1 Introduction
The Outline of the 14th Five-Year Plan (2021–2025) for National Economic and Social Development and the Long-Range Objectives Through the Year 2035 clearly puts forward the need to strengthen environmental noise pollution control
[1]. The problems caused by urban noise are becoming increasingly serious, which has greatly affected residents' daily life and their physical and mental health
[2] [3]. Among them, traffic noise is the main source that is characterized with high intensity, wide influence range, and long duration
[4]. The negative impacts of it on residents include sleep disorders
[5], reduction of helpful behaviors, decreased attention to the peripheral elements of the environments passing through
[6], increase in problematic behaviors of children and adolescents (hyperactivity, inattention, etc.)
[7] [8] and driving violations
[9], etc.
At present, in order to better explore urban environmental design strategies to mitigate the negative effects of traffic noise, the research of its impact on psychology and physiology has mainly concerned with the noise annoyance under different conditions
[10]~
[14]. As one of the key components constituting the environment, behavior of spatial users is an outward manifestation of physiological and psychological perceptions
[15]. In environmental assessment, individual behavior usually refers to the attitude or performance of a person in a given situation, which is greatly influenced by the environment and often tends to be random
[16]. In contrast, the crowd behavior refers to the performance of the overall crowd in a specific environment, which tends to show a certain regularity
[17] [18]. The latter can be further classified as active and passive behaviors according to motivation
[16], or participatory and non-participatory behaviors by their purposes
[19]. In addition, according to the mobility of research objects, it can also be divided into static behavior (sitting, standing, etc.) and movement behavior (walking around, passing by, etc.)
[20]~
[22]. In comparison, movement behaviors can better reflect the overall tendency of the crowd
[20]. Currently, traffic noise related studies focus more on the effects on individual behaviors, but less on the effects on crowd behaviors.
Traditional control methods for urban traffic noise often use physical means to reduce the noise level, such as sound barriers and sound-absorbing materials. However, recent studies have shown that only reducing the noise level or the number of noise sources does not necessarily improve human comfort
[4]. Meanwhile, the research on multisensory interactions provides new possibilities for improving noise control. In recent years, the positive effects of olfactory sensory factors on people's perception have gradually received scholarly attention, and it has been demonstrated that positive odor stimuli such as fragrances of plants or food can alleviate the negative effects of traffic noise on residents' self-reported perception
[23]~
[25]. Studies on the combined effects of auditory and olfactory on crowd behaviors using typical sounds and odors in urban environments as sensory variables have shown that food odor can bring crowds closer to the odor source, which could become more significant after introducing either positive or negative sounds
[26]. It is also worth exploring whether the negative effects of traffic noise on residents' behaviors will be ameliorated by the presence of positive plant odor.
In this study, a typical urban landscape plant, lilac (Syringa spp.), was used as an odor affecting factor, and field observations of crowd behaviors were taken place under the combinations with different sound pressure level (SPL) conditions of traffic noise. This study aims to investigate how plant odors (concentration gradients) and traffic noise (low and high SPLs) affect crowd behaviors (movement path and speed). The results of this study will offer new ideas to mitigate the negative impacts of traffic noise on residents from the perspective of multisensory interaction, so as to improve the quality of urban multisensory environments.
2 Research Methods
2.1 Study Area and the Selection of Plant Odor–Traffic Noise Sources
When selecting the study area, the following factors were taken into account: 1) traffic noise existed in the area, with varied SPLs at different hours throughout the day; 2) both plant odor and no plant odor conditions existed in the area for the control of odor variables; and 3) no restaurants, cafes, markets, factories, or other odoriferous places existed in the area to emit affecting odors.
Based on the above considerations, Minsheng Road in Xiangfang District, Harbin City, China was selected as the study area (Fig.1), where trees of lilac (
Syringa amurensis) were planted uniformly along the sidewalks of the road at each 1.5 m. These trees were then used as the odor source in this study. Lilac is a deciduous shrub or small tree that is commonly planted in urban greening in China, due to the pleasant shape and odor. Studies have found that factors such as building layout and edge effect at road intersections can affect people's olfactory perception, resulting in a stronger plant odor in the central part and gradually decreasing towards the edge
[23]. Therefore, the study area fits the plant odor and traffic noise conditions for the research.
2.2 Plant Odor–Traffic Noise Measurement Methods
Quantitative measurement and recording of odor concentration is difficult. Although specific chemical components can be detected instrumentally, methods that simply measure certain compounds cannot adequately explain or grade odor concentration. Because such measurements are usually of tracer compounds that are present in limited amounts in the gas, it is not possible to directly measure the odor concentration itself or to correlate the odor concentration with its properties
[27]. Furthermore, in urban environments, measurements within a single time period or location are not representative of the odor situation in the entire area
[28]. Therefore, this study mainly adopted the method of measuring odor concentration with pedestrians as "sensors, " which is also commonly used in related studies
[29] [30].
Before the experiment, the study area was first divided into 3 m × 3 m grids, labeled as numbers 1 to 10 from left to right. A questionnaire pre-experiment on plant odor concentration was conducted, which adopted a Likert 7-point scale, with 1 indicating no odor and 7 indicating very strong odor. Both the pre-experiment and the formal experiment were performed on 10 June 2019, when lilacs were in full bloom and the odor was noticeable within in the study area. The experimental day was sunny with a wind speed of about 0.3 m/s and a relatively stable odor condition. In the pre-experiment, at least 30 valid questionnaires were collected from each grid and their mean concentrations were calculated. The study used Cronbach's α coefficient to test the internal consistency of the questionnaire (with a result of 0.925) and KMO coefficient to test the structural validity (with a result of 0.803), which indicated that the questionnaire had a sound reliability and validity. Meanwhile, during the pre-experiment, participants all reported that they did not smell any other odors other than lilac within the study area, suggesting that the plant odor was the only dominant source and the participants were not affected by other odor sources in the experiment.
In this study, a portable 4-channel acquisition front-end was used to measure traffic noise
[31][32]. A 1-min recording was made at the center of each grid, and the SPLs of the recordings were subsequently analyzed using HEAD ArtemiS software. During the recording, the crowd density remained less than 0.05 person/m
2, thus the interference of pedestrians on the sound environment could be ignored
[33]. Traffic noise was classified into low and high SPLs according to the actual traffic conditions at different hours: the SPL at the noisier hours within the study area was set as the high SPL, determined based on the field observations and SPL measurements in the pre-experiment; the difference between the high and low SPLs was 10 dB or more to ensure that the participants could clearly distinguish between the two. Through A-weighted SPL analysis, the low SPL was finally defined as 55.6 dB (L
10 = 50.3 dB; L
50 = 54.6 dB; L
90 = 58.5 dB) and the high SPL was 70.5 dB (L
10 = 68.9 dB; L
50 = 70.4 dB; L
90 = 71.7 dB). Calculated from the observation videos recorded by drone during the experiment, the average driving speeds in the low and high SPL conditions were 46 km/h and 53 km/h, respectively. The average hourly traffic flow was 1, 089 and 696 vehicles, respectively. The evaluation results of odor concentration and SPL distributions within the study area are shown in Fig.2.
2.3 Observation of Crowd Behaviors
This study focused on the movement behavior of crowds. Since the study area was close to an urban road, the crowd behaviors were mostly walking, so pedestrians' movement path and speed can better reflect the pattern of crowd behaviors, which two thus were selected as the analysis indicators for the research—both of them are also common indicators in relevant studies
[34][35]. Specifically, crowd path refers to the walking trajectory of the participants within the study area, while crowd speed refers to the straight-line displacement of the participant in unit time. It is important to note that pedestrians in both directions were included in the samples in the experiment (Tab.1).
The experimental times for low and high SPLs corresponded to the time periods of 13:00 ~ 15:00 and 15:00 ~ 17:00, respectively. Both time periods were set in the afternoon to minimize the difference in environmental conditions caused by temporal factors. The SPL tests were conducted first in both conditions, followed by the observation of crowd behaviors. In the experiment, the drone used to record the crowd behaviors flew at 100 m above to avoid the device sound from interfering with the experimental results
[34]. In addition, to ensure the randomness of the crowd behavior, three sets of measurement were taken under both low and high SPL conditions, and each set of video shooting lasted 15 ~ 20 min
[34].
Subsequently, the study took screenshots of the drone-shot videos every 2 s. The bottom-left endpoint of the leftmost grid was set as the origin of the coordinate system. The longitudinal coordinates of the intersection between the path and the longitudinal grid line were recorded as
y. So each path had 11
y-values, from left to right were
y1 ~
y11. And the path of each sample pedestrian was mapped by connecting the
y points on the coordinate. After that, the mean values of
y1 ~
y11 of all paths in the given time period were calculated. The 5th and 95th percentiles of these
y-values were also calculated, and finally a path range was determined by the means and the percentiles
[26]. At the same time, the speed of each sample pedestrian (
N = 211) was calculated as the distance displaced between the video screenshots divided the unit time (2 s)
[26].
3 Research Results and Discussion
3.1 Analysis of Plant Odor–Traffic Noise Impacts on Crowd Path
The path ranges under low and high SPL conditions are shown in Fig.3. The upper and lower boundary lines and mean values were more downward when the SPL was high. The noise with high SPL may be the cause of the crowd moving towards the opposite direction to the sound source. At low SPL, the crowd paths showed a gradual movement towards the odor source as the odor concentration grow across the grids; this trend was more significant at high SPL. Thus, the lilac odor may be a potential cause of crowd attraction, which became more obvious as the odor concentration increased and eventually remained stable.
In order to deeply investigate the combined effect of plant odor and traffic noise on crowd path, this study used an all-factor model to conduct a repeated measurement ANOVA on y-values of paths. As the odor concentration grow from left to right, with y1 ~ y11 distributed in the grids of different concentrations, y could represent the overall trend of the odor concentration gradient: an independent variable that indicates the odor concentration (denoted as "odor") were generated with the repeated measurement ANOVA model that fit the variables of y1 ~ y11; another independent variable was SPLs (denoted as "noise"). The final test results are shown in Tab.2. In detail, the effects of odor and odor ∩ noise on crowd paths were all significant (p < 0.05), proving the interaction of odor concentration and SPL on crowd path, with the effect of odor concentration on crowd path varying with SPLs.
The study calculated the estimated marginal means of the
y-values for the two conditions: the single effect of odor concentration and interaction of odor concentration and SPL (Fig.4). The estimated marginal mean was used to calculate the average change in a variable for a given condition, with larger values indicating that the crowd was closer to the noise source and farther away from the odor source. The results showed that under the single condition of odor concentration, the crowd path in the grids with strong odor was closer to the odor source (Fig.4). A two-by-two comparison of the 11
y-values under this condition revealed that
y1,
y2, and
y3 (no odor) were not significantly different from each other (
p > 0.05), while they and the rest
y values were significantly different from each other (
p < 0.01);
y4,
y5, and
y6 (weak odor) were different from any of the
y values (
p < 0.01), but
y7 ~
y11 (strong odor) were not significantly different from each other (
p > 0.05). These findings indicated that odor could attract crowds to a certain extent, where sensory attraction seems to be motivated by human biological nature and it may further affect emotion
[36].
The
y values for the condition of odor concentration–SPL interaction are shown in Fig.4. The difference between the low and high SPL conditions was significant (
p < 0.01). In detail, compared with low SPL, the crowd moved farther away from the noise source at high SPL. As the odor concentration increased, the crowd path at low SPL also moved slightly closer to the odor source, though not statistically significant (
p > 0.05). At high SPL, the crowd path tended to move closer to the odor source more significantly, with the maximum longitudinal distance difference between
y1 and
y11 being about 0.3 m. A two-by-two comparison of the 11
y-values at high SPL revealed that the differences between
y1 ~
y4 were not significant (
p > 0.05), while the differences between them and the rest were significant (
p < 0.01). Meanwhile, the differences between
y5 ~
y11 were not significant (
p > 0.05), but the differences between them and the rest were significant (
p < 0.01). This suggested that pedestrians tend to move away from the noise source and get closer to the odor source in high SPL environments. Previous research on the combined effects of food odor and sounds on crowd behaviors indicated that the presence of noise may enhance the attraction of odor to crowds
[26]. Similarly, this study suggests that there may also be some augmented effect of traffic noise on the attraction of plant odor, thus making crowds tend to move away from the noise source towards the odor source of plants.
3.2 Analysis of Plant Odor–Traffic Noise Impacts on Crowd Speed
The study also used a full-factor model for crowd speed under both conditions of odor concentration and odor concentration–SPL (Tab.3). The effects of odor and odor ∩ noise were both significant (p < 0.05), which indicated an interaction of odor concentration and SPL on crowd speed, with the effect of odor concentration on crowd speed also varying with the change of SPL.
The estimated marginal means of crowd speeds are shown in Fig.5, which had an overall mean of 1.17 m/s. In terms of the single effect of odor concentration, the crowd speeds showed a gradual decrease and a stabilized trend as the odor concentration increased (Fig.5). A two-by-two comparison of the crowd speeds in the 10 grids under the single effect of odor concentration revealed that there was no significant difference between the crowd speeds in grids 1 to 3 (no odor, p > 0.05), while they and the rest differed significantly from each other (p < 0.01); grids 4 to 6 (weak odor) differed from any of the grids (p < 0.01); grids 7 to 10 (strong odor) did not see significant difference among each other (p > 0.05). These results suggest that the higher the odor concentration, the slower the crowd speed.
The estimated marginal means of crowd speeds for the odor concentration–SPL interaction are shown in Fig.5. The difference between the low and high SPL conditions was significant (
p < 0.01). For the overall trend, at high SPL, the crowd speed was significantly accelerated compared with low SPL, which is consistent with previous findings
[37] [38]—this accelerated behavior may be pedestrians' spontaneous response to noise. At low SPL, there was no significant difference in crowd speeds between all the grids (
p > 0.05). When the SPL was high, there was no significant difference between grids 1 and 2 (
p > 0.05), but they were significantly different from the other grids (
p < 0.01); grids 3 to 5 differed from any of the grids (
p < 0.01); grids 6 to 10 were not significantly different between each other (
p > 0.05), but their differences with the other grids were significant (
p < 0.01). These results represent that the crowd speeds were more significantly affected by odor concentration under high SPL condition, with the most significant effect found around grids 3 to 5.
The analysis results of the mean values show that the crowd speeds across the grids rarely changed when the SPL was low (Fig.6). Nevertheless, the changes were more significant at high SPL, with the fastest average speed in grids 1 and 2 (1.24 m/s) and the slowest average speed in grid 10 (1.18 m/s). But the speeds in grids 5 to 10 were relatively stable, indicating that changes in odor concentration did not significantly affect crowd speed. Overall, although the crowd speeds under low and high SPL conditions showed varied patterns, the maximum difference in speeds between grids was not obvious, being only 0.06 m/s.
3.3 Correlation Analysis Between Crowd Path and Crowd Speed
This study analyzed the crowd path and crowd speed under two SPL conditions (103 pedestrians at low SPL and 108 pedestrians at high SPL) and tested the correlation between the two under the odor concentration–SPL interaction (Tab.4). Since the trend of odor concentration was the same in the grids under both low and high SPL conditions, the study calculated the correlation coefficients between the average y value and the average crowd speed using Pearson correlation analysis. The results showed that under low SPL condition, the crowd path showed a negative correlation with the crowd speed, though not significant (
p > 0.05). While under the high SPL condition, the crowd path showed a significant negative correlation with the crowd speed (
p < 0.01), indicating that the closer to the side of noise source (i.e., the farther away from the side of odor source), the slower the crowd speed. This contradicts the previous findings that higher noise SPL leads to higher crowd speeds
[38]. It may be due to that such studies usually focus on the overall speed change pattern in the direction parallel to noise source and do not examine the longitudinal distance differences between the noise source and the pedestrians. In addition, people are often more careful when walking closer to the road, thus the shorter the longitudinal distance to the road, the slower the crowd speed.
4 Conclusions and Implications
In this study, a covert behavioral observation experiment was conducted on a typical urban road with evenly planted lilac trees to investigate the combined effects of plant odor concentration and traffic noise on crowd behaviors. The results confirm that the positive effect of plant odor can mitigate the negative impact of traffic noise, offering new ideas for traffic noise control. The exploration of auditory–olfactory combined effects on crowd behaviors can provide a more comprehensive evaluation on the quality of urban landscapes, which helps improve the traditional visually-dominated design paradigm of urban spaces. The main findings of this study include as follow.
1) For crowd path, the path range under high SPL condition was farther away from the noise source compared with that at low SPL; the crowd path in the area with strong plant odor tended to be closer to the direction of odor source, which was even more obvious under the high SPL condition.
2) For crowd speed, as the plant odor concentration increased, the crowd speed at high SPL tended to slow down first and then remained stable; but at low SPL, crowd speed was almost unaffected by plant odor.
3) Compared with the low SPL condition, the crowd speed slowed down when the crowd path was closer to the noise source at high SPL.
The above findings are instructive for the improvements of the landscape quality of urban environments. For instance, plants with fragrant scent can be planted in areas with traffic noise or other negative sound sources in urban environments to alleviate the negative impacts.
Furthermore, the present study, as a field experiment on crowd behaviors, is not possible to control the environmental variables completely and precisely. Future studies can be optimised in the following aspects.
1) The study area was adjacent to an intersection, and even though it was about 15 m from the nearest end of the crosswalk, this may still cause some interference to the crowd. So future research is recommended to choose central sections of urban roads as the study area, to avoid possible impacts.
2) This study explored the common behavioral trends of pedestrians in the study area; however, there may be differences among pedestrians' walking behaviors in different directions. Therefore, subsequent studies could be carried out to investigate the disparity in different walking directions.
3) Syringa amurensis was selected as the odor source for this study; however, its strong and distinctive odor may lead to individual preference differences. Other fragrant plants could be selected for subsequent studies.
4) Although the participants reported that they did not smell any odor other than the plant odor, the study area may be potentially affected by other odors such as traffic odors. Subsequent studies may consider more precise control of experimental variables to take odor perception thresholds into account.
5) The preliminary research found that the pedestrian densities of crowd during the two selected time periods were relatively consistent, thus the experimental settings under low and high SPL conditions were considered consistent. However, the crowd density and behavioral characteristics on urban roads may vary at different times, thus future studies can put efforts in the effects of temporal factors and crowd density on crowd behaviors.
6) Plants themselves may also have an effect on crowd behavior (e.g. visual attraction), thus future studies can control the multisensory factors of vision, audition, and olfaction separately, so as to explore the combined effects between the three.
PDF
(2730KB)
