Personal assessment of urban heat exposure: a systematic review

The following multi-step criteria (figure 1) are established to select a representative study sample for review and evaluation. First, to obtain a workable sample size, the evaluations are limited to peer-reviewed studies that used sensing, modeling, or data analysis for assessing thermal comfort, physiological/heat strain, or heat-health impacts. In this framework, other environmental stressors as well as assessments that solely focus on the heat impact on urban energy, plant ecology, and pollutant emissions are excluded.

Zoom In Zoom Out Reset image size

Second, we assess the criteria for 'personal heat exposure'. Here, we define personal heat exposure studies as an evaluation of heat exposure either in the immediate environment of individuals and where people mostly reside in cities (extrinsic) or with the inclusion of physiological and/or behavioral and subjective responses in individuals (intrinsic) together with the environmental parameters. Accordingly, we excluded traditional publications that aimed to obtain a 'one-size-fits-all' analysis or prediction of heat exposure impact by analyzing a group of participants. Additionally, studies that only relied on supervised participants' surveys (unaccompanied with other sensing/modeling methods) are excluded based on the following factors: there has been little progress made in paper-based or supervised momentary feedback assessments of heat exposure, and due to not being scalable, they have limited ability to cover the time variability and range of places and conditions where people are or assess physiological and behavioral responses, which is core to our definition of personalized heat assessment. Lastly, the focus on urban and built environments is emphasized, excluding studies that rely on data from surrogate weather stations placed away from urban areas or studies that have limited relevance to urban and built environments (such as agricultural studies or intense fire/battlefield conditions). In this context, all aspects of indoor, outdoor, and transition spaces are considered to provide a comprehensive outlook on personalized heat exposure. Non-English language publications were also excluded.

2.2.1. Targeted outcomes and objectives in personal heat exposure assessments

The impact of heat exposure on human life is multifaceted and manifold, covering various aspects of physical and mental health as well as human comfort, productivity, and performance. Considering such a wide range of impacts, it is unsurprising that emerging methodologies and efforts to monitor personal heat exposure stem from distinct disciplines, spinning fields of environmental science and engineering, urban climate, building technology, public health, occupational health and sports science (figure 2). This diversity in the research perspectives has resulted in a wide range of desired and targeted outcomes in the literature sample. For instance, in the Building Technology discipline, increasing energy efficiency as well as human performance and productivity are considered as the main motivations for assessing personal heat exposure [34–36], while research in Environmental and Occupational Health has focused on health outcomes of heat and implication on occupational safety in the built environment [37, 38]. Accordingly, to provide an in-depth analysis of state-of-the-art, it is paramount that we distinguish the targeted outcomes in the literature sample. Additionally, even within each field of research, the objectives for monitoring personalized heat exposure vary greatly, responding to the limitations of conventional heat exposure studies that range from spatial and temporal misrepresentations to exclusion of physiological and psychological responses [8]. This results in a diverse set of objectives in personal heat exposure studies, such as overcoming technical challenges with sensor technologies and scalability [39, 40], obtaining real-time and momentary assessments [25, 41], or developing novel methodologies and frameworks for the integration of heat exposure assessment in people's life and workforce [34]. These multi-perspective approaches to personal heat exposure necessitate a more detailed classification of the literature sample based on the targeted outcomes and objectives posed (table 1). Methodologies and scales of personal heat exposure are further classified and discussed in section 3.

Zoom In Zoom Out Reset image size

Table 1. Summary of literature sample classifications. All selected publications (figure 1) are reviewed in detail to identify the methodologies used at various scales as well as targeted outcomes (up to two) and objectives (up to three).

Targeted outcomes	Objectives	Methods	Scales
Human Performance (including physical and cognitive performance and workers' productivity) Health (including mortality, morbidity, and well-being) Occupational Safety Energy Efficiency	Enhancing thermal comfort Monitoring personalized heat impact Real-time monitoring of heat exposure Spatial analysis of heat exposure Inclusion of human activity or adaptive behavior Inclusion of physiological responses Momentary assessment of subjective feedback Accuracy assessments and standardization	Wearables IoT sensors/Smartphones Mobile measurements IR-thermography Statistical modeling Machine learning Mobility data analytics	Indoor Outdoor Transition

In this section, we provide a more in-depth analysis of the driving aims and objectives in the literature sample, focusing on the underlying links as well as distinctions that exist in this multidisciplinary field. A co-occurrence map of keywords is generated for this purpose (figure 3), which provides a visual representation of keywords in the field, their co-occurrence and relatedness across different disciplines, and the key subtopics and themes that emerge. We used a combination of two keywords to assess the bibliographic network: (a) author-defined keywords in published articles, and (b) KeyWords Plus [43] generated in Clarivate Analytics databases based on words or phrases that frequently appear in the abstract and titles of an article's references, but do not appear in the title of the article itself. It is, however, worth noting that the formation of clusters observed here was obtained even without the inclusion of keyword plus.

Zoom In Zoom Out Reset image size

In this bibliographic network map (figure 3), each cluster (indicated by the color) represents a set of closely related keywords that are grouped together based on their relatedness and co-occurrences. Three clusters are formed in this analysis, demonstrating the major subtopics in the personal heat exposure assessments, centered around these key areas: (a) assessment of thermal comfort in the built environment (mainly focused on the buildings) using a combination of novel methodologies (such as IoT, machine learning, and IR thermography); (b) assessment of urban temperature and urban heat island, exacerbated by climate change, impacting the physical and mental health of urban dwellers, and (c) assessment of heat stress and strain, particularly through evaluation of physiological responses and human thermoregulation, as they further contribute to occupational and sporting safety.

This bibliometric visualization further informed the classification of the literature sample (section 2.2 and table 1), such that the body of work is categorized based on the targeted outcomes and objectives (figure 4). Although the assessment of personal heat exposure is distributed across various disciplines (figure 2) in these three emerging themes, we observe that the majority of publications (62%) focus on human performance (physical and cognitive) and workers' productivity as one of the key drivers for assessing heat exposure at the human scale. This focus highly overlaps with indoor studies of thermal comfort that are motivated by building energy efficiency (33%). Compared to the dominance of thermal comfort studies (36% in both indoor and outdoor scales), research that focused on occupational safety represents the smallest fraction of personal heat exposure assessments. This gap in knowledge calls for immediate attention to the personal assessment of occupational heat exposure. This is particularly important due to the diverse characteristics of different professions (such as constructions, transport, manufacturing, and agriculture) in the expected level of heat exposure indoors and outdoors as well as the type, intensity, and duration of occupational activity that may aggravate heat stress impacts (section 4). More importantly, as reportedly noted [19], psychological factors (such as autonomy, motivation, and competence), as well as racial and socioeconomic disparities (contributing to the lack of access to risk-mitigating options) [46] in many exertional workers result in grave heat stress impacts not only on work safety but also on human health and wellbeing.

Zoom In Zoom Out Reset image size

Research objectives identified in the literature sample are further assessed based on the established scientific criteria, leading to several observations. First, although collected articles stem from distinct research fields, the close link between different emerging themes as well as overlapping objectives in different disciplines, signal a positive trend in taking an interdisciplinary approach to urban heat exposure observed in the literature. This is particularly seen in considering the physiological responses to heat exposure together with environmental monitoring: 46% of publications monitored at least one physiological parameters (such as heart rate, skin temperature, or sweat rate) in their evaluation of thermal comfort or heat strain. This represents a shift in research methodology, particularly in environmental science and urban climate fields where physiological and psychological responses were traditionally neglected and one-size-fits-all approaches are considered to assess urban heat exposure. Additionally, over the last 3–4 years, a growing number of studies set out to integrate human activities (either through real-time monitoring, modeling, or mobility data analytics) in urban heat exposure assessment. The assessment of momentary feedback in an unsupervised and unobtrusive way is lagging behind, though novel methodologies using smartphones [47] or wearable devices [25] are gaining traction in this area. Lastly, despite the rapid developments in sensing and modeling methodologies over the last decade (section 3.3), there are very few studies that focus on accuracy assessments and standardization of data collection, comparison, and analytics. This represents a concerning trend in a rapidly evolving field, particularly as many studies deploy off-the-shelf and low-cost sensing solutions or rely on mobile sensing methodologies (including wearables) that are yet to be tested in unsupervised and dynamic use in realistic urban settings.

Although studies that detailed the need for personalized heat exposure assessments date back to the early 2000s [48, 49], the majority of progress in this field is made in the last decade (figure 5), with a significant rise in the number of studies and citations observed in the last 5–6 years. This trend coincides with a notable increase in urban heat island [29] and thermal comfort [29, 50] studies but, arguably, is mostly enabled due to technological advancements in environmental and physiological sensing (including the rise of consumer health wearables [26]), data communication and analytics (namely the evolution, progress, and rapid deployment of IoT architecture [51]), and cloud-centric visualization and computing. The literature suggests that stakeholder awareness of urban heat impact in certain (mostly developed) countries and industries has also contributed to the rise in personal heat studies (particularly as it relates to human performance and occupational safety), though there is no quantitative analysis of this effect available in the literature.

Zoom In Zoom Out Reset image size

Considering the multifaceted impacts of urban heat exposure, the spatial representation of this evolving field should also be evaluated, focusing on the scale of studies (indoor, outdoor, and transition as shown in figure 5) as well as the geographic focus of case studies. Review and classification of the literature sample indicate that the largest fraction of personal heat exposure studies (43.1%) is focused on indoor spaces, enabled by significant progress made in the field of indoor environmental quality (IEQ). This is often motivated by the argument that in developed countries, people on average spend more than 90% of their time indoors. However, a deeper look at this subset reveals that a significant majority of indoor studies are done in office spaces, with only six articles identified in the literature sample that focused on residential buildings. This represents a significant gap in personal heat exposure research (particularly in the post-COVID era), as people are more likely to be impacted by indoor heat exposure at home [52]. Additionally, the strongest protective factor for heat morbidity and morbidity is considered access to air conditioning [52], which is more widely available in office and commercial buildings. This disparity in research focus is likely caused by a long history of building operation standards, guidelines, and certifications (such as LEED and WELL certificates) that consider various aspects of IEQ for well-being and health in office buildings that are traditionally neglected for residential properties. Additionally, privacy concerns tend to make participant recruitment and data collection more difficult in residential buildings, motivating more methodologies to be implemented and tested in public or commercial spaces.

Unlike the assessment of indoor environment that is rooted in well-established thermal comfort literature, the focus on the transition spaces, i.e. transition of people and sensors from indoor to outdoor spaces, is observed more recently, mainly enabled by the rise of wearable and mobile devices used for environmental and physiological monitoring (79% of studies categorized as 'transition' used a form of wearable devices, discussed in detail in section 3.3). This is arguably one of the main progress made in addressing the challenges of conventional heat exposure studies, such that a larger spatial and temporal distribution of heat exposure can be obtained with lower costs compared to fixed or mobile measurements with centralized efforts.

Figure 6 takes this spatial analysis further to evaluate the geographic focus of the literature sample. In order to demonstrate the relative importance of heat exposure studies in different locations, the regional distribution of case studies in this review is represented in the context of maximum global heat stress, characterized by the monthly maximum value of daily maximum Universal Thermal Climate Index (UTCIxx), calculated from the open-source historical dataset on human thermal comfort [53]. Comparing the global map of UTCIxx with the regional location of case studies reveals that, not only there exists a general bias in the literature towards the northern-hemisphere and developed countries, but also there is a complete lack of representation in some areas with very strong to extreme heat stress with dense populations. Despite the growing number of publications in English in the field, there are only a handful of case studies focused in developed countries and regions such as the middle east, central and south-east Asia, Africa, and South America. Comparing this map with the global map of urban population (UN World Urbanization Prospects 2018) demonstrates that the higher frequency in focus is not necessarily linked to a larger percentage of the urban population in these regions, as many countries underrepresented in the field had 70%–90% urban population as of 2018. We note the exclusion of non-English articles is a limitation of this analysis that may have contributed to the geographic coverage (particularly in Latin America). However, this observation is similarly observed in previous research on urban heat that included foreign-language literature [29] and reflect a pattern reported previously in academic research. Accordingly, this observation calls for a new direction to address this critical geographic (and potentially cultural and socio-economic) knowledge gap in this field.

Zoom In Zoom Out Reset image size

Review of the literature sample further reveals a diverse set of methodologies used and developed for personal assessment of urban heat exposure (table 1). Figure 7 shows the distribution of different methodologies used, noting that a combination of sensing and modeling methods are seen in the literature with 52% of publications using more than one methodology to assess urban heat exposure. An in-depth review of this multidisciplinary field further reveals that methodologies are mainly evolved to address one of the two key perspectives:

Zoom In Zoom Out Reset image size

(a) Cover spatial and temporal variabilities in urban heat exposure, focusing on the immediate thermal environment (extrinsic)

In this perspective, research aims to evaluate the high variability of heat exposure within an urban landscape and address the heat impact misclassifications, i.e. when lack of comprehensive meteorological datasets (for instance missing radiation or wind speed data), or the use of meteorological stations remote from the city, can lead to misrepresenting the intensity, duration, or impact of heat exposure. Responding to this shortcoming, new methodologies focus on assessing the thermal environment variability experienced by individuals as they go about their lives, describing when and where heat exposure may happen. Sensing and modeling methodologies are then developed to evaluate the magnitude, duration, and frequency of heat exposure through an array of factors (such as air temperature, radiative exchange, humidity, and air velocity) that impose a risk on the physiological strain and perceived thermal comfort.

In sensing the immediate thermal environment of city dwellers, we observe a key paradigm shift from highly accurate and scientifically maintained weather sensing units (that are often sparse and at high cost) to low-cost, citizen-science, or crowd-sourced sensors and devices that are deployed more extensively [54, 55] or are mobile [56–58]. The rise of IoT in environmental sensing [59] has further enabled this paradigm shift, leading to the emergence of ubiquitous and 'nearable' sensors [60, 61] (i.e. in the immediate environment of individuals), that cover a wider range of spatial distributions at reduced costs and can be simultaneously deployed in various cities across the world [54]. Such fine-scale assessment of heat exposure further facilitated evaluating the role of urban fabric and morphology, as commonly characterized by 'local climate zones' (LCZ) [62], on determining the diurnal and annual cycles of heat exposure and distinguishing inter and intra-urban variability in heat exposure [58, 63]. However, despite tremendous efforts made in closing the thermal environment measurement gaps, data gathered from low-cost sensors appear to have limitations in two key aspects: (a) comprehensive representation of the thermal environment. sensing is largely focused on monitoring the urban temperature and relative humidity as the proxy for the thermal environment (while neglecting the critical impact of wind speed and MRT on heat exposure), and (b) accuracy assessment and standardization: to date, limited work is done to address the accuracy concerns and develop guidelines for deploying and more importantly comparing IoT data [64] or mobile measurements. A novel methodology that addresses the accuracy of sensing urban temperature at 'hyperlocal' scale is presented by Venter et al [65], integrating remote sensing and LiDAR data with crowd-sourced air measurements to achieve air temperature at 10–30 m resolution with sub-degree accuracy. This unique example further points to the need for comprehensive assessments of personal heat exposure that integrate relevant methodologies across different scales.

Modeling of the thermal environment in cities, on the other hand, is rooted in decades of climate modeling at micro- and meso-scale (several review articles have previously detailed modeling of UHI [66] and thermal comfort [67] as well as Computational Fluid Dynamics analysis of urban microclimate [68] that go beyond the scope of our review). However, access to data obtained from smartphones [69] and surrogate data sources (such as vehicle-based [69, 70] and mobility data [16, 71]) has provided more accurate and representative input parameters to inform modeling of the thermal environment, following detailed patterns of people's mobility in the built environment and accomplishing the visions of personal heat exposure assessments.

(b) Integrate and assess the human factor, focusing on the exposure impact of thermal environments (intrinsic)

This perspective aims to address the heat exposure impact on urban dwellers, providing information that explains how and to whom heat exposure is realized [8]. However, unlike the thermal environment that can be quantified in four environmental parameters, the human factor is far more complex with parameters non-linearly altering the relationship between heat exposure and outcomes. These personal factors include, but are not limited to, age, sex, body mass and surface area, pre-existing health conditions, activity and fitness level, psychological state, climatic background and acclimatization [72], and socioeconomic status (SES) [12, 46, 49]. Additionally, adaptive behaviors and efforts for heat management at the personal level (through a change in clothing, location, or hydration) or infrastructural level (through climate-responsive design or climate-control equipment) play a crucial role in alleviating the outcomes [73]. A combination of these factors determines how different individuals and subpopulations not only experience different heat exposures in urban areas [74], but also experience different levels of discomfort, physiological strain or psychological stress when subjected to similar thermal environments.

Responding to this complexity, methodologies to monitor heat exposure impact can be distinct in their integration of the human factor, focusing on one of the three critical subsets of heat exposure impact: (a) thermal discomfort and heat stress (through monitoring of physiological responses such as heart rate, heart rate variability, skin temperature, and conductance as well as subjective feedback using automated surveys obtained from smartphones, smartwatches, and interactive screens) [75], (b) heat-wellbeing impact (through monitoring of self-reported wellbeing factors as well as indirect contributors to health impacts such as sleep quality [32] and exposure to natural environments [76]), or (c) heat injury caused by physiological strain (through monitoring of physiological responses, particularly core temperature) [10].

Two emerging sensing methods stand out in all three categories (figure 3): (a) wearables technologies (i.e. devices that can be worn or integrated into clothing) that measure the physiological response (such as heart rate, skin temperature, and skin conductance) [37, 38], human activities [77], or subjective feedback [25], and (b) infrared thermography (i.e. cameras detecting infrared energy emitted from objects), which monitor physiological responses (mainly skin temperature at various body parts) [41, 78] and human activity pattern [79]. Wearable methodologies include a range of devices worn on the wrist [25, 80], chest, head [81], or shoes [82] as well as a subset of smart garments [83]. Although monitoring similar parameters (such as skin temperature and activity patterns), these two methodologies are distinct in the scales they cover and opportunities they provide: while wearables can be carried by individuals in a range of indoor, outdoor, and transition spaces, IR-thermography in the field is often confined to indoor environments. Additionally, bioengineering advances in noninvasive sensor technologies (including miniaturization, reduced cost, power requirements, and comfort) [37] make it more feasible for wearables to be deployed at a larger scale to produce individual monitoring and predictions of thermal comfort and heat stress.

Informed by advancement made in personal heat exposure sensing (through the rise of IoT and wearable technologies), modeling algorithms and machine learning techniques [34, 84] are then developed based on physical and physiological models to predict heat exposure impact and close the loop in achieving a targeted outcome in the face of the dynamic thermal environment. These models are mainly focused on predicting thermal comfort and physiological strain. Focusing on the former, personal comfort models [34] are developed using a combination of IoT or wearable sensing and machine learning algorithms that predict an individual's thermal comfort response, instead of the average response of a large population. In order to monitor and assess physiological strain of individuals (particularly focused on avoiding heat injuries and performance degradation in outdoor workers), many have turned to investigating and employing the methods of thermoregulatory models. However, the most reliable indicator of thermal work strain is body core temperature (Tc), which is often invasive, inconvenient and impractical to directly measure during work or physical activities. Accordingly, heat stress indices and models are used to predict and quantify heat strain [85]. These modeling methods range from empirical analysis to focusing on heat balance and thermoregulatory systems, with the latter considering a detailed physics of clothing, skin evaporation and blood flow regulation of the human body [86]. However, as the validity of the heat-balance model predictions rely on the combination of clothing, climate, and physical workloads an individual is experiencing, they are often more suitable for assessing the average response of the population [48]. Accordingly, empirical prediction of Tc based on a combination of heart rate (and heart rate variability), skin temperature (at various body parts), and sweat rate monitoring is proposed [37, 48] with prediction accuracy of 0.3 °C–0.65 °C Root Mean Square Error (RMSW) reported [24, 87]. A recent study [24] further combined air temperature measurements at wrist with heart rate and wrist skin temperature to achieve a higher percentage of attainment rate within 0.3 °C, which is more suitable for detecting heat strain in high-risk individuals in real time. However, to be incorporated in a larger scale in heat strain and thermal comfort predictions, future research should assess performance of all proposed personal models in a larger scale and with more representative subject profiles (including vulnerable groups such as elderly, young children, and people with pre-existing health conditions), which is currently lacking in the literature.

Focusing on the extrinsic factors, one of the main challenges in this field is providing a more accurate representation of urban heat exposure across the city and to do so as it happens in real-time. This is particularly critical is personal heat exposure assessments: while it is not feasible to provide a personalized assessment of heat exposure for every vulnerable individual in the society, it may be feasible to cover and classify as many heat exposure conditions as possible across the built environment.

These efforts to provide fine-scale assessments of urban heat exposure are done across three scales: city-scale, building-scale, and human-scale [59]. At the city level, the variation in heat exposure in heterogeneous urban environments is of focus, using fixed IoT and crowdsourced/citizen-science networks, as well as mobile sensors that are handheld, mounted on bikes, or implemented in various vehicles. The focus of these assessments is still largely on monitoring temperature and humidity as a proxy for the thermal environment, which is less valid in outdoor environments where dynamic changes in radiation and wind speed dominate heat exposure [88]. This is mainly caused by the prohibitive cost and challenges of sensing MRT and wind speed in large-scale deployments, as well as accuracy concerns based on the location and resolution of sensors. There are very limited examples of crowdsourcing or IoT wind speed measurements at the city-scale that comprehensively assess and remove the uncertainties in data [89], revealing much-needed attention in the field. This further suggests the need for integrating crowdsourced or IoT data with other modeling or sensing methodologies (such as remote sensing [65]) to provide a more comprehensive outlook on fine-scale urban heat exposure. An example of such methodologies is the calculation of sky view factor (and further estimating MRT) using large datasets of street-level images (such as Google Street View) [90–92] that can complement monitoring of thermal environment at a finer scale.

Quantifying the intra- and inter-urban variability in outdoor thermal environments is another focus of city-scale monitoring, identifying considerable temperature differences (0.5 °C–2.5 °C) observed between different LCZ classes [56]. However, while LCZ classifications (governed under the World Urban Database and Access Portal Tools [93]) strictly follow a standardized procedure for classifying urban neighborhoods across the globe (based on their morphology, land cover, and land use types), the monitored data on the thermal environment is obtained using a range of different sensors with a large range of accuracy, resolutions, and setups (for instance, deployed with or without radiation shields, installed at different heights, combined mobile and fixed measurements, or considered at different resolutions). Providing guidelines and potential means to standardize data collection at finer resolutions are needed in future research to scale the analyses and classification of heat exposure in different cities and neighborhoods. The outcome can then be linked with urban design practices, providing feedback to urban planners and decision-makers to incorporate climatic considerations into the planning process and ultimately mitigate outcomes of heat exposure.

At the building scale, the emergence of IoT and low-cost sensors has enabled real-time and spatially variable monitoring of heat exposure, although the body of research appears to have a significant bias towards commercial spaces. To provide a more representative classification of heat exposure, particularly as it relates to the frequency and intensity of heat exposure, measurements need to be extended to a variety of indoor spaces including residential housings with different socioeconomic profiles, aged care facilities, and homeless shelters as well as educational and vocational spaces such as schools and campuses, factories, and healthcare centers. This is particularly needed in developing countries or areas with lower SES, as one of the most critical risk factors is the ill-suited living conditions and lack of resources to cope with urban heat exposure. More focus on quantifying and classifying the indoor thermal environment can further quantify the level of energy assistance needed to deal with energy poverty challenges observed around the world [94], further minimizing adverse heat exposure impacts.

At the human scale, research has focused on monitoring individually experienced temperatures (IET) and humidity. A range of sensors (such as wearable devices or sensor mounted on backpacks and clothing) are carried by students [95], traffic officers [96], and outdoor workers [97] to look at the variability in the thermal environment experienced by individuals, and reduce misclassification of exposure by extending the measurements at a human scale. Diurnal intra-urban human movement and mobility data is also deployed in this category (using mobile data or mobility pattern models [98, 99]), to better predict residents' exposure to heat. Similar to the city- and building-scale monitoring, however, there is limited discussion on the accuracy of data collection in dynamic use and particularly based on where sensors are placed. To yield reliable data, sensors need to be exposed to atmospheric conditions and away from heat sources and interferences. To be used in realistic and unsupervised scenarios, however, requires that sensors are compliant with the participants' comfort. Sensing solutions that achieve both conditions include miniaturized sensors mounted on wristbands or clothing (such as on shoes) that do not interfere with the individual's activity. However, consistent guidelines are needed, and yet to be determined, as differing placements may result in skewed heat exposure measurements at the human scale [100]. More technological advancements are also needed to address sensor robustness and drift in dynamic use, while ensuring wireless communication, efficient power consumption, and processes that comply with individuals' data security [101]. A sensing solution that achieves (or scientifically examines) these criteria has not been achieved or comprehensively discussed in the literature. This is particularly relevant when human (physiological, behavioral, and psychological) responses are recorded together with the thermal environment (section 4.2). Accordingly, it is paramount that future research not only investigates individually experienced thermal environments in more unsupervised settings but also determine what factors contribute to the uncertainties of data collections and how they can be accounted for in classifying heat exposure.

The health and wellbeing of individuals exposed to heat are linked to their thermal comfort and heat strain, encompassing intrinsic factors that should be integrated in assessing the ensuing impacts. Thermal comfort can be degraded due to passive heat stress where an individual may experience a change in skin temperature or sweat rate but often with little change in body core temperature. Heat strain, on the other hand, is often induced by extreme thermal conditions as well as exertional heat stress (due to work or exercise) that result in elevation of body core temperature. Individuals become at risk of heat strain not only when extreme heat exposure is realized, but also when they are exposed to chronic and prolonged (milder) heat exposures [102] that increase the body's core temperature beyond what is tolerable for physiological functioning. Traditional heat-health studies have primarily assessed these concerns focused on mortality, particularly as it relates to extreme heat events, followed by morbidity in a range of acute and chronic conditions that can be statistically assessed in the population. Research is now trying to better link the heat-related health effects to the measured heat exposure through various types of health datasets, surveys, interviews, and monitors across different scales. Personal heat exposure studies are at the forefront of these endeavors, creating a shift from focusing on 'after-the-fact' phenomena and instead, assessing the impact of chronic and extreme heat exposures on different subsets of the society.

When focusing on human comfort, personal comfort models present a novel methodology in the field, bringing the attention back to identifying personal differences in heat exposure impact [34]. These models have the ability to close the 'monitor, predict, control' loop such that heat exposure monitoring results in personalized recommendation engines that mitigate the impact of heat exposure. However, given that it is not feasible to identify every person exposed to heat, clustering of thermal comfort 'personalities', i.e. redefining the thermal comfort zones based on a cluster of occupants and residents [25, 59], might present a more effective approach to addressing human comfort challenges. To achieve this objective, two areas should arguably be prioritized in future research. First, particularly given the subjective nature of thermal comfort, further investigations into obtaining momentary feedback of individuals in an unobtrusive and unsupervised way is needed, such that human assessments of the thermal environment is included as one of the key parameters in identifying the heat impact. Second, most consistent and evidence-based physiological monitoring is needed to predict thermal comfort. Although physiological monitoring appears to be a critical aspect of analyzing personal thermal comfort, it is yet unclear what are the key physiological parameters that fully represent and predict thermal sensation or pleasure. Personal thermal comfort studies have focused on skin temperature, skin conductivity, sweat rate, and heart rate as measures for prediction of thermal comfort, with the most scalable approach being the skin temperature and heart rate at wrist or face (obtained from wearable and IR thermography, respectively). Future research should provide a more comprehensive review of physiological responses determining thermal comfort in a more diverse set of populations than office workers. More accurate and consistent physiological monitoring in response to heat can further bridge the gap between thermal comfort and heat-health studies and specify the personal boundaries between thermal discomfort and heat strain.

Focusing on the health impacts of heat exposure, outdoor workers appear to be one of the most vulnerable groups that can benefit from personal heat exposure assessments and the use of mobile and wearable devices has made it feasible to monitor their daily thermal comfort and heat adaptations. When individually experienced thermal conditions are monitored for outdoor workers, it was shown that they indeed experienced more adverse thermal conditions than what was reported at weather stations [96, 97] and more importantly, are much more prone to heat strain due to the exertional load and unsuitable clothing for heat adaptation. A range of professions have been the subject of personal heat exposure analyses, including construction, agricultural, municipal, and ground workers [83, 97, 103–105] as well as soldiers [106], but the extent of work is not remotely sufficient to monitor, predict, and control heat exposure in such a diverse subpopulation with a highly variable exertional load and SES. Comparing this urgency with limited representation in the field (section 3.1) demonstrates that there is an urgent need to shift the attention and use cases of research in personal heat exposure studies from office and commercial to outdoor workers to address both social and environmental aspects of this challenge.

Besides threatening the health of the worker, occupational heat exposure can also degrade performance or work capacity [6]. Performance is defined as the amount of work that can be produced within a fixed time while capacity refers to the maximum amount of work done till exhaustion. The former is similar to paying by piece rate such that the worker can produce a high workload, and therefore can be exposed to higher heat strain which places him/her at a high risk of incurring heat illness. The latter assumes a fixed metabolic heat production so in situations when the individual cannot determine the work duration, heat illness can ensue too. This particularly is a concern in workers from low- and middle-income tropical countries experiencing excessive heat exposure that often degrades performance and well-being [19, 107]. These workers often have competing motivations as they cannot afford to stop working despite the disadvantages presented, and therefore, are more likely to underreport heat strain or heat illnesses unless severe. There are recognized international standards to protect the workers in the face of extreme heat conditions but such standards are not considered for all vulnerable groups (for instance taxi drivers or seasonal ground workers). More importantly, they have been developed mostly in temperate western settings, thus their relevance and validity are questioned for workers from a different geographical, cultural, or socio-economic background [107]. The assessment of immediate thermal environments is therefore encouraged to implement self-pacing and work breaks to reduce heat strain [108]. However, it is important to acknowledge and address that these practices may lead to reduced work rates that employers do not favor and therefore discourage long-term adoption. The use of wearable sensors to monitor workers (and limit activity only when needed) can be considered as a more effective compromise for addressing the performance loss. Personal assessment of heat exposure in these subpopulations (particularly when including physiological and psychological responses) can provide an unprecedented knowledge regarding the extent of heat exposure impact on both health and human performance. If a digitally enhanced workforce will become standard practice in the field, the quantification of human performance degradation in response to various levels of heat exposure can be achieved [109]. It is, however, likely that many employers are not able to purchase wearable or personalized heat monitoring devices for their workers due to the associated costs, or may disregard the need due to lack of understanding of the issue. Research focused on workers [105] suggests that there is a mismatch between worker perception of heat and prevalence of heat strain, which leads to limited adaptation actions to protect themselves during the non-heatwave period and at night. This further details the need for more education and communication with regards to heat impacts, which is indeed not unique to occupational workplaces.

Similarly, heat exposure affects athletes, mainly due to their high exertional load. The risk of heat injuries is ever present in exercise in heat due to the growing number of competitive sports in the face of climate change. Despite knowledge of exertional heat injuries and better treatment available for prevention, recognition and treatment [20], it is not uncommon for an individual performing an intense exercise to suffer in performance and health. It is shown that heat stress reduces power output during self-paced exercise even in highly trained men, whereby this decrease is associated with factors related to body temperature rather than metabolic capacity [110]. Personal heat exposure monitoring for this group of people is strongly advocated through the use of wearable sensors, such that personalized recommendations can be achieved. However, personal heat exposure may still differ according to sensor placement as certain sensors need to be administered in regions with high vascular density and superficial vascular depth in order to obtain reliable physiology data for continuous measurements of health during heat stress [77, 100]. It has been shown that the physiological data acquired by the sensors at the chest are most accurate during sport [111], motivating the development of specialized sensors weaved or integrated into clothing in order to acquire health information more seamlessly [112].

More related to the general population than specific professions, it is evident that assessment of heat exposure impact on wellbeing is overlooked in the literature [24]. Research suggests that heat exposure affects various aspects of human life and wellbeing beyond the direct heat-health relationship, including physical activity [30], food intake [31], sleep quality [32] and mental health [33] and exposure to outdoor environments, ultimately contributing to creating obesogenic environments. Ambient temperature is one of several factors hypothesized to influence obesity, a significant public health concern. Evidence of the role of ambient temperature relating to obesity (through energy imbalance of intake and expenditure) has been limited to observational studies, model organism studies, and few randomized control trials. The results of observational studies are limited to mere associations to generate hypotheses [113]. Future direction in research can measure and quantify the thresholds and conditions that alter food intake, activity levels, and time spent outdoors linked with temperatures experienced to gain better insight into how these effects may affect subsequent wellbeing of individuals in the built environments.