Introduction

Natural world heritage sites represent some of the earth’s most valuable natural assets, and their protection has evolved into a significant global initiative. Established in 1972, the World Heritage Convention is a landmark achievement in heritage conservation1. To qualify for inclusion on the World Heritage List, a natural site must meet at least one of the criteria outlined in Paragraphs vii–x of the Operational Guidelines, which include natural aesthetic value (vii), geomorphic features (viii), ecosystems (ix), and natural habitats conservation (x), along with fulfilling authenticity and integrity requirements2. Only sites that meet these criteria are recognized as possessing Outstanding Universal Value (OUV).

Despite their significance, the global conservation status of natural world heritage sites is precarious due to threats from climate change, earthquakes, wars, and population pressures3. Research indicates that 60% of these sites are vulnerable to geological hazards, including earthquakes, landslides, and tsunamis4. Restoring the damaged landscapes and ecosystems of natural heritage sites poses considerable challenges, necessitating urgent protective measures. For instance, the Ms 7.0 earthquake in 2017 severely impacted Sparkling Lake in Jiuzhaigou, a World Natural Heritage Site in China, resulting in dam collapses, vegetation loss, and lake desiccation5,6. Subsequent disasters, such as heavy rainfall or further earthquakes, could worsen downstream dam failures, endangering Jiuzhaigou’s heritage landscape. In response, local authorities initiated conservation and restoration projects with targeted artificial interventions to facilitate phased recovery7. As the first case of world natural heritage landscape restoration, the restoration work of Sparkling Lake has important scientific research value and reference significance. Similar challenges may arise globally, yet critical questions remain unanswered: How do earthquakes influence the aesthetic value and OUV of world natural heritage? Can artificial restoration enhance these values and restore the authenticity and integrity of natural landscapes? What should be prioritized in post-disaster restoration?

Landscape Aesthetic Value (LAV), a critical criterion in evaluating natural world heritage sites, is defined by the Operational Guidelines as areas of exceptional natural beauty or aesthetic significance. Current LAV assessments incorporate both subjective methods (e.g., psychological preference surveys) and objective approaches (e.g., physiological index testing). Traditional research often relies on static images to document and convey landscape changes. As early as 1977, Shafer employed photo experiments to measure the relationship between different natural landscape elements and public preferences8. Recent advancements have further elucidated aesthetic preferences9, including the characteristics of aesthetic audiences10 and correlation analyses11. However, studies focusing on aesthetic preferences specific to world natural heritage sites are limited. Some scholars have utilized methodologies such as UNESCO’s global framework approach and questionnaire surveys to evaluate landscape aesthetics12,13,14. While these methods effectively reveal participants’ subjective preferences, they are susceptible to various influencing factors, including respondent disparities and idealized perceptions10. This has prompted the exploration of physiological index testing, such as eye-tracking, to gain more objective insights into aesthetic values15.

By conducting joint eye-tracking experiments for evaluating landscape supplementation, this technology effectively quantifies visual attention patterns, such as the direction of saccades, duration of fixations, and areas visited16,17,18,19. Saliency maps derived from eye-tracking data reliably highlight focal areas within landscapes20,21. This method offers objective insights into the relationship between aesthetic preferences and gaze patterns by capturing participants’ visual attention and fixation behaviour, thus integrating human perception with behavioural actions. Similarly, the aesthetic value of natural landscapes contributes to the inherent heritage value of world natural heritage sites, which arises from the interplay between biophysical landscape characteristics and human perception22. However, research addressing whether LAV evaluations can measure damage or restoration in world natural heritage landscapes remains limited. Nonetheless, combining aesthetic preference questionnaires with eye-tracking technology has been empirically validated as an effective methodology for assessing architectural heritage23.

This study focuses on Sparkling Lake in Jiuzhaigou, a natural world heritage site severely impacted by the 2017 earthquake. Restoration efforts commenced in early 2019 and concluded in 2021. We evaluate aesthetic preferences and the world natural heritage value of Sparkling Lake across pre-earthquake (2015), post-earthquake and pre-restoration (2017), early restoration (2019), and late restoration (2021) stages. The primary research questions include: (1) How did aesthetic preferences and the world natural heritage value shift before and after the earthquake and restoration? (2) What relationships exist between aesthetic preferences and the world natural heritage value? (3) Which landscape elements are most favoured by people, influencing their aesthetic preferences? By analyzing the LAV of this world natural heritage site, the study assesses whether restoration efforts successfully reinstated Sparkling Lake’s exceptional natural beauty and aesthetic significance, providing actionable recommendations for future conservation and management of world natural heritage sites.

Methods

Study area

Jiuzhaigou, located in Jiuzhaigou County, Aba Tibetan and Qiang Autonomous Prefecture, Sichuan Province, China, is recognized as a World Natural Heritage site. It features a remarkable natural environment, emphasizing the conservation of travertine lakes, shoal flows, waterfall landscapes, karst water systems, and forest ecosystems as primary protective objectives24. In 1992, the site was inscribed on the World Heritage List by the United Nations Educational, Scientific and Cultural Organization (UNESCO) for its exceptional natural phenomena and rare beauty, fulfilling the seventh selection criterion for world heritage.

Within Jiuzhaigou lies Sparkling Lake (geographic coordinates: 103°54′1″E, 33°12′13″N), situated in Shuzheng Valley, one of Jiuzhaigou’s three main valleys. At an elevation of 2211 m, Sparkling Lake spans 294 m in length, 232 m in width, and reaches a depth of 16 m, with a water storage capacity of 45 × 104 m3. It ranks among the principal lakes in Jiuzhaigou.

On August 8, 2017, an Ms 7.0 earthquake struck the area, causing a collapse of the dam at the outlet of Sparkling Lake (Fig. 1). This resulted in a gap measuring 40 m long, 15 m wide on the east side, 12 m wide on the west side, and approximately 13–15 m deep. Consequently, the lake dried up and the surrounding landscape suffered significant damage. Restoration efforts for Sparkling Lake commenced in 2019.

Fig. 1
Fig. 1
Full size image

Location of the study site.

Following thorough deliberation and expert assessments, it was determined that restoring Sparkling Lake was not only feasible but essential for preserving Jiuzhaigou’s unique natural heritage. A comprehensive restoration plan was developed, focusing on reconstructing the collapsed dam, replenishing the lake’s water, and rehabilitating the surrounding ecosystem to its former state7. This meticulous approach aimed to harmonize restoration efforts with the natural environment, thereby safeguarding the area’s aesthetic and ecological integrity for future generations. The restoration was completed in 2021 (Fig. 2).

Fig. 2
Fig. 2
Full size image

Restoration of Sparkling Lake.

Participants

This study selected college students as experimental participants. Previous research indicates that students typically possess strong aesthetic abilities and diverse professional backgrounds. By choosing students for the eye-tracking experiment, we aimed to encompass a wide range of preferences and aesthetic perspectives, making the selection both feasible and representative25,26. Participants were required to have uncorrected or corrected visual acuity of 1.0 or above, normal colour vision, and no visual impairments such as colour blindness, strabismus, amblyopia, or astigmatism. After distributing volunteer recruitment advertisements on campus, we randomly selected 75 college students to participate in the experiment.

Photographic stimuli

Research on landscape visual quality assessment and eye-tracking has demonstrated that landscape photos offer ease of operation, strong experimental controllability, and similar effects to viewing actual landscapes19,27,28,29. This study utilized photographs from four pivotal stages of landscape change: the pre-earthquake (2015), post-earthquake and pre-restoration (2017), early restoration (2019), and late restoration (2021) stage. To enhance for participants’ understanding of Sparkling Lake’s landscape, we selected one aerial viewpoint to capture aerial views and four tourist viewpoints to acquire landscape photographs as research stimuli (Fig. 3). And the AOI was defined as the entire displayed landscape photographs. Sparkling Lake features two primary tourist routes. After visiting Lying Dragon Lake, visitors proceed on foot to Sparkling Lake and then continue to Double-dragon Lake Bus Station via either Tourist Route 1 or Tourist Route 2 (Fig. 4). Consequently, the aerial and tourist viewpoint locations were strategically selected along these two designated tourist routes.

Fig. 3
Fig. 3
Full size image

Photographic stimuli.

Fig. 4
Fig. 4
Full size image

Tourist routes.

The aerial views were sourced from three-dimensional maps in Google Earth (GE), with a resolution of 8192 × 4320 pixels (8 K UHD). To mitigate significant colour variances and the impact of cloud interference across different periods, all images were adjusted for brightness and colour balance using Adobe Photoshop CC630. The aerial views incorporated key landscape elements including geomorphic features, vegetation, water and some other landscape elements.

The landscape photographs from four tourist viewpoints were obtained through author-conducted field photography. Each landscape photographs set was carefully standardized to maintain similar shooting angles or landscape elements. The first group of landscape photos was taken at the upstream viewpoint of Sparkling Lake, with the core landscape elements being the lake, embankment, and vegetation. The second group was captured at the entrance viewpoint of Sparkling Lake, featuring the embankment, vegetation, and lake as the primary elements. The third group was photographed at the downstream viewpoint, highlighting the travertine dam and terraced waterfalls. The fourth group was taken at the dam crest viewpoint, with the key elements being vegetation and the boardwalk. Due to the impacts of the earthquake, it was impossible to collect photographs taken from identical angles across all stages. Consequently, two photographs exhibit partial discrepancies.

  • First substitution: No pre-earthquake image of the travertine dam was available, so a photo of the remnant dam on the western side of Sparkling Lake was used as a substitute.

  • Second substitution: No ground-level photo from the upstream viewpoint during the early stage of landscape restoration could be located, prompting the use of a drone-captured image from the similar perspective.

Although minor perspective differences exist, both images retain the same core landscape elements and effectively document the transitional changes in Sparkling Lake across different stages.

Before the earthquake, the entire tourist area had intact terrain and abundant vegetation. Numerous lakes were separated by travertine embankments, forming diverse travertine dams and terraced waterfall landscapes. After the earthquake, the Sparkling Lake dam collapsed, vegetation was washed away, embankment landscapes were damaged, and the lake and terraced waterfall scenes disappeared, leaving the travertine landforms exposed. In the early stage of landscape restoration, artificial intervention reconstructed the dam of Sparkling Lake, providing a growth environment for plants. At this stage, pioneer plant species were introduced for gradual vegetation restoration, and the lake began to refill with water. The embankment’s form became continuous, yet it remained exposed to the atmosphere without being submerged underwater. In the late stage of landscape restoration, the lake was fully replenished, and the embankment landscape was restored, now accompanied by various vegetation. The terraced waterfall landscape reappeared with abundant water flow. Additionally, the boardwalk was successfully reconstructed (Table 1).

Table 1 Characteristics and differences of stimulus variables
Full size table

Questionnaire

The questionnaire consists of three sections: participant information, evaluation of aesthetic preferences, and assessment of world natural heritage value. The first section gathers basic participant information, including name, gender, age, education level, previous visits to Jiuzhaigou, and whether they have landscape-related expertise. The second section seeks to measure participants’ aesthetic perceptions of the presented photographs through a straightforward question: “Does this photo give you a sense of beauty?” The final section evaluates the world natural heritage value of the site. According to the Operational Guidelines for the Implementation of the World Heritage Convention, criteria for assessing OUV are defined in Articles 77 (vii)\(-\)(x), 78, 86, and 90–952. Based on these criteria, we identified seven influencing factors related to world natural heritage value: natural beauty, geomorphic features, vegetation quantity, water landscape, natural habitat conservation, landscape integrity, and landscape authenticity. Responses for both aesthetic preference and world natural heritage value evaluations utilized a 7-point Likert scale, with 1 indicating “strongly disagree” and 7 indicating “strongly agree.”

Experimental procedure

The experiment was conducted in the laboratory of the College of Landscape Architecture at Sichuan Agricultural University from April 10th to April 15th, 2023. This laboratory is a quiet and independent space, with indoor lighting controlled at 300–400 Lux, and the temperature and relative humidity maintained at approximately 21 °C and 55%, respectively30.

Before the experiment, each participant received a thorough briefing on its purpose and procedures. Participants were required to sign a written informed consent form to confirm their awareness and agreement to the terms of participation. The eye-tracking experiment involved participants entering the lab sequentially and wearing the Tobii Pro Glasses 2 eye-tracking device (https://www.tobii.cn/products, Tobii Pro, Stockholm, Sweden) (Fig. 5). Participants were instructed to keep their heads still during the experiment and were prohibited from wearing frame glasses, false eyelashes, or eyeliner. A laptop connected to the Tobii Pro Glasses 2 was used to collect real-time eye-tracking data via the Tobii Pro Controller software, capturing 20 types of data, including visit, fixation, and glance data. An EPSON CB-X06E projector, with a resolution of 1024 × 768 dpi, displayed the photographic stimuli on a 140 × 190 cm projection screen.

Fig. 5: Experimental procedure.
Fig. 5: Experimental procedure.
Full size image

a Explain the experimental procedure. b Participants wear eye-tracking devices and calibration equipment. c Play eye-tracking experiment slides. d Participants close their eyes to rest and relax. e Watch the slides and fill out the questionnaire. f Check and save experimental data for analysis.

Participants began with a pre-experiment phase, observing two interfering photos to familiarize themselves with the process, followed by the main experiment where they viewed 20 experimental photographs and 19 blank images. Each photo was displayed for 15 s, and blank images for 5 s, with the presentation order randomized. Participants were free to observe the photos without specific tasks. The total duration of the eye-tracking slide presentation was 470 s. A 2-min rest was provided after the eye-tracking session to alleviate visual fatigue. Subsequently, the staff presented the evaluation questionnaire using a computer, and participants filled it out. Aesthetic preference evaluations lasted 15 s per aerial view, and world natural heritage value evaluations were allotted 30 s. Only aesthetic preferences for the landscape photos from different perspectives were evaluated for 15 s each. The total duration for the questionnaire presentation was 445 s. Finally, staff checked and saved the experimental data. The entire experimental procedure was completed within 30 min.

Data analysis

A total of 75 initial data sets were collected. The Tobii Pro Controller software filtered and discarded data with eye-tracking sampling rates below 80%31. Out of these, 14 data sets were deemed unqualified, leaving 61 valid experimental sets. In landscape evaluation studies, having more than 30 sets of experimental data constitutes valid data32,33, thus meeting our requirements. The 61 valid eye-tracking data sets were imported into Tobii Pro Lab (version 1.194.41215, TOBII, Sweden) software, which exported the eye-tracking indicators as Excel files and generated visualized heat maps.

Data were organized using Microsoft Excel 2023, whereas variance, regression, and correlation analyses were conducted using IBM SPSS Statistics 25 software. Normality was tested using the Shapiro-Wilk test with a significance level of p = 0.05. Homogeneity of variance test was examined using Levene’s test, setting the significance level to 0.05. Then differences among different temporal scenes were analyzed by one-way ANOVA. Stepwise multiple linear regression was performed to establish predictive models between seven influencing factors of world natural heritage value and aesthetic preferences. Pearson bivariate correlation analysis was conducted to examine relationships between aesthetic preferences and eye-tracking indicators.

All measured values are presented as mean ± standard error. The 61 valid datasets comprised 26 males and 35 females (a ratio of approximately 1:1.35), with a landscape professional to non-landscape professional ratio of 1:1.77. The reliability of the questionnaire was assessed using Cronbach’s alpha coefficient34, resulting in an alpha coefficient of 0.909 for aesthetic preferences and 0.926 for the influencing factors of world natural heritage value, indicating high data reliability.

Results

Aesthetic preference and world natural heritage value of Sparkling Lake

The analysis of aesthetic preferences for Sparkling Lake across different periods reveals that the aesthetic preference score in the post-earthquake stage (2017) is significantly lower than in the pre-earthquake stage (2015) (p < 0.01). In the early stage of landscape restoration (2019), there is no significant change in aesthetic preference scores. However, during the late stage of landscape restoration (2021), the aesthetic preference score improved significantly and was significantly higher than that in the pre-earthquake stage (p < 0.05) (Fig. 6a).

Fig. 6: Aesthetic preferences and the world natural heritage value.
Fig. 6: Aesthetic preferences and the world natural heritage value.
Full size image

a Score of aesthetic preferences at different times. b Score of the world natural heritage value at different times. c Relationship between aesthetic preferences and world natural heritage value. Note: Within the same group, different lowercase letters indicate significant differences (p < 0.05), and different uppercase letters indicate extremely significant differences (p < 0.01).

Similarly, the evaluation of world natural heritage value mirrors these trends (Fig. 6b). The world natural heritage value in the pre-earthquake stage (2015) and the late stage of landscape restoration (2021) is significantly higher than in both the post-earthquake stage (2017) and the early stage of restoration (2019) (p < 0.01). Following the initiation of artificial interventions in 2019, the world natural heritage value significantly increased compared to that in the post-earthquake period (p < 0.05). In the late stage of restoration (2021), the world natural heritage value continues to rise, with no significant difference compared with the pre-earthquake period (2015).

The Pearson correlation coefficient between aesthetic preferences and world natural heritage value is calculated at 0.967, which is statistically significant at the 0.05 level, indicating a strong positive correlation between these two variables.

Influencing factors of the world natural heritage value

This study evaluates seven influencing factors on world natural heritage value: natural beauty, geomorphic features, vegetation quantity, water landscape, natural habitat conservation, landscape integrity, and landscape authenticity (Fig. 7). Assessment results for these factors reveal a consistent pattern, indicating no significant differences between the late stage of landscape restoration (2021) and the pre-earthquake stage (2015) (p > 0.05). Both stages register significantly higher scores than those in the early stage of restoration (2019) and the post-earthquake stage (2017) (p < 0.01).

Fig. 7: Score of influencing factors of the world natural heritage value.
Fig. 7: Score of influencing factors of the world natural heritage value.
Full size image

Within the same group, different lowercase letters indicate significant differences (p < 0.01).

Comparatively, the evaluation of the post-earthquake stage (2017) and the early stage of landscape restoration (2019) shows that six influencing factors (natural beauty, geomorphic features, vegetation quantity, water landscape, natural habitat conservation, and landscape integrity) have significantly higher scores in the early stage of restoration (2019) than in the post-earthquake stage (2017) (p < 0.01). However, no significant difference is observed in landscape authenticity scores between the early stage of restoration (2019) and the post-earthquake stage (2017) (p > 0.05).

The relationship between aesthetic preferences and influencing factors of the world heritage value

This study conducted a correlation analysis of aesthetic preferences against seven influencing factors of value: natural beauty, geomorphic features, vegetation quantity, water landscape, natural habitat conservation, landscape integrity, and landscape authenticity. The results, presented in Table 2, showed strong correlations, with correlation coefficient values of 0.819, 0.772, 0.619, 0.739, 0.816, 0.786, and 0.452, respectively. All correlation coefficients were greater than 0, indicating a highly significant positive correlation between aesthetic preferences and the seven influencing factors of world natural heritage value (p < 0.01). This suggests that as the scores of the seven influencing factors rise, aesthetic preference scores also increase; however, the specifics of their relationship warrant further regression analysis.

Table 2 Relationship between aesthetic preference and seven influencing factors of the world natural heritage value
Full size table

In this study, the seven influencing factors of world natural heritage value were treated as independent variables, while the corresponding aesthetic preferences served as dependent variables, leading to the establishment of a stepwise multiple linear regression model. Tests on residuals, variance analysis, and multicollinearity were conducted, yielding an F statistic of 219.284 and p = 0.000 < 0.05, indicating that the model was valid. All variance inflation factor (VIF) values were below 5, suggesting no issues with multicollinearity. The Durbin–Watson statistic was approximately 2 (ranging between 1.7 and 2.3), indicating absence of autocorrelation and confirming that the model was well constructed.

According to the regression analysis results in Table 3, the R2 value for the factors of natural beauty, geomorphic features, natural habitat conservation, and landscape integrity was 0.786, indicating that these four factors explain 78.6% of the variance in aesthetic preferences. The established regression model is as follows:

Table 3 Regression coefficient analysis of aesthetic preference and influencing factors (n = 244)
Full size table

aesthetic preference = −0.386 + 0.445 * natural beauty + 0.148 * geomorphic features + 0.287 * natural habitats conservation + 0.125 * landscape integrity

Eye-tracking results

As detailed in Table 4, the analysis of eye-tracking indicators reveals significant changes in visual engagement at Sparkling Lake across different stages. When comparing the pre-earthquake stage (2015) to subsequent periods, the average duration of visits, average duration of fixations, and average duration of glances all significantly decreased during the post-earthquake stage (2017), early stage of landscape restoration (2019), and late stage of landscape restoration (2021). Conversely, the number of fixations increased significantly during these periods.

Table 4 Comparison of eye-tracking indicators for Sparkling Lake landscape evaluation in different years
Full size table

In the post-earthquake (2017) and early landscape restoration (2019) stages, the number of visits was significantly higher compared to the pre-earthquake stage (2015). However, in the late stage of landscape restoration (2021), the number of visits decreased to a level that was not significantly different from the pre-earthquake stage (2015).

The duration of first fixation was significantly lower in the post-earthquake stage (2017) compared to that of the pre-earthquake stage (2015), but it increased in both the early (2019) and late (2021) stages of restoration, showing no significant difference relative to the pre-earthquake stage (2015).

Relative to the pre-earthquake stage (2015), the number of glances continued to rise after the earthquake, although the increase was not significant in the post-earthquake stage (2017). However, there was a significant increase in the number of glances during both the early (2019) and late stages (2021) of restoration.

As for the number of saccades, no significant differences were observed among the post-earthquake stage (2017), the early stage of landscape restoration (2019), and the late stage of restoration (2021) when compared to the pre-earthquake stage (2015). However, the number of saccades was significantly lower in both the early (2019) and late stages (2021) of restoration compared to the post-earthquake stage (2017).

Changes in features and elements of heat maps

Following the organization of eye-tracking data from 61 participants who viewed photographs of Sparkling Lake across different periods, visual focus areas were depicted in heat maps (Fig. 8). In these heat maps, red, yellow, and green areas represent varying frequencies of visual attention, with red indicating the most frequently viewed areas and green reflecting the least.

Fig. 8
Fig. 8
Full size image

Heat maps of landscape photographs at different stages.

The analysis of the colour proportions representing different visual focus areas revealed the following insights:

  1. 1.

    Pre-earthquake stage (2015): Visual focus was primarily on the boardwalks, roads, vegetation, water, and travertine landforms. In viewing photos showcasing natural scenery, the red visual focus area highlighted prominent plants, water features, and travertine landscapes. In photos that included artificial facilities, the red areas also included the boardwalks and roads.

  2. 2.

    Post-earthquake stage – before landscape restoration (2017): The visual focus shifted primarily to the travertine landforms and water landscapes damaged by the earthquake. Compared to the pre-earthquake stage (2015), the proportion of the red visual area increased, demonstrating a concentrated focus on the areas affected by the earthquake. The overall increase in the proportions of red, yellow, and green areas indicated a broader visual focus and heightened engagement with the landscape.

  3. 3.

    Early stage of landscape restoration (2019): Visual attention remained largely concentrated on the earthquake-damaged travertine landforms, restored vegetation, and some water landscapes. Similar to the post-earthquake stage (2017), there was still a high degree of visual focus on the areas impacted by the earthquake, with an increased proportion of the red visual area.

  4. 4.

    Late stage of landscape restoration (2021): By this stage, visual focus areas returned to primarily emphasize vegetation, water, and boardwalks, resembling patterns seen in the pre-earthquake stage (2015). Compared to the early stage of restoration (2019), the total proportions of red, yellow, and green areas decreased in the late stage, indicating a reduction in the overall visual focus area. Attention became more concentrated on the restored waterfalls, lakes, and vegetation. While the repaired boardwalks drew visual interest, the proportion of red colour allocated to them was only 0.3%.

Correlation between aesthetic preferences and eye-tracking indicators

A correlation analysis was performed between aesthetic preference scores from 2015 to 2021 and eye-tracking data collected during this period. The analysis revealed significant relationships between aesthetic preference scores and certain eye-tracking indicators, as summarized in Table 5. Key findings include:

  • Average duration of visit: There was a significant positive correlation with aesthetic preference scores (p = 0.029 < 0.05). This suggests that longer visits correlate with higher aesthetic preferences.

  • Number of visits: A significant negative correlation was found (p = 0.020 < 0.05), indicating that more frequent visits are associated with lower aesthetic preference scores.

  • Average duration of glances: An extremely significant positive correlation was observed (p = 0.004 < 0.01). This indicates that longer average durations of glances are associated with increased aesthetic preference scores.

Table 5 Relationship between aesthetic preference score and eye-tracking indicators for Sparkling Lake
Full size table

Discussion

This study demonstrates a positive correlation between aesthetic preferences and the world natural heritage value, aligning with findings from other research on these relationships35,36. Earthquakes significantly impact landscapes, reducing both aesthetic preferences and heritage value. However, data indicates that these values gradually recover through artificial intervention and restoration efforts. By the end of landscape restoration in 2021, the world natural heritage value was statistically indistinguishable from that of the pre-earthquake stage (2015), while aesthetic preference scores were significantly higher. This suggests that restoration efforts effectively rehabilitate natural beauty, confirming the role of artificial restoration as crucial for recovering both damage caused by natural disasters and landscape aesthetics.

Further analysis of the seven factors influencing the world natural heritage value of Sparkling Lake reveals that the earthquake severely damaged natural beauty, geomorphic features, vegetation quantity, water landscape, habitat conservation, landscape integrity, and authenticity. Nevertheless, by the late restoration stage in 2021, scores for these factors showed no significant difference compared to the pre-earthquake stage, suggesting successful restoration in the eyes of observers. It highlights that Sparkling Lake’s landscape integrity and authenticity have been effectively restored to their prior state.

It is noteworthy that while there was no significant improvement in landscape authenticity between the early stage of restoration (2019) and the post-earthquake stage (2017), there was a marked improvement by the late stage, comparable to pre-earthquake conditions. This underscores that, although early restoration efforts began improving heritage value, sustained efforts are essential for achieving comprehensive restoration.

In the context of restoring natural landscapes, especially world natural heritage sites, concerns often arise over the authenticity of restored landscapes. Some argue that restoration alters core authenticity, rendering nature artificially constructed. However, scholars contend that restoration does not inherently damage nature, and well-intentioned restoration is desirable37. Others argue that humans, as part of nature, do not act externally but integrate into natural processes; thus, artificial intervention can be a natural act38. These views, supported by our data, advocate for artificial restoration when motivated by good intentions and when benefits outweigh drawbacks. Our findings also indicate that the authenticity of the restored landscape parallels that of the original landscape, supporting restoration efforts when executed thoughtfully and purposefully.

There is a highly significant positive correlation (p < 0.01) between aesthetic preferences and the seven influencing factors of world natural heritage value. Further regression analysis indicates that natural beauty, geomorphic features, natural habitat conservation, and landscape integrity are the four primary factors influencing landscape aesthetic preferences. This suggests that individuals favour landscapes characterized by natural beauty, unique terrain and landforms, well-preserved natural habitats, and high landscape integrity when appreciating scenery. Consequently, these factors should be emphasized in the process of restoring world natural heritage landscapes. Presently, many countries and regions primarily attract tourists by highlighting the aesthetic values of heritage sites to the public. By enhancing and prioritizing natural aesthetic value, there may be a powerful impetus for environmental protection39. Restoration efforts focusing on natural beauty, geomorphic features, natural habitat conservation, and landscape integrity can foster greater awareness of the significance of protecting world natural heritage.

Heat maps serve as effective and intuitive tools for displaying visual attention areas among multiple participants40. Generally, people tend to concentrate more on the foreground and central areas of photographs41,42. In images featuring linear spatial progression, points of interest are often located near the vanishing point of the perspective43. In this study, the primary areas of visual focus and key elements during the pre-earthquake stage (2015) and the late stage of landscape restoration (2021) were nearly identical, with water, plants, and boardwalks (or roads) receiving significant visual attention. This observation aligns with previous findings that note greater attention propensity toward water and vegetation in landscape assessments44,45,46,47. Moreover, reflected images on the water surface can amplify the proportion of natural landscape elements, thereby enhancing the visual appeal of aquatic features48. Consequently, some studies identify water as the most attractive element among various landscape components14,29.

Conversely, the boardwalks also attracted considerable visual attention, potentially because these artificial elements distinctively contrast with the surrounding natural landscape48, disrupting the authenticity and integrity of the original environment. Research by Amati et al. lends support to this hypothesis49, as similar studies have shown that artificial structures such as roads and buildings can diminish the attractiveness of vegetation, adversely impacting aesthetic preferences and landscape evaluations48,50. Human visual attention is thus not drawn exclusively to beautiful elements; it is also influenced by unattractive components51,52. This finding enhances our understanding of human perception concerning landscape aesthetics.

The visual attention area during the pre-earthquake stage primarily consisted of the boardwalk and tree trunks. This observation can be attributed to the fact that the original vegetation had flourished for many years prior to the earthquake, reaching considerable heights. From a human perspective, the visible tree trunks were a significant focus. Following landscape restoration, the height of the vegetation was notably lower, allowing for greater visibility of the foliage. A larger number of green plants enhanced human visual attention, which is reflected in the reduction of the proportion of red areas in the visual attention map to just 0.3% after restoration. This increase in foliage effectively mitigated the visual impact of the boardwalk for observers. Previous studies concerning the proportions of tree trunks and leaves have shown similar preferences, indicating that images with a balanced ratio of leaves to trunks are more appealing to viewers53.

The visual focus areas of the restored landscape closely resemble those observed before the earthquake, suggesting that restoration efforts successfully achieved their intended visual objectives, alleviating the earthquake’s impact on the landscape. This aligns with findings by Liu and Neisch, which indicate that increased green spaces enhance visual appeal and overall visual quality54.

In contrast, the areas of visual attention during the post-earthquake stage (2017) and early stage of landscape restoration (2019) were higher than those in the pre-earthquake stage (2015) and late stage of restoration (2021). Visual hotspots were concentrated on the fractured travertine dam, protruding travertine formations, vegetation on exposed slopes, and desiccated waterfalls. This phenomenon suggests that damaged areas from earthquakes, as well as textures indicating significant alterations from restoration efforts, tend to attract more visual attention than harmonious and intact natural landscapes. In the early restoration stage, even though some vegetation had been revived, Sparkling Lake remained water-stressed and revealed exposed travertine deposits, creating stark visual contrasts with the surrounding environment. The prominent seismic damage to the travertine deposits left a lasting impression, thus prioritizing visual attention still on these affected areas.

The findings of this study can inform future natural landscape restoration designs, highlighting the importance of prioritizing attractive water features and vegetation during restoration efforts. Attention should also be given to minimizing the visual impact of artificial elements within the landscape. In preserving natural landscapes, considerations about the integrity of the landscape must be key factors, as artificial components can compromise this integrity. Ensuring that artificial elements harmonize with natural features such as lakes, mountains, and forests is crucial for maintaining the aesthetic value of these landscapes. Conversely, another study emphasizes that when designing forest landscapes, integrating iconic or whimsical elements can enhance visual attraction19. Thus, in future preservation of natural heritage landscapes, it is advisable to minimize the construction of artificial facilities or implement measures to ensure visual integration of these elements, such as utilizing materials that match the surrounding environmental colours.

The average duration of visits and glances was highest during the pre-earthquake stage (2015), with significant differences compared to other periods. This may be attributed to the natural environment’s ability to attract involuntary attention, often referred to as “fascination.” This intrinsic quality of the environment can draw attention effortlessly55,56,57, explaining why participants spent more time visiting and glancing at landscapes in the pre-earthquake stage (2015).

Natural environments, rich in fascinating elements such as water, vegetation, and rocks, tend to generate greater interest and receive more favourable aesthetic evaluations58,59. Other studies have similarly shown that individuals spend more time engaging with preferred natural landscapes, indicating higher aesthetic preferences60.

However, after viewing photos of the same location repeatedly and familiarizing themselves with the site, people tend to focus on observable differences over time, such as the area affected by earthquake damage. This inclination is reflected in the significantly higher number of fixations during the post-earthquake stage (2017) and early stage of landscape restoration (2019) compared to the pre-earthquake stage (2015). This may relate to perceptual fluency, as previous research suggests that visual systems process natural scenes more smoothly than urban environments61. The stark contrast presented by landscapes damaged by earthquakes and those in early restoration stages versus the original natural landscape can disrupt perceptual fluency, likely leading to multiple repeated fixations62. While the number of visits increased, aesthetic preference scores remained relatively low. During both the pre-earthquake stage (2015) and later stage of restoration (2021), there was a tendency toward harmonious natural landscapes characterized by high perceptual fluency and fewer fixations. Environmental psychology research indicates that less eye movement occurs when viewing aesthetically pleasing natural landscapes63,64,65. Future research could explore the landscape perception process more thoroughly by analyzing eye movement trajectories in greater detail.

Several crucial limitations are associated with this study. First, previous research has indicated that participant characteristics, such as gender, age, education level, occupation, and living environment, influence landscape evaluations66,67. The current study primarily involved students, predominantly young individuals with higher education levels, which limits the generalizability of the findings. Second, this study uses photographs as visual surrogates for landscape evaluation, introducing inherent limitations in photo selection and scene simulation. The extended temporal scope and multiple observational perspectives significantly increases the difficulty of photo collection. The abrupt nature of the earthquake makes it particularly challenging to obtain temporally matched stimuli—photographs taken from identical viewpoints and positions across different periods at multiple viewpoints before and after the seismic event. Although prior research has demonstrated that variations in viewing angles, backgrounds, and distances can affect visual behaviour, potentially introducing aesthetic deviations when viewing scenes composed differently63. This study mitigates these effects by selecting photos from the same viewpoints, controlling for the same core landscape elements, and simulate tourist viewing perspectives to reduce compositional discrepancies that could affect evaluation outcomes. Similarly, in situ experiments, dynamic landscapes can be affected by natural changes or human factors and thus have an impact on the landscape evaluation. Therefore, perceptions and valuation of a landscape are performed by analyzing changes of composition or structure in the scenary in the same landscape29. Furthermore, future research should more rigorously control specific visual elements within the scenes evaluated before and after restoration to obtain more precise and insightful data regarding restoration assessments. It is hoped that these evaluations of restored landscapes will continue to inform and inspire concrete actions aimed at protecting natural world heritage sites and enhancing their contributions to human well-being.

In summary, natural disasters pose significant threats to world natural heritage sites, necessitating timely and effective restoration interventions. When restoration efforts are well intended and produce favourable outcomes that outweigh negative impacts, artificial intervention becomes an urgent necessity. This study evaluated the landscapes of Sparkling Lake across different periods and demonstrated that landscape restoration can successfully restore the authenticity and integrity of world natural heritage from a visual perception perspective. However, achieving these outcomes relies on long-term, sustained restoration efforts. Emphasizing factors such as natural beauty, geomorphic features, natural habitat conservation, and landscape integrity is crucial, as these elements significantly impact aesthetic preferences. Measures should also be implemented to mitigate the effects of artificial elements that disrupt landscape integrity and visual aesthetics.

This research represents a pioneering evaluation of artificial restoration applied to a world natural heritage landscape, equipping landscape designers and heritage management units with a deeper understanding of post-damage conditions and critical restoration issues. It establishes effective methodologies for the evaluation, restoration, and protection of natural world heritage landscapes. The findings highlight the necessity of considering ecological and aesthetic factors during the restoration of damaged natural landscapes and stress the importance of continuous monitoring and adaptive management to ensure the sustainability of restoration efforts. Ultimately, this study addresses existing research limitations that struggle to quantify restoration effects in natural world heritage sites, fills gaps in the understanding of aesthetic values in previous restoration cases, and serves as a reference for future restoration, protection, and post-management activities related to natural world heritage sites.