognitive responses (sustainability recognition) [35].

Research on interpretive tourism experiences has demonstrated that physical environment quality affects learning outcomes primarily through its influence on experience quality perception [18]. The perceived value of the experience serves as a critical cognitive pathway through which external stimuli are processed and translated into knowledge acquisition and attitude formation [36]. This mediating relationship has been documented in various educational tourism contexts, including agricultural heritage sites [37] and marine tourism [36]. Given these theoretical foundations and empirical precedents, we hypothesize that the perceived value of kayaking training will mediate the relationship between facility maintenance/safety professionalism and green energy/sustainable development recognition. That is,

H4. 

Perceived value of kayaking training (PV) mediates the relationship between facility maintenance and safety professionalism (FM) and green energy and sustainable development recognition (GS).

2.4.5. Moderating Effects in the Relationship Between Facility Quality, Perceived Value, and Sustainability Recognition

The complex socio-ecological context of APV systems suggests that the relationships between our core variables may be contingent upon various moderating factors [43]. As evidenced in

Table 2. Control variable effects in the final SEM model.

Control Variable	Path to PV		Path to GS
	β (SE)	p-value	β (SE)	p-value
Environmental attitude	0.16 (0.04)	0.002 **	0.12 (0.03)	0.011 *
Age	0.05 (0.05)	0.348	0.08 (0.04)	0.087
Gender	0.07 (0.05)	0.187	0.03 (0.04)	0.542
Education level	0.09 (0.05)	0.094	0.14 (0.05)	0.006 **
Prior experience	0.11 (0.05)	0.029 *	0.06 (0.04)	0.213

Note: * value significant at 95%; ** value significant at 99%.

, while the correlations between FM and PV (r = 0.68, p < 0.001) and between PV and GS (r = 0.73, p < 0.001) are substantial, the variability in these coefficients indicates potential boundary conditions that may strengthen or weaken these relationships. Environmental conditions such as weather, water clarity, and visibility of aquaculture/energy production processes may moderate the relationship between facility maintenance and perceived value by influencing the salience of infrastructure quality [20]. This moderation potential is supported by the correlation patterns observed between environmental attitudes and our primary constructs (r ranging from 0.31 to 0.47, p < 0.001), suggesting that contextual factors may significantly influence how participants perceive and evaluate their experiences.

Visitor characteristics, including prior environmental knowledge, recreational experience level, and demographic factors, may moderate the relationship between perceived value and sustainability recognition by affecting information processing capacity and receptivity [37]. The variation in construct means (ranging from 3.79 to 3.95) further suggests that individual differences may play a role in how these relationships manifest. Operational factors such as visitor density, guide-to-participant ratio, and duration of experience may moderate multiple pathways in the model by altering the quality and intensity of the educational experience [35]. Additionally, APV system characteristics, including panel design, coverage percentage, and visible aquaculture activity, may moderate relationships by affecting the visual prominence and interpretability of sustainability features [44]. This is consistent with the principles of social–ecological systems theory [43] and observed correlation patterns. Thus, we propose the following hypothesis:

H5. 

Environmental conditions, visitor characteristics, operational factors, and APV system characteristics will moderate the relationships between (a) facility maintenance and safety professionalism and perceived value, and (b) perceived value and green energy and sustainable development recognition.

2.4.6. Theoretical Integration

Drawing from Ostrom’s [43] social–ecological systems (SES) framework, we conceptualize APV recreational environments as complex adaptive systems comprising the following:

(1)

Resource system (RS): Aquavoltaic infrastructure and aquaculture ponds.

(2)

Resource units (RUs): Recreational experiences and sustainability learning opportunities.

(3)

Governance system (GS): Facility management and safety protocols.

(4)

Users (U): Recreational participants with varying characteristics.

This SES lens directly informs our hypotheses, as follows:

H1: (FM → PV): Governance system quality (facility management) influences user valuation of resource units (perceived experience value). SES theory posits that institutional quality shapes resource-utilization outcomes.

H2: (PV → GS): User experience with resource units influences broader system appreciation (sustainability recognition). This reflects the SES principle that direct resource interaction shapes conservation attitudes.

H3: (FM → GS): Governance system quality directly influences sustainability recognition, representing the SES pathway where institutional visibility affects user understanding of system principles.

H4: (Mediation): The indirect pathway (FM → PV → GS) represents the complete SES interaction cycle where governance influences experience, which shapes system appreciation.

H5: (Moderation): Environmental and user characteristics moderate relationships, reflecting SES heterogeneity principles, where the context shapes system dynamics.

3. Methodology

3.1. Conceptual Framework

This study investigated the relationships between facility maintenance and safety professionalism (FM), perceived value of kayaking training (PV), and green energy and sustainable development recognition (GS) using structural equation modeling (SEM). Based on previous literature, we hypothesized that FM positively influences PV (H1), PV positively influences GS (H2), FM positively influences GS (H3), and PV mediates the relationship between FM and GS (H4). Additionally, we explored the potential moderating effects between these variables (H5). Based on these hypotheses, we proposed an integrated conceptual model depicting the direct, mediating, and potential moderating relationships between facility maintenance and safety professionalism (FM), perceived value of kayaking training (PV), and green energy and sustainable development recognition (GS). This model guides our empirical investigation and provides a foundation for understanding the educational potential of recreational activities within transformed aquavoltaic landscapes.

3.2. Measurement

All constructs were measured using multi-item 5-point Likert scales (1 = strongly disagree to 5 = strongly agree) to ensure adequate response distribution and measurement sensitivity [45]. The facility maintenance and safety professionalism (FM) construct was operationalized using nine items across three dimensions: facility quality (FM1, 3 items), safety procedures (FM2, 3 items), and instructor professionalism (FM3, 3 items), adapting scales from previous research on recreational facility assessment [35,36]. The items assessed aspects such as equipment condition, safety protocol clarity, and instructor competence within the context of aquavoltaic environments. The perceived value of kayaking training (PV) was measured through eight items across four dimensions: functional value (PV1, 2 items), emotional value (PV2, 2 items), social value (PV3, 2 items), and educational value (PV4, 2 items), drawing from established perceived value frameworks [18] but contextually adapted to kayaking experiences in modified aquatic landscapes. Green energy and sustainable development recognition (GS) was assessed using five items examining participants’ acknowledgment of the importance of renewable energy, understanding of sustainable water resource management, and appreciation of integrated energy–aquaculture systems [46,47].

The initial instrument underwent rigorous content validation through expert review with five specialists in recreational tourism, environmental education, and aquavoltaic systems. Subsequently, the questionnaire was pilot tested with 50 participants (30 local residents and 20 tourists) at Cigu APV recreational sites to evaluate comprehensibility, cultural appropriateness, and completion time (Figure 1). Item wording was refined based on cognitive interviews and feedback analysis, resulting in the final 22-item measurement instrument. Data collection was conducted at purposively selected kayaking launch points and rest areas within APV recreational sites in Cigu District, Tainan City, between May 2024 and January 2025. This timeframe enabled sampling across seasonal variations in visitor patterns and environmental conditions [3]. Stratified random sampling was employed to ensure representation across weekdays (40% of collection sessions), weekends (45%), and public holidays (15%), with data collection distributed across morning (9:00–12:00), afternoon (13:00–16:00), and evening (16:00–18:00) timeframes to minimize temporal sampling bias. Trained research assistants approached potential participants using a systematic intercept method, explaining the research objectives and obtaining written informed consent before questionnaire administration. The participants completed the self-administered surveys either before (42% of the sample) or after (58%) their kayaking activities. The research assistants remained available to clarify questions without influencing responses. The average completion time was 16.8 min (SD = 4.2).

The pilot testing involved cognitive interviews with the participants to assess item comprehension, cultural appropriateness, and response time. Feedback analysis revealed minor wording adjustments needed for three items related to APV system understanding. The questionnaire completion time averaged 16.8 min (SD = 4.2), which is considered appropriate for field administration. To enhance response rates and data quality, the participants received a locally produced sustainable souvenir (valued at approximately TWD 150) as a token of appreciation upon questionnaire completion. Quality control procedures included a daily review of completed questionnaires, with immediate follow-up for missing data when participants were still on-site. Non-response bias was assessed by comparing early and late respondents across key demographic variables, revealing no significant differences (p > 0.05). Multi-wave data collection across different seasons and weather conditions further mitigated potential sampling biases [20].

3.3. Sample Selection and Data Collection

Data were collected from 613 participants who had engaged in recreational kayaking activities. The sample demographics are presented in Table 2. The research model consisted of three main constructs: facility maintenance and safety professionalism (FM), perceived value of kayaking training (PV), and green energy and sustainable development recognition (GS). The FM construct comprised three first-order dimensions (FM1, FM2, and FM3), each measured by three indicators. The PV construct consisted of four first-order dimensions (PV1, PV2, PV3, and PV4), each measured by two indicators. The GS construct was measured directly by five indicators. The data were analyzed using structural equation modeling (SEM) techniques. First, we assessed the measurement model through confirmatory factor analysis (CFA) to ensure the reliability and validity of the constructs. Second, we evaluated the structural model to test the hypothesized relationships. Finally, we conducted a mediation analysis to examine the mediating effect of PV in the relationship between FM and GS.

Underwater video recordings were systematically conducted using an array of 16 high-definition cameras (GoPro HERO10 Black with custom waterproof housings, GoPro, San Mateo, CA, USA). These cameras were deployed in four distinct APV aquaculture ponds situated within Cigu District, Taiwan, over a six-month period, from August 2024 to January 2025. Each pond was stocked with whiteleg shrimp (P. vannamei) at typical commercial densities, ranging from 25 to 30 individuals/m². At the commencement of the study, the average individual weight of the shrimp was recorded as 12.6 g (±2.3 g).

Cameras were strategically positioned at four standardized locations within each pond to capture a range of environmental conditions: directly beneath APV panel arrays (representing 100% shading), in partial shade zones (experiencing 40–60% shading), in open-water zones (with 0% shading), and along the pond margins (<1 m from the embankment). Video recording followed a stringent, standardized protocol for each experimental event. This protocol comprised a 60-min pre-disturbance baseline period (serving as the control), a 30-min vessel passage period (representing the experimental treatment), and a 120-min post-disturbance period for monitoring recovery. Two distinct vessel treatments were implemented using a randomized crossover design, with a minimum washout period of 48 h between exposures to minimize carryover effects. The vessel treatments included the use of non-motorized kayaks (standard recreational polyethylene sit-on-top kayaks, 3.3 m in length, operated by two paddlers) and small motorized vessels (aluminum work boats, 4.2 m in length, equipped with 15 HP outboard motors). Each treatment was replicated six times per pond, resulting in a total of 24 replicates for each vessel type. Standardized vessel movement patterns, maintaining consistent speed, distance from cameras, and duration of passage, were strictly adhered to across all replicates.

Control observations, where no vessel passage occurred, were conducted on alternating days throughout the study. The behavioral responses of the shrimp were quantitatively assessed through continuous focal sampling of 30 randomly selected individuals per video frame, resulting in the analysis of 120 individuals per treatment event. Behavioral metrics were precisely defined and coded based on established ethogram protocols for penaeid shrimp, adapted from the methodologies described by [48] and [49]. The key behavioral metrics quantified were the startle response rate (percentage of observed individuals exhibiting sudden directional change via rapid tail-flipping or jumping), duration of startle response (time in seconds from initiation to cessation of elevated activity), recovery time (minutes until the return to pre-disturbance behavioral patterns such as feeding and normal swimming), latency to first response (seconds elapsed between vessel approach and the first observable behavioral change), distance-triggered response (minimum distance in meters between the vessel and shrimp at the onset of a response), vertical displacement (maximum vertical movement in centimeters during an escape response), feeding interruption (duration in minutes of suspended substrate probing behavior), and group synchronization index (a scale from 0 to 1 representing the proportion of individuals exhibiting synchronized responses).

Behavioral categories were validated through comparison with established ethograms for penaeid shrimp [50]. Initial pilot observations (n = 50 shrimp, 10 h of video) were independently scored by three trained observers to establish inter-rater reliability. Only behavioral categories achieving Cohen’s κ > 0.80 were retained for analysis.

Video analysis was performed using BORIS v7.9.8 behavioral observation software [51]. To ensure high inter-rater reliability, 20% of the video recordings were double-coded by independent observers, achieving a Cohen’s κ value exceeding 0.85. Statistical analyses were conducted using R v4.2.1 [52]. One-way ANOVA with Tukey’s post-hoc tests was employed to compare continuous behavioral variables between the different vessel types. For proportional behavioral data, chi-square tests were utilized. Two-way ANOVA models were constructed to investigate potential interactions between vessel type and various environmental factors, including water depth, clarity, APV coverage, and time of day. The statistical significance threshold was set at α = 0.05, with Bonferroni correction applied to account for multiple comparisons and minimize the risk of Type I errors.

Data analysis employed a two-stage structural equation modeling (SEM) approach using AMOS 26.0 software [53]. Prior to analysis, the data were screened for multivariate normality, outliers, and missing values following established protocols [54]. Multiple imputation techniques were applied to the minimal missing data (<1.8%), and cases with systematic response patterns (e.g., straight-lining) were excluded from analysis (n = 14), resulting in 613 valid responses. The measurement model was first evaluated using confirmatory factor analysis (CFA) to assess construct validity and reliability. Model fit was assessed using multiple indices, including the chi-square ratio (χ²/df < 3.0), the root mean square error of approximation (RMSEA < 0.08), the comparative fit index (CFI > 0.95), the Tucker–Lewis index (TLI > 0.95), and the standardized root mean square residual (SRMR < 0.06) [55]. Local fit was examined through standardized factor loadings (>0.70), modification indices, and standardized residuals. Reliability was assessed through Cronbach’s alpha (α > 0.70) and composite reliability (CR > 0.70), while convergent validity was evaluated using average variance extracted (AVE > 0.50). Discriminant validity was established by comparing the square root of AVE values with inter-construct correlations and through the heterotrait–monotrait (HTMT) ratio methodology [56], with values below 0.85 indicating sufficient discrimination between constructs. Missing data analysis revealed a completely random pattern (Little’s MCAR test: χ² = 23.45, p = 0.432). Multiple imputation using the expectation maximization algorithm was applied to the minimal missing data (<1.8%). Cases exhibiting systematic response patterns (straight-lining across more than 80% of items) were identified using standard deviation and variance criteria and subsequently excluded from analysis.

Upon establishing a valid measurement model, the structural model was tested to evaluate hypothesized relationships. Mediation effects (H4) were examined using bootstrap analysis with 5000 resamples and bias-corrected 95% confidence intervals (Preacher and Hayes, 2008) [57]. Moderation effects (H5) were tested through interaction terms and multi-group analyses based on theoretically derived moderators [44]. Potential multicollinearity was assessed using variance inflation factors (VIF < 3.0), while common method bias was evaluated using Harman’s single-factor test and the common latent factor approach [58]. Additionally, a marker variable technique was employed to quantify and control for potential method effects.

3.4. Control Variable Integration in SEM Control Variables Were Integrated into the Structural Model Using a Hierarchical Approach

(1)

Direct effects model: Environmental attitude, age, gender, education, and prior experience were modeled as direct predictors of both PV and GS constructs (Table 2).

(2)

Covariate control: All control variables were allowed to correlate with each other and with the main constructs to account for shared variance.

(3)

Model specification:

-

Environmental attitude → PV: β = 0.16, SE = 0.04, p = 0.002.

-

Environmental attitude → GS: β = 0.12, SE = 0.03, p = 0.011.

-

Gender → PV: β = 0.07, SE = 0.05, p = 0.187 (n.s.).

-

Education → GS: β = 0.14, SE = 0.05, p = 0.006.

(4)

Model comparison:

-

Model without controls: χ² = 387.2, df = 165, CFI = 0.931.

-

Model with controls: χ² = 421.7, df = 175, CFI = 0.952.

-

Δχ² = 34.5, Δdf = 10, p < 0.001 (significant improvement).

4. Results and Discussion

4.1. Descriptive Statistics and Demographics

A total of 613 valid responses were obtained from kayaking participants at aquavoltaic (APV) recreational sites in Cigu District, Tainan City. The demographic profile of the respondents is presented in Table 2. The gender distribution was relatively balanced, with a slight majority of females (53.3%, n = 327) compared to males (46.7%, n = 286). The age distribution showed a concentration in the 25–44 age range, with 30.0% (n = 184) aged 25–34 years and 25.8% (n = 158) aged 35–44 years, reflecting the typical demographic profile of active outdoor recreation participants in Taiwan.

Educational attainment was notably high, with 76.7% of the participants possessing a bachelor’s degree or higher, including 22.5% (n = 138) with graduate degrees. This educational profile aligns with previous studies suggesting that eco-tourists and participants in educational recreational activities tend to have above-average educational qualifications [37]. Regarding occupational distribution, the professional/technical (34.2%, n = 210) and business/management (20.8%, n = 127) categories were the most represented, followed by students (19.2%, n = 118). In terms of kayaking experience, the participants were classified into three categories: novices with no prior kayaking experience (42.5%, n = 261), occasional kayakers with one to five previous experiences (38.3%, n = 235), and experienced kayakers with more than five previous outings (19.2%, n = 117). This distribution enabled analysis of how prior experience potentially moderates the relationships between key constructs. Environmental attitudes, as measured using a validated scale adapted from the New Ecological Paradigm [58,59], showed generally pro-environmental orientations (M = 3.92, SD = 0.68 on a five-point scale), consistent with expectations for individuals selecting nature-based recreational activities [36].

Descriptive statistics for the main constructs are presented in

Table 3. Path analysis results and interaction effects.

Dependent Variable	Independent Variable	Coefficient (β)	t-Value	p-Value	95% CI
Perceived value (PV)	Facility maintenance and safety professionalism (FM)	0.68 ***	9.24	<0.001	[0.56, 0.79]
	Gender	0.07	1.32	0.187	[−0.03, 0.17]
	Age	0.05	0.94	0.348	[−0.06, 0.16]
	Prior experience	0.11 *	2.18	0.029	[0.01, 0.21]
Green energy and sustainable development recognition (GS)	Perceived value (PV)	0.56 ***	7.83	<0.001	[0.42, 0.70]
	Facility maintenance and safety professionalism (FM)	0.29 ***	4.12	<0.001	[0.15, 0.43]
	Education level	0.14 **	2.76	0.006	[0.04, 0.24]
	Environmental attitude	0.16 **	3.05	0.002	[0.06, 0.26]
Mediation effect	FM → PV → GS (indirect effect)	0.38 ***	6.57	<0.001	[0.26, 0.45]
Interaction effects
Perceived value (PV)	FM × APV coverage	0.17 **	3.12	0.002	[0.06, 0.28]
	FM × guide expertise	0.21 ***	3.85	<0.001	[0.10, 0.32]
	FM × weather conditions	0.14 *	2.46	0.014	[0.03, 0.25]
Green energy and sustainable development recognition (GS)	PV × prior knowledge	0.19 **	3.44	<0.001	[0.08, 0.30]
	PV × visibility of APV operations	0.22 ***	4.06	<0.001	[0.11, 0.33]
	FM × PV	0.15 **	2.72	0.007	[0.04, 0.26]

Note: * p < 0.05, ** p < 0.01, *** p < 0.001.

. All constructs were measured on five-point Likert scales, with means ranging from 3.79 to 3.95, indicating generally positive perceptions across all dimensions. Green energy and sustainable development recognition (GS) exhibited the highest mean score (M = 3.95, SD = 0.87), suggesting strong acknowledgment of sustainability principles among the participants. Facility maintenance and safety professionalism (FM) received the lowest mean rating (M = 3.79, SD = 0.75), though still well above the scale midpoint, reflecting generally positive but somewhat more critical evaluations of infrastructure and safety protocols (Figure 2).

Examining the component dimensions of each construct revealed additional patterns. Within the FM construct, safety procedures (FM2) received the highest ratings (M = 3.90, SD = 0.75), while facility quality (FM1) was rated lowest (M = 3.75, SD = 0.78). For the PV construct, among the PV dimensions, emotional value (PV2) received the highest endorsement (M = 3.90, SD = 0.87), whereas social value (PV3) was rated lowest (M = 3.70, SD = 0.93). Within the GS items, recognition of “sustainable development importance for preserving water resources” received the highest endorsement (M = 4.02, SD = 0.86). Table 3 presents the correlation coefficients among the latent variables, with diagonal elements representing the square root of the average variance extracted (AVE) for each construct. All correlations were positive and statistically significant (p < 0.001), indicating meaningful associations between constructs while remaining distinct enough to demonstrate discriminant validity (as evidenced by diagonal values exceeding off-diagonal elements in corresponding rows and columns).

The strongest correlation was observed between the perceived value of kayaking training (PV) and green energy and sustainable development recognition (GS) (r = 0.73, p < 0.001), suggesting a robust relationship between the participants’ valuation of their experience and their recognition of sustainability principles. Facility maintenance and safety professionalism (FM) showed a strong correlation with perceived value (r = 0.68, p < 0.001) and a moderate-to-strong correlation with green energy and sustainable development recognition (r = 0.59, p < 0.001). Environmental attitude demonstrated moderate correlations with all three primary constructs (r ranging from 0.31 to 0.47, p < 0.001), underscoring its relevance as a control variable in subsequent analyses. Control variables included gender, age, education level, prior experience, and environmental attitude. Goodness-of-fit indices: χ²/df = 2.18, CFI = 0.952, TLI = 0.943, RMSEA = 0.048 (90% CI [0.041, 0.055]), and SRMR = 0.042. R² values: PV = 0.46, GS = 0.54. All VIF values <2.5, indicating no significant multicollinearity issues.

These correlation patterns provide preliminary support for the hypothesized relationships in our conceptual model. The stronger correlation between PV and GS compared to FM and GS suggests the potential presence of mediation effects, warranting further investigation through structural equation modeling. All constructs were measured on five-point Likert scales (1 = strongly disagree to 5 = strongly agree). Composite reliability (CR) values: FM = 0.92, PV = 0.94, GS = 0.93, environmental attitude = 0.89. All CR values exceeded the recommended threshold of 0.70, indicating good internal consistency reliability. The heterotrait–monotrait (HTMT) ratio values between all construct pairs were <0.85, providing additional evidence of discriminant validity (

Table 4. Correlation coefficients of the latent variables.

Construct	Mean	SD	1	2	3	4
1. FM (facility maintenance and safety professionalism)	3.79	0.75	0.88
2. PV (perceived value of kayaking training)	3.80	0.84	0.68 ***	0.90
3. GS (green energy and sustainable development recognition)	3.95	0.87	0.59 ***	0.73 ***	0.86
4. Environmental attitude	3.92	0.68	0.31 ***	0.42 ***	0.47 ***	0.83

Note: *** p < 0.001.

).

4.2. Behavioral Responses of Whiteleg Shrimp (Penaeus vannamei) to Vessel Disturbances in Aquavoltaic Environments: Implications for Recreational Integration

4.2.1. Differential Behavioral Responses to Vessel Types

Our underwater camera analysis revealed substantial differences in Penaeus vannamei behavioral reactions between motorized and non-motorized vessel disturbances. As presented in

Table 5. Behavioral startle response rates and associated metrics in whiteleg shrimp during vessel passage.

Response Measure	Control Period	During Kayaking	Recovery Period (0–30 min)	Recovery Period (30–60 min)
Non-motorized kayaks
Startle response (% of individuals)	3.2 ± 1.5 ᵃ	18.7 ± 4.3 ᵇ***	8.4 ± 2.6 ᵇ**	4.1 ± 1.8 ᵃ
Swimming activity (movements/min)	2.8 ± 0.9 ᵃ	4.6 ± 1.2 ᵇ**	3.5 ± 1.0 ᵃᵇ*	2.9 ± 0.8 ᵃ
Feeding behavior (substrate probing/min)	7.3 ± 1.7 ᵃ	3.1 ± 1.4 ᵇ***	5.2 ± 1.5 ᶜ**	6.8 ± 1.6 ᵃᶜ
Motorized boats (comparison)
Startle response (% of individuals)	3.4 ± 1.4 ᵃ	42.6 ± 7.8 ᵇ***	27.3 ± 5.4 ᶜ***	16.8 ± 4.2 ᵈ***
Swimming activity (movements/min)	2.7 ± 0.8 ᵃ	6.9 ± 1.7 ᵇ***	5.4 ± 1.5 ᶜ***	4.2 ± 1.2 ᵈ**
Feeding behavior (substrate probing/min)	7.5 ± 1.6 ᵃ	1.2 ± 0.8 ᵇ***	2.4 ± 1.1 ᶜ***	4.3 ± 1.4 ᵈ**

Note: Values represent the mean ± standard deviation. Different superscript letters (a, b, c, d) indicate significant differences between treatment conditions based on Tukey’s post-hoc tests following one-way ANOVA. Asterisks indicate significant differences from the control period: * p < 0.05, ** p < 0.01, *** p < 0.001.

, non-motorized kayaks elicited startle responses in 18.7% (±4.3%) of observed individuals, significantly exceeding the baseline rate of 3.2% (±1.5%) during control periods (F = 143.6, p < 0.001), but markedly lower than the 42.6% (±7.8%) observed during motorized vessel passages. This pattern was consistent across all behavioral metrics, with motorized vessels inducing 2.2–3.4 times greater response magnitudes compared to non-motorized alternatives. The temporal dimension of behavioral alterations differed significantly between vessel types. While kayak-induced responses typically resolved within 24.6 min (±7.2), behavioral modifications following motorized vessel exposures persisted for 82.3 min (±16.4) on average (t = 18.7, p < 0.001). Inter-rater reliability analysis revealed strong agreement across all behavioral categories: startle response (κ = 0.87, 95% CI [0.82, 0.92]), substrate probing (κ = 0.83, 95% CI [0.78, 0.88]), and displacement response (κ = 0.85, 95% CI [0.80, 0.90]). This extended recovery period represents a substantial impact on feeding opportunity and energy expenditure (

Table 6. Comprehensive behavioral ethogram for P. vannamei.

Behavioral Category	Operational Definition	Scoring Criteria	Duration Threshold
Startle response	Rapid tail-flip movement >30° from baseline position with immediate directional change	Visible flexion of abdomen with propulsive movement	Initiated within 2 s of stimulus
Substrate probing	Directed movement of maxillipeds or walking legs toward/into substrate	Observable sediment displacement or appendage contact with bottom	Continuous for ≥3 s
Feeding interruption	Cessation of substrate probing behavior during active feeding periods	Complete withdrawal of feeding appendages from substrate	Duration from cessation to resumption
Displacement response	Horizontal movement away from original position following disturbance	Measurable change in position ≥1 body length

). These findings align with [60], who documented similar recovery timeframes for P. vannamei following mechanical disturbances, and support the mechanistic explanation proposed by [61] regarding the relationship between disturbance intensity and recovery trajectory in decapod crustaceans (

Table 7. Complete effect size analysis.

Comparison	Mean Difference	Cohen’s d	95% CI	Eta-Squared (η2)	Interpretation
Behavioral responses
Kayak vs. control (startle %)	15.50%	1.24	[0.89, 1.59]	0.133	Large effect
Motorized vs. control (startle %)	39.40%	2.87	[2.45, 3.29]	0.391	Very large effect
Kayak vs. motorized (recovery time)	−57.7 min	−2.14	[−2.56, −1.72]	0.224	Large effect
SEM path coefficients
FM → PV	β = 0.68	-	[0.56, 0.79]	R2 = 0.46	Large effect
PV → GS	β = 0.56	-	[0.42, 0.70]	R2 = 0.54	Large effect
Interaction Effects
Vessel type × water depth	-	-	-	η2 = 0.14	Large effect
Vessel type × APV coverage	-	-	-	η2 = 0.12	Medium–large effect

).

4.2.2. Qualitative Response Patterns

Beyond quantitative differences in response magnitude, our analysis revealed distinct qualitative patterns in shrimp behavioral reactions to different vessel types, as detailed in

Table 8. Proportion of shrimp exhibiting different categories of behavioral responses during vessel passage.

Response Category	Non-Motorized Kayaks (%)	Small Motorized Vessels (%)	Statistical Significance
Startle response with short displacement	11.2 ± 2.8	15.4 ± 3.1	p < 0.05
Startle response with extended displacement	7.5 ± 1.9	27.2 ± 5.3	p < 0.001
Instantaneous substrate burying behavior	3.8 ± 1.2	12.6 ± 2.9	p < 0.001
Prolonged movement cessation	8.3 ± 2.1	22.4 ± 4.2	p < 0.001
Erratic swimming pattern	6.2 ± 1.7	18.5 ± 3.8	p < 0.001
No observable response	81.3 ± 4.3	57.4 ± 7.8	p < 0.001
Persistent behavioral alterations (>30 min)	4.1 ± 1.3	19.7 ± 4.2	p < 0.001

. Non-motorized kayaks predominantly elicited short-displacement startle responses (11.2% of individuals), characterized by brief tail-flip movements covering distances of <30 cm. In contrast, motorized vessels triggered significantly higher rates of extended displacement responses (27.2% of individuals), involving multiple sequential tail-flips and distances frequently exceeding 1 m. The values represent the mean percentage of individuals ± standard deviation exhibiting each response category. The categories are not mutually exclusive, as individuals may display multiple response types sequentially. Statistical significance was determined using independent t-tests comparing vessel types for each response category (

Table 9. Comprehensive measurement model validity.

Construct	CR	AVE	α	MSV	ASV	Validity Status
FM	0.92	0.78	0.91	0.46	0.31	✓ Valid
PV	0.94	0.81	0.93	0.53	0.38	✓ Valid
GS	0.93	0.74	0.89	0.53	0.41	✓ Valid
EA	0.89	0.69	0.87	0.22	0.18	✓ Valid

,

Table 10. Discriminant validity assessment.

	FM	PV	GS	EA
FM	0.88
PV	0.68	0.9
GS	0.59	0.73	0.86
EA	0.31	0.42	0.47	0.83

Note: Diagonal elements (bold) are √AVE; off-diagonal elements are construct correlations.

and

Table 11. Heterotrait–monotrait (HTMT) ratios.

	FM	PV	GS	EA
PV	0.74
GS	0.67	0.81
EA	0.38	0.49	0.53

Note: All HTMT ratios < 0.85 threshold, confirming discriminant validity.

). Substrate burying behavior, indicating higher-intensity threat perception, was observed in only 3.8% of individuals during kayak passages but increased to 12.6% during motorized vessel exposures (χ² = 24.3, p < 0.001). This behavioral distinction is particularly noteworthy, as [62] identified substrate burying as an energetically costly stress response in penaeid shrimp that can significantly impact metabolic rate and subsequent growth efficiency.

Similarly, prolonged movement cessation (>30 s of immobility) occurred in 22.4% of individuals following motorized disturbance compared to 8.3% after kayak passages (χ² = 31.7, p < 0.001), consistent with the “freeze” component of invertebrate threat responses documented by [63]. Most significantly, persistent behavioral alterations continuing beyond 30 min post-disturbance were observed in 19.7% of individuals following motorized vessel passages but only 4.1% after kayak exposures (χ² = 28.4, p < 0.001). This difference supports the theoretical framework in [64] regarding disturbance intensity thresholds and chronic stress induction in aquacultured crustaceans, with important implications for long-term production impacts.

4.2.3. Environmental and Operational Moderators

Several environmental and operational factors significantly moderated the relationship between vessel type and behavioral responses, as presented in

Table 12. Variation in startle response rates according to environmental and operational factors.

Factor	Level	Non-Motorized Kayaks (%)	Small Motorized Vessels (%)	Significance of Interaction
Water depth	Shallow (<1.2 m)	24.3 ± 5.2	51.8 ± 8.4	p < 0.01
	Medium (1.2–1.6 m)	17.8 ± 4.1	43.5 ± 7.6
	Deep (>1.6 m)	12.4 ± 3.2	36.2 ± 6.7
Water clarity	Clear (<5 NTU)	21.6 ± 4.8	47.2 ± 8.1	p < 0.05
	Moderate (5–15 NTU)	18.2 ± 4.2	42.3 ± 7.5
	Turbid (>15 NTU)	15.7 ± 3.8	38.4 ± 7.1
Vessel speed	Slow	13.5 ± 3.4	32.6 ± 6.5	p < 0.001
	Moderate	19.8 ± 4.5	44.8 ± 7.9
	Fast	23.7 ± 5.1	58.3 ± 9.2
APV coverage	None (0%)	24.5 ± 5.3	48.7 ± 8.2	p < 0.01
	Partial (20–40%)	15.6 ± 3.7	40.2 ± 7.3
	High (>40%)	14.2 ± 3.5	37.8 ± 7.0
Time of day	Morning	17.2 ± 4.0	41.8 ± 7.4	p > 0.05
	Midday	18.9 ± 4.4	42.3 ± 7.6
	Afternoon	19.2 ± 4.5	43.2 ± 7.8

. Water depth emerged as a particularly important moderator, with shallow areas (<1.2 m) exhibiting amplified response rates for both vessel types. The vessel type × water depth interaction was significant (F = 8.72, p < 0.01, η² = 0.14), representing a large effect size according to Cohen’s guidelines. In shallow areas (<1.2 m), kayaks elicited startle responses in 24.3 ± 5.2% of individuals compared to 12.4 ± 3.2% in deep areas (>1.6 m), yielding a Cohen’s d of 1.23 (95% CI [0.89, 1.57]), indicating a significant practical difference. Notably, APV infrastructure exerted a significant moderating influence, with shrimp beneath panel arrays exhibiting markedly reduced startle responses (14.2% for kayaks and 37.8% for motorized vessels) compared to those in unshaded areas (24.5% for kayaks and 48.7% for motorized vessels). This moderating effect was proportionally greater for kayaks (42.0% reduction) than for motorized vessels (22.4% reduction). These findings support and extend observations [65] regarding the sensory buffer effect of overhead structures in aquaculture environments. As [66] noted, “physical structures above the water surface create differential stimulus penetration profiles, with visual stimuli experiencing greater attenuation than acoustic or hydrodynamic disturbances”. Our results provide empirical confirmation of this effect in the specific context of APV infrastructure.

The values represent the mean percentage of individuals ± standard deviation exhibiting startle responses under different environmental and operational conditions. Significance of interaction refers to the vessel type × factor interaction term in the two-way ANOVA models. Vessel speed significantly influenced response rates, with faster passages triggering higher response frequencies for both vessel types. This aligns with [67], who documented similar speed-dependent stress responses in P. vannamei, attributing this to increased water displacement and vibration intensity. The interaction between vessel type and speed was particularly pronounced (F = 16.38, p < 0.001), suggesting that speed management represents a particularly effective mitigation strategy for kayaking activities. Time of day showed no significant interaction with vessel type (F = 0.87, p > 0.05), indicating that the relative difference between vessel impacts remains consistent across diurnal cycles. This contrasts with the findings in [68] regarding diurnal variations in disturbance sensitivity for other aquacultured species, suggesting that P. vannamei may exhibit more temporally stable response patterns. This temporal consistency has important practical implications for recreational management in APV systems, as it suggests that time-of-day restrictions may be less critical than other management interventions.

The substantially different behavioral response patterns observed between vessel types likely reflect multiple underlying sensory mechanisms. Motorized vessels generate complex multimodal stimuli, including louder sounds (35–45 dB higher underwater sound pressure levels than kayaks), stronger pressure waves, more intensive water column mixing, and larger visual profiles. Our findings suggest that acoustic and hydrodynamic stimuli may dominate shrimp response patterns, given the significantly higher startle rates, greater response distances, and longer recovery times associated with motorized vessels. This interpretation is consistent with the characterization in [69] of crustacean sensory hierarchies, which positions mechanosensory and vibrational stimuli as primary threat detection modalities. Non-motorized kayaks primarily create visual stimuli (moving shadows and hull presence) and minimal hydrodynamic disturbance from paddle movements. The observed moderate increase in startle responses above baseline (18.7% vs. 3.2%) confirms that even these limited stimuli are detected by P. vannamei’s sensory systems, consistent with previous research demonstrating visual sensitivity in decapod crustaceans [32]. However, the relatively rapid recovery (24.6 min) and predominance of short-displacement responses suggest lower perceived threat intensity, potentially falling below what [70] termed the “allostatic loading threshold” for significant physiological impact.

The differential moderating effect of APV structures provides further insights into underlying mechanisms. The disproportionate reduction in kayak-induced responses beneath panel arrays suggests that visual stimuli (specifically moving shadows) may constitute a larger component of kayak disturbance effects. This aligns with the findings in [71] regarding shadow avoidance behavior in penaeid shrimp and extends their work by demonstrating how anthropogenic overhead structures can modulate this response. Motorized vessel impacts, while also moderated by APV coverage, showed less relative reduction, consistent with their more diverse stimulus profile, including substantial acoustic and hydrodynamic components that penetrate shaded areas. The ecological significance of these behavioral differences depends on their frequency, duration, and cumulative energetic consequences.

Based on observed behavioral patterns, kayaking activities resulted in feeding behavior interruptions averaging 24.6 min per exposure event (95% CI [21.3, 27.9]). While these interruptions represent measurable disturbances, translating behavioral responses into production impacts requires additional empirical validation. The current findings suggest the following: (1) behavioral impact—significant but brief feeding interruptions occur during kayak passages; (2) recovery pattern—normal feeding behavior typically resumes within 30 min; (3) production impact—unknown, requiring direct measurement of growth rates and feed conversion efficiency. The production impact falls well below the 5% threshold identified by [72] as significant for production impacts, suggesting that moderate kayaking activity can be incorporated into APV systems without substantial yield penalties. In contrast, regular motorized vessel traffic (equivalent frequency) would result in feeding interruptions totaling 4.3–8.7% of potential feeding time, approaching or exceeding critical thresholds. This distinction provides empirical support for established management preferences for non-motorized recreation in sensitive production environments [36]. It also aligns with the energy budget models for P. vannamei in [73], which indicate that behavioral disturbances affecting less than 3% of the total activity budget are unlikely to significantly impact growth trajectories or production efficiency.

The observed moderating effects of environmental and operational factors suggest promising avenues for impact mitigation through strategic management. Maintaining slower paddling speeds, following deeper-water routes, and utilizing APV-shaded areas for primary kayaking corridors could substantially reduce behavioral disturbance while maintaining recreational access. These findings complement and extend the work of [74], who demonstrated similar benefit–risk optimization approaches for recreational activities in marine protected areas, and suggest that evidence-based spatial management can effectively balance multiple use objectives in transformed aquatic landscapes.

4.2.4. Behavioral vs. Physiological Stress Assessment

Our study focused exclusively on behavioral indicators of disturbance responses, providing valuable insights into immediate reactions but representing only one component of comprehensive stress assessment. The behavioral indicators used were (1) startle responses: immediate threat perception [63]; (2) displacement behaviors: avoidance and escape responses [61]; (3) feeding interruptions: resource-acquisition impacts [71]. The literature support for behavioral–physiological links includes research on P. vannamei that demonstrates correlations between the following: (1) tail-flipping frequency and cortisol levels (r = 0.72, p < 0.001) [49]; (2) feeding cessation duration and metabolic rate changes [48]. (3) displacement distance and immune function suppression [64]). Regarding study limitations, we acknowledge that comprehensive stress assessment requires physiological measures, including (1) cortisol or other stress hormone concentrations, (2) metabolic rate measurements, (3) immune function indicators, and (4) growth performance metrics.

4.3. Moderation Analysis

Having thoroughly examined the ecological implications of recreational activities by analyzing the behavioral responses of whiteleg shrimp to vessel disturbances, our research next focused on the human dimensions of this integrated aquavoltaic system. Thus, prior to testing structural relationships, we evaluated the measurement model to ensure construct validity and reliability of the survey data, thereby establishing a robust foundation for analyzing participant perceptions. As reported in

Table 13. Factor loading and composite reliability.

Measure Dimensions	Variables and Items	SFL	SMC	CR	AVE	Cronbach’s α
Facility maintenance and safety professionalism (FM)				0.92	0.78	0.91
Facility quality (FM1)				0.89	0.73	0.87
	FM1_1: The kayaking facilities are well-maintained	0.85	0.72
	FM1_2: The equipment provided is in good condition	0.88	0.77
	FM1_3: The facilities meet safety standards	0.83	0.69
Safety procedures (FM2)				0.91	0.76	0.89
	FM2_1: Safety procedures are clearly communicated	0.87	0.76
	FM2_2: Emergency protocols are well-established	0.90	0.81
	FM2_3: Safety equipment is readily available	0.85	0.72
Instructor professionalism (FM3)				0.88	0.71	0.85
	FM3_1: Instructors are knowledgeable about safety measures	0.82	0.67
	FM3_2: Instructors demonstrate professional skills	0.84	0.71
	FM3_3: Instructors effectively communicate instructions	0.88	0.77
Perceived value of kayaking training (PV)				0.94	0.81	0.93
Functional value (PV1)				0.92	0.78	0.90
	PV1_1: The kayaking training provides good value for money	0.89	0.79
	PV1_2: The training delivers expected benefits	0.91	0.83
Emotional value (PV2)				0.90	0.75	0.88
	PV2_1: The kayaking experience is enjoyable	0.87	0.76
	PV2_2: The training creates positive emotions	0.89	0.79
Social value (PV3)				0.87	0.69	0.85
	PV3_1: The training provides opportunities for social interaction	0.84	0.71
	PV3_2: Participating in kayaking improves social image	0.86	0.74
Educational value (PV4)				0.89	0.72	0.86
	PV4_1: The training provides valuable knowledge about nature	0.85	0.72
	PV4_2: The experience enhances environmental awareness	0.88	0.77
Green energy and sustainable development recognition (GS)				0.93	0.74	0.89
	GS1: I recognize the importance of green energy in outdoor recreation	0.83	0.69
	GS2: Sustainable development is essential for preserving natural water resources	0.87	0.76
	GS3: I understand how kayaking relates to environmental conservation	0.85	0.72
	GS4: Green energy practices should be integrated into recreational activities	0.82	0.67
	GS5: Sustainable approaches enhance the quality of recreational experiences	0.84	0.71
Control variable
Environmental attitude				0.89	0.69	0.87
	EA1: Human intervention in nature often produces disastrous consequences	0.78	0.61
	EA2: The balance of nature is very delicate and easily upset	0.81	0.66
	EA3: Plants and animals have as much right as humans to exist	0.86	0.74
	EA4: Humans are severely abusing the environment	0.83	0.69
	EA5: If things continue on their present course, we will soon experience a major ecological catastrophe	0.79	0.62

Note: N = 613. SFL = standardized factor loading; SMC = squared multiple correlation; CR = composite reliability; AVE = average variance extracted. All factor loadings are significant at p < 0.001.

, all constructs demonstrated excellent reliability, with composite reliability (CR) values ranging from 0.89 to 0.94, substantially exceeding the recommended threshold of 0.70 [54]. Convergent validity was established through average variance extracted (AVE) values (represented by squared diagonal elements in Table 13), all of which exceeded the 0.50 threshold, indicating that the constructs explain more than 50% of the variance in their respective indicators [75]. Discriminant validity was confirmed through multiple methods. First, the Fornell–Larcker criterion was satisfied, as the square root of AVE for each construct (diagonal elements in Table 13) exceeded all corresponding inter-construct correlations. Second, the heterotrait–monotrait (HTMT) ratios remained below the conservative threshold of 0.85 for all construct pairs, providing additional evidence of construct distinctiveness [56]. Higher-order constructs (FM and PV) were modeled as second-order constructs with their respective dimensions as first-order factors. Environmental attitude items were adapted from the New Ecological Paradigm [58]. Confirmatory factor analysis (CFA) yielded excellent fit indices: χ²/df = 2.18, CFI = 0.952, TLI = 0.943, RMSEA = 0.048 (90% CI [0.041, 0.055]), and SRMR = 0.042, all meeting or exceeding recommended benchmarks [55]. All standardized factor loadings exceeded 0.70 (ranging from 0.72 to 0.91), further supporting the measurement model’s validity. These results collectively establish a robust foundation for testing the structural relationships proposed in our hypotheses.

4.4. Examining the Relationships Between Facility Maintenance, Perceived Value, and Sustainability Recognition in Aquavoltaic Recreational Settings

The structural equation modeling analysis provided significant evidence regarding the relationships proposed in our research framework.

Table 14. Summary of the hypothesis testing results.

Hypothesis	Path	Standardized Path Coefficient (β)	t-Value	p-Value	Result
H1	FM → PV	0.68	9.24	<0.001	Supported
H2	PV → GS	0.56	7.83	<0.001	Supported
H3	FM → GS	0.29	4.12	<0.001	Supported
H4	FM → PV → GS	0.38 (indirect effect)	6.57	<0.001	Supported
H5a	Environmental conditions moderate FM → PV	-	-	-	Partially supported
H5a-1	APV coverage × FM → PV	0.17	3.12	0.002	Supported
H5a-2	Guide expertise × FM → PV	0.21	3.85	<0.001	Supported
H5a-3	Weather conditions × FM → PV	0.14	2.46	0.014	Supported
H5a-4	Visitor density × FM → PV	0.08	1.53	0.127	Not supported
H5b	Visitor characteristics moderate PV → GS	-	-	-	Partially supported
H5b-1	Prior knowledge × PV → GS	0.19	3.44	<0.001	Supported
H5b-2	Visibility of APV operations × PV → GS	0.22	4.06	<0.001	Supported
H5b-3	Educational background × PV → GS	0.07	1.42	0.156	Not supported
H5b-4	Age × PV → GS	0.05	1.18	0.239	Not supported

presents a summary of the hypothesis testing results, illustrating the statistical significance and empirical support for each proposed relationship. Hypothesis 1 proposed that facility maintenance and safety professionalism (FM) would positively influence the perceived value of kayaking training (PV). This relationship was strongly supported (β = 0.68, t = 9.24, p < 0.001), indicating that participants’ perceptions of facility quality, safety procedures, and instructor professionalism significantly enhanced their valuation of the kayaking experience. This finding aligns with the research in [36] on marine tourism experiences, which demonstrated that infrastructure quality and safety management are primary determinants of perceived experiential value.

Hypothesis 2 postulated that the perceived value of kayaking training (PV) would positively influence green energy and sustainable development recognition (GS). This hypothesis was strongly supported (β = 0.56, t = 7.83, p < 0.001), confirming that participants who perceived higher value in their kayaking experience demonstrated greater recognition and appreciation of green energy and sustainable development principles. This relationship supports the theoretical framework in [76], connecting experiential value with knowledge internalization and attitude formation in sustainability contexts.

Hypothesis 3 proposed that facility maintenance and safety professionalism (FM) would directly influence green energy and sustainable development recognition (GS). This direct relationship was supported (β = 0.29, t = 4.12, p < 0.001), though with a relatively lower coefficient compared to the relationship mediated through perceived value. This finding is consistent with the research in [77] on photovoltaic landscapes, which identified physical environment quality as a significant but not predominant factor in visitors’ sustainability perceptions. Hypothesis 4 predicted that the perceived value of kayaking training (PV) would mediate the relationship between facility maintenance and safety professionalism (FM) and green energy and sustainable development recognition (GS). Bootstrap analysis with 5000 resamples revealed a significant indirect effect (β = 0.38, t = 6.57, p < 0.001, 95% CI [0.26, 0.45]). Since the direct effect of FM on GS remained significant after accounting for the indirect effect, the results indicate partial mediation.

While our cross-sectional design precludes definitive causal mediation inference, the pattern of associations (β = 0.38, 95% CI [0.26, 0.45]) is consistent with theoretical predictions of mediation. The significant indirect effect suggests that PV serves as an important associational pathway linking FM to GS. However, establishing temporal precedence requires longitudinal data to confirm causal mediation. The magnitude of the indirect effect (0.38) exceeds that of the direct effect (0.29), suggesting that a substantial portion (approximately 57%) of the influence of facility maintenance and safety professionalism on sustainability recognition operates through the enhancement of perceived experiential value. This finding extends the stimulus–organism–response model of experiential learning [78] by quantifying the relative contributions of direct and mediated pathways in sustainability education contexts.

Hypothesis 5 proposed the existence of moderating effects between key variables in the model. This hypothesis was partially supported, with significant moderation effects identified for some but not all proposed moderators (see Table 14). Three key moderators demonstrated significant interaction effects showing that APV coverage significantly moderated the relationship between FM and PV (β = 0.17, t = 3.12, p = 0.002), with the effect being stronger in environments with moderate coverage (20–40%) compared to high coverage (>40%) or no coverage. This finding suggests that an optimal level of APV infrastructure creates a distinctive environment that enhances the perceived connection between facility quality and experience value. Guide expertise moderated the FM → PV relationship (β = 0.21, t = 3.85, p < 0.001), with stronger effects observed when guides demonstrated specific knowledge about aquavoltaic systems. This aligns with the work in [18], highlighting the critical role of human interpretation in mediating visitor experiences of novel technological landscapes. Our findings specifically apply to aquavoltaic recreational systems in subtropical coastal environments similar to Cigu, Taiwan. The theoretical framework may be relevant to other contexts, but empirical validation is required before generalizing beyond the following geographic scope: (1) similar climate zones (subtropical/tropical), (2) coastal aquaculture regions, and (3) Asian cultural contexts with comparable environmental attitudes. Visibility of APV operations moderated the PV → GS relationship (β = 0.22, t = 4.06, p < 0.001), with stronger effects when operational components of the aquavoltaic system were highly visible and incorporated into the educational narrative. This supports the findings in [79] regarding the importance of tangible demonstration in sustainability education. However, several hypothesized moderators, including participant age, educational background, and weather conditions, did not exhibit significant interaction effects, suggesting that the core relationships in our model are relatively robust across diverse participant characteristics and environmental conditions. Our findings make several important theoretical contributions to the understanding of sustainability education in novel technological landscapes. First, the strong support for our mediation hypothesis (H4) extends experiential learning theory by empirically demonstrating the critical mediating role of perceived value in translating physical environment qualities into sustainability knowledge and attitudes. This advances beyond the general premises of experiential education to specify the cognitive pathways through which learning occurs in sustainability contexts. The significant direct effect of facility maintenance on sustainability recognition (H3), even after accounting for mediation, suggests a more complex relationship than previously theorized. This finding aligns with the dual processing model of environmental learning in [74], which posits that physical environment qualities can influence attitudes through both rational–evaluative pathways (mediated by perceived value) and affective–associative pathways (direct influence). Our results provide empirical validation for this dual-pathway model in the specific context of aquavoltaic recreational environments.

The moderating effect of APV coverage contributes to emerging theories regarding optimal stimulation levels in technological nature experiences [80]. The inverted U-shaped relationship we observed—with moderate coverage (20–40%) producing stronger effects than either low or high coverage—suggests that environmental design for educational effectiveness may require balancing novelty and familiarity, technological presence, and natural elements. This extends beyond simple presence/absence considerations to more nuanced optimization approaches.

5. Conclusions

This study offers several significant theoretical contributions. First, we advance experiential learning theory by empirically validating a dual-pathway model, highlighting perceived value’s crucial mediating role between facility quality and sustainability recognition, while also revealing direct learning mechanisms. Second, our research quantified the differential ecological impacts of motorized versus non-motorized vessels on key aquaculture species. Underwater video demonstrated that kayaking elicits significantly lower startle responses in whiteleg shrimp (18.7%) compared to motorized boats (42.6%), with faster recovery, providing empirical parameters for disturbance thresholds in production landscapes. Third, the observed inverted U-shaped moderating effect of APV coverage advances theories on optimal environmental design for sustainability education, emphasizing a balance between technological and natural elements. Finally, by integrating aquavoltaic development, recreational tourism, and sustainability education, this study proposed an integrated theoretical framework for the “fishery–energy–recreation” nexus, offering interdisciplinary insights for sustainable coastal transitions.

The “fishery–energy–recreation” nexus exemplified in Cigu offers valuable insights for similar transitions globally. By carefully designing and managing these systems to optimize both ecological compatibility and visitor experiences, we can create landscapes that not only produce renewable energy and sustainable food but also nurture the environmental consciousness needed to support broader sustainability transitions. In this way, transformed production landscapes can become powerful catalysts for transforming mindsets, creating virtuous cycles of sustainability that extend far beyond their physical boundaries.

Several limitations constrain causal interpretation of our findings: (1) The cross-sectional design means temporal precedence cannot be established, limiting causal mediation inference; (2) common method variance indicates that self-report measures may inflate associations; and (3) the single context means that the findings are specific to Cigu APV systems. Future research should employ (1) longitudinal designs with repeated measures, (2) experimental manipulation of facility quality, and (3) multi-site replication across different APV configurations.