2.1. Ice and snow tourism

Ice and snow tourism refers to travel that centers around destinations with winter-related activities, such as skiing, snowboarding, snowshoeing, ice climbing, and various winter leisure pursuits. This type of tourism is especially popular in colder climates, where natural elements such as snowfall and low temperatures create unique opportunities for outdoor recreation. It is distinct from other forms of tourism due to its heavy reliance on environmental factors, primarily consistent snow cover and cold temperatures, which determine both the seasonality and attractiveness of the destinations. The viability of ice and snow tourism is largely tied to the reliability of climatic conditions, making it particularly vulnerable to the impacts of climate change. Research has demonstrated that warming temperatures and fluctuating snowfall patterns are diminishing the length and reliability of ski seasons in many regions, highlighting growing climate-related challenges for winter tourism destinations, as reported by Scott et al. [7]. Research has demonstrated that warming temperatures and fluctuating snowfall patterns are diminishing the length and reliability of ski seasons in many regions, highlighting growing climate-related challenges for winter tourism destinations, as reported by Steiger et al. [8]. This shift not only challenges tourist satisfaction but also poses significant economic risks for destinations that depend heavily on winter tourism. The potential economic consequences of these changes are particularly concerning for communities where tourism constitutes a major source of income, as emphasized by Timoshenko [9].

Adaptive measures such as the development of snowmaking technologies, the implementation of energy-efficient infrastructure, and the diversification of tourism products through year-round recreational activities have been proposed to mitigate the impacts of climate change. Such adaptation strategies can help maintain the attractiveness and competitiveness of tourism destinations under changing environmental conditions, as highlighted by Hao [10]. Similarly, integrating local cultural experiences and promoting off-season tourism activities can diversify tourism offerings and reduce dependence on snow reliability, as emphasized by Richards [11].

Furthermore, studies have explored the importance of destination resilience, emphasizing the need for strategic planning that incorporates both environmental sustainability and the diversification of tourism experiences. Diversified tourism products, including cultural tours, winter festivals, and eco-tourism initiatives, can enhance the resilience of ice and snow tourism destinations by reducing dependence on a single tourism resource. The incorporation of such offerings can not only strengthen destination attractiveness but also enrich tourists’ overall experiences, as argued by Dawson and Scott [12].

Although ice and snow tourism has been widely examined in mature winter tourism destinations such as Canada and Norway, existing findings are largely grounded in long-established recreational cultures and stable market structures. Research in these contexts has primarily focused on climate-related risks, destination adaptation, and the sustainability of winter tourism systems, as demonstrated by Scott et al. and Steiger et al. [7,8]. In these contexts, winter tourism has evolved gradually over decades, with repeat visitors, well-developed infrastructure, and relatively predictable demand patterns. As a result, prior studies often conceptualize tourist satisfaction primarily in terms of service quality, snow reliability, and destination image within relatively stable operating environments. In contrast, China’s ice and snow tourism has developed under markedly different institutional and market conditions. Its rapid expansion has been strongly influenced by national policy initiatives, including the post-Olympic development agenda and large-scale participation campaigns. These measures have accelerated destination development and tourism demand growth within a relatively short period, as highlighted by Jiang [1], Such policy-driven development can generate seasonal congestion, capacity mismatches, and heightened service pressure, potentially intensifying tourists’ emotional responses during peak periods. Moreover, visitor structure differs substantially. While winter tourism markets in Canada and Norway are characterized by high levels of repeat participation and skill-based recreation, China’s ice and snow tourism market includes a large proportion of first-time or infrequent participants, many of whom originate from regions without natural snow conditions. Research in mature winter tourism markets has primarily focused on climate-related risks, destination adaptation, and the sustainability of winter tourism systems, as demonstrated by Scott et al. [7]. This distinctive market structure contributes to unique patterns of destination perception and tourist behavior, as noted by Liu and Guo [13]. These tourists may rely more heavily on symbolic and experiential cues such as festival atmosphere, snow landscapes, and entertainment activities when forming satisfaction judgments. Consequently, emotional reactions to crowding, pricing, and service delivery may play a more prominent role in shaping overall evaluations. In addition, the experiential configuration of ice and snow tourism in China extends beyond sport-centered activities. Whereas research in traditional winter destinations focuses predominantly on skiing and snow-based recreation, Chinese ice and snow destinations frequently integrate large-scale festivals, themed architecture, and immersive entertainment environments. This hybrid experiential structure suggests that satisfaction formation may involve stronger affective and symbolic components, rather than being driven solely by functional performance attributes. Finally, the high penetration of online platforms in China has resulted in extensive user-generated content that records tourists’ emotional expressions in real time. This digitally mediated environment provides both a distinctive behavioral context and a rich empirical basis for analyzing how emotions dynamically shape satisfaction. Taken together, these contextual differences indicate that conclusions derived from mature Western winter tourism markets may not be directly transferable to China. A dedicated empirical investigation is therefore necessary to examine whether and how emotional dynamics influence tourist satisfaction within this rapidly developing and policy-driven ice and snow tourism system.

However, a key factor influencing the sustainability of ice and snow tourism lies in tourists’ satisfaction. Beyond snow conditions, satisfaction is shaped by service quality, infrastructure, and emotional experiences. Understanding these perceptions is essential for maintaining destination appeal in the face of climate uncertainties.

2.2. Tourists’ satisfaction about ice and snow tourism

Tourist satisfaction is a nuanced and multifaceted concept in ice and snow tourism, given its reliance not only on environmental conditions but also on a complex interplay among service quality, infrastructure, and emotional experience. Traditional research has predominantly employed structured questionnaires, interviews, and panel data to explore this phenomenon. Even relatively small variations in snow depth can significantly influence tourist satisfaction, as demonstrated through dynamic panel data analysis by Falk [14]. User-generated photographs have also been shown to serve as effective proxies for evaluating destination image and aligning tourists’ expectations, as demonstrated by Kim and Stepchenkova [15] Furthermore, positive emotional experiences characterized by joy, affection, and surprise are closely associated with tourist satisfaction and word-of-mouth intentions, based on the multidimensional emotional framework developed by Hosany et al. [16]. The importance of understanding the relationship between emotional responses and tourism experience evaluation has been further emphasized by Zhang [17]. However, these methods exhibit several noteworthy limitations: surveys and interviews are typically conducted post-experience and so lack the capacity to capture real-time fluctuations in tourist sentiment; the convenience-based sampling often produces low sample sizes that compromise representativeness; and psychometric scales, despite rigorous design, fail to uncover subtle or emergent emotions embedded in spontaneous online reviews. Indeed, textual analyses of ice and snow tourism reviews suggest that sentiment dynamics during tourism experiences often associated with factors such as pricing transparency, service quality, and on-site delays are considerably more complex and fluid than can be captured by traditional survey-based approaches. Tourist satisfaction and evaluation are influenced by multiple interrelated experiential factors that evolve dynamically throughout the travel process, as emphasized by Puh and Babac [18]. Moreover, while sentiment analysis models (e.g., IDCAN-BiLSTM) developed for this niche have achieved over 92% classification accuracy, they underscore the need for computational processing of large-scale textual data to decipher granular emotional patterns in near real-time. These shortcomings signal a clear methodological gap: conventional tools are ill-suited to track the temporal evolution and contextual nuances of tourist satisfaction in ice and snow environments. As such, there is a growing imperative to adopt more scalable, dynamic, and data-driven approaches to better understand tourist experiences in this specialized sector.

2.3. Big data analysis in ice and snow tourism

The emergence of big data analytics has profoundly transformed tourism research, providing novel insights into tourists’ behavior, demand forecasting, and satisfaction modeling. Li and Law [19] highlight the potential of big data in revolutionizing tourism marketing by offering a more granular understanding of consumer preferences and behaviors. Mariani [20] provide a thorough overview of the challenges and opportunities associated with big data in tourism and hospitality, noting that despite its potential, methodological and ethical issues remain significant barriers to its widespread adoption.

The impact of user-generated content (UGC) on destination image formation has become a central topic in contemporary tourism research. Online reviews and social media content have become critical components of destination marketing and play an important role in shaping tourist perceptions, as demonstrated by Chemin et al. [21]. Machine learning techniques have been increasingly employed to predict the sentiment of online reviews, providing a novel framework for comparative sentiment analysis, as proposed by Budhi, Chiong [22]. Furthermore, network analysis has been applied to tourism big data to uncover key research trends and emerging thematic clusters, thereby mapping the evolving landscape of tourist preferences and behaviors, as demonstrated by Li et al. [23].

Big data technologies have demonstrated considerable potential to enhance both theoretical understanding and practical applications in tourism research, as highlighted by Li and Law [19]. Topic-modeling-based sentiment classification has been shown to improve the accuracy of satisfaction assessment in consumer-generated online reviews, as demonstrated by Ali, Omar [24]. Furthermore, large-scale data analytics combining sentiment analysis and Latent Dirichlet Allocation (LDA) can provide valuable insights into the factors influencing tourist satisfaction and destination competitiveness, as evidenced by the study of Harbin’s ice and snow tourism conducted by Jiang et al. [25].

Despite a growing body of literature on ice and snow tourism, current research remains fragmented and methodologically limited, particularly in leveraging diverse data modalities and real-time analysis. Existing studies often rely on single-source data such as surveys or static textual reviews, which fall short in capturing the dynamic, multi-dimensional nature of tourist satisfaction. In response to these gaps, this study proposes an integrated analytical framework that combines user-generated reviews and sentiment analysis techniques to more accurately and dynamically assess tourist satisfaction in ice and snow tourism contexts. By addressing both spatial and emotional dimensions, the study provides a more nuanced understanding of tourist behavior and offers practical insights for destination management and service optimization.

Existing studies on ice and snow tourism and tourist satisfaction can be broadly categorized into three strands. The first strand focuses on climate dependency and sustainability, emphasizing the impacts of snow reliability, climate change, and adaptive strategies in traditional winter tourism destinations such as ski resorts in Europe and North America. The second strand examines tourist satisfaction determinants using survey-based or interview-based methods, highlighting the roles of service quality, infrastructure, and destination image, but largely treating satisfaction as a static outcome measured ex post. The third strand applies big data and sentiment analysis to tourism research, demonstrating the value of user-generated content for identifying experiential themes and emotional tendencies, yet often relying on cross-sectional designs and descriptive correlations. Despite these advances, several limitations remain. First, prior studies rarely integrate thematic cognition and emotional valence within a unified analytical framework, leaving their interaction insufficiently explored. Second, most existing sentiment-based studies focus on static associations and do not model the temporal dynamics or causal relationships between emotions and satisfaction. Third, research on ice and snow tourism has predominantly examined mature Western markets, while rapidly developing, policy-driven contexts such as China remain underrepresented. Consequently, there is a lack of evidence on how emotional fluctuations shape satisfaction formation in large-scale, highly seasonal, and digitally mediated tourism environments. To address these gaps, the present study proposes an integrated framework combining topic modeling, domain-adapted sentiment analysis, and vector autoregression. By leveraging large-scale longitudinal review data, this study explicitly models the dynamic and asymmetric effects of tourist emotions on satisfaction, thereby extending existing literature from static evaluation to a process-oriented understanding of satisfaction formation in ice and snow tourism.

2.4. Research gaps and study contributions

Existing studies on ice and snow tourism and tourist satisfaction have generated important insights into climate dependency, destination image, service quality, and tourist emotional experiences. However, several important research gaps remain.

First, previous studies have predominantly relied on survey-based or interview-based approaches, which often treat tourist satisfaction as a static post-consumption outcome. Such approaches have limited ability to capture the dynamic and evolving nature of tourists’ emotional experiences in highly seasonal tourism contexts.

Second, although recent studies have introduced sentiment analysis and text mining techniques into tourism research, most existing analyses remain descriptive and cross-sectional in nature. Limited attention has been paid to the temporal dynamics and causal interrelationships between emotional fluctuations and tourist satisfaction.

Third, prior research has largely focused on mature winter tourism markets in Europe and North America, while rapidly developing and policy-driven ice and snow tourism systems such as China remain underexplored. Consequently, the emotional mechanisms underlying tourist satisfaction formation in emerging winter tourism markets are still insufficiently understood.

To address these gaps, the present study proposes an integrated analytical framework combining LDA topic modeling, domain-adapted sentiment analysis, and Vector Autoregression (VAR) modeling. By leveraging large-scale longitudinal user-generated content from China’s ice and snow tourism destinations, this study contributes to the literature in three ways. First, it advances tourist satisfaction research from static evaluation toward a dynamic process-oriented perspective. Second, it integrates thematic cognition and emotional responses within a unified computational framework. Third, it empirically reveals the asymmetric and temporal effects of emotional fluctuations on tourist satisfaction in rapidly developing ice and snow tourism contexts. Table 1 summarizes the representative literature, major methodologies, key findings, and remaining limitations in existing studies related to ice and snow tourism and tourist satisfaction.

Download:

• PNG
  larger image
• TIFF
  original image

Table 1. Summary of prior studies on ice and snow tourism and tourist satisfaction.

https://doi.org/10.1371/journal.pone.0358229.t001

3.1. Research area overview

China’s integration of ice and snow tourism into its national development strategy, accelerated by the 2022 Beijing Winter Olympics and the “300 million participants in ice and snow sports” initiative, has transformed the sector into a key driver of regional economic revitalization, particularly in northern and western regions, as highlighted by Tang et al [26].To analyze tourist satisfaction dynamics across diverse contexts, this study examines fifteen representative destinations across China. Deliberately selected for spatial representativeness and thematic diversity—encompassing urban attractions, ecological snowfields, alpine resorts, and rural villages—these sites include key locations in the Northeast (e.g., Harbin, Changchun), Northwest (Xinjiang), and North-Central regions (e.g., Hebei, Beijing). Their prominence on major travel platforms ensures high tourist traffic and robust data. This selection enables the investigation of spatial heterogeneity in satisfaction determinants, providing a nationally relevant framework for advancing seasonal tourism theory and practice in emerging markets.

3.2. Data sources and processing

The empirical foundation of this study is a large-scale dataset of user-generated content (UGC) collected from Ctrip (https://www.ctrip.com), one of the most influential online travel service and booking platforms in China. Following the data collection approach adopted by Xia et al. [27], Ctrip reviews were selected because they provide rich, large-scale, and experience-oriented textual information that is well suited for tourism satisfaction analysis. Fifteen iconic ice-and-snow tourism destinations across China were selected for analysis based on their established prominence in national tourism statistics, high digital visibility on travel platforms, and their geographical and experiential representativeness of the winter tourism sector (e.g., Harbin Ice and Snow World exemplifying large-scale ice architecture; Yabuli Ski Resort representing major ski destinations). Table 2 presents the full list of these destinations along with their respective review counts and proportional distributions, demonstrating the substantial coverage of China’s major ice and snow tourism markets.

Download:

• PNG
  larger image
• TIFF
  original image

Table 2. Distribution of online tourist reviews across selected ice and snow tourism destinations in China.

https://doi.org/10.1371/journal.pone.0358229.t002

We collected 32,634 Ctrip reviews (2018–2025) using a Python crawler. After deduplication, length filtering, and text normalization (removing non-Chinese elements, HTML, and emojis), domain-specific tokenization and stop-word removal were applied. A custom lexicon standardized tourism terminology (e.g., unifying festival/service synonyms). The final cleaned corpus of 32,516 reviews—annotated with ratings, timestamps, and destinations—was prepared for downstream analysis.

3.3. Research framework based on the CRISP-DM process

To improve methodological transparency and reproducibility, the analytical workflow of this study was conceptually aligned with the Cross-Industry Standard Process for Data Mining (CRISP-DM), a widely recognized framework for structuring data-driven research processes. The CRISP-DM methodology organizes analytical procedures into interconnected phases, including business understanding, data understanding, data preparation, modeling, evaluation, and deployment. Following this logic, the present study integrates text mining, sentiment analysis, and time-series modeling into a unified analytical framework for investigating tourist satisfaction dynamics in ice and snow tourism contexts.

Fig 1 illustrates the overall research framework. The analytical workflow operationalizes the major phases of the CRISP-DM process through three interconnected stages: data collection and preparation, computational modeling, and evaluation and interpretation.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 1. Research Framework Flowchart.

https://doi.org/10.1371/journal.pone.0358229.g001

In the data collection and preparation stage, large-scale user-generated reviews were collected from the Ctrip platform and subjected to systematic preprocessing procedures, including duplicate removal, invalid text filtering, text normalization, tokenization, and stop-word removal. After data cleaning, 32,516 valid reviews from 2018 to 2025 were retained for subsequent analysis. The broad temporal coverage enabled the study to capture seasonal fluctuations and policy-related dynamics in tourists’ experiential evaluations.

In the computational modeling stage, the cleaned textual corpus underwent an integrated analysis combining LDA topic modeling and SnowNLP-based sentiment analysis. LDA was employed to identify latent experiential themes embedded in tourists’ online reviews, while sentiment analysis was used to quantify emotional valence associated with different tourism experiences. Through iterative coherence evaluation and semantic interpretation, ten dominant experiential dimensions were extracted, including natural ecology, polar wildlife, rural village landscapes, accommodation and transportation, mountain snow scenery, skiing facilities and equipment, service quality, ice and snow entertainment, snow-themed architecture, and fairy-tale atmosphere. Simultaneously, a domain-adapted SnowNLP model trained on manually labeled tourism reviews generated sentiment scores ranging from 0 (highly negative) to 1 (highly positive), allowing emotional characteristics to be associated with thematic dimensions.

In the evaluation and interpretation stage, the study further extended from descriptive text mining to dynamic causal analysis by incorporating topic-level sentiment indicators into a Vector Autoregression (VAR) framework. Monthly time-series variables were constructed to capture temporal fluctuations in tourist satisfaction dimensions. Granger causality tests and impulse response analyses (IRFs) were subsequently employed to examine the dynamic interrelationships and temporal effects among emotional experiences and satisfaction drivers. This integrated framework enables systematic interpretation of tourist satisfaction formation in China’s ice and snow tourism context and provides data-driven insights for destination management and tourism service optimization.

3.4. Research methods

3.4.1. LDA model.

A Bayesian generative model was employed to identify latent topics within large document collections. This approach represents documents as mixtures of topics and has been widely adopted in topic modeling research, as originally proposed by Blei et al. [28]. It represents each document as a mixture of topics, while each topic is characterized by a probability distribution over words, as further discussed by Chang et al. [29]. In the present study, this flexible and noise-tolerant approach was applied to online reviews of ice and snow tourism destinations in Northeast China, following the application framework demonstrated by Marrese-Taylor et al. [30]. The analysis generated both document-topic and topic-term distribution matrices, providing the basis for subsequent thematic interpretation. Topic coherence optimization determined the optimal number of topics and their weights, enabling macro-level analysis of tourism characteristics to support subsequent sentiment analysis. As shown in Table 3, the coherence score reached its maximum at ten topics, which was therefore selected for subsequent analysis. Specifically, the Cv coherence metric was adopted because it has been widely recognized as an effective indicator for evaluating semantic interpretability in short tourism-related texts. Compared with perplexity-based evaluation, coherence scores better reflect human understanding of topic quality. The number of topics was determined through an iterative comparison of alternative models, followed by semantic validation to ensure conceptual clarity and practical interpretability.

Download:

• PNG
  larger image
• TIFF
  original image

Table 3. Topic coherence scores for candidate LDA models.

https://doi.org/10.1371/journal.pone.0358229.t003

3.4.3. VAR.

Vector Autoregression (VAR) is utilized to assess the temporal dynamics and interrelationships among sentiment-enhanced thematic variables. This method is particularly appropriate in the present research context because it treats all variables as endogenous and models their dynamic interactions over time without imposing strong a priori assumptions regarding causal directionality, consistent with the VAR framework developed by Sims [33]. Specifically, we specify a VAR(p) model of the form

(1)

Where is the k-dimensional vector ofendogenous variables in month t (k = 1 + 3 × 10 = 31); St denotes the overall satisfaction score in month t; represent the positive,neutral,and negative sentiment scores, respectively, for theme i in month t, all derived from text-based sentiment analysis of online reviews; c is a k × 1 intercept vecto; Aj (j = 1,⋯,p) are k × k coefficient matrices capturing own-lag and cross-lag effects.

This setup treats all sentiment and satisfaction variables as jointly endogenous, allowing each series’ own lags and other series’ lags to explain its dynamics. Granger causality tests, originally proposed by Granger [34], were subsequently employed to determine the existence and direction of predictive relationships between variables. Generalized impulse response functions (IRFs), following the approach of Pesaran and Shin [35], were then used to visualize the effects of shocks in one variable (e.g., a decline in facility satisfaction) on other dimensions of tourist experience (e.g., service-quality perception) over a specified time horizon.The combination of these three methods provides a robust, interpretable, and policy-relevant approach for analyzing the complexity of tourist satisfaction in multi-dimensional and seasonally sensitive tourism contexts.

4.1. The core experiential topics of ice and snow tourism

Latent Dirichlet Allocation (LDA) analysis of 32,516 online reviews identified ten core experiential dimensions characterizing tourists’ experiences in China’s ice and snow tourism destinations, including travel experience, ice and snow sports, architectural style, service facilities, natural landscape, interactive tours, snow landscape, outdoor scenery, accommodation and transportation, and parent-child entertainment.

Fig 2 presents the thematic review distributions for several representative ice and snow tourism destinations selected from the study sample. Clear differences in thematic emphasis can be observed across destinations, reflecting their unique tourism resources and market positioning. Harbin Polar Park received a particularly large number of reviews related to architectural style and interactive tourism experiences, indicating the importance of immersive attractions and visitor engagement activities. Saint Sophia Cathedral was strongly associated with architectural style and cultural landscape themes, highlighting its role as a landmark heritage attraction. In contrast, destinations characterized by natural resources, such as Changbai Mountain and Volga Manor, attracted a greater proportion of reviews related to snow landscapes, outdoor scenery, and travel experiences. China Snow Town exhibited a strong concentration of reviews associated with accommodation and transportation as well as travel experience, suggesting that infrastructure and visitor logistics play a significant role in shaping tourists’ evaluations.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 2. Thematic review distributions for representative ice and snow tourism destinations selected from the full sample of fifteen study sites.

https://doi.org/10.1371/journal.pone.0358229.g002

Overall, the thematic distributions reveal substantial heterogeneity in tourist experiences across ice and snow tourism destinations. While some destinations are primarily valued for their natural and scenic resources, others derive their attractiveness from cultural heritage, recreational activities, interactive experiences, or family-oriented tourism products. These findings highlight the multidimensional nature of tourist satisfaction in China’s rapidly developing ice and snow tourism market.

Due to space limitations and readability considerations, only representative destinations are visualized in Fig 2, while the complete thematic distributions for all study destinations are provided in the Supplementary Material.

Crucially, the co-occurrence of activity-based (e.g., entertainment), environmental (scenery/ecology), cultural (village), and operational themes (transport/accommodation) in LDA results demonstrates that tourist satisfaction is driven by dynamic interactions among multidimensional factors—challenging traditional single-factor models. These findings highlight the need for holistic management strategies that integrate experience design and infrastructure development, particularly in addressing transportation and service bottlenecks during peak tourism seasons. Such improvements are essential for maintaining destination competitiveness and enhancing overall visitor satisfaction. Furthermore, the observed thematic disparities suggest that destination managers should adopt tailored regional development strategies that capitalize on unique local resources and experiential attributes, consistent with the experience economy framework proposed by Sun et al. [36]. Topic coherence analysis identified ten interpretable topics from the corpus of online reviews. Table 4 presents the LDA-derived topic categories and their representative high-probability keywords. The identified themes encompass both destination attributes and experiential dimensions, including travel experience, ice and snow sports, architectural style, service facilities, natural landscape, interactive tour, snow landscape, accommodation and transportation, and parent-child entertainment. Topic labels were assigned based on semantic interpretation of the highest-weighted keywords and validated through expert review.

Download:

• PNG
  larger image
• TIFF
  original image

Table 4. The results of topic model and keywords.

https://doi.org/10.1371/journal.pone.0358229.t004

4.2. Seasonal sentiment analysis

Sentiment analysis of 32,516 validated reviews revealed a robust positive skew across China’s ice-snow tourism sector (68.4% positive, 24.1% negative, and 7.5% neutral), indicating generally favorable tourism experiences. This distribution suggests that tourist satisfaction is closely associated with the degree of consistency between expected and actual experiences, a relationship that can be explained by the expectation-disconfirmation framework proposed by Oh et al. [37]. Significant seasonal heterogeneity was observed across experiential dimensions, demonstrating that satisfaction drivers vary substantially over time and that infrastructure and service systems play different roles across seasons, as illustrated in Fig 3.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 3. Seasonal sentiment distrbution by topic.

https://doi.org/10.1371/journal.pone.0358229.g003

Spring: Spring was characterized by strong positive sentiment toward Travel Experience, Natural Landscape, Interactive Tour, and Architectural Style. These results suggest that tourists highly valued the combination of natural scenery and cultural attractions during the shoulder season. However, comparatively higher negative sentiment was observed for Ice and Snow Sports and Service Facilities, indicating that some tourism facilities and winter-related recreational services were unable to fully satisfy visitor expectations after the peak winter season. This pattern is consistent with expectation-disconfirmation theory, which suggests that dissatisfaction may arise when actual service performance falls below anticipated standards, as explained by Oliver [38].

Summer: Summer exhibited a mixed sentiment pattern. Positive evaluations remained high for Travel Experience, Service Facilities, and Architectural Style, reflecting the benefits of lower visitor pressure and improved service accessibility during the off-season. In contrast, Natural Landscape and Interactive Tour attracted relatively higher levels of negative sentiment compared with other topics. The reduced availability of snow-related tourism resources and the discrepancy between winter-oriented destination images and actual summer conditions may contribute to this dissatisfaction. Such findings highlight the importance of destination image management and expectation alignment, as discussed by De Bloom et al. [39].

Autumn: Autumn: Tourists expressed particularly positive sentiment toward Natural Ecology, Village Landscape, and Snow Mountain Scenery, suggesting strong appreciation for transitional seasonal landscapes. Nevertheless, Accommodation and Transportation and Service Evaluation attracted comparatively higher negative sentiment. Resource constraints and preparations for the upcoming winter season may reduce service efficiency, reinforcing the importance of transportation accessibility and service quality as key determinants of tourist satisfaction, as emphasized by Buttle and Francis [40].

Winter: Winter displayed the strongest emotional polarization among all seasons. Travel Experience, Accommodation and Transportation, and Service Facilities generated exceptionally high levels of positive sentiment, reflecting tourists’ immersion in winter recreational activities and destination experiences. At the same time, these dimensions also received relatively high levels of negative sentiment, indicating that crowding, capacity limitations, and operational pressure during the peak tourism season may adversely affect visitor experiences. This coexistence of intense positive and negative emotions highlights the critical role of service quality management in maintaining tourist satisfaction during periods of peak demand, as discussed by De Bloom et al. [39].

Destination-level analysis further contextualized these seasonal dynamics. Harbin Ice and Snow World generated particularly strong positive sentiment regarding Travel Experience and Ice and Snow Sports, reflecting its distinctive winter recreation attractions. Harbin Polar Park was characterized by high levels of positive sentiment associated with Interactive Tour experiences, whereas destinations such as Changbai Mountain and Saint Sophia Cathedral received consistently favorable evaluations for Natural Landscape and Architectural Style, respectively. Cross-thematic analysis further suggested that positive evaluations of Travel Experience were frequently accompanied by positive perceptions of Natural Landscape and Service Facilities, indicating that immersive tourism experiences may enhance tourists’ overall evaluations of destination attractiveness.

4.3. Empirical results of VAR

Vector Autoregression (VAR) modeling reveals dynamic interdependencies among tourism satisfaction factors, with satisfaction measured using ratings. Granger causality tests establish that satisfaction with “Ice and Snow Sports” is a significant Granger cause for satisfaction with “Natural Landscape” (p < 0.01), “Snow Landscape” (p < 0.05), and “Service Facilities” (p < 0.01), indicating winter activities systematically enhance environmental and service evaluations. Crucially, bidirectional causality links satisfaction with “Natural Landscape” and “Ice and Snow Sports” (p < 0.01), empirically validating the activity-environment reciprocity mechanism where recreational immersion and environmental appreciation reinforce each other. Meanwhile, satisfaction with “Accommodation and Transportation” functions as both driver and outcome—significantly Granger-causing satisfaction with “Service Facilities” (p < 0.01) and “Travel Experience” (p < 0.05), while being significantly Granger-caused by satisfaction with “Ice and Snow Sports” (p < 0.05) and “Snow Landscape” (p < 0.01), highlighting infrastructure’s critical role in the satisfaction valuation process.

Impulse response analysis demonstrates that a one-standard-deviation shock to aggregate tourist sentiment generates a sustained satisfaction increase, stabilizing at +0.82 rating points by period 5 (p < 0.001), as visualized in Fig 4. This confirms sentiment as a temporally antecedent driver and establishes the sentiment-rating inertia mechanism (IRF Δ = +0.82, p < 0.001). The persistent positive trajectory indicates that initial sentiment improvements create self-reinforcing satisfaction loops, with effects remaining significant throughout the observed horizon.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 4. Impulse Response Analysis.

https://doi.org/10.1371/journal.pone.0358229.g004

These findings establish satisfaction as a multidimensional construct driven by recursive sentiment-rating linkages (inertia). The activity-environment synergy (where snow sports catalyze spillovers) and infrastructure’s critical mediation reveal inherent temporal sensitivity, with rating stabilization demonstrating experiential reevaluation inertia. Holistic management must therefore synchronize activity-environment integration with preemptive mitigation of logistical bottlenecks during peak demand to capitalize on cross-dimensional synergies. Regional LDA disparities (e.g., Harbin’s entertainment-architecture vs. Sayram Lake’s scenery-village focus) further modulate these pathways, reframing satisfaction as a spatiotemporally contextualized outcome of recursive environment-infrastructure-sentiment interactions.