Tourism is an important economic driver for numerous regions, attracting more than one billion visitors annually. While economically significant, excessive numbers of visitors lead to local overcrowding, which negatively impacts visitors’ experience and safety, and causes environmental harm. This paper proposes a practical approach to empowering destination management organizations (DMOs) to manage tourist flows. We advocate for a rule-based approach that models visitor occupancy based on easily understandable influence factors like weather and date. As a central component, an ontology-guided knowledge graph ensures compatibility with diverse touristic data models and allows seamless integration into existing infrastructures. By digitizing DMOs’ experiential knowledge, we facilitate the implementation of lean and cost-effective visitor guidance. We demonstrate our approach by implementing two applications for two different use cases. The results of our qualitative evaluation reveal the compelling potential for rule-based occupancy modeling approaches serving as a baseline for future visitor management systems.
1 Introduction
Tourism has long been a vital economic driver for numerous regions in the world, attracting visitors to explore inspiring landscapes and cultural offerings. Despite many benefits, the rapid growth of global tourism comes at the price of overcrowding, which is well-known in popular places such as Venice, Barcelona, and Dubrovnik [1] but also increasingly affects smaller rural areas [2]. Overcrowding not only leads to decreased visitor satisfaction [3] but also poses a higher risk to visitor safety and adversely impacts the surrounding landscape, nature protection efforts, and biodiversity [4, 5]. This raises the need for effective visitor management.
Data-driven methods hold potential for this task [6]. However, while advanced technologies enable the accurate tracking and forecasting of the expected visitor traffic [7], many destination management organizations (DMOs) still struggle with hurdles in implementing such measures, often caused by a lack of technical expertise and financial resources [8, 9], making widespread adoption of expensive technologies infeasible. Even without expensive technology, however, DMOs possess invaluable experience in their regions’ high-traffic conditions and peak periods. Therefore, one viable option would be to digitize this knowledge as a digital asset and use it for visitor management.
Anzeige
This paper proposes a lean reference concept to empower DMOs to digitize their experiential insights on visitor traffic, making them available for various touchpoints. We advocate for a rule-based approach based on easy-to-grasp factors like weather, temperature, season, date, and holidays, which is cost-effective and accessible to DMOs with limited technical expertise and financial resources. Employing a knowledge graph as our central element fosters compatibility with a wide range of touristic data models and seamlessly integrates into existing infrastructure. We demonstrate how manually captured occupancy rules can be used on exemplary touchpoints to guide visitors with tailored recommendations.
2 Background
For DMOs, overcrowding of specific areas [4] is a well-known issue. Overcrowding can be described from two standpoints: the subjective visitor perception of the crowdedness of an area [10, 11] and the objective capacity of an area to offer hospitality services (i.e., parking lots, hotels, availability of adjacent free-time activities) [4, 12]. Besides nature protection agencies and emergency services that manage access to certain places due to specific situations, such as wildlife protection or natural events (e.g., avalanches and floods), DMOs manage visitor traffic, focusing on the visitors’ experiences [13]. Considering multiple stakeholders, DMOs adopt a more holistic perspective and act on various aspects, including visitors’ impact on the environment [14], visitor safety, and visitor satisfaction, all by guiding visitors to mitigate overcrowding [15, 16].
DMOs’ visitor guidance varies from stricter closures [17] to nudging (i.e., a behavioral economics concept where subtle suggestions influence human decision-making) [18] by delivering recommendations to alternatives in order to distribute visitor traffic evenly [13]. Nudging is especially effective due to preserving visitors’ feeling of agency and satisfaction while empowering the tourist to choose a more sustainable option [19, 20] and the higher willingness of today’s tourists to actively search for better experiences using digital tourist solutions [21]. A delivery system for recommendation nudges can be a brochure, information board or a QR code, the latter having the advantage of delivering dynamic information, making it the optimal low-cost scalable solution [22].
Real-time and future occupancy estimates are key requirements for fast responses indispensable for managing overcrowding and ensuring the overall resilience of the destination [7, 23]. Highly dynamic data such as tourist arrivals, bookings, parking availability, weather, holidays, and web traffic have proven successful in measuring and predicting occupancy in real-time, even for areas only partly covered by sensors [7]. Nonetheless, many DMOs still struggle with implementing such measures due to a lack of technical and financial resources [8, 9] and data privacy concerns [21, 24, 25].
Anzeige
Our approach addresses several gaps left in the current literature. Past studies focus on the effectiveness and acceptance of visitor guidance practices and leveraging digital technologies for spatiotemporal occupancy prediction but fall short of exploring cost-effective, scalable methods. Installing sensors can be expensive [26], and the data architecture [27] and maintaining prediction models requires further financing. In the following sections, we introduce a cost-efficient approach based on digitizing DMO experiences concerning spatiotemporal occupancy that can be used out-of-the-box even by DMOs with a limited level of digitization. DMOs with a higher level of digitization can integrate our approach with existing sensors or other occupancy prediction systems, striking the perfect balance between accessible, accurate, and scalable.
3 Method
Our overall research process follows a design science paradigm [28] since it is an effective approach for generating and evaluating practical solutions. We implement the six-step framework of [29], comprising (i) motivation, (ii) solution objectives, (iii) conceptualization, (iv) demonstration, (v) evaluation, and (vi) communication (covered by this publication). Guided by the motivational scenario and solution objectives outlined above, we conceptualized our framework and developed prototypes in multiple iterations, incorporating continuous expert feedback.
Within the scope of four expert workshops with different German DMOs, ranging from alpine to seaside tourism, we determined key factors to be incorporated into a rule-based occupancy model. Considering the widespread adoption of knowledge graphs and ontologies in tourism (e.g., for cross-DMO data sharing) [30‐32], we designed our approach to extend existing knowledge graphs by occupancy values (i.e., measurements, predictions, and rule-based occupancies). Based on the determined factors for occupancy rules, we developed the tourism occupancy ontology (TOO) following best practices, including the NeOn ontology engineering framework [33]. To demonstrate and evaluate our work, we identified two relevant use cases within the domain of German outdoor tourism. We developed two simple, easy-to-use prototypes: (i) a web-based user interface for DMOs to incorporate occupancy rules and (ii) a progressive web app (PWA) for on-site visitors that can be opened by scanning stationary-mounted QR codes. The choice of a QR-code-based application as an exemplary touchpoint resulted from the selected use cases. Our approach was evaluated as a workshop with each use-case-specific DMO, discussing our end-to-end solution in the context of the DMO’s specific requirements.
4 From Rule-Based Occupancy to Visitor Guidance
Our proposed approach consists of different components. DMO representatives capture rules based on certain influence factors and describe different occupancy levels as a function of these factors (e.g., when it is sunny, beach A is crowded). The resulting rule sets are persisted in a knowledge graph alongside touristic entities (e.g., POIs and tours) and optionally other occupancy descriptions stemming from alternative sources (e.g., sensors, predictions). A rule inference component evaluates the captured occupancy rule sets based on a given context (e.g., sunny), turning them into realized occupancies relating to specific entities. The contextual data is gathered from local or third-party sources (e.g., weather API). Based on this information, recommendations (e.g., beach A is likely to be crowded, go to beach B) are generated and displayed to visitors using one or multiple touchpoints.
4.1 Overall Process
The end-to-end process, from capturing occupancy rules to deploying visitor guidance, is summarized in Fig. 1. Its components are outlined below.
Fig. 1.
Communicating occupancy rules to visitors (overall process). Solid arrows indicate the flow of data and information, the dashed line indicates the knowledge graph’s data semantics.
×
Occupancy Rules.
To turn their experiences into formal rules, DMO representatives leverage a set of pre-defined influence factors. During our expert workshops in the first step of our research, the most influential factors described comprised (i) weather conditions (e.g., cloudy, sunny, rainy), (ii) temperature ranges, (iii) temporal conditions including months and seasons, as well as (iv) public and school holidays. Notably, we found that offering additional influence factors would be counterproductive, as it contradicts the simplicity of the rule-based approach and complicates collaboration.
Occupancy Knowledge Graph.
Knowledge graphs (KGs) [34] have evolved as a de facto standard for data management in tourism [31] and are used in a wide range of projects, including for open data sharing (e.g., [35, 36]). They offer a flexible data structure comprising nodes (i.e., instances) and edges (i.e., relationships) linking them. KGs are often guided by ontologies [37], formal conceptualizations of real-world entities that capture the semantics of entities in the graph and serve as the graphs’ underlying schema [31]. The occupancy KG extends existing touristic KGs by incorporating occupancy as an additional concept, which can be linked to geographic objects, such as POIs, areas, or hiking tours. Its additional semantics are contained in the touristic occupancy ontology (TOO). The occupancy KG serves as an application-agnostic data source. Instead of being bound to one specific application, it can serve multiple applications that may realize different visitor touchpoints.
Rule Inference.
The rule inference component evaluates formal rules given a specific context (e.g., current weather and time of the year). To determine the validity of configured rules, it acts on top of the attached TOO following the semantics described. As a result, it passes entities with their determined occupancies to different touchpoints that process these results and display them to visitors.
Touchpoints.
Visitor touchpoints refer to any point of contact with current or potential visitors (e.g., destination websites, digital kiosks, flyers, tourist information points). In previous work [38], we identified high-impact touchpoints in collaboration with DMOs suitable for visitor guidance concerning the different stages of a visitor’s travel. Within the scope of Fig. 1, touchpoints are standalone applications that realize services to visitors, including the recommendation of less-crowded touristic offerings. These applications act on top of the occupancy KG and the contextual occupancies inferred by the rule inference component and may further process this data.
4.2 Tourism Occupancy Ontology
The occupancy knowledge graph is a crucial component and contains formal rules to determine the crowdedness of geographic objects (e.g., POIs, hiking tours, and areas). While existing touristic ontologies offer a wide range of relatively static data types (e.g., POIs, events, hiking tours), they fall short of modeling dynamic context, such as spatiotemporal occupancy [31]. Therefore, we developed the tourism occupancy ontology (TOO) by following best practices of ontology engineering [33]. The TOO provides an abstraction of geographic entities, which makes it compatible with a wide range of existing related ontologies, including schema.org and its extensions [39], as well as the FIWARE Smart Data Models [40].
Figure 2 depicts an abstract overview of the TOO, including its mapping to schema.org. In its current version, the TOO comprises three components (i.e., geometry, occupancy, and occupancy rules) and supports modeling the occupancies of arbitrary geographic objects represented by the \(GeoObject\) class. Specifications of this class are basic geometric building blocks and include points (\(GeoPoint\)), lines (\(GeoLineString\)), and polygons (\(GeoPolygon\)), each of which can have arbitrary relationships to each other (e.g., polygon intersections and point containments). Geometric objects are identified by appropriate attributes (e.g., coordinates). Moreover, these objects serve as links to more concrete touristic entities, such as hiking tours or POIs, which can be imported using other well-established ontologies, as exemplarily shown for schema.org using the \(owl:equivalentClass\) axiom. An occupancy is characterized by different values, including a score between 0 and 1 and additional metadata (e.g., creation date). While the TOO aims at supporting different types of occupancies, namely measured occupancy (e.g., occupancy from sensors), predicted occupancy (e.g., occupancy derived from machine learning algorithms), and rule-based occupancy, the rest of this paper will only regard rule-based occupancies. Each occupancy type has conditions for when a given occupancy is considered valid. In rule-based occupancy, this validity is determined by whether its respective occupancy rule set matches a given context.
Fig. 2.
Components of the TOO including an exemplary mapping to schema.org concepts. Dashed arrows denote sub-class relationships, solid arrows refer to non-taxonomic relations. Geometric relations and attributes are omitted to ensure better readability.
×
Occupancy rule sets are specifications of when instances of \(RuleBasedOccupancy\) apply. They are collections of logically conjunct conditions, that is, a specific occupancy is triggered if and only if all conditions hold. Following our expert workshops for rule-based occupancy modeling, we identified five specific condition categories: weather conditions, temperature, date and time, public holidays, and school holidays.
5 Visitor Guidance for German Outdoor Tourism
To showcase the efficiency of our concept, we developed a web-based prototype for applying visitor guidance to the exemplary domain of German outdoor tourism. We instantiate our solution on two specific use cases, namely (i) navigating visitors around a lake using a route as under-occupied as possible and (ii) recommending low-occupancy POIs and hiking tours nearby. We detail both use cases (UC) in the following.
UC1: UC1 relates to a popular bike trail in Germany. A notable hotspot along the trail is a lake, whose northern shore tends to suffer from severe overcrowding on certain days. Thus, the goal of UC1 is to recommend on-site visitors to opt for the southern shore on occupied days.
UC2: UC2 refers to one of Germany’s most popular alpine tourist destinations, well-known for its famous castles. During the summer months (i.e., June to September), the destination becomes a significant hotspot with only a few less-occupied alternatives. Besides local POIs and hiking tours, overnight guests should partly be recommended less crowded places in a larger radius to relax local overcrowding while offering visitors the best memories possible.
In line with the overall data flow described in Fig. 1, our demonstration consists of three parts. First, occupancy rules are captured by DMO employees using a web-based application. Second, an inference API evaluates the pre-defined rules within a given context and returns the actual occupancy condition to the requesting client. Third, a mobile application with a UI configurable specific to the use case displays recommendations for less-crowded alternatives to on-site visitors.
Fig. 3.
Occupancy rule management based on polygons. The map1 shows the selection of an exemplary polygon with two rule sets attached (see right panel).
×
5.1 Occupancy Rule Management
The first application (Fig. 3) implements a simple web-based user interface (UI) for DMO employees. It covers all necessary aspects of the TOO. The rule-based occupancy configuration is based on visually drawn polygons \(p\in GeoPolygon\), which can have attached multiple rule-based occupancies \(rbo\in RuleBasedOccupancy\) with a rule set \(rs\in OccupancyRuleSet\) that comprises logically conjunct conditions \(c\in Condition\) (e.g., weather, time). Instead of a continuous value score \(s\in [0, 1]\), the UI has been restricted to allow only a set of categorical occupancy values \(v\in \{low, medium, high\}\), which map to \(0\), \(0.5\), and \(1\), respectively. This results from our initial expert workshops and the intention to keep the UI simple and tangible.
Polygons can receive different rule-based occupancies at once, each resulting from the applicability of different rules. To apply the specified rules to touristic objects, POIs, and hiking tours are identified within the polygon’s boundary by evaluating the geometric relations specified by the TOO (e.g., point containment and intersection). Besides rules and conditions, a short description of each polygon’s rule set can be attached for better collaboration of different co-workers of the same DMO.
5.2 QR Code Touchpoint for Local Visitor Guidance
In both use cases, we mainly plan to reach on-site visitors via stationary-mounted QR codes to keep the DMOs’ efforts small and minimize the risk of vandalism. Scanning a QR code leads to a location-aware progressive web app (PWA), which can be configured to the use case’s specifics. Figure 4 showcases the resulting PWAs.
For UC1, a QR code is intended to be installed on a physical sign on the inside wall of a ferry positioned shortly before the lake, which most visitors of the trail use. In line with the workshop key outcomes, the linked PWA highlights two alternative routes: one via the northern shore and the other along the southern shore. The occupancy of each route is modeled using a polygon surrounding the respective route and having a set of rules attached. After rule evaluation, routes are colored green, orange, or red, referring to a low, medium, or high occupancy, respectively. To nudge visitors, interesting POIs (defined by the DMO) are shown alongside the less crowded route. If the conditions of at least one occupancy rule set hold, a generated statement is shown that explains that overcrowding is expected.
Fig. 4.
PWA with maps2 configured for UC1 (left) and UC2 (right).
×
Regarding UC2, QR codes are meant to be printed on flyers or displayed in the lobbies of local hotels. In contrast to UC1, where only two local route alternatives are shown, UC2 intends to guide visitors within a radius of 45 km. Thus, the occupancy is modeled in a much wider area. While the DMO decided not to directly display the determined occupancy polygons, the recommender only suggests POIs and hiking tours with low to medium occupancy while giving low-occupancy objects higher ranks. Moreover, due to the large number of possible recommendations, a recommendation dialog may be opened to answer questions regarding personal preferences, including favorite activities, fitness level, and cultural interest (the latter may lead to routes along castles or historical buildings).
6 Evaluation
Based on the two use cases outlined above, we evaluated our end-to-end solution in use-case-specific workshops with the respective DMOs, aiming to assess the feasibility and potential of our approach for managing tourist flows in overcrowded destinations. The workshops involved various experts and key stakeholders, including DMO representatives in leading positions and individuals from independent academic and research institutes. Each workshop consisted of two parts. First, to give participants a general understanding of our approach, we gave a complete overview of the use-case-specific instantiation of our concept, that is, the occupancy rule management UI and the accordingly configured PWA. Second, we processed a comprehensive catalog containing all built-in functionalities of our exemplary applications step-by-step, discussing the use-case-specific relevance of each functionality to map the collected feedback to our more general concept.
Despite demonstrating two very different instantiations of the same concept, the overall perception of our demonstrations across all workshops was remarkably positive, and our solution generated great interest among the involved DMOs. What participants found particularly insightful was the simplicity and scalability of the concept of digitizing DMOs’ experiences for overcrowding. Also, the participants were convinced that the accuracy of a rule-based occupancy should be sufficient for most of their use cases concerning visitor guidance, which primarily requires rough occupancy estimates rather than concrete numbers.
While the experts of UC1 argued that a rule-based occupancy assessment is specifically helpful for them due to the occurrence of significant but relatively infrequent crowding events, where expensive sensors might not pay off, UC2 suffers from many high-traffic hotspots. It recognizes the usefulness of our approach, particularly in widespread adoption. Although the destination of UC2 already possesses a relatively large number of sensors at the most frequented spots, the intended area with a radius of 45 km cannot be covered by sensors alone; the DMO, therefore, strives for a rule-based solution stating, however, that other DMOs must be involved in the capturing process of relevant occupancy, as not every single part of this area falls under their responsibility. This underscores the need for decoupling information and data sharing, a key competency of knowledge graphs, which are an integral part of our approach.
Another aspect attracting considerable interest during the evaluation workshops was our PWA and the fact that it is activated using a stationary-mounted QR code. The participants agreed that QR codes are affordable, easy to maintain, and highly scalable. They are suitable for ad-hoc experiments while offering sufficient room for personalization. According to UC2, QR codes could be tailored to the hotels’ needs by adding their design and related service offers (e.g., transportation to the recommended places and lunch packages), which may also increase the hotels’ willingness to adopt such a measure. This finding confirms the necessity of decoupling individual touchpoints from the knowledge graph’s data basis.
7 Conclusion, Limitations, and Future Research
In this paper, we proposed a theoretical concept and a practical solution for a lean, cost-effective, and privacy-preserving reference concept for multi-touchpoint visitor guidance with low onboarding requirements, delivering accessible solutions to DMOs in all stages of digitization. Our approach is based on digitizing DMO’s experiences on the occupancy of different areas of their destination into rule-based descriptions, constituting a novel combination of experiential knowledge with data-driven methodologies, enriching traditional visitor management strategies [41] and serving as a baseline for occupancy-aware methods [13, 42]. Furthermore, we provide the TOO, a touristic ontology capable of modeling occupancy data, a known research gap [31]. Practically, we provide DMOs with a feasible and scalable solution for mitigating overcrowding situations cost-efficiently and two elaborated examples for a concrete implementation. Our approach is highly accessible, even for DMOs without technical expertise. DMOs with a higher level of digitization can incorporate our approach into their existing infrastructure without significant disruption.
Like other work, this work suffers from certain limitations. First, our initial workshops to determine occupancy influence factors, as well as the investigated use cases and their evaluation, are subject to the domain of German outdoor tourism and may miss out on other perspectives. Second, while valuable, our evaluation is based on qualitative workshops limited to two use cases and needs further investigation, including additional use cases and (quantitative) user studies.
For future research, we plan to strengthen our results by deploying our prototype at scale, incorporating real users, and measuring the deployed application’s impact on visitors’ (re-)distribution. Furthermore, since the occupancy of our rule-based approach only reflects the assessment of the capturing DMO, we aim at a more granular investigation of the trade-off between simplicity and factors that should be included to obtain a sufficiently accurate occupancy estimation. Moreover, we motivate future research to investigate multistakeholder occupancy calibration to collect the occupancy-related perceptions of visitors and further stakeholders to calibrate the possible thresholds of occupancy levels. Finally, future applications should incorporate values from different sources (e.g., sensors and predictions) to obtain a more diverse view of overcrowding.
Acknowledgements
This work is a result of a cooperation between Outdooractive AG, the Technical University of Munich, and the University of Applied Sciences Kempten. This study was supported by the AIR research project (67KI21005B), funded by the German federal ministry for the environment, nature conversation, nuclear safety, and consumer protection.
Open Access This chapter is licensed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license and indicate if changes were made.
The images or other third party material in this chapter are included in the chapter's Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the chapter's Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder.
