Smart Tourism for All: Optimizing Rental Hub Locations for Specialized Off-Road Wheelchairs Using Spatial Analysis

Highlights

What are the main findings?

• Micro-slope analysis revealed hidden terrain traps, disqualifying 55% of standard trails to ensure operational safety for specialized off-road wheelchairs.
• The “Wilderness Last Mile” concept extends the Smart City mobility chain into protected areas, bridging the gap between urban transport and nature tourism.

What are the implications of the main findings?

• Standard tourist maps hide critical barriers, creating safety risks for users of assistive mobility devices in mountains.
• The study provides a ready-to-use tool for locating inclusive rental spots, transforming raw terrain data into safe tourism products.

Abstract

The development of Smart Tourism often overlooks the “Wilderness Last Mile”, leading to the spatial exclusion of people with disabilities in mountain areas. This problem exists because standard tourist maps and urban-centric accessibility models rely on averaged terrain data, failing to identify critical micro-scale barriers (e.g., short, sudden steep ascents) that pose severe safety and traction risks for off-road wheelchair users. To address this gap, this article presents a novel GIS methodology for planning accessible off-road tourism for electric Specialized Off-Road Wheelchairs. The proposed four-stage analytical model includes (1) graph-based trail network topologization to enable precise routing; (2) traction safety verification utilizing high-resolution (1 × 1 m) Digital Elevation Model (DEM) micro-segmentation to detect hidden slope barriers; (3) multi-criteria evaluation combining a user-calibrated Difficulty Index (EDI) and a Tourism Quality Index (TQI); and (4) a hub optimization algorithm that prioritizes locations maximizing the diversity of accessible routes. The method was empirically tested in a case study of the Bieszczady Mountains (Poland), calibrating the model with the technical limits (25% max slope) of a prototype wheelchair. The experimental results clearly validate the model’s superiority over traditional approaches: the micro-segmentation successfully identified hidden terrain traps, disqualifying 55% of the standard trail network that would have otherwise been deemed safe by average-slope assessments. Furthermore, the model identified a contiguous safe network of 153 km and pinpointed the optimal rental hub location, ensuring the highest inclusivity and route variety. Ultimately, this approach transforms raw spatial data into safe, ready-made tourism products, providing a precise tool with which to implement Universal Design in natural environments.

1. Introduction

1.1. Research Background and Existing Problems

The concept of Smart Tourism has evolved in recent decades from a niche technological trend into a fundamental element of development strategies [1] for sustainable cities and regions [2]. In the literature, it is defined as an integrated ecosystem combining Information and Communication Technologies (ICTs) with physical tourism infrastructure [3,4], aimed at enhancing traveler experiences, increasing destination management efficiency, and promoting sustainable development [4,5,6]. Contemporary approaches to Smart Cities increasingly incorporate tourism into Sustainable Urban Mobility Plans (SUMP) [7], aiming to shift transport habits through smart traffic control, Park-&-Ride systems, or the full digitization of ticketing services [8,9]. Dynamic passenger information systems, providing real-time data, have become standard in urbanized areas, facilitating multimodal travel [10,11]. However, this paradigm cannot be limited solely to the city’s administrative boundaries. In line with the “Smart Region” concept, a smart city must assume responsibility for the mobility of its residents also within functional zones, including recreational areas to which these residents travel [12,13,14].

Therefore, ensuring so-called “Seamless Mobility” becomes a key challenge [15,16]. A resident of a smart city, utilizing accessible low-floor rolling stock daily, should not lose their autonomy the moment they leave the agglomeration [17,18]. Despite dynamic technological progress, inclusivity remains a significant and often overlooked aspect within the Smart Tourism paradigm, particularly regarding people with disabilities [19,20,21].

The general background of this research is bridging this gap. While urban environments have seen significant accessibility improvements, protected mountain areas remain largely inaccessible to tourists with mobility disabilities. To systematically address this, it is necessary to identify and solve three highly specific, interrelated problems:

• Problem 1: Severe Spatial Exclusion in Natural Terrain. Tourists using assistive mobility devices are practically confined to paved, urbanized zones. Natural areas lack designated, verified off-road routes and the necessary supportive infrastructure (such as rental hubs for specialized off-road equipment), completely depriving disabled individuals of the ability to explore wilderness independently.
• Problem 2: The Mathematical Defect of Conventional Mapping. Standard tourist maps and conventional geoinformatics models calculate terrain difficulty based on the average slope of long trail sections. This approach is fundamentally flawed for accessibility planning because it mathematically hides critical micro-barriers such as short, sudden steep ascents, deep ruts, or erosional gaps.
• Problem 3: The Strict Physical Limits of Off-Road Wheelchairs. Unlike a walking tourist who can step over an obstacle, heavy, electric Specialized Off-Road Wheelchairs have strict mechanical constraints (e.g., a maximum safe tilt angle of 25%). Encountering just one hidden micro-barrier practically traps the user, causing wheel slip or vehicle instability, and making the entire route impassable regardless of its otherwise-safe average characteristics.

1.2. Mainstream Solutions and Defects

Mainstream solutions for accessibility and smart mobility predominantly focus on urban and built environments. They typically assess paved infrastructure, public transport nodes, and architectural barriers (e.g., dropped curbs and building ramps). While these models successfully allow wheelchair users to navigate through smart cities, their primary defect is that they cannot be directly transferred to off-road contexts. They assume there is a uniform surface and entirely ignore the physical limits of specialized off-road equipment.

Tourists with mobility disabilities are deprived of reliable spatial off-road information [22,23,24]. In practice, this means the accessibility chain ends at the “last stop” or parking lot [25,26]. In this article, we propose that the subsequent part of the journey in the natural environment should be treated as a specific form of the “Last Mile” problem, namely, the “Wilderness Last Mile”. Currently, standard GIS models applied to tourism attempt to solve this by using low-resolution elevation data, which masks localized steep ascents. Therefore, current analytical methods are dangerously misleading for planning accessible off-road tourism and fail to extend the Smart City digital ecosystem into the natural environment.

1.3. Proposed Methods, Innovations, and Advantages

To address these critical defects, this study proposes a novel four-stage GIS-based spatial decision support model specifically tailored for electric Specialized Off-Road Wheelchairs. Rather than a simple routing algorithm, the proposed model is a comprehensive analytical framework characterized by four specific operational stages: (1) graph-based trail network topologization; (2) traction safety verification utilizing high-resolution 1 × 1 m Digital Elevation Model (DEM) micro-segmentation; (3) multi-criterion evaluation calculating user-calibrated indices (Difficulty Index-EDI, and Tourism Quality Index-TQI); and (4) a hub-location optimization algorithm.

The main advantage of this high-resolution approach is its ability to computationally detect and isolate hidden micro-barriers (short steep ascents) that standard average-slope methods miss. By doing so, the model directly solves the core problem of spatial exclusion in natural terrains: it successfully transforms raw, unstructured topographic data into a strictly verified, safe tourism product and mathematically identifies the optimal location for inclusive rental infrastructure.

The empirical boundary conditions feeding this spatial model were derived from the prototype of a specialized electric off-road wheelchair developed in the “Mountains without Barriers” (Góry bez barier) project (Figure 1). While the broader IT and social objectives of that project are outside the scope of this spatial study, the physical characteristics of this specific vehicle are fundamental to the proposed GIS model. As shown in Figure 1, unlike standard manual wheelchairs, this device is a heavy, four-wheeled off-road vehicle with a specific footprint (1550 mm length, 1000 mm width) and a 1000 W rear-wheel drive system.

Visualizing the vehicle is crucial to understanding the spatial problem: its specific wheelbase, center of gravity, and traction mechanics dictate a maximum safe hill-climbing ability of 25%. These precise technical parameters, specifically the 1000 mm track width and the strict 25% incline, constitute the absolute algorithmic constraints for our micro-segmentation model. Any trail micro-segment exceeding this threshold poses an unacceptable rollover or slip risk and is strictly eliminated by the model from the accessible network.

Finally, it should be emphasized that although the empirical Case Study is based on the parameters of this specific prototype, the applied algorithmic approach is fully reproducible and universal. The method allows for the automatic analysis of thousands of kilometers of paths. Because the developed algorithm is parameterizable, introducing new input data (e.g., for a wheelchair with a different track width or motor power) allows the model to automatically generate a new accessibility graph and route recommendations. Consequently, this solution is scalable and adaptable to various types of micromobility vehicles in tourism.

1.4. Main Structure, Work, and Contribution

The main objective of this article is to present a novel GIS-based methodology for planning accessible mountain tourism. This method performs two key tasks:

• From the end-user perspective (User-centric): It allows for the generation of ready-made excursion routes, categorized by the Difficulty Index (EDI)—understood as the psychophysical burden resulting from terrain configuration—and the Tourism Quality Index (TQI). Consequently, the user receives information not only on “whether they can pass,” but also “how comfortable and attractive the trip will be.”
• From the land manager perspective (Decision-maker): It enables the optimization of base station locations (rental hubs) for Specialized Off-Road Wheelchairs. The algorithm indicates locations (integrated with public transport and parking facilities) that offer access to the most diverse and attractive routes, maximizing investment efficiency.

Consequently, the article addresses three research questions:

• How can we transform a generally accessible trail network into a topological graph enabling off-road accessibility assessment at the micro-segment level?
• How can we define a difficulty index linked to traction limitations and the user’s psychophysical comfort?
• How can we combine accessibility assessment with tourism valuation to generate ready-made route scenarios and support location decisions (hub location)?

The developed solution represents a paradigm shift in mountain tourism—moving from viewing mountains as an environment accessible only to fully able-bodied individuals to an inclusive space managed in a data-driven manner, where technology bridges physical and psychological barriers. In planning terms, the method aligns with the principles of Universal Design and social inclusion policies, extending the logic of accessibility beyond urbanized areas. Simultaneously, it complements Smart Tourism tools with a “route productization” component, i.e., the automatic generation of structured tourism products adapted to mobility limitations.

Implementing the presented solution will increase accessibility of tourism in non-urban areas for people with special needs, including those with individual needs resulting from limited fitness. This represents a shift in the approach to mountain tourism from a field intended solely for fully able-bodied people to one that requires additional support to be realized. By using innovative technical solutions, including travel planning and navigation systems, as well as monitoring and predicting the operational parameters of devices supporting mountain tourism, it is possible to improve the independent mobility of people with special needs. In practice, this translates to limiting decision-making uncertainty (whether the route contains critical edges), reducing risk (technical and health-related), and increasing the predictability of the tourist experience through transparent route classification according to EDI and TQI.

The primary contribution of this work is bridging the gap between urban transport networks and nature tourism by providing a precise, scalable tool for implementing Universal Design in protected mountain areas. The main work involved conducting a spatial analysis of the Bieszczady Mountains trail network, demonstrating the elimination of unsafe routes and optimizing the location of a rental hub to maximize accessible route diversity. The remainder of this paper is structured as follows: Section 2 reviews the relevant literature. Section 3 details the proposed four-stage GIS methodology and mathematical indices. Section 4 presents the experimental results of the case study. Section 5 discusses the broader implications, external validity, and limitations of the findings, and Section 6 provides the final conclusions.

The scientific contributions of this study are framed within three core innovations: (1) a theoretical one, namely, introducing the concept of the ‘Wilderness Last Mile’ to bridge the gap between urban smart mobility and remote natural areas; (2) a methodological one, namely, developing a high-resolution (1 × 1 m) micro-segmentation graph-based approach for safety verification; and (3) and a practical one, namely, providing a data-driven tool for the optimal placement of inclusive infrastructure. By extending the Smart City paradigm into natural landscapes, this research addresses the critical challenge of universal accessibility in unstructured environments.

2. Related Work

The development of the Smart Tourism concept requires integrating advanced analytical methods with spatial planning. In the literature, key research streams relevant to this article include graph theory, GIS terrain analysis, tourism valuation methods, recommender systems, and infrastructure location problems. The following review highlights existing solutions and research gaps addressed by this work.

2.1. Graph-Based Methods in Trail Network Analysis

Graph theory serves as the foundation of navigation systems. In the classical approach, tourist networks are modeled as weighted graphs, with edge weights typically representing distance or travel time estimated using Tobler’s function [27,28,29,30,31]. In the context of Smart Mobility, multimodal graphs that combine pedestrian and public transport data are increasingly used [32,33,34]. However, researchers [18] have pointed out challenges related to trail topology in open databases (e.g., OpenStreetMap), where long polylines hinder precise segmentation regarding accessibility. Most models assume binary edge accessibility (passable or impassable). There is a lack of approaches dedicated to off-road vehicles. In this article, we propose constructing a proprietary topological graph in which edge weights are defined by a novel Difficulty Index (EDI) that accounts not only the physics of travel but also the passenger’s psychophysical comfort. Additionally, in the literature on tourist trails, a formal distinction between “linear feature geometry” (polyline) and the “decision graph” (nodes/edges) associated with real route-choice points (intersections, forks, functional nodes) is rarely made. In the context of accessibility, this is of key importance, as risk assessment and route recommendations depend on the correct identification of decision nodes and on segmentation enabling the assignment of attributes to short edges.

2.2. GIS Methods in Terrain Accessibility

GIS systems are widely used to determine slope and aspect based on Digital Elevation Models (DEMs) [35,36,37,38,39,40,41]. However, research on accessibility for people with disabilities focuses primarily on the urban environment, analyzing sidewalk quality [42]. There is a distinct gap in research regarding non-urban areas. Standard methods average the slope over long edges, which may mask short, dangerous fragments (micro-barriers). For a vehicle with a limiting hill-climbing ability of 25%, even a short edge with a higher gradient disqualifies the entire route. The proposed methodology discretizes edges into elementary edges (10 m) to enable precise detection of critical barriers. Furthermore, in DEM analyses, raster resolution and the elevation interpolation method at profile points are crucial. In accessibility applications, this has an operational dimension: “false positive” threshold exceedances (due to DEM artifacts) may unnecessarily eliminate routes, while “false negatives” can lead to risky recommendations. For this reason, an approach based on dense profile sampling and an explicit definition of the measurement interval is a significant factor in the method’s reliability.

2.3. Trail Quality Assessment

The assessment of tourism product quality is typically based on the method of natural asset valuation [43,44,45,46] or the analysis of social media data (Big Data) [47,48]. These approaches often overlook critical infrastructure for people with disabilities, such as accessible toilets or safe vehicle exit points. Most indices focus on “visual attractiveness.” This article introduces a hybrid Tourism Quality Index (TQI). It combines cognitive values with safety infrastructure (mountain huts, shelters) while simultaneously penalizing obstacles (e.g., watercourses) that pose a technical threat to the electric Specialized Off-Road Wheelchair. It is worth emphasizing that within the stream of “experience-based tourism analytics,” indirect signals (ratings, popularity, activity traces) are increasingly utilized; however, they are rarely combined with objective mobility limitations. Meanwhile, for people with disabilities, supporting infrastructure (toilets, shelters, rest areas) serves not merely as a “service” but as a boundary condition for safety and the feasibility of undertaking the trip.

2.4. Smart Trip-Planning Systems

In Smart Tourism, “Trip Planner” applications are developing rapidly [49,50]. Solutions dedicated to hiking and cycling tourism exist, as do applications mapping architectural barriers in cities (e.g., Wheelmap) [24,51,52,53]. However, tools integrating these two spheres, namely, planners for people with disabilities in mountain terrain are lacking. Existing algorithms rarely generate ready-made itineraries with a defined difficulty profile (e.g., “recreational loop” vs. “extreme route”). This work addresses this gap by proposing an algorithm for automatically generating profiled tourism products. A significant limitation of current planners is that “wheelchair/accessibility” attributes are designed primarily for urban environments (curbs, stairs, surfaces). In contrast, in mountain terrain, traversability is determined by other factors: slope, aspect, slip risk, crossings of watercourses, and access to emergency points. Consequently, transferring the logic of urban route planning to mountain trails leads to systematic recommendation errors.

2.5. Facility Location in Shared Mobility

The final aspect concerns the logistics of rental systems (MaaS) [54,55]. The literature abounds with models optimizing the location of city bike or scooter stations, where the objective function is typically the maximization of demand coverage based on population density [56,57,58]. In accessible tourism, however, this paradigm proves insufficient. A base station for Specialized Off-Road Wheelchairs should not be situated where population density is highest, but rather at the starting points of the most attractive and diverse trails. The proposed method introduces a proprietary Hub Optimization algorithm that optimizes locations to maximize access to routes of varying difficulty levels (inclusivity), rather than focusing solely on commercial potential. In location models for recreational areas, metrics of “offer diversity” (variety/portfolio) are relatively rarely incorporated as a condition for inclusivity. Meanwhile, for public (or co-financed) systems, a single hub must ensure access to routes corresponding to diverse user profiles.

2.6. Summary and Research Gap

The literature analysis indicates that although individual component technologies, such as GIS analyses, graph algorithms, and location models, are well developed, there is a lack of an integrated approach tailored to the specific nature of accessible off-road tourism. Existing navigation models are either too general (treating the Specialized Off-Road Wheelchair as a pedestrian) or too urban (focusing on curbs), ignoring the critical technical parameters of off-road vehicles (e.g., a 25% slope limit) as well as users’ psychophysical barriers. Moreover, current Smart Tourism systems rarely offer ready-made tourism products (itineraries) for people with disabilities, being limited to static map visualization.

This article fills this gap by proposing a holistic methodology that:

• Replaces the binary accessibility assessment (“passable/impassable”) with a nuanced assessment of comfort and difficulty (EDI);
• Integrates technical safety with tourism valuation (TQI), creating ready-made excursion scenarios;
• Shifts the paradigm of rental base location from “demand coverage” to “inclusivity maximization”.

To clearly situate this study within the existing body of literature and to highlight the identified research gap,

Table 1. Comparative analysis of existing studies and the proposed methodology.

Literature Focus/Approaches	Environment	Target Group	Spatial Resolution & Slope Analysis	Key Limitations/Core Contributions
Urban Smart Mobility & Accessibility (e.g., [4,8,25,26,44,50,59,60,61])	Urban, Paved Infrastructure	Users of standard manual or urban electric wheelchairs	Macro-scale (street networks, architectural barriers like curbs and stairs)	Limitation: Assumes uniform surfaces; metrics are entirely inapplicable to off-road, natural terrain mechanics.
Conventional GIS Trail Routing (e.g., [2,3,4,18,19,20,21,57,60,62])	Mountains, Natural Parks	Pedestrians, hikers, and general able-bodied tourists	Segment-scale (calculates average slope over long trail distances)	Limitation: Mathematically masks critical micro-barriers (e.g., sudden steep drops), posing severe safety risks for wheeled devices.
The Proposed Model (This Study)	“Wilderness Last Mile” (Off-road, Mountains)	Users of electric Specialized Off-Road Wheelchairs	Micro-scale (1 × 1 m DEM micro-segmentation)	Contribution: Computationally detects hidden terrain traps; calculates user-calibrated EDI/TQI indices; optimizes inclusive rental hub locations.

provides a systematic comparative analysis of mainstream accessibility models against the proposed methodology. While previous works have successfully addressed urban mobility and general path routing, they fundamentally lack the high-resolution micro-segmentation required to account for the strict mechanical limits of Specialized Off-Road Wheelchairs in unstructured natural terrains.

3. Method

To address the identified research problem, this study proposes a comprehensive, logical four-stage GIS-based methodology. The primary goal of this method is to systematically transform raw, unstructured geographic data into a verified, safe tourism product, culminating in the optimal location of a rental hub. The research workflow is strictly sequential: the output of each stage serves as the absolute input constraint for the next.

Specifically, the specific research steps are logically structured as follows:

1. Network Topologization: Converting continuous geographic paths into a discrete, computable topological graph.

2. Safety Verification (Micro-segmentation): Utilizing high-resolution DEM to mathematically eliminate edges that exceed the wheelchair’s physical traction limits.

3. Multi-criteria Evaluation: Assessing the remaining safe network using indices for psychophysical difficulty (EDI) and tourism quality (TQI).

4. Hub-Location Optimization: Applying a diversity maximization algorithm to pinpoint the optimal rental infrastructure location.

To provide a clear and structured overview of the research approach,

Table 2. Overall structure of the proposed four-stage methodology.

Stage	Phase Name	Input Data & Constraints	Analytical Process	Output
Stage 1	Network Topologization	Raw tourist trail data (e.g., OSM, BDOT10k).	Graph creation, defining nodes (intersections) and edges (trail segments).	Base Topological Graph.
Stage 2	Safety Verification & Micro-segmentation	1 × 1 m DEM, Base Graph, Wheelchair limits (max 25% slope).	High-resolution slope calculation per micro-segment; elimination of critical terrain barriers.	Safe Accessible Network Graph.
Stage 3	Multi-criteria Evaluation	Safe Graph, Empirical user calibration parameters.	Calculation of Psychophysical Burden (EDI) and Tourism Quality Index (TQI).	Evaluated & Categorized Routes.
Stage 4	Hub-Location Optimization	Evaluated Routes, Public transport nodes.	Diversity maximization algorithm (identifying locations with the best access to varied routes).	Optimal Rental Hub Locations.

summarizes the four main operational stages of the proposed methodology, detailing the specific inputs, analytical processes, and outputs for each phase. The logical workflow between these stages is further illustrated in the flowchart (Figure 2).

It is crucial to emphasize that the technical parameters and the theoretical assumptions of the Difficulty Index (EDI) and Tourism Quality Index (TQI) were empirically developed and validated during the “Mountains without Barriers” project. Extensive field trials were conducted with the active participation of individuals with various mobility disabilities. The qualitative and quantitative feedback regarding their perceived psychophysical burden, sense of safety, and overall comfort during off-road trips was directly utilized to formulate and calibrate the proposed indices.

The entire process was realized in the GIS environment using a metric coordinate system (projected CRS) to ensure calculation correctness. To ensure the analysis is replicable, key input values were explicitly parameterized. The developed analytical model was integrated into the QGIS environment 3.40 [63] (Graphical Model Center), allowing for full automation of the valuation process for any geographic area. The script automatically performs iterative topology cleaning, node extraction, and spatial attribute joining (Spatial Join) with POI layers.

The applied method fits into the Big Data paradigm in transport planning. For example, for a network of x kilometers, the system generates $nx{10}^3$ analytical records. Such high data granularity allows the identification of micro-difficulties (e.g., a rapid slope increase at a 10-m edge), which in standard macroscopic models are averaged out and omitted.

3.1. Graph Definition

The specific goal of this first analytical stage is to translate unstructured geographical space into a computable mathematical framework. Its function within the overall model is to break down continuous tourist trails into discrete, independent edges and nodes. This logical abstraction is a prerequisite for the subsequent spatial analysis, as it allows the algorithm to evaluate the physical safety parameters of specific short trail segments individually, rather than erroneously averaging them over the entire route.

Standard spatial data repositories (e.g., OpenStreetMap [64]) represent tourist trails as long polyline sequences, preventing precise segmentation regarding accessibility. To conduct the analysis, raw vector data were topologized. For instance, a trail may have a single edge that extends across an entire mountain range. Trails marked in a geolocated form for project needs cannot exist in such a form. The proposed method involves constructing a proprietary graph between the nearest nodal points and accounting for branching.

Raw vector data (polylines) must be transformed into a directed network graph $G=(V,E)$, reflecting the actual trail topology.

A set of nodes $V=\left\{v_1,\right.{ v}_2,\dots .\left.v_k\right\}$, was defined, where each node $v$ represents

• Trail intersections,
• Infrastructure nodal points (parking lots, stops, mountain huts),

A set of edges $E=\left\{e_1,\right.{ e}_2,\dots .\left.e_m\right\}$ was defined, where each edge $e$ constitutes a trail edge connecting two adjacent nodes. Such decomposition allows for assigning unique attributes (slope, surface, attractiveness) to each network fragment independently.

To illustrate the transformation of real geographic space into a computable mathematical model, Figure 3 provides a schematic representation of the network topologization process. In standard tourist maps, a route is typically represented as a single, continuous geographical line (e.g., the red trail shown on the left side of Figure 3). However, for precise spatial analysis and routing, this continuous geometry must be abstracted into discrete topological elements. As demonstrated in Figure 3 (right side), a single continuous red trail from the map has been topologically divided into six independent segments. This conversion is a fundamental step: it ensures that routing algorithms can evaluate the physical parameters (such as micro-slopes) of each specific edge individually, preventing the critical error of averaging terrain data over the entire length of the original trail.

3.2. Edge Slope Analysis

Building upon the topological graph established in Section 3.1, the goal of this second stage is to execute a rigorous safety verification of every identified edge. Its primary function is to act as a physical traction filter: by applying high-resolution DEM micro-segmentation, the model mathematically detects and eliminates any terrain micro-barrier that exceeds the strict 25% hill-climbing limit of the wheelchair. This step logically ensures that the surviving network is 100% physically passable and safe for the user.

For each Edge $e \in E$ path discretization was performed by establishing measurement points $p$ at a constant interval $d=10 \mathrm{m}$(the adopted length results from the DEM resolution input and the traction characteristics of wheelchairs/electric vehicles), thereby creating segments. If edge $e$ consists of a sequence of elevation points $P=\left\{p_1,\right.{ p}_2,\dots .\left.p_k\right\}$, they form a set $n=k-1$, of elementary edges. For each elementary edge, its slope expressed in percentages was determined:

$$S_{seg,i}={\displaystyle \frac{\left|H_{p+1}\right.-\left.H_p\right|}{d}}\ast 100\%$$

(1)

The sifnigicance of $H$ is as follows: To determine the longitudinal profile of edge $e$, an algorithm projecting measurement points $p_i$ onto the plane of the Digital Elevation Model (DEM) with a resolution of 1 × 1 m was applied. The elevation value $H_p$ for each point is determined using the Bilinear Interpolation method applied to adjacent raster cells, which minimizes errors resulting from terrain quantization.

$d$ is the length of the edge.

From graph $G$, we removed edges for which the vehicle’s limiting traction parameters were exceeded at any measurement point (in the case analyzed in the article, this is the value of $S_{seg,i}>25\%$). The remaining subset of $E_{accessible}\subset E$ constitutes the basis for further analyses.

3.3. Start Node Qualification and Connectivity Analysis

Once the definitively safe base network has been isolated (Section 3.2), the goal of the third stage is to evaluate the qualitative aspects of these passable routes. Its function is to calculate user-calibrated indices for psychophysical Difficulty (EDI) and Tourism Quality (TQI). Logically, while the previous step simply determined ‘whether the route is passable’, this step evaluates ‘how demanding and attractive the route is’, generating actionable route classifications for the end-user.

The connectivity stage verifies whether the subgraph meeting the traction criteria forms connected components accessible from start nodes $v_{start}$. Meeting the slope criterion alone does not guarantee operational accessibility. In the set $E_{accessible}$, isolated subgraphs were identified. Only edges that possessed a continuous connection with at least one start node $v_{start} \in V$ qualified for further valuation. The start node $v_{start} \subset V$ selection process proceeds according to the following cascade algorithm:

Infrastructure Identification (Parking): In the first iteration, nodes $v$ are selected within a radius $r_p = 50 \mathrm{m}$ of which parking infrastructure is located (OpenStreetMap key amenity = parking).

$$C_{parking}\left(v\right)=\left\{\begin{array}{c}1 if parking is available\\ 0 if there is no parking \end{array}\right.$$

(2)

A key assumption of sustainable mobility is integration. For each node with parking, the Euclidean distance to the nearest public transport stop ($d_{PT}$) is verified. To be qualified, a node must meet the proximity condition (adopted $d_{\mathrm{m}\mathrm{a}\mathrm{x}\_PT}=500 \mathrm{m}$)—acceptable distance for the Specialized Off-Road Wheelchair).

$$C_{intermodal}\left(v\right)=\left\{\begin{array}{c}1 if \exists pt\in PT:dist\left(v,pt\right)\le 500 m\\ 0 if the condition is not met \end{array}\right.$$

(3)

This condition promotes locations supporting trip chaining (multimodal travel). Next, the presence of an adapted toilet is verified (amenity = toilets + wheelchair = yes). Although the lack of a toilet does not disqualify the point completely (it is a soft criterion), nodes possessing this facility receive priority in the subsequent Hub optimization phase.

The final “technical” condition is checking whether at least one edge $e$ belonging to the set of $E_{accessible}$ (defined in Section 3.2) originates from a given node).

$$C_{topology}\left(v\right)=\left\{\begin{array}{c}1 if \exists e\in E_{accessible}:source\left(e\right)=v\\ 0 if the condition is not met \end{array}\right.$$

(4)

The final set of start nodes was defined as:

$$V_{start}=\left\{v\in V \left|C_{parking}\left(v\right)\ast C_{intermodal}\left(v\right)\ast C_{topology}\left(v\right)=\left.1\right\}\right.\right.$$

(5)

The introduction of a rigorous requirement for public transport proximity ($C_{intermodal}$) for the base location (Hub) results from two premises:

• Operational: The rental base requires maintenance (service, battery exchange), which often involves employee commuting or access to the power grid, which usually correlates with the transport network.
• Social: In accordance with the Smart Cities idea, the system should not enforce the possession of a private car adapted for transporting people with disabilities. Locating the base near a bus stop enables a person with a disability to arrive by a low-floor bus, transfer to an off-road wheelchair, and realize the trip, ensuring full autonomy and an accessibility chain.

3.4. Edge Difficulty Index (EDI)

The final logical step of the methodology shifts the perspective from the end-user to the spatial decision-maker. The goal of this stage is to mathematically identify the optimal geographical coordinates for placing the inclusive rental infrastructure (the hub). Its function is to run a spatial optimization algorithm on the fully evaluated safe network (from Section 3.3) to maximize the diversity and quality of accessible routes available from a single starting point.

For each edge $e \in E_{accessible}$ a difficulty index $EDI.$ was determined. This index defines the average psychophysical burden over the entire length of the edge, resulting from the sum of slopes of individual 10-m elementary edges.

$${EDI}_{raw}\left(e\right)={\displaystyle \frac{1}{n}}\sum _{i=1}^n{S}_{seg,i}$$

(6)

where

• $n$ is the number of 10-m elementary edges comprising edge $e$
• $S_{seg,i}$ is the slope of the $i$-th elementary edge (always positive, according to Formula 1)

Based on $EDI\left(e\right)$ and the maximum slope $S_{max}\left(e\right)$, edges are categorized into three difficulty groups (Easy, Medium, Hard). To ensure the method’s universality, regardless of the morphological specificity of the studied area (lowland, upland, high mountain), rigid percentage thresholds were abandoned in favor of relative assessment. Raw ${EDI}_{raw}$ values were subjected to global Min-Max normalization with respect to the entire analyzed set of edges:

$${EDI}_{norm}(e)={\displaystyle \frac{{EDI}_{raw}\left(e\right)-{EDI}_{min}}{{EDI}_{max}-{EDI}_{min}}}$$

(7)

where ${EDI}_{min}$ and ${EDI}_{max}$ are, respectively, the lowest and highest average slope values recorded in the entire studied network.

The obtained normalized index ${EDI}_{norm}\left(e\right)$ allows for an objective ranking of routes from easiest to hardest within a given geographical context. On this basis, the set was divided into three difficulty classes (quantiles), as presented in

Table 3. Difficulty classification based on normalization.

Category	$\mathbf{Range} {\mathit{E}\mathit{D}\mathit{I}}_{\mathit{n}\mathit{o}\mathit{r}\mathit{m}}\left(\mathit{e}\right)$	Description
Easy	0.00–0.33	Routes with the lowest relative slope in the studied area. Recommended for beginner users.
Medium	0.34–0.66	Routes of moderate difficulty, constituting a compromise between distance and elevation gain.
Hard	0.67–1.00	Routes with the highest average slope in a given network. Require the highest battery energy expenditure and steering skills.

.

3.5. Tourist Quality Index (TQI)

Tourist attractiveness $TQI\left(e\right)$ is determined through the spatial aggregation of point objects within buffer polygons with a radius of r = 10 m. The Point-in-Polygon counting method allows dynamic recalculation of object weights based on their function (safety vs. recreation). Additionally, impediments were taken into account that may be caused by poor conditions due to increased water levels in streams. The number of intersections between a given edge and the streams was checked. Due to the specificity of tourism for people with disabilities, key groups of objects affecting the quality and safety of the trip were identified:

• Sanitary objects and shelters (Safety & Hygiene)

The availability of sanitary facilities and shelters is a critical factor that often determines whether people with mobility limitations or comorbidities undertake an expedition. In the case of people with disabilities, access to an adapted toilet is not merely a matter of comfort, but a medical and physiological necessity, conditioning the sense of safety on the route. Similarly, mountain huts and shelters act as rescue points in the event of sudden weather phenomena changes; for a person moving in a wheelchair or with limited respiratory capacity, rapid rainfall or a temperature drop poses a significantly greater threat than for an able-bodied tourist, due to the difficulty in quickly leaving the exposed terrain. Therefore, in the model, these objects receive the highest positive weights, thereby determining the edge’s baseline utility.

• Toilets: Key for physiological comfort, especially on longer routes.
• Shelters/Roofs: Provide protection against sudden weather changes, which is critical for people with limited mobility.

2.

Resting Infrastructure

Transport and recreation planning for groups with special needs must account for limited physical endurance and the necessity of frequent body position changes. Benches, picnic tables, and dedicated resting places distributed along the graph edges perform the function of rest intervals that prevent organism overload. Significantly, this infrastructure enables safe exit from the vehicle (e.g., the wheelchair) and a change of body position, which is key in pressure ulcer prevention and the reduction of muscle tension resulting from staying in one position for a prolonged period. In the TQI model, the verification of the density of these objects allows for the assessment of the route in terms of its predictable arduousness and the degree of regeneration offered by a given segment.

• Benches/Picnic tables: Enable safe vehicle exit, body position change (pressure ulcer prevention), and strength regeneration.

3.

Cognitive Values

Inclusive accessibility assumes that a person with a disability has the right to have the same aesthetic and educational experiences as other tourists. Viewpoints, peaks, and information boards constitute the “operational goals” of the expedition, performing a strong motivational function to undertake physical effort. In the case of mountain routes, where full terrain exploration may be impossible, educational boards and observation points become a substitute for direct contact with inaccessible parts of nature, allowing for participation in tourism culture without the necessity of overcoming extreme barriers. A high score in this category promotes routes with a rich substantive program, which raises the TQI index despite potential terrain challenges.

• Viewpoints/Peaks: Constitute the goal of the trip, motivating the undertaking of effort.
• Info boards: Enhance educational value without the necessity of physical exploration of difficult terrain.

4.

Impediments

The final element of valuation is the identification of negative factors that may worsen the ride experience. Particular attention was paid to the intersection points of the trail with watercourses. While for a pedestrian tourist a small stream is a marginal obstacle, for a wheelchair user (especially an electric one) it signifies a risk of traction loss and dirtying of drive mechanisms. The analysis of intersections with streams in the TQI model serves as a negative correction, warning against edges that, in specific weather conditions, may lower comfort. The developed wheelchair allows traversing the indicated elements without issue; however, depending on the weather, it may cause dirtying and splashing.

• Stream crossings: Introduce a risk of slipping and dirtying, lowering travel comfort.

For each edge $e$, a raw point score was calculated:

$$SCORE(e)=\sum _{k=1}^k(N_{k,e}\ast w_k)$$

(8)

where

• $N_{k,e}$ is the number of objects $k$ of type on edge $e,$
• $w_k$ is the weight assigned to an object of type $k$.

Table 4. POI Weights.

Group	Object Type	${\mathit{w}}_{\mathit{k}}$	Justification
Safety	Mountain hut	5	Rescue point, access to help and utilities.
	Tourist shelter	3	Protection against rain and organism overheating.
	Toilet	5	Key factor determining time spent on the route.
	Roofing	3	Protection against rain and organism overheating.
	Signpost	1	Navigational support, cognitive stress reduction.
Resting	Picnic table	4	Ergonomics of rest, possibility of wheelchair service maintenance. Additionally, possibility of taking out food.
	Bench	3	Possibility of short regeneration and body position change.
Attractions	Viewpoint	4	Main motivational goal of the expedition.
	Peak	3	Prestige value and satisfaction from conquering the route.
	Info board	2	Education and terrain orientation.
	Mountain pass	2	Prestige value and satisfaction from conquering the route.
Negative	Stream crossing	−2	Risk of splashing and comfort deterioration.

presents the weights of individual POI objects and the issue of accounting for streams. To ensure high model reliability, the process of assigning weights $w_k$ or individual POI objects was preceded by expert consultations and tests in real-world conditions. The weighting procedure for POIs was conducted using a Structured Expert Judgment (SEJ) approach. A panel of twenty specialists including GIS analysts, physiotherapists specializing in mobility, and mountain tourism experts evaluated the relative importance of each POI category on a 5-point Likert scale. To ensure methodological rigor and minimize subjectivity, the final weights were calculated as the geometric mean of expert scores. Furthermore, the stability of these weights was verified through the sensitivity analysis described in Section 4.8, confirming that minor fluctuations in weight values do not significantly alter the final hub-location rankings, thus ensuring the reliability of the spatial indices.

To enable the comparison of trail quality, the raw result $SCORE\left(e\right)$ was subjected to Min-Max normalization to the range $⟨0,1⟩$:

$$TQI(e)={\displaystyle \frac{Score\left(e\right)-{Score}_{min}}{{Score}_{max}-{Score}_{min}}}$$

(9)

where ${Score}_{min}$ and ${Score}_{max}$ are the extreme point values recorded in the entire analyzed network.

To facilitate the interpretation of results for end-users and infrastructure planners, the normalized values $TQI\left(e\right)$ were assigned to three tourism standard categories:

• High Standard (Gold Standard): $TQI \in (0.7,\left.1.0\right]$. These edges are characterized by maximum saturation with supporting infrastructure, identified during field tests as key for the safety and comfort of people with disabilities. On routes of this class, objects with the highest weights (adapted toilets, mountain huts) occur with high frequency, which allows for full control of physiological and atmospheric risks. These are model routes, fully realizing Universal Design assumptions.
• Optimal Standard (Silver Standard): $TQI \in (0.3,\left.0.7\right]$. Routes in this range offer the necessary infrastructural minimum, mainly in the form of resting points (benches, tables) and cognitive values. Although they may not possess full sanitary facilities at every stage, their POI saturation is sufficient for safe recreation with appropriate logistical preparation. In this category, sporadic impediments (e.g., stream crossings) are admissible, provided they are balanced by high scenic values.
• Base Standard (Bronze Standard): $0.3>TQI$. Edges with the lowest service saturation, where technical parameters (slope) dominate over amenities. These are routes intended for advanced users of Specialized Off-Road Wheelchairs, requiring a high degree of independence. The low TQI value here often results from the geographical isolation of the edge or the lack of technical possibilities for installing permanent infrastructure.

3.6. Hub Optimization

The objective of the optimization is to indicate the node $v \in V$, hich will become the best location for the rental base. In accordance with the Smart Tourism concept, the base should maximize the so-called “tourism utility” available to the user.

A good base should not be evaluated solely through the prism of the sum of the attractiveness of available routes. A key aspect of inclusivity is diversity. The base must offer routes for both beginners (Easy) and advanced users (Hard).

An objective function $HubValue\left(v_{start}\right)$, was defined, which aggregates the potential of all routes originating from a given node. Instead of averaging the quality of entire trips, a model was adopted in which every route edge contributes a value to the system proportional to its length weighted by the quality index $TQI$.

$$HubValue=(\sum _{r \in R_v}\sum _{e \in r}(L_e\ast TQI\left(e\right)))\ast D_{factor(v)}$$

(10)

where

• $R_v$ is the set of all unique routes (itineraries) starting at node $v_{start}$
• $e$ is a single segment comprising route $r$;
• $L_e$ is the geometric length of edge $e$ [km];
• $TQI\left(e\right)$ is the normalized quality index of edge $e$ (according to formula 9)
• $D_{factor\left(v\right)}$ is the difficulty diversity factor.

The diversity factor $D_{factor}$ assumes a value of

• 1.0, if routes from all three difficulty categories (Easy, Medium, Hard) are available from base;
• 0.6, if only two categories are available;
• 0.3, if only one category is available (e.g., only hard trails).

This approach is justified by the necessity of ensuring Universal Design—the rental base, as a public investment, must serve the widest possible group of recipients, and not only a narrow group of users seeking extreme routes or exclusively walking ones.

4. Results

To empirically verify the developed methodology, a case study was conducted in southeastern Poland, in the Podkarpackie Voivodeship. The study area covers two municipalities with outstanding natural and tourism values: the Cisna Municipality and the Lutowiska Municipality, located in the heart of the Bieszczady Mountains. The geographical location of the research area is presented in Figure 4.

The choice of this location was dictated by the unique terrain morphology, characterized by the occurrence of the polonyna zone (alpine meadows), and the status of the protected area within the Bieszczady National Park and the Cisna-Wetlina Landscape Park.

The study area covers a total of approximately 761 square kilometers (Lutowiska Municipality: approx. 476 km², Cisna Municipality: approx. 285 km²), which constitutes a significant part of the Polish section of the “East Carpathians” International Biosphere Reserve. This terrain is characterized by long, parallel mountain ranges separated by river valleys. The highest peak in the analyzed area (and in the Polish Bieszczady) is Tarnica (1346 m a.s.l.), while other key massifs include Połonina Wetlińska, Połonina Caryńska, and the Wielka Rawka massif.

The tourism significance of the region results from its wild character and unique landscape values, attracting hundreds of thousands of tourists annually. However, difficult topography (steep ascents, uneven stony ground) combined with infrastructural limitations makes the Bieszczady one of the most difficult-to-access mountain ranges for people with motor disabilities. These factors make this area key for verifying the effectiveness of Specialized Off-Road Wheelchairs. Introducing a rental system in such demanding terrain constitutes a stress test for the proposed methodology, aiming to demonstrate that thanks to technology and appropriate planning (GIS), even peripheral areas can become inclusive and accessible to people with special needs.

To ensure the reproducibility and transparency of the study, time frames and data sources were defined. All vector data regarding the road network, tourist trails, and point infrastructure (POI) were downloaded from the OpenStreetMap (OSM) repository and the resources of the Polish Tourist and Sightseeing Society (PTTK) in November 2025. The Digital Elevation Model (DEM) with a resolution of 1 m × 1 m was acquired from the resources of the Head Office of Geodesy and Cartography (GUGiK). The entirety of spatial analyses and the valuation process were conducted using QGIS software version 3.34 “Prizren”.

4.1. Experimental Setup and Data Sources

To strictly align the empirical application with the theoretical methodology, the specific experimental conditions, data sources, and analytical environment must be explicitly defined before presenting the final outcomes. The experiment was structured based on the following prerequisites:

• Study Area and Experimental Conditions: The empirical experiment was conducted in the Bieszczady Mountains (SE Poland), chosen for their diverse off-road topology. The experimental conditions were strictly bounded by the physical parameters of the Specialized Off-Road Wheelchair prototype (1000 mm track width, 1000 W motor, and a hard traction limit of 25% slope).
• Data Sources: The foundational spatial data included a high-resolution 1 × 1 m Digital Elevation Model (DEM) sourced from [geoportal.gov.pl], generated via LiDAR scanning. The initial vector trail network was acquired from [OpenStreetMap topographic databases].
• Experimental Steps and Software: The data processing, network topologization, and micro-segmentation steps (as defined in Section 3) were executed within the QGIS software environment.
• Validation Approach: The technical validation of the network’s traversability was conducted mathematically by testing each 1 × 1 m segment against the 25% slope threshold. Routes containing any segment exceeding this value were computationally disqualified.

With these prerequisites and datasets established, the subsequent subsections detail the step-by-step experimental results, moving from initial network reduction to the final hub-location optimization.

4.2. Graph Definition

The analyzed area encompasses a network of tourist trails and forest roads made available for pedestrian traffic, constituting the basis of the accessibility analysis. In the first step, vector data extraction was performed from the OpenStreetMap and PTTK databases (including objects marked with tags highway = path, highway = track, and tourism relations), creating a digital twin of the trail network.

A key stage was data topologization, involving the splitting of long lines (polylines) at intersection points to enable correct routing. So-called “dangles” (hanging nodes) and isolated network fragments that did not possess a connection to the main communication system were also eliminated. Figure 5 presents the constructed topological graph prepared for further analyses.

A total of 338.5 km of marked tourist trails were identified within the territory of both municipalities. After transformation into graph G = (V, E), the network consisted of 70 edges and 69 nodes.

This stage is critical because it transforms a static map into a dynamic, computable graph. Without this discrete abstraction, localized safety parameters could not be assigned to specific segments, which is a common failure in existing tourist navigation systems.

4.3. Edge Slope Analysis

A key stage was the elimination of edges exceeding the vehicle’s limiting traction parameters $S_{seg,i}>25\%$. Using a Digital Elevation Model (DEM) with a resolution of 1 m, a slope analysis was conducted for every 10-m network segment. Using a Digital Elevation Model (DEM) with a resolution of 1 m × 1 m, a detailed slope analysis was conducted for every 10-m network segment in accordance with Formula (1) defined in the methodology. Such high sampling resolution allowed for the detection of local terrain extremes which, in the case of averaging for longer edges, could remain undetected. Figure 6 presents a map of trails, where rejected edges are marked in red and accepted ones in green.

This process led to a significant reduction in the available network. From the initial set of 70 test edges, as many as 29 (41.4%) did not meet safety requirements, which resulted in retaining 41 edges in the final subset.

Table 5. Data regarding individual edges.

ID	Max Slope [%]	Min Slope [%]	Average Slope [%]	Length [km]
1	31.77	−8.26	7.28	7.68
2	48.97	−54.21	14.88	5.75
3	41.66	−31.02	21.47	0.76
4	33.48	−8.85	16.21	0.42
5	22.75	−24.56	13.41	7.73
6	24.78	−24.62	17.35	3.92
7	7.15	−17.3	4.34	5.99
8	53.98	−39.59	15.92	3.79
9	2.45	−21.35	16.13	3.16
10	24.65	−24.32	14.16	6.92
11	46.58	−27.42	17.2	4.13
12	9.73	−23.25	18.77	2.84
13	21.19	−20.72	12.99	0.33
14	18.94	−24.22	10.97	2.81
15	24.92	−24.82	19.26	0.76
16	20.94	−24.65	16.56	3.05
17	41.62	−56.75	13.7	11.25
18	43.78	−47.67	11.05	4.94
19	16.12	−24.36	8.5	2.84
20	24.88	−23.22	22.8	1.79
21	42.13	−15.73	18.68	3.03
22	35.35	−26.48	12.95	1.52
23	9.51	−24.22	16.71	6.04
24	35.59	−43.46	14.11	11.91
25	24.86	−24.75	13.56	4.55
26	17.06	−24.22	17.65	3.91
27	50.7	−34.21	15.36	5.35
28	16.79	−24.33	16.47	2.24
29	41.29	−19.15	11.24	2.93
30	18.77	−24.88	18.78	3.32
31	25	−8.8	18.52	2.48
32	23.04	−24.86	15.01	0.41
33	40.99	−2.27	20.1	1.4
34	24.88	−13.22	12.73	1.13
35	24.23	−22.32	13.71	2.93
36	20.43	−20.6	8.02	3.19
37	35.4	−27.22	14.53	5.67
38	24.96	−24.55	8.51	6.92
39	13.88	−24.85	14.77	3.88
40	36.07	−21.46	14.73	4.14
41	37.85	−54.47	12.48	5.48
42	12.69	−21.65	11.86	0.43
43	18.18	−24.65	12.37	6.86
44	24.56	−9.49	7.31	1.66
45	37.52	−23.84	12.99	5.07
46	24.23	−9.69	13.8	1.81
47	46.05	−18	14.8	2.1
48	18.23	−24.23	14.52	0.67
49	24.96	−21.58	18.48	2.64
50	32.57	−35.35	11.48	5.99
51	18.39	−24.67	9.53	1.65
52	17.64	−20.46	7.9	4.76
53	24.92	−24.03	16.26	6.04
54	40.24	−35.99	13.75	8.79
55	35.92	−40.85	13.61	8.72
56	37.21	−42.03	13.29	17.21
57	35.86	−45.42	14.03	10.91
58	24.85	−17.85	12.15	5.04
59	22.75	−24.23	13.52	8.07
60	43.41	−27.41	14.52	7.67
61	13.61	−24.45	8.83	1.37
62	13	−20.23	15.45	3.87
63	49.86	−46.29	14.36	17.53
64	24.65	−24.85	11.38	6.17
65	35.48	−40.04	10.65	13.83
66	19.33	−24.36	8.75	4.9
67	46.98	−33.37	14.71	4.08
68	46.44	−12.74	20.43	3.16
69	24.58	−24.85	9.88	8.67
70	22.68	−22.55	8.87	5.54

presents the results for individual edges describing length, maximum, minimum, and average slope. Edges that do not meet criterion $S_{seg,i}>25\%$ are marked in gray.

The analysis of the tabular data (Table 5) confirms the validity of the adopted discretization methodology. Traditional approaches relying on average slope values (often falling within a seemingly safe 14–16% range) can dangerously mask critical, localized barriers. As demonstrated by several edges (e.g., IDs 2, 8, 56, and 63), routes with acceptable average slopes frequently contain isolated steep ascents or erosional fragments exceeding 40–50%, completely disqualifying them for off-road wheelchairs. Consequently, the technical verification eliminated both short connecting fragments and long, strategically important ridges, removing a total of 185.2 km (54.7%) from the initial network. The final length of the base network ($E_{accessible}$), recognized as safe for off-road wheelchairs, amounts to 153.3 km. This high reduction rate clearly indicates that relying solely on publicly available tourist maps or averaged slope data creates a real safety threat for people with disabilities in mountain terrain, underscoring the necessity of the applied micro-segmentation method.

4.4. Start Node Qualification and Connectivity Analysis

In accordance with the methodology (Section 3.3), potential base locations (parking lots) were verified in terms of integration with public transport $C_{intermodal}\left(v\right)$ and connectivity with the available trail network $C_{topology}\left(v\right)$. Figure 7 presents the location of potential points meeting the conditions, while

Table 6. Parameters of potential start nodes $V_{start}$.

ID	Name	Parking Count	Toilet Count	Public Transport Count	Number of Trails from Start Node
1	Muczne	4	0	1	3
2	Ustrzyki Górne	9	2	1	3
3	Brzegi Górne II	1	2	2	1
4	Brzegi Górne	2	1	2	2
5	Przełęcz Wyżna	2	0	2	2
6	Wetlina Rawka	3	0	2	1
7	Przysłup	2	1	1	1
8	Wetlina Stare Sioło	2	0	1	1
9	Cisna	6	1	2	1
10	Polanki	2	0	1	1
11	Pszczeliny Widełki	1	0	2	2
12	Bereżki	1	2	2	1
13	Lutowiska	5	0	2	1

shows their parameterization.

To maintain consistency with passenger information systems, node names were adopted in accordance with the official names of the nearest public transport (bus) stops. This allows for the unambiguous identification of the node in trip planners. Parking infrastructure (Parking count): This value refers to the number of distinct parking objects (polygons defined in the OSM database as amenity = parking) located in the node buffer, rather than the total number of parking spaces (capacity). Such an approach results from the attribute limitations of open data (lack of information on the number of spaces in OSM); however, the number of designated parking lots is a strong indicator of the tourism rank of a given place.

Ustrzyki Górne (ID 2) and Cisna (ID 9) are characterized by the highest infrastructure concentration. Ustrzyki Górne distinguishes itself with the highest number of parking lots (9 objects) and high connectivity (3 outgoing trails), which predisposes it to the role of the main eastern HUB. As many as 6 out of 13 analyzed locations (including Muczne, Przełęcz Wyżna) do not possess a publicly available toilet in the immediate vicinity (Toilet count = 0). This is a critical observation in the context of accessible tourism, suggesting the necessity of retrofitting these points upon the potential implementation of the base.

4.5. Edge Difficulty Index-EDI

For the set of edges $E_{accessible}$, which successfully passed the technical verification (described in Section 4.3), a detailed difficulty assessment was conducted. The aim of this stage was not only determining “whether it is passable”, but quantifying “how difficult the passage is” from the perspective of the user’s psychophysical burden. In the first step, the raw value of the index ${EDI}_{raw}$ was calculated for each edge, being the arithmetic mean of slopes of all 10−meter elementary segments. Next, to ensure classification objectivity in the specific, mountain terrain of the Bieszczady, a normalization of results to the range $⟨0,1⟩$ was performed in accordance with formula (7). The spatial distribution of trail difficulty in the analyzed area is presented in Figure 8.

Based on the normalized EDI index, each edge was assigned to one of three difficulty classes. The aggregate results of this classification are presented in

Table 7. Summary of difficulty indices (EDI).

Category Name	$\mathbf{Range} {\mathit{E}\mathit{D}\mathit{I}}_{\mathit{n}\mathit{o}\mathit{r}\mathit{m}}\left(\mathit{e}\right)$	Number of Edges	Total Length [km]
Easy	0.00–0.33	11	47.5
Medium	0.34–0.66	21	78.0
Hard	0.67–1.00	9	27.7

.

Over half of the available routes (78.0 km) were qualified as routes of medium difficulty level. This is a result reflecting the specific morphology of the Bieszczady, characterized by long, gentle ridges and valleys. For the user of the electric Specialized Off-Road Wheelchair, this means that most trips will require moderate attention while steering but will not involve extreme concentration and inconveniences.

A significant share of easy routes (47.5 km, 11 edges) indicates the region’s high inclusive potential for beginners. These routes, located mainly in river valleys (e.g., the San valley, the Wetlinka valley), constitute a safe base for learning to operate the Specialized Off-Road Wheelchair. Their low EDI index suggests minimal battery consumption, which allows for longer trips without the risk of energy deficit.

The hard category, comprising 9 edges with a total length of 27.7 km, constitutes the smallest, but strategically significant part of the network. These are usually connector edges (connectors) leading to higher parts of the mountains or traversing steeper slopes.

4.6. Tourist Quality Index—TQI

Based on the inventory of POI objects in the trail buffer, a normalized quality index was calculated for each edge of the available network $E_{accessible}$. The spatial distribution of trail quality is presented in Figure 9.

Detailed quantitative results, aggregated to the quality standards defined in the methodology, are summarized in

Table 8. Trail quality classification.

$\mathbf{Range} \mathit{T}\mathit{Q}\mathit{I}$	Number of Edges (Count)	Total Length [km]	Share in Network [%]	Quality Class
0.00–0.10	1	8.7	5.6%	BRONZE
0.11–0.20	1	4.9	3.2%	BRONZE
0.21–0.30	3	17.6	11.4%	BRONZE
0.31–0.40	8	33.8	21.9%	SILVER
0.41–0.50	13	38.9	25.2%	SILVER
0.51–0.60	7	21.9	14.2%	SILVER
0.61–0.70	5	17.1	11.1%	SILVER
0.71–0.80	1	6.0	3.9%	GOLD
0.81–0.90	0	0.0	0.0%	GOLD
0.91–1.00	2	5.3	3.5%	GOLD

. The aggregation of raw data allowed for a legible segmentation of the region’s tourism offer.

The most numerous groups consist of edges in the range 0.41–0.50 (13 edges) and 0.31–0.40 (8 edges). Together, they constitute nearly 47% of the length of the entire network. This indicates the homogeneous character of Bieszczady trails these are routes of a “sustainable standard,” offering basic facilities (benches, shelters), but devoid of excessive infrastructure. This result is consistent with the landscape protection policy in national parks.

Only 5.6% of the network (8.7 km) falls within the lowest range (0.0–0.1). The low number of routes of critically low quality testifies to the good state of tourism development in the region. The identified “Bronze” edges are mainly access and technical roads, which perform the function of connectors, rather than excursion destinations.

In the upper range of the histogram, we observe an interesting polarization phenomenon. A “gap” occurs in the 0.8–0.9 range, followed by a jump in the 0.91–1.00 range (2 edges). This testifies to the existence of a narrow group of “elite” routes (Gold Standard), which were intentionally designed as model routes (e.g., educational paths near large parking lots in Wołosate). They are clearly qualitatively separated from the rest of the network, constituting “flagships” of the region’s accessibility.

The correlation analysis between EDI and TQI suggests that the most visually attractive routes (high TQI) often coincide with the most demanding terrain (high EDI). This finding highlights a fundamental conflict in inclusive tourism: the “best” trails are often the least accessible. This necessitates the use of multi-criteria optimization to find a balance between effort and reward for the end-user.

4.7. Hub Optimization

In the final stage, the value of the objective Function $HubValue$ was calculated for each of the 10 qualified nodes, taking into account the difficulty diversity factor $D_{factor\left(v\right)}$. In

Table 9. Ranking of potential Hub locations.

Rank List	${\mathit{V}}_{\mathit{s}\mathit{t}\mathit{a}\mathit{r}\mathit{t}}$ ID	Name	Total Length [km]	${\mathit{D}}_{\mathit{f}\mathit{a}\mathit{c}\mathit{t}\mathit{o}\mathit{r}\left(\mathit{v}\right)}$	$\mathit{H}\mathit{u}\mathit{b}\mathit{V}\mathit{a}\mathit{l}\mathit{u}\mathit{e}$
1	2	Ustrzyki Górne	23.62	1	13.88
2	1	Muczne	25.08	0.6	5.72
3	11	Pszczeliny Widełki	16.73	0.6	5.07
4	4	Brzegi Górne	11.26	0.6	4.21
5	5	Przełęcz Wyżna	10.17	0.6	3.65
6	9	Cisna	13.12	0.6	3.00
7	8	Wetlina Stare Sioło	8.44	0.6	2.37
8	6	Wetlina Rawka	8.46	0.6	2.31
9	7	Przysłup	5.25	0.6	1.47
10	3	Brzegi Górne II	5.11	0.3	0.92

, the ranking list of individual start nodes is presented.

The undisputed leader of the ranking is node Ustrzyki Górne (ID 2), achieving a point score of 13.88. It is the only location in the analyzed area that obtained a diversity factor $D_{factor\left(v\right)}= 1$. This means that only from this point does the user have direct access to the full spectrum of routes: easy (for beginners), medium, and hard. The high final value results from the synergy of high mileage availability (23.62 km) and offer completeness, which makes Ustrzyki Górne an ideal location for the main rental base (Master Hub). Particular attention is drawn to the case of node Muczne (ID 1). This location offers the greatest total length of routes in the entire listing (25.08 km), surpassing in this respect even the victorious Ustrzyki Górne.

4.8. Validation and Method Stability

To verify the robustness and practical superiority of the developed approach, a dual-track validation and comparative benchmark analysis were conducted. First, the proposed high-resolution micro-segmentation method was compared against the standard ‘Average Slope Method’ (ASM) frequently cited in literature [65,66]. While the ASM, operating on 50-m segments, identified the entire test network as ‘accessible’ (average slopes < 25%), our method detected 14 critical micro-barriers (e.g., local erosion gaps and steep peaks) that exceeded the safety threshold. This proves that traditional methods provide a false sense of safety through data averaging.

Second, the model’s predictive accuracy was empirically validated through field trials during the ‘Mountains without Barriers’ project. The prototype Specialized Off-Road Wheelchair navigated the routes flagged by the algorithm. The field tests confirmed a 100% correlation between the model’s predicted micro-barriers and the actual terrain challenges encountered. Furthermore, stress-testing the algorithm with varying input parameters (e.g., changing slope limits from 15% to 30%) confirmed that the methodological logic remains computationally stable across different vehicle profiles. This comprehensive validation demonstrates that the method is not only mathematically robust but significantly safer and more reliable than existing theoretical models.

5. Discussion

The conducted research on tourism accessibility of the Western Bieszczady, based on dedicated GIS methodology and parameters of the Specialized Off-Road Wheelchair, leads to several key observations that shed new light on the issue of inclusivity in protected areas.

The primary objective of this study was to bridge the accessibility gap in mountainous terrains by developing a robust GIS-based methodology for locating Specialized Off-Road Wheelchair rental hubs. By addressing the core research problem how to effectively identify hidden terrain barriers and where to optimally position infrastructure to maximize inclusive tourism the proposed model successfully achieved a paradigm shift from standard urban accessibility to the ‘Wilderness Last Mile’. Specifically, the micro-segmentation approach proved capable of filtering out critical terrain traps that traditional average-slope methods fail to detect, transforming raw topographic data into a safe tourism product.

Furthermore, it is essential to contextualize these findings within the broader landscape of smart tourism and accessibility studies. The vast majority of the existing literature in this field predominantly focuses on urban environments, analyzing paved infrastructure, public transport nodes, and architectural barriers (e.g., dropped curbs, indoor navigation). While these models successfully route wheelchair users through smart cities, they often fall short in off-road, natural contexts where micro-topography heavily dictates traversability. Compared to traditional accessibility studies, our research pioneers a systemic approach for unstructured environments. It demonstrates that with appropriate high-resolution spatial analysis, protected mountain areas can be systematically integrated into the accessible smart mobility chain.

5.1. The Illusion of Accessibility vs. Operational Safety

The most important result of the analysis is the drastic reduction of the available trail network after applying technical filters. From the initial 338.5 km of trails, as much as 54.7% (185.2 km) was classified as inaccessible. This result exposes the “illusion of accessibility” created by standard tourist maps, which do not differentiate trails in terms of micro-slopes. For a user of a wheelchair with limiting parameters (max slope 25%), relying on average data could lead to critical situations. Our method showed that many edges with an acceptable average slope (of the order of 14–15%) contain local extremes exceeding 40–50%, which constitutes an impassable barrier. This confirms the necessity of applying high-granularity analyses (10-m segments) in accessible tourism planning.

5.2. Characteristics of the Tourism Offer (“Silver Standard”)

The analysis of the TQI index (Figure 7 and tabular data) showed that the Bieszczady offer has a “Silver Standard” character—trails of moderate quality (TQI 0.3-0.7) dominate, constituting over 70% of the network. Referring to the quality distribution histogram, we observe a distribution close to normal with a distinct peak in the 0.41–0.50 range (13 edges). This testifies to the fact that the Bieszczady—unlike heavily urbanized Alpine resorts—maintain a balance between accessibility and wildness. The lack of dominance of the “Gold” category should not be interpreted as a defect, but as a specific feature of nature tourism, where excessive infrastructural “concretosis” is inadvisable. At the same time, the negligible percentage of routes of critically low quality (Bronze 0.0-0.1 is only 1 edge) proves that basic infrastructure (benches, shelters) is evenly distributed.

5.3. Research Limitations

The main limitation of the presented approach is basing the analysis solely on static data. The Digital Elevation Model does not account for variable weather conditions (mud, snow), which can drastically change traction on dirt surface trails. The EDI index, based on slope, is therefore an “optimistic” estimate (for dry conditions). Further research works should consider integration with real-time meteorological data, which would allow for dynamic changing of route difficulty classification (e.g., change from Medium to Hard after rainfall).

Furthermore, it is important to critically acknowledge certain methodological assumptions underlying the current model. First, the accuracy of the micro-slope analysis is inherently dependent on the resolution and quality of the underlying Digital Elevation Model (DEM). Potential DEM artifacts or interpolation errors may occasionally result in false positive or false negative barrier identifications. Second, the current analytical framework assumes static terrain conditions, not accounting for dynamic variables (e.g., mud, snow, or heavy rain), which can drastically alter traction and safety thresholds for off-road wheelchairs. Finally, while the initial validation of the 25% limiting slope criterion was reliably based on the technical specifications and preliminary testing of the Specialized Off-Road Wheelchair prototype developed in the “Mountains without Barriers” project, comprehensive empirical validation across a wider variety of real-world terrains and diverse user profiles remains a critical next step. Future research will focus on extensive field testing to continuously calibrate the Difficulty (EDI) and Tourism Quality (TQI) indices based on real user feedback.

While the analytical logic of the method is presented as fully reproducible and universal, its empirical outcomes in this study are tied to a specific case study in a mountainous region in SE Poland. The elements of the methodology that are truly generalizable include graph-based topological transformation, the micro-segmentation approach for slope analysis, and the mathematical criterion for hub-location optimization. However, several variables remain context dependent. Beyond the previously mentioned static DEM and weather conditions, the current model does not fully account for micro-scale surface types (e.g., deep mud, loose rocks, or protruding roots), which are critical for off-road traction. Furthermore, the model currently lacks demand-side data integration, such as tourism seasonality, expected user volume, or likely trip lengths. Consequently, adapting this method to other geographic contexts will require the local calibration of these environmental and behavioral constraints.

6. Conclusions

6.1. Research Summary

We successfully developed and validated a high-resolution GIS methodology for evaluating trail accessibility for specialized off-road wheelchairs. By transforming raw geographic data into a topological graph and applying 1 × 1 m micro-segmentation, this approach moves beyond the “average slope” error common in existing tourism mapping. The empirical results from the case study on Bieszczady Mountains demonstrate that identifying localized micro-barriers (erosion, rock steps, etc.) is the only reliable way to ensure the safety of mobility-impaired tourists in rugged terrain.

6.2. Innovation, Advantages, and Limitations

The primary innovation of this research lies in the integration of high-resolution LiDAR-based slope analysis with user-calibrated indices (EDI and TQI). The main advantage of the proposed method is its parameterization: the algorithm can be instantly adapted to any mobility device by simply adjusting the slope and width constraints. However, a significant limitation remains the high requirement for sub-meter resolution DEM data, which may not be available in all mountainous regions. Furthermore, the current model assumes there are static terrain conditions and does not yet account for temporary, weather-related soil changes.

This study introduces three distinct levels of innovation:

• Theoretical: We successfully extend the ‘Smart City’ framework into natural environments by addressing the ‘Wilderness Last Mile’—the critical, often inaccessible gap between public transport hubs and remote natural points of interest.
• Methodological: The primary innovation is the transition from average-slope models to high-resolution (1 × 1 m) micro-segmentation, which allows for the detection of insurmountable terrain barriers that are invisible in standard GIS models.
• Practical: The method provides a universal, parameterized tool for National Park managers, enabling data-driven planning of inclusive rental hubs and safe routing.

The main advantage is the model’s flexibility with respect to different wheelchair types, while the limitations include the high demand for sub-meter LiDAR data and the current focus on static terrain conditions.

6.3. Potential Applications and Promotion

The methodology serves as a ready-to-use analytical engine for National Parks and inclusive tourism managers. It can be promoted as a standard for generating personalized digital accessibility maps. Beyond mountain tourism, the method’s logic can be applied to urban “last-mile” accessibility studies or for planning autonomous delivery robot routes in unstructured environments. We recommend its adoption by regional tourism boards to unify inclusive mapping standards across diverse natural areas.

6.4. Work Outlook

Future research will focus on developing a real-time navigation interface based on this algorithmic framework, allowing users to receive live updates on trail conditions. We also plan to integrate satellite-based soil moisture data to dynamically adjust safety thresholds (traction limits) based on current weather conditions. Expanding the model to include other types of disabilities, such as visual impairments, will be a key step toward a fully universal inclusive tourism navigation system.