Assessing human emotional responses to the design of public spaces around subway stations: a human factors research

Transit-oriented development (TOD) is a powerful urban planning strategy to enhance sustainability and provide socioeconomic benefits for cities. The human-centered design of public spaces around subway stations is a critical issue in TOD. In this study, a socio-technical system-based perspective was adopted to investigate the impact of using different design strategies in public spaces around subway stations on human emotional responses. The novelty of this study lies in performing a human factors experiment to examine human emotional reactions to outdoor public spaces surrounding transit stations using a comprehensive method combining physiological assessment and subjective self-report. Thirty-four participants were recruited for the experiment, which was conducted at the catchment areas of two subway stations in Nanjing, China. Urban design characteristics related to open space enclosure and visual elements, natural elements, pedestrian access, surrounding buildings or walls, and land use and activities were tested in both semi-underground and above-ground public spaces. Additionally, human emotions were assessed using the pleasure, arousal, and dominance model and by investigating the four response systems that predominantly reflect people’s emotional states. These results led to design and policy suggestions that could assist practitioners and researchers in selecting relevant approaches for human-oriented and place-based TOD planning.


Introduction
The importance of transport to sustainable development was first acknowledged at the 1992 United Nations (UN) Earth Summit (Pronk and Haq 1992).Sustainable transportation, including public transit, was integrated into several sustainable development goals and objectives of the 2030 Agenda for Sustainable Development (UN-Habitat 2016).On this basis, transit-oriented development (TOD) has drawn increasing interest from scholars and practitioners as a means of promoting the use of sustainable transport modes and providing socioeconomic benefits to cities by integrating residential, workplace, commercial, and public open spaces in a walkable environment (Calthorpe 1993).As transit stations attract large crowds of people across the city, these places have become valued gathering places (Bertolini 2006).Improving the urban design of public spaces can provide high-quality meeting places and, in turn, promote urbanity and the use of public transport (Jensen 2014).Ibraeva et al. (2020) advocated for more studies to explore the possible advantages of TOD for community life at a micro-scale, in relation to human comfort and aesthetic appeal of an area.
Integrating systems science and urban design can bring new insights and shape smart cities (Yang and Yamagata 2020).Recently, the complex socio-technical systems theory has been applied into urban design and planning for managing the complexity of the interdependent urban systems, including main streets and open spaces (Stevens 2016;Stevens et al. 2021).This theory essentially states that the design and performance of any organizational system can only be understood and improved if both the "social" (human part) and "technical" (non-human part) aspects are considered together and treated as interdependent parts of a complex system (Trist 1981).Consequently, stations and catchment areas can be taken as a socio-technical system wherein people interact with the physical environment, which includes the station and the surrounding public space (Norman and Stappers 2015;Trist 1981;Yang et al. 2019).
To inform the development of socio-technical systems, human-centered design is a practical approach (International Organization for Standardization 2010) that adheres to the idea of basing the design upon precise knowledge of people and their interaction with the physical environment.Human emotions have been widely examined in studies of human-environment interactions as emotions influence choices and serve as a foundation for social interaction (Keltner and Kring 1998).Moreover, emotions are conveyed with awareness of the situational environment in which they occur.Hence, there has been an increasing research interest in examining human emotional responses to stations and public spaces (Qiao et al. 2021;Tang and Auffrey 2018).
Accordingly, to achieve sustainable transportation goals and improve the performance of TOD, there is a need to plan for more integrated transport stations and public space systems that put humans at the system's center.This study addresses this challenge by studying human emotional responses to public spaces around subway stations.The emotional reactions of people to different urban designs of public places (including semi-underground and above-ground open spaces) around these stations were both quantitatively and qualitatively evaluated.Consequently, results reveal design and policy suggestions that could assist practitioners and researchers in selecting relevant approaches for human-oriented and place-based TOD planning.

Public spaces design in TOD planning
Initially promoted by urban planners as a fundamental principle of New Urbanism, the TOD strategy encourages individuals to use public transport in preference to private vehicles and complement public transport with non-motorized travel for shorter distances (Congress for the New Urbanism 2000).To guide TOD planning, Ewing and Cervero (2010) proposed a 5D principle that highlights the importance of design, density, diversity, destination, and distance.Prior studies (Peimani and Kamalipour 2020;Furlan et al. 2019;Lamour et al. 2019;Loo and du Verle 2017;Jacobson and Forsyth 2008) have highlighted the importance of urban design conditions in evaluating TOD and advocated applying a combination of transit-oriented, place-based, and people-oriented design theories.
The design of public spaces affiliated with transit stations has attracted considerable attention.To this end, a great body of literature investigated the issues of crowd control, evacuation management, and air quality improvement of indoor underground spaces (Feng et al. 2021;Soltani and Kashi 2022).However, few studies examined above-ground or semi-underground public open spaces surrounding transit stations (Peimani and Kamalipour 2020;Stojanovski 2020) such as plazas and streets (Moughtin 2007).There is an urgent need for improvement of these open spaces by designing a vivid and connected pedestrian environment in line with using place-making approaches (Cervero 2009).
Among these open spaces, the plaza acts as a key place that encourages various forms of interaction between people while walking (Carmona et al. 2019).The design features of these public spaces can be categorized into five dimensions: open space features, natural elements, pedestrian access, surrounding buildings, and land use and activity.Appendix A summarizes several essential design features for each dimension.Specifically, the D/H ratio ("D" represents the depth/ width of the public space, and "H" represents the height of its surrounding buildings) is widely used in urban design to assess the open space closures.For example, the D/H ratios of 2/1 and 3/1 are frequently employed in urban design practice, especially for enclosed open areas (Sitte et al. 1965;Ashihara 1981;Lynch and Hack 1984;Jacobs 1993).

Human-centered design and evaluation of emotional responses to public spaces
Designing with humans in mind has long been a norm in both transportation engineering and public space design (Harris 1966;Lynch 1960).Engagement of the public in urban design and planning may date back to the 1970s that aims to increase participants' mutual awareness of design processes or to urge the adoption of a design project (Yang et al. 2021;Vainio 2016;Sanoff 1990).In recent years, humans' emotional responses to the urban environment have been used as a new information layer within planning processes, leading to a new kind of participatory planning (Zeile et al. 2015).Such a human-centered planning approach (Zeile et al. 2015) has been applied by designers and planners to analyze people's momentary satisfaction with urban spaces (Kaklauskas et al. 2021;Qiao et al. 2021;Weijs-Perrée et al. 2020;Tang and Auffrey 2018;Li et al. 2016).Nevertheless, there remains a need for considering human emotion in the plan of human-centered transport and public space systems.
Emotional responses refer to "a selection of an appropriate response and the inhibition of other less appropriate responses from a more or less broad behavioral repertoire" (Thayer and Lane 2000, p. 203).Generally, discrete and dimensional models are used to measure a person's emotional state.Discrete models assume that all human emotions can be derived from a defined number of innate basic emotions (Ekman 1999), while dimensional models hold that emotional responses are organized according to several basic dimensions (Smith and Ellsworth 1985).From the perspective of environmental psychology, Mehrabian and Russell (1974) proposed a dimensional emotion model to link environmental stimuli with the emotional responses of pleasure, arousal, and dominance (PAD).These three dimensions have been used as crucial evaluation indicators for emotional responses (Lang et al. 1997;Mauss and Robinson 2009).Pleasure is defined as a continuous dimension ranging from pleasantness (e.g., joy) to unpleasantness (e.g., sadness) (Osgood et al. 1957).Arousal represents the extent of a person's excitement or motivational activation (Bradley et al. 2001).Dominance indicates a feeling of control and influence in the situation, varying from controlling (i.e., the motivation to approach stimuli) to controlled (i.e., the tendency to avoid triggers) conditions.
Typically, human psychological studies examined emotional responses to urban spaces using qualitative methods such as field surveys (Rios-Rodriguez et al. 2021;Weijs-Perrée et al. 2020).A major current focus in investigating the impacts of urban plans on user emotion is adopting integrated research approaches, which include case studies, controlled experiments (Yin 2014), self-report questionnaires, visualization techniques, and human sensors (Weijs-Perrée et al. 2020;Zeile et al. 2015) to find impartial support for these causal linkages.Investigations of individuals' emotional responses on public spaces by applying the latest technologies such as video neuroanalytics (Kaklauskas et al. 2021) and emotion-tracking sensors (Li et al. 2016;Zeile et al. 2015) have generated considerable recent research interest.For example, Tang and Auffrey (2018) used eye tracking with virtual reality (VR) technology to understand human wayfinding behavior in indoor subway stations and accordingly, improve the design of stations.Qiao et al. (2021) examined emotional reactions to the spatial quality of urban green spaces using subjective assessments and face recognition measures.

Gaps and aims
Existing studies have recognized the need to improve the quality of the public spaces surrounding transport stations.However, more efforts are needed regarding the microscale and human-centered design of these public spaces; particularly, human emotional responses to the designs of such spaces need to be investigated further.Additionally, existing design theories on public spaces need to be tested through experiments and case studies with the support of latest technologies.
The aim of this study, therefore, is to investigate how urban design characteristics affect human emotions when they explore the public spaces surrounding transit stations by taking a socio-technical systems perspective.To address the research question, a human factors experiment was conducted to observe the interaction between different designs of public spaces and people's physiology and subjective experience.This method enables us to identify the most effective characteristics that contribute to human's positive emotional responses.The present study is intended to serve as a starting point for understanding the design standards for public open spaces that could potentially be used as a design guide for urban designers and transport specialists to build an integrated transport, spaces, and humans system.
Two subway stations in Nanjing city, China were used as case studies.Human emotional responses were examined by collecting quantitative physiological data on electrodermal activity (EDA), heart rate variability (HRV), electroencephalogram (EEG), and eye tracking, as well as qualitative selfreport data.As this is a multidisciplinary research issue, a research team from urban design, architecture, and human factors engineering fields was assembled.
The remainder of this article is organized as follows.The Methodology and methods section describes the methodology, characteristics of the two case study areas, and the design of the experiments providing details of the indicators and devices of physiological measurements.Then, in the Results section, the impacts of urban design on emotional responses are presented, followed by the Discussions and Conclusion sections where general urban design guidelines are elucidated.

Methodology
The research question was addressed using the methodology showing below (Fig. 1).By introducing the socio-technical system theory, TOD planning was reconsidered as an integrated design of the technical part (station and public spaces) and social part (humans) of the transport, spaces, and humans system.To achieve such an integrated design, the interactions between people and the physical environment need to be investigated.To this end, the theory of human-centered design was adopted in this study wherein the human emotional response to urban spaces is a crucial issue.
To explore the interplay between the technical and social parts of the transport, spaces, humans system, the two parts were initially described based on TOD and public space design principles and the PAD model, as shown in Fig. 1.From the technical part, urban design characteristics of open space features, natural elements, pedestrian access, surrounding buildings/walls, land use and activities (Appendix A) were investigated.From the social part, people's autonomic nervous system, central nervous system, affect-modulated startle, and subjective experience which predominantly reflect emotional states were examined (Mauss and Robinson 2009) (Table 1).
More specifically, first, the autonomic nervous system is part of the peripheral nervous system and is responsible for regulating vital functions such as heartbeat and breathing (Öhman et al. 2000).The system is most commonly assessed based on cardiovascular responses (HRV is a standard measurement) and electrodermal responses (EDA is a common measure) (see Kreibig (2010) for a review).Second, the central nervous system comprises the brain, spinal cord, and neurons.To recognize emotion from it, brain activity is usually quantified using positron emission tomography and EEG.EEG signals are very useful for evaluating emotions (Chen et al. 2015).Third, startle is a primitive reflex that helps avoid injury and threats (Bradley et al. 1999).The startle response involves several motor actions such as tensing of the back and eye blinking (Landis and Hunt 1939).Last, people's subjective experience is typically reflected by filling questionnaire after the experiment because self-reports of current emotional responses are more valid than those of past or future experiences (Robinson and Clore 2002).
Human emotional responses to various urban design characteristics of public spaces surrounding stations were then Root mean square of successive differences between NN intervals pNN50 (%) Percentage of adjacent NN intervals differing by more than 50 ms LF power /HF power Index of "sympathovagal balance": the ratio of low to high frequency EDA SC Skin conductance Tonic activity (SCL) Slower components and background characteristics of the signal Phasic activity (SCR) Faster-changing elements of the signal that can be associated with a stimulus EEG Alpha wave activity (dB) Average power of the Alpha wave Theta wave activity (dB) Average power of the Theta wave Frontal alpha asymmetry for five electrode site pairs: F3-F4, F7-F8, C3-C4, P3-P4, O1-O2 Left (L) hemisphere vs. right (R) hemisphere: (L-R) / (L + R) Eye blink Eye blink frequency (n/min) Average frequency of blinking explored by conducting human factors experiments, from which objective physiological data and subjective self-report data were collected and analyzed.In the end, the results were concluded in a series of urban design and policy recommendations for TOD planning.

Physiological measurements of emotion
Table 2 presents the physiological measurement indicators used in this study.HRV is an index of the ability of a person's heart to adapt to changing conditions by showing how the body varies between adjacent heartbeats over a while (McCraty and Shaffer 2015).HRV in both time and frequency domains was examined by assessing: (1) the heart rate (HR), standard deviation of all normal-to-normal intervals (SDNN), root mean square of successive differences (RMSSD), and pNN50 (percentage of adjacent normalto-normal intervals differing by more than 50 ms) for time domain analysis; and (2) the ratio of the low frequency (LF) to high frequency (HF) in the frequency domain analysis to reflect the balance between the sympathetic and parasympathetic nerves (Appelhans and Luecken 2006;Malliani et al. 1991).HR is a fundamental indicator of the pleasure dimension of emotions and is positively correlated with the degree of pleasantness (Cuthbert et al. 1988;Lang 1993;Acharya et al. 2006).However, according to a previous review (Kreibig 2010), other HRV indicators negatively correlate with the degree of pleasure.EDA refers to the autonomic changes in skin electrical properties and skin conductance (SC) is the most extensively studied property.The SC can be obtained by applying an electrical potential between two contact points on the skin and measuring the electrical current produced between them (Staats 2012).The EDA complexes include background tonic activity (SC level [SCL]) and fast phasic activity (SC response [SCR]) elicited by sympathetic neural activity (Braithwaite et al. 2013).EDA is arguably the most useful indicator of changes in sympathetic arousal that can be controlled by emotional and cognitive states.SC positively correlates with emotional arousal (Cuthbert et al. 1988;Lang 1993).The SCL and SCR are also correlated positively with the arousal state (Kreibig 2010).
EEG measures the electrical activity of brain neurons.The neurons work together to produce a constant rhythmic signal, which can be divided into several bands according to frequency.Here, the absolute power and relative power of alpha (α) (8-14 Hz) and theta (θ) (4-8 Hz) bands were captured.The alpha band is a fast wave that has been extensively studied (Bronzino 2014;Kandel et al. 2000).This wave is believed to represent the cerebral cortex's main electrical activity when a person is in a wakeful and quiet state.The theta band is a slow wave that is usually more obvious when humans are in a sleepy state, especially when they are frustrated or feel depressed (Kandel et al. 2000).Earlier studies found a positive correlation between the power of the theta wave and the arousal level (Krause et al. 2000).
This study also measured "frontal asymmetry," which compares the alpha power in the left (L) and right (R) hemispheres using the formula (L-R) / (L + R) because this asymmetry-based index was found significant for pleasure Fig. 2 Location of the stations and dominance.Greater left-sided activation indicates more intense pleasant emotions and dispositional tendencies toward approach rather than avoidance (Sutton and Davidson 1997).For EEG measurements, this study applied the International 10/20 System using the Modified Combinatorial Nomenclature to position 16 electrodes on an individual's scalp.The odd-numbered electrodes refer to the electrodes placed on the left hemisphere, whereas the even-numbered ones denote those on the right hemisphere.
Eye blink rate has been used to investigate the effects of dopamine on attention allocation and frontal cortex in perceptual tasks; thus, it is a valuable measure of emotional pleasure (Maffei and Angrilli 2019).Eye blink rate increases in response to unpleasant images or film clips owing to reduced processing of sensory information to prepare a behavioral defensive reaction; however, the blink rate decreases in response to pleasant images (Bradley and Lang 2000).

Study area and participants
The experiment was conducted during December 15-22, 2021, in Nanjing, the capital city of Jiangsu province in China.It is a megacity in the Yangtze River Delta city cluster and has a humid subtropical climate.
Researchers selected two subway stations in the central area of Nanjing (Fig. 2): Site1 is Daxinggong subway station and Site2 is Gulou subway station-both were constructed during the same period.Semi-underground public spaces and above-ground public spaces affiliated with station exits were used as the case study subjects.Characteristics of these spaces are shown in Fig. 3 and Table 3.
The semi-underground open space of Daxinggong station is more enclosed with solid walls than Gulou station.Moreover, the ground-floor open space of Daxinggong is arranged along a motorway with street trees on one side, while the ground space of Gulou station is a small corner garden.
Thirty-four healthy voluntary participants aged 14-55 years were recruited by distributing a volunteerrecruitment flyer through social media.The adult participants and guardians of the teenage participants were notified of the content and rules of the experiment and signed informed consent forms.This study conformed to the guidelines within the Declaration of Helsinki.
A preliminary experiment was conducted on December 16, 2021, with four participants to test the rationality of experiment design and implementation.Then, the remaining 30 participants, comprising 16 females and 14 males (Table 4), were grouped into six groups and joined the experiment on one of the days during December 17, 19-22 (wintertime).Although the case study areas were not intentionally described to participants, they had some previous knowledge about the stations as they are residents.Data from 30 participants were collected and analyzed.
The BAPPU-Time physical environment measurement software was used (ELK GmbH, Krefeld, Germany) to record the microclimate at all research sites to assess the impact of environmental conditions on human comfort.The experiments were also conducted under a similar situation during the sunny wintertime with a temperature of 2-6 °C, relative humidity of 4-5%, gentle breezes, and a noise level of 4-6dBA.

Study protocol
The protocol shown in Fig. 4 describes what participants did in this station.First, they were gathered at one station, asked to wear the required experimental instruments, and then informed of the experimental procedure.Subsequently, all participants were requested to wear two wireless physiological recording devices for collecting HRV and EDA data.A group member in each group (six in total) also wore an apparatus for recording brain activity and glasses for eye tracking.The researchers selected these six participants beforehand to include one teenager, four young people, and one middle-aged (male/female is 1/1) for covering the target age span of this research.Simultaneously, the experimenters recorded the environmental conditions, including temperature, humidity, and noise, to recognize and avoid any microclimatic impact on the emotional responses.Then, when participants were in a resting state, the researchers clicked "start" on the ErgoLAB wearable device to start recording participants' physiological data for three minutes.Participants were then guided to the underground exit of the subway station and let them walk to the ground exit.Next, they were asked to wander in the semi-underground space for five minutes and then wander in the above-ground open space for five minutes.Finally, the researchers stopped recording and asked each participant to complete a 12-item questionnaire after completing the tasks.An example items was: "Which space (underground/semi-underground/aboveground) makes you feel pleasant?"Participants were also asked to score the surrounding area in five aspects: space quality, comfort, aesthetics, identifiability, and usability.The answer to each question was scored on a five-point scale (− 2, − 1, 0, 1, and 2).
After completion, they took the subway and went to the next station.The researchers changed the order and starting time of the experiments at Daxinggong station and Gulou station to avoid the influence of fatigue-resulting from the participants walking a long time outdoors and traveling by subways-on the experimental results.

Devices and statistical analysis
ErgoLAB portable wireless physiological recording devices (Kingfar International Inc. Beijing, China) were used to collect HRV (based on photoplethysmography) and EDA data.Each participant wore an ear clip sensor in the middle of their left earlobe with a 64 Hz frequency and a 0-100% sampling range to measure the HRV.Everyone also wore two electrodes of the EDA sensor, which has a 64 Hz frequency and a 0-30 μS sampling range, on the palm of their left hand.One participant in each group wore: a BitBrain semi-dry EEG cap (Bitbrain, Zaragoza, Spain) with 16 channels and a 256 Hz sampling frequency for recording the real-time EEG signals; and Tobii Glasses 2 (Tobii, Stockholm, Sweden) for tracking eye movements (Fig. 5).
Collected physiological data were processed before statistical analysis on the ErgoLAB V3.0 Human-Machine-Environment synchronization platform developed by Kingfar International Inc. Beijing, China, which was verified by researchers in related fields (Zhang et al. 2018;Zou et al. 2017).First, HRV was processed using filtering methods such as wavelet denoise, high-/ low-pass denoise, band pass; then, the ectopic interval values were corrected.EDA was processed using a Gaussian smoother and low-pass denoise.Since the HRV and EDA data would be affected by the baseline status of sweat glands and sympathetic and parasympathetic nerves, which vary with each individual and differ within a day, the degree of change was utilized to analyze these data.The baseline average values of the indicators were removed from the dataset collected in the semi-underground and above-ground spaces (Eq.1).
(1) ΔHRV = HRV Semi-under(Above-)ground − HRV Resting Fig. 5 The experiment process (participants were wearing red hats while the ones with blue hats were also wearing devices of Tobii Glasses2 and BitBrain Semi-dry EEG) EEG data were processed using high-/low-pass denoise and band-stop filters.Regarding the eye tracking data, a velocity-threshold identification algorithm was employed to distinguish the types of eye movements (eye blinks, saccades, and fixations).The blink data of both eyes were also processed and defined the maximum and minimum time thresholds as 350 ms and 75 ms.
All data were expressed in the form of mean ± standard deviation.Stata17 was employed to conduct t-tests, Wilcoxon signed-rank tests, and sign tests for within-group comparisons.First, we checked whether the data followed the normal distribution: if so, then we used t-tests; if not, we attempted to transform the data.If data still did not meet the normal distribution, we chose Wilcoxon signedrank or sign tests.To report the statistics, we followed the American Psychological Association's (2010) 6th edition of statistical standards, where p < 0.05 is considered significant, and a p-value between 0.05 and 0.1 is considered marginally substantial.The latter is especially common in the field of experimental psychology (Olsson-Collentine et al. 2019).

Impact of urban design features on HRV
T-tests were conducted to compare each station's semiunderground and ground-level spaces, whereas the Wilcoxon signed-rank test was applied to compare the public spaces between different stations.Figure 6 shows that the HRV values measured by the HR, SDNN, RMSSD, and pNN50 metrics differ across different areas.
Concerning the HR, it was higher (p < 0.01) in the ground space than it was in the semi-underground area of Gulou station.The HR was significantly larger (p < 0.001) at Gulou than at Daxinggong station, for both the semi-underground and ground spaces.Contrastingly, the SDNN, RMSSD, and pNN50 values were generally smaller (p < 0.01) in the ground-level area than in the semi-underground area in Gulou station; the values of these three indicators were lower (p < 0.01) in Gulou than in Daxinggong station.According to the Methodology and methods section, a larger HR value and smaller SDNN, RMSSD, pNN50, and LF/HF values indicate that people feel more pleasure.Therefore, people felt more pleasure in the ground space of Gulou station than in the semi-underground area.Furthermore, participants were generally more pleased with the ground-level public spaces than the semi-underground public spaces.
The comparison results of the two spaces in Daxinggong station were not prominent.The SDNN and RMSSD values are slightly smaller (0.1 < p < 0.2), and the LF/HF ratio was much smaller (p < 0.05) in the ground space than in the semi-underground space; the opposite was true for pNN50.

Impact of urban design features on EDA
A sign test was conducted to analyze the influence of public space characteristics on EDA. Figure 7 shows that the SC and SCL values differ across different spaces.In Daxinggong station, these indicators were generally larger for the semi-underground areas than the above-ground spaces, indicating a higher level of emotional arousal in the semiunderground spaces, in accordance with the literature in the Methodology and methods section.A comparison of the two stations reveals that the ground space in Gulou triggered a higher (p < 0.1) level of SC and SCL (emotional arousal) than that in Daxinggong station.In contrast, the semiunderground space in Gulou only led to a slightly larger (0.1 < p < 0.2) degree of SC and SCL (emotional arousal) than that in Daxinggong station did.The analysis results were non-significant when comparing the two public spaces in Gulou station.

Impact of urban design features on EEG
A Wilcoxon signed-rank test was applied to compare the impact of public space characteristics on EEG signals (Fig. 8).First, the power of the theta wave differed across the two stations, while the power of the alpha wave differed slightly between the different public spaces.There was typically a higher level of brain activity (for theta wave, p < 0.05) for the semi-underground space of Gulou station than there was for that of Daxinggong station.This indicates that the former space provided greater emotional arousal than the latter did, which is similar to the findings of EDA.The ground space of Gulou gives rise to brain activities (emotional arousal) compared with the counterpart space of Daxinggong.This also agrees with the EDA results.Within-station comparisons were not very significant, thus showing that participants were slightly more aroused in the ground space of Daxinggong station than in the semi-underground space.
Second, the alpha wave asymmetries were similar for within station comparisons (Fig. 9).When comparing the semi-underground spaces of Daxinggong and Gulou station, the asymmetry indicator of F3-F4 was more prominent (p < 0.05) in Daxinggong than in Gulou station, which suggests that people may feel more dominant in the former space.When comparing the ground space of the two stations, a few asymmetry indicators, including F7-F8, F3-F4, C3-C4, and P3-P4, presented higher values at Daxionggong than at Gulou station.That is to say, people feel much more dominant in the ground-level spaces at Daxinggong station than in those at Gulou station.

Impact of urban design features on eye blink rate
A Wilcoxon signed-rank test was introduced to compare the eye blink frequency in different spaces.Figure 10 indicates that the value differs between the two spaces of Daxinggong station; the semi-underground space triggered less eye blink (p < 0.1), meaning it was more pleasing to people (see the Methodology and methods section).This result is consistent with the HRV results measured by pNN50.Thus, we conclude that the semi-underground space of Daxinggong station was more pleasing to humans than its ground-floor space.People feel slightly more pleasure (with less eye blink) in the ground space of Gulou than in the ground space of Daxinggong station, which agrees with the HRV results.

Subjective experience
By comparing participants' subjective feelings shown in Fig. 11, the semi-underground space was more pleasing than the above-ground space at Daxinggong station.The inverse could be observed at Gulou station.As for the semi-underground space, that at Daxinggong was more satisfying than that at Gulou, while for the above-ground spaces, that at Gulou was more satisfying.Regarding the two stations' physical environment, both were good in terms of space quality and comfort, although their usability and identifiability were relatively poor.Interestingly, the aesthetic value differs between the two stations: Daxinggong station is more aesthetically well-designed than Gulou station.

General correlations between urban design features and emotional responses
In agreement with humans' subjective experiences, quantitative assessment indicates that at Daxinggong station, people's emotions were more aroused, and they may feel more pleased in the semi-underground space than in the above-ground area.The former space is more enclosed by painted walls and trees with a lower level of D/H (2/1).At Gulou station, the above-ground space was more pleasing and possibly triggered a higher emotional arousal in people than the semi-underground space did.Such aboveground space features a street park and small paving open spaces with a low D/H ratio (1/1).
When comparing the two stations, physiological measurements show that people were more pleased and Fig. 8 Influence of four public spaces on alpha and theta waves."U" represents the semi-underground spaces, and "G" indicates the aboveground spaces; "1" refers to Daxinggong and "2" to Gulou station emotionally aroused in the public spaces at Gulou station than in the corresponding areas at Daxinggong.However, the two public spaces in Daxinggong proved more approachable than those in Gulou station.This indicates that semi-underground spaces equipped with landscape stairs and a gentle slope, transparent surrounding walls, as well as a small store may increase users' pleasantness and emotional arousal but could lead to a feeling of out of control.Moreover, above-ground spaces that is more enclosed and full of vegetation may have similar impacts.
Although the qualitative assessment agrees with the finding that the ground-level public space at Gulou station was more pleasant than that at Daxinggong station, the results for the semi-underground spaces are different.The participants' subjective experiences show that the semi-underground space of Daxinggong station was more pleasing than that of Gulou.This finding contrasts with the quantitative analysis results-the better aesthetic design of Daxinggong station is a probable reason for these results.

Recommendations
The results can be summarized in the following recommendations for architects, urban designers, and planners to promote an integrated design of transport stations, public spaces, and humans systems: • Open space enclosure and pedestrian access: First, increasing the enclosure by decreasing the D/H of public spaces can bring much more pleasure and trigger greater emotional arousal to people.Particularly, a D/H ratio between 1/1 and 2/1 is recommended which confirms Sitte et al. (1965) and Jacobs' (1993) findings.Second, when designing a semi-underground open space, under a defined D/H, to lower the enclosure of the space by changing one side of the surrounding walls to green slopes/landscape stairs or replacing the solid walls with glass walls can increase human pleasure and emotional arousal; however, such places may make people feel less

Urban design guidelines and policy initiatives
Based on these results, the following urban design guidelines and policy initiatives could be generated (also see Table 5).
• Spatial scale and perception: (1) Control the enclosure of the public space within a certain range, preferably a D/H between 1/1 and 1/2; (2) The surrounding buildings should use more transparent materials or greenery to relieve the oppressive feeling of the enclosure; (3) The enclosed interface of the semi-underground space can be combined with elements such as slopes, stairs, entrances, and businesses to increase its spatial richness; (4) Moderately increase the opening area of the underground public space facing the ground.(2) The usability and identification of public space should be enhanced.For example, urban furniture such as seats, street lamps, and advertisements can be added.
(3) Combining specific station cultural characteristics and public art elements to enhance space art and cultural attributes.

Other contributions
This study adopted a socio-technical systems perspective on subway stations.This approach emphasizes the technical side of present engineered systems and, most importantly, highlights the role that humans play in determining the value and future of infrastructure.Hence, these results can be of paramount importance for TOD planning by making stations more attractive and pleasant to people and benefiting the surrounding areas.For example, a busy station that is a hub for social interaction for the community may attract businesses and increase revenue in the surrounding area, thus contributing to the broader economic growth of the surrounding area.
Although this study was initiated with assessments of two specific subway station catchment areas in Nanjing, the Fig. 10 Influence of four public spaces on eye blink frequency."U" represents the semi-underground spaces, and "G" indicates the aboveground spaces; refers to "1" implying Daxinggong and "2" to Gulou station public space patterns and urban design characteristics that were observed here are commonly encountered in many other cities.In addition, participants in this experiment were of different ages, professions, and sexes; therefore, the emotional response measurements of these people are expected to represent the reactions of a wide range of citizens to different space designs within the limitations of the experiment assumptions and the accuracy of the measurement devices.
Compared with the indoor experiments using VR techniques to simulate different urban spaces, the outdoor experiments provide insights into people's feelings when exploring real-life urban spaces using multiple senses such as sight, smell, touch, and hearing.Moreover, they can walk, pass through, or sit at different levels of the public spaces.By letting the participants first step from the exit of a subway station, their real feelings when walking out from an indoor underground space to an open space were captured.

Limitations and future research
This study highlights three possible research avenues that can be explored in the future:

Conclusion
TOD is a powerful urban planning strategy to enhance cities' sustainability and provide socio-economic benefits.Existing literature has revealed that more research is needed in the human-centered design of urban spaces around transport stations.Human emotional responses to stations and public spaces needs to be investigated using controlled experiments, case studies, and the latest technologies.To address this gap, this study took a socio-technical systems viewpoint to examine the impacts of using different design strategies in public spaces around subway stations on human emotional responses.A case study was conducted in Nanjing, China, to obtain quantitative physiological data and qualitative, subjective experience data from a human factors experiment at the catchment areas of two subway stations (Daxinggong and Gulou).As it is fundamentally a multidisciplinary research issue, a research team was built among those in the urban design, architecture, and human factors engineering fields.This study ultimately leads to design suggestions that can offer an understanding of design criteria for urban designers, planners, and other decision-makers to design humanoriented and place-based TOD.Furthermore, the findings will benefit practitioners and researchers in selecting relevant design recommendations to trigger positive human emotions.The D/H of the public space be designed between 1/1 and 1/2

Appendix A
The public spaces' surrounding buildings use transparent materials or greenery The semi-underground public space be equipped with slopes, stairs, entrances, and businesses The underground public space be partly opened to face the ground Walking path experience The continuity of the walking path be increased Visual centers be added at different nodes of the walking path Human flow and urban vitality guidance Small businesses and gray space be added to the public space Pavement design of the public space be changed Space management and art experience improvement The space utilization restrictions imposed by subway station management rules be lessened Usability and identification of public space be enhanced Specific station cultural characteristics and public art elements be designed

Fig. 3
Fig. 3 Spatial pattern of the surrounding areas of the two stations

Fig. 4
Fig. 4 Procedure for human factors experiment

Fig. 6
Fig.6Influence of four public spaces on HRV indicators."U" represents the semi-underground spaces, and "G" indicates the above-ground spaces; "1" refers to Daxinggong and "2" to Gulou station

Fig. 9
Fig.9Influence of four public spaces on alpha wave asymmetry."U" represents the semi-underground spaces, and "G" indicates the aboveground spaces; "1" refers to Daxinggong and "2" to Gulou station (A) Comparison with more case studies under different conditions.To further eliminate bias of the two case studies, more case studies are needed in future research to get more general correlations between the physical environment and human emotions.Future experiments can also explore the interaction under different weather conditions and normal/emergency situations.(B) Examination of the link between specific design elements and emotion.Considering that it is infeasible to use all the urban design strategies in a real public space, and that it was not appropriate to ask the participants to experience too many spaces (the increased time required would tire them), only certain essential design strategies for the public spaces around the station were tested and compared.The next step is to use VR experiments to test additional urban design strategies and to discover the relationships between specific urban design elements and physiological measurements in public areas to supplement this study.(C) More indicators for human emotional states measurement.In this study, people's physiological states were assessed using four commonly used indicators, HRV,

Fig. 11
Fig. 11 Subjective experience of participants at different public spaces

Table 1
(Mauss and Robinson 2009)ensions of emotion using the PAD model(Mauss and Robinson 2009)HRV heart rate variability, EDA electrodermal activity, EEG electroencephalogram

Table 2
Physiological measurement indicators used in this study

Table 4
Demographic characteristics of the 30 participants Urban design characteristics of open public spaces.

Table 5
Urban design and policy recommendations