The influence of visual and auditory environments in parks on visitors' landscape preference, emotional state, and perceived restorativeness

Rapid urbanization diminishes natural spaces and contributes to stress-related health issues, creating demand for restorative urban environments such as parks. Theoretical foundations (Stress Recovery Theory, Attention Restoration Theory) posit that natural environments alleviate stress and support cognitive and emotional restoration. Landscape preference (LP) and emotional state (ES) are key drivers of perceived restorative benefits (PRS), yet prior research often isolates visual or auditory factors and rarely models their combined pathways. Because environmental perception is multisensory, integrating visual and auditory stimuli is essential to understand the mechanisms by which parks support restoration. This study addresses gaps by modeling the relationships among preference, emotion, and restoration in complex audiovisual settings and examining how specific landscape features and sound types influence and interact to affect LP, ES, and PRS.

- Audio-visual interaction and landscape preference: People tend to prefer nature-related elements (water bodies, vegetation). Auditory cues can strongly shape preferences; in some contexts sound exerts stronger influence than vision. Coherence between seen landscapes and heard soundscapes (e.g., birdsong, insects, water; soft music/temple bells) enhances immersion and preference; better sound quality in green spaces increases popularity. - Audio-visual interaction and emotional state: Visual green/blue spaces typically promote positive emotions; certain visual characteristics (open, less crowded scenes) reduce stress. Natural sounds (e.g., birdsong, flowing water) reduce tension, anxiety, and pain and support autonomic regulation. Emotions shape perceived soundscape quality, and acoustics can more strongly affect ES than visuals in some contexts. Visual-auditory combinations can modulate emotional processing (e.g., vegetation enhancing noise reduction via emotional processes). - Audio-visual environment and perceived restorativeness: Natural visual and acoustic environments improve physiological and psychological recovery (lower stress, better attention, benefits to heart rate/blood pressure/HRV). High-quality soundscapes are key to restorative environments; some natural sounds can mask or reduce negative perceptions of target noise. Interactions between visuals and sounds (e.g., water and vegetation with birdsong) can enhance restorative potential; greater diversity/complexity may aid restoration. - Research objectives: Develop a moderated mediation model to: (1) test ES as mediator between LP and PRS; (2) assess direct impacts of visual and auditory elements on LP, ES, PRS; (3) test moderating roles of sound source types and landscape features on ES and PRS.

Study area: Five popular, publicly accessible parks in Hangzhou, China (Xiaoqiao Creek Ecological Park; North City Sports Park; Riverside Park; Xixi National Wetland Park; Hangzhou Forest Park) representing diverse audiovisual settings. In each park, 10–13 locations were selected. Sampling and survey: On-site, face-to-face paper surveys were conducted from Sept 20 to Nov 5, 2023, 9:00–18:00 on weekdays and weekends, after peak tourist season. Inclusion criteria: normal (or corrected) vision and hearing, no psychological disorders, and at least 5 minutes in the target area. Of 922 questionnaires, 861 valid responses were obtained (93% response rate). Demographics included gender, age, education, occupation, park visit frequency, and visit duration. Audiovisual data collection: - Visual: Panoramic 360° images captured at 1.6 m height using Insta360 One (8K, 6080×3040). DeepLab v3+ was used to segment and quantify proportions of five visual categories: trees, water, sky, ground vegetation, buildings. - Auditory (subjective): Based on the Swedish Soundscape Quality Protocol, respondents rated perceived presence (1–5 Likert; 1 = not heard, 5 = dominates) of nine sound sources grouped as: Nature (birdsong, wind, flowing water, insects); Human (conversation, children playing, footsteps); Mechanical (traffic, construction). - Auditory (objective): Environmental acoustics measured per ISO/TS 12913-2:2018 using HS6298 analyzer: A-weighted equivalent sound pressure, sound exposure level (Lae), and percentile levels (LA95, LA90, LA50, LA10, LA5). Psychoacoustic recordings (SQope binaural headset) analyzed per ISO 2017 to derive loudness, sharpness, roughness, and fluctuation strength. The Environmental Acoustic Index (EAI) used in SEM measurement comprised loudness, Lae, and LA95. Measures: - Emotional State (ES): Ten positive items adapted from STAI SAI (calm, secure, comfortable, content, self-confident, relaxed, steady, satisfied, at ease, pleasant), 6-point scale (1–6), higher scores indicate more positive mood/lower state anxiety. - Perceived Restorativeness (PRS): 20 items covering Fascination, Being-Away, Compatibility, Extent (1–7 agreement scale). - Landscape Preference (LP): Six items based on Kaplan et al. with additions for landscape quality and willingness to experience (1–5 agreement scale). Data analysis: - Preliminary: Normality tests indicated non-normal distributions; Spearman correlations between audiovisual attributes and LP, ES, PRS; Pearson correlations for demographics. PCA with varimax rotation reduced acoustic quality dimensionality. - PLS-SEM: Due to non-normal data and model complexity, PLS-SEM was used. Measurement model reliability evaluated via Cronbach’s alpha, composite reliability, rho_A; convergent validity via factor loadings and AVE; discriminant validity via Fornell-Larcker, cross-loadings, and HTMT. PRS modeled as a second-order reflective construct using latent scores of its four dimensions. VIF assessed multicollinearity; bootstrapping (5,000 resamples) tested significance. Moderation analyses used bias-corrected 97.5% CIs.

Sample and preliminary results: - N = 861; response rate 93%. - Spearman correlations: Birdsong and flowing water positively associated with LP, ES, PRS; wind and footsteps showed no correlation with ES/PRS; insect sounds negatively correlated with ES; conversation, children playing, traffic, construction negatively correlated with LP, ES, PRS. Tree, ground vegetation, and water views positively correlated with LP, ES, PRS; buildings and sky views negatively correlated; ground view not correlated. Age positively correlated with LP, ES, PRS. Measurement model: - First-order constructs showed acceptable reliability and validity: standardized loadings 0.599–0.923; AVE 0.504–0.731; Cronbach’s alpha and rho_A > 0.70. - PRS second-order construct: loadings 0.891–0.949; AVE 0.858; Cronbach’s alpha 0.945; rho_A 0.946; VIF 3.151–4.332. Structural model: - Explained variance: R² (LP) = 0.415; R² (ES) = 0.659; R² (PRS) = 0.761. - Direct effects on PRS: LP (β = 0.130, p < 0.001) and ES (β = 0.192, p < 0.001) positively predicted PRS; EAI negatively predicted PRS (β = -0.085, p < 0.001). Age had a small positive effect (β = 0.038, p = 0.037). - Visual predictors: Tree (β to PRS = 0.224, p < 0.001), ground vegetation (β = 0.163, p < 0.001), and water views (β = 0.122, p < 0.001) positively influenced PRS (and were also positive for LP and ES). Buildings view negatively influenced PRS (β = -0.076, p = 0.010) but was not significant for LP or ES; sky view effects on PRS and LP were not significant; sky view negatively affected ES (β = -0.217, p < 0.001). - Sound predictors: Birdsong and flowing water sounds positively influenced LP, ES, and PRS (e.g., PRS: birdsong β = 0.267, p < 0.001; flowing water β = 0.373, p < 0.001). Children playing (PRS β = -0.096, p = 0.022), traffic (β = -0.210, p < 0.001), and construction (β = -0.098, p = 0.001) negatively affected PRS (and generally LP, ES). Conversation sounds negatively affected PRS (β = -0.100, p < 0.001) but were not significant for LP or ES. Mediation effects (selected, p < 0.05): - ES mediates LP → PRS (indirect positive effect). Trees, ground vegetation, and water views increased PRS via ES (e.g., tree view → ES → PRS β = 0.033, p = 0.042) and via LP then ES/PRS (e.g., tree view → LP → ES β = 0.035, p = 0.002; → LP → PRS β = 0.022, p = 0.010). - Sky view had a negative indirect effect on PRS via ES (β = -0.069, p < 0.001). - Birdsong and flowing water had positive indirect effects on PRS via ES and via LP (e.g., birdsong → ES → PRS β = 0.043, p < 0.001; flowing water → ES → PRS β = 0.068, p = 0.001). - Children playing negatively influenced PRS via LP (β = -0.009, p = 0.021) and ES via LP (β = -0.014, p = 0.009). Traffic negatively influenced PRS via ES (β = -0.029, p = 0.001). Conversation and construction sounds showed no significant indirect effects via LP/ES on PRS. - EAI exerted indirect effects through sound sources: poorer acoustic environments decreased birdsong contributions (EAI → birdsong → PRS β = -0.062, p < 0.001; → ES β = -0.048, p < 0.001) and increased negative effects via traffic and children’s sounds (e.g., EAI → traffic → PRS β = -0.059, p < 0.001) while supporting flowing water perception positively (EAI → flowing water → LP β = 0.109, p < 0.001; → ES β = 0.110, p < 0.001). Moderation (audio-visual interactions): - PRS: Birdsong positively moderated effects of water (β = 0.100, p < 0.001) and ground vegetation (β = 0.103, p = 0.035) on PRS, but negatively moderated sky view → PRS (β = -0.078, p = 0.002). Flowing water sounds positively moderated tree (β = 0.121, p = 0.029) and water views (β = 0.126, p < 0.001) on PRS. Conversation and construction sounds strengthened negative building view effects on PRS (conversation × buildings β = -0.038, p = 0.043; construction × buildings β = -0.062, p = 0.012). - ES: Birdsong enhanced water view → ES (β = 0.128, p = 0.005). Flowing water enhanced tree (β = 0.166, p = 0.028) and water (β = 0.069, p = 0.001) effects on ES. Conversation sounds enhanced ES with tree views (β = 0.128, p = 0.008) but reduced ES with ground vegetation (β = -0.080, p = 0.011). Traffic reduced ES with buildings (β = -0.078, p = 0.002) and water views (β = -0.087, p = 0.002). Construction reduced ES in building-dominant views (β = -0.079, p = 0.006).

The findings confirm that LP and ES are central mechanisms through which urban park environments deliver perceived restorative benefits. Natural visual elements (trees, ground vegetation, water) and natural sounds (birdsong, flowing water) consistently elicit higher preferences and more positive emotions, which in turn enhance PRS. Built visual elements and human/mechanical sounds tend to undermine these outcomes, though conversation sounds primarily reduce PRS rather than LP or ES, likely reflecting context and expectations. The results align with SRT and ART: nature’s esthetic and semantic affordances (moderate complexity, structure, presence of water) are easier to process, reduce stress, and enable attention recovery. The study extends the literature by demonstrating mediated pathways (LP → ES → PRS) and by revealing moderating audio-visual interactions: congruent combinations (e.g., birdsong with water/vegetation; flowing water with high tree/water coverage) enhance restoration and mood, whereas incongruent or urban-dominated combinations (e.g., birdsong with expansive sky-only views; traffic or construction with building views) diminish them. The observed dependency of birdsong’s benefits on acoustic quality suggests that not all natural sounds are equally restorative; properties of calls and environmental acoustics can alter appraisals from positive to potentially aversive. These results emphasize the importance of sensory coherence and context in shaping restorative experiences and provide actionable guidance for park design that integrates both visual and auditory planning.

This study proposes and tests a moderated mediation model that integrates environmental characteristics, visual features, LP, ES, and PRS in urban parks. It demonstrates that LP facilitates perceived restoration and that ES mediates the preference–restoration link. Natural views and sound types directly and indirectly bolster LP, ES, and PRS, while human and mechanical sounds generally impede them (with conversation mainly affecting PRS). Significant audio-visual interactions show that coherence between sounds and landscapes strengthens restorative outcomes. Practically, the results advise increasing natural visual elements, enhancing congruent natural soundscapes (e.g., birdsong with water/vegetation; flowing water with tree/water-rich scenes), and minimizing human/mechanical sounds in urbanized vistas. Future work should refine acoustic parameter thresholds (e.g., for birdsong), incorporate additional senses (smell, touch), and account for socio-cultural and person-level factors (e.g., noise sensitivity, place attachment) to deepen understanding of multisensory restorative mechanisms.

- Sensory scope: The study focused on vision and audition; olfaction and haptics, known to interact with hearing and influence perception, were not measured. - Acoustic parameterization: While identifying that birdsong benefits depend on acoustic quality, the study could not define optimal parameter ranges or thresholds for specific species or contexts. - Contextual/person-level factors: Potential influences such as personality traits, place attachment, cultural background, and noise sensitivity were not modeled, potentially limiting generalizability across populations and settings. - Observational field design: Although extensive, the cross-sectional field approach precludes causal inference beyond modeled pathways and may be influenced by unmeasured site characteristics. - Visual categorization granularity: Vegetation types were grouped (e.g., ground vegetation) without finer structural distinctions that may differentially affect safety perceptions and emotions.