An Integrated Climate–Spatial Analytical Framework for Assessing 3S Tourism Resilience on the Mediterranean Island of Vis, Croatia

Abstract

Small Mediterranean islands relying on the sun–sea–sand (3S) tourism model face growing climate risks that threaten their tourism-dependent economies. This study evaluates climate suitability for 3S tourism on the Island of Vis by integrating the Climate Index for Tourism (CIT) with land- use and land-cover (LU/LC) spatial analysis. The integration is operationalized by overlaying CIT-derived seasonal suitability windows with LU/LC-based spatial vulnerability maps, enabling identification of micro-zones where natural buffers (forest cover and elevation) can offset thermal discomfort during peak heat stress periods. Observed data reveals declining ideal 3S conditions from July to October, with the island already exceeding 50 days per year of Physiologically Equivalent Temperature (PET) above 35.1 °C, increasing by 0.7 days per year. Regional climate models tend to exhibit a cold bias over small Adriatic islands, largely related to their limited spatial horizontal resolution (12.5 km grid spacing). However, they robustly reproduce the direction of recent and projected warming trends. Future projections indicate that the annual number of strong heat stress days with PET above 35.1 °C increase from approximately one per year in the reference period to six under RCP4.5 and nine under RCP8.5, with both scenarios reducing ideal peak-summer conditions while extending favorable periods into transitional seasons. Spatial analysis shows that coastal zones have higher sealed surfaces and less forest cover, reducing natural shade and cooling capacity, while the island interior offers higher elevations, forest buffers, hiking trails, and a UNESCO Global Geopark. Drawing on social–ecological resilience theory, we conceptualize the island’s tourism system as an adaptive unit whose long-term viability depends on spatially diversified resource use and temporally extended seasonality. The integrated analytical framework identifies not only when conditions deteriorate but where alternative tourism resources exist, enabling more targeted adaptation planning and supporting diversification toward outdoor tourism forms. The novelty of this study lies in the systematic spatial integration of bioclimatic suitability assessments (CIT and PET) with LU/LC analysis at the micro-island scale. Such an approach moves beyond temporally focused climate–tourism indices to produce actionable, location-specific adaptation strategies.

1. Introduction

Small Mediterranean islands are attractive destinations and key hubs in the global tourism system, yet their vulnerability stems from limited spatial and natural resources and tourism-dependent economies. Coastal areas are often clustered with settlements and visitor infrastructure, creating a paradox of ecological fragility (Davenport & Davenport, 2006). The Mediterranean basin has been recognized as a “hotspot” of climate change, with projected warming rates exceeding global averages and pronounced changes in precipitation patterns, sea level rise, and the frequency of extreme events (IPCC, 2023). These changes cause climate hazards, such as prolonged droughts, forest fires, heat waves, heavy rains with stormy winds, tidal waves, and coastal flooding, which alter patterns of tourism activities. For Mediterranean island tourist destinations, these changes manifest water shortages; devastation of landscape and resources; reduced thermal, aesthetic, and physical comfort; coastal erosion; and temporary interruption of the island’s connections with the mainland (ESOTC, 2024; Lionello et al., 2014; Lionello & Scarascia, 2018; Vousdoukas et al., 2020). These complex stressors do not act in isolation. Linked synergistically through cascading impact chains, they disrupt the natural resource base. As Agulles et al. (2022) and Monioudi et al. (2023) revealed, climate change elevates the risk of loss of tourism attractiveness—beaches, thermal comfort, and coastal landscapes on which 3S tourism depends. Consequently, these shifts directly affect tourism resilience and destination competitiveness (MedECC, 2024), which requires appropriate adaptation responses. Research on tourism adaptation has identified a range of strategic responses to climate change, from risk reduction through diversification to transformative destination repositioning (Gössling et al., 2023; Hall et al., 2015). For small island destinations, adaptive capacity is further constrained by resource limitations and the dependency of the 3S model (Scott et al., 2019), making early warning analytical frameworks particularly valuable. The remote Croatian Adriatic islands, including the Island of Vis, exhibit specific vulnerabilities related to their geographical isolation and concentrated tourism development in narrow coastal zones. Long-term meteorological analyses conducted on the Mediterranean region and small islands show a statistically significant increase in the average, maximum, and number of warm days, with the strongest trends being during peak summer months (Kotsias & Lolis, 2025; Speranza, 2026). Water supply challenges are increasing as peak tourist loads coincide with drier warm periods (Grofelnik & Maradin, 2023), and changes in wind regimes affect island supply chains and tourism activities, like nautical tourism (Lionello et al., 2014; Ruti et al., 2016). The Croatian Island of Vis, located in the central Adriatic Sea, exemplifies this dual reality. With approximately 3500 residents and seasonal overnights exceeding 100,000, the Island of Vis faces intensified pressures on ecosystem services and infrastructure during peak summer months (CBS, 2022, 2025).

Understanding how small island tourism destinations can navigate these compounding climate pressures requires a theoretical framework capable of capturing the dynamic, interconnected nature of humans—environment systems. Resilience thinking rooted in the analysis of socio-ecological systems provides such a platform, offering an interdisciplinary basis for examining how destinations could sustain development in the face of both expected and unexpected change (Folke et al., 2010; Walker et al., 2004). In this framework, tourism destinations are understood as complex adaptive systems embedded within the biosphere, where social, economic, and ecological subsystems are tightly coupled and co-evolve across spatial and temporal scales (Folke, 2006). Resilience encompasses three complementary capacities: persistence (the ability to absorb disturbance and retain core functions), adaptability (the capacity of actors and institutions to adjust strategies under changing conditions), and transformability (the ability to create fundamentally new system configurations when existing structures become untenable) (Folke et al., 2010; Walker et al., 2004). For Mediterranean island destinations such as Vis, where tourism is the primary economic activity and the ecological base is under simultaneous pressure from climate change and land-use intensification, these capacities constitute the practical objectives of any meaningful adaptation strategy.

A variety of analytical tools are available to quantify climate impacts on tourism systems. Particularly useful are assessments based on climate–tourism indices, such as the Climate Index for Tourism (CIT), which provide insights into the suitability of weather conditions for tourism activities (Cvitan, 2024; de Freitas et al., 2008; Nicholls & Cazenave, 2010). This index integrates thermal (T), aesthetic (A) and physical (P) atmospheric components (de Freitas et al., 2008). The physical component addresses whether an activity can occur under conditions such as rain or strong wind, while the aesthetic component captures the importance of visual weather experience for positive destination perception (de Freitas et al., 2008). The CIT also employs a biometeorological framework that assesses perceived thermal comfort. It recognizes that tourist experience depends on the interaction of weather variables, not their isolated values. This enables assessment across a wide range of tourism activities, thereby assisting climate adaptation decision-making (Cvitan, 2024).

Existing tourism indicators and metrics, such as the Travel and Tourism Development Index, the GSTC Destination standard, and Copernicus indicators for tourism, consider LU/LC primarily through impervious or built-up land cover and forest cover. The former typically indicates tourism-related development that causes deforestation, habitat loss, or pressure on natural land cover. Research on climate vulnerability in tourism systems shows that changes are strongly mediated by physical landscape characteristics, including geomorphology, coastal configuration, elevation gradients, and LU/LC, including sea-level rise (Nguyen et al., 2016; Nicholls & Cazenave, 2010; Ružić et al., 2019). Assessing the combined effects of LU/LC and climate change is crucial for understanding ecosystem responses and hydrological dynamics (Cabral et al., 2024). Similarly, integrating bioclimatic and spatial analyses is essential for interpreting tourism patterns (Papageorgiou, 2025) and for planning adaptive transformations that ensure long-term destination sustainability. Addressing this gap is essential for advancing assessments of tourism resilience within specific destinations and for developing more comprehensive climate change adaptation strategies.

Climate–tourism indices, however, are often conducted independently of spatial analyses, creating a critical gap in contemporary tourism and climate research. Although climate indices have been used to assess Mediterranean tourism suitability at regional scales (Amelung & Viner, 2006; Perch-Nielsen et al., 2010) and LU/LC dynamics examined in coastal tourist areas (Papageorgiou, 2025), to our knowledge no study has systematically integrated both approaches at the island micro-scale to generate spatially differentiated adaptation strategies. This gap is particularly pronounced for small islands, where heterogeneous microclimates and land-use patterns create highly localized vulnerability profiles that aggregate-level analyses cannot capture.

This study presents an integrated climate–spatial analytical framework combining bioclimatic indices (CIT and PET) with LU/LC spatial analysis, designed to operationalize resilience assessment for 3S tourism destinations on small Mediterranean islands. Such an integrated analytical approach allows for spatial mapping of climate-related risk “hotspots” within a destination for a targeted type of tourism. Combining CIT data with LU/LC data, the suitability of 3S tourism was assessed for the Island of Vis, establishing a basis for identifying spatially heterogeneous vulnerabilities and implementing climate change adaptation measures for tourism in similar destinations. By integrating CIT S3 analysis with LU/LC data, this approach provides a replicable methodological structure applicable to comparable island contexts. Accordingly, this study is guided by three research questions:

• RQ1: How have observed and projected changes in bioclimatic suitability (CIT 3S and PET) altered the temporal distribution of ideal and acceptable conditions for beach tourism on the Island of Vis, and what implications do these shifts hold for the traditional peak-season tourism model?
• RQ2: How does spatial integration of bioclimatic suitability assessments (CIT and PET) with LU/LC analysis enable the identification of macrozones with different vulnerabilities and adaptive capacities, and what specific diversification and spatial adaptation strategies can be applied for the Island of Vis?
• RQ3: What theoretical and methodological prerequisites are needed to transfer the integrated climate–spatial analytical framework to other small Mediterranean island destinations, and under what conditions does the framework retain validity and utility?

2. Materials and Methods

2.1. Data Description

The analysis was done for the meteorological station Komiža (lon = 16.096304, lat = 43.042671). To evaluate thermal comfort using PET and CIT 3S, we used daily data for 2 m maximum temperature, 2 m mean relative humidity, mean cloud cover, 10 m mean wind speed, and precipitation amount (Bafaluy et al., 2014), covering the period of 1 January 1981 to 31 December 2020.

For the regional climate change assessment, we used four simulations produced with the RegCM model at a horizontal resolution of 12.5 km. RegCM was driven by four global climate models (EC EARTH, CNRM CM, MPI ESM, and HadGEM) under the RCP4.5 and RCP8.5 scenarios. This multi-model ensemble framework partly accounts for uncertainty in climate projections by capturing variability across different global climate model forcings, while the consideration of two RCP scenarios reflects a range of possible future emission pathways. The same set of meteorological variables used for the station analysis was extracted from the simulations for both historical (1981–2020) and future (2031–2060) periods. The historical period was constructed by extending the historical simulation beyond November 2005, which marks the end of the historical simulations defined by the EURO-CORDEX protocol, using data from the initial years of the RCP4.5 simulations. This approach is appropriate for the present study, as the early years of scenario simulations are only weakly influenced by greenhouse gas emission pathways and primarily reflect the continuation of present-day climate conditions. As a result, near-term RCP4.5 simulations can be treated as a physically consistent extension of the historical climate when analyzing climatological means and distributions, particularly for small-scale regional assessments where longer continuous time series are required (Kotlarski et al., 2014; Vautard et al., 2013). Variables for Komiža were obtained using bilinear interpolation to the model grid point closest to the station location. However, the relatively coarse spatial resolution (12.5 km) and the representation of the land–sea mask introduce inherent uncertainties, especially in coastal regions, which may affect the accuracy of the results. These issues are expected to be reduced in next-generation, convection-permitting simulations with a substantially finer spatial resolution (e.g., 3 km).

Being an island, albeit a small one, because of its remoteness and unique characteristics, Vis can be considered a region of its own within Adriatic Croatia. For spatial analysis of local-to-regional scale, data from (EU) the Copernicus program, (Croatian) the State Geodetic Administration, and OpenStreetMap were accessed and geo-processed. Copernicus GLO-30 DEM with a 30 m resolution was sufficient for basic geomorphological analysis, and the Corine Land Cover (CLC) Backbone 2023 raster with a resolution of 10 m was essential for LU/LC analysis.

2.2. Climate Index for Tourism (CIT) Assessment

The methodological approach of this study is based on CIT developed by de Freitas et al. (2008), which integrates the most relevant atmospheric conditions influencing tourism activities. The CIT combines thermal (T), aesthetic (A), and physical (P) components of weather conditions (de Freitas et al., 2008; Matzarakis, 2007):

CIT = f [(T, A) × P]

This structure enables an integrated evaluation of climatic suitability for specific tourism activities (Matzarakis & Amelung, 2008).

The thermal component represents the human energy balance between the body and the surrounding atmosphere and is expressed using a bioclimatic indicator reflecting perceived thermal comfort rather than energy flux values (de Freitas, 2003). In this research, thermal conditions were assessed using the Physiologically Equivalent Temperature (PET), a widely applied human–biometeorological index developed by Höppe (1993), Matzarakis and Mayer (1996) and Matzarakis et al. (1999). PET values were calculated using the RayMan Pro Version 2.2 model, which simulates short- and long-wave radiation fluxes in simple and complex environments. Standard physiological parameters, 0.9 clo (clothing), sitting position, and a fixed metabolic rate of 80 W, were applied according to Höppe (1993) and Matzarakis (2007). The model requires input data on air temperature, relative humidity, wind speed, cloud cover, and geographic coordinates to compute the mean radiant temperature and subsequently PET values. Thermal sensation was classified into nine categories based on PET values (

Table 1. Thermal sensation for PET.

Thermal Sensation	PET (°C)
Very cold	<4
Cold	4–8
Cool	8–13
Slightly cool	13–18
Comfortable	18–23
Slightly warm	23–29
Warm	29–35
Hot	35–41
Very hot	>41

), following Matzarakis et al. (1999).

The original CIT methodology defines seven classes of tourism climate suitability derived from weather-type matrices (de Freitas et al., 2008; Bafaluy et al., 2014). To simplify subsequent discussions in line with established theoretical and applied approaches in the literature these classes were aggregated into three operational categories:

• Unacceptable—climatic conditions limiting tourism activities (CIT = 1, 2, 3);
• Acceptable—conditions allowing tourism participation (CIT = 4, 5);
• Ideal—optimal climatic conditions for tourism (CIT = 6, 7).

2.3. Methodological Framework for LU/LC Data Analysis

The Island of Vis is a rather small Mediterranean island (89 km²) and, in the context of Croatia’s spatial development, a remote one too, being situated roughly 45 km (24 nmi) from the mainland. To examine its resilience further, an LU/LC analysis was conducted. Our approach was influenced by studies of climate and coastal vulnerability indices (Bukvic et al., 2020; Lobeto et al., 2024) that consider numerous variables such as coast type, elevation, slope, population, land cover, road network, infrastructure and many others. Because the research was conducted on a relatively small island, spatial scale represented an important consideration in the analysis (McLaughlin & Cooper, 2010). A key geographical challenge was the definition of the coastal zone of the Island of Vis, which is of particular importance for the assessment of 3S tourism. Therefore, the methodological framework was designed in accordance with the spatial resolution and thematic consistency of the available geospatial datasets, ensuring methodological consistency. All spatial data processing and analysis were performed by using the QGIS 4.0 software.

3. Results

3.1. Thermal Stress Assessment Based on Measured Data

The PET distribution at Komiža shows a clear seasonal cycle consistent with Mediterranean maritime climate (Figure 1). Winter months are characterized by mild cold stress, with only episodic occurrences of extreme cold. Spring transitions rapidly toward comfortable and warm conditions, driven by increasing insolation and sea-surface warming. Summer is dominated by the hot and warm PET categories, with July and August recording nearly 50% hot conditions reflecting the stable anticyclonic regime of the Adriatic. Autumn cooling is gradual, delayed by the thermal inertia of the sea. Comparing the two climatological periods (1981–2010 vs. 1991–2020), a clear warming signal emerges (Figure 2): colder categories decline in winter and spring, hot and very hot categories intensify in summer, and comfortable and slightly warm conditions extend further into autumn. This pattern is consistent with the broader regional warming trend documented for the eastern Adriatic.

Moving towards the more recent climatological reference period (1991–2020; Figure 1, right), PET conditions reveal a clear signal of warming that affects both the seasonal distribution and the frequency of thermal categories. Positive differences (red) in Figure 2 indicate an increase in “warm”, “hot” and “very hot” classes during summer and less cold categories in colder months. The coldest categories from November to early spring have negative differences (blue), which means that these months are also getting warmer and therefore more pleasant. In the referent period described as dominantly mild, winter in 1991–2020 has become even less dominated by cold stress. Spring warms earlier and more strongly. The decline in cold categories was replaced by more comfortable and warm conditions. Summer warming is seen as intensification of the hottest categories. The warm conditions in autumn persist for longer, extending into October and November. This reflects the thermal inertia of the Adriatic Sea, one of the key components of Mediterranean climate.

3.2. Assessment of the Touristic Potential for Sun–Sea–Sand Climate Index Based on Observational Data

Having established a detailed bioclimatic characterization of thermal conditions through the PET analysis, the next step was to examine how these thermal patterns translate into the broader context of coastal tourism suitability. The intra-annual variability in the relative frequency of CIT 3S categories reflects a clear seasonal cycle in climatic suitability, with a pronounced peak in July and August (Figure 3). In both observed climate periods (1981–2010 and 1991–2020), CIT 3S values remain predominantly unacceptable (CIT 3S 1–3) throughout winter and early spring. This is climatologically consistent with maritime climate conditions on the Island of Vis, where thermal and aesthetic factors remain unfavorable for 3S tourism until May. During spring, ideal conditions (CIT 3S 6–7) begin to emerge with rising PET values. The period of optimal weather for 3S activities extends from June through September, with the summer peak being in July and August, when the combined share of acceptable (CIT 3S 4–5) and ideal (CIT 3S 6–7) conditions exceeds 90%. This is consistent with the conceptual framework proposed by de Freitas et al. (2008), where optimal conditions correspond to warm, but not excessively hot, thermal states. Suitability for 3S tourism declines through October. By November, the distribution shifts back toward predominantly unacceptable conditions.

Although the distribution of the CIT 3S categories in both periods (Figure 3) appears broadly similar at first glance, comparison with the more recent climatological period reveals a reduction in the frequency of ideal conditions, particularly during the warmest part of the year (Figure 4). The differences in CIT 3S frequency between the two climatological periods (1991–2020 vs. 1981–2010; Figure 4) indicate a consistent shift in the seasonal suitability of weather conditions for 3S tourism. The most prominent signal is a reduction in the frequency of ideal conditions during the warmest part of the year. Declines in the ideal category are most evident between July and October (from −1.8% to −3.0%), suggesting a reduced occurrence of optimal conditions during the peak summer season. This pattern is consistent with higher summer temperatures (Figure 2). At the same time, the acceptable category increases, especially in summer (June–September), with the largest increase in August (+3.5%). This compensatory change indicates that while many days remain sustainable for beach tourism activities, they increasingly fall into lower comfort categories compared to the reference period.

3.3. Results of the Regional Multi-Model Ensemble

Daily simulated variables from each model were interpolated to the nearest grid point to Komiža. Based on this data, daily PET and CIT 3S values were calculated separately for each model. The analysis was performed for both the historical baseline period and the future 30-year period (2031–2060) under two emission scenarios (RCP4.5 and RCP8.5). Finally, a multi-model mean ensemble was constructed and discussed.

3.3.1. Historical Thermal Stress Assessment Based on Ensemble

The relative frequency of PET during the year obtained from the ensemble of modeled data for the historical climate (Figure 5) in the same two periods as in the analysis of measured data shows the same distribution shape with the modeled data lacking the warmest categories or having them occur less frequently. This clearly indicates that our ensemble of models gives a significantly cooler bioclimate compared to the actual climatic conditions in Komiža (Figure 1). The “very hot” category is not represented in the model ensemble for both periods, and only a slight positive increase in frequency for the “hot” category is observed towards the second period (Figure 5). A pronounced increase in the frequency of “warm” conditions is observed from June to September, accompanied by a reduction in “cold” and “very cold” conditions between December and April (Figure 6). The modeled data further suggests weaker spring warming compared to the referent climate, while a noticeable slowdown in autumn cooling is also evident.

3.3.2. Assessment of the Historical Touristic Potential for CIT 3S Based on Ensemble

The intra-annual variability in the relative frequency of CIT 3S categories based on ensemble for the two historical periods (1981–2010 and 1991–2020) (Figure 7) shows a seasonal pattern of climatic suitability. In both periods, conditions from December to March are entirely unacceptable for 3S activities. A similarly unfavorable situation persists in November and April, and more favorable conditions begin to emerge in May. With the beginning of summer, the frequency of suitable weather increases rapidly. This pattern reflects the combination of warm, but not excessively hot, thermal conditions, abundant sunshine, and minimal cloud cover, fully consistent with the conceptual definition of optimal CIT 3S categories. From September onward, climatic suitability declines rapidly.

A comparison of the two modeled historical periods (Figure 8) shows that the seasonal structure of CIT 3S suitability remains broadly consistent, with only modest (1–3%) but systematic changes. The most notable differences appear during the warm season: in June, the share of ideal conditions increases, while July and August show small but coherent increases in the ideal category, indicating slightly improved peak-summer suitability. September also shifts toward more favorable conditions, suggesting a modest extension of warm-season suitability. In contrast, winter and early spring show no meaningful change.

The increase in ideal 3S conditions in the model ensemble for 1991–2020 compared to 1981–2010 contradicts observational findings. This discrepancy could largely be attributed to the coarse spatial resolution of the regional climate models (RCMs), which struggle to resolve the highly indented Croatian Adriatic coastline and its complex archipelago. Consequently, the interpolation to the grid points closest to Komiža frequently samples sea-surface conditions rather than island land surface, introducing a systematic cold bias through lower simulated PET values. This maritime cold bias causes modest modeled warming to be interpreted as improved 3S suitability, diverging from observed thermal conditions.

3.3.3. Future Thermal Stress Assessment Based on Ensemble

Building on the observed changes in PET and CIT within the historical climate, we extended the analysis to assess how these indices may evolve in the near future. For the future climate, the period of 2031–2060 was examined and compared with the reference climate (1981–2010). The procedure for constructing the future-climate ensemble followed the same steps as for the historical period. Finally, changes were quantified by calculating differences in the relative frequencies between the future ensemble and the reference climate. Climate change signals were evaluated under two emission scenarios, RCP4.5 and RCP8.5 (Figure 9).

While RCP4.5 and RCP8.5 differ substantially in the strength of the warming signal, the underlying structure of change remains the same: a clear intensification of thermal conditions during the warm season and a marked reduction in cold-related categories throughout the winter months. The most substantial differences emerge in the core summer period. Under RCP8.5, the ensemble projects a pronounced shift toward the warmest PET classes in June, July, and August, with particularly strong increases in the warm and hot categories. RCP4.5 follows the same direction but with a systematically lower amplitude. A significant feature in both scenarios is the extension of warm-season characteristics into early autumn, which is consistent with the observed trend of delayed autumn cooling across the Adriatic (Lobeto et al., 2024). Conversely, winter and early spring shift towards milder PET categories, improving outdoor thermal comfort and expanding opportunities for outdoor recreation, however without reaching the thresholds required for 3S tourism.

To further compare real and modeled bioclimate conditions, we examined the annual number of days with PET exceeding 35.1 °C, corresponding to severe and extreme heat stress (Figure 10). Measured data shows a high and increasing frequency of days with PET above 35.1 °C (severe/extreme heat stress): from 49 days/year in the period of 1981–2010 to 57 days/year in the period of 1991–2020, with a linear trend of 0.7 days/year for both periods. In contrast, climate models substantially underestimate the absolute number of heat stress days, simulating around 1 day/year in the historical period and increasing to 6 and 9 days/year under the RCP4.5 and RCP8.5 scenarios, respectively. Although modest in absolute modeled terms, these projections suggest a relative increase within the model framework. However, given the pronounced underestimation of heat stress in the modeled baseline, these projections should be interpreted with caution, particularly when compared to the observed conditions. The ensemble trends are negligible in the historical simulations but increase to approximately 0.2 days/year under RCP4.5 and 0.4 days/year under RCP8.5. While this indicates a tendency toward intensification of heat stress in future projections, the magnitude of this change remains uncertain due to model limitations.

3.3.4. Assessment of the Future Touristic Potential for CIT 3S Based on Ensemble

To obtain climate change of CIT 3S, the difference in the relative frequency of CIT 3S categories for the future period of 2031–2060 (for both emission scenarios) and referent period of 1981–2010 was calculated (Figure 11). The analysis shows that both scenarios reduce the share of ideal 3S conditions during peak summer months, as stronger warming pushes thermal conditions beyond the optimal range. Scenario RCP8.5 amplifies this effect but also projects a more pronounced increase in ideal conditions in June and favorable conditions in September, indicating a clearer extension of the 3S tourism window compared to RCP4.5.

For a small island community, these future shifts in bioclimatic conditions carry profound socio-economic implications. Tourism revenues, which constitute the primary source of income for the majority of households, are at risk from reduced peak-season suitability. Simultaneously, the adaptation investments required shade infrastructure, water management, and diversification of attractions. Such a situation places financial burdens on the limited financial resource of the community. This asymmetry between climate risk exposure and adaptive capacity highlights the need for targeted regional and national policy support for small island tourism communities.

3.4. Results of LU/LC Analysis

As mentioned before, setting the boundary of the coastal zone was both important and challenging. The common distance (buffer) from the coastline approach did not seem to be adequate for defining a coastal zone on an island on which, according to our measurements for Vis, one cannot get farther than 3.5 km from the sea. The steep Adriatic karst coastline is also problematic for the definition of the coastal zone. The criterion quite often used in vulnerability studies of elevation measuring 5, 10, or even 20 m ASL for low coastal areas is also not suitable for places like Vis. After a closer inspection of the island, its coast, and settlements, a contour line for an elevation of 50 m ASL was created from GLO-30 DEM and used for the creation of the coastal zone in our study. This, still rather narrow coastal zone (area) clearly illustrates how steep and quite dramatic the coast is on the Island of Vis (Figure 12).

In the LU/LC analysis itself, special attention was given to determining the abundance (amount) of tree cover in the coastal zone. The tree-covered coastline in Croatia is a great resource because it mitigates the negative impacts of extreme heat and solar radiation, i.e., tree canopies offer tourists much desired shade on sunny and hot summer days. Unfortunately, this biomass is also a fueling source for wildfires and is especially vulnerable during the peak summer tourist season. In contrast to forests, sparsely and non-vegetated areas and especially built-up (artificial) or sealed areas heat-up easily and radiate heat—something that most are trying to avoid during heat waves.

The LU/LC analysis (Figure 13) was based on CLC Backbone (2023) raster data. The share of land-cover classes was calculated for the entire Island of Vis and compared to the coastal zone, an area defined by the 50 m contour line (

Table 2. The share of land cover classes for the Island of Vis and coastal zone of the island.

Coastal Zone (%)	Island of Vis (%)	CLC
6.91	1.58	Sealed
26.36	35.45	Woody needle leaved trees
none	≈0	Woody broadleaved deciduous trees
32.03	38.54	Woody broadleaved evergreen trees
17.54	15.91	Low-growing woody plants
2.72	3.76	Permanent herbaceous
1.04	2.11	Periodically herbaceous

).

Additional measurements were carried out to determine the infrastructure for the potential expansion or transformation of tourism on the Island of Vis outside the narrow coastal belt, which would be suitable for outdoor tourism (cycling, hiking, and recreation). The length of tracks, footways, pedestrian paths, residential streets, and steps from the OpenStreetMap database were measured and mapped (Figure 14). According to these feature classes there are some 217 km of tracks/paths or streets that are more suitable for walking than driving on the island. Officially, there are some 45.3 km on six hiking trails that are listed and mapped on the mountaineering association’s website (HPD, 2026).

4. Discussion

4.1. Temporal Shifts in Bioclimatic Suitability of 3S Tourism

The eastern Adriatic Sea, including the Croatian coastal and island system, represents a particularly sensitive sub-region due to its complex coastline, numerous islands, shallow northern shelf, and strong influence of terrestrial inputs (Malinović-Milićević et al., 2025). Observations and projections consistently indicate accelerated atmospheric and sea-surface warming in the eastern Adriatic, especially during summer, accompanied by reduced precipitation and more frequent droughts (IPCC, 2023; EEA, 2023). These changes pose substantial risks to 3S tourism by reducing thermal comfort and the climatic suitability of beach environments (Parete et al., 2024), particularly during peak summer months.

In our study, observed and projected bioclimatic changes on the Island of Vis reveal a clear temporal redistribution of favorable 3S tourism conditions, with direct implications for the traditional peak-season model (RQ1). The island already experiences more than 50 days per year, with PET exceeding 35.1 °C (strong heat stress), increasing at a rate of 0.7 days per year. Concurrently, the CIT 3S analysis demonstrates declining ideal conditions, especially in summer (June–September), with the largest increase in August (+3.5%). Future projections under both RCP4.5 and RCP8.5 scenarios indicate that annual strong heat stress days will increase from approximately one per year in the reference period to six and nine days respectively. These findings align closely with bioclimatic assessments across the Mediterranean basin. Amengual et al. (2011) applied CIT projections to Platja de Palma, Mallorca, and documented a similar pattern: fewer ideal 3S days during the summer peak and higher frequencies of acceptable conditions in spring and autumn, signaling a fundamental shift away from the current peak season (Amengual et al., 2011). PET-based assessments in Mediterranean tourism destinations corroborate this trend, detecting an upward shift toward stronger heat-stress classes and progression to extreme PET values under high-emission scenarios (Nastos et al., 2023). A similar change in beach-favorable thermal conditions has been found for places in Dalmatia in the warmer part of the year (Agapito et al., 2023). Although the model results consistently yield lower PET values than those observed, this situation does not degrade the CIT response but rather leads to an apparent improvement during the summer months. This pattern is primarily attributed to the limited representation of small Adriatic islands, such as Vis, at a horizontal resolution of 12.5 km, where local land–sea contrasts and island-scale processes are insufficiently resolved (EEA, 2023; Parete et al., 2024). Previous evaluations have shown that regional climate models, including RegCM, exhibit a modest but persistent summer cold bias over the Adriatic, particularly along the coast and on more remote islands. This bias has been linked to several factors, including the smoothing of complex coastlines at coarse grid spacing, the partial or complete sub-grid representation of small islands, excessive turbulent mixing within the planetary boundary layer, and uncertainties in sea-surface temperature forcing (Belušić et al., 2018; Denamiel et al., 2021; Reale et al., 2020). Consequently, near-surface air temperatures and derived thermal comfort indices such as PET tend to be underestimated during summer daytime conditions, while seasonal variability and relative changes remain robust.

The modeled and observed trends in severe heat-stress days highlight an important distinction between absolute and relative model behavior. The observed climate already exhibits a high frequency of days exceeding the PET threshold (35.1 °C), with a clear upward trend. Although the models underestimate the absolute number of such days, they consistently reproduce the positive trend signal. The relative increases projected for the future (from one to six days under RCP4.5, and from one to nine days under RCP8.5) represent a notable intensification within the modeled climate system. However, given the substantial underestimation of the baseline conditions, these projections should be interpreted with caution when compared to observations. When considered in the context of observed conditions, where the baseline already exceeds 50 days, even modest relative increases in the model simulations may correspond to considerable changes in actual thermal stress. In this sense, the models are better interpreted as indicators of directional change rather than precise predictors of absolute PET values, suggesting a tendency toward more frequent and prolonged heat stress in the Adriatic region.

When assessing CIT, it is important to acknowledge its limitations. In general, climate indices focus exclusively on climate. They are often static, not considering human adaptations (e.g., wearing appropriate clothing and using air conditioning) or destination-level adaptations (e.g., providing shade areas and water activities). They do not consider non-climatic factors that influence tourism decisions, such as infrastructure, accessibility, cultural attraction features, safety, costs, or marketing efforts. Although climate indices aim to provide a comprehensive measure of climate suitability for tourism, they often oversimplify complex climate conditions, failing to capture the complex interaction of local factors that influence thermal comfort (Scott et al., 2016). Vargas-Pérez et al. (2023) highlighted that while CIT may indicate a decline in optimal conditions, it does not adequately account for the adaptive capacity of destinations or the potential for changes in tourism patterns. This limits its ability to fully predict tourism demand or satisfaction. Furthermore, CIT’s reliance on fixed thresholds for assessing climate suitability may lead to misleading conclusions about destination attractiveness, especially given the significant interannual variability and changing nature of tourist preferences (Amelung & Viner, 2006). For example, a survey of French tourists (n = 1643) conducted by Dubois et al. (2016) showed that tourists have a high tolerance for heat and heat waves, while precipitation was a strong deterrent. The thresholds for perceived excessive heat varied by 2–3 °C, implying that indices using fixed thermal thresholds may misestimate impacts on tourist satisfaction and behavior.

It is obvious that climate adaptation will become a structural requirement for maintaining destination competitiveness. In practical terms, this means that given the presently degraded thermal environment and the robust direction of future warming, the Adriatic coastline can expect continued deterioration of optimal 3S conditions during peak summer, reinforcing the need for proactive planning and climate-sensitive tourism management. Amelung and Viner (2006) indicate a decline in summer climate suitability in the Mediterranean, shifting attractiveness to the pre-season and jeopardizing regional economic sustainability. Additionally, this study confirmed the projections of the Holiday Climate Index against Antalya visitation data, showing links between changes in climate suitability and tourist arrivals (Demiroğlu et al., 2020). Furthermore, Vargas-Pérez et al. (2023) provided results from simulations showing that climate impacts reduce the market share of island destinations, while anticipatory adaptation policies significantly mitigate these losses. Our findings for the Vis Island align with these nuances: while peak-summer ideal conditions decline, the extension of the favorable season into spring and autumn represents a genuine opportunity that adaptation strategies should explicitly target.

4.2. Spatial Heterogeneity, LU/LC Patterns and Differential Vulnerability

The integration of LU/LC analysis (Figure 13 and Table 2) with bioclimatic indices reveals marked spatial heterogeneity in vulnerability and adaptive capacity across the Island of Vis. This research goes beyond time-based climate assessments to gain insights into site-specific adaptation. Our spatial analysis shows that coastal areas exhibit higher closed land cover (6.91%) compared to the island average (1.58%), with reduced forest cover. Furthermore, the results of the analysis reveal less vegetation along the coast and beaches—woody needle leaved trees (26.36% coastal vs. 35.45% island-wide) and woody broadleaved evergreen trees (32.03% coastal vs. 38.54% island-wide). This means less natural shade and cooling effects are provided by forest cover in coastal areas. Most beaches on Vis are located in or around the two largest settlements—Komiža and Vis—which are characterized by limited capacity and difficult access (TBK, 2026). These land-use patterns concentrate thermal vulnerability precisely where tourism infrastructure and beach activities are most intensive. This creates a spatial mismatch between heat exposure and natural buffering capacity. In contrast, the interior of the island offers higher altitudes, extensive forest buffer zones, 217 km of trails and paths, and 45.3 km of official hiking trails (Figure 14). The area, declared a Vis Archipelago UNESCO Global Geopark (2019), provides resources for thermal refuges and alternative tourism products. This spatial differentiation is crucial: while coastal areas experience the double pressure of increasing heat stress and limited natural cooling, the interior retains microclimatic advantages and diversification potential that are underutilized in the current tourism model.

This spatial differentiation is crucial: while coastal areas experience the double pressure of increasing heat stress and limited natural cooling, the interior retains microclimatic advantages and diversification potential that are underutilized in the current tourism model. The importance of LU/LC integration is well established in the climate–tourism literature. García et al. (2025) linked tourism-driven urbanization in Cyprus to expanding urban heat hotspots and localized surface warming, demonstrating the spatial imprint of land-use change on thermal conditions. Their findings underscore that sealed surfaces and vegetation loss amplify heat stress in tourist zones. Olya and Alipour (2015) applied GIS-based tourism climate indices to Northern Cyprus. With this approach, they demonstrated the operational utility of spatially explicit climate–tourism diagnostics, which can be considered a highly useful tool for tourism destination management.

The spatial heterogeneity documented in this study supports a differentiated adaptation strategy. To mitigate concentrated heat stress, coastal zones require targeted interventions, such as green infrastructure, shading structures, permeable surfaces, and microclimate management. Simultaneously, the island interior should be actively developed as a complementary tourism space, leveraging its natural cooling capacity, trail networks, and geopark designation to diversify the product mix beyond beach-centric offerings. This spatial diversification aligns with resilience principles that emphasize functional redundancy and response diversity. This implies a distribution of tourist activities in several environments with different thermal profiles, reducing the systemic vulnerability to climate disturbances (Folke et al., 2010; Walker et al., 2004).

4.3. Resilience, Diversification, and the Transition Beyond 3S Dependence

Based on the temporal and spatial findings of this study, there is a strategic imperative: tourism on the Island of Vis should shift from a concentrated, seasonal 3S model towards a diverse, spatially distributed, year-round tourism system (RQ2). Within the framework of socio-ecological resilience (Folke et al., 2010; Walker et al., 2004), the island currently retains the capacity for gradual adaptation. This is reflected in adaptations within the existing system, but continued overreliance on a single climate-vulnerable product risks crossing thresholds that would require a far more disruptive transformation. The island’s spatial assets directly enable this transition. The network of inland trails, the geopark, cultural heritage, wine tourism and gastronomy offer active and experiential products that are not thermally limited and can function effectively in transitional seasons when coastal conditions remain favorable. Diversification does not mean abandoning 3S tourism but rather its inclusion in a broader portfolio that reduces the system’s dependence on climate. Čavlek et al. (2019) showed that destinations that proactively redirect towards activities like cultural, wellness and outdoor active tourism significantly outperform those that wait for crises. Vargas-Pérez et al. (2023) confirmed this with agent-based simulations of European island tourism, showing that islands implementing operational resilience measures maintain competitiveness even in adverse climate scenarios. Integrating spatial dimensions with CIT assessments enables more targeted adaptation planning, identifying not only when conditions deteriorate but where on the island alternative tourism resources exist and how they can be activated.

4.4. Framework Transferability, Methodological Considerations and Limitations

An integrated framework combining CIT/PET bioclimatic assessment with GIS-based LU/LC analysis and regional climate projections is designed to be transferable to comparable small Mediterranean island destinations (RQ3). From a theoretical standpoint, this study advances the application of social–ecological resilience theory to small island tourism systems by providing an empirically grounded, spatially explicit operationalization of adaptive capacity assessment. However, three prerequisites must be met. First, the following data requirements must be satisfied: high-quality meteorological observations (temperature, humidity, wind speed, and cloudiness) with a sufficient temporal resolution, spatially explicit LU/LC datasets with a resolution ≤10 m, and bias-corrected regional climate projections. Second, calibration represents an important methodological step. Namely, regional climate model outputs should be evaluated and corrected for local biases before being used in index calculations. The cold bias of our study on small Adriatic islands illustrates the consequences of omitting this step. In addition to the presented framework, the authors believe that it is equally important for future research to ensure local validation of the results. Dubois et al. (2016) have shown that the categories of “ideal” and “acceptable” vary across tourist populations and cultural contexts and need to be calibrated through visitor surveys or stakeholder consultations (Dubois et al., 2016). Third, stakeholder engagement is key. Matzarakis (2013) established the Climate Tourism/Information Transfer Scheme (CTIS) as a portable planning tool, emphasizing that local adaptation of index results and co-production with destination stakeholders are essential for legitimacy and acceptance.

There are several limitations of the current study. Relying on a single meteorological station (Komiža) and climate model results with a 12.5 km grid may not fully capture microclimate variability across the island’s topography. Higher resolution simulations and remotely sensed land surface temperature data would improve spatial accuracy. The Copernicus LU/LC dataset, while comprehensive, cannot resolve fine-scale features such as individual shade structures or small vegetation areas that affect local thermal comfort. Furthermore, this study was not designed to quantify tourists’ perceptions and behavioral responses to changing conditions. Given that the effectiveness of adaptation ultimately depends on visitor perceptions and local community knowledge, this element is important to include in future research. We believe that despite these limitations, the framework provides a robust, evidence-based foundation for adaptation planning and a replicable methodological template for small Mediterranean islands facing similar climate tourism challenges.

5. Conclusions

This study presents an integrated climate–spatial analytical framework for assessing the resilience of 3S tourism on the Mediterranean island of Vis in Croatia, combining bioclimatic suitability assessments (CIT 3S and PET) with spatial LU/LC analysis. The key findings are threefold. First, bioclimatic suitability for 3S tourism is seasonally declining: the island already experiences more than 50 days per year of severe heat stress (PET > 35.1 °C, with an increase of 0.7 days per year), ideal CIT 3S conditions are decreasing from July to October, and future projections indicate further intensification under RCP4.5 and RCP8.5, while favorable conditions extend into the transitional seasons. Second, the spatial analysis reveals a critical vulnerability mismatch. Coastal areas, where tourism is concentrated, show larger closed areas, reduced forest cover, and limited natural cooling capacity, while the interior of the island offers forest buffer zones, an extensive network of trails, and a UNESCO Global Geopark that remains underutilized as a tourism resource. Third, the integration of these two analytical dimensions—temporal bioclimatic trends and spatial LU/LC patterns—allows for the consideration of more specific adaptation strategies that neither approach achieves on its own. By applying this integrated approach, it is possible to identify not only when conditions are deteriorating but also where alternative tourism adaptation resources exist.

The primary contribution of this study is the methodological advancement of analytical approaches aimed at detecting the degree of climate resilience of small remote islands, which is crucial for informed decision-making on possible adaptation strategies. The systematic spatial integration of CIT/PET bioclimatic indices with LU/LC analysis at the micro-island scale goes beyond aggregate climate tourism assessments to create actionable site-specific strategies. For Vis, these strategies prioritize season extension, spatial diversification that exploits the island’s interior resources, targeted coastal interventions (vegetation preservation and shading infrastructure), and planning processes that co-design adaptation pathways with local stakeholders. This climate–spatial analytical framework could be transferred to comparable Mediterranean island destinations, depending on the availability of local data, correction of climate model output bias, and calibration of index thresholds. Future research should address the identified limitations by incorporating visitor behavior data and local knowledge regarding climate issues and multi-hazard perspectives. Longitudinal monitoring of bioclimatic trends, LU/LC dynamics, and tourism flows will be essential to assess the effectiveness of adaptation over time.

Small Mediterranean islands like Vis face increasing pressures from climate-related tourism, but their spatial diversity, environmental management resources, and potential for community engagement, if strategically mobilized, can support a resilient tourism future. The integrated framework developed here offers a replicable, evidence-based basis for this process.