Introduction

Landscape contributes enormous and diversified benefits to humans in myriad ways, and collectively, these benefits have been defined as ecosystem services (MEA 2005). These services are normally divided into providing, regulating, supporting, and cultural ecosystem services (CESs). The effectiveness of these services largely depends on the spatial distribution, pattern of occurrence, and interactions among them (Chicago et al. 2022; Zoeller et al. 2022). CESs are defined as nonmaterial benefits with characteristics of ‘intangible’ and ‘subjective’ and thus, are relatively more difficult to investigate than other services (Mandle et al. 2021; Huynh et al. 2022). Nevertheless, CESs play a pivotal role in fostering human well-being through the direct human-nature interactions, especially within human-dominated landscapes which emphasize nature-society interactions (Wu 2006; Daniel et al. 2012). Hence, exploring how CESs are delivered from landscapes to human society helps understanding the tradeoffs and synergies between ecosystem services and landscape sustainability (Cheng et al. 2022; Huynh et al. 2022).

Although the importance of CESs is increasingly recognized, there is no consensus among researchers, policymakers, and stakeholders regarding how CESs are effectively delivered (Huynh et al. 2022; Xu and Peng 2022). In the fields of ecology, conservation science, and urban planning, studies aimed at CES enhancement mainly focus on the role of landscape or socio-demographic attributes in shaping perceptions of CESs (Daniel et al. 2012; Dou et al. 2021; Zhang et al. 2023), such as landscape features (i.e. size and connectivity) and land use contexts (i.e. surrounding construction density) (Wang et al. 2022). Besides, socio-demographic attributes like age (Gai et al. 2022), gender (Zhou et al. 2020), and education level (van Zanten et al. 2016) have been found to moderate the relationship between humans and nature. Although these findings have unraveled the relationship between CES and some landscape or personal factors, the underlying mechanism linking external natural environment and internal mental benefits remains unclear. For example, how we comprehend, develop, and enact the human-nature relationships is rather critical but poorly understood (Pramova et al. 2021). This inadequacy hinders the understanding of how people perceive CESs from landscape attributes and the development of interventions aimed at CES enhancement (Gould et al. 2020).

Human-nature interactions are created through the dynamic interplay of the mind, body, and the surrounding environment (Cooke et al. 2016). How people respond to nature does not exclusively depend on the external environment, but also on the internal dimension of psychological experience (Malhotra 1984; Agapito et al. 2013). A central challenge thus lies in understanding how people perceive landscape attributes from the perspective of human psychological experience (Jones et al. 2022a). Fortunately, these internal realms have been well investigated in the field of psychology (Gifford 2014). A wealth of theories have been proposed to understand how people perceive the surrounding environment, such as stimulus-organism-response theory (Mehrabian and Russell 1974), cognitive appraisal theory (Lazarus 1966), and multisensory integration theory (Calvert et al. 2004). Notably, psychology and ecology have divergent emphases, with the former focusing on biological and clinical significance, while the latter seeks to enhance human benefits from nature (i.e. CESs) through landscape planning and design. Yet, the convergence of established psychological theories into specific landscape planning is scarce and challenging. For instance, sensation, cognition, and affect are essential factors that determine people’s perceptions and responses to environment, while how to use these critical concepts to enhance CESs remains underexplored. This lack of operational clarity could be due to the disciplinary gaps between psychological concepts and ecological practice. Consequently, the rational interpretation and effective application of psychological principles within the realm of ecology offers promising prospects for bridging the gaps. For instance, linking the concept of sensation with landscape sensory attributes might be a valid interdisciplinary integration (Chen and Lin 2023). This interdisciplinary integration holds the potential to enhance landscape management, improve human well-being, and advance sustainability (Wu 2021; Zhou et al. 2021).

Sensation is regarded as the essential modality for mankind to appreciate and comprehend the world directly (Bratman et al. 2019), linking external physical nature and internal human psychology. Ecologists understand the role of sensation mainly using landscape sensory attributes (Chen and Lin 2023). Existing research identified critical landscape sensory attributes that significantly influence the perception of CESs, such as landscape visual features like color diversity, landscape heterogeneity, and vegetation functioning (Vaz et al. 2019), and other sensory features like feel of sunlight and sound of animals (Zheng et al. 2020). However, further investigations are required to achieve a deeper understanding of how people respond to landscapes and how landscape sensory features can be optimized to enhance CESs (Bratman et al. 2019). These guiding questions highlight the need to clarify how landscape sensation contributes to CESs.

From the perspective of psychology, sensation cannot be directly regarded as nonmaterial benefits, but only the starting point to receive environmental stimuli (Huppert 2009; Gifford 2014). The pathway from sensation to nonmaterial benefits has been regarded to engender cognitive and affective responses (hereafter ‘cognition’ and ‘affect’) (Bratman et al. 2021a, 2021b). More specifically, the sensory organs firstly act as the signal system for the physical body to receive external stimuli (sensation); then the signals are sent to the brain and produce cognitive and affective responses; finally, nonmaterial benefits are obtained (Bratman et al. 2012). Sensation, cognition, and affect are interrelated processes, each interpreted differently by psychologists (Leventhal and Scherer 1987). In this paper, we consider ‘cognition’ to be how we process sensory information, leading to outcomes like aesthetic judgment (Pessoa 2008). Affect, on the other hand, is an emotional response to experiences, resulting in feelings like pleasure or excitement (Pramova et al. 2021). Affect generation is considered to depend on cognitive outcomes and physiological arousal (Weiner 1985). Based on these theories, psychologists commonly hold the viewpoint that cognition and affect play different roles in the pathway to nonmaterial benefits (Pessoa 2008; Lähteenmäki et al. 2015). For example, cognition was found to bring about experiences of life meaning, generate positive affect, and thereby enhance well-being (del Bosque and San Martín 2008). Understanding the different roles of cognition and affect in the pathway from sensation to CESs can drive a deeper exploration of how sensation contributes to CESs. These psychological concepts also suggest new directions for further landscape interventions aimed at CES enhancement.

It remains a challenge to comprehensively assess CESs (Jones et al. 2022a), as CESs encompass a wide range of nonmaterial benefits, varying with the type, duration, and frequency of human-nature interactions (Bratman et al. 2012; Yao et al. 2024). Recent studies introduced the term satisfaction to assess the quality and effectiveness of CESs, such as those provided by species groups (McGinlay et al. 2017), bird communities (Cumming and Maciejewski 2017) and urban forests (Wang et al. 2022). Satisfaction has its foundations in social science to evaluate service quality (Tam 2004), as it comprehensively assesses the feelings and attitudes towards services across various aspects, including their diversity, accessibility, and value of services (Fornell et al. 1996). Therefore, satisfaction, driven by service quality, serves as a useful indicator for assessing CESs. Some studies have proposed to understand human-nature interactions from the perspective of the sensory, affective and cognitive dimensions (Pramova et al. 2021), while the interactive effects of these psychological experiences on CESs have not been explicitly investigated (Norwood et al. 2019). Although the pathway from sensation to cognition and affect, and ultimately to satisfaction is logically sound in psychology, systematic investigation is imperative to scrutinize and validate these intricate relationships within the context of human-nature interactions.

This study aims to explore the pathways from landscape sensation (hereafter ‘sensation’) to CESs with an interdisciplinary framework (Fig. 1). We utilize the psychological terms ‘cognition’ and ‘affect’ to express how benefits from nature are delivered. Natural attractions are chosen as our study area because they provide diverse CESs. We seek to tackle the following scientific questions: (1) How are the effective pathways linking landscape attributes to CESs? (2) What are the roles of cognition and affect in the CES-delivery pathway? and (3) How can CESs be integrated into landscape planning and management using an interdisciplinary framework? Addressing these questions will help to establish a win–win relationship between humans and nature, thus boosting the nonmaterial contributions of landscapes to human societies.

Fig. 1
figure 1

Schematic illustration of interdisciplinary processes from sensation to CESs

Full size image

Methods

A framework for assessing possible pathways from landscape sensation to CESs

The cognitive-affective framework proposed by del Bosque and San Martín (2008) was used to abstract and simulate the processes of the nonmaterial benefit delivery from nature to humans. Landscape sensation, which is a fundamental environmental stimulus for humans (Pramova et al. 2021), was used as the exposure variable. Satisfaction acted as the outcome variable to assess CESs. Cognition and affect were mediator variables. The overarching hypothesis was that cognition influences affect according to the attribution theory of emotion in psychology (Weiner 1985).These four variables were included in the conceptual model, and six hypotheses elaborating on the interaction between these variables were proposed and tested (Fig. 2). Based on the six hypotheses, a conceptual framework was developed, unveiling the possible interaction profiles. Specifically, there were four possible paths from sensation to satisfaction (Fig. 2): (i) from sensation, directly to satisfaction, (ii) from sensation, mediated by cognition, to satisfaction, (iii) from sensation, mediated by affect, to satisfaction, and (iv) from sensation, mediated by cognition, followed by affect, to satisfaction. The research flowchart can be found in Fig. S1.

Fig. 2
figure 2

Conceptual framework of possible pathways to CES outcome

Full size image

Measurement

We used social media comment data to develop the measurement scale of cognition with eight indicators and draw on existing scales to measure the other three latent variables: sensation with twelve indicators, affect with two indicators, and satisfaction with four indicators. These indicators were used to design questions in the questionnaire for survey data collection.

Landscape sensation

The landscape sensation was measured with twelve indicators, which reflected landscape diversity, biodiversity, vegetation heterogeneity, vegetation density, color, natural soundscape, noise control, music, temperature, sunshine, natural flavor, and smell. These indicators covered four dimensions of landscape sensory features: visual, auditory, tactile, and olfactory sensations (Zheng et al. 2020). Both natural and artificial sensations were taken into consideration in the measurement scale to fully unveil the interaction between multidimensional sensory and the authentic sceneries of thematic scenes (Aletta et al. 2016). Generally, questions were designed to indicate each sensory dimension as presented in the questionnaire. Unless otherwise stated, the items are regarded to positively correlate with sensation. For example, the participants were required to answer questions such as “I think natural soundscape could be heard in the scenic spot, such as birds chirping, frogs croaking, and/or underwater sound” with five choices on a 5-point Likert scale, ranging from 1 (completely disagree) to 5 (completely agree). A higher score indicated a greater perceptual quality of sensation. For the negative items (e.g., unpleasant odor), participants were asked questions like “I think bad odor could be smelled in the scenic spot, such as car exhaust, the smell of garbage, etc.” with the same five point scale to choose from. This measuring scale was then reversed so that lower scores reflected better sensation.

Cognition

Since cognitive outcomes are highly dependent on the context (Valtchanov and Ellard 2015), there are no existing scales available for us to draw upon. We therefore developed a suitable scale specifically for the purpose of this study. Big data analysis of social media comments was used to extract cognitive outcomes in natural attractions. To enhance representativeness, we selected six attractions based on different natural resources (see Table S1) according to the classification “China National Standard for the Classification of Tourism Resources”. Secondly, we used a Python-based (version 3.9) web crawler to scrape online comments on these selected destinations made by visitors on a popular travel website (https://ctrip.com/) from January 2018 to March 2022. To protect the privacy of these visitors, we implemented strict privacy protection during data collection and processing. Following the data minimization principle suggested by Di Minin et al. (2021), we obtained the comment texts useful for this research. We did not collect usernames, ID numbers, or geographic locations. All identifiable contents in the comment texts were anonymized. In addition, we used the Wilcoxon rank-sum test to examine any differences between the two groups of data before and during the pandemic of Covid. The results indicated no significant differences with a Wilcoxon rank-sum test of 21,414.0 and a P-value of 0.993. It’s reasonable to conclude that the pandemic had no significant impact on the data we used. Furthermore, we excluded comments unrelated to the natural experience. Semantic segmentation was then performed to extract lexemes (phrases) from the texts (sentences). As the frequency of words in textual data normally indicated the prominence of keywords in the entire text (Song et al. 2020), we conducted frequency analysis to designate the main cognitive factors. Following word frequency analysis, we incorporated the lexemes with relatively high frequency into the question formulation to measure the cognitive experience of the participants. Details of this method are included in Table S2.

Affect and satisfaction

Affect was measured with the two-dimensional model of emotion (Russell and Barrett 1999) and characterized by two dimensions: valence and arousal, representing whether this emotion is positive or negative (pleased-unpleased) and the intensity of the emotional state (sleepy-excited) (Russell and Barrett 1999; Russell 2003, 2009). The higher the scores, the more favorable the affect. To investigate satisfaction with CESs, four indicators were adopted: worthwhile, overall satisfaction, expectations, and correct decisions (Fornell et al. 1996). The same techniques were implemented for questioning and scoring as described above.

Survey data collection

We chose natural attractions as study sites to focus on visitors who have had CES experiences. An online survey was conducted to sample people with experience in any scenic spots during the past three months. The questionnaire was designed to infer the information of the participants from five aspects: landscape sensation, cognition, affect, satisfaction, and to obtain personal attributes (see Text S1). The survey was conducted through popular social media in China (e.g., WeChat and Sina Weibo) during April and May 2022. Screening questions were also included to ensure that the participants met the research requirement of possessing an experience of a natural attraction within the past three months. The investigation ultimately yielded a total of 550 replies. The protocol was approved by the Research Ethics Committee of the Institute of Urban Environment, Chinese Academy of Sciences (Registration Number: HSR-23–000221). We obtained informed consent from each participant, and implemented strict privacy protection measures, which ensured the anonymity of all participants and strict confidentiality of relevant information about the participants (see Text S1). Invalid samples were excluded if the completion time was less than thirty seconds (average completion time was two minutes), answers were completely same (e.g., all the selections were A), or the questionnaire was incomplete. A total of 503 valid questionnaires were finally obtained.

Partial least squares structural equation modeling

Partial least squares structural equation modeling (PLS-SEM) was employed to test the hypotheses. Compared to the traditional structural equation model, PLS-SEM performs better especially for small-sized samples and exploratory studies (Hair et al. 2019). Complete PLS-SEM was performed with a two-step modeling technique: measurement and structural models. The measurement model was assessed using reliability and validity to confirm the predictive potential of the observed variables for their respective latent constructs. The structural model was evaluated to assess the predictive capability among the latent constructs through hypothesis testing and mediation analysis. The number of surveys (N = 503) was sufficient to conduct PLS-SEM data analysis as the required minimum number was suggested to be greater than 10 times the questionnaire items (we had 26 items in this study) (Hair et al. 2019). We employed SmartPLS 3.0 to run the models and perform statistical analysis.

Results

Establishment of the cognition measurement scale

A total of 18,500 pieces of comment texts were obtained by using web crawler. After eliminating irrelevant comments, 15,987 pieces of comment data, comprising 1.42 million Chinese characters, formed the initial text material. The sample consisted of 2,986 pieces of comments from Enshi Grand Canyon, 2,971 from West Lake, 2,988 from Kulangsu, 1,174 from Rime Island, 2,876 from Base of Giant Panda, and 2,992 from The Century Park (see Table S1). After semantic segmentation and word frequency analysis, the cognitive words with high frequencies related to the research topic were filtered out and categorized synonymously (Fig. 3). Given the research focusing on landscapes, some highly frequent but irrelevant words such as ‘ticket price’, ‘cuisine’, and ‘accommodation’ were excluded. It should be noted that all the raw text material was in Chinese. After its translation into English, the high-frequency words with similar lexical meanings were further categorized to summarize the indicators of cognition. For example, the high-frequency words of beautiful, joyful, and magnificent corresponded to the cognitive judgment of aesthetic value were classified into ‘aesthetic value’ (Table 1). Ultimately, a cognition measurement scale was established with eight indicators, incorporating aesthetic value, cultural diversity, uniqueness, weather, environmental quality, sense of comfort, cleanliness, and sense of safety. Based on the eight indicators, we formulated eight questionnaire items to assess cognition (referring to the questionnaire in Text S1: items from 13 to 20) (Fig. 4)

Fig. 3
figure 3

Illustration of high-frequency words resulted from social media content analysis. The words within the central circle are lexemes with high frequency: the larger the font size, the higher the frequency. The segments of the doughnut outlier present the respective study sites and valid comment counts

Full size image
Table 1 Results of word frequency analysis
Full size table
Fig. 4
figure 4

Results of the structural model evaluation (*** indicates p < 0.001)

Full size image

Survey data statistics

This study categorized the demographic characteristics of participants based on five aspects: gender, age, educational background, occupation, and monthly income. More females (62.0%) than males (38.0%) participated in the survey, young (under 30) and middle-aged (from 30 to 59) dominated the age groups, and occupation was dominated by students (61.1%), followed by administrative staff (17.1%) and government employees (6.2%). The monthly income was concentrated between US $300 and US $1,200 (see Table S3). Overall, the samples were taken from a broad range of social strata and were reasonably representative of the background population, thus meeting the criteria of the research. The skewness values ranged from -1.67 to -0.31 and kurtosis values ranged from -0.96 to 3.30. Both skewness and kurtosis values were within the required cut-off points (-2 < skewness < 2; -7 < kurtosis < 7), and thus the normality of the survey data was acceptable (Hair et al. 2014).

Measurement model evaluation

The quality of the measurement model was evaluated through reliability, exploratory factor analysis (EFA) and confirmatory factor analysis (CFA) (Table 2 and Table S4). The values of Cronbach’s alpha and Dijkstra-Henseler’s rho were all above 0.7, indicating excellent reliability of the measurement scale (Hair et al. 2019). The values of the average variance extraction (AVE) were all greater than 0.5, suggesting a high correlation under the same measuring dimension (Fornell and Larcker 1981). The values of the Heterotrait-Monotrait ratio (HTMT) were much smaller than 1, implying that items under different constructs had significant differences (Hair et al. 2019). Collectively, these results showed that the overall model fitness of the measurement scale established in this study was valid.

Table 2 Results of the reliability and validity test
Full size table

Common method bias

If a study involves two or more factors originated from a self-reported survey, common method bias (CMB) may occur (Podsakoff and Organ 1986). To minimize CMB as much as possible, the independent factors were separated from the dependent variables on purpose when designing the questionnaire (Kock 2015). The variance inflation factors (VIFs) are a common index used for checking the CMB in PLS-SEM (Kock 2015). After a full collinearity test, all the VIF values in the model were found to be lower than 3.3 (Table 3), indicating that the model is considered free of CMB.

Table 3 Results of the full collinearity test
Full size table

Structural model evaluation

The validity of the six proposed hypotheses was determined with significance tests and the structural model was estimated by using a PLS bootstrapping approach with 50,000 bootstraps (Hair et al. 2019) (Table 4). Sensation had a significant positive impact on cognition (β = 0.803, p < 0.001), but no significant direct impact on satisfaction (p = 0.807) and affect (p = 0.311). In addition, cognition possessed a significant effect both on satisfaction (β = 0.500, p < 0.001) and affect (β = 0.723, p < 0.001). Moreover, affect significantly and directly influenced satisfaction (β = 0.435, p < 0.001).

Table 4 Results of hypothesis tests
Full size table

Mediation analysis

Two mediating effects passed the mediation test (cognition as a single mediator, and cognition and affect as two serial mediators). One mediating effect did not pass the test (affect as a single mediator) (Table 5). The total mediation effect was 0.654, among which the effect of cognition as a single mediator was 0.401 (p < 0.001), and the effect of cognition and affect as two serial mediators was 0.253 (p < 0.001). The coefficient of indirect effect for path ii (cognition as a single mediator) was 0.148 greater than that of path iv (cognition and affect as two serial mediators). These results supported hypotheses 2, 4, 5, and 6.

Table 5 Mediation test results
Full size table

Discussion

The key contribution of this study has been to identify two valid pathways of how people get nonmaterial benefits through landscape sensation, suggesting that the delivery of CESs from a landscape was through well-structured lenses of sensation, cognition, and affect.

Psychological mechanism of CES pathway

The nonmaterial contribution of nature to humans is well accepted but poorly understood (Lapointe et al. 2021; Jones et al. 2022b), as is the pathway of how these benefits are delivered (Huynh et al. 2022). This study indicated that CESs are delivered from sensation to satisfaction, through cognition as single mediator, or cognition and affect as two serial mediators. Previous studies have concurred that CESs require the engagement of sensation and brain activity to comprehend the environmental information (Cai et al. 2022; Chen and Lin 2023). Pramova et al. (2021) proposed that cognition and affect play critical roles in delivering intangible benefits from nature. This study was built on the previous studies to differentiate the distinct roles of sensation, cognition, and affect, and revealed that sensation failed to directly trigger affect and satisfaction. This finding is supported by the attribution theory of emotion, which contends that emotion is not triggered directly by mere physiological arousal but by the cognitive process (Schachter and Singer 1962). The reasons behind the statistical rejection of a direct pathway from sensation to satisfaction remain obscure. One plausible explanation could be that sensation is unable to reach the threshold for the formation of satisfaction (Biedenweg et al. 2017) and thus, highlighting the inclusion of mediating functions of cognition and affect (in this study). Evidence indicates that multisensory stimuli of enjoyable landscape features are likely to enable people to generate a positive cognitive assessment (Liu et al. 2023), thereby significantly boosting their emotional state and satisfaction with CESs.

Implications for facilitating CESs

The psychological mechanism underlying the pathways of CES delivery further implies how CESs can be sustained. Although the results showed that landscape sensation did not have a direct link to satisfaction, we found a positive indirect correlation between landscape sensation and satisfaction. Therefore, people who perceive a high-quality sensory environment are more likely to be satisfied than those who do not. Research from different disciplines has proposed what constitutes a high-quality sensory environment. In psychology, the notion of ‘soft fascination’ suggests that beneficial stimuli for people should allow them to relax and provide an opportunity to restore cognitive capacity (Kaplan 1995). Given that nature offers a multisensory experience (Franco et al. 2017), studies have proposed targeted measures from the perspectives of multiple senses. The visual experience has garnered the most attention (Hartig et al. 2003). For example, Tveit et al. (2006) identified nine key visual concepts for visual quality, such as stewardship, coherence, disturbance, historicity, visual scale, imageability, complexity, naturalness, and ephemera. Natural sound contributed to restorative experience, while the technological sounds have been found to be disturbing (Cerwén et al. 2016). Additionally, specific smells from woodlands (e.g. earthy) and garden (e.g. citrus) have the potential to evoke feelings of peace and calmness (Pálsdóttir et al. 2021). Tactile contact with sunlight has also been proven to have beneficial effects on affect, especially for those who are in greater need of recovery (Beute and De Kort 2018). This study emphasizes that sensory stimuli from nature is the starting point of CES delivery and forms the foundation for intangible benefits.

The hypothesis validation indicated that cognition and affect contributed significantly to satisfaction, with cognition having a greater impact than affect. Previous studies have pointed out the importance of cognition in relation to CESs. For instance, Russell et al. (2013) indicated that cognition constitutes a crucial channel to attaining nonmaterial well-being. Teff‐Seker and Orenstein (2019) showed that people obtain diverse CESs through complex and multi-level cognitive experiences. Our findings corroborate previous observations. These observations suggest that affirmative cognitive experiences can potentially help people obtaining CESs (Huang et al. 2023). Furthermore, past psychological research has provided various cognition-based interventions, such as interpretive signs and boards (Ablett and Dyer 2009), educational guided tours and explanations (Ritchie et al. 2015; Huang et al. 2022), and environmental education programs (Ballantyne et al. 2018). Our findings also emphasize the indirect relationship between landscape sensation and affect. Existing psychological research may offer valuable insights for affect-based interventions aimed at promoting CESs. For example, Matlin and Foley (1992) found that auditory features can easily cause fluctuations in people’s affect, and Gordon et al. (2013) discovered that tactile stimulation leads to the release of chemicals that are beneficial to promote affect, such as endorphins and dopamine. Thus, we encourage future work to explore additional measures to improve CES quality from the perspective of cognition and affect enhancement.

Furthermore, this study presents a potential indicator system for cognitive experience within the context of natural attractions, encompassing eight indicators (aesthetic value, cultural diversity, uniqueness, weather, environmental quality, sense of comfort, cleanliness, and sense of safety). As different ecosystems provide varied CESs through divergent cognitive experiences (Nowak-Olejnik et al. 2022), we employed social media methods to identify critical cognitive experience specific to natural attractions. These eight indicators we extracted were largely aligned with the CES themes derived by Teff-Seker et al. (2022). However, some common benefits of interacting with nature were not identified, such as sense of place. One possible explanation is that both Teff-Seker et al. (2022) and this study focus on the short-term interactions with nature, such as tourism experience and nature walks. Some cognitive benefits of interacting with nature, such as sense of place and environmental familiarity, are more likely to be generated in human habitats rather than recreational areas due to difference in exposure duration (Berman et al. 2008). The indicator system from this study is valid in the context of natural attractions. We also encourage future studies to ameliorate CESs according to the particular background of their study areas, taking the type and duration of human-nature interactions into consideration (Yan et al. 2024). This study also underscores the potential of the eight indicators applying to CES’s enhancement.

Limitations and recommendations

We acknowledge that our research has some limitations. The first limitation is the subjectivity issues of the survey methods. As the concepts under investigation are difficult to research, we have used a questionnaire survey to collect self-reported data. We have done our best to minimize biases, such as accurately conveying the meaning by explaining concepts thoroughly to participants, and randomly sampling individuals rather than designating a specific site to represent broader CES users. However, some biases were inevitably neglected, such as interpretive biases caused by respondents’ incomplete understanding of the concepts and the lack of representativeness. Additionally, the social media data also has a limitation of representativeness because not all of society uses social media equally. Therefore, we suggest that future research consider incorporating other methods to solve some of these issues, such as stratified sampling.

On the other hand, the broader conclusions of this study may be biased owing to the fact that only natural attraction was selected as the representative background of CESs. In other words, natural attraction only delineates a subset of CESs and may not be readily applied to CESs as a whole. It is also worth noting that the concept of CESs is a subject of ongoing debate. Definitions of CESs vary among academic fields but generally concurred with the prevailing logic of MEA: CESs refers to the ‘intangible’ or ‘nonmaterial’ benefits people obtain from ecosystems (MEA 2005). It is somehow unrealistic to find a topic that can completely represent CESs. Nevertheless, by taking natural attractions as a representative CESs topic, this study showcases a paradigm to study the interaction mechanism between sensation and CESs. Additional research on CESs should explicitly explore how multisensory stimuli yield valuable information for a broad range of CES topics, including educational values, cultural heritage, and green space utilization.

Conclusions

This study has shed light on the underlying mechanism of how landscape sensation proceeds to CES outcomes. Our findings showcase an interdisciplinary framework encompassing sensation, cognition, affect, and satisfaction. Eight novel indictors were developed using social media data analytics for cognition measurement. The SEM results revealed that landscape sensory features were capable of generating satisfaction, where CESs were acquired, through the mediating routes of (i) cognition to satisfaction or (ii) cognition, followed by affect, to satisfaction. These findings enrich our understanding of human well-being initiated by sensory stimuli of landscape elements and provide valuable information for boosting CESs from landscapes.