*Article* **The Physiological Restorative Role of Soundscape in Different Forest Structures**

**Xin-Chen Hong 1,2,†, Shi Cheng 1,†, Jiang Liu <sup>2</sup> , Emily Dang <sup>3</sup> , Jia-Bing Wang <sup>2</sup> and Yuning Cheng 1,\***


**Abstract:** Natural soundscape is considered a dominant type of hearing in forested areas and contributes to health and recovery effects from exposure to the biophilic outdoor environment. This study focuses on the different forest structures, and aims to explore the relationship between perceived soundscape and acoustical parameters, observe physiological indicators, and model the physiological restorative role of soundscape. Questionnaires and measuring equipment were used to gather psychophysical and physiological information at 20 observation sites in urban forested areas. Back-propagation neural network techniques were conducted to determine the forecasting model from psychophysical to physiological parameters. Our results suggested that LAeq and L<sup>10</sup> are important factors that influence questionnaire responses. Our findings also showed that electromyogram (EMG) signals were the most obvious and sensitive in physiological parameters. Additionally, we found that L10–90 played the most important role among all physical parameters in the physiological restorativeness soundscape model. This can facilitate the understanding of the physiological restorative role of soundscape in different forest structures when proposing suitable forest-based health care strategies.

**Keywords:** soundscape; restorativeness; acoustic parameters; physiological parameters; pleasantness

#### **1. Introduction**

In 2020, the public found themselves in an unprecedented situation. With COVID-19 spreading, lockdowns were enforced, and outdoor activities subsequently declined. This led to a rise in both psychological stress and mortality by suicide [1]. There has been a growing demand by the public, especially in high-density cities, for relaxation and entertainment. This public demand contributes to an important effect for public health and work efficiency, and thus the public pays increasing attention to this issue [2,3]. Fortunately, urban forested areas, which are an important part of the urban green infrastructure, are natural places that provide relaxation, entertainment and perceived restoration to the public [4–6]. Furthermore, forest landscapes contribute ecosystem services to the public [7].

Urban forested areas potentially play a key role in the construction of healthy cities [8]. The World Health Organization (WHO) announced goals for healthy cities, which aim to continuously improve the health and quality of life of city dwellers [9]. Various studies have explored physiological and psychological relationships, including taste [10], touch [11], smell [12], vision [13], hearing [14] and other senses [15,16]. In general, vision is considered the most important driver for sensory and cognition effects in environmental exposure. However, the second important driver, hearing, also plays a key role in cognition and behavior. This includes tracking functions, such as spatial cognition without visuals [17]; positioning and connecting functions, such as judging sound sources and audiovisual relationships [18]; and focusing and memory functions, such as understanding of the

**Citation:** Hong, X.-C.; Cheng, S.; Liu, J.; Dang, E.; Wang, J.-B.; Cheng, Y. The Physiological Restorative Role of Soundscape in Different Forest Structures. *Forests* **2022**, *13*, 1920. https://doi.org/10.3390/f13111920

Academic Editor: Chi Yung Jim

Received: 28 September 2022 Accepted: 14 November 2022 Published: 15 November 2022

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2022 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

environment combining visual information [19]. Thus, hearing potentially occupies an important position in the perception of urban forested areas.

Soundscape is described in the ISO as 'acoustic environment as perceived or experienced and/or understood by a person or people, in context' [20]. In urban forested areas, natural soundscape is considered a dominant type as it performs in natural sound occurrences [21], perceived geophony and biophony [22], and birdsong identification [23]. Exposure to the biophilic outdoor environment contributes to health and recovery, and is applied especially in forest-based health care [24]. Previous studies used questionnaire responses to explore the perception of forest soundscape in national parks [25], urban parks [18], and forest parks [26]. The Perceived Restorativeness Soundscape Scale (PRSS) was developed and tested to assess a soundscape's potential to provide psychological restoration [27]. The PRSS focuses on the dimensions of psychological restoration, such as curiosity, interest, concentration, and demand for soundscape in context. However, the dimensions of physiological restoration for forest soundscapes are lacking, especially in different forest structures. Another potential research gap is how to simulate advanced mental processes that elicit physiological responses and contribute to the modelling of psychophysical parameters to physiological parameters in forested areas.

This study was conducted to fill these gaps and aims to: (1) explore the relationship between perceived soundscape and acoustical parameters in different forest structures; (2) observe physiological indicators in different forest structures; and (3) model the physiological restorative role of soundscapes in forested areas.

## **2. Methodology**

## *2.1. Study Area*

Our study was conducted in the arboretum (8,469,323 m<sup>2</sup> ) of Fuzhou National Forest Park (57,439,074 m<sup>2</sup> ) in Fuzhou, Fujian, China. Fuzhou National Forest Park is located to the north of downtown Fuzhou. It has a subtropical oceanic monsoon climate with an average annual rainfall of about 1438.5 mm. The average wind speed is 1.8 m/s, relative humidity is 75%, and average annual sunshine is 1848 h. The arboretum consists of well-maintained paths, various tree species, and high forest coverage (65.54%). We found that the arboretum was a suitable site for soundscape research as it contains potential sources of both natural sounds and man-made sounds.

Based on different forest structures and previous research [28,29], 20 observation sites were chosen in the arboretum (see Figure 1), comprising five in bamboo forests, five in broad-leaved forests, five in coniferous forests, and five in coniferous/broad-leaved mixed forests. The acoustic environmental conditions at each site were measured for 5 min, and included LAeq, L10, L<sup>90</sup> and L10-L90. The measured LAeq ranged from 43.9 dBA to 76.8 dBA, L<sup>10</sup> from 47.7 dBA to 78.6 dBA, L<sup>90</sup> from 41.8 dBA to 63.1 dBA, and L10-L<sup>90</sup> from 1.9 dBA to 22.1 dBA. *Forests* **2022**, *13*, x FOR PEER REVIEW 3 of 14

**Figure 1.** Observation sites in study area. *2.2. Physiological and Soundscape Information* **Figure 1.** Observation sites in study area.

ological parameters contributed by stress inducement.

2.2.1. Physiological Parameters

and out of the lungs [37].

1 rpm for RESP.

To observe participant responses to stress inducement, we examined the variations

Previous studies have shown that four physiological parameters potentially reflect the role of physiological restoration [30–32]. The first parameter is electromyogram (EMG) and is influenced by frontal muscle activity. Frontal muscle activity decrease when participants are exposed to positive influences from environmental scenes, and increase when exposed to negative influences [33]. This influences EMG levels. The second parameter is electrodermal activity (EDA), a representative measure for mood changes in biometrics research [34]. EDA is affected by exocrine sweat gland activity based on the sympathetic nervous system increasing secretion from sweat glands [35]. The third parameter is photoplethysmography (PPG), which uses low-intensity infrared (IR) light to detect blood volume changes in the microvascular bed of tissue [36]. The fourth parameter is respiration (RESP), the movement of respiratory gases (such as oxygen and carbon dioxide) into

Physiological parameters were measured using ErgoLAB [38,39], a wearable polygraph with 2048 Hz sampling rate, 16-bit resolution and a wireless communication frequency of 2.4 GHz. We found that each of the physiological parameters had different signal accuracies: 0.183 μV for EMG with 16-bit resolution, 0.01 μs for EDA, 1% for PPG, and

To observe the role of physiological restoration in different forest structures, we gathered the pre-test value (Pr) for physiological parameters in bamboo forests, broad-leaved forests, coniferous forests, and coniferous/broad-leaved mixed forests. We then gathered the post-test value (Po) in the same forest structures. The absolute value of Pr-Po reflected how relaxed the participants were and the degree of physiological experienced restoration. Pr-Po was represented by ΔEMG, ΔEDA, ΔPPG, and ΔRESP, respectively.

Five scales were selected to represent the degree to which soundscapes affect physiological restorative role (PRR) [40]. These included 'extremely restorative', 'very restorative', 'moderately restorative', 'slightly restorative', and 'not restorative at all'. To match the five scales, the intervals of ΔEMG, ΔEDA, ΔPPG, and ΔRESP were derived. In green space, EMG, EDA and RESP values are the same order of magnitude, with maximum around or less than 10 [41]. PPG value is another order of magnitude with maximum

#### *2.2. Physiological and Soundscape Information*

#### 2.2.1. Physiological Parameters

To observe participant responses to stress inducement, we examined the variations in the physiological parameters between baseline value (BL) with eye mask and earmuffs, and pre-test value (Pr) after stress inducement. Pr-BL represented the variation in physiological parameters contributed by stress inducement.

Previous studies have shown that four physiological parameters potentially reflect the role of physiological restoration [30–32]. The first parameter is electromyogram (EMG) and is influenced by frontal muscle activity. Frontal muscle activity decrease when participants are exposed to positive influences from environmental scenes, and increase when exposed to negative influences [33]. This influences EMG levels. The second parameter is electrodermal activity (EDA), a representative measure for mood changes in biometrics research [34]. EDA is affected by exocrine sweat gland activity based on the sympathetic nervous system increasing secretion from sweat glands [35]. The third parameter is photoplethysmography (PPG), which uses low-intensity infrared (IR) light to detect blood volume changes in the microvascular bed of tissue [36]. The fourth parameter is respiration (RESP), the movement of respiratory gases (such as oxygen and carbon dioxide) into and out of the lungs [37].

Physiological parameters were measured using ErgoLAB [38,39], a wearable polygraph with 2048 Hz sampling rate, 16-bit resolution and a wireless communication frequency of 2.4 GHz. We found that each of the physiological parameters had different signal accuracies: 0.183 µV for EMG with 16-bit resolution, 0.01 µs for EDA, 1% for PPG, and 1 rpm for RESP.

To observe the role of physiological restoration in different forest structures, we gathered the pre-test value (Pr) for physiological parameters in bamboo forests, broadleaved forests, coniferous forests, and coniferous/broad-leaved mixed forests. We then gathered the post-test value (Po) in the same forest structures. The absolute value of Pr-Po reflected how relaxed the participants were and the degree of physiological experienced restoration. Pr-Po was represented by ∆EMG, ∆EDA, ∆PPG, and ∆RESP, respectively.

Five scales were selected to represent the degree to which soundscapes affect physiological restorative role (PRR) [40]. These included 'extremely restorative', 'very restorative', 'moderately restorative', 'slightly restorative', and 'not restorative at all'. To match the five scales, the intervals of ∆EMG, ∆EDA, ∆PPG, and ∆RESP were derived. In green space, EMG, EDA and RESP values are the same order of magnitude, with maximum around or less than 10 [41]. PPG value is another order of magnitude with maximum around or less than 40. Then, the maximum values were split into twenty parts to observe the variation of parameters. The interval length of ∆EMG, ∆EDA, and ∆RESP was 0.5, and that of ∆PPG was 2. The value of EMG, EDA and RESP dropped gradually without stress inducement in general [42]. For PPG, due to be affected by factors other than stress inducement, we took a symmetric interval distribution. Thus, the scales corresponding to the intervals of ∆EMG, ∆EDA, and ∆RESP were [−∞, −1.5), [−1.5, −1.0), [−1.0, −0.5), [−0.5, 0.0), and [0.0, +∞]. ∆PPG was [−∞, −3.0), [−3.0, −1.0), [−1.0, 1.0), [1.0, 3.0), and [3.0, +∞].

#### 2.2.2. Soundscape Parameters

In this study, questionnaires and measuring equipment were used to gather soundscape parameters in different forest structures [43]. Questionnaires were conducted to inquire about the pleasantness of perceived soundscape (PL): not pleasant at all (+1), slightly pleasant (+2), moderately pleasant (+3), very pleasant (+4), and extremely pleasant (+5).

Soundscape parameters were collected via measurements from Type-1 sound level meters (AWA 6228+) at 1.5 m height. This included measuring LAeq, L10, L90, and L10-L90. LAeq was the A-weighted equivalent sound pressure level. L<sup>10</sup> and L<sup>90</sup> were statistical levels that represented the levels that exceeded 10% and 90%, respectively. L10-L<sup>90</sup> measured temporal variability and represented the difference between L<sup>10</sup> and L90.

#### 2.2.3. Stress Inducement

Stress inducement came mainly through mathematical calculations for participants and consisted of two parts. The first part involved asking participants to add two three-digit random numbers. The results were a four-digit number, such as '571 + 815 = 1386'. The second part involved asking them to multiply a two-digit and a one-digit random number. The results were a three-digit number, such as '89 × 5 = 445<sup>0</sup> . There were ten sets in total, with five sets in each part.

#### *2.3. Physiological Restorativeness Soundscape Modeling*

To simulate psychophysical processes that elicited physiological responses, a backpropagation neural network was created to determine the forecasting model from psychophysical parameters to physiological parameters [44]. For the back-propagation neural network, LAeq, L<sup>10</sup> L90, L10-L90, and PL were selected as input variables, while ∆EMG, ∆EDA, ∆PPG, ∆RESP, and PRR were selected as output variables.

There were two hidden layers in the physiological restorativeness soundscape model (PRS model), which included 5 neurons and 4 neurons in the first and second hidden layers, respectively. Hyperbolic tangent functions were used for all neurons in each hidden layer.

## *2.4. Procedure*

#### 2.4.1. Participants and Equipment Measuring

Physiological and soundscape information were gathered on weekdays with sunny weather between 9:00 and 17:00 in the months of February and March 2020. Young adults make up the majority of urban forest visitors [45]. Thus, we randomly recruited staff and graduate students from local universities in Fuzhou. A total of 48 participants (male = 25, female = 23, average 29.5 ± 5.1) with normal hearing abilities were recruited to respond to questionnaires and gather physiological information in a sitting position. Before the test began, all participants were required to sign a consent form outlining the details of the study, including content, purpose and methodology. Furthermore, participants could quit the study at any point if they felt uncomfortable during the process.

In this study, the measuring process included five steps with a total duration of 15 min (See Figure 2). Due to the limited number of ErgoLAB devices (24 sets), all participants were divided into two groups and required to complete the measuring process separately by single group. In the group, half of the participants were tested at the same time. *Forests* **2022**, *13*, x FOR PEER REVIEW 5 of 14

**Figure 2.** Measurements time stamp for gathering parameters. **Figure 2.** Measurements time stamp for gathering parameters.

Step 1: Preparatory work. We spent 2 min on participants putting on ErgoLAB and placing acoustical equipment at an observation site. Step 1: Preparatory work. We spent 2 min on participants putting on ErgoLAB and placing acoustical equipment at an observation site.

Step 2: Peaceful statement**.** Participants spent 2 min on maintaining a peaceful state, wearing eye masks and earmuffs. Meanwhile, we used ErgoLAB to gather their BL values in this step. Step 3: Stress inducement**.** We spent 5 min on inducing stress in participants . Mean-Step 2: Peaceful statement. Participants spent 2 min on maintaining a peaceful state, wearing eye masks and earmuffs. Meanwhile, we used ErgoLAB to gather their BL values in this step.

> while, we used ErgoLAB to gather their Pr values in this step. Participants were required to wear earmuffs before soundscape exposure (Step 4). This contributed to a reduction in

> servation site for 5 min. Meanwhile, we used ErgoLAB to gather their Po values, and used sound-level meters to gather acoustical information. To focus on the physiological restorative role of soundscape, participants were required to wear eye masks during soundscape exposure. This contributed to a reduction in memory attenuation caused by visual

Step 5: Questionnaire process. Participants wearing earmuffs spent 1 min on filling

After this, we conducted tests to analyze reliability and validity for physiological and psychological parameters. Our results suggested that Cronbach's alpha of physiological and psychological parameters was 0.87, and Cronbach's alpha of each parameter ranged from 0.71 to 0.93. Then, we found that KMO of physiological and psychological parame-

To explore the physiological restorative role of soundscape in different forest structures, various statistical analyses were used. Pearson's correlation was conducted to analyze the relationship between acoustic parameters and perceived soundscape. T-test was conducted to analyze: 1) EMG, EDA, PPG, and RESP at tranquility and stress-inducement state; and 2) EMG, EDA, PPG, and RESP during pre-test and post-test in different forest structures. Principal components analysis (PCA) was used to determine the different contents of PRR in psychophysical and physiological parameters. The statistical analysis was carried out in SPSS 26.0. To forecast psychophysical parameters to physiological parame-

Figure 3 shows the distribution of soundscape pleasantness degree (PL) and acoustical parameters, as well as L10, L90, L10-L90 and LAeq, at the study sites. Figure 3a shows that

auditory short-term memory effects resulting from non-experimental procedures.

out questionnaires. We gathered the values of perceived soundscape in this step.

ters was more than 0.75, suggesting an acceptable reliability and validity [28].

ters, a back-propagation neural network was conducted using Matlab R2021a.

*3.1. Relationship between Perceived Soundscape and Acoustical Parameters*

distraction, and an increase in the level of their auditory attention [46].

2.4.2. Statistical Analyses

**3. Results**

Step 3: Stress inducement. We spent 5 min on inducing stress in participants. Meanwhile, we used ErgoLAB to gather their Pr values in this step. Participants were required to wear earmuffs before soundscape exposure (Step 4). This contributed to a reduction in auditory short-term memory effects resulting from non-experimental procedures.

Step 4: Soundscape exposure. Participants were exposed to the soundscape at an observation site for 5 min. Meanwhile, we used ErgoLAB to gather their Po values, and used sound-level meters to gather acoustical information. To focus on the physiological restorative role of soundscape, participants were required to wear eye masks during soundscape exposure. This contributed to a reduction in memory attenuation caused by visual distraction, and an increase in the level of their auditory attention [46].

Step 5: Questionnaire process. Participants wearing earmuffs spent 1 min on filling out questionnaires. We gathered the values of perceived soundscape in this step.

After this, we conducted tests to analyze reliability and validity for physiological and psychological parameters. Our results suggested that Cronbach's alpha of physiological and psychological parameters was 0.87, and Cronbach's alpha of each parameter ranged from 0.71 to 0.93. Then, we found that KMO of physiological and psychological parameters was more than 0.75, suggesting an acceptable reliability and validity [28].

#### 2.4.2. Statistical Analyses

To explore the physiological restorative role of soundscape in different forest structures, various statistical analyses were used. Pearson's correlation was conducted to analyze the relationship between acoustic parameters and perceived soundscape. T-test was conducted to analyze: (1) EMG, EDA, PPG, and RESP at tranquility and stress-inducement state; and (2) EMG, EDA, PPG, and RESP during pre-test and post-test in different forest structures. Principal components analysis (PCA) was used to determine the different contents of PRR in psychophysical and physiological parameters. The statistical analysis was carried out in SPSS 26.0. To forecast psychophysical parameters to physiological parameters, a back-propagation neural network was conducted using Matlab R2021a.

#### **3. Results**

#### *3.1. Relationship between Perceived Soundscape and Acoustical Parameters*

Figure 3 shows the distribution of soundscape pleasantness degree (PL) and acoustical parameters, as well as L10, L90, L10-L<sup>90</sup> and LAeq, at the study sites. Figure 3a shows that L<sup>10</sup> was distributed during interval [51.1, 62.2] dBA, and the distribution of PL was concentrated at response 4 ('very pleasant'). As L<sup>10</sup> further increased, the distribution of PL fluctuated from response 2 to 4 ('slightly pleasant' to 'very pleasant'). Figure 3b shows that L<sup>90</sup> was distributed during interval [43.0, 54.8] dBA, and the distribution of PL was during interval [3, 4]. As L<sup>90</sup> further increased, the distribution of PL fluctuated from response 2 to 3 ('slightly pleasant' to 'moderately pleasant'). Figure 3c shows that L10-L<sup>90</sup> was distributed during interval [5.3, 14.7] dBA, and the distribution of PL was concentrated at response 4 ('very pleasant'). As L<sup>90</sup> further increased, the distribution of PL fluctuated during interval [2, 3]. Furthermore, Figure 3d shows that LAeq was distributed during interval [44.7, 55.9] dBA, and the distribution of PL was concentrated at response 4 ('very pleasant'). The distribution of PL decreased from [3, 4] to [2, 3] as LAeq increased from [55.9, 62.9] dBA to more than 62.9 dBA.

In general, there was a negative tendency between PL and acoustical parameters. We conducted the Pearson correlation analysis to explore the different relationships between these parameters in different forest structures (See Table 1). Based on a total of 960 sets of data, our results showed that the value of perceived soundscape significantly correlated with all acoustical parameters in bamboo forests, and with L10, L<sup>90</sup> and L10-L<sup>90</sup> in other forest structures.

*Forests* **2022**, *13*, x FOR PEER REVIEW 6 of 14

from [55.9, 62.9] dBA to more than 62.9 dBA.

L10 was distributed during interval [51.1, 62.2] dBA, and the distribution of PL was concentrated at response 4 ('very pleasant'). As L10 further increased, the distribution of PL fluctuated from response 2 to 4 ('slightly pleasant' to 'very pleasant'). Figure 3b shows that L90 was distributed during interval [43.0, 54.8] dBA, and the distribution of PL was during interval [3, 4]. As L90 further increased, the distribution of PL fluctuated from response 2 to 3 ('slightly pleasant' to 'moderately pleasant'). Figure 3c shows that L10-L90 was distributed during interval [5.3, 14.7] dBA, and the distribution of PL was concentrated at response 4 ('very pleasant'). As L90 further increased, the distribution of PL fluctuated during interval [2, 3]. Furthermore, Figure 3d shows that LAeq was distributed during interval [44.7, 55.9] dBA, and the distribution of PL was concentrated at response 4 ('very pleasant'). The distribution of PL decreased from [3, 4] to [2, 3] as LAeq increased

**Figure 3.** Distribution of soundscape pleasantness level and (**a**) L10, (**b**)L90, (**c**) L10-L90 and (**d**) LAeq, in urban forests. **Figure 3.** Distribution of soundscape pleasantness level and (**a**) L10, (**b**)L90, (**c**) L10-L<sup>90</sup> and (**d**) LAeq, in urban forests.

conducted the Pearson correlation analysis to explore the different relationships between these parameters in different forest structures (See Table 1). Based on a total of 960 sets of **Table 1.** Relationship between acoustic parameters and perceived pleasantness in different forest structures, where Pearson correlation coefficients are shown in each cell.

In general, there was a negative tendency between PL and acoustical parameters. We


#### L10 −0.825 \*\* −0.666 \* −0.821 \*\* −0.689 \* L90 −0.562 \*\* −0.441 −0.314 −0.279 *3.2. Physiological Indicators in Different Forest Structures*

L10-L90 −0.720 \* −0.595 \* −0.967 \*\* −0.793 \*\* 3.2.1. Effect of Stress Inducement for Physiological Indicators

LAeq −0.847 \*\* −0.753 \*\* −0.676 \* −0.709 \* Table 2 shows the variations of electromyography (EMG), electrodermal activity (EDA), photoplethysmography (PPG) and respiration (RESP) at tranquility and stress inducement. Our results showed that the values of EMG, EDA, PPG, and RESP rose at stress inducement based on the difference between Pr and BL. This suggested that the process of stress inducement increased physiological activity such as prefrontal muscle contraction, vigorous activity of exocrine sweat glands, accelerated pulse, and shortness of breath. Furthermore, results of paired-sample t-tests showed significant changes for all physiological indicators, which suggested that the process of stress inducement was effective for physiological indicators in this study.


**Table 2.** The *t*-test of EMG, EDA, PPG, and RESP at tranquility and stress inducement.

\*\* *p* < 0.01.

#### 3.2.2. Variation Degree of Physiological Indicators

We conducted t-tests of EMG, EDA, PPG, and RESP in different forest structures based on stress inducement (see Table 3) during pre-test and post-test conditions. For EMG, EDA, and RESP, there were negative tendencies in different forest structures. These physiological indicators showed the most obvious drop in values while in bamboo forests. The decline of EMG and EDA were not obvious in broad-leaved forests. Furthermore, for the PPG of participants, results showed a negative tendency in bamboo forests.

**Table 3.** The *t*-test of EMG, EDA, PPG, and RESP at pre-test and post-test in different forest structures.


\* *p* < 0.05, \*\* *p* < 0.01; Pre-test value (Pr), Post-test value (Po), Po-Pr value (∆EMG, ∆EDA, ∆PPG, ∆RESP).

For ∆EMG and ∆EDA (See Figure 4), most participants recorded 'moderately restorative' in different forest structures. The total proportion of answers that included 'moderately restorative' and above was more than 65%. This suggested that all forest structures played a role in EMG and EDA for the participants. Furthermore, our results showed that 'slightly restorative' and 'not restorative at all' amounted to a small proportion of answers when in bamboo forests, compared to other forest structures. Few participants answered 'extremely restorative' in broad-leaved forests, which was consistent with the above results for the decline of EMG.

*Forests* **2022**, *13*, x FOR PEER REVIEW 8 of 14

ΔRESP).

sults for the decline of EMG.

\* *p* < 0.05, \*\* *p* < 0.01; Pre-test value (Pr), Post-test value (Po), Po-Pr value (ΔEMG, ΔEDA, ΔPPG,

For ΔEMG and ΔEDA (See Figure 4), most participants recorded 'moderately restorative' in different forest structures. The total proportion of answers that included 'moderately restorative' and above was more than 65%. This suggested that all forest structures played a role in EMG and EDA for the participants. Furthermore, our results showed that 'slightly restorative' and 'not restorative at all' amounted to a small proportion of answers when in bamboo forests, compared to other forest structures. Few participants answered 'extremely restorative' in broad-leaved forests, which was consistent with the above re-

**Figure 4.** Proportion of ΔEMG (**top left**), ΔEDA (**top right**), ΔPPG (**bottom left**), and ΔRESP (**bottom right**) in different forests Scheme 4. Our results showed that 'slightly restorative' and 'not restorative at all' accounted for more than 40% of responses in broad-leaved forests. In bamboo forests, we found that only 7% of participants answered 'slightly restorative' and 'not restorative at all' for ΔPPG, which suggested a consistency of above results for the decline of physiological indicators. Furthermore, our findings showed the restorative effect of RESP was limited in urban forests, with an average that was more than 38%. **Figure 4.** Proportion of ∆EMG (**top left**), ∆EDA (**top right**), ∆PPG (**bottom left**), and ∆RESP (**bottom right**) in different forests Scheme 4. Our results showed that 'slightly restorative' and 'not restorative at all' accounted for more than 40% of responses in broad-leaved forests. In bamboo forests, we found that only 7% of participants answered 'slightly restorative' and 'not restorative at all' for ∆PPG, which suggested a consistency of above results for the decline of physiological indicators. Furthermore, our findings showed the restorative effect of RESP was limited in urban forests, with an average that was more than 38%.

#### *3.3. Modelling the Physiological Restorative Role of Soundscape 3.3. Modelling the Physiological Restorative Role of Soundscape*

#### 3.3.1. Relationship between Psychophysical and Physiological Parameters 3.3.1. Relationship between Psychophysical and Physiological Parameters

Psychophysical and physiological datasets were combined to create a PRR model. The model could be applied to different forest structures to explore the relationship between psychophysical and physiological parameters. Principal components analysis (PCA) was conducted to reduce the dimensionality of Psychophysical and physiological datasets were combined to create a PRR model. The model could be applied to different forest structures to explore the relationship between psychophysical and physiological parameters.

psychophysical and physiological parameters, and to combine the original variables into potential restorative factors [47]. Table 4 shows the PCA results of the psychophysical and physiological dataset. Two components obtained by PCA showcased the differences between psychophysical and physiological parameters: component 1 showed that 66.22% of the variance in functional parameter was due to its large capacity for loading most of the psychophysical and physiological parameters; component 2 showed that 21.68% of the variance in background sound was due to a high factor of loading LAeq and L90. As restorative factors for the public, these components affected human perception and response to the soundscape in different forest structures. Thus, we suggested a potential interaction between psychophysical parameters and physiological parameters. Principal components analysis (PCA) was conducted to reduce the dimensionality of psychophysical and physiological parameters, and to combine the original variables into potential restorative factors [47]. Table 4 shows the PCA results of the psychophysical and physiological dataset. Two components obtained by PCA showcased the differences between psychophysical and physiological parameters: component 1 showed that 66.22% of the variance in functional parameter was due to its large capacity for loading most of the psychophysical and physiological parameters; component 2 showed that 21.68% of the variance in background sound was due to a high factor of loading LAeq and L90. As restorative factors for the public, these components affected human perception and response to the soundscape in different forest structures. Thus, we suggested a potential interaction between psychophysical parameters and physiological parameters.

**Table 4.** Summary of principal component analysis (PCA) on physiological restorative parameters.


#### 3.3.2. Back-Propagation Neural Network for PRS Model

In our physiological restorativeness soundscape model (PRS model), 960 samples were used and divided into three randomly chosen sets: training set (672 samples, 70.0%), test set (144 samples, 15.0%) and validation set (144 samples, 15.0%). Three-fold cross validation was also conducted.

Table 5 shows the accuracy of the PRS model based on the validation set. The accuracy percentage of both the training and testing sets was more than 90%. After training the PRS model, classified soundscape data results indicated that the accuracy of ∆EMG, ∆EDA, ∆PPG, ∆RESP, and PR were 81.1%, 87.2%, 95.6%, 92.9% and 86.2%, respectively. As shown in Figure 5, ∆EMG, ∆EDA, and ∆RESP maintained stable accuracy during interval [−2.5, 2.5], while ∆PPG and PRR maintained stable accuracy during interval [−5.0, 10.0] and [−2.5, 3.5], respectively. Accuracy decreased when values were outside these intervals.

**Table 5.** Accuracy of parameters in the PRR model based on the validation set.


**Figure 5.** Prediction error for (**a**) ΔEMG, (**b**) ΔEDA, (**c**) ΔPPG, (**d**) ΔRESP and (**e**) PRR in ANNs. **Figure 5.** Prediction error for (**a**) ∆EMG, (**b**) ∆EDA, (**c**) ∆PPG, (**d**) ∆RESP and (**e**) PRR in ANNs.

**Figure 6.** Importance of input indicators in determining the output value of soundscape.

When Figure 3a,d were combined, our findings showed that LAeq and L10 were important drivers that influenced questionnaire responses in urban forests. This was similar to previous studies conducted in urban green areas [48,49]. We also found that LAeq and L10 displayed overlapping intervals, since there were fewer sources of mechanical noise in urban forests and animals did not need to raise their volume to communicate [50]. As our observation sites were in areas with some tourist activity, the maximum LAeq was higher

Figure 6 showcases the importance of input variables for determining outputs. Re-

physiological parameters.

**4. Discussion**

Figure 6 showcases the importance of input variables for determining outputs. Results showed that PL impacted accuracy the most and accounted for more than 35% of the independent variable importance. This suggested that perception was a main driver for physiological parameters. sults showed that PL impacted accuracy the most and accounted for more than 35% of the independent variable importance. This suggested that perception was a main driver for physiological parameters.

**Figure 5.** Prediction error for (**a**) ΔEMG, (**b**) ΔEDA, (**c**) ΔPPG, (**d**) ΔRESP and (**e**) PRR in ANNs.

Figure 6 showcases the importance of input variables for determining outputs. Re-

*Forests* **2022**, *13*, x FOR PEER REVIEW 10 of 14

(**a**) (**b**) (**c**)

(**d**) (**e**)

**Figure 6.** Importance of input indicators in determining the output value of soundscape. **Figure 6.** Importance of input indicators in determining the output value of soundscape.

#### **4. Discussion 4. Discussion**

When Figure 3a,d were combined, our findings showed that LAeq and L10 were important drivers that influenced questionnaire responses in urban forests. This was similar to previous studies conducted in urban green areas [48,49]. We also found that LAeq and L10 displayed overlapping intervals, since there were fewer sources of mechanical noise in urban forests and animals did not need to raise their volume to communicate [50]. As our observation sites were in areas with some tourist activity, the maximum LAeq was higher When Figure 3a,d were combined, our findings showed that LAeq and L<sup>10</sup> were important drivers that influenced questionnaire responses in urban forests. This was similar to previous studies conducted in urban green areas [48,49]. We also found that LAeq and L<sup>10</sup> displayed overlapping intervals, since there were fewer sources of mechanical noise in urban forests and animals did not need to raise their volume to communicate [50]. As our observation sites were in areas with some tourist activity, the maximum LAeq was higher when compared to previous studies [51], which suggests more various soundscape exposure contributing to more significant PRR in our research. Furthermore, our results showed a similarity between the distribution of L<sup>90</sup> and L10-L<sup>90</sup> (see Figure 3b,c). Most of these values fluctuated and dropped after values reached the threshold, which suggested that the threshold may conduct enough physiological and psychological phenomena on an individual to change them from a steady state to an excited state. Table 4 showed that L<sup>90</sup> and L10-L<sup>90</sup> influenced the data for the parameters of physiological restoration from different components. Therefore, there should be more attention and control over L<sup>90</sup> and L10-L90, especially in forest-based health care [52]. In practice, L<sup>90</sup> that is lower than 54.8 dBA is beneficial in the creation of quiet areas in urban areas, while L10-L<sup>90</sup> lower than 14.7 dBA is effective in weakening the negative effects of eventual and unexpected sound events [53,54].

Our findings showed that EMG was the most sensitive physiological parameter in our data set (see Table 3 and Figure 4). This suggested an optimal effect of physiological restorativeness because of the dual sensory channels of input and output in the muscular system, contributing to the cognition and response of participants in forest soundscapes [55,56]. However, the accuracy of EMG was lower than the other parameters in the artificial neural network (ANN). This suggested that EMG was influenced by other environmental drivers and the individual's senses. We also found that the coefficients ranking of parameters in the principal components analysis (PCA) and the accuracy ranking of parameters in the ANN testing group were potentially consistent (see Tables 4 and 5): ∆EMG > ∆EDA > ∆RESP > ∆PPG. As shown in Figure 5, our ANN results suggested that L10–90 played the most important role of all physical parameters in determining the output value of soundscape. L10–90 was also consistent with the highest coefficient in the functional parameter. These results suggested that PCA can be used as a pre-experiment method for the creation of a model for the physiological restorative role of soundscape (PRR model). Figure 6 showed that soundscape questionnaire responses were the most important for input indicators and suggested that physiological responses were based on the cognitive basis of soundscape [57]. Furthermore, soundscape pleasantness contributed

to the enhancement of attention to sound sources [58], which promoted a physiological response. Thus, we found that questionnaires were one of the most important methods for gathering physiological information. Furthermore, previous research suggests different absorption and radiation of the leaves and woods among different forest structures [55,59]. This helps us to understand the physiological restorative role of soundscapes in different forest structures strategies when proposing suitable forest-based health care.

In our study, some limitations may have been presented. Although we tried to minimize the effect of vision on physiological responses by using eye masks, participants were potentially affected by somatosensory effects including the variation of temperature and humidity. Additionally, audio-visual interaction was expected in urban forests, but we did not consider audio-visual drivers in this study.

#### **5. Conclusions**

Urban forested areas contribute to favorable exposure to the biophilic outdoor environment, which is beneficial to public health and recovery. This study revealed that psychophysical parameters jointly function in the physiological restorative role of soundscape in urban forested areas. Our findings showed that: (1) LAeq and L<sup>10</sup> were important drivers that influence questionnaire responses; (2) EMG was the most sensitive physiological parameter; and (3) L10-L<sup>90</sup> played the most important role of all physical parameters in the PRS model.

We suggest that the biophilic outdoor environment may offer physiological restorative potential for therapy after COVID-19. Furthermore, other potential drivers such as audiovisual interaction in forested areas may be considered in future studies to further explore physiological restorative patterns in different forest structures.

**Author Contributions:** Conceptualization, X.-C.H. and S.C.; methodology, X.-C.H. and J.L.; software, E.D.; formal analysis, X.-C.H. and J.-B.W.; investigation, X.-C.H. and S.C.; data curation, X.-C.H.; writing—original draft preparation, X.-C.H., S.C. and J.L.; writing—review and editing, E.D. and J.-B.W.; supervision, Y.C. All authors have read and agreed to the published version of the manuscript.

**Funding:** This project was supported by the National Natural Science Foundation of China (52208052, 51838003), the National Key Research & Development Program of China (2019YFD1100405), and the Program of Humanities and Social Science Research Program of Ministry of Education of China (Grant No. 21YJCZH038).

**Informed Consent Statement:** Informed consent was obtained from all subjects involved in the study.

**Conflicts of Interest:** The authors declare no conflict of interest.

## **References**


**Jian Xu 1,2,3,4, Muchun Li 1,\*, Ziyang Gu <sup>1</sup> , Yongle Xie <sup>1</sup> and Ningrui Jia <sup>1</sup>**


**Abstract:** The purpose of this study is to explore the audio-visual preferences of exercisers in urban forest parks in China and to make practical suggestions for park landscape design. Taking Beigushan Forest Park in Lianyungang City, Jiangsu Province as a case, based on field research and questionnaire survey, this study analyzed the audio-visual preference characteristics of exercisers in the park, revealed the correlation between audio-visual preference and exercisers' behaviors and individual characteristics, and explored the influence of audio-visual preferences on exercise feelings by establishing a structural equation model. It was found that (1) the forest and its avenue landscape and birdsong are most preferred by exercisers; (2) the audio-visual preferences of people with different exercise forms differ, for example, people who slowly walk, run, and briskly walk have stronger preferences for natural soundscape and visual landscape, while people who use fitness equipment have stronger inclusiveness for human activity sound and prefer public facility-based landscapes. In addition, some individual characteristics such as exercise intensity and exercise frequency significantly affect exercisers' audio-visual preferences; (3) visual landscape preferences have a greater direct impact on exercise feelings, with natural waterscape having the greatest direct impact, but overall soundscape preferences do not have a high degree of direct impact on exercise feelings, with natural sound still having a strong positive impact. These findings provide a more quantitative basis for the landscape design of urban forest parks from the perspective of exercise behavior.

**Keywords:** urban forest park; exercise behaviors; audio-visual preferences; correlation analysis

## **1. Introduction**

As a new type of park arising from urbanization in the era of ecological civilization [1], urban forest parks have multiple functions such as recreation, recuperation, and avoiding the heat, which help improve air quality [2], reduce noise [3], and provide a pleasant environment for people to promote physical and mental health development [4]. Forest parks can meet the needs of urban residents for environment, leisure, and sports, thus attracting an increasing number of urban residents for physical exercise close to nature [5]. In China, where urbanization is accelerating, the relationship between the health level and quality of life of urban residents and urban forest parks has become increasingly important [6], especially in the context of COVID-19, where the demand for parks and outdoor green spaces has increased rather than decreased [7]. Studies have shown that physical exercise in green spaces can release stress and enhance the body's ability to fight infectious diseases [7–9]. Therefore, it can be predicted that in the future, the demand for forest parks and exercise activities will continue to grow and more urban residents can

**Citation:** Xu, J.; Li, M.; Gu, Z.; Xie, Y.; Jia, N. Audio-Visual Preferences for the Exercise-Oriented Population in Urban Forest Parks in China. *Forests* **2022**, *13*, 948. https://doi.org/ 10.3390/f13060948

Academic Editors: Xin-Chen Hong, Jiang Liu and Guang-Yu Wang

Received: 21 May 2022 Accepted: 14 June 2022 Published: 17 June 2022

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2022 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

benefit from the services of forests and parks [10]. In this context, exercisers in urban forest parks and their audio-visual preferences become the focus of this study.

In the studies on exercise behavior in urban forest parks, many scholars have devoted their research to exploring the main factors that influence the attractiveness and satisfaction of exercise in parks. For example, Li et al. found that open activity space with waterscape, landscape sketch, can attract more people toward exercise activities [11]; McCormack et al. summarized qualitative studies about the relationship between park use and physical exercise and found that safety, aesthetics, park facilities, and landscape maintenance were important factors influencing satisfaction with park use [12]. Although some scholars have also emphasized the importance of individual perceptual factors in influencing park use satisfaction [13,14], the majority of scholars have focused on objective factors such as facilities and public services provided by parks, and have not paid enough attention to people's underlying psychological motivations and preferences. Therefore, an investigation of people's exercise satisfaction from the perspective of their visual and auditory preferences at the psychological level would provide a relevant complement to the current research on the factors influencing exercise satisfaction in parks.

People's perceptions of the landscape initially originate from human intuitive experiences, in which people rely on their eyes to obtain 87% of the information from the outside world, and 75–90% of human activities is visually induced [15]. Studies have shown that individual visual aesthetic preferences influence people's perceptions of ecological and aesthetic values, which further influence their behavioral choices. [16]. For example, Ma et al. explored the influence of the degree of visual landscape heterogeneity on landscape aesthetic quality and public visual perception effects [17]; Zhang et al. used eye-tracking to explore the visual preferences of different types of visitors to trail landscapes and revealed the reasons for the differences in visitors' landscape gaze time [18]. Visual factors also affect people's perception and evaluation of soundscapes, and the relationship between these two is the focus of this study. In this regard, many scholars have made significant research contributions. Cassina et al., proposed a linear model for predicting perceived tranquility in different environments based on visual and acoustic features [19]. Romero et al. found that visual factors such as ocean visibility can affect the perception of the soundscape quality in the areas with road traffic [20], and they also found there are color associations between people and different urban soundscapes [21]. Preis et al. found that the addition of visual information increases the noise annoyance assessment [22]. Moreover, numerous studies have also demonstrated that people's visual preferences affect the perception of landscape and environmental behavior [23,24].

Soundscape is the acoustic environment perceived by an individual, group, or community in a given scene [25] and is highly relevant to people's health [26]. With the increasing concern about health, urbanization, and globalization, more and more studies are focusing on soundscape. It has been shown that people's perceptions and preferences of soundscapes can play a key role in the construction of related landscapes in urban forest parks [27]. Currently, most of the soundscape studies focus on people's perception. Among them, scholars have found that people's perception of soundscape is related to the type of soundscape, people's personal preferences and sensitivities, and demographic indicators related to soundscape [28–32]. For example, Fang et al. found that five main dimensions of social, demographic, and behavioral attributes (age and familiarity of site, educational and economic condition, companion and type of recreational use, gender, and length of stay) were associated with people's soundscape perceptions and preferences [33]. In addition, scholars have expanded their research in related fields, such as Hong et al. who analyzed the relationship between each soundscape element that has an impact on forest park soundscapes and its physical stimulus amount and people's soundscape preferences [34]. Subdivided into the field of soundscape preference research, some scholars have found that natural sounds are more preferred by people [35]; some scholars have put people's soundscape preferences in the context of COVID-19 and found that individual characteristics such as age, occupation, education level, and life happiness are the

main factors affecting soundscape preferences [36]; in addition, the frequency of visits to destinations also affects people's preferences for beautiful soundscapes [37]. Although the above studies have addressed different influencing factors of soundscape preference, few studies have focused on soundscape preference among a population with specific behaviors; therefore, the variability in soundscape preference cannot be explained in a more behavioral characteristic sense.

At present, many research results in the field of audio-visual perception are directed to the applied science fields such as medicine and engineering, but there are still few reports in the natural science fields such as landscape and ecology, as well as natural and social interdisciplinary subjects. Among the studies on landscape and ecological environment that focus on audio-visual perception, there are mostly studies on people's single-sensory perception and preference, but there is a lack of studies on multisensory preference and its interaction. Therefore, this study investigates the audio-visual preferences of the exercisers, which will be helpful to explore and improve the research on the audio-visual field of the exercisers in urban forest parks. In this study, 406 exercisers in Beigushan Forest Park, Lianyungang, Jiangsu Province, were surveyed according to the subjective evaluation method. The purpose of this study is to explore the audio-visual preferences of exercisers in China's urban forest parks, and to reveal the correlation between these preferences and exercisers' behaviors and individual characteristics, so as to put forward practical suggestions for park landscape design. Unlike most previous studies, this study is novel in that it focuses on audio-visual preferences among a specific behavioral population and reveals differences in audio-visual preferences from an environmental behavioral perspective. However, the restrictions on pedestrian flow in the park under the influence of the epidemic and the impact of some precautionary measures on people's landscape evaluation pose certain challenges to this study. Overall, the study helps to maximize the usefulness of natural resources and provide auxiliary visual and acoustic landscape design for urban forest park designers and planners, while providing better exercise experience for exercise groups and improving people's quality of life and happiness.

This paper is divided into four parts. The method part mainly introduces the study area, questionnaire design, and field research. In the result part, firstly, the reliability and validity test results of the collected questionnaires and the statistical results of the respondents' personal characteristics are analyzed, then the audio-visual preference characteristics of the exercisers are revealed, and the correlation between the audio-visual preferences and exercisers' behaviors as well as their personal characteristics is discussed through correlation analysis. Finally, the influence of audio-visual preferences on the exercise feeling is explored by establishing a structural equation model. On the basis of comparing and summarizing the similarities and differences between this study and the existing scholars' research, this discussion section explores the landscape design of urban forest parks from the perspective of exercise behavior. In the conclusion section, the full text and its important points are reviewed, while the limitations and future work of this study are also summarized.

This study explores the following issues to be addressed:


#### **2. Materials and Methods**

#### *2.1. Study Area*

Beigushan Forest Park in Lianyungang City, Jiangsu Province, China, was selected as the case site for the study. The location map of Beigushan Forest Park is shown in Figure 1. The park has three major ecosystems: marine, forest, and wetland, with a forest coverage of 86.3%. Beigushan Forest Park is a mountainous forest park, which is the most common type of forest park in China. The average annual temperature is 14 ◦C, which is similar to China's

average annual temperature of 13 ◦C. The park has good accessibility and experiences strong demand from residents; the trail around the mountain was officially opened to the public in March 2017 with new facilities. It is 5000 m in length, which circles around Beigu Mountain, and won the "2017 Jiangsu Most Beautiful Running Route" award. The exercisers account for 19% of visitors to this park, which is comparable to the percentage of Chinese nationals exercising in urban forest parks [38], and is highly representative as a case study for this study. ilar to China's average annual temperature of 13 °C. The park has good accessibility and experiences strong demand from residents; the trail around the mountain was officially opened to the public in March 2017 with new facilities. It is 5000 m in length, which circles around Beigu Mountain, and won the "2017 Jiangsu Most Beautiful Running Route" award. The exercisers account for 19% of visitors to this park, which is comparable to the percentage of Chinese nationals exercising in urban forest parks [38], and is highly representative as a case study for this study.

Beigushan Forest Park in Lianyungang City, Jiangsu Province, China, was selected as the case site for the study. The location map of Beigushan Forest Park is shown in Figure 1. The park has three major ecosystems: marine, forest, and wetland, with a forest coverage of 86.3%. Beigushan Forest Park is a mountainous forest park, which is the most common type of forest park in China. The average annual temperature is 14 °C, which is sim-

*Forests* **2022**, *13*, x FOR PEER REVIEW 4 of 26

**Figure 1.** Geographical location of Beigushan Forest Park in Lianyungang. **Figure 1.** Geographical location of Beigushan Forest Park in Lianyungang.

Before conducting the questionnaire survey, through fieldwork, Beigushan Forest Park can be divided into a plaza and artificial building area, a hillside and waterscape Before conducting the questionnaire survey, through fieldwork, Beigushan Forest Park can be divided into a plaza and artificial building area, a hillside and waterscape area, and a mountain rim trail area. There are 17 main types of landscapes in Beigushan Forest Park, including natural waterscape (streams, ponds, and lakes), topographic landscape

(lawns, avenues, hillsides, and lakesides), natural vegetation (shrubs, ornamental flowers, and forests), artificial landscape (rockery, parterres, fountains, sculptures, bridges, and pavilions), artificial facilities (fitness equipment, squares, and public buildings). We often hear 16 different kinds of sounds, including natural sounds (sound of wind, birdsong, cry of insects, rustle of leaves, and water flow sound), human activity sounds (conversational voice, sound of children playing, footstep, and exercise sound), and artificial sounds (traffic sound, entertainment sound, device music, construction noise, machine noise, and broadcast).

## *2.2. Questionnaire Design*

Respondents were asked to fill out the questionnaires created by "Questionnaire Star" using the tablet PCs provided to them by the researchers ("Questionnaire Star" is a professional, unlimited free online questionnaire, assessment, voting platform, focusing on providing users with a powerful, user-friendly online questionnaire design; free to use the program, it provides powerful, fast, easy to use, and low-cost obvious advantages [39]. The questionnaire star program has released a total of 154 million questionnaires, which can fully meet the number of questionnaire research and question type setting requirements; https://www.wjx.cn/, accessed on 25 January 2022). In addition, the researchers also prepared a certain number of paper questionnaires for the elderly who cannot use electronic devices skillfully.

The questionnaire consisted of four parts, with 18 questions in total. The first part was designed to collect demographic information about the respondents, such as age, gender, number of participants in exercise activities, distance of residence from the target park, and activities performed in the park other than exercise; the second part focused on the exercise profile of the respondents, including exercise mode, exercise time, exercise duration, exercise frequency, driving factors for exercising, reasons for choosing the park as an exercise site, specific location of exercise, and exercise frequency. The second part focused on the respondents' exercise patterns, exercise time, exercise duration, exercise frequency, exercise site, exercise intensity, overall feelings of exercise, and willingness to exercise in parks in the future. In the third and fourth sections, the three types that constitute a soundscape as defined by Kraus [40] (abiotic natural sounds from the physical environment, nonhuman biological sounds emitted by all organisms in a given habitat, and anthropomorphic sounds emitted by stationary and moving man-made objects) were used as the basis for classifying the types of soundscapes in the questionnaire. On this basis, the scales used in the study of landscape perception by scholars such as Zheng Zhao [41] and the scales involved in the study of urban forest park soundscape by Wei Zhao [42] and Banu Chitra [43], respectively, were used, and the scales used in this study were appropriately adjusted and modified by combining the ISO [44] definition and classification of soundscape and the actual situation of Beigushan Forest Park in Lianyungang. The third part focused on the understanding of individual visual landscape preferences, mainly using a five-point Likert scale (strongly dislike (−2), dislike (−1), average (0), like (1), and like very much (2)) to illustrate their overall preferences for visual landscapes. In the fourth section, a selection of the frequency of occurrence of 16 soundscapes and a five-point rating of the soundscapes (from very dislike (−2) to very like (2)) were included to illustrate the overall preference of respondents for common soundscapes. After a pre-research test with 20 people, the average response time was 4 min and 39 s, all of whom had no objections to answer the questionnaire questions. The relevant contents of the questionnaire are shown in Table 1.


**Table 1.** Exerciser landscape and soundscape preference system.

## *2.3. Field Research*

The sampling sites covered three subdivisions in the park, including three sampling sites in the mountain rim trail area, two sampling sites in the hillside and water area, and one sampling site in the plaza and artificial building area, for a total of six sampling sites, the specific locations of which are shown in Figure 2. The survey was conducted during the daytime in April 2021 under sunny weather, and each survey lasted for eight hours (from about 9:00 to 17:00); when the average temperature is about 17 ◦C, the climate is suitable, the vegetation is abundant, the residents are willing to travel more, and the number of exercise activities is higher. Before each part of research, attention was paid to temperature, relative humidity and wind speed, and similar weather was used for the research to avoid differences in audio-visual preferences of respondents due to climate effects. Additionally, to reduce any bias due to the selection of respondents at a specific time, each sampling site was surveyed twice on different days. The research was conducted using anonymous random interviews, where respondents were first explained the purpose and procedures of the survey, which did not mention positive or negative sounds, noise pollution, etc. They were then informed that their responses would be anonymous. To avoid distractions from other participants during their stay on the site, those who wished to participate in the survey were given a tablet containing the questionnaire and invited to fill it out individually. Respondents were invited to go to a secluded place near their location while

the questionnaire was being filled out, and the ambient sound was tested using a decibel meter to ensure that there were no significant sound disturbances in the surroundings. Due to the short duration of the questionnaire and the fact that only a limited number of sounds may be present during a given time period, participants were asked to respond for the length of time chosen in the questionnaire, based on their long-term experience in the park. In addition to this, for the issue of hearing impairment, specific questioning was conducted prior to the study and observations were made to ensure that respondents did not have any significant hearing impairment. near their location while the questionnaire was being filled out, and the ambient sound was tested using a decibel meter to ensure that there were no significant sound disturbances in the surroundings. Due to the short duration of the questionnaire and the fact that only a limited number of sounds may be present during a given time period, participants were asked to respond for the length of time chosen in the questionnaire, based on their long-term experience in the park. In addition to this, for the issue of hearing impairment, specific questioning was conducted prior to the study and observations were made to ensure that respondents did not have any significant hearing impairment.

sounds, noise pollution, etc. They were then informed that their responses would be anonymous. To avoid distractions from other participants during their stay on the site, those who wished to participate in the survey were given a tablet containing the questionnaire and invited to fill it out individually. Respondents were invited to go to a secluded place

*Forests* **2022**, *13*, x FOR PEER REVIEW 7 of 26

**Figure 2.** Sampling point distribution map. **Figure 2.** Sampling point distribution map.

In order to reduce the influence of the order effect on the accuracy of the questionnaire results due to the single form of questions, the researcher randomly switched the order of scoring questions to improve the accuracy of the questionnaire results before interviewing the respondents, and used the method of setting irrelevant interfering items to filter the questionnaire (eliminating the questionnaire with irrelevant interfering items), so as to guarantee the authenticity and credibility of the questionnaire results to the greatest extent. For the final collection of 406 questionnaires, the number of actual valid ques-In order to reduce the influence of the order effect on the accuracy of the questionnaire results due to the single form of questions, the researcher randomly switched the order of scoring questions to improve the accuracy of the questionnaire results before interviewing the respondents, and used the method of setting irrelevant interfering items to filter the questionnaire (eliminating the questionnaire with irrelevant interfering items), so as to guarantee the authenticity and credibility of the questionnaire results to the greatest extent. For the final collection of 406 questionnaires, the number of actual valid questionnaires was 344, and the questionnaire efficiency was 84.7%.

tionnaires was 344, and the questionnaire efficiency was 84.7%. The research framework is shown in Figure 3. The research framework is shown in Figure 3.

**Figure 3.** Research framework. **Figure 3.** Research framework.

#### **3. Results**

#### **3. Results**  *3.1. Data Testing and Demographic Analysis*

*3.1. Data Testing and Demographic Analysis*  After the 344 valid questionnaires were collected and sorted, SPSS 26.0 software was used to test the reliability and validity of the questionnaire data. In this study, Cronbach's alpha was used to analyze the reliability of the questionnaire, and α ≥ 0.7 represents reliable results [46]. The results of the questionnaire were calculated to meet this reliability criterion: natural water features (0.8), topographic landscapes (0.862), natural vegetation (0.772), artificial landscapes (0.891), and public facilities (0.849); natural sounds (0.839), activity sounds (0.936), and artificial sounds (0.921). The reliability of the overall perception factor was 0.914. Thus, it can be seen that the reliability of the questionnaire meets the survey requirements. In this study, the validity of KMO was tested by factor analysis, and KMO = 0.92, which satisfied the condition of factor analysis (KMO ≥ 0.6) (Lu, 2004), indicating that the validity of the questionnaire also met the requirements. Bartlett's ball test approximated a chi-square value of 9331.808, corresponding to a probability value of 0.000 After the 344 valid questionnaires were collected and sorted, SPSS 26.0 software (IBM, Armonk, NY, USA) was used to test the reliability and validity of the questionnaire data. In this study, Cronbach's alpha was used to analyze the reliability of the questionnaire, and α ≥ 0.7 represents reliable results [46]. The results of the questionnaire were calculated to meet this reliability criterion: natural water features (0.8), topographic landscapes (0.862), natural vegetation (0.772), artificial landscapes (0.891), and public facilities (0.849); natural sounds (0.839), activity sounds (0.936), and artificial sounds (0.921). The reliability of the overall perception factor was 0.914. Thus, it can be seen that the reliability of the questionnaire meets the survey requirements. In this study, the validity of KMO was tested by factor analysis, and KMO = 0.92, which satisfied the condition of factor analysis (KMO ≥ 0.6), indicating that the validity of the questionnaire also met the requirements. Bartlett's ball test approximated a chi-square value of 9331.808, corresponding to a probability value of 0.000 (*p* < 0.01), indicating that the questionnaire measures significant correlation of the question items and that the data are valid.

(*p* < 0.01), indicating that the questionnaire measures significant correlation of the question items and that the data are valid. The personal characteristics of the respondents are shown in Figure 4. The proportions of respondents were 48% and 52% for men and women, respectively, which were relatively equal Figure 4(a), but there were fewer respondents over the age of 60, and the respondents were mainly the young and middle-aged group Figure 4(b). In terms of travel mode Figure 4(c), respondents traveled in a variety of ways, and exercising with three to five friends was the composition of the largest number of exercisers. In terms of the distance of the respondents' addresses from the park Figure 4(d), 500-3000 m (36% for 500- 1500 m and 28% for 1500-3000) accounted for the majority, and the majority of exercisers were living near the park. Figure 4(e) About 58% of the respondents exercised for 1-2 h, and the overall frequency of exercise was low Figure 4(f), generally concentrated on once a month, 2-3 times a month and 1-2 times a week, and more respondents (39%) exercised The personal characteristics of the respondents are shown in Figure 4. The proportions of respondents were 48% and 52% for men and women, respectively, which were relatively equal (Figure 4a), but there were fewer respondents over the age of 60, and the respondents were mainly the young and middle-aged group (Figure 4b). In terms of travel mode (Figure 4c), respondents traveled in a variety of ways, and exercising with three to five friends was the composition of the largest number of exercisers. In terms of the distance of the respondents' addresses from the park (Figure 4d), 500–3000 m (36% for 500–1500 m and 28% for 1500–3000) accounted for the majority, and the majority of exercisers were living near the park (Figure 4e) About 58% of the respondents exercised for 1–2 h, and the overall frequency of exercise was low (Figure 4f), generally concentrated on once a month, 2–3 times a month and 1–2 times a week, and more respondents (39%) exercised during the time period of 18:00–21:00 (Figure 4g), with 61% of respondents only exercising lightly (Figure 4h), exercisers generally exercised less intensely.

during the time period of 18:00-21:00 Figure 4(g), with 61% of respondents only exercising

lightly Figure 4(h), exercisers generally exercised less intensely.

*Forests* **2022**, *13*, x FOR PEER REVIEW 9 of 26

**Figure 4.** Statistical results of respondents' personal characteristics (**a–h**: gender, age, companion, the distance from home to the park, duration, frequency, and the time period of exercise and exercise intensity). **Figure 4.** Statistical results of respondents' personal characteristics (**a**–**h**): gender, age, companion, the distance from home to the park, duration, frequency, and the time period of exercise and exercise intensity. the distance from home to the park, duration, frequency, and the time period of exercise and exercise intensity). *3.2. Audio-visual Preference Characteristics of Exercisers in Urban Forest Park* 

#### *3.2. Audio-visual Preference Characteristics of Exercisers in Urban Forest Park 3.2. Audio-Visual Preference Characteristics of Exercisers in Urban Forest Park* As shown in Figure 5, the most preferred visual landscapes for people exercising in

As shown in Figure 5, the most preferred visual landscapes for people exercising in urban forest parks are, in order of preference: avenues, forests, streams, lakesides, bridges and pavilions, ornamental flowers, and lawns, and less preferred landscapes are rockery, public buildings, and sculptures. In terms of overall categories, people prefer topographic landscapes and natural landscapes, and have a lower preference for artificial landscapes. As shown in Figure 5, the most preferred visual landscapes for people exercising in urban forest parks are, in order of preference: avenues, forests, streams, lakesides, bridges and pavilions, ornamental flowers, and lawns, and less preferred landscapes are rockery, public buildings, and sculptures. In terms of overall categories, people prefer topographic landscapes and natural landscapes, and have a lower preference for artificial landscapes. urban forest parks are, in order of preference: avenues, forests, streams, lakesides, bridges and pavilions, ornamental flowers, and lawns, and less preferred landscapes are rockery, public buildings, and sculptures. In terms of overall categories, people prefer topographic landscapes and natural landscapes, and have a lower preference for artificial landscapes.

**Figure 5.** Average landscape preferences of exercisers in urban forest park. **Figure 5.** Average landscape preferences of exercisers in urban forest park. **Figure 5.** Average landscape preferences of exercisers in urban forest park.

As shown in Figure 6, the most preferred sounds for the exercisers were birdsong, water flow sound, and rustling of leaves, and the least preferred sounds for landscapes were construction noise, machine noise, traffic sounds, and broadcasts. The exercise population prefers nature-related landscapes and soundscapes more, and were less fond of sounds and landscapes generated or created by people. As shown in Figure 6, the most preferred sounds for the exercisers were birdsong, water flow sound, and rustling of leaves, and the least preferred sounds for landscapes were construction noise, machine noise, traffic sounds, and broadcasts. The exercise population prefers nature-related landscapes and soundscapes more, and were less fond of As shown in Figure 6, the most preferred sounds for the exercisers were birdsong, water flow sound, and rustling of leaves, and the least preferred sounds for landscapes were construction noise, machine noise, traffic sounds, and broadcasts. The exercise population prefers nature-related landscapes and soundscapes more, and were less fond of sounds and landscapes generated or created by people.

sounds and landscapes generated or created by people.

tivity sound

Artificial sound

Entertainment

Construction

**Figure 6.** Average soundscape preferences of exercisers in urban forest park. **Figure 6.** Average soundscape preferences of exercisers in urban forest park.

*3.3. The Effect of Exercisers' Exercise Style and Venue Choice on Audio-visual Preference 3.3. The Effect of Exercisers' Exercise Style and Venue Choice on Audio-Visual Preference* 3.3.1. Effects of Exercise Modality and Exercise Site Selection on Soundscape Preference

3.3.1. Effects of Exercise Modality and Exercise Site Selection on Soundscape Preference As the results of the correlation analysis (Table 2) show, different exercise methods affect people's preference for soundscape. Those who chose jogging and brisk walking had a higher potential preference for natural sounds, while this group tended to choose exercise sites that were close to the natural landscape in the form of lawn, forest, and As the results of the correlation analysis (Table 2) show, different exercise methods affect people's preference for soundscape. Those who chose jogging and brisk walking had a higher potential preference for natural sounds, while this group tended to choose exercise sites that were close to the natural landscape in the form of lawn, forest, and lakeside footpath.


lakeside footpath. **Table 2.** Correlation between different exercise forms and soundscape preferences.

playing 0.024 −0.009 0.009 0.075 −0.016 −0.048 −0.025 0.020 \* Correlation is significant at the 0.05 level. \*\* Correlation is significant at the 0.01 level.

Footsteps 0.046 0.058 −0.004 0.013 −0.063 −0.141\*\* −0.073 0.014 Exercise sound −0.042 0.101 −0.009 0.096 0.059 −0.029 0.096 0.036 Traffic sound 0.028 −0.033 0.005 0.015 −0.066 −0.019 −0.039 0.017 sound −0.017 −0.024 0.008 0.017 −0.031 0.059 −0.005 −0.024 Since the venues for dance and gymnastics are limited by the mountainous terrain and are far from the mountains and forests, those who choose these activities have less exposure to natural sounds [47]. There is a positive correlation between the preference for device music and the crowd of square dancers and gymnasts, but they have a stronger aversion to activity sound and artificial noise. People using fitness equipment showed a higher acceptance for activity sound on a potential level.

Device music −0.069 −0.051 −0.016 0.005 0.019 0.033 −0.012 0.017 noise 0.043 −0.007 0.000 −0.010 −0.148 \*\* 0.012 0.021 0.019 The data in Table 3 indicate that there is a positive correlation between the choice of footpath in forests and lawns and preference for natural sounds, especially sounds in the forest, and a negative correlation between preference for activity sounds and artificial noise.

\* Correlation is significant at the 0.05 level. \*\* Correlation is significant at the 0.01 level.

Machine noise 0.020 0.006 −0.037 0.012 −0.095 −0.005 −0.001 0.056

There was a positive correlation between the choice of lakeside footpath and preference for water flow sounds. People exercising in square open spaces and fitness equipment venues had a higher tolerance for the sound of human activity and the sound of playing music. Almost all people in different exercise areas have different levels of aversion to traffic, construction, and machine sounds.


**Table 3.** Correlation between different exercise site selections and soundscape preferences.

\* Correlation is significant at the 0.05 level. \*\* Correlation is significant at the 0.01 level.

3.3.2. Effects of Exercise Modality and Exercise Site Selection on Visual Landscape Preference

As shown in Table 4, in terms of visual landscape, there is a significant positive correlation between slow walking, jogging, and brisk walking crowds and preference for natural landscape, with the people who run showing a lower preference for artificial landscape compared to the other two categories. In contrast, the fitness crowd and the square dancing crowd have a higher preference for fitness equipment and squares. There is a positive correlation between the choice of carrying out gymnastics activities, Chinese martial arts, ball games, and the preference for artificial landscapes, but a lower degree of relationship with the preference for natural landscapes, in which those who perform ball games do not show a positive preference for natural landscapes.

The data in Table 5 illustrate that there is a strong positive correlation between the choice of fitness equipment site and the preference for natural landscape, artificial landscape, and communal facilities. Weak correlations exist between the choice of being on a lawn or forest trail and the preference for artificial landscapes, while positive correlations exist with natural landscapes. There is a strong positive correlation between the choice of lakeside footpath and preference for fountains, bridges and pavilions, and parterres.

3.3.3. Effects of Choice of Activity Type Other Than Exercise on Audio-Visual Preference

Considering that exercise is not the only purpose for which people visit urban forest, and that most exercisers engage in concurrent activities, such as family, social, and leisure activities, it is necessary to explore the visual landscape and soundscape preferences for these activities as well.


#### **Table 4.** Correlation between different exercise forms and visual landscape preferences.

\* Correlation is significant at the 0.05 level. \*\* Correlation is significant at the 0.01 level.

**Table 5.** Correlation between different exercise site selections and visual landscape preferences.


\* Correlation is significant at the 0.05 level. \*\* Correlation is significant at the 0.01 level.

The data in Tables 6 and 7 illustrate that there is a positive correlation between those who perform family activities and all visual landscape preferences after exercise, but such activities show a weaker correlation with natural water features, hillside, and rockery preferences, while there is a positive correlation with natural sound preferences, which are more averse to noise. There is a positive correlation between the choice of social activities, leisure activities and group activities and preference for communal facilities, with a higher tolerance for activity sound. In contrast, there was a positive correlation between the choice of quiet–type activities and the preference for avenues, hillsides, and forests, and a negative correlation between the preference for artificial sound and activity sound.


**Table 6.** Correlation between the choices of other types of activities other than exercise and visual landscape preferences.

\* Correlation is significant at the 0.05 level. \*\* Correlation is significant at the 0.01 level.

**Table 7.** Correlation between the choices of other types of activities other than exercise and soundscape preferences.


\* Correlation is significant at the 0.05 level. \*\* Correlation is significant at the 0.01 level.

#### *3.4. The Influence of Individual Characteristics of Urban Forest Park Exercisers on Audio-Visual Preferences Preferences*  To explore the differences in soundscape and visual landscape preferences under

*Forests* **2022**, *13*, x FOR PEER REVIEW 15 of 26

To explore the differences in soundscape and visual landscape preferences under other exercise-related factors, we conducted an ANOVA between soundscape and visual landscape preferences under different individual characteristic indicators and plotted radar plots. If significant differences were presented (*p* < 0.05 or *p* < 0.01), the specific differences were described by specifically comparing the mean size; if no significance was presented, it means that there were no significant differences in audio-visual preferences under different individual characteristics. The analysis revealed that distance from home to the park and time period did not have significant effects on audio-visual preferences (*p* > 0.05), so the effects of the major individual characteristic factors of gender, age, number of companions, frequency, and exercise intensity were mainly explored. other exercise-related factors, we conducted an ANOVA between soundscape and visual landscape preferences under different individual characteristic indicators and plotted radar plots. If significant differences were presented (*p* < 0.05 or *p* < 0.01), the specific differences were described by specifically comparing the mean size; if no significance was presented, it means that there were no significant differences in audio-visual preferences under different individual characteristics. The analysis revealed that distance from home to the park and time period did not have significant effects on audio-visual preferences (*p* > 0.05), so the effects of the major individual characteristic factors of gender, age, number of companions, frequency, and exercise intensity were mainly explored. 3.4.1. The Relationship between Gender and Audio-visual Preference

*3.4. The Influence of Individual Characteristics of Urban Forest Park Exercisers on Audio-visual* 

#### 3.4.1. The Relationship between Gender and Audio-Visual Preference As shown in Figure 7(a), females generally preferred natural sounds more than

As shown in Figure 7a, females generally preferred natural sounds more than males, while males had a higher acceptance for activity and artificial sounds. Males and females showed significant differences in their preferences for traffic sound (*p* = 0.006 \*\*), construction noise (*p* = 0.038 \*), and mechanical noise (*p* = 0.007 \*\*), with females showing more significant aversions to these three types of sounds. Figure 7b shows that females preferred visual landscapes in the park more than males, significantly in terms of preference for ornamental flowers (*p* = 0.049 \*), fountains (*p* = 0.02 \*), and bridge and pavilion (*p* = 0.039 \*). males, while males had a higher acceptance for activity and artificial sounds. Males and females showed significant differences in their preferences for traffic sound (*p =* 0.006 \*\*), construction noise (*p* = 0.038 \*), and mechanical noise (*p* = 0.007 \*\*), with females showing more significant aversions to these three types of sounds. Figure 7(b) shows that females preferred visual landscapes in the park more than males, significantly in terms of preference for ornamental flowers (*p* = 0.049 \*), fountains (*p* = 0.02 \*), and bridge and pavilion (*p* = 0.039 \*).

**Figure 7.** Radar map of soundscape (**a**) and visual landscape (**b**) preferences distribution of exercisers at different genders. **Figure 7.** Radar map of soundscape (**a**) and visual landscape (**b**) preferences distribution of exercisers at different genders.

#### 3.4.2. The Relationship between Age and Audio-visual Preference 3.4.2. The Relationship between Age and Audio-Visual Preference

ANOVA results show that people aged 40-59 years have a higher preference for natural sounds, particularly birdsong (*p* = 0.009 \*\*), cry of insects (*p* = 0.047 \*), and water flow sound (*p* = 0.018 \*). Older people are more tolerant of activity sounds, significantly for the sound of children playing (*p* = 0.02 \*) and entertainment sounds (*p* = 0.032 \*). The visual-ANOVA results show that people aged 40–59 years have a higher preference for natural sounds, particularly birdsong (*p* = 0.009 \*\*), cry of insects (*p* = 0.047 \*), and water flow sound (*p* = 0.018 \*). Older people are more tolerant of activity sounds, significantly for the sound of children playing (*p* = 0.02 \*) and entertainment sounds (*p* = 0.032 \*). The visualized mean data are shown in Figure 8a.

ized mean data are shown in Figure 8(a). According to the ANOVA results, people aged 40–59 years showed a more significant preference for streams (*p* = 0.006 \*\*), ponds and lakes (*p* = 0.001 \*\*), lawns (*p* = 0.000 \*\*), avenues (*p* = 0.000 \*\*), hillsides (*p* = 0.002 \*\*), lakesides (*p* = 0.000 \*\*), ornamental flowers (*p* = 0.001 \*\*), forests (*p* = 0.000 \*\*), parterres (*p* = 0.001 \*\*), and fitness equipment (*p* = 0.02 \*) compared to other age groups. This is also shown in Figure 8b, where middle-aged people have a higher preference for the overall park landscape compared to other age groups.

*Forests* **2022**, *13*, x FOR PEER REVIEW 16 of 26

**Figure 8.** Radar map of soundscape (**a**) and visual landscape (**b**) preferences distribution of exercisers at different ages. **Figure 8.** Radar map of soundscape (**a**) and visual landscape (**b**) preferences distribution of exercisers at different ages. compared to other age groups. This is also shown in Figure 8(b), where middle-aged people have a higher preference for the overall park landscape compared to other age groups.

According to the ANOVA results, people aged 40-59 years showed a more significant 3.4.3. The Relationship between Companion Number and Audio-Visual Preference 3.4.3. The Relationship between Companion Number and Audio-visual Preference

preference for streams (*p* = 0.006 \*\*), ponds and lakes (*p* = 0.001 \*\*), lawns (*p* = 0.000 \*\*), avenues (*p* = 0.000 \*\*), hillsides (*p* = 0.002 \*\*), lakesides (*p* = 0.000\*\*), ornamental flowers (*p* = 0.001 \*\*), forests (*p* = 0.000 \*\*), parterres (*p* = 0.001 \*\*), and fitness equipment (*p* = 0.02 \*) compared to other age groups. This is also shown in Figure 8(b), where middle-aged people have a higher preference for the overall park landscape compared to other age groups. 3.4.3. The Relationship between Companion Number and Audio-visual Preference According to the results of the ANOVA, those who went with small families showed According to the results of the ANOVA, those who went with small families showed a significant preference for the four natural sounds of birdsong (*p* = 0.005 \*\*), insects (*p* = 0.045 \*\*), leaves (*p* = 0.046 \*), water flow (*p* = 0.028 \*\*) and the sound of children playing (*p* = 0.001 \*\*) compared to the rest of the population. In addition, according to Figure 9a, people traveling with small families were more averse to artificial noise and entertainment equipment, but no significant differences were found in the ANOVA. Those who were accompanied by a companion showed a higher preference for soundscapes compared to those who were alone, while being more averse to noise. According to the results of the ANOVA, those who went with small families showed a significant preference for the four natural sounds of birdsong (*p* = 0.005 \*\*), insects (*p* = 0.045\*\*), leaves (*p* = 0.046 \*), water flow (*p* = 0.028 \*\*) and the sound of children playing (*p* = 0.001 \*\*) compared to the rest of the population. In addition, according to Figure 9(a), people traveling with small families were more averse to artificial noise and entertainment equipment, but no significant differences were found in the ANOVA. Those who were accompanied by a companion showed a higher preference for soundscapes compared to those who were alone, while being more averse to noise.

a significant preference for the four natural sounds of birdsong (*p* = 0.005 \*\*), insects (*p* = 0.045\*\*), leaves (*p* = 0.046 \*), water flow (*p* = 0.028 \*\*) and the sound of children playing (*p*

**Figure 9.** Radar map of soundscape (**a**) and visual landscape (**b**) preferences distribution of exercisers at different companions. **Figure 9.** Radar map of soundscape (**a**) and visual landscape (**b**) preferences distribution of exercisers at different companions.

**Figure 9.** Radar map of soundscape (**a**) and visual landscape (**b**) preferences distribution of exercisers at different companions. According to the results of the ANOVA and in conjunction with Figure 9b, the visual landscape preferences of the exercisers showed significant differences for different numbers of companions, except for the bridges and pavilions (*p* = 0.132). Those who traveled in small families showed more significant preferences for streams (*p* = 0.003 \*\*), ponds and lakes (*p* = 0.008 \*\*), lawns (*p* = 0.000 \*\*), avenues (*p* = 0.011 \*), hillsides (*p* = 0.004 \*\*), lakesides (*p* = 0.001 \*\*), shrubs (*p* = 0.047 \*), ornamental flowers (*p* = 0.006 \*\*), forests (*p* = 0.007 \*\*), rockeries (*p* = 0.01 \*), parterres (*p* = 0.003 \*\*), fountains (*p* = 0.001 \*\*), fitness equipment

(*p* = 0.002 \*\*), and public buildings (*p* = 0.02 \*), while those who traveled in pairs showed more significant preferences for sculptures (*p* = 0.029 \*) and squares (*p* = 0.003 \*\*). ment (*p =* 0.002 \*\*), and public buildings (*p =* 0.02 \*), while those who traveled in pairs showed more significant preferences for sculptures (*p =* 0.029 \*) and squares (*p =* 0.003 \*\*).

According to the results of the ANOVA and in conjunction with Figure 9(b), the visual landscape preferences of the exercisers showed significant differences for different numbers of companions, except for the bridges and pavilions (*p =* 0.132). Those who traveled in small families showed more significant preferences for streams (*p =* 0.003 \*\*), ponds and lakes (*p =* 0.008 \*\*), lawns (*p =* 0.000 \*\*), avenues (*p =* 0.011 \*), hillsides (*p =* 0.004 \*\*), lakesides (*p =* 0.001 \*\*), shrubs (*p =* 0.047 \*), ornamental flowers (*p =* 0.006 \*\*), forests (*p =*  0.007 \*\*), rockeries (*p =* 0.01 \*), parterres (*p =* 0.003 \*\*), fountains (*p =* 0.001 \*\*), fitness equip-

#### 3.4.4. The Relationship between Exercise Frequency and Audio-Visual Preference 3.4.4. The Relationship between Exercise Frequency and Audio-visual Preference

*Forests* **2022**, *13*, x FOR PEER REVIEW 17 of 26

According to the variance results and in conjunction with the results shown in Figure 10a, people who exercise more frequently show a significant preference for natural sounds such as wind (*p* = 0.012 \*), birdsong (*p* = 0.04 \*), cry of insects (*p* = 0.014 \*), rustle of leaves (*p* = 0.007 \*\*), and water flow (*p* = 0.024 \*), and a more significant tolerance for noise: device music (*p* = 0.024 \*), entertainment sound (*p* = 0.022 \*), and traffic sound (*p* = 0.045 \*). People with lower activity frequencies show opposite trends in sound preference to those with higher activity frequencies. According to the variance results and in conjunction with the results shown in Figure 10(a), people who exercise more frequently show a significant preference for natural sounds such as wind (*p =* 0.012 \*), birdsong (*p =* 0.04 \*), cry of insects (*p =* 0.014 \*), rustle of leaves (*p =* 0.007 \*\*), and water flow (*p =* 0.024 \*), and a more significant tolerance for noise: device music (*p =* 0.024 \*), entertainment sound (*p =* 0.022 \*), and traffic sound (*p =* 0.045 \*). People with lower activity frequencies show opposite trends in sound preference to those with higher activity frequencies.

**Figure 10.** Radar map of soundscape (**a**) and visual landscape (**b**) preferences distribution of exercisers at different exercise frequencies. **Figure 10.** Radar map of soundscape (**a**) and visual landscape (**b**) preferences distribution of exercisers at different exercise frequencies.

The results of the ANOVA showed that those who exercised more frequently showed a significant preference for ponds and lakes (*p =* 0.015 \*), avenues (*p =* 0.021 \*), shrubs (*p =*  0.045 \*), parterres (*p =* 0.015 \*), and fitness equipment (*p =* 0.027 \*), while those who exercised less frequently showed a lower preference for all these landscapes. The results of the ANOVA showed that those who exercised more frequently showed a significant preference for ponds and lakes (*p* = 0.015 \*), avenues (*p* = 0.021 \*), shrubs (*p* = 0.045 \*), parterres (*p* = 0.015 \*), and fitness equipment (*p* = 0.027 \*), while those who exercised less frequently showed a lower preference for all these landscapes.

#### 3.4.5. The Relationship between Exercise Intensity and Audio-Visual Preference

3.4.5. The Relationship between Exercise Intensity and Audio-visual Preference According to the ANOVA results, the light exercisers showed a significant preference for natural sounds such as wind (*p =* 0.047 \*), birdsong (*p =* 0.022 \*) and leaves (*p =* 0.000 \*\*) compared to the strenuous exercisers, while the strenuous exercisers showed a more significant tolerance for noise such as traffic (*p =* 0.015 \*), construction (*p =* 0.03 \*), and According to the ANOVA results, the light exercisers showed a significant preference for natural sounds such as wind (*p* = 0.047 \*), birdsong (*p* = 0.022 \*) and leaves (*p* = 0.000 \*\*) compared to the strenuous exercisers, while the strenuous exercisers showed a more significant tolerance for noise such as traffic (*p* = 0.015 \*), construction (*p* = 0.03 \*), and machine noise (*p* = 0.009 \*\*). These trends are also shown in Figure 11a.

machine noise (*p =* 0.009 \*\*). These trends are also shown in Figure 11a. Using ANOVA, we found that the light exercise group showed a significant preference for the categories of streams (*p* = 0.001 \*\*), ponds and lakes (*p* = 0.015 \*), avenues (*p* = 0.025 \*), lakesides (*p* = 0.003 \*\*), parterres (*p* = 0.05 \*), and squares (*p* = 0.021 \*). Figure 11b also reflects that the heavy exercisers showed a lower preference for these types of landscapes compared to the light exercisers.

**Figure 11.** Radar map of soundscape (**a**) and visual landscape (**b**) preferences distribution of exercisers at different exercise intensity. **Figure 11.** Radar map of soundscape (**a**) and visual landscape (**b**) preferences distribution of exercisers at different exercise intensity.

#### Using ANOVA, we found that the light exercise group showed a significant preference for the categories of streams (*p =* 0.001 \*\*), ponds and lakes (*p =* 0.015 \*), avenues (*p = 3.5. The Effect of Audio-Visual Preference on Exercise Perception among Urban Forest Park Exercisers*

0.025 \*), lakesides (*p =* 0.003 \*\*), parterres (*p =* 0.05 \*), and squares (*p =* 0.021 \*). Figure 11b also reflects that the heavy exercisers showed a lower preference for these types of landscapes compared to the light exercisers. *3.5. The Effect of Audio-visual Preference on Exercise Perception among Urban Forest Park Exercisers*  Structural equation modeling was first developed by Swedish statisticians as a mul-Structural equation modeling was first developed by Swedish statisticians as a multivariate statistical analysis method for analyzing the complex structure of relationships between multi-indicator variables, SEM, which has similar aims to regression analysis but has two advantages over regression analysis [48]. Firstly, SEM is able to take into account the estimated residuals of the observed variables, which gives a more realistic picture of the sample information [49]. Secondly, SEM allows the reader to understand the relationship between variables in a more intuitive way by presenting the results in a simple graphical output [50].

tivariate statistical analysis method for analyzing the complex structure of relationships between multi-indicator variables, SEM, which has similar aims to regression analysis but has two advantages over regression analysis [48]. Firstly, SEM is able to take into account the estimated residuals of the observed variables, which gives a more realistic picture of the sample information [49]. Secondly, SEM allows the reader to understand the relationship between variables in a more intuitive way by presenting the results in a simple graphical output [50]. In this paper, in order to explore the influence of soundscape and visual landscape preferences on the perceptions of exercising people in the park, methods that can reveal the relationships that exist between multiple variables need to be used. Combining the properties and advantages of SEM, we ultimately used SEM to explore the influence of In this paper, in order to explore the influence of soundscape and visual landscape preferences on the perceptions of exercising people in the park, methods that can reveal the relationships that exist between multiple variables need to be used. Combining the properties and advantages of SEM, we ultimately used SEM to explore the influence of relationships between the variables. The visual and acoustic landscapes were divided into previously classified categories and the mean scores were used to calculate the scores for each dimension. The mean score is the most commonly used dimensional induction treatment. For the accuracy of the model, a multicollinearity test was performed to ensure that there was no multicollinearity between the variables. If tolerance ≤ 0.1 or VIF ≥ 10, it exist multicollinearity. According to what is shown in Table 8, none of the observed variables in the model are subject to multicollinearity.

relationships between the variables. The visual and acoustic landscapes were divided into


Topography 0.329 3.042 Natural vegetation 0.321 3.120 Manufactured Landscapes 0.330 3.030

previously classified categories and the mean scores were used to calculate the scores for **Table 8.** Collinearity diagnostics.

Model plotting was performed in SPSS AMOS software (IBM, Armonk, NY, USA), and relevant data were imported for computational analysis; the final model plot is shown in Figure 12. The oval in the figure represents the latent variable and the rectangle represents the observed variable, which is the measurement item in the questionnaire. Each measurement term must have a residual term, which is a circular term from e1–e9 in the figure. Artificial sound 0.882 1.133 Model plotting was performed in SPSS AMOS software, and relevant data were imported for computational analysis; the final model plot is shown in Figure 12. The oval in the figure represents the latent variable and the rectangle represents the observed variable, which is the measurement item in the questionnaire. Each measurement term must have a residual term, which is a circular term from e1-e9 in the figure.

Communal facilities 0.422 2.369 Natural sound 0.588 1.701 Human activity sound 0.984 1.016

*Forests* **2022**, *13*, x FOR PEER REVIEW 19 of 26

**Figure 12.** Structural equation model of audio-visual preferences and exercise feelings. **Figure 12.** Structural equation model of audio-visual preferences and exercise feelings.

In the model, soundscape and visual landscape preferences are latent variables, and the dependent variable exercise perception is the observed variable. Natural sound, human activity sound, and artificial sound are used as observed variables for soundscape preference, and natural waterscape, topographic landscape, natural vegetation, artificial landscape, and communal facilities are used as observed variables for visual landscape preference. According to the principles of SEM, residual terms need to be added to the latent variables and double arrows added between the exogenous latent variables. The model fit indices were as follows, CMIN/DF = 2.595(<3), GFI = 0.958(>0.9), AGFI = 0.925(>0.9), RMR = 0.023(<0.05), CFI = 0.973(>0.9), reflecting the overall goodness of fit of the model. The validation level α = 0.05 and each estimated parameter is significant. The results of the optimal model path coefficient estimation are shown in Table 9. In the model, soundscape and visual landscape preferences are latent variables, and the dependent variable exercise perception is the observed variable. Natural sound, human activity sound, and artificial sound are used as observed variables for soundscape preference, and natural waterscape, topographic landscape, natural vegetation, artificial landscape, and communal facilities are used as observed variables for visual landscape preference. According to the principles of SEM, residual terms need to be added to the latent variables and double arrows added between the exogenous latent variables. The model fit indices were as follows, CMIN/DF = 2.595(<3), GFI = 0.958(>0.9), AGFI = 0.925(>0.9), RMR = 0.023(<0.05), CFI = 0.973(>0.9), reflecting the overall goodness of fit of the model. The validation level α = 0.05 and each estimated parameter is significant. The results of the optimal model path coefficient estimation are shown in Table 9.

Correlation analysis of each visual landscape preference and exercise perceptions found that they were all significantly correlated. Figure 12 of the structural equation model shows that the overall visual landscape preference (0.65) has a greater direct effect on the exercise population's perception in the park and has the highest direct effect status for natural waterscape (0.98), followed by topographic landscape (0.91), and less direct effect for artificial landscape (0.78) and communal facilities (0.74). Additionally, the linear regression of the single term shows that sculptures have a significant negative effect on exercise perception.


**Table 9.** Path analysis of modified structural equation model.

The correlation analysis between soundscape preference and exercise perception found that most of them had significant correlation with perception, except for conversational sounds, footsteps, music, and entertainment equipment sounds that belonged to personal activity sounds. The structural equation model showed that the overall soundscape preference (0.06) did not have a high degree of direct effect on perception, and the activity sound had a relatively minor effect, which was also consistent with the correlation analysis results. However, noise still plays a negative influence in it. There is still a strong positive influence of natural sound, and in the regressions of individual items, bird song and water flow sound are found to have a more significant influence, which has some connection with landscape preference.

#### **4. Discussion**

This discussion section explores the landscape design of urban forest parks from the perspective of exercise behavior on the basis of comparing and summarizing the similarities and differences between this study and existing scholars' research views.

#### *4.1. Audio-Visual Preference Characteristics of Exercisers in Urban Forest Parks*

This study found that the exercise population had a higher preference for natural soundscapes and visual landscapes and a lower preference for man-made landscapes and human-generated sounds, and developed an aversion to noise in particular, which is consistent with the findings of Jeon et al. [51–53]. Among them, forests and footpaths and bird songs in it were most preferred by the exercising population, followed by streams and the water flow sounds they produce. Significant aversion was shown toward construction noise, traffic sounds, and machine noise, confirming the findings of Fang [33]. Park designers can install trails within the natural landscape or in the surrounding areas to increase the frequency of exercisers' contact with the natural landscape and optimize people's exercise experience in forest parks. At the same time, noise needs to be controlled. Some studies have shown that the use of bird calls to mask noise may improve people's soundscape perception [29]. On the one hand, park designers can consider artificially setting up some bird nests or bird feeders in the forest to attract birds to nest, thereby increasing bird calls; on the other hand, sound insulation panels can be installed next to noise sources for noise reduction.

## *4.2. The Effect of Exercise Style and Individual Characteristics on Audio-Visual Preferences* 4.2.1. Effects of Exercise Modality Choices on Audio-Visual Preferences

Among the different exercise activities, it is worth attention that the crowd of square dancers prefer squares and fitness equipment sites, probably because square dancing requires flat and open sites. They also prefer ornamental flowers and parterres compared to the crowd of other activities. The researchers' observation and questioning found that most of this group were middle-aged and older women who preferred flowers. There was a positive correlation between the square dancing and gymnastics crowd and the preference for music sound played by the equipment, but they had a strong aversion to activity sound and artificial noise. After questioning, it was found that they did not like the activity sound and artificial noise of people around them to interfere with the music sound they played. People using fitness equipment showed a higher acceptance for activity sound at the potential level, because there are often more people gathered at the fitness equipment in the park, and there are many children playing around.

#### 4.2.2. Effects of Exercise Site Selections on Audio-Visual Preferences

Those who choose forest and lawn footpaths are more concerned about the purity of the surrounding natural landscape and do not want artificial landscapes in the natural landscape, while being very averse to fire announcements in the forest. Those who choose lakeside footpath are more interested in small bridges and pavilions, fountains, parterres, waterscapes, and water sounds. Compared with other groups, those who choose the square open space are more interested in artificial landscape, and those who choose the fitness equipment site are very receptive to the landscape in the park. Both of these two types of exercise groups have a higher tolerance for activity sound as well as music sound.

Chen et al. showed that when the surrounding conditions can satisfy park users to engage in active health behaviors, they will still engage in exercise behaviors even if the landscape preference is weak, while when the conditions are not satisfied, they will not engage in exercise behaviors despite the strong landscape preference [54], which partly explains the differences in landscape preferences among people with different exercise behaviors. Park designers can try to provide adequate exercise conditions by installing some fitness equipment and trails at the interface area between natural and man-made landscapes. When planning forest footpaths, the purity of the surrounding natural landscape can be ensured by minimizing the involvement of man-made landscape and sound elements along the trails.

#### 4.2.3. Effects of Factors Other Than Exercise on Audio-Visual Preferences

Most of the population will also perform some other activities, and there is a relationship between these activities and audio-visual preferences. Those who engage in family activities prefer visual landscapes, but have a lower preference for natural waterscapes, hillsides, and rockeries, perhaps because parents are concerned that their children may be harmed in these areas. Park designers may consider putting signs and hints in prominent places to indicate potential hazards. Moreover, noise is more uncomfortable for them, probably because parents are more concerned about their children's feelings, so designers can consider creating special areas for children's activities away from noise.

The recreational crowd with public participation type of activities prefers to be near communal facilities and artificial landscapes, while having a higher tolerance for activity sound. Some studies have shown that people who come to play with children and gather with family and friends prefer to stay in areas that can accommodate group activities, such as lawns and recreation areas [55]. Designers can consider providing them with some resting benches or tables and chairs in areas close to human activity sites. Quiet-type activity groups prefer to be in less crowded avenues, hillsides, and forests, away from artificial landscapes and places with dense activity sounds, and designers can consider installing resting facilities in these areas for them to use.

## 4.2.4. Effects of Individual Characteristics on Exercisers' Audio-Visual Preferences

Among the different individual characteristics, the distance from home to the park and the time period of exercise had a low degree of influence on audio-visual preference, while gender, age, number of companions, exercise frequency, and intensity factors had an influence on audio-visual preference.

The present study found that gender has an effect on soundscape preference, which fits with Hedblom's findings [56]. Among them, female exercisers preferred natural sounds more than males, probably because females are more sensitive to some sounds and can easily perceive sounds that males ignore [57]. Moreover, similar to Gozalo's view in this study is that men have a higher acceptance for activity and artificial sounds, probably because it is usually women who take care of children and they are closer to artificial sound sources and thus need to endure the distress caused by this part of the sound [58]. In addition, females prefer visual landscapes in parks more than males, although males have a greater preference for forest landscapes.

The effect of age on audio-visual preferences also deserves our attention. The present study showed that as residents age, their evaluation of natural sounds increases and their tolerance for musical and activity sounds increases, and these results are generally consistent with Zhou's findings [59], which may result from the decline of human perception of high-frequency sounds with age [60]. Compared to middle-aged and young adults, older adults have a stronger perception of natural sounds. Some studies have shown that older adults are more likely to derive a sense of calm from natural sounds [56]. This study found that middle-aged exercisers had the highest preference for various park landscapes, and in line with Paneerchelvam et al.'s study, it was found that older adults enjoyed parks more than younger people, with a particular preference for natural landscapes [61]. Park designers may consider installing facilities within or around natural landscapes that are convenient for the elderly, such as handrails along the footpaths or adding anti-slip features to the trails.

This study found that the number of companions showed an effect on both auditory and visual landscape preferences. The audio-visual preference characteristics were similar for people on small family outings and those in the parent-child category. Accompanied exercisers are more likely to resent noise than those who are alone, possibly because the presence of noise tends to interfere with the activity they are carrying out or their interaction with their partner. People who traveled in pairs and small families had a stronger preference for the visual landscape in the park, and those who moved alone and in large groups appeared to be less concerned about the visual landscape.

The present study found a significant effect of exercise frequency on audio-visual preference, which was not mentioned in previous studies. People who exercised more frequently had a greater preference for the sounds of birds, insects, and water flow, and had a higher tolerance for noise, probably because this group of people was more accustomed to the surrounding soundscape environment. People with low exercise frequency are more sensitive to the noise, and are susceptible to the negative effects of noise. People with high exercise frequency also have more preference for visual landscape in the park, while people with low frequency do not perceive visual landscape significantly.

Exercise intensity also had a significant effect on exercisers' audio-visual preferences, and a correlation between greenway built environment and exercise intensity has been demonstrated by Dong et al. [62]. The present study found a higher degree of preference for natural sounds and natural landscapes among people with mild exercise intensity by further investigation. However, the phenomenon that the moderate and heavy exercise population found in this study had lower preference for each soundscape and visual landscape still needs further discussion. The author believes that this group is more concerned with the process of exercise itself to motivate themselves to achieve the desired exercise effect, so they may not pay attention to the surrounding sound and visual scenery.

#### *4.3. Effect of Audio-Visual Preferences on Exercisers' Perception of Exercise*

In this study, natural landscapes and soundscapes had the effect of enhancing people's exercise perceptions, which is consistent with Stigsdotter and Watts. Natural landscape spaces representing the tranquil type were rated as having the most restorative potential [63]. The inclusion of natural soundscapes in the environment helps to create a tranquil atmosphere and stimulate positive and pleasant emotions [64]. Interestingly, the overall soundscape preference did not have a high degree of direct effect on exercise perception, a finding that needs to be verified in further discussions. It is possible that the heavier breathing sounds produced during exercise or the preference of some exercisers to wear

headphones to listen to music during exercise may have affected their ability to capture the surrounding sounds.

#### **5. Conclusions**

This study was based on a field questionnaire in Beigushan Urban Forest Park, Lianyungang City, Jiangsu Province, China. The survey examined the different preferences for sound and visual landscapes among people exercising in urban forest parks. In terms of overall audio-visual preferences, natural soundscapes and visual landscapes were generally preferred, but artificially generated sounds and constructed landscapes were preferred to a relatively lesser extent. In terms of specific factors, the study analyzed the relationship between exercise style, exercise venue, other activities, and audio-visual preferences, and showed that these factors differed in terms of people's audio-visual preferences. People's audio-visual preferences for their surroundings influence their choice of exercise venue. Other types of activities outside of exercise also have an effect on the audio-visual preference of exercisers, for example, the crowd of family activities are more concerned about the possible harm caused by rockeries, hillsides, waterscapes, and noise to children. The study also used ANOVA to investigate whether different individual characteristics had a significant effect on audio-visual preferences. The results show that audio-visual preferences vary by gender, age, number of companions, frequency, and intensity of exercise, which provides important information for park planners to consider when designing their parks. However, when studying the effect of age on audio-visual preferences, due to the fact that some elderly people and children are not proficient in the use of electronic devices and require substantial manpower to assist in the research, and that this group has a weaker perception of the landscape, a large amount of insensitive data were screened out, resulting in a small sample size for these two groups. The sample size of the elderly and children will be further increased in the future for more in-depth supplementary research.

By constructing a structural variance model between soundscape preference, visual landscape preference and exercise perception, it was found that visual landscape has a more direct impact on exercise perception than soundscape for exercisers, and that soundscape planning can be taken into account when planning parks while ensuring visual landscape planning.

**Author Contributions:** Conceptualization, J.X., M.L., Z.G. and Y.X.; data curation, J.X. and Z.G.; funding acquisition, J.X. and M.L.; methodology, J.X. and Z.G.; project administration, J.X. and M.L.; resources, J.X., Z.G., Y.X. and N.J.; supervision, M.L.; validation, J.X. and M.L.; visualization, Z.G; writing—original draft, J.X., Z.G., Y.X. and N.J.; writing—review and editing, J.X., Z.G. and Y.X. All authors have read and agreed to the published version of the manuscript.

**Funding:** Project funding was provided by "The Study on Rural Tourism Revitalization Planning and Sustainable Living Coordination Mechanism", the International Cooperation Open Project of State Key Laboratory of Subtropical Building Science, South China University of Technology [grant numbers: 2019ZA02].

**Institutional Review Board Statement:** The study was conducted in accordance with the Declaration of Helsinki, and approved by the Institutional Review Board of The Department of Tourism Management, South China University of Technology.

**Informed Consent Statement:** Informed consent was obtained from all subjects involved in the study.

**Data Availability Statement:** The data for this study can be accessed through the uploaded affiliated files or by contacting the relevant author.

**Acknowledgments:** We would especially like to thank Jian Xu and Muchun Li for their assistance in proofreading the manuscript and the International Cooperation Open Project of State Key Laboratory of Subtropical Building Science, South China University of Technology. Furthermore, we are grateful to the 406 park exercisers for their participation in our questionnaire survey.

**Conflicts of Interest:** The authors declare no conflict of interest.

## **References**

