Landscape Characteristics Influencing the Spatiotemporal Dynamics of Soundscapes in Urban Forests

Abstract

The acoustic environment of urban forests is indispensable for urban residents’ nature-based recreation opportunities and experience of green spaces, and the perceptual and physical sound features in time and space serve as determinants during this process. However, their spatiotemporal variation mechanisms and influential landscape characteristics are still underexplored in urban forests. Thus, this study aims to explore the spatiotemporal variability of perceptual and physical sound features and their relationship with landscape characteristics in urban forests. For this purpose, we measured perceptual sound features using the indicators of the sound harmonious degree (SHD) and soundscape pleasantness and eventfulness. The physical acoustic features were determined using sound-level parameters for measuring the sound level intensity (L_Aeq, L₁₀, L₉₀) and fluctuation (L_10–90). Perceptual and physical sound data collection was based on on-site questionnaire surveys and acoustic instrument measurements, respectively. The landscape characteristics were classified using the principal components of four main categories, including the terrain, area proportion of land cover types, distance to land cover types, and landscape patterns. The results showcase that significant spatiotemporal variation was found in most perceptual and physical sound features, whereas soundscape pleasantness and eventfulness did not vary significantly across time. In general, the variabilities of both perceptual and physical sound features were affected more by the types of spatial functions than by diurnal patterns. Human activities that generate sounds (e.g., hawking, playing, and exercise) may be the key drivers for spatiotemporal changes in physical acoustic features. The components of landscape patterns, including landscape structural diversity and shape complexity persistently, affected specific sound features in all periods. However, no landscape component had persistent cross-spatial influences on the sound features. This study offers critical insights into the spatiotemporal patterns of the acoustic environment and its relationship with landscape characteristics in urban forests. The findings underscore the practical importance and implications of integrating acoustic considerations into urban forest management. By providing a scientific foundation, these results can usefully inform dynamic resource management, functional zoning optimization, and sustainable landscape development in urban forests.

1. Introduction

Urban forests, as an indispensable form of public green space within urban areas, support the dwellers’ engagement with nature and contribute to people’s well-being and quality of life [1]. They perform a series of landscape and ecological functions for urban areas, such as mitigating the heat island effect [2], reducing air and noise pollution [3,4], sequestering carbon [5], and providing multisensory experiences [6]. These benefits also emphasize the importance and contribution of urban forests to the goal of healthy cities proposed by the World Health Organization [7]. The acoustic environmental features of urban forests have been identified as one of the most essential determinants for the various functions of urban forests [8,9]. They play a vital role, especially in people’s perception, experience, and understanding of a landscape, and are strongly associated with the benefits that people gain from the urban forest ecosystem [10,11]. They provide numerous recreational ecosystem services for visitors, such as landscape aesthetic values and nature-based recreation [12]. Such contributions can also promote human physiological and psychological health, for example, by heart rate reduction, mental restoration, and stress recovery [13,14]. Moreover, acoustic environmental features also reflect a series of ecological values; they are a necessary channel for daily communication between vocal creatures, especially the passerine bird species [15].

It is essential to combine the considerations of both physical and perceptual dimensions to achieve a holistic understanding of an acoustic environment [16]. The former refers to the physical sound level that is usually used to measure noise exposure [17]. It usefully captures the physical sound intensity exerted by the sound sources in an environment. The latter refers to the features of the acoustic environment, as perceived or experienced by a person or people in context, as stated in ISO 12913-1 [18], which highlights the perceived attributes of certain sound types and the overall sound environment [19]. Such perceptual sound features can help to understand the interaction between humans and the acoustic environment [20]. Thus, this measure is meaningful for understanding the changing mechanisms of both the physical and perceptual acoustic environments and the factors influencing them, allowing planners to obtain better results in the landscape planning, design, and management of urban forests [21,22]. These physical and perceptual sound features shape the distinct spatiotemporal patterns of the acoustic environment within urban forests [23]. They tend to be influenced by various landscape characteristics of urban forests because these sound features originally emerged from the landscape [24].

However, recent studies focusing on the acoustic environment in urban forested areas have mostly failed to consider the spatiotemporal pattern of both perceptual and physical dimensions simultaneously. Furthermore, they rarely addressed the various impacts of landscape characteristics on these two features across times and spaces. Most of the studies only concerned either the perceptual sound features (e.g., perceived sound frequency, intensity, and soundscape preference) [25,26,27,28] or the physical acoustic features (e.g., sound pressure levels) [29], and did not explore their spatial or temporal variability. There were some investigations considering spatiotemporal variations [9,30]. Still, the authors concentrated on physical acoustic features (indicated by, e.g., acoustic indices), and neglected perceptual sound features. Recently, a few studies have begun to take into account both the perceptual and physical features of the acoustic environment and the spatial and/or temporal variability of urban forested areas [21,31]. Bian et al. explored the relationship between the physical and perceptual features of the acoustic environments in urban forests, but they solely examined the temporal changes found in these two features [31]. Hong et al. studied the spatial and temporal variations of perceptual sound elements and sound pressure levels in an urban forest [21]. However, their analyses and results regarding spatial and temporal variabilities were only based on subjective observations instead of statistical calculations when establishing the significance of the changes. This flaw also exists in other, similar studies [32,33,34,35], which makes such results hard to adopt as a scientific reference for the practice because such variabilities may be caused by random fluctuations in the analyzed data [36]. More importantly, how urban forest landscape characteristics affect the perceived and physical sound characteristics at specific times or in specific spaces remains underexplored. Mostly, the existing studies have not comprehensively considered spatial and temporal factors in exploring the relationship between acoustic environmental features and landscape characteristics [32,33,35,37]. It is difficult to summarize those landscape characteristics that have a significant impact on the acoustic environment at a specific time or space in urban forests, solely based on such results. This would give an insufficient scientific basis to support the prioritization of influential landscape characteristics in landscape planning and design at specific times or spaces, for proposing concrete measures for the dynamic management and sustainable development of sound resources, or for implementing spatiotemporal acoustic environment impact assessments.

To fill these research gaps, this study aims to explore the spatiotemporal dynamics of perceptual and physical soundscapes and their relationships with spatial landscape characteristics in urban forests. For this purpose, the present study examines whether there are significant variabilities in both the perceptual and physical features of the acoustic environment in urban forests and also identifies how landscape characteristics influence spatiotemporal perceptual and physical sound features, based on a case study of Fuzhou National Forest Park, China. The perceptual and physical sound features were measured separately at different periods and functional zones within urban forests. The measured data were first processed and then used to examine the significance of spatial and temporal variabilities in the perceptual and physical sound features, using statistical methods. A series of multiple stepwise regression models were employed to identify the impact of landscape characteristics on perceptual and physical sound features in different times and spaces. This study will advance the understanding of the spatiotemporal mechanisms of the acoustic environment in urban forests and their relationship with landscape characteristics, which can usefully inform both landscape planning and design as well as dynamic resource management in urban forested areas.

2. Materials and Methods

2.1. Study Area

The study area (covering 218 ha) is located within the Fuzhou National Forest Park, a famous tourist destination in Fuzhou, China. The park experiences a subtropical maritime climate with mild temperatures and abundant rainfall, and its ecosystem is dominated by subtropical evergreen broadleaf forests with over 1700 tree species. The park also shares rich ecological and cultural resources that serve as tourism assets, aligning with the themes of forest tourism and promoting natural education and ecological awareness to the public. These inherent conditions bestow its status as a representative example of an urban forest, allowing the current study to provide useful results and instructive information that can contribute to urban forest research.

Before the formal field survey, a pilot field investigation in the study area was conducted to predetermine the functional zoning, sampling sites, and sound source categories. The functional zoning process was based on the resources, features, and main functions of the areas that are described in

Table 1. Information on the functional zones and sampling sites within the study area.

Functional Zone	Code	Area (ha)	Number of Sampling Sites	Brief Description
Forest landscape	FL	75	5	This area is full of natural resources, especially forest landscapes.
Recreation	RC	37	7	This area contains several specialized botanical gardens, places for outdoor activities (e.g., barbecues, picnics, and children playing), and a few museums.
Waterside	WS	47	4	This area is centered on the “Bayi” Reservoir, a famous reservoir in Fuzhou, and its surrounding waterfront landscapes.
Cultural landscape	CL	59	5	This area is characterized by rich historical elements and cultural landscapes, such as historical temples and monumental sites.

. A total of 21 sampling sites were selected in the four functional zones of the study area (Figure 1). The selection of sampling sites was based on three main aspects: (1) being spatially accessible to ensure ease of visitor movement within the selected areas; (2) areas with relatively high visitor flow and density or long stay durations; and (3) representative high-interest areas in each functional zone (e.g., landscape nodes and scenic spots). This selection process ensured that each sampling site had enough potential participants to take part in the survey and that the sites adequately reflected the spatiotemporal variations of the soundscape features experienced by visitors in the urban forests. The number of sampling sites was set with reference to previous studies [34,38], ensuring that the sampling size was sufficient to capture the spatiotemporal variability of sound features in the area. Afterward, 21 common sound sources were identified around all the selected sampling sites and were then categorized into natural sounds, human activity, and mechanical sounds (

Table 2. Common sounds identified in the study area.

Main Category Sound	Code	Sub-Category Sound
Natural sounds	NS	Birds, insects, frogs, wind-induced vegetation, wind, waterfall, stream, water drops
Human activity sounds	HS	Talking, footsteps, hawking, recreational activity, playing, exercise
Mechanical sounds	MS	Broadcasting, phones ringing, alarms, construction, traffic, musical instruments, music on the radio

), based on the taxonomy of sound sources in ISO 12913-2 [19].

2.2. Data Collection and Processing

2.2.1. Field Survey

The formal field surveys were conducted during the weekends in October and November 2020, specifically on 24, 25, and 31 October, and 1, 7, 8, 14, 15, 21, and 22 November, for a total of 10 days. October and November were chosen because the weather in Fuzhou during these months is relatively stable, with mild temperatures compared to the summer or winter months, which is favorable for people engaging in outdoor activities and is, thus, also conducive to on-site data collection and measurement. Each survey was conducted over three time periods: morning, 8:00–11:00 (P1); afternoon, 12:00–15:00 (P2); and evening, 16:00–19:00 (P3). These periods have been used in previous studies to present the temporal variations in sound features effectively [21,31,33]. The specific time points of each period were slightly adjusted to better align with the actual conditions (e.g., the flow of visitors) in the study area. The surveys were systematically organized by the second author and two master’s degree students and were collaboratively conducted with the assistance of six additional master’s degree students. All of them majored in landscape architecture and had expertise in the soundscape field. The observers were divided into four groups for data collection purposes, with one three-person group located in the recreation zone and three two-person groups in the other three zones, divided according to the number of sampling sites and the flow of visitors. Each group moved to each sampling site within the responsible functional zone across the three sampling periods within the same day, to collect perceptual and physical sound data simultaneously.

2.2.2. Perceptual Sound Features

This study considers the perceptual acoustic environment in sound source perceptions and overall soundscape perceptions. The sound source perceptions were measured using a seven-point scale, scoring the perceived occurrences (POS) (1—never to 7—very frequently), the perceived loudness (PLS) (1—extremely quiet to 7—extremely loud), and preferences (PRE) (1—extremely dislike to 7—extremely like) [28,39]. The overall soundscape perceptions were evaluated using the semantic differential method, also using a seven-point scale (1—extremely disagree to 7—extremely agree) for six perceptual attributes: pleasant, harmonious, comfortable, eventful, diverse, and vivid [32,39].

The perceptual sound data were collected by conducting questionnaire surveys on-site. The surveys were voluntary and anonymous and were taken with the consent of all participants. The visitors appearing around the sampling points would be asked if they would like to participate in this survey. The maximum distance around these points was approximately 150 m because the areas farther away were limited in accessibility and were not used for human activities (e.g., without trail facilities or with extreme terrain or high-density vegetation), especially the sampling sites in the forest landscape zone. This ensured that the data collected reflected the perceptual sound features that were accessible in the sampling areas. The participants were asked to fill out the questionnaires depending on their current feelings. Each participant could stop the survey and leave at any time without giving any reason. Finally, 814 valid questionnaires were returned after eliminating any invalid ones (e.g., incomplete answers or questionnaires with all the same answers). The results of the reliability and validity tests were good for perceived sound sources and overall soundscape perception. Specifically, the Cronbach’s alpha coefficients were 0.827 (>0.7) and 0.953 (>0.8), respectively, which was considered acceptable for reliability [40]. The values from the Kaiser–Meyer–Olkin (KMO) calculations were 0.761 (>0.6) and 0.744 (>0.6), respectively, showing an acceptable level in terms of the adequacy of the sample size in factor analysis [41]. The p-values of Bartlett’s sphericity test were 0.000 (<0.01), suggesting that the collected data were appropriate for factor analysis [42].

The sound harmonious degree (SHD) indicator in Equation (1) was employed to comprehensively analyze the perceptual features of sound sources in this study. SHD is a two-dimensional indicator that was further developed based on a previous indicator [28]. Its rationality and effectiveness for reflecting and evaluating perceptual sound features have been indicated when used in urban green spaces in previous studies [39,43]. SHD integrates the sound dominant degree (SDD) and the orientation of PRE. SDD refers to the dominance of a specific sound source in the acoustic environment, which depends on POS and PLS, as shown in Equation (2). The orientation of PRE describes the degree to which the dominance of a sound source matches people’s preferences, calculated by Equation (3):

$${\mathrm{S}\mathrm{H}\mathrm{D}}_{\mathrm{j}\mathrm{i}}=\left[{\displaystyle \frac{1}{\mathrm{e}\mathrm{x}\mathrm{p}\left({\overrightarrow{\mathrm{P}\mathrm{R}\mathrm{E}}}_{\mathrm{j}\mathrm{i}}\right)+1}}-0.5\right]\times {\mathrm{S}\mathrm{D}\mathrm{D}}_{\mathrm{j}\mathrm{i}}$$

(1)

$${\mathrm{S}\mathrm{D}\mathrm{D}}_{\mathrm{j}\mathrm{i}}={\mathrm{P}\mathrm{O}\mathrm{S}}_{\mathrm{j}\mathrm{i}}\times {\mathrm{P}\mathrm{L}\mathrm{S}}_{\mathrm{j}\mathrm{i}}$$

(2)

$${\overrightarrow{\mathrm{P}\mathrm{R}\mathrm{E}}}_{\mathrm{j}\mathrm{i}}={\displaystyle \frac{1}{\mathrm{n}}}{\sum }_{\mathrm{j}=1}^{\mathrm{n}}{\mathrm{P}\mathrm{R}\mathrm{E}}_{\mathrm{j}\mathrm{i}}-{\mathrm{P}\mathrm{R}\mathrm{E}}_{\mathrm{j}\mathrm{i}}$$

(3)

where j is the jth sample, i is the ith source, and n is the total sample size. The values for SDD and SHD in each main category sound were the mean values for SDD and SHD in all sub-category sounds, respectively [32,43]. Eventually, the SHDs of natural sounds (SHD_NS), human-activity sounds (SHD_HS), and mechanical sounds (SHD_MS) were included in the analyses.

2.2.3. Physical Acoustic Features

The features of the physical acoustic environment were examined using four acoustic parameters: L_Aeq, L₁₀, L₉₀, and L_10–90. These physical parameters have been used in past studies for measuring and presenting the different aspects of sound pressure levels [33,43]. L_Aeq is the A-weighted sound pressure level (SPL), which represents the sound intensity of the acoustic environment. L₁₀ indicates the foreground sound of the acoustic environment, indicating that the sound level exceeds this value for 10% of the sampling time. L₉₀ represents the sound level that exceeds this value 90% of the sampling time, which can be understood as the background sound. L_10–90 is the fluctuating differences in sound levels between L₁₀ and L₉₀, which can quantify the dynamic variability of the acoustic environment of urban forests. L_10–90 was calculated based on the values of L₁₀ and L₉₀.

The physical acoustic features (L_Aeq, L₁₀, and L₉₀) were measured directly, based on a series of 3-min recordings. This recording duration aligns with the optimal timeframe for people’s psychological engagement and stability in response to environmental conditions, as referenced in prior studies [44,45], making this duration suitable for reflecting the physical features of the acoustic environment that people will experience and focus on. These recordings were obtained simultaneously with the perceptual sound data noted by the group members. The measurement device was placed at a height of around 1.5 m above the ground, which is consistent with the average height of the human ear [46]. The field measurements utilized BSWA308 sound level meters, which are octave sound level meters that use a single-chip ARM. The device updates all fixed-point calculations to a float-point, thereby usefully improving the accuracy and stability of measurement. It has a single range covering the 123 dB/122 dB dynamic range for general environmental noise measurements, a linearity range between 20 dBA and 134 dBA for Class 1 measurements, and between 25 dBA and 136 dBA for Class 2 measurements.

2.2.4. Landscape Characteristics

The indicators of landscape characteristics include four aspects: terrain, area proportion of land cover types, distance to land cover types, and spatial landscape patterns, as referred to in previous studies concerning the relationship between landscape characteristics and acoustic environmental features [47,48]. The study terrain was characterized by elevation, slope, and aspect, which were calculated based on the digital elevation model. The area proportion of and distance to land cover types were calculated directly based on the raster data of vegetation cover type and the vector data of road networks in Fuzhou city. The spatial landscape patterns were indicated by eight landscape metrics, the concrete description and formula of which are listed in Table S1 in the Supplementary Materials. These metrics were also measured based on the raster data of vegetation cover types. All data were processed and analyzed in ArcGIS 10.7 and Fragstats 4.2. More details about the data source and processing steps of each landscape indicator are shown in Table S2 in the Supplementary Materials.

2.3. Statistical Analysis

Principal component analysis (PCA) was utilized to tackle the problem of multicollinearity among the variables and to summarize the information gathered on the variables of overall soundscape perception and each category of landscape characteristics, using the varimax rotation method. The value of each principal component was calculated through the regression method (

Table 3. The extracted components of overall soundscape perception and the four categories of landscape characteristics. More detailed results are shown in Tables S3 and S4 in the Supplementary Materials.

Main Category	Extracted Principal Components	Code	Explained Variance (%)	Cumulative Variance (%)
Overall soundscape perception	Pleasantness	PLE	45.8	91.1
	Eventfulness	EVE	45.3
Terrain	Composite terrain features	CTF	66.8	66.8
Area proportion of land cover type	Area ratio of artificial land cover to forests	Prop_AtF	72.6	92.2
	Area ratio of shrublands to artificial land cover	Prop_StA	19.6
Distance to land cover type	Distance to natural land cover	Dist_NLC	22.5	83.5
	Distance to artificial land cover	Dist_ALC	35.8
	Distance to urban transportation	Dist_UT	25.2
Landscape: spatial patterns	Landscape structural diversity	LSD	63.1	92.5
	Landscape shape complexity	LSC	29.4

).

A multivariate analysis of variance (MANOVA) was used to explore the main and interactive effects of time and space on perceptual and physical sound features. The three sampling periods and four functional zones were set as the categorical independent variables, including SHD_NS, SHD_HS and SHD_MS, PLE, and EVE, as well as L_Aeq, L₁₀, L₉₀, and L_10–90, which were the continuous dependent variables. Multivariate tests for the main and interaction effects of time and space were all significant (Tables S6 and S7, Supplementary Materials). Partial eta squared values of 0.01 to 0.06, 0.06 to 0.14, and greater than 0.14 indicate small, medium, and large effects, respectively [49], highlighting the ability of the factor to explain the variance of the dependent variables. The Games–Howell post hoc test using a Bonferroni correction was employed to conduct pairwise comparisons of each dependent variable between the different sampling periods and functional zones, respectively. Bonferroni correction is a commonly used correction method for multiple comparisons, which can effectively decrease the probability of false positive results (type I errors) in multiple pairwise comparisons [50].

Stepwise multiple linear regression analysis was employed to develop a series of models across times and spaces to examine the impacts of landscape characteristics on the spatiotemporal variability of the sound features. The perceptual and physical sound features were dependent variables, while the landscape characteristics were independent variables. The tolerance (TOL) and variance inflation factor (VIF) methods were applied to conduct collinearity diagnostics and reduce the collinearity issue of the independent variables [51]. TOL > 0.1 and VIF < 10 indicate that there is no collinearity issue [32,52]. The calculations are shown in Equations (4) and (5):

$$\mathrm{T}\mathrm{O}\mathrm{L}=1-{\mathrm{R}}_{\mathrm{j}}^2$$

(4)

$$\mathrm{V}\mathrm{I}\mathrm{F}={\displaystyle \frac{1}{1-{\mathrm{R}}_{\mathrm{j}}^2}}$$

(5)

where ${\mathrm{R}}_{\mathrm{j}}^2$ represents the coefficient of determination for the regression models on all independent variables, except for the examined jth variable. The collinearity independent variables (TOL < 0.1 or VIF > 10) in stepwise regression were removed to ensure the validity of the results. All the data were normalized using the Z-score method before analysis. The statistical analyses were performed with the IBM SPSS Statistics software package version 24.0 (IBM Corp., Armonk, NY, USA).

3. Results

3.1. Spatiotemporal Variability of the Perceptual and Physical Sound Features in Urban Forests

The results of the MANOVA in

Table 4. Tests of the between-subject effects of the MANOVA used for assessing temporal and spatial effects on the perceptual sound attributes.

Sound Attribute	Effect	Type III Sum of Squares	df	Mean Square	F	Sig.	Partial Eta Squared
SHD_NS a	Time	55.238	2	27.62	6.38 **	0.002	0.016
	Space	392.186	3	130.73	30.19 ***	0.000	0.101
	Time × Space	56.869	6	9.48	2.19 *	0.042	0.016
SHD_HS b	Time	196.259	2	98.13	13.15 ***	0.000	0.032
	Space	67.604	3	22.53	3.02 *	0.029	0.011
	Time × Space	39.927	6	6.65	0.89	0.500	0.007
SHD_MS c	Time	144.527	2	72.26	25.91 ***	0.000	0.061
	Space	52.605	3	17.54	6.29 ***	0.000	0.023
	Time × Space	75.964	6	12.66	4.54 ***	0.000	0.033
PLE d	Time	0.244	2	0.12	0.13	0.878	0.000
	Space	43.258	3	14.42	15.35 ***	0.000	0.054
	Time × Space	10.854	6	1.81	1.93	0.074	0.014
EVE e	Time	0.681	2	0.34	0.35	0.702	0.001
	Space	22.708	3	7.57	7.86 ***	0.000	0.029
	Time × Space	16.663	6	2.78	2.88 **	0.009	0.021

^a Adjusted R² = 0.127; ^b adjusted R² = 0.050; ^c adjusted R² = 0.117; ^d adjusted R² = 0.061; ^e adjusted R² = 0.037. * p < 0.05, ** p < 0.01, and *** p < 0.001.

show that only the SHDs of three sound types varied significantly across the times, while all perceptual sound features exhibited evident variations across spaces. This shows that the overall soundscape perception tended to be more stable than the SHDs across time. In addition, the differences in the types of spatial functions could make the perceptual sound features more variable than diurnal changes generally. The interaction effect between time and space was significant for the variations of SHD_NS, SHD_MS, and EVE, but not for the changes in SHD_HS and PLE. This means that SHD_NS, SHD_MS, and EVE may be affected by time and space simultaneously.

The main and interaction effects of time and space were significant for all physical acoustic features with mostly large effect sizes (partial eta squared > 0.14) (

Table 5. Tests of the between-subject effects from the MANOVA regarding temporal and spatial effects on the physical acoustic features: time-sampling period and space-functional zone.

Acoustic Feature	Effect	Type III Sum of Squares	df	Mean Square	F	Sig.	Partial Eta Squared
LAeq a	Time	3925.767	2	1962.88	64.13 ***	0.000	0.138
	Space	6450.619	3	2150.21	70.24 ***	0.000	0.208
	Time × Space	5618.213	6	936.37	30.59 ***	0.000	0.186
L10 b	Time	5228.180	2	2614.09	98.62 ***	0.000	0.197
	Space	7269.518	3	2423.17	91.42 ***	0.000	0.255
	Time × Space	5190.332	6	865.06	32.64 ***	0.000	0.196
L90 c	Time	3274.579	2	1637.29	144.31 ***	0.000	0.265
	Space	3117.913	3	1039.30	91.60 ***	0.000	0.255
	Time × Space	1055.648	6	175.94	15.51 ***	0.000	0.104
L10–90 d	Time	266.106	2	133.05	11.53 ***	0.000	0.028
	Space	1149.717	3	383.24	33.21 ***	0.000	0.110
	Time × Space	3185.377	6	530.90	46.00 ***	0.000	0.256

^a Adjusted R² = 0.454; ^b adjusted R² = 0.532; ^c adjusted R² = 0.531; ^d adjusted R² = 0.380. *** p < 0.001.

), indicating that the physical acoustic features were more amenable to changes in space and time than the perceptual sound features, especially in terms of temporal effect. The main effect of time had the largest effect size on the variations of L_Aeq and L₁₀, and the main effect of space had the largest effect size on L₉₀, while the effect size of interaction on L_10–90 was the highest. This implies that the individual temporal or spatial factor has a better ability to explain the sound intensity, while the interaction effect accounts more notably for the dynamic changes in sound levels.

Figure 2 shows that, in general, more significant spatial variabilities of the perceptual sound features were detected than temporal variabilities, except for SHD_HS. This echoes the results of the MANOVA to some extent, showing the more influential role of spatial function types. Temporally speaking, SHD_NS was significantly low in P2, and SHD_MS was significantly high in P1. SHD_HS was significantly high in P1 and P2, showing that people performed more daytime exercises, such as walking, running, or tai chi, in the study area. Spatially speaking, SHD_NS was considerably increased in the FL and CL zones than in the RC and WS zones. This is because the flow of visitors was more concentrated in the RC and WS zones, offering more opportunities for people to experience natural sounds, so as to increase the SHD_NS in the FL and CL zones. SHD_MS was visibly decreased in the FL zone than in other zones, which may be associated with the relatively high density of vegetation in the FL zone, thereby reducing the intensity of MS. The results show that diurnal changes did not cause significant variability in PLE and EVE, showing a stable temporal pattern of overall soundscape perception, as found in the MANOVA (Table 5). However, PLE significantly improved in the FL zone compared to other zones, and EVE was also evidently higher in the FL zone than in the RC and WS zones. This is associated with more natural features and elements in the FL zone than in the other three zones, providing more natural and diverse sound compositions, thereby constituting a more pleasant and eventful soundscape.

Similarly, L_Aeq, L₁₀, L₉₀, and L_10–90 also showed more significant variability across spaces than across times (Figure 3), indicating the important role of the type of spatial function in the variability of physical acoustic features. The mean differences in the physical acoustic features in P2 were all found to be significantly higher than in P1 and P3, indicating that the overall sound events in the urban forests were most intense and changeable in the afternoon. This offered an overall perspective from which to understand the sound activities in urban forests compared to the perception of individual sound sources. In addition, all physical acoustic features in the FL zone were significantly lower than in other zones. This also reflects the same information as the results of perceptual sound features from a different perspective, which can also indicate that the FL zone contains relatively good ecological conditions (e.g., a high density and variety of vegetation) and less intensity concerning human activities. All of these features provide the premise for a quieter and more stable acoustic environment existing in the FL zone than the other zones. Interestingly, the means of four features were even slightly higher in the CL zone than in the RC and WS zones. This may be because the sound compositions in the CL zone were more complicated; the sound compositions here were dominated by natural and human activities, while the sound compositions in the other three zones were either nature-dominated or human-dominated.

3.2. Influences of Landscape Characteristics on the Perceptual and Physical Sound Features Across Time and Space

All the significant (p < 0.05) and non-collinearity (TOL > 0.1 and VIF < 10) independent variables with their standardized coefficients (Beta) from each model are listed in

Table 6. Results of the regression models for perceptual sound attributes in the different sampling periods.

Dependent Variable	Sampling Period	Independent Variable	Beta	Tolerance	VIF	Adjusted R2	F
SHD_NS	P1	Dist_UT	0.202 ***	0.873	1.145	0.154	27.403 ***
		LSD	−0.415 ***	0.873	1.145
	P2	CTF	0.214 **	0.422	2.368	0.173	24.259 ***
		LSD	−0.213 **	0.425	2.351
		LSC	−0.128 *	0.969	1.032
	P3	Prop_AtF	0.466 *	0.149	6.719	0.226	14.862 ***
		LSD	−0.735 ***	0.145	6.899
		LSC	−0.208 **	0.850	1.177
SHD_HS	P1	CTF	0.407 ***	0.244	4.093	0.095	6.245 ***
		Prop_StA	−0.264 **	0.402	2.490
		Dist_UT	0.526 ***	0.307	3.261
		LSC	0.156 *	0.717	1.394
		LSD	0.257 **	0.321	3.120
	P2	None	-	-	-	-	-
	P3	Prop_StA	0.207 *	0.949	1.053	0.086	7.253 ***
		LSD	0.166 *	0.949	1.053
SHD_MS	P1	Dist_ALC	0.262 *	0.186	5.387	0.068	11.069 ***
		LSD	0.472 ***	0.186	5.387
	P2	Prop_AtF	−0.366 **	0.226	4.434	0.037	6.673 ***
		LSD	0.241 *	0.226	4.434
	P3	None	-	-	-	-	-
PLE	P1	LSD	−0.287 ***	1.000	1.000	0.082	27.148 ***
	P2	Prop_AtF	0.373 **	0.226	4.434	0.075	14.236 ***
		LSD	−0.538 ***	0.226	4.434
	P3	LSD	−0.170 *	1.000	1.000	0.029	4.596 *
EVE	P1	Prop_StA	−0.190 ***	1.000	1.000	0.036	11.325 ***
	P2	CTF	0.108 *	1.000	1.000	0.024	4.201 ***
		Prop_StA	−0.11 *	1.000	1.000
	P3	Prop_AtF	0.571 **	0.171	5.862	0.147	13.230 ***
		LSD	−0.821 ***	0.171	5.862

* p < 0.05, ** p < 0.01, *** p < 0.001.

,

Table 7. Results of the regression models for perceptual sound attributes in different functional zones. Only the significant independent variables are listed.

Dependent Variable	Functional Zone	Independent Variable	Beta	Tolerance	VIF	Adjusted R2	F-Statistic
SHD_NS	FL	None	-	-	-	-	-
	RC	Dist_NLC	−0.525 ***	0.745	1.342	0.211	42.195 ***
		LSD	0.343 ***	0.745	1.342
	WS	Dist_ALC	−0.350 ***	1.000	1.000	0.122	24.096 ***
	CL	Prop_StA	0.272 *	0.205	4.872	0.174	25.595 ***
		Dist_NLC	0.641 ***	0.205	4.872
SHD_HS	FL	Dist_ALC	0.267 *	1.000	1.000	0.071	5.607 *
	RC	Dist_UT	0.142 *	1.000	1.000	0.020	6.506 *
	WS	Prop_StA	−0.166 *	1.000	1.000	0.028	4.910 *
	CL	Prop_AtF	0.264 ***	1.000	1.000	0.070	18.235 ***
SHD_MS	FL	None	-	-	-	-	-
	RC	Prop_StA	0.410 **	0.119	8.393	0.177	16.818 ***
		Prop_AtF	1.059 ***	0.106	9.398
		Dist_UT	−0.569 ***	0.124	8.062
		LSD	−0.935 ***	0.134	7.480
	WS	Dist_NLC	−0.316 ***	1.000	1.000	0.100	19.226 ***
	CL	Prop_AtF	0.229 ***	1.000	1.000	0.053	13.548 ***
PLE	FL	Prop_StA	0.269 *	1.000	1.000	0.072	5.702 *
	RC	CTF	−0.320 ***	0.819	1.220	0.094	16.267 ***
		Dist_NLC	−0.235 ***	0.819	1.220
	WS	Dist_ALC	−0.496 ***	1.000	1.000	0.246	56.586 ***
	CL	Dist_UT	−0.239 ***	1.000	1.000	0.057	14.721 ***
EVE	FL	None	-	-	-	-	-
	RC	LSD	0.175 **	1.000	1.000	0.031	10.019 **
	WS	Prop_StA	−0.324 **	0.417	2.397	0.291	35.278 ***
		Dist_NLC	0.250 *	0.417	2.397
	CL	Dist_NLC	0.133 *	1.000	1.000	0.018	4.415 *

* p < 0.05, ** p < 0.01, *** p < 0.001.

,

Table 8. Results of the regression models for physical acoustic features in different sampling periods. Only the significant independent variables are listed.

Dependent Variable	Sampling Period	Independent Variable	Beta	Tolerance	VIF	Adjusted R2	F-Statistic
LAeq	P1	Prop_StA	0.250 ***	0.467	2.140	0.383	37.160 ***
		Prop_AtF	−0.599 ***	0.107	9.344
		Dist_UT	−0.173 *	0.357	2.802
		LSD	0.735 ***	0.102	9.838
		LSC	−0.519 ***	0.387	2.584
	P2	Prop_StA	0.268 ***	0.299	3.348	0.554	71.300 ***
		Dist_NLC	0.496 ***	0.294	3.401
		Dist_ALC	0.810 ***	0.110	9.090
		Dist_UT	0.490 ***	0.369	2.711
		LSD	0.631 ***	0.109	9.208
		LSC	0.629 ***	0.382	2.618
	P3	Prop_StA	−0.368 ***	0.385	2.597	0.507	31.052 ***
		Dist_NLC	0.489 ***	0.329	3.040
		Dist_UT	0.329 ***	0.525	1.904
		LSD	0.188 **	0.811	1.233
		LSC	0.638 ***	0.318	3.149
L10	P1	CTF	−0.135 *	0.951	1.052	0.163	19.566 ***
		Prop_StA	0.275 ***	0.760	1.316
		LSC	−0.126 *	0.762	1.312
	P2	Dist_NLC	0.462 ***	0.556	1.797	0.464	66.458 ***
		Dist_ALC	0.225 ***	0.930	1.076
		Dist_UT	0.732 ***	0.785	1.275
		LSC	0.532 ***	0.511	1.956
	P3	CTF	−0.281 ***	0.911	1.098	0.504	51.760 ***
		Dist_NLC	0.615 ***	0.552	1.812
		LSC	0.764 ***	0.593	1.686
L90	P1	Prop_AtF	−0.535 ***	0.199	5.026	0.251	33.674 ***
		Dist_UT	0.216 ***	0.780	1.282
		LSD	0.731 ***	0.179	5.590
	P2	Dist_NLC	0.284 ***	0.556	1.797	0.477	79.097 ***
		Dist_ALC	0.130 **	0.930	1.076
		Dist_UT	0.768 ***	0.785	1.275
		LSC	0.391 ***	0.511	1.956
	P3	Prop_AtF	−0.255 ***	0.841	1.189	0.519	32.557 ***
		Prop_StA	−0.322 **	0.372	2.688
		Dist_NLC	0.599 ***	0.292	3.419
		Dist_UT	0.546 ***	0.543	1.843
		LSC	0.686 ***	0.328	3.051
L10–90	P1	Prop_StA	0.380 ***	0.504	1.985	0.182	22.323 ***
		Dist_UT	−0.320 ***	0.495	2.018
		LSC	−0.296 ***	0.721	1.387
	P2	CTF	−0.596 ***	0.594	1.684	0.251	38.890 ***
		Dist_ALC	0.571 ***	0.588	1.700
		LSC	0.194 ***	0.934	1.071
	P3	Dist_NLC	0.568 ***	0.601	1.665	0.523	55.818 ***
		LSD	0.411 ***	0.974	1.027
		LSC	0.694 ***	0.590	1.694

* p < 0.05, ** p < 0.01, *** p < 0.001.

and

Table 9. Results of the regression models for physical acoustic features in different functional zones. Only the significant independent variables are listed.

Dependent Variable	Functional Zone	Independent Variable	Beta	Tolerance	VIF	Adjusted R2	F-Statistic
LAeq	FL	Dist_ALC	0.551 *	0.140	7.140	0.236	11.094 ***
		LSC	0.951 **	0.140	7.140
	RC	CTF	0.543 ***	0.319	3.132	0.682	133.712 ***
		Prop_StA	0.364 ***	0.148	6.760
		Dist_NLC	0.747 ***	0.263	3.796
		Dist_ALC	−0.451 ***	0.451	2.216
		Dist_UT	−0.442 ***	0.197	5.085
	WS	Dist_UT	0.782 ***	1.000	1.000	0.612	272.430 ***
	CL	Prop_StA	0.254 ***	1.000	1.000	0.065	16.894 ***
L10	FL	Prop_StA	0.264 *	0.818	1.222	0.263	12.817 ***
		Dist_UT	−0.566 ***	0.818	1.222
	RC	CTF	0.606 ***	0.319	3.132	0.686	136.417 ***
		Prop_StA	0.38 ***	0.148	6.760
		Dist_NLC	0.784 ***	0.263	3.796
		Dist_ALC	−0.409 ***	0.451	2.216
		Dist_UT	−0.412 ***	0.197	5.085
	WS	Prop_AtF	−0.68 ***	1.000	1.000	0.462	148.835 ***
	CL	Prop_StA	0.301 ***	1.000	1.000	0.091	24.301 ***
L90	FL	Prop_StA	0.333 **	1.000	1.000	0.111	9.081 **
	RC	CTF	−0.513 ***	0.524	1.907	0.760	197.947 ***
		Prop_StA	0.871 ***	0.247	4.046
		Dist_ALC	−0.905 ***	0.495	2.020
		Dist_UT	0.523 ***	0.182	5.482
		LSC	0.672 ***	0.258	3.878
	WS	CTF	−0.159 *	0.857	1.166	0.440	67.537 ***
		Prop_AtF	−0.587 ***	0.857	1.166
	CL	Prop_StA	0.127 *	1.000	1.000	0.016	3.999 *
L10–90	FL	Dist_ALC	1.587 ***	0.140	7.148	0.582	33.014 ***
		Dist_UT	−0.467 ***	0.484	2.067
		LSC	1.470 ***	0.119	8.384
	RC	Dist_ALC	−0.702 ***	0.569	1.758	0.442	82.941 ***
		Dist_UT	−0.694 ***	0.689	1.452
		LSD	−0.353 ***	0.775	1.290
	WS	LSD	0.647 ***	0.669	1.494	0.295	36.068 ***
		LSC	0.250 **	0.669	1.494
	CL	CTF	0.463 ***	0.369	2.709	0.288	49.134 ***
		LSD	0.825 ***	0.369	2.709

* p < 0.05, ** p < 0.01, *** p < 0.001.

. Generally, the model results show the complex and varied impacts of landscape components on the sound features in different times and spaces. Considering the space limitations, and to avoid redundant elaboration, this section mainly concentrates on the results that are representative, interesting, or capable of reflecting overall patterns.

Table 6 shows that LSD had continuous impacts on SHD_NS and PLE in all sampling periods; however, for different functional zones, no specific landscape components were detected to consistently influence the perceptual sound features (Table 7). This phenomenon may be associated with the differences between the temporal and spatial variabilities of perceptual sound features. When combining the results in Section 3.1, this further indicates that the greater variabilities in the perceptual sound features may lead to less possibility of the persistent influence of landscape characteristics in urban forests. In addition, EVE was negatively affected by Prop_StA in P1 and P2 but positively influenced by Prop_AtF in P3. These results exhibit the positive correlations found between EVE and the area proportion of artificial land cover, which implies that the area proportion of artificial land cover is one of the main drivers of creating eventful soundscapes.

Although the cross-spatial impacts of landscape characteristics on the perceptual sound features were more complicated and changeable than the cross-temporal impacts, there are still some similarities regarding the impacts of the main categories of landscape characteristics among perceptual sound features within a specific functional zone. For instance, the SHDs of the three main category sounds were affected more by the distance to land cover types in the RC zone and by the area proportion of land cover types in the CL zone. The overall soundscape perceptions were influenced more by the distance to land cover types in both the WS and CL zones. This informs the choice of management measures for the acoustic environment in urban forests, which can prioritize these categories of landscape characteristics in the development of the overarching framework; however, during implementation, adaptive and explicit adjustments should be made according to the type of spatial function, especially for those areas with frequent human activities.

Table 8 indicates that in the different periods, L_Aeq was consistently affected by LSD, Prop_StA, and Dist_UT, L₁₀ and L_10–90 were consistently influenced by LSC, and L₉₀ was consistently affected by Dist_UT. These findings imply that the influence of landscape components on the physical acoustic features may be more regular and stable than the perceptual features across sampling periods. It is a reasonable assumption because both the acoustic features and landscape characteristics share the same nature of physicality and objectivity.

Similar to the perceptual sound features, there was no specific landscape component that consistently affected the physical sound features in all spaces (Table 9). However, it is interesting that L₁₀ and L₉₀ were affected by the same components of the area proportion of land cover types within the same functional zones. Specifically, they were affected by Prop_StA in the FL, RC, and CL zones, and by Prop_AtF in the WS zone. This reveals that the ratio between natural and artificial land cover is one of the essential determinants for both foreground and background sounds. Generally, in open spaces, the foreground sounds are more related to human sounds, while the background sounds are composed of more natural sounds [53], so that they also reflect the activity patterns of vocal organisms and people in some way.

Overall, the temporal and spatial regression models of physical acoustic features had higher R² values than those of perceptual sound features. This is also associated with their same inherent attributes; therefore, objective landscape characteristics can better capture the variance of physical acoustic features. This also implies that the operating landscape measures used in the regulation and management of the acoustic environment in urban forests may be more effective for capturing the physical aspects.

4. Discussion

4.1. Spatiotemporal Variation Mechanisms of the Perceptual and Physical Soundscape Features in Urban Forests

The study results indicate that the spatiotemporal variabilities of most perceptual and physical sound features were generally significant. In addition, there were always more significant variabilities found between the different functional zones than between different sampling periods for both perceptual and physical sound features, except for SHD_HS (Figure 2 and Figure 3). This echoes the finding that the spatial factor generally had a better explanatory ability for the variance in sound features than for the temporal factors (Table 4 and Table 5). This also indicates that both the perceptual and physical acoustic environments in urban forests are influenced more strongly by the spatial function type than by the diurnal pattern. Such findings suggest that during the processes of landscape planning, design, and management, spatiotemporal patterns of the sound features should be considered to better support decision-making regarding the construction and maintenance of a good-quality acoustic environment in urban forests. The type of spatial or landscape function should be prioritized over diurnal variations.

The results of SDD and SHD calculations of natural and human sounds reveal the spatiotemporal patterns of the behaviors of vocal organisms and humans in urban forests. The significantly high SDD_NS and SHD_NS values recorded in the morning and evening may be attributed to vocal activities, especially of passerine bird species, as in the “dawn chorus” or “dusk chorus” [54,55]. Therefore, it is suggested that the measures for protecting natural sound resources should focus on the morning and evening periods when vocal animals are generally more active, especially by minimizing human interference during these periods. In addition, both SDD_HS and SHD_HS values were significantly high in the morning, which may be associated with people’s morning exercise, such as walking, tai chi, or dancing [34,56]. However, this did not result in obvious reductions in SDD_NS and SHD_NS (Figure 2 and Figure S1). This finding essentially reveals the importance of appropriate functional zoning planning in urban forests, which is useful for reducing the conflict between natural and human-activity sounds and for maintaining their respective harmony in the environment. The reasonable allocation of public open spaces (e.g., recreational and waterside areas) within urban forests can help guide the flow of people, provide tourists with suitable activity venues, and mitigate excessive human disturbance in nature-dominated areas (e.g., forest landscape areas).

In addition, we found that human activity and mechanical sounds were considered relatively harmonious across times and spaces (Figure 2), although they were generally not thought to be positive in previous studies [33,57,58]. We infer that the sound harmonious degree shares a similar core of subjective landscape preferences, which also depends on the use that people make of the landscape [59]. It strongly relates to the appropriateness of sound type and spatial function within the areas. Most of the human and mechanical sounds were incidental elements of human activities, such as the sounds of hawking, playing, broadcasting, and playing musical instruments, so the users were more likely to accept them [60]. This indicator of the harmonious degree of sound reflects similar information to that in the soundscape to some extent [61], but with more focus on individual sound sources. The harmonious degree of sound shows more than simply pleasant or unpleasant feelings from the sounds and is a combination of sound dominance and preference. Using this indicator can comprehensively record the perceived quality of sound sources at a specific time or space, usefully informing planners and decision-makers to better consider people’s well-being, as offered by the acoustic environment of urban forests.

The result of there being no significant temporal variability of soundscape pleasantness and eventfulness is different, in some way, from previous studies [32,33,34], but we also found similar spatiotemporal trends between pleasantness and eventfulness that were consistent with previous evidence [32,33]. The different result regarding significant temporal variability was found because our study employed statistical methods to quantitatively examine the significance of the variations, while previous studies identified significant variations solely based on subjective visual observation, as we mentioned in Section 1. Such “significant” results may be caused by fluctuations in the measured data that are not sufficiently scientific and reference-based to contribute to planning and management practice. [36]. In contrast, our statistically tested results can make acoustic or soundscape research more rigorous by controlling erroneous inferences, reducing subjective biases, and providing reliable and evidence-based references for soundscape planning and acoustic environment management. They can also contribute to a deeper understanding of the spatiotemporal variations in overall soundscape perception.

We found that both the temporal and spatial variabilities of the physical acoustic environment in the studied urban forests were closely associated with the patterns of human activities because they were similar to those of SDD_HS (Figure S1). This implies that the significantly high values for the physical sound feature may be attributed temporally to increased human flow and activities, especially in the afternoon, and, spatially, to those human activities chiefly concentrated in the recreation, waterside, and cultural landscape zones, such as hawking, playing, walking, and exercise. These associations indicate the dominant role of human-related sounds in the spatiotemporal pattern of the physical acoustic environment in urban forests. Such dominance was also found in previous studies [34,62]. This suggests that the implementation of noise management measures in urban forests, if necessary, should focus on areas such as diverting or limiting the flow of tourists, especially in the afternoon and in those areas with recreational and sightseeing functions.

4.2. Understanding the Influence of Landscape Characteristics on Spatiotemporal Sound Features for Landscape Planning and Management in Urban Forests

The temporal and spatial model results can be understood from the global and local perspectives, respectively. The cross-temporal results essentially reflect the influence of the landscape characteristics of the studied urban forests on sound features in different sampling periods at the global level. The cross-spatial results present the impacts of landscape characteristics on the sound features of specific functional zones within urban forests, which tend to reflect the local relationships. This implies that a sound feature was inevitably affected by varied landscape characteristics; sometimes, the directions of influence were even different. Such findings were difficult to detect in previous studies that did not consider both temporal and spatial factors, which also highlights the innovation and significance of our study to some extent. Both global and local results can be combined to better develop specific planning objectives and management measures in practice. In the temporal models, we found that some of the main components of landscape characteristics had persistent effects on perceptual and physical sound features. However, no landscape component had a persistent effect on the sound features across spaces. This is because the acoustic environmental features varied significantly across times and spaces, as presented in our results above, while the landscape characteristics commonly showed significant differences only between spaces, and they did not vary considerably in the short term (e.g., in one day) [63].

Temporally, LSD negatively influenced SHD_NS and PLE but positively affected L_Aeq in all periods. This result suggests that the low diversity of landscape structures in urban forests can usefully improve the diurnal natural harmonious degree of sound and the soundscape pleasantness, while reducing the sound levels simultaneously. This is because a landscape with a simple arrangement and typology of composition can provide more room for natural sound propagation and exposure [15], which is also useful for promoting the overall soundscape pleasantness [8,12]. Also, such a landscape structure is often relatively even and contiguous, which may effectively reduce the impact of the reflection and diffraction of sound waves [64], so it is difficult to gather high sound pressure levels within the selected areas. Similarly, past evidence also indicated that a landscaped area with a uniform structure or on a large scale can usefully reduce the perception of traffic sounds [33,35]. In addition, LSC also had persistent impacts on L₁₀ and L_10–90 across times. Interestingly, the impacts on these two features were negative in the morning but positive in the afternoon and evening. This may be related to the varied composition of the high-intensity sound sources. This speculation can be supported by the results of SDDs to a large extent. Each main category sound had a unique temporal trend of SDD (Figure S1), which means they can combine into different dominant sound compositions at each period. Both L₁₀ and L_10–90 contain information about foreground sounds that are highly related to the dominant sounds, so they were affected by the landscape components in different directions across time.

Spatially, Dist_NLC was found to negatively affect SHD_NS and PLE in the recreation zone but to positively affect SHD_NS and EVE in the cultural landscape zone. These results suggest that within the recreational area, those human-activity spaces close to natural land cover can usefully improve the visitors’ perceived harmony of natural sounds and soundscape pleasantness when they are performing their activities. However, a different measure should be applied in those areas with cultural elements. According to the results, we found that people may generally prefer culture- or history-related elements compared to natural components in such areas, and they consider the peaceful and comfortable attributes to be more important [65]. Accordingly, for those areas with cultural resources, it is suggested to keep the cultural elements a suitable distance from the natural land cover, to maintain the harmony of natural sounds and create a calm soundscape. Furthermore, both L₁₀ and L₉₀ were positively affected by Prop_StA in the forest landscape, recreation, and cultural landscape zones, while they were negatively influenced by Prop_AtF in the waterside zone. Combining our analysis of the physical acoustic features in Section 4.1, these results indicate that for those areas with recreational functions, waterside spaces, or cultural and historical elements, increasing the area ratio of shrubs or forests to pavements may attract more human gatherings and activities. We speculate that this attraction is associated with the thermal comfort of the environment. Although the temperatures during survey periods were lower than in the summer, the weather at that time in Fuzhou city was still relatively sultry. Vegetation coverage can effectively reduce the ambient temperature, thus offering a more comfortable environment [66]. For those areas containing a high proportion of vegetation, the results suggest that an increase in the ratio of shrubs to pavements can lead to an improved intensity of natural sounds because natural sounds dominated (L₁₀) in the forest landscape zone (Figure S1). In addition, the results illustrate that an increase in the area of shrubs or forests can also significantly increase the intensity of background sounds (L₉₀). This is because the increase in vegetation coverage can generate more subtle natural sounds (e.g., wind-induced vegetation sounds), which are key components of background sounds. Such increases can serve as useful measures for improving the overall soundscape quality in landscape planning and design in urban forests.

4.3. Limitations and Suggested Research Topics in the Future

There are still some limitations that can be addressed in future studies. First, the findings of this study were only derived from the Fuzhou National Forest Park, China, so that future studies are encouraged to conduct more comparative investigations in different countries. As such, people’s perceptual outcomes in these studies may also differ from those in the present research due to differentiated social and cultural backgrounds. This raises a second flaw, in that demographic characteristics (e.g., gender, age, education, or occupation) were not included in the analyses. Future research is expected to explore the main effects of demographic characteristics and their interactions with other factors on perceptual sound attributes in urban forests. Despite these limitations, this study successfully followed the research objective and addressed the research questions, exploring the spatiotemporal variability and influential landscape characteristics of perceptual and physical features in urban forests. The study findings offer a deeper understanding of the spatiotemporal pattern of the perceptual and physical acoustic environment and advance the state of knowledge of both temporal and spatial relationships between the sound features and landscape characteristics seen in urban forests. They also provide statistical evidence and scientific references to usefully inform landscape planning and management in urban forests.

5. Conclusions

This study statistically examined the spatiotemporal variability of perceptual and physical sound features and identified the cross-temporal and cross-spatial impacts of landscape characteristics on these sound features. The results reveal that the variabilities of perceptual and physical sound features were both affected more strongly by spatial function types than diurnal patterns. Most sound features varied significantly across times and spaces, while soundscape pleasantness and eventfulness did not show significant temporal variation. We found that the harmonious degree of sound is largely associated with the appropriateness of the sound type and spatial functions. In addition, those human activities that generate sounds were the main drivers for the spatiotemporal variation of physical acoustic features in urban forests. The results of the temporal and spatial regression models essentially reflect the global and local relationships between landscape characteristics and the acoustic environment in urban forests. The main components of landscape patterns (i.e., landscape structural diversity and landscape shape complexity) persistently influenced the sound features across all times, but no cross-spatial impact of the landscape component was found. This study helps improve the understanding of the mechanism of spatiotemporal variations in acoustic environments and their relationship with landscape characteristics, which informs landscape planning and management in urban forests. The study findings can provide scientific references and empirical evidence to support the development of management measures for spatiotemporal dynamic acoustic environmental quality, and to emphasize the priority given to their most influential landscape characteristics in urban forests. The findings can also inform the appropriate arrangement of landscape types and functional zoning planning, usefully promoting the coordination of acoustic environment quality and spatial function types in urban forests. Future research could incorporate more case studies of urban forests from different regions or countries for comparison and consider the influence of the public’s demographic features on the perception of urban forest soundscapes.