Applied Geography 184 (2025) 103764

Contents lists available at ScienceDirect

Applied Geography
journal homepage: [Link]/locate/apgeog

Riding through the heat: Assessing heat health risk of cycling based on
diurnal temperature and shared-bike big data
Kun Zeng a , Feng Gao a,b,c,* , Yihuan Peng d, Chang Liu a, Wangyang Chen a , Guanyao Li a
a
Guangzhou Urban Planning & Design Survey Research Institute Co.,Ltd., Guangzhou, 510060, China
b
Guangzhou Collaborative Innovation Center of Natural Resources Planning and Marine Technology, Guangzhou, 510060, China
c
Guangdong Enterprise Key Laboratory for Urban Sensing, Monitoring and Early Warning, Guangzhou, 510060, China
d
Guangdong Center for Marine Development Research, Guangzhou, 510220, China

A R T I C L E I N F O A B S T R A C T

Handling Editor: J Peng Fine-scale heat health risks assessment is critical to improve resilience to global warming and extreme heat
events. This study proposed a framework to estimate heat health risks of cyclists using massive bike-sharing
Keywords: trajectories and diurnal dynamic temperature data. This study proposed two indicators, accumulated heat
Heat health risks health risks (CycHeat) and accumulated excessive heat health risks (CycExcessHeat), to estimate the heat health
Heat exposure
risks of bike-sharing cyclists in Beijing, Shanghai, and Xiamen based on massive GPS trajectories and diurnal
Bike-sharing
dynamic temperature data. Results show that there is no significant difference in terms of the CycHeat between
Mobility
ECOSTRESS the morning peak and evening peak in both three cities. While the CycExcessHeat in the morning peak is
significantly higher than that in the evening peak in Beijing, Shanghai, and Xiamen (p < 0.05). That is, the
indicator that focuses on extreme heat is more reflective of spatiotemporal heterogeneity in cyclists’ heat health
risks. Finally, the relationship between CycExcessHeat and the built environment was regressed, and the model
results align with the literature focusing on cycling frequency. This study provided a framework to estimate heat
health risks of outdoor mobility, which provides management implications for transportation authorities and
urban planning in the context of SDGs 3 and 11.

1. Introduction human health and urban environments, such research enhances resil­
ience to global warming and extreme heat events, leading to more
With global urbanization and the consequent intensification of the livable, equitable, and sustainable cities.
urban heat island effect, extreme heat has become a global problem that Heat health risks have emerged as a critical research focus in geog­
seriously endangers the health of urban residents (Biardeau et al., 2019; raphy, environmental science, and related disciplines, emphasizing the
Gasparrini et al., 2017; Yang, Peng, Jiang, et al., 2024; Yang & Peng, interaction between humans and their environment (Yao et al., 2024;
2025). In the United States, 20 percent of hazard deaths are attributed to Zha et al., 2024). This interdisciplinary field examines the health im­
extreme heat events, which is as many as the total number of deaths pacts of human exposure to high temperatures, integrating data on land
caused by floods as hurricanes (Borden & Cutter, 2008; National surface temperature (LST) and population dynamics (Baqa et al., 2025;
Weather Service, 2018). Excessive heat exposure can lead to energy Foroutan et al., 2024). Meteorological station observation data, and
imbalance and cause body temperature to exceed the safe temperature, remote sensing data are the most widely used sources in heat exposure
which in turn affects human thermal comfort, physical and mental research (Kuras et al., 2017). However, station-based observation data,
health, and cognitive ability (Borden & Cutter, 2008; Cheng & Sha, due to their scattered and sparse distribution, cannot fully capture the
2024; Ho et al., 2018; Tian et al., 2025). Therefore, research on heat spatial heterogeneity of urban LST (Cao et al., 2020; Gasparrini et al.,
health risks is imperative for mitigating the deleterious effects of 2015; Ho & Wong, 2019; Noelke et al., 2016; Wang et al., 2018).
excessive thermal stress, thereby contributing significantly to the Although remote sensing satellites, such as MODIS and Landsat, provide
attainment of Sustainable Development Goals (SDGs) 3, 11, and 13 (Tian extensive spatial coverage, their temporal resolution is often insufficient
et al., 2025; Watkins et al., 2021). By clarifying the impacts of heat on for high-frequency analysis (Guo et al., 2019; Jenerette et al., 2016;

* Corresponding author. Guangzhou Urban Planning & Design Survey Research Institute, China.
E-mail address: 2111801048@[Link] (F. Gao).

[Link]
Received 14 March 2025; Received in revised form 30 July 2025; Accepted 23 August 2025
Available online 4 September 2025
0143-6228/© 2025 Elsevier Ltd. All rights are reserved, including those for text and data mining, AI training, and similar technologies.
K. Zeng et al. Applied Geography 184 (2025) 103764

Orusa et al., 2023; Pearsall, 2017; Wu et al., 2022; Yasumoto et al., bike-sharing travel trajectories to investigate the potential heat envi­
2019). Neither MODIS nor Landsat can simultaneously meet the de­ ronmental health risks of active commuting during morning and evening
mands for high spatial and temporal resolution, which is crucial for peak hours. Furthermore, we identify high - risk areas through threshold
understanding dynamic human exposure to heat. adjustments, employ spatial regression analysis to examine the drivers
Existing studies on environmental health risks or exposure have of high-risk distribution, and propose heat-mitigation planning and
primarily concentrated on static population distributions when assessing strategies tailored to cycling behaviors by integrating the results of heat
heat health risks, with limited incorporation of human mobility and environmental health risk identification and driver modeling. The
travel patterns (Cao et al., 2020; Kuras et al., 2017; Yasumoto et al., findings of this research can offer data-driven insights for urban plan­
2019; Yuan et al., 2022). For example, Yuan et al. (2022) delved into the ning and decision-making, as well as spatial patterns of heat exposure
diurnal dynamics of heat exposure across different local climate zones derived from cross-city comparisons. Ultimately, this contributes to the
by integrating ECOSTRESS LST with static population data in Xi’an, sustainable and resilient development of the relationship between resi­
China. Yasumoto et al. (2019) probed into individual daily heat expo­ dents and their environment.
sure in Dhaka, Bangladesh, leveraging MODIS LST data alongside travel
questionnaire data, yet this approach failed to precisely mirror people’s 2. Study areas and data
movement characteristics. Similarly, Cao et al. (2020) investigated heat
exposure risk in Guangzhou, China, utilizing meteorological station data 2.1. Study areas
as well as hourly social media location data, thereby accomplishing 24-h
temporal monitoring of heat exposure, but it still fell short in capturing Beijing, Shanghai, and Xiamen were selected as study areas (Fig. 1)
individual mobility-based heat health risks. In terms of research on to capture diverse climatic conditions and urban characteristics. Beijing,
environmental health risk assessment concentrating on active travel and in northern China, has a temperate continental climate. Shanghai, on the
mobility, diverse methodologies have been employed to collect travel eastern coast, experiences a subtropical monsoon climate. Xiamen, in
trajectories and estimate personal exposure. Initially, predefined travel the southeast, features a subtropical maritime climate. These cities are
routes enable the measurement of exposure values along designated prominent in heat environment research within China (Guo et al., 2019;
paths (Okokon et al., 2017; Zhou & Lin, 2019). For example, Kim and Li et al., 2011; Tan et al., 2010) due to their high population densities
Brown (2021) selected specific paths within New York’s High Line Park and varying heat exposure patterns. Additionally, the availability of
to examine pedestrian thermal comfort conditions and factors associated bike-sharing data for these cities provided a robust foundation for
with walking behaviors. Nevertheless, this predefined path method is analyzing cyclist mobility and heat health risks.
constrained by small sampling sizes and predetermined trajectory
coverage areas, and the preset paths might deviate from actual routes
(Okokon et al., 2017), thus leading to estimates that diverge from re­ 2.2. Data
ality. Subsequently, another approach involves extracting origins and
destinations from user travel surveys and employing path simulation to In this study, we employed bike-sharing datasets, ECOSTRESS LST
identify potential travel routes (Cao et al., 2021; Jiang et al., 2022). For datasets, and road network datasets. The bike-sharing trajectories data
example, Vallamsundar et al. (2022) utilized Dijkstra’s pathfinding were provided by the local operators from July to August 2018
method to determine the shortest travel routes between OD pairs and recording the riding trips in the three study areas during the morning
further measured air pollution exposure during the journey. Qian and peak and night peak on weekdays. It recorded the information of trip ID,
Wu (2019) drew on Google Maps to ascertain possible route choices for time stamp, origin coordinate, and destination coordinate. To analyze
bike-sharing OD pairs and gauged PM2.5 exposure along the travel the cyclists’ heat health risks accurately, the acquisition time of bike-
trajectories. sharing datasets and ECOSTRESS LST datasets were consistent.
Overall, research on thermal environmental health risks and expo­ The diurnal dynamic temperature data, ECOSTRESS Level-2 Land
sure assessment has evolved from being based on static population Surface Temperature and Emissivity product were obtained from the
distribution to focusing on population dynamic mobility (Liu et al., NASA-LPDAAC ([Link] The product provides sur­
2024; Yang et al., 2025). Despite the fact that prior work concentrating face temperature and emissivity retrieved from five thermal infrared
on active travel-related environmental health risk assessment has made bands at a spatial resolution of 70 m × 70 m using a physics-based
outstanding contributions in terms of data sources and improved accu­ temperature emissivity separation algorithm (Hulley, 2019; Hulley
racy, there are still limitations to be addressed. First, regarding the issue et al., 2019). Relevant studies have shown that ECOSTRESS products
of heat health risk exposure in active travel behaviors, introducing have a very high consistency with the surface temperature data gener­
large-scale GPS trajectory data serves as an efficient and low-cost ated by existing thermal infrared instruments, with RMSE of 1.07K, MAE
method. However, comprehensive comparisons across multiple cities of 0.40K, and R2 > 0.988 at 14 verification points globally (Hook et al.,
as study areas are still rare, with most literature focusing on single cities, 2020; Hulley et al., 2022; Silvestri et al., 2020). To explore the cyclists’
and the transferability of methods as well as the universality of findings heat health risks, we focused on the hottest month of the year in the
remain uncertain. Second, while a wealth of relevant literature has three study areas. Therefore, we obtained available ECOSTRESS LST
achieved precise spatial mapping of heat health risks for active travel imageries from July to August 2018 and selected six scenes with cloud
behaviors, including cycling, walking, and jogging, there has been rates lower than 5 % during the morning peak (8:00–10:00) and night
relatively limited exploration of the socioeconomic and built environ­ peak (17:00–19:00). The selected imageries were acquired at different
ment drivers underlying the spatial distribution of these thermal envi­ times of the day: 9:54 and 18:10 in Beijing; 10:01 and 17:54 in Shanghai;
ronmental risks. Such understanding is crucial for identifying the root 10:04 and 17:43 in Xiamen.
causes of risk distribution and devising mitigation strategies following In terms of influencing factors, the analysis included OpenStreetMap
the identification of risk patterns. road data, POI data, building data, and NDVI data (Table 1). Road data
To sum up, this study aims to address the following research ques­ were sourced from OpenStreetMap (2018), excluding non-cyclable
tions: How can we dynamically identify the spatiotemporal character­ routes like highways and tunnels. POI data, obtained via Baidu Maps
istics of heat environmental health risks associated with active travel API, covered categories such as corporate entities, residential areas, and
behaviors? How can we further pinpoint the spatial layout of high-risk public services. Building data from Baidu Maps provided vector
areas and their underlying socioeconomic and built environment boundaries and floor counts. NDVI data were acquired from the National
drivers? Are these findings consistent across different cities? In order to Earth System Science Data Center of China, with a 30-m spatial reso­
tackle these questions, this study combines ECOSTRESS LST data with lution. All datasets correspond to 2018, aligning with the bicycle data.

2
K. Zeng et al. Applied Geography 184 (2025) 103764

Fig. 1. Study areas.

conceptually anchored in the distinction between general heat exposure

Table 1
and excessive heat exposure. CycHeat captures the cumulative effect of
Summary statistics for all variables.
ambient temperature on cyclists over their journey, reflecting the
Data Source Format Time overall thermal burden. In contrast, CycExcessHeat specifically targets
Road OpenStreetMap Vector 2018 exposures above a critical temperature threshold, highlighting the pe­
network riods and routes where cyclists face significantly elevated health risks.
POI Baidu Maps API Vector 2018
This conceptual distinction allows for a more nuanced understanding of
Buildings Baidu Maps API Vector 2018
NDVI National Earth System Science Data Center Raster 2018
how different thermal conditions impact cyclist well-being.
of China (30m) Finally, the framework extends to explore the association between
identified heat health risks and the urban built environment. This part of
the framework is theoretically motivated by the need to uncover the
3. Methods underlying drivers of heat risk distribution in cities. By employing a
suite of regression models, including OLS, SLM, and SEM, the framework
3.1. Research framework examines how various built environment features, such as road density,
land use patterns, and vegetation cover (NDVI), influence the spatial
This study develops a comprehensive framework (Fig. 2) to investi­ distribution of heat health risks. This analytical step is designed to reveal
gate and address the heat health risks faced by cyclists in urban envi­ the complex interplay between urban design and thermal exposure, of­
ronments, integrating dynamic cyclist mobility patterns with high- fering valuable insights for urban planning and infrastructure
resolution remote sensing data. Conceptually, the framework is built optimization.
upon the integration of cyclist behavior and environmental exposure, Overall, the framework represents a systematic and conceptually
aiming to bridge the gap between human mobility and environmental integrated approach to addressing the heat health risks of cyclists. It
health risk assessment (see Fig. 3). moves beyond mere data processing and analysis steps to establish a
The framework initiates with the reconstruction of detailed cycling clear conceptual pathway from data collection to risk assessment and
trajectories through the fusion of bike-sharing Origin-Destination (O-D) finally to planning implications.
big data and road network information, capturing the nuanced spatio­
temporal patterns of urban cycling activity. This step is theoretically
grounded in the understanding that individual mobility patterns are 3.2. Preprocess of shared bike trajectories
crucial for accurate exposure assessment, moving beyond static popu­
lation assumptions. Subsequently, the framework incorporates the pre­ We reconstructed bike trajectories from origin-destination co­
processing of ECOSTRESS LST data, which involves rigorous time ordinates using an online map API (see Fig. 3). First, we cleaned data by
consistency checks and advanced cloud removal techniques. This en­ removing trips over 2 h, over 10 km or under 100 m, or with speeds over
sures that the land surface temperature data aligns precisely with the 30 km/h (Cao et al., 2021; Gao, Li, Tan, Wu, et al., 2021). Next, we used
temporal resolution of cycling activities, providing a robust environ­ a Python script to call Baidu Map’s cycling route service with the co­
mental data foundation for subsequent risk assessment. ordinates. The returned coordinate chains were then converted from
At the core of the framework is the quantification of cyclists’ heat Baidu to WGS84 coordinates using ArcPy and matched to the road
health risks through two novel indicators: CycHeat and CycExcessHeat. network. Finally, bike volumes on each road segment were measured
These indicators are not merely computational tools but are using ArcGIS Pro’s spatial join tool.

3
K. Zeng et al. Applied Geography 184 (2025) 103764

Fig. 2. The research framework of this study.

3.3. Preprocess of ECOSTRESS LST data health risks (CycHeat) for each travel can be assessed as,
∫ t1
The ECOSTRESS LST data provided by NASA-LPDAAC has not been CycHeat = Tt (lont , latt , t)dt (1)
preprocessed. Therefore, we removed clouds and corrected the geo­ t0

location of ECOSTRESS LST data with the corresponding L2_CLOUD and

where CycHeat is the accumulated heat health risks of each cyclist riding
ECOSTRESS L1B_GEO, which are the cloud mask data and the geo­
bike sharing; t0 and t1 is the start and end timestamp of a trajectory,
location files for the products. Finally, the six scenes were extracted by
respectively; Tt is the value of ECOSTRESS LST at the time t and location
the corresponding study area boundaries.
(lont , latt ). Considering the spatial resolution of the ECOSTRESS data,
this study accessed the Tt every 70 m by splitting each trajectory into
3.4. Cyclists’ heat health risks estimation multiple 70 m-segments and overlaid with the Tt map. Therefore, the
accumulated cyclists’ heat health risks (◦ C⋅min) would be,
As one of the highest-resolution LST datasets, ECOSTRESS LST data ∑n
can obtain surface temperature distribution at multiple times CycHeat = i=0
di T i (2)
throughout the day with a spatial resolution of 70 m (Chang et al., 2021;
Wang et al., 2023). Based on the spatial and temporal distribution of where n is the amount of 70 m-segments of each trajectory; di is the time
bike sharing trajectories and the LST data, the accumulated cyclists’ heat spent by a cyclist in the ith segment; Ti is the mean value of ECOSTRESS

4
K. Zeng et al. Applied Geography 184 (2025) 103764

Fig. 3. The reconstruction of bike-sharing trajectories.

LST along the ith segment. The above two formulas aim to quantify the active travel behavior and heat health risks (Huang et al., 2023), the
accumulated heat health risks of cyclists with high-resolution LST data. following variables were selected (Table 2): from the road infrastructure
In addition, this study further estimated the extreme heat health risks perspective, main road density, secondary road density, branch road
of cyclists by focusing on the surface temperature above 30 ◦ C. The density, and intersection density (Gao et al., 2024; Li et al., 2020); from
excessive high temperature was extracted from the Tt map, then the bike the public transportation perspective, the number of bus stops and metro
trajectories within the extreme temperature areas were selected to stations (Chen et al., 2022); and from the destination accessibility
perform the excessive heat health risks assessment. The accumulated perspective, floor area ratio, the number of companies, the number of
cyclists’ excessive heat health risks (CycExcessHeat) can be expressed by residential communities, the number of recreational facilities (including
the following formulas: shopping, dining, and cultural entertainment), and the number of public
∫ t1 service facilities (Chen et al., 2022). These were incorporated as inde­
CycExcessHeat = ETt (lont , latt , t)dt (3) pendent variables in the analysis. Furthermore, the NDVI was selected as
t0
an independent variable to quantify green space coverage, which has
∑n been documented in the literature to exhibit a significant correlation
CycHExcesseat = i=0
di ETt (4) with cycling behavior (Gao, Li, Tan, Zhang, et al., 2021; Wang et al.,
2020).
where CycExcessHeat is the accumulated cyclists’ excessive heat health
risks; t0 and t1 is the start and end timestamp of a trajectory, respec­
tively; ETt refers to the value above 30 ◦ C at the time t and location (lont , 3.6. Regression analysis
latt ). For instance, if the mean LST of the ith segment is 35 ◦ C, then the
excessively high temperature is 5 ◦ C. The above two formulas help to Before statistical analysis, multicollinearity was assessed using the
quantify the accumulated excessive heat health risks of [Link] is variance inflation factor (VIF), with all variables showing VIF values of 4
important to note that the efficiency of cyclists isn’t particularly high, or lower, except for intersection density, which was omitted. An ordi­
with most cyclists operating at an efficiency of around 20 %, with the nary least squares (OLS) regression was initially used to examine the
majority of energy being converted into heat (Andy, 2025). Given this, it relationship between the built environment and CycExcessHeat.
is evident that the extreme heat thresholds used for assessing the heat
risks of cycling commuters shouldn’t directly adopt general thresholds, Table 2
such as the meteorological definition of high temperature being 35 ◦ C. Summary statistics for all variables.
Due to the high intensity of metabolic heat production during cycling, Variables Beijing Shanghai Xiamen
this study selects 30 ◦ C as the threshold. This allows for a more sensitive
Mean SD Mean SD Mean SD
capture of the initial turning point of heat stress in cyclists compared to
other general thresholds that apply to the human body under normal Main road density 3.011 3.254 3.251 3.301 2.561 2.579
(km/km2)
conditions.
Secondary density 6.879 10.014 8.014 9.167 4.789 5.019
Furthermore, the heat health risks differences between the morning (km/km2)
peak and evening peak of the three study areas were examined. The non- Branch density 8.912 9.147 10.047 11.002 6.014 7.812
parametric Kruskal-Wallis one-way analysis, which can compare mul­ (km/km2)
Intersection 26.124 34.781 21.014 25.193 16.416 20.291
tiple independent groups, was adopted in this study to explore the
density (/km2)
differences. Number of bus 4.561 4.715 5.014 7.913 5.114 6.921
stop (N)
Number of metro 0.510 0.621 0.461 0.502 0.263 0.344
3.5. Built environment factors station (N)
Floor area ratio 0.810 0.647 0.714 0.522 0.621 0.418
To further investigate the factors associated with cyclists’ heat health Company POI (N) 16.877 22.813 22.014 30.147 15.697 24.011
Residential POI 3.124 2.787 2.614 2.547 3.567 3.601
risks, 11 potential built environment determinants were selected for the
(N)
three study areas. Firstly, since CycExcessHeat will subsequently serve Entertainment POI 4.108 7.461 6.872 10.295 5.014 5.415
as the dependent variable in a series of regression analyses, this metric is (N)
derived from two critical components: LST and segment-level cycling Public service POI 6.714 7.241 10.011 16.782 7.047 9.963
volume. Consequently, the independent variables were deliberately (N)
NDVI 0.512 0.507 0.623 0.714 0.615 0.603
chosen to exclude these two aspects. Drawing on literature related to

5
K. Zeng et al. Applied Geography 184 (2025) 103764

Table 3
Results of SLM models.
Variables Beijing Shanghai Xiamen

Coef. (SE) p-Value Coef. (SE) p-Value Coef. (SE) p-Value

Main road density 0.715 (0.054) 0.011a 0.612 (0.104) 0.005b 0.354 (0.050) 0.012a
Secondary density 1.023 (0.005) 0.005b 0.978 (0.002) 0.003b 1.218 (0.005) 0.003b
Branch density 2.518 (0.451) <0.001c 4.514 (1.012) <0.001c 4.014 (1.125) <0.001c
Number of bus stop 0.005 (0.004) 0.120 0.002 (0.002) 0.125 0.001 (0.000) 0.474
Number of metro 0.058 (0.011) <0.001c 0.068 (0.014) <0.001c 0.060 (0.000) <0.001c
Floor area ratio 0.021 (0.007) 0.008b 1.012 (0.051) 0.100 0.054 (0.012) 0.014a
Company POI 1.821 (0.004) <0.001c 1.514 (0.019) <0.001c 2.001 (0.005) 0.005b
Residential POI 2.529 (0.214) 0.003b 1.860 (0.013) 0.003b 2.208 (0.006) <0.001c
Entertainment POI 1.002 (0.008) 0.011a 1.345 (0.010) 0.005b 1.787 (0.007) 0.012a
Public service POI 0.005 (0.005) 0.271 0.011 (0.009) 0.305 0.009 (0.002) 0.269
NDVI − 3.014(0.000) <0.001c − 2.143(0.000) <0.001c − 1.015(0.000) <0.001c
R2 0.521 ​ 0.651 ​ 0.567 ​
LL − 1869 ​ − 1969 ​ − 1751 ​
AIC 3841 ​ 3148 ​ 2973 ​

Note.
Coef. = Coefficient; SE = Standard error; LL = Log-likelihood; AIC = Akaike information criterion.
a
p < 0.05.
b
p < 0.01.
c
p < 0.001.

However, OLS results may be biased due to spatial dependence in spatial (Fig. 7).
data. To address this, spatial lag model (SLM) and spatial error model
(SEM) were employed (Anselin & Rey, 1991). The SLM assumes spatial 4.2. Diurnal temperature characteristic during rushing hours
dependence arises from autocorrelation in the dependent variable, while
the SEM attributes it to autocorrelation in the error term (Anselin & Rey, Fig. 8 shows the spatial distribution of ECOSTRESS LST during
1991). The SLM was found to be more suitable than the SEM based on morning and evening rush hours in three cities. In the morning, tem­
Lagrange Multiplier (LM) diagnostics. Significant spatial dependence peratures exceeded 30 ◦ C in most areas of Beijing and Shanghai, except
was identified in CycExcessHeat (Moran’s I > 0.5; p < 0.001). The SLM for southern Xiamen Island. High temperatures in Beijing were more
effectively mitigates spatial effects, and consistency between OLS and contiguous, likely due to its inland location. In the evening, high tem­
SLM results confirms the robustness of the identified relationships. peratures were contiguous on Xiamen Island but dispersed in Beijing and
Shanghai. Fig. 9 reveals that morning peak temperatures were higher
4. Results than evening peaks: average morning temperatures were 34.28 ◦ C in
Beijing, 33.42 ◦ C in Shanghai, and 30.98 ◦ C in Xiamen; evening averages
4.1. Shared bike usage characteristic during rushing hours were 28.85 ◦ C, 30.8 ◦ C, and 29.69 ◦ C, respectively.

Fig. 4 shows the spatial distribution of bike-sharing usage during the

4.3. Heat health risks of cyclists during rushing hours
morning peak and evening peak on the workdays in the central areas of
Beijing, Shanghai, and Xiamen. From the spatial distribution of bike
Fig. 10A presents the box plot of CycHeat in three cities, with mean
flows on road segments, bike-sharing usage in different periods and
values ranging from 170 (◦ C⋅min) to 330 (◦ C⋅min). Shanghai had a mean
different cities was spatially heterogeneous. In terms of usage, Beijing
CycHeat of 253.68 (◦ C⋅min) in the morning peak and 326.22 (◦ C⋅min) in
had the highest bike-sharing usage (9957 in the morning peak and 8662
the evening peak, while Beijing and Xiamen showed similar values in
in the evening peak), followed by Shanghai (9340 in the morning peak
both peaks: 185.44 (◦ C⋅min) and 178.36 (◦ C⋅min) in Beijing, and 192.1
and 8959 in the evening peak) and Xiamen (7055 in the morning peak
(◦ C⋅min) and 191.84 (◦ C⋅min) in Xiamen. Fig. 10B shows CycExcessHeat,
and 5373 in the evening peak). The bike-sharing usage in the morning
with means ranging from 2 (◦ C⋅min) to 28 (◦ C⋅min). Greater differences
peak was higher than that in the evening peak in all three cities (Fig. 5).
were observed between morning and evening peaks: 23.74 (◦ C⋅min) and
In terms of spatial distribution, it can be seen from Figs. 4 and 5 that
1.99 (◦ C⋅min) in Beijing, 27.52 (◦ C⋅min) and 9.45 (◦ C⋅min) in Shanghai,
the bike-sharing usage at the road section scale in Shanghai has a sig­
and 9.89 (◦ C⋅min) and 4.31 (◦ C⋅min) in Xiamen. The Kruskal-Wallis test
nificant spatial agglomeration characteristic with the higher global
revealed no significant difference in CycHeat between morning and
Moran’s I of 0.68 and 0.61 in the morning and evening peak, respec­
evening peaks in all cities. However, CycExcessHeat was significantly
tively, compared with Beijing and Xiamen.
higher in the morning peak (p < 0.05), indicating that this indicator
In terms of usage characteristics, users in Shanghai had the longest
better reflects temporal variations in cyclists’ heat health risks.
riding duration (12.45 min in the morning peak and 18.5 min in the
The CycHeat and CycExcessHeat of each trajectory on the same seg­
evening peak), compared with those in Beijing (8.54 min in the morning
ments were accumulated by spatial join. Figs. 11 and 12 present the
peak and 9.31 min in the evening peak) and Xiamen (8.8 min in the
maps of CycHeat and CycExcessHeat. The biggest difference between the
morning peak and 9.26 min in the evening peak). It is worth noting that
two indicators is that the spatial distribution of CycHeat is consistent
the morning peak riding duration was shorter than the evening peak in
with the numerical statistics (Fig. 10A), that is, the pattern of morning
three cities (Fig. 6), which is consistent with relevant literature that the
peak and evening peak is similar. In contrast, the CycExcessHeat focuses
morning peak period of working days is relatively short and concen­
on extreme heat (over 30 ◦ C), reflecting a more significant difference
trated, and most cyclists are on their way to work; However, the evening
between morning and evening peaks (Fig. 12).
peak period lasts longer, and most cyclists are usually less rushed to
In short, the most potential practical significance of the CycExces­
cycle home after work or to eat or other activities. Correspondingly, the
sHeat lies in the rapid and accurate identification of the most severe heat
riding distance of cyclists in Shanghai was longer than that in Beijing
health risks cycling sections. For example, we show the representative
and Xiamen in both morning peak and evening peak on weekdays
road sections in three cities in Fig. 13: Middle Fuxing Road, Shanghai,

6
K. Zeng et al. Applied Geography 184 (2025) 103764

Fig. 4. Spatial distribution of bike-sharing usage in Beijing, Shanghai, and Xiamen.

Fig. 5. The bike-sharing usage and the global Moran’s I of its distribution during morning peak and evening peak.

7
K. Zeng et al. Applied Geography 184 (2025) 103764

Fig. 6. The frequency distribution of riding duration during morning peak and evening peak.

Fig. 7. The frequency distribution of riding distance during morning peak and evening peak.

8
K. Zeng et al. Applied Geography 184 (2025) 103764

Fig. 8. Spatial distribution of ECOSTRESS LST in the morning and evening peak period.

Fig. 9. The violin plots of temperature in the morning and evening peak periods.

9
K. Zeng et al. Applied Geography 184 (2025) 103764

Fig. 10. The box plot of CycHeat (A), and CycExcessHeat (B).

Fig. 11. Spatial distribution of the accumulated heat health risks of cyclists (CycHeat).

10
K. Zeng et al. Applied Geography 184 (2025) 103764

Fig. 12. Spatial distribution of the accumulated excessive heat health risks of cyclists (CycExcessHeat).

the segment with the highest CycExcessHeat during morning peak with heat health risks. This is likely attributable to the predominance of
(Fig. 13A); Changling Road, Xiamen, the segment with the highest cycling activity on lower-category roads, such as branch roads or sec­
CycExcessHeat during morning peak (Fig. 13B); West Chang ’an Avenue, ondary roads, within these urban areas. Regarding public transportation
Beijing, the segment with the highest CycExcessHeat during evening accessibility, the model results across the three cities revealed no sig­
peak (Fig. 13C). CycExcessHeat is based on the bike volume, tempera­ nificant association between the number of bus stops and heat health
ture, and the riding duration along the segment of the road. These three risks, whereas subway station density demonstrated a significant posi­
sections show exactly three different characteristics: Middle Fuxing tive correlation. In terms of destination accessibility, the distribution of
Road had a relatively longer riding duration; the bike volume along employment and residential locations exerted a more pronounced pos­
Changling Road was relatively greater; the temperature along West itive impact on heat health risks compared to other types of POIs, such as
Chang ’an Avenue was relatively higher than other parts of Beijing recreational or public service POIs. Furthermore, the models consis­
during the evening peak. tently identified a significant negative association between green space,
as measured by NDVI, and heat health risks. This suggests that areas
with elevated cycling-related heat health risks are often characterized by
4.4. Association between heat health risks and built environment insufficient green space coverage in their vicinity.

Table 3 delineates the relationships between the built environment

and average CycExcessHeat. Given that the SLM outperformed the OLS
model in terms of goodness-of-fit metrics (R2, log-likelihood, and AIC),
the analysis primarily focuses on the SLM results. In the context of road
infrastructure, road hierarchy appears to exhibit a negative correlation

11
K. Zeng et al. Applied Geography 184 (2025) 103764

Fig. 13. Three examples at Middle Fuxing Road, Shanghai, the segment with the highest CycExcessHeat during morning peak (A); Changling Road, Xiamen, the
segment with the highest CycExcessHeat during morning peak (B); West Chang ’an Avenue, Beijing, the segment with the highest CycExcessHeat during evening peak
(C). (Their locations can be found in the bright green circles of Fig. 13). (For interpretation of the references to color in this figure legend, the reader is referred to the
Web version of this article.)

5. Discussions higher than evening peak temperatures. Second, while the cumulative
heat health risk indicator (CycHeat) showed no significant difference
5.1. Major findings compared with related works between morning and evening peaks via Kruskal-Wallis test, the excess
cumulative heat health risk indicator (CycExcessHeat) revealed signifi­
The field of heat exposure and heat health risk assessment has pre­ cantly higher risks during morning peaks than evening peaks. These
dominantly focused on risk evaluations based on population distribution findings suggest that morning peak cycling behavior is largely commute-
at specific time points, such as day and night or hourly intervals (Cao driven, where efficiency considerations outweigh other factors like
et al., 2020; Yasumoto et al., 2019; Yuan et al., 2022). These studies greenery and environment. This aligns with active travel literature
often rely on static, cross-sectional population data for single or multiple indicating higher cycling frequency during morning peaks due to the
time snapshots, resulting in static risk assessments. Such an approach shorter, more urgent time window for getting to work (Faghih-Imani &
has limitations for urban planning and practical applications since urban Eluru, 2016; Meng et al., 2023; Shen et al., 2018; Xu et al., 2019). In
populations are highly mobile. Focusing solely on population distribu­ contrast, evening peaks offer more flexibility, leading to relatively lower
tion during working or living periods fails to capture the dynamic heat health risks. The consistent results across the three cities imply that
interplay between population mobility and heat health risks (Li & Wang, this phenomenon may be common in large cities.
2021).
Compared to similar studies on environmental health risks associ­
5.2. Urban planning implications and nature-based solutions
ated with active travel (Li & Wang, 2021; Liu et al., 2024; Yang et al.,
2025; Zhou & Lin, 2019), this research extends previous work in both
This study offers significant insights for urban planning and man­
depth and breadth. In terms of research depth, we have delved deeper
agement, particularly in mitigating heat health risks through nature-
into the drivers behind the spatial distribution of heat health risks. By
based solutions (NbS). Our findings highlight the crucial role of green
employing OLS-SLM-SEM regression models, we quantified how built
spaces in reducing heat exposure for cyclists, as evidenced by the
environment variables influence the layout of heat health risks, using
negative correlation between NDVI and heat health risks. This un­
cycling commuting heat health risks as the response variable. The model
derscores the importance of integrating NbS into urban design to
results revealed an interesting pattern: that is, the modeling results with
enhance thermal comfort and promote sustainable urban development.
cycling heat risk as the dependent variable are generally consistent with
By incorporating more green infrastructure, such as shading trees
the existing literature that takes cycling frequency as the dependent
and permeable pavements, into high-risk areas identified by CycExces­
variable (Gao & Fang, 2025a, 2025b; Zhuang et al., 2022). The results
sHeat, urban planners can create cooler and more comfortable envi­
show a positive correlation between heat health risks and factors such as
ronments for cyclists and pedestrians alike (Fig. 14). These solutions not
lower-category roads, employment and residential locations, and prox­
only improve thermal conditions but also contribute to biodiversity
imity to subway stations, which is consistent with most studies on
conservation and better air quality. For instance, increasing tree
cycling frequency modeling (Gao & Fang, 2025c; Li et al., 2020; Wang
coverage along cycling routes can provide shade and lower surface
et al., 2020; Xu et al., 2023). However, green space coverage exhibited a
temperatures, directly reducing heat exposure. Additionally, green
negative correlation with heat health risks in this study, differing from
spaces can act as natural heat sinks, absorbing and dissipating heat more
the general findings in cycling frequency modeling literature. This
effectively than urban surfaces.
discrepancy may be because while many cyclists choose routes with
Our results also suggest that urban planning strategies should bal­
more green space for sun protection or aesthetic reasons, others priori­
ance connectivity and thermal safety. While branch roads with high
tize commuting destinations, making their route choices less influenced
connectivity attract more riders, they often lack shade and can exacer­
by green space availability.
bate heat retention. Prioritizing the introduction of shading facilities in
In terms of research breadth, this study broadens the scope by con­
branch road updates can help address this issue. Furthermore, the
ducting a comparative analysis across three major Chinese cities. Unlike
implementation of colored bike lanes made from materials with high
previous studies limited to single cities due to data constraints, we ob­
reflectivity and thermal conductivity can provide better thermal comfort
tained peak-hour cycling commuting data from the core areas of three
compared to conventional materials. Bike-sharing operators can also
representative large Chinese cities. Despite differences in latitude, all
contribute by replacing heat-absorbing materials on bicycles with
three cities experience high summer temperatures. Consistent patterns
reflective and breathable alternatives. Integrating real-time temperature
were found across the cities: first, morning peak temperatures were
data into navigation apps can empower cyclists to make informed

12
K. Zeng et al. Applied Geography 184 (2025) 103764

Fig. 14. Management and urban planning strategies based on heat health risks assessment.

decisions about their routes and help mitigate the negative impacts of identified in our study are representative for a substantial majority of
heat exposure. short to medium distance commuters in the studied city cores. The
In summary, this research provides a foundation for developing remaining commuters using other active or motorized modes would
comprehensive urban planning strategies that combine NbS with data- likely experience different thermal profiles; however, the pronounced
driven insights to enhance urban resilience and sustainability. By concentration of cycling trips within the central districts implies that our
focusing on both immediate and long-term solutions, cities can effec­ findings meaningfully reflect the heat health risks encountered by a
tively address the challenges posed by urban heat islands and extreme dominant segment of urban commuters during peak hours.
heat events.
5.4. Limitations and prospects
5.3. Scope and representativeness of the sample
This study integrates extensive remote sensing data (from ECO­
STRESS) and cycling trajectories big data to analyze cycling-related
Although our data do not provide the exact share of shared-bicycle
health risks, differing from traditional heat exposure studies focused
commuters among all rush-hour travelers, authoritative statistics indi­
on thermal comfort (Yang, Peng, Liu, et al., 2024). Traditional studies
cate that the permitted parking zones of bike-sharing fleets in Beijing,
often construct detailed thermal comfort models by considering multiple
Shanghai and Xiamen effectively serve 61.4 %, 61.9 % and 57 % of the
factors like radiation temperature and vegetation cover. They guide
commuting population within 100 m buffers (Caupd et al., 2024). This
personal decisions and micro-environment designs but require
high service coverage (Fig. 15) suggests that the heat exposure patterns
high-precision meteorological data, limiting their application. Our
method, using remote sensing data for LST and trajectories big data,
offers a macro-level analysis of health risk patterns. Though not
capturing individual thermal sensations, it reveals overall heat risk
distributions, aiding urban and health policy planning. Unlike tradi­
tional methods with high data requirements, our approach with
open-source remote sensing and crowdsourced trajectory data ensures
easier data access. The computational framework is simple yet logical,
ensuring stronger operability and replicability. Our study’s macro-level
approach, while revealing overall heat risk patterns, has limitations,
notably the exclusion of micro-environmental factors like tree shade and
building shadows that affect cyclists’ actual exposure. LST, a surface
metric, might not accurately reflect cyclists’ thermal experiences,
especially in shaded areas. Future research could benefit from inte­
grating our analysis with detailed environmental data, such as shading
from Baidu Map, to provide a more nuanced assessment of heat health
risks for cyclists.
Due to data availability, the diurnal dynamic temperature data
during the morning peak and evening peak used in this study were not
acquired on the same day, which is also the limitation of relevant studies
with ECOSTRESS data (Chang et al., 2021, 2023; Huang et al., 2024; Wei
Fig. 15. The proportion of the commuting population served within a 100-m & Sobrino, 2024). Future work should consider incorporating other data
radius of the permitted parking spots for shared bikes. sources, such as drone remote sensing and Sentinel satellite data to

13
K. Zeng et al. Applied Geography 184 (2025) 103764

improve the data consistency. In addition, although the bike-sharing References

trajectory big data used in this study has valuable advantages in terms
of temporal and spatial accuracy, user attribute data such as gender, age, Andy, T. (2025). Heat adaptation: Why training in hot conditions, especially indoors, can
boost your fitness [WWW Document] Cyclingnews. URL [Link]
socioeconomic status, and other information should be further intro­ com/features/heat-adaptation-why-training-in-hot-conditions-especially-indoors-
duced in the future to carry out heat health risks assessment for different can-boost-your-fitness/.
social groups. Anselin, L., & Rey, S. (1991). Properties of tests for spatial dependence in linear
regression models. Geographical Analysis, 23, 112–131. [Link]
j.1538-4632.1991.tb00228.x
6. Conclusions Baqa, M. F., Lu, L., Guo, H., Song, X., Alavipanah, S. K., Nawaz-ul-Huda, S., Li, Q., &
Chen, F. (2025). Investigating heat-related health risks related to local climate zones
using SDGSAT-1 high-resolution thermal infrared imagery in an arid megacity.
This study proposed a novel framework to estimate individual heat International Journal of Applied Earth Observation and Geoinformation, 136, Article
health risks of cyclists based on bike-sharing trajectory big data and the 104334. [Link]
diurnal dynamic temperature data. The developed framework provides Biardeau, L. T., Davis, L. W., Gertler, P., & Wolfram, C. (2019). Heat exposure and global
air conditioning. Nature Sustainability, 3, 25–28. [Link]
new insights into understanding the heat health risks of human outdoor
019-0441-9
mobility, which helps policymakers and urban planners to understand Borden, K. A., & Cutter, S. L. (2008). Spatial patterns of natural hazards mortality in the
individual heat health risks in daily commuting and urban sustainable United States. International Journal of Health Geographics, 7, 64. [Link]
development. This study proposed two indicators, accumulated heat 10.1186/1476-072X-7-64
Cao, Z., Gao, F., Li, S., Wu, Z., Guan, W., & Ho, H. C. (2021). Ridership exceedance
health risks (CycHeat) and accumulated excessive heat health risks exposure risk: Novel indicators to assess PM2.5 health exposure of bike sharing
(CycExcessHeat), to estimate the heat health risks of bike-sharing cyclists riders. Environmental Research, 197, Article 111020. [Link]
in Beijing, Shanghai, and Xiamen based on massive GPS trajectories and envres.2021.111020
Cao, Z., Wu, Z., Li, S., Guo, G., Song, S., Deng, Y., Ma, W., Sun, H., & Guan, W. (2020).
diurnal dynamic temperature data. Results show that there is no sig­ Explicit spatializing heat-exposure risk and local associated factors by coupling
nificant difference in terms of the CycHeat between the morning peak social media data and automatic meteorological station data. Environmental Research,
and evening peak in both three cities. While the CycExcessHeat in the 188, Article 109813. [Link]
Caupd, Mohurd, & DiDi. (2024). Annual report on sharing bikes and sharing electric bikes
morning peak is significantly higher than that in the evening peak in riding in major Chinese cities.
Beijing, Shanghai, and Xiamen (p < 0.05). That is, the indicator that Chang, Y., Xiao, J., Li, X., Middel, A., Zhang, Y., Gu, Z., Wu, Y., & He, S. (2021).
focuses on extreme heat is more reflective of spatiotemporal heteroge­ Exploring diurnal thermal variations in urban local climate zones with ECOSTRESS
land surface temperature data. Remote Sensing of Environment, 263, Article 112544.
neity in cyclists’ heat health risks. The CycExcessHeat maps in this study [Link]
help identify sections of roads at risk of severe cycling heat health risks. Chang, Y., Xiao, J., Li, X., & Weng, Q. (2023). Monitoring diurnal dynamics of surface
Finally, the relationship between CycExcessHeat and the built environ­ urban heat island for urban agglomerations using ECOSTRESS land surface
temperature observations. Sustainable Cities and Society, 98, Article 104833. https://
ment was regressed, and the model results align with the literature
[Link]/10.1016/[Link].2023.104833
focusing on cycling frequency. This study provided a framework to es­ Chen, L., Lu, Y., Ye, Y., Xiao, Y., & Yang, L. (2022). Examining the association between
timate individual heat health risks of human outdoor mobility, which the built environment and pedestrian volume using street view images. Cities, 127,
helps policymakers and planners to understand individual heat health Article 103734. [Link]
Cheng, Q., & Sha, S. (2024). Resisting the heat wave: Revealing inequalities in matching
risks in daily commuting and urban sustainable development, and pro­ between heat exposure risk and healthcare services in a megacity. Applied Geography,
vides management implications for transportation authorities and bike- 167, Article 103291. [Link]
sharing operators. Faghih-Imani, A., & Eluru, N. (2016). Incorporating the impact of spatio-temporal
interactions on bicycle sharing system demand: A case study of New York CitiBike
system. Journal of Transport Geography, 10.
CRediT authorship contribution statement Foroutan, E., Hu, T., Zhang, F., & Yu, H. (2024). Assessing heat vulnerability in
Philadelphia using geographically weighted principal component analysis (GWPCA):
A geospatial big data-driven approach. International Journal of Applied Earth
Kun Zeng: Writing – review & editing, Writing – original draft, Observation and Geoinformation, 127, Article 103653. [Link]
Methodology, Investigation, Funding acquisition, Formal analysis, Data jag.2024.103653
curation, Conceptualization. Feng Gao: Visualization, Validation, Su­ Gao, F., Chen, X., Liao, S., Chen, W., Feng, L., Wu, J., Zhou, Q., Zheng, Y., Li, G., & Li, S.
(2024). Crafting a jogging-friendly city: Harnessing big data to evaluate the
pervision, Resources, Project administration, Funding acquisition, runnability of urban streets. Journal of Transport Geography, 121, Article 104015.
Formal analysis, Data curation, Conceptualization. Yihuan Peng: Su­ [Link]
pervision, Software, Methodology, Investigation, Funding acquisition, Gao, M., & Fang, C. (2025a). Pedaling through the cityscape: Unveiling the association of
urban environment and cycling volume through street view imagery analysis. Cities,
Data curation. Chang Liu: Validation, Supervision, Software, Resources,
156, Article 105573. [Link]
Project administration, Methodology. Wangyang Chen: Validation, Gao, M., & Fang, C. (2025b). Deciphering urban cycling: Analyzing the nonlinear impact
Supervision, Software, Resources, Project administration. Guanyao Li: of street environments on cycling volume using crowdsourced tracker data and
Visualization, Validation, Supervision, Project administration, machine learning. Journal of Transport Geography, 124, Article 104179. [Link]
org/10.1016/[Link].2025.104179
Methodology. Gao, M., & Fang, C. (2025c). Decoding the impact of audiovisual street environment
features on cycling volumes: Insights from street view imagery and machine
Declaration of competing interest learning. Transportation Research Part A: Policy and Practice, 199, Article 104586.
[Link]
Gao, F., Li, S., Tan, Z., Wu, Z., Zhang, X., Huang, G., & Huang, Z. (2021a). Understanding
The authors declare that they have no known competing financial the modifiable areal unit problem in dockless bike sharing usage and exploring the
interests or personal relationships that could have appeared to influence interactive effects of built environment factors. International Journal of Geographical
Information Science, 35, 1905–1925. [Link]
the work reported in this paper. 13658816.2020.1863410
Gao, F., Li, S., Tan, Z., Zhang, X., Lai, Z., & Tan, Z. (2021b). How is urban greenness
Acknowledgments spatially associated with dockless bike sharing usage on weekdays, weekends, and
holidays? ISPRS International Journal of Geo-Information, 10, 24. [Link]
10.3390/ijgi10040238
This work was supported by Collaborative Innovation Center for Gasparrini, A., Guo, Y., Hashizume, M., Lavigne, E., Zanobetti, A., Schwartz, J.,
Natural Resources Planning and Marine Technology of Guangzhou (No. Tobias, A., Tong, S., Rocklöv, J., Forsberg, B., Leone, M., De Sario, M., Bell, M. L.,
Guo, Y.-L. L., Wu, C., Kan, H., Yi, S.-M., De Sousa Zanotti Stagliorio Coelho, M.,
2023B04J0301), Key-Area Research and Development Program of
Saldiva, P. H. N., … Armstrong, B. (2015). Mortality risk attributable to high and low
Guangdong Province (NO. 2020B0101130009) and Guangdong Enter­ ambient temperature: A multicountry observational study. The Lancet, 386, 369–375.
prise Key Laboratory for Urban Sensing, Monitoring and Early Warning [Link]
(No. 2020B121202019). Gasparrini, A., Guo, Y., Sera, F., Vicedo-Cabrera, A. M., Huber, V., Tong, S., de Sousa
Zanotti Stagliorio Coelho, M., Nascimento Saldiva, P. H., Lavigne, E., Matus
Correa, P., Valdes Ortega, N., Kan, H., Osorio, S., Kyselý, J., Urban, A.,
Jaakkola, J. J. K., Ryti, N. R. I., Pascal, M., Goodman, P. G., … Armstrong, B. (2017).

14
K. Zeng et al. Applied Geography 184 (2025) 103764

Projections of temperature-related excess mortality under climate change scenarios. Orusa, T., Viani, A., Moyo, B., Cammareri, D., & Borgogno-Mondino, E. (2023). Risk
The Lancet Planetary Health, 1, e360–e367. [Link] assessment of rising temperatures using landsat 4–9 LST time series and meta®
(17)30156-0 population dataset: An application in Aosta Valley, NW Italy. Remote Sensing, 15,
Guo, X., Huang, G., Jia, P., & Wu, J. (2019). Estimating fine-scale heat vulnerability in 2348. [Link]
Beijing through two approaches: Spatial patterns, similarities, and divergence. Pearsall, H. (2017). Staying cool in the compact city: Vacant land and urban heating in
Remote Sensing, 11, 2358. [Link] Philadelphia, Pennsylvania. Applied Geography, 79, 84–92. [Link]
Ho, H. C., Knudby, A., Chi, G., Aminipouri, M., & Lai, D. Y.-F. (2018). Spatiotemporal [Link].2016.12.010
analysis of regional socio-economic vulnerability change associated with heat risks Qian, X., & Wu, Y. (2019). Assessment for health equity of PM2.5 exposure in bikeshare
in Canada. Applied Geography, 95, 61–70. [Link] systems: The case of Divvy in Chicago. Journal of Transport & Health, 14, Article
apgeog.2018.04.015 100596. [Link]
Ho, H. C., & Wong, M. S. (2019). Urban environmental influences on the Shen, Y., Zhang, X., & Zhao, J. (2018). Understanding the usage of dockless bike sharing
temperature–mortality relationship associated mental disorders and in Singapore. International Journal of Sustainable Transportation, 12, 686–700.
cardiorespiratory diseases during normal summer days in a subtropical city. [Link]
Environmental Science & Pollution Research, 26, 24272–24285. [Link] Silvestri, M., Romaniello, V., Hook, S., Musacchio, M., Teggi, S., & Buongiorno, M. F.
10.1007/s11356-019-05594-0 (2020). First comparisons of surface temperature estimations between ECOSTRESS,
Hook, S. J., Cawse-Nicholson, K., Barsi, J., Radocinski, R., Hulley, G. C., Johnson, W. R., ASTER and landsat 8 over Italian volcanic and geothermal areas. Remote Sensing, 12,
Rivera, G., & Markham, B. (2020). In-Flight validation of the ECOSTRESS, landsats 7 184. [Link]
and 8 thermal infrared spectral channels using the Lake Tahoe CA/NV and salton sea Tan, J., Zheng, Y., Tang, X., Guo, C., Li, L., Song, G., Zhen, X., Yuan, D., Kalkstein, A. J.,
CA automated validation sites. IEEE Transactions on Geoscience and Remote Sensing, Li, F., & Chen, H. (2010). The urban heat island and its impact on heat waves and
58, 1294–1302. [Link] human health in Shanghai. International Journal of Biometeorology, 54, 75–84.
Huang, X., Li, L., Yan, X., Ji, W., Zhao, K., & Zhao, X. (2024). Assessment of heat [Link]
exposure risk for urban populations and spatio-temporal patterns: A perspective of Tian, P., Zhang, F., Yan, Y., Liu, Y., Zhang, H., & Li, J. (2025). Spatial inequalities in
urban functional zones in Xi’an, China. Urban Climate, 55, Article 101992. https:// global population exposure to extreme heats and heatwaves. Applied Geography, 174,
[Link]/10.1016/[Link].2024.101992 Article 103474. [Link]
Huang, H., Ma, J., & Yang, Y. (2023). Spatial heterogeneity of driving factors for urban Vallamsundar, S., Jaikumar, R., & Venugopal, M. (2022). Exploring the spatial-temporal
heat health risk in Chongqing, China: A new identification method and proposal of dynamics of travel patterns and air pollution exposure of E-scooters. Journal of
planning response framework. Ecological Indicators, 153, Article 110449. [Link] Transport Geography, 105, Article 103477. [Link]
org/10.1016/[Link].2023.110449 jtrangeo.2022.103477
Hulley, G. (2019). ECOsystem Spaceborne Thermal Radiometer Experiment on Space Station Wang, R., Lu, Y., Wu, X., Liu, Y., & Yao, Y. (2020). Relationship between eye-level
(ECOSTRESS) mission: Level 2 product user guide. greenness and cycling frequency around metro stations in Shenzhen, China: A big
Hulley, G. C., Gottsche, F. M., Rivera, G., Hook, S. J., Freepartner, R. J., Martin, M. A., data approach. Sustainable Cities and Society, 59, Article 102201. [Link]
Cawse-Nicholson, K., & Johnson, W. R. (2022). Validation and quality assessment of 10.1016/[Link].2020.102201
the ECOSTRESS Level-2 land surface temperature and emissivity product. IEEE Wang, Q., Wang, X., Meng, Y., Zhou, Y., & Wang, H. (2023). Exploring the impact of
Transactions on Geoscience and Remote Sensing, 60, 1–23. [Link] urban features on the spatial variation of land surface temperature within the diurnal
TGRS.2021.3079879 cycle. Sustainable Cities and Society, 91, Article 104432. [Link]
Hulley, G., Shivers, S., Wetherley, E., & Cudd, R. (2019). New ECOSTRESS and MODIS scs.2023.104432
land surface temperature data reveal fine-scale heat vulnerability in cities: A case Wang, C., Zhang, Z., Zhou, M., Wang, P., Yin, P., Ye, W., & Zhang, L. (2018). Different
study for Los Angeles county, California. Remote Sensing, 11, 2136. [Link] response of human mortality to extreme temperatures (MoET) between rural and
10.3390/rs11182136 urban areas: A multi-scale study across China. Health & Place, 50, 119–129. https://
Jenerette, G. D., Harlan, S. L., Buyantuev, A., Stefanov, W. L., Declet-Barreto, J., [Link]/10.1016/[Link].2018.01.011
Ruddell, B. L., Myint, S. W., Kaplan, S., & Li, X. (2016). Micro-scale urban surface Watkins, L. E., Wright, M. K., Kurtz, L. C., Chakalian, P. M., Mallen, E. S., Harlan, S. L., &
temperatures are related to land-cover features and residential heat related health Hondula, D. M. (2021). Extreme heat vulnerability in Phoenix, Arizona: A
impacts in phoenix, AZ USA. Landscape Ecology, 31, 745–760. [Link] comparison of all-hazard and hazard-specific indices with household experiences.
10.1007/s10980-015-0284-3 Applied Geography, 131, Article 102430. [Link]
Jiang, Y., Song, G., Wu, Y., Zhai, Z., Zhang, Z., & Yu, L. (2022). The impact of cold-start apgeog.2021.102430
emissions on air pollution exposure during active travel. Transportation Research Part Wei, L., & Sobrino, J. A. (2024). Surface urban heat island analysis based on local climate
D: Transport and Environment, 112, Article 103469. [Link] zones using ECOSTRESS and Landsat data: A case study of Valencia city (Spain).
trd.2022.103469 International Journal of Applied Earth Observation and Geoinformation, 130, Article
Kim, Y. J., & Brown, R. D. (2021). A multilevel approach for assessing the effects of 103875. [Link]
microclimatic urban design on pedestrian thermal comfort: The high line in New Wu, X., Liu, Q., Huang, C., & Li, H. (2022). Mapping heat-health vulnerability based on
York. Building and Environment, 205, Article 108244. [Link] remote sensing: A case study in Karachi. Remote Sensing, 14, 1590. [Link]
buildenv.2021.108244 10.3390/rs14071590
Kuras, E. R., Richardson, M. B., Calkins, M. M., Ebi, K. L., Hess, J. J., Kintziger, K. W., Xu, Y., Chen, D., Zhang, X., & Tu, W. (2019). Unravel the landscape and pulses of cycling
Jagger, M. A., Middel, A., Scott, A. A., Spector, J. T., Uejio, C. K., Vanos, J. K., activities from a dockless bike-sharing system. Computers, Environment and Urban
Zaitchik, B. F., Gohlke, J. M., & Hondula, D. M. (2017). Opportunities and challenges Systems, 75, 21.
for personal heat exposure research. Environmental Health Perspectives, 125, Article Xu, X., Wang, J., Poslad, S., Rui, X., Zhang, G., & Fan, Y. (2023). Exploring intra-urban
085001. [Link] human mobility and daily activity patterns from the lens of dockless bike-sharing: A
Li, J., Song, C., Cao, L., Zhu, F., Meng, X., & Wu, J. (2011). Impacts of landscape structure case study of Beijing, China. International Journal of Applied Earth Observation and
on surface urban heat islands: A case study of Shanghai, China. Remote Sensing of Geoinformation, 122, Article 103442. [Link]
Environment, 115, 3249–3263. [Link] Yang, Z., & Peng, J. (2025). Efficiency-oriented phased urban green space planning
Li, X., & Wang, G. (2021). Examining runner’s outdoor heat exposure using urban framework to mitigate heat-stress exposure. npj Urban Sustain, 5, 57. [Link]
microclimate modeling and GPS trajectory mining. Computers, Environment and 10.1038/s42949-025-00247-3
Urban Systems, 89, Article 101678. [Link] Yang, Z., Peng, J., Jiang, S., Yu, X., & Hu, T. (2024a). Optimizing building spatial
compenvurbsys.2021.101678 morphology to alleviate human thermal stress. Sustainable Cities and Society, 106,
Li, W., Wang, S., Zhang, X., & Jia, Q. (2020). Understanding intra-urban human mobility Article 105386. [Link]
through an exploratory spatiotemporal analysis of bike-shar. International Journal of Yang, Z., Peng, J., Liu, Y., Jiang, S., Cheng, X., Liu, X., Dong, J., Hua, T., & Yu, X.
Geographical Information Science, 34, 2451–2474. [Link] (2024b). GloUTCI-M: A global monthly 1 km universal thermal climate index dataset
13658816.2020.1712401 from 2000 to 2022. Earth System Science Data, 16, 2407–2424. [Link]
Liu, Y., Zhang, G., & Yang, W. (2024). Dynamic estimation of urban heat exposure for 10.5194/essd-16-2407-2024
outdoor jogging: Combining individual trajectory and mean radiant temperature. Yang, W., Zhang, G., Liu, Y., & An, Z. (2025). Heat exposure assessment and comfort path
Urban Climate, 55, Article 101871. [Link] recommendations for leisure jogging based on street view imagery and GPS
Meng, F., Zheng, L., Ding, T., Wang, Z., Zhang, Y., & Li, W. (2023). Understanding trajectories. Sustainable Cities and Society, 119, Article 106099. [Link]
dockless bike-sharing spatiotemporal travel patterns: Evidence from ten cities in 10.1016/[Link].2024.106099
China. Computers, Environment and Urban Systems, 104, Article 102006. [Link] Yao, X., Jin, S., Zhao, Z., Sun, R., Wang, C., & Yu, Z. (2024). A novel integrated socio-
org/10.1016/[Link].2023.102006 ecological-economic index for assessing heat health risk. Ecological Indicators, 169,
National Weather Service. (2018). 78-year list of severe weather fatalities. Article 112840. [Link]
Noelke, C., McGovern, M., Corsi, D. J., Jimenez, M. P., Stern, A., Wing, I. S., & Yasumoto, S., Jones, A. P., Oyoshi, K., Kanasugi, H., Sekimoto, Y., Shibasaki, R.,
Berkman, L. (2016). Increasing ambient temperature reduces emotional well-being. Comber, A., & Watanabe, C. (2019). Heat exposure assessment based on individual
Environmental Research, 151, 124–129. [Link] daily mobility patterns in Dhaka, Bangladesh. Computers, Environment and Urban
envres.2016.06.045 Systems, 77, Article 101367. [Link]
Okokon, E. O., Yli-Tuomi, T., Turunen, A. W., Taimisto, P., Pennanen, A., Vouitsis, I., compenvurbsys.2019.101367
Samaras, Z., Voogt, M., Keuken, M., & Lanki, T. (2017). Particulates and noise Yuan, B., Zhou, L., Hu, F., & Zhang, Q. (2022). Diurnal dynamics of heat exposure in
exposure during bicycle, bus and car commuting: A study in three European cities. Xi’an: A perspective from local climate zone. Building and Environment, 222, Article
Environmental Research, 154, 181–189. [Link] 109400. [Link]
envres.2016.12.012 Zha, F., Lu, L., Wang, R., Zhang, S., Cao, S., Baqa, M. F., Li, Q., & Chen, F. (2024).
Understanding fine-scale heat health risks and the role of green infrastructure based

15
K. Zeng et al. Applied Geography 184 (2025) 103764

on remote sensing and socioeconomic data in the megacity of Beijing, China. Transportation Research Part D: Transport and Environment, 76, 305–322. [Link]
Ecological Indicators, 160, Article 111847. [Link] org/10.1016/[Link].2019.09.004
ecolind.2024.111847 Zhuang, C., Shaoying, L., Zhangzhi, T., Feng, G., & Zhifeng, W. (2022). Nonlinear and
Zhou, S., & Lin, R. (2019). Spatial-temporal heterogeneity of air pollution: The threshold effects of traffic condition and built environment on dockless bike sharing
relationship between built environment and on-road PM2.5 at micro scale. at street level. Journal of Transport Geography, 102, Article 103375. [Link]
10.1016/[Link].2022.103375

16