Patterns and associations of summer thermal comfort and students’ physical activity in campus green spaces

Abstract

Campus green spaces (CGSs), as integral components of campus life, serve as primary venues for student activities and learning. However, with the increasing frequency of urban extreme heat events, creating thermally comfortable spaces for physical activities (PA) through campus green space design has become a crucial issue. Taking the West Campus of Yangtze University in China as a case study, we analyzed thermal comfort and physical activities across 10 locations representing three types of campus green spaces (plaza-type, rest-type, and leisure-type) through field measurements, 982 questionnaire surveys and 3864 recorded activity events during summer. Our findings reveal that: (1) thermal comfort and physical activities showed an inversely related, phase-dependent pattern across space typologies, with the association generally attenuated in the high-temperature period (11:30 AM–15:30 PM) and significantly amplified in the cooling period (15:30–19:30 PM); (2) different types of campus green spaces revealed divergent thermal and usage characteristics, with plaza-type green spaces being heat-vulnerable during the high-temperature period and showing the lowest usage levels, rest-type green spaces consistently providing more favorable thermal conditions yet only medium utilization, and leisure-type green spaces sustaining the highest overall usage rates but with suppression of activity intensity during heat events; (3) low-intensity activities (e.g., walking or board games) showed the greatest thermal sensitivity, particularly in plaza-type green spaces, where each 1 °C temperature increase corresponded to a decrease of 0.80 persons in activity participation (p < 0.001). This study provides scientific evidence for optimizing campus green space design and enhancing the suitability of outdoor activities under thermal stress conditions.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-026-37253-z.

Introduction

Campus green spaces, student health, and the campus heat-risk context

University campuses are not only cradles for knowledge transmission and innovation but also crucial physical carriers for the healthy development of young students¹. Within this specific social ecosystem, Campus Green Spaces (CGSs), as an indispensable component, play multiple key roles^2,3. Beyond optimizing the overall campus landscape with their natural aesthetic value and providing platforms for rest⁴, socialization⁵, and cultural exchange⁶, they serve as the most direct and accessible outdoor venues for promoting student Physical Activity (PA)^7,8. Extensive research has confirmed that regular PA has profound and positive effects on college students’ comprehensive development. Physiologically, it helps reduce the risk of chronic diseases such as obesity and cardiovascular diseases while enhancing physical fitness and immunity^9,10. Psychologically, outdoor activities have been proven effective in alleviating academic stress, reducing anxiety and depression, and improving subjective well-being and emotional stability^11,12. Cognitively, moderate physical exercise can enhance attention, executive function, and memory, thus positively impacting academic performance¹³. However, the extent to which these benefits can be realized in daily student life increasingly depends on how campus outdoor environments perform under heat stress. Therefore, creating campus environments that encourage and support daily PA has become a core issue in environmental construction and health promotion for modern higher education institutions. CGSs, with their openness, accessibility, and natural attributes, should serve as the core physical environment carrying this function.

Currently, the challenges posed by global climate change increasingly threaten the effectiveness and safety of campus outdoor environments, with extreme heat events being particularly prominent^14,15. According to World Meteorological Organization data, global average temperatures have repeatedly broken records in recent years, with heat waves showing significant increases in intensity, duration, and frequency¹⁶. When this macro-climate background is contextualized to the micro-environment of university campuses, it often intensifies the urban heat island effect due to dense building clusters¹⁷ and extensive hardscape surfaces¹⁸, resulting in significantly higher local temperatures compared to surrounding areas. When students are exposed to excessive heat loads during outdoor activities, they face various health risks, including but not limited to heat stress, dehydration, heat stroke, and potentially cardiovascular emergencies^19,20. These risks not only directly diminish students’ willingness and ability to participate in outdoor activities but also pose potential threats to their physical health. Against this background, students unconsciously incorporate environmental thermal comfort into their considerations when choosing whether to engage in outdoor activities and what types of activities to pursue. The PAs discussed in this paper can be roughly categorized into four types based on their metabolic equivalents (MET) and energy consumption levels^21,22: static activities (e.g., sitting, reading), low-intensity activities (e.g., walking, board games), moderate-intensity activities (e.g., jogging, badminton), and high-intensity activities (e.g., basketball, fast cycling). Different intensity activities have varying sensitivities to thermal environments, and understanding these differences is crucial for refined campus outdoor space design.

Taken together, campus green spaces simultaneously represent a primary setting for health-promoting physical activity and a frontline environment of heat-related risk. As heat stress becomes more frequent and intense, students’ daily activity choices are increasingly filtered through thermal conditions, and different activity intensities are unlikely to respond in the same way. Understanding how thermal environments in campus green spaces shape patterns of use across activity types is therefore a prerequisite for creating outdoor settings that can sustain their educational and health-supporting roles under a warming climate.

Cooling mechanisms and typology differences of green spaces

Fortunately, both scientific and design communities widely recognize that green spaces themselves are effective natural solutions for addressing urban heat and improving microclimate^23,24. Through vegetation’s ecological processes, green spaces can create significant Park Cool Island effects. The cooling mechanism primarily stems from two aspects: First, large tree canopies provide extensive shading, effectively blocking most direct solar radiation from reaching the ground, thus significantly reducing surface temperature and the Mean Radiant Temperature (MRT) perceived by human bodies²⁵, and improving outdoor thermal comfort (OTC) indices such as physiologically equivalent temperature (PET), the predicted mean vote (PMV), and universal thermal climate index (UTCI)²⁶. Second, vegetation through Evapotranspiration converts liquid water into water vapor and releases it into the atmosphere, a strongly endothermic process that directly reduces surrounding air temperature while increasing air humidity to some extent, jointly regulating the microclimate²⁴. Therefore, well-designed campus green spaces can create locally cool and comfortable microenvironments through optimized vegetation configuration and spatial layout, providing valuable “Thermal Refugia” for students during hot summer days, thus ensuring and encouraging continued outdoor physical activity.

Although green spaces have recognized cooling potential, recent granular analyses of UHI–heatwave interactions indicate that extreme-heat amplification is highly heterogeneous across urban form and time, and that vegetation’s cooling capacity can weaken during intense heatwaves, suggesting that green infrastructure alone may be insufficient²⁷. Consistent with this, campus green spaces often include multiple subtypes with different functions and forms, showing significant differences in their microclimate regulation efficiency^28,29. For example, plaza-type spaces dominated by hardscape with low vegetation coverage might have thermal environments similar to or worse than typical urban built areas¹⁸. Rest-type spaces characterized by dense tree shade and resting facilities might provide optimal cooling effects²³. Between these extremes, leisure-type spaces with open lawns and scattered trees present different thermal comfort characteristics.

The cooling mechanisms and typological contrasts described above indicate that campus green spaces do not function as equally effective thermal refugia, especially under more frequent and intense heat events. Rather, their cooling and usage profiles likely form a gradient shaped by underlying surfaces, vegetation structure, spatial form, and function, positioning campus green spaces as thermally heterogeneous infrastructures rather than uniformly cooling backdrops.

Knowledge gap, measurement approaches and study questions

The preceding discussion indicates that current research has begun to focus on cooling differences among urban green spaces at the macro level. However, systematic comparisons of thermal comfort performance and intrinsic differences among different types of green spaces in university campuses, a specific functional zone, remain relatively scarce^30,31. More importantly, the existing studies have not clearly revealed how the thermal comfort differences caused by spatial type variations specifically affect the actual physical activities of student groups, including whether they cause the redistribution of active populations among different spaces, and whether they alter the types and intensities of activities people engage in. Clarifying these “space–climate–behavior” linkages is the key to filling the knowledge gap in the current research field and guiding the precise campus environment design.

To quantify and explain these complex interactions, researchers have developed a relatively mature set of measurement and survey methods. For thermal comfort assessment, research typically combines objective physical measurements with subjective perception surveys^32,33. Objective measurements rely on equipment such as portable weather stations to monitor multiple microclimate parameters including air temperature (Ta), relative humidity (RH), wind speed (v), and globe temperature (Tg), then calculate comprehensive thermal indices such as Physiological Equivalent Temperature (PET) or Universal Thermal Climate Index (UTCI) to comprehensively evaluate environmental thermal physiological stress on human bodies^33,34. Subjective surveys mainly collect users’ direct perceptions of current thermal environments (Thermal Sensation Vote, TSV) and comfort evaluations through questionnaires^35,36. In physical activity research, methods are equally diverse³⁷. Direct observation methods, such as Behavior Mapping or specific observation systems (like SOPARC), can accurately record information about users’ numbers, activity types, gender, age, etc., in spaces, though they are time and labor-intensive^38,39. Questionnaire surveys can efficiently collect large-scale self-reported data about individual activity habits, preferences, frequency, and influencing factors^40,41. This research integrates these proven methods, combining “physical measurement-questionnaire survey” approaches to comprehensively and deeply interpret the complex relationship between campus green space thermal comfort and student physical activity.

This study’s core objective is to address the identified knowledge gap by empirically explaining the thermal comfort differences among different types of campus green spaces in summer and revealing how these differences influence students’ physical activity choices and distribution. To achieve this goal, we established three core Research Questions to guide the research: (1) Under typical summer weather conditions, what are the differences in thermal comfort performance (with PET as the main indicator) among plaza-type, rest-type, and leisure-type green spaces on university campuses? (2) Do the thermal comfort differences lead to significant variations in the number of students (i.e., activity occupancy rate) engaging in activities in different types of green spaces? (3) How do changes in thermal comfort conditions, especially temperature increases, specifically affect students’ participation in various physical activities (static, low, moderate, high intensity) in different types of spaces? The research results will provide scientific basis for thermal mitigation strategies in campus green space design, contributing to climate-adaptive campus construction.

Method and data

Study area

This study selected the West Campus of Yangtze University as the research area (Fig. 1). The campus is located in Jingzhou District, Jingzhou City, Hubei Province, China (112°23’E, 30°21’N), within the subtropical monsoon climate zone characterized by distinct seasons and synchronous rainfall and heat patterns. The campus covers approximately 644.6 mu (about 43 hectares) with a greening rate of 35%, featuring beautiful landscape gardens and serving as a demonstration unit for garden-style universities in Hubei Province. Based on preliminary field investigations and pilot studies, our research team selected 10 representative green spaces as basic data collection points. Specifically, site selection followed three principles: (1) typicality of location and spatial scale, covering the campus’s primary functional zones (academic, residential, sports, and waterfront) and accommodating multiple size classes while avoiding bias toward peripheral or atypical spaces; (2) diversity of spatial content and structure, encompassing varied spatial morphologies, vegetation types, ground-surface compositions, and adjacent buildings; (3) behavioral richness, ensuring high activity frequency and heterogeneity in user groups and activity types, in order to obtain more reliable behavioral observations. The observation points were systematically categorized into three spatial types: plaza-type, leisure-type, and rest-type green spaces (Table 1), guided by their spatial attributes (such as underlying surface composition, vegetation structure, greening coverage), functional uses, and Sky View Factor (SVF). Based on the survey and statistical analysis of the attribute characteristics at each point, the classification followed a predefined decision sequence: when a hardscaped forecourt enclosed by buildings primarily serves circulation and assembly functions, it is categorized as plaza-type. Such spaces exhibit high hardscape coverage, low vegetation proportion, low greening coverage, predominantly herbaceous vegetation or sparse tree-herb mixes, and high openness (SVF typically 0.60–1.00). When vegetation structure and stay-oriented facilities delineate the area as a “room” or “garden,” it is categorized as rest-type, characterized by low hardscape coverage, high vegetation proportion and greening coverage, commonly with a multilayer tree-shrub-herb structure, and low openness (SVF typically 0.00–0.30). The remaining green spaces used for informal recreational activities were categorized as leisure-type. Identification of these spaces primarily relied on use functions (e.g., recreation, sports), supplemented by site features (e.g., open lawns, proximity to water bodies). Their morphologies are diverse rather than characterized by a single canonical form, and to meet user needs, their openness typically falls between the plaza-type and rest-type, which is at a medium level (SVF typically 0.30–0.60).

Fig. 1.

Study areas. (a) Hubei Province. (b) Jingzhou City. (c) Location of the field survey in West Campus of Yangtze University.

Table 1.

Classification and attribute characteristics of green spaces in campus.

Microclimate parameters acquisition and thermal comfort assessment

To accurately assess the thermal environment of campus green spaces and its impact on users’ thermal comfort, systematic microclimate parameter measurements were conducted. Following international microclimate observation standards, PC-8 automatic weather stations (accuracy: temperature ± 0.1℃, relative humidity ± 2%, wind speed ± 0.2 m/s) and RS-HQ black globe thermometers (accuracy: ±0.5℃) were installed at each measurement point at 1.5 m above ground level (approximate to human body center height)⁴². These instruments simultaneously collected basic meteorological elements including air temperature (Ta), relative humidity (RH), wind speed (Va), and globe temperature (Tg). Observations were conducted during typical summer days (clear, cloudless weather) from 7:00–19:00, with sampling intervals set at 1 min to ensure data continuity and representativeness.

Based on the acquired microclimate parameters, this study employed the internationally recognized RayMan pro model to calculate the Physiological Equivalent Temperature (PET) as the core indicator for thermal comfort assessment⁴³. PET was selected because it integrates air temperature, mean radiant temperature, wind speed, and relative humidity within the Munich Energy-balance Model for Individuals (MEMI) and has been widely validated for outdoor thermal comfort assessment in urban green and open spaces, enabling consistent comparison across campus space typologies and daytime phases. PET is theoretically founded on the human body heat balance equation:

1

where is the metabolic rate (internal energy production by oxidation of food); is the physical work output; is the net radiation of the body; is the convective heat flow; is the latent heat flow to evaporate water diffusing through the skin; is the sum of heat flows for heating and humidifying inspired air; is the heat flow due to evaporation of sweat; and is the storage heat flow for heating or cooling the body mass. The mean radiant temperature (Tmrt), a crucial input parameter for PET calculation, was derived from the measured globe temperature using the following equation:

2

where is mean radiant temperature (°C); is globe temperature (°C); is air velocity (m/s); is air temperature (°C); is the emissivity of the globe (0.95 for black-colored globe); and is the globe diameter (m). Using these measured and calculated parameters, the RayMan model computed PET values by simulating the human body thermal conditions of a standardized person (25 years old, 1.75 m height, 75 kg weight) wearing typical business clothes (0.9 clo) and having a moderate activity level (80 W) (Table 2).

Table 2.

Parameter list of microclimate measurement instruments.

Instrument	Measurement parameters	Relevant parameters	Temporal interval
Weather station PC-8	Air temperature Ta (°C)	Measurement range: −40 to 80 °C Resolution: 0.1 °C Accuracy: ±0.1 °C	60s
	Relative humidity RH (%)	Measurement range: 0–100% Resolution: 0.1% Accuracy: ±2%	60s
	Wind speed Va (m/s)	Measurement range: 0–70 m/s Resolution: 0.1 m/s Accuracy: +0.3 + 0.03 V/ms	60s
Globe temperature logger RS-HQ	Globe temperature Tg (°C)	Standard accuracy: ±0.2 °C (25 °C) High accuracy: ±0.1 °C (25 °C) Temperature resolution: 0.1 °C Globe diameter: 0.15 m	60s

Notably, PET was computed under the MEMI framework using constant reference-person settings. Keeping these settings fixed ensured comparability across sites and daytime phases (morning warming, midday peak, afternoon cooling), but any departure between the assumptions and actual clothing or activity can shift absolute PET values. Uncertainty in the microclimatic inputs likewise propagates to PET: in heterogeneous outdoor settings, PET is particularly sensitive to Tmrt and Va, while measurement accuracy and short-term variability in Ta and RH also contribute. To reduce measurement artifacts, Ta and RH probes were radiation-shielded, the anemometer was sited in unobstructed exposure, and 60-s logs were quality-checked. Nevertheless, because Tmrt is derived from a 0.15 m black globe whose thermal response is on the order of minutes, brief short- or long-wave fluctuations (e.g., transient shading near trees or facades) may still be partially smoothed and short-term PET excursions slightly damped. For these reasons, we primarily analyzed PET as a continuous variable, while using locally adapted PET heat stress categories for reports, and we emphasized relative patterns across space typologies and daytime phases.

Physical activity level survey

This survey included a questionnaire with four main items related to respondents’ activities (Table 3): activity type, activity frequency, time of occurrence, and duration. Behavioral observations recorded more detailed activity types, activity frequency per unit time, and other on-site behavioral characteristics. To ensure consistency and comparability of field data collection and to minimize recall bias and participant reactivity, each observation site was staffed with one behavioral observer and one interviewer. The interviewer administered and collected an anonymous, neutrally worded questionnaire on site during each participant’s current activity period, while the observer, from a non-intrusive location, refrained from interacting with users and systematically recorded activities using standardized definitions and recording forms, taking concurrent photographs. Behavioral observations and on-site questionnaires were conducted in fixed 60-minute windows, and PET was averaged to the same hourly windows for analysis. A total of 982 valid questionnaires were collected, and 3864 activity events were recorded. Without accounting for duplicates, the combined count of respondents and individuals recorded indicates 1975 males (51.1%) and 1889 females (48.9%). The gender distribution of the sample was balanced, thus avoiding possible bias and data errors due to gender differences. Questionnaire and observational data were matched and merged by site and time period for subsequent analysis. It should be noted that prior to commencing any research-related activities, written informed consent was obtained from all participants. They were apprised of the voluntary nature of their involvement, with the liberty to decline or discontinue involvement at any point without providing a reason. Additionally, the study protocol, including the questionnaire and behavioral observations, was reviewed and approved by the Yangtze University, Institute of Science and Technology Development Institutional Review Board (ISTD-IRB) committee (approval number: 202501039). All procedures were conducted in accordance with the ethical principles of the Declaration of Helsinki (as revised in 2000) as well as relevant guidelines and regulations.

Table 3.

Physical activities questionnaire.

Question	Description	Options
Q1	What is the main activity you are here?	recreation/extracurricular activities/reading and study/exercise/communication/club activities/walking/other
Q2	How frequently do you active here?	once a week/2–3 times a week/once a day/2–3 times a day/more than 3 times a day
Q3	What time of day do you usually active here?	before 7:30/7:30–09:30/9:30–11:30/11:30–14:30/14:30–16:30/16:30–19:30/after 19:30
Q4	How long do you usually active here?	Less than 15 min/15–30 min/30–60 min/more than 60 min

Activity frequency and intensity were identified as key indicators for quantitatively assessing population physical activity levels. The frequency of activities per unit time not only directly reflects the intensity of green space usage but also serves as an important reference for evaluating spatial quality and user preferences^21,22. Considering the variations in energy expenditure across different types of activities, the study established a systematic activity intensity classification system based on their metabolic equivalents (MET) (Table 4). To mitigate the effects of climate and other external factors and ensure the natural occurrence of physical activity, the field investigation was conducted from 14 to 16 June 2025, representing the typical summer period in this region. Based on preliminary meteorological data analysis and thermal variation patterns, the daily observation period was divided into three thermal phases: the warming period (7:30–11:30), the high-temperature period (11:30–15:30), and the cooling period (15:30–19:30). This temporal division ensures comprehensive coverage of diurnal thermal variations and their potential impact on human activities.

Table 4.

Categorization of physical activity intensity.

Intensity level	Metabolic equivalent (MET)	Typical activities
Static	1.0–1.5	Sitting quietly, reading, conversing, fishing, photography, dining, sunbathing
Low	1.6–2.9	Leisure walking, dog-walking, children’s play, board/card games, singing
Medium	3.0–5.9	Cycling, brisk walking, badminton, tennis, square dancing, resistance training, Tai Chi, aerobics
High	≥ 6.0	Jogging, inline skating, skateboarding, rope skipping

Correlation and regression analysis

To thoroughly investigate the relationship between thermal environment and physical activity levels, this study constructed a systematic statistical analysis framework. Initially, Pearson correlation coefficient analysis was employed to examine the correlation and significance levels between PET and activity frequencies of different intensities⁴⁴. The Pearson correlation coefficient (r) is calculated as:

3

where and represent the thermal environment and physical activity levels respectively. Subsequently, considering the direct influence of thermal comfort on physical activity, a simple linear regression model was established⁴⁵:

4

where is the dependent variable representing physical activity frequencies of different intensities; is the independent variable of PET. The regression analysis was performed using SPSS 26.0 statistical software, with the significance level set at p < 0.05. Through this model, we can quantitatively assess how thermal comfort influences physical activity levels and predicted activity patterns under different thermal conditions, thereby providing scientific evidence for campus green space management and utilization.

Results

Spatiotemporal variation of PET in different CGSs

The results revealed distinct temporal dynamic patterns in the PET across various green space types during different periods of the day (Fig. 2). Starting at 7:30 AM, PET values for all green space types showed a gradual increasing trend during the Heating period (7:30–9:30 AM). Plaza-type green spaces (G1–G3) exhibited the most rapid warming, with PET rising quickly from an initial 28 °C to 38 °C. During the High-temperature period (11:30 AM–1:30 PM), all green space types reached their respective daily peak PET values. Plaza-type green spaces recorded the highest peak, approaching 42 °C, while the peak PET for rest-type green spaces (A1–A3) was relatively lower, maintained around 32–33 °C. Pie chart data indicated that during the high-temperature period, approximately 52% of the areas experienced severe heat stress, and 40% experienced extreme heat stress, reflecting significant thermal pressure challenges in campus green spaces during summer. In the Cooling period (3:30–6:30 PM), PET values for all types gradually declined, dropping to around 30 °C by 6:30 PM. However, distinct temperature gradients persisted between the different green space types.

Fig. 2.

Analysis of PET variation in different green spaces. (Codes denote observation points by typology: G = plaza-type, X = leisure-type, A = rest-type; see Table 1 for full site names and attributes.).

Through detailed statistical analysis of PET values across the three periods, the study found significant differences in the thermal environment among the different green space types (Fig. 3). During the Heating period, the average PET values for the three types were 31.5 °C, 29.1 °C, and 34.2 °C, respectively, corresponding to mild to moderate heat stress. Upon entering the High-temperature period, all types showed a marked increase in heat stress. Nevertheless, rest-type (A) green spaces demonstrated the best thermal regulation capacity, with an average PET approximately 5–6 °C lower than that of plaza-type green spaces (G). Specifically, the average PET during the high-temperature period was 35.9 °C and 38.7 °C for leisure (X) and plaza-type green spaces (G), both indicating severe heat stress but at different intensities. This disparity may be related to differences in spatial structure, vegetation coverage, and underlying surface characteristics among the green space types²⁵. During the Cooling period, PET values for all types gradually decreased, and the differences between them correspondingly diminished, primarily manifesting as moderate heat stress with average PET maintained between 30 and 32 °C. Notably, rest-type green spaces (A) maintained relatively stable and lower PET values throughout the observation period, indicating their distinct advantage in thermal environment regulation, particularly their stronger heat mitigation capacity during high-temperature weather.

Fig. 3.

Analysis of differences in PET across different periods in various types of green spaces (95% CI). (Codes denote observation points by typology: G = plaza-type, X = leisure-type, A = rest-type; see Table 1 for full site names and attributes.).

Usage patterns and activity intensity of different CGSs

We conducted a statistical analysis of green space usage across different periods (Fig. 4a). The results showed significant variations in usage frequency among the three green space types at different times. During the Heating period, a total of 513 activity occurrences were recorded. Leisure-type (X) green spaces had the highest usage frequency (approximately 200 occurrences), followed by rest-type (A) and plaza-type green spaces (G) (both around 150 occurrences). During the High-temperature period, total activity occurrences decreased to 152, yet leisure-type green spaces (X) still had the highest usage rate (approximately 75 occurrences). In the Cooling period, activity occurrences significantly rebounded to 844, with leisure-type green spaces (X) reaching the highest frequency (approximately 420 occurrences), far exceeding rest-type (about 230 occurrences) and Plaza-type (about 180 occurrences). This pattern may also be related to student lifestyles, which favor afternoon activities. These results indicate that leisure-type green spaces (X) maintained high usage rates across all periods, although participation was suppressed during the high-temperature period.

Fig. 4.

Usage patterns of CGSs across different periods (a) and activity intensities (b) (95% CI). (Codes denote observation points by typology: G = plaza-type, X = leisure-type, A = rest-type; see Table 1 for full site names and attributes.).

Furthermore, regarding activity intensity distribution (Fig. 4b), low-intensity activities dominated across all green space types, totaling 915 occurrences, significantly higher than moderate-intensity (144 occurrences) and high-intensity activities (24 occurrences). Sedentary activities also accounted for a considerable proportion, reaching 463 occurrences. The distribution of activity intensity across different periods showed significant variations (Fig. 5). During the Heating period (7:30–9:30 AM), low-intensity activities were most prevalent (315 occurrences), primarily occurring in leisure-type green spaces (X), followed by sedentary activities (approximately 139 occurrences). Although overall activity frequency significantly decreased during the High-temperature period (11:30 AM–1:30 PM), low-intensity activities still dominated (approximately 102 occurrences), and usage differences between green space types narrowed. During the Cooling period (3:30–6:30 PM), activity frequency notably increased, particularly in leisure-type green spaces (X) where low-intensity activities surged from 50 occurrences in the high-temperature period to 258 occurrences. These data suggest that the intensity of green space usage is closely related to the time of day and temperature, and that different green space types exhibit clear differences in their capacity to support activities of varying intensities. Leisure-type green spaces (X) demonstrated the strongest activity carrying capacity, especially for low-intensity and sedentary activities. High temperatures significantly inhibited all activity types, but this effect was more pronounced for moderate and high-intensity activities.

Fig. 5.

Usage patterns of CGSs by activity intensity during different periods (95% CI). (a) Heating period. (b) High-temperature period. (c) Cooling period. (Codes denote observation points by typology: G = plaza-type, X = leisure-type, A = rest-type; see Table 1 for full site names and attributes.).

Exploration of influence mechanisms between PET and PA

Correlation analysis further revealed the relationship between PET and physical activities of different intensities (Fig. 6). The results showed a significant negative correlation between PET and overall activity level (r = −0.52, p < 0.001), indicating that overall activity participation markedly decreased as heat stress increased. Specifically, sedentary activities (r = −0.48, p < 0.001) and low-intensity activities (r = −0.47, p < 0.001) exhibited the most significant negative correlations with PET, suggesting these activities are most sensitive to thermal environment changes. Moderate-intensity activities showed a relatively weaker negative correlation with PET (r = −0.29, p < 0.01), while high-intensity activities displayed the weakest negative correlation (r = −0.18, p < 0.05). Significant correlations also existed between activities of different intensities, with the strongest positive correlation observed between sedentary and low-intensity activities (r = 0.96, p < 0.001), indicating these activities often co-occur or transform into each other.

Fig. 6.

Correlation matrix between PET and PA.

Regression analysis of PET versus activity levels within different green space types revealed that all three types exhibited a trend of decreasing activity with increasing PET, but the slopes of decline differed significantly (Fig. 7). For overall activity level, X-type green spaces showed the steepest decline (y = −1.78x + 74.57), indicating their activity levels were most sensitive to temperature changes. In contrast, A-type green spaces showed a relatively flatter decline (y = −2.33x + 83.85), suggesting better thermal environment adaptability. Regarding activity intensity, sedentary and low-intensity activities showed clear negative trends across all green space types. For leisure-type green spaces (X), the regression equation for sedentary activities was y = −0.58x + 24.15, and for low-intensity activities it was y = −0.95x + 42.30, both indicating high temperature sensitivity. Moderate-intensity activities showed the most significant decline in plaza-type green spaces (G) (y = −0.41x + 16.19). High-intensity activities showed weak correlations across all types, with regression coefficients generally close to zero, likely due to their inherently low occurrence frequency.

Fig. 7.

The regression relationship between PET and PA across different CGSs (95% CI). (a) The relationship between PET and Total number participants. (b–e) The relationship between PET and different activity intensities. (Codes denote observation points by typology: G = plaza-type, X = leisure-type, A = rest-type; see Table 1 for full site names and attributes.).

Further analysis of the relationship between PET and activity levels across different periods revealed significant time-dependent characteristics (Fig. 8). During the Heating period, all green space types exhibited relatively mild negative trends, with rest-type green spaces (A) showing the fastest activity decline with rising temperature (y = −2.35x + 80.29). During the High-temperature period, the sensitivity of activity levels to PET generally decreased, reflected in flatter regression curves, possibly because participants had adapted to the high-temperature environment. However, during the Cooling period, the sensitivity of activity levels to PET significantly increased, particularly in rest-type green spaces (A) which showed the steepest decline (y = −11.08x + 362.45). Leisure-type (X) and plaza-type (G) green spaces also showed clear negative correlations (y = −2.57x + 109.75 and y = −2.53x + 95.32, respectively). This temporal variation indicates significant changes in people’s response strategies to the thermal environment across different periods. Particularly during the Cooling period, higher sensitivity to temperature changes may reflect an expectation for more comfortable activity conditions.

Fig. 8.

The regression relationship between PET and PA across different periods (95% CI). (Codes denote observation points by typology: G = plaza-type, X = leisure-type, A = rest-type; see Table 1 for full site names and attributes.).

Discussion

This study systematically integrates microclimate measurements and behavioral observations to reveal the mechanism by which the thermal comfort of CGSs influences students’ PA. Moving beyond traditional single-dimensional environmental assessment models, it quantifies the thermal differentiation among three CGS types during heat events and their differential regulation of PA intensity. This provides a scientific thermo-physiological and behavioral coupling basis for campus landscape design in hot-humid climates. The theoretical value lies in validating the applicability of the thermal adaptation behavior model at the microscale: individual activity decisions are not only constrained by absolute temperature values but are mediated through perceived thermal comfort, forming spatial preferences.

Mechanistic interpretation of thermal comfort differences and physical activity patterns across space typologies

We found that rest-type green spaces (A) exhibited significantly superior ability in maintaining low PET values (32–33 °C) during high-temperature periods compared to plaza-type spaces (42 °C). The contrast can be explained by the cooling effect of vegetation highlighted in the introduction: trees intercept incoming solar radiation through shading, suppress surface and near-surface air heating, and, via evapotranspiration, convert radiative energy into latent heat, thereby cooling leaf surfaces and the surrounding air and creating local cool areas beneath the canopies^46,47. Consistent with this mechanism, our finding also aligns with the vegetation thermal buffering theory proposed by Chen et al.⁴⁸: canopy-driven shading and evapotranspiration can reduce the probability of thermal stress in Type A spaces by 40%.

Concurrently, we noted that recreation-type spaces (X), despite having the highest daily utilization rate, exhibited a “high quantity, low quality” characteristic of PA intensity during heat events, dominated by sedentary activities (61.2%). This pattern is consistent with the behavioral compensation theory⁴⁹ and aligns with the introduction’s framing that thermal conditions shape activity choices and intensities: when the thermal environment exceeds comfort thresholds, individuals spontaneously reduce activity intensity rather than completely avoiding the space. Contrasting with findings in urban parks, this indicates that students in campus settings exhibit stronger thermal tolerance due to schedule constraints.

Differential thermal sensitivity of low- and high-intensity activities and interpretable mechanisms and factors

Low-intensity PA showed a strong negative correlation with PET (r = −0.47, p < 0.001), with its temperature sensitivity being 2.6 times higher than high-intensity activities. Although this appears to contradict the simple expectation that higher metabolic rates should increase thermal sensitivity, the discrepancy can be reconciled by the metabolic cost theory of thermoregulation: low-energy activities like walking are more susceptible to thermal interference, as the imbalance between heat dissipation demands and exercise-induced heat production triggers behavioral cessation⁵⁰. In plaza-type green spaces (G), each 1 °C temperature increase led to a reduction of 0.80 persons engaged in low-intensity activities (p < 0.001), highlighting the thermal vulnerability of unshaded hardscapes. This is consistent with the positive correlation between shade coverage and activity persistence found in Middle Eastern campuses⁵¹.

In addition, methodological factors and behavioral mechanisms specific to field observations likely account for our results to a further extent. Laboratory and self-report studies that link metabolic load to heightened thermal sensation typically assess subjective responses under controlled exercise conditions^52,53. In contrast, our study is based on behavioral observations conducted in uncontrolled outdoor settings. Thus, additional physiological and behavioral pathways in outdoor field settings may explain our findings. Physiologically, vigorous movement generates self-generated airflow that enhances convective and evaporative cooling, partly offsetting metabolic heat⁵⁴; clothing permeability, coverage, and hydration routines typical of sport may increase evaporative efficiency⁵⁵; and relatively fitter or heat-acclimatized individuals can sustain effort at higher core or skin temperatures⁵⁶. Behaviorally, first, individuals engaging in high-intensity outdoor activities (e.g., exercise or sports practice) are typically more motivated and goal-oriented, enabling them to persist under higher thermal load; second, under heat stress, users may engage in micro-site selection (more shade or airflow), time-shifting within the day, or session truncation (maintaining intensity but shortening bouts) rather than discontinuing high-intensity activity altogether^57,58. These behavioral mechanisms are supported by our evidence. In our data, high-intensity activities were infrequent, yet activity counts remained relatively stable as PET increased across space types, consistent with limited heat-related attrition among high-intensity participants. Moreover, during the high-temperature period, the PA–PET regression slope was flatter, suggesting adaptive behavioral regulation—users maintained activity levels by shortening dwell times rather than reducing intensity under sustained heat stress.

Measurement and statistical considerations may also accentuate the apparent contrast between high- and low-intensity activities: high-intensity events are rarer, which can make fitted slopes statistically flatter and less stable; by contrast, the larger baseline and longer dwell times of low-intensity uses mean that modest absolute reductions translate into steeper slopes in count-based regressions. Furthermore, when vigorous bouts are shortened but persist, interval-aggregated counts can understate substitution relative to time-weighted metrics. These points are offered as hypotheses consistent with the observed associations rather than causal claims, and they motivate future work using wearable sensing, session-level duration measures, and micro-exposure mapping.

Spatiotemporal optimization strategies for campus based on the study findings

Based on our findings, we propose spatiotemporal optimization strategies that directly target the observed typology-phase patterns. Plaza-type green spaces (G) exhibit higher PET during the high-temperature period and a notable decline in PET–PA, identifying unshaded hardscapes as heat-vulnerable nodes. Accordingly, we recommend integrating tree matrices to deliver at least 60% effective shade along desire lines and activity clusters while preserving airflow through spacing and canopy porosity. Consistent with the within-day PET dynamics and with the greater PET sensitivity of sedentary and low-intensity uses across types—which is particularly evident in leisure-type green spaces (X), we propose time-specific functional zoning: guide heat-sensitive activities to shaded, well-ventilated pockets during the midday peak heat and activate more open spaces during the cooler periods (after 15:30). Although rest-type green spaces (A) provide favorable thermal conditions, their medium utilization suggests constraints beyond heat, including accessibility, aesthetics and acoustics, perceived safety and legibility, seating supply and placement, and social programming and amenity provision (e.g., Wi-Fi, power, water). Thus, access and draw should be enhanced through improved path connectivity and visual exposure, visible and well-distributed seating, targeted programming, and upgrades to aesthetics and acoustics (e.g., planting and material choices that improve visual quality and provide sound buffering).

At the campus scale, we further advocate thermal risk mapping to prioritize shade-and-airflow interventions and to align with climate-adaptive campus landscape principles⁵⁹. To aid implementation, actions can be phased as immediate operational measures (retiming heat-sensitive programming toward the cooling period, deploying movable seating in shaded and breezy pockets, and using temporary shade), near-term minor works (modular canopies, targeted planting, and bench relocation to micro-shade and breeze paths), and medium-term projects (selective opening of breeze corridors, cool-pavement resurfacing, and utility stubs to support seasonal cooling devices).

Limitations and directions for future research

Furthermore, this study has three limitations: (1) Observations focused on summer (June–August), missing seasonal impact on thermal perception–behavior relationships; future work should include spring transition data; (2) PET as the thermal indicator did not integrate radiation non-uniformity (e.g., abrupt solar radiation changes under tree shade); future studies should optimize microenvironment assessment such as using Local Climate Zone (LCZ) theory⁶⁰; and (3) Social factors (e.g., peer effects) influencing thermal tolerance thresholds were not quantified, potentially explaining the high comfort–medium utilization paradox in Type A spaces. Subsequent research could integrate wearable sensors and machine learning to build multidimensional behavioral prediction models.

Conclusion

This study combined PET monitoring with behavioral observations and on-site questionnaires across ten campus green spaces to quantify how summer thermal comfort relates to students’ PA. Key findings demonstrate: (1) across space typologies, PET was negatively associated with PA, and this association was phase-dependent—weaker in the high-temperature period and stronger in the cooling period; (2) plaza-type spaces emerged as heat-vulnerable nodes during the high-temperature period and showed the lowest usage levels, rest-type spaces consistently provided more favorable thermal conditions yet only medium utilization, and leisure-type spaces sustained the highest overall usage rates but with suppression of activity intensity during heat events; and (3) sedentary and low-intensity activities were more susceptible to thermal conditions, while high-intensity activities remained comparatively stable. These patterns advance understanding of thermal–behavior interactions in campus green spaces and provide a concise, typology- and phase-resolved evidence base to guide climate-responsive campus design and operations. The study’s inferences are associational and pertain to summer conditions at a single campus, with PET used as the core index and hourly aggregation aligning activity counts and thermal data; generalization beyond similar contexts should therefore be made cautiously.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

Conceptualization, S.X. and X.G.; methodology, S.X. and X.G.; software, S.X. and B.L.; validation, S.X. and B.L.; formal analysis, S.X.; investigation, S.X. and B.L.; resources, X.G.; data curation, S.X.; writing—original draft preparation, S.X.; writing—review and editing, S.X. and X.G.; visualization, S.X.; supervision, X.G. and J.B.; project administration, X.G.; funding acquisition, X.G. All authors have read and agreed to the published version of the manuscript.

Funding

This research was supported by The National Natural Science Foundation of China (Youth Program) (NO. 51908063); 2025 Higher Education Science Research Planning Project (No. 25JZ0301); 2023 Hubei Provincial Undergraduate Teaching Reform Research Project (No. 270).

Data availability

The datasets generated and/or analysed during the current study are available in the figshare repository, https://doi.org/10.6084/m9.figshare.30316159.v1.

Declarations

Competing interests

The authors declare no competing interests.