Spatial assessment of heat-health risk among vulnerable populations in Greater Perth: integrating hazard, exposure and vulnerability indicators Open Access

The effects of heatwaves vary among individuals, with certain people being more vulnerable because of factors such as demographics, socio-economic status, health and the biological and physical features of their environment (Mason et al., 2022; Tong et al., 2010; Zhu and Yuan, 2023). During heatwaves, vulnerable population subgroups have a greater need for health services (Çağlak and Matzarakis, 2024; Mason et al., 2022). Over the last decade, it has been shown that certain groups of people, such as the elderly and those living in poverty, exhibit heightened sensitivity (Wu et al., 2022). Elderly individuals are more prone to cardiovascular problems caused by heat (Abrar et al., 2022; Alonso and Renard, 2020). For example, in Brisbane, Australia, the death rate of those aged over 65 years rises by 7% for each 1°C increase in mean summer daytime temperature (Adnan et al., 2022). Research has also shown an inverse relationship between heat stress and income level (Adnan et al., 2022; Alonso and Renard, 2020; Cutter et al., 2003; Yoon, 2012). Individuals residing in suburban areas characterized by low and moderate socioeconomic levels have heightened susceptibility to heat-related conditions (Adnan et al., 2022).

Understanding the causes of increased temperature is also essential for effective risk assessment and targeted interventions aimed at protecting vulnerable population groups during heatwave events (Estoque et al., 2020; Maragno et al., 2020; Rao, Gupta, et al., 2021). For example, temperature and thermal behaviour exhibit significant variation depending on the type of land use (Kim et al., 2022; Njoku and Tenenbaum, 2022; Saha et al., 2021). Areas that include green land use and adequate green areas (GAs) have been found to mitigate heat absorption and promote balance in urban environments by supporting cooling effects (Alavipanah et al., 2015; Galalizadeh et al., 2024). However, the cooling effect decreases with distance from a park and with a reduction in the percentage of canopy cover (Galalizadeh et al., 2024). Thus, it is possible to include proximity to GAs and the percentage of canopy cover as contributing factors in the assessment of heat risk.

Western Australia has a lengthy record of experiencing severe hot weather and associated hazards such as heat waves, heat-induced droughts and bushfires (Akompab et al., 2013; Jyoteeshkumar Reddy et al., 2021; Trancoso et al., 2020). There are two approaches to heatwave mapping and heatwave warnings in Australia, conducted by the Department of Health and Bureau of Meteorology, respectively. In the first approach, a Perth metropolitan heatwave impact map has been generated at local government area scales in which heat impact is classified into different ranges from least to highest. It measures health-related impacts by contracting the rate of hospital admissions and emergency room visits for particular health problems during heatwave and non-heatwave days (Western Australia Department of Health, 2022). The second strategy, known as the national heatwave service, forecasts the expected highest and lowest temperatures and the timing of the heatwave’s peak or decline and identifies the communities located in the high-risk zone (Bureau of Meteorology, 2023). The only index used in this approach is heatwave, which refers to a period of three or more consecutive days with unusually high daytime and night-time temperatures, relative to the local long-term climate and recent past.

Beyond these operational measures at the state level, recent research has undertaken broader heat-risk assessments. At a national scale, Wang et al. (2023) introduced a comprehensive heat-risk evaluation in which heat hazard, heat exposure and heat vulnerability indices were developed based on IPCC’s risk assessment framework to produce an overall Heat Health Risk Index (HHRI) at the Small Areas Level 1 scale (SA1) across Australian capital cities. In their study, the heat hazard index was derived from land surface temperature (LST), which does not always correlate directly with air temperature. For example, in open landscapes such as shrublands, grasslands and croplands, LST can be 10°C to 20°C higher than the corresponding air temperatures, especially under higher temperature conditions (Mildrexler et al., 2011). Nevertheless, LST can serve as a reliable proxy for ambient heat when validated and adjusted against local meteorological data. Additionally, the study by Wang et al. was conducted at a national scale, focused on identifying high-risk areas but did not specify the relative contributions of each index to the overall heat risk in the cities studied. For instance, high heat risk in an area may primarily stem from vulnerability, while in another, it might result from exposure. As a result, it is essential to provide detailed insights into the dominant drivers of heat risk at the local level. These gaps motivate a city-scale analysis that integrates hazard, exposure and vulnerability and reports their relative contributions to local heat health risk.

Implementing a risk assessment approach, the research reported on here provides a thorough evaluation of the risks posed by heat waves to vulnerable individuals in Greater Perth, Western Australia. The assessment incorporates indicators that have been previously examined, as well as new indicators that cover both indoor and outdoor heatwave risks. This study aims to develop an SA1-level heat health risk assessment for Greater Perth, maps the composite risk surface and quantifies the relative contribution of each component. This study investigates three interrelated research questions that address the spatial patterns and determinants of heat health risk in Greater Perth. First, it examines which SA1 neighbourhoods display the highest composite heat health risk. Second, it evaluates, within these high-risk neighbourhoods, whether hazard, exposure, or vulnerability exerts the strongest influence on local risk levels. Third, it explores how proximity to GAs and the extent of tree canopy cover are associated with the distribution of exposure hotspots and zones of elevated heat health risk.

The results of this study are significant for implementing location-specific adaptation strategies that effectively mitigate heat-related diseases and deaths. This holistic approach ensures that adaptation strategies not only target vulnerable populations but also address the underlying environmental factors driving heat stress. As a result, policymakers can develop tailored interventions to enhance the resilience of communities to heatwaves. By situating the assessment within the frameworks of Western Australia’s Heatwave Management Plan and Australia’s National Climate Resilience and Adaptation Strategy, and by aligning with global imperatives such as the United Nations Sustainable Development Goals and the World Health Organization’s guidance on heat-health action plans, this study contributes simultaneously to regional practice and to the broader international discourse on climate adaptation.

The spatial risk assessment, based on the IPCC (2014) framework used in this study, follows a systematic approach (see Figure 1). The methodology begins with comprehensive data collection, integrating diverse sources to construct multi-dimensional indicators of heat risk components. Data analysis involved the spatial classification of indicators, which were then normalized on a scale from 0 to 1, representing the least and most significant values, respectively. The heat risk layer was ultimately created by combining all the indices. The methodology ensures a holistic understanding of the interplay between environmental, demographic, health-related and socio-economic indicators in assessing and mapping heat-related risks. A quantitative, indicator-based methodology was selected because it enables systematic integration of heterogeneous data sets, generates objective and reproducible metrics (Mehrotra et al., 2019; Zhou and Lai, 2025) and aligns with the IPCC risk assessment framework, making it particularly suited to evaluating spatial heat risk in complex urban environments. This research did not involve any direct human participants or personal data. It relied exclusively on aggregated, publicly available data sets (e.g. meteorological records and census statistics). Consequently, no individual informed consent was required, and formal institutional ethics approval was not applicable.

Perth, located on the Indian Ocean coast (Figure 2), serves as the capital of Western Australia and is home to a population of over 2.2 million individuals (ABS, 2022). The region experiences a hot Mediterranean climate during the summer, with mild and highly seasonal rainfall primarily occurring in the winter. Winters are frequently characterized by cold and humid conditions, whereas summers (December to March) are marked by hot weather (Duncan et al., 2019; Zijlema et al., 2019). Studies have demonstrated that heatwaves in Australia, specifically in Perth, are increasing in frequency, duration and intensity in the 21st century (Cowan et al., 2014). According to Adnan et al. (2022), the cost of hospital health care for diseases caused by heat in Perth is expected to rise from AU$79.5m between 2006 and 2012 to AU$125.8m–129.1m between 2026 and 2032 because of forecasted climate change scenarios.

The research is grounded in the IPCC risk model, which defines risk as a function of hazard, exposure and vulnerability (IPCC, 2014). This theoretical framework guided the indicator selection and analysis, ensuring a coherent and rigorous approach to assessing heatwave risk. A thorough literature review identified the most effective indicators, validated through empirical studies, ensuring their suitability for urban heat risk assessments. Each indicator was rigorously assessed for its relevance and ability to accurately reflect the complex interactions underpinning heat vulnerability. As a result, the first step in the heat risk assessment was collecting data from multiple sources to develop multi-dimensional indicators.

The indicators were split into three components, including hazard, exposure and vulnerability, which collectively influence heat risk (Maragno et al., 2020). The starting point of the risk assessment is heat hazard, which is characterized by high temperatures or extreme heat events (Bansal and Kianmehr, 2022). Exposure increases the risk by considering where and how individuals interact with the hazard, such as proximity to hot spots (Liu et al., 2023). Vulnerability further influences this interaction by accounting for sociodemographic variables, such as age, health and access to cooling services, which affect an individual’s ability to tolerate heat stress (Maragno et al., 2020).

The analysis was conducted at the Australian Bureau of Statistics SA1 to align with available health, demographic and travel-to-work census indicators and to enable neighbourhood-scale mapping (typical SA1 population = 200–800). However, working at SA1 can mask within-neighbourhood heterogeneity and introduce aggregation bias (i.e. the modifiable areal unit problem/ecological fallacy). As such, an indicator-based approach was designed as a screening tool to identify areas warranting closer attention, not to infer risk for individual households or parcels. Downscaling to parcels, blocks, or facility-level assets should be undertaken in subsequent, local analyses where finer data and ethics approvals are available.

The term “heat hazard” refers to the potential for experiencing high temperatures that could have detrimental effects on one’s health, whether as a result of natural phenomena or human activity (Hammer et al., 2020). Similar to previous studies (Faisal et al., 2021; Loughnan et al., 2012; Mohammad and Goswami, 2022; Willett and Sherwood, 2012; Williams et al., 2012), this research identified that the places with significant heat hazards are those that are above the threshold air temperature of 35°C. Because the current meteorological stations do not encompass the entirety of the study area, remotely sensed LST data at the scale of the entire study area was used. However, LST represents surface temperature, while air temperature is measured at 1.5–2 m above ground level. This difference can cause discrepancies, especially in urban areas where materials such as asphalt retain heat more than vegetation (Iqbal and Ali, 2022).

While LST does not provide a direct measurement of air temperature, numerous studies have demonstrated a significant association between these two distinct data sets (Chen et al., 2018; Eugenio Pappalardo et al., 2023; Faisal et al., 2021; Mohammad et al., 2022). The maximum daily temperature of seven Bureau of Meteorology weather stations (Table 1) located in various parts of Greater Perth was collected from January 1 to the end of March from 2001 to 2024. LST data was also extracted for each of the weather station locations from Landsat 7 and 8 Level 2, Collection 2, Tier 1 products on Google Earth Engine (GEE) with a spatial resolution of 30 m and a temporal resolution of 16 days. GEE is a cloud-based computational platform that provides pre-processed images automatically (Rodrigues de Almeida et al., 2023). The seven Bureau of Meteorology stations (Table 1) were purposively chosen to represent the main climatic gradients and urban–rural transition zones across Greater Perth, ensuring that the calibration between LST and near-surface air temperature is robust across microclimates, while the 30-m Landsat imagery provides the fine spatial resolution necessary to detect intra-urban thermal variability. Daily maximum air temperature and LST were then compared to determine whether changes in air temperature trends could be explained by LST. It verifies which value of LST could be used as a proxy for an air temperature threshold of 35°C. Finally, based on the climate extremes framework (Bureau of Meteorology, 2023), which uses thresholds such as maximum air temperature > 40°C and maximum air temperature > 35°C to define very hot and hot days, respectively, the proxy LST map for Greater Perth was classified similarly. The remaining areas were then categorized into three distinct classes: very low, low and moderate.

Greater Perth encompasses a diverse mix of land uses, ranging from natural to built areas, which can serve as the environmental indicators influencing LST. In this study, a comprehensive analysis was conducted to ascertain the significance of different land use in the dynamics of LST in the study area. To achieve that, the land use map was extracted from Landsat 8 satellite images with a spatial resolution of 30 m. Supervised classification was then applied to the satellite images, with 2200 sampling points used for training. Training points were proportionally stratified across all land-use categories to ensure representative sampling and to reduce classification bias, while a separate set of 300 validation points was reserved to provide an unbiased assessment of mapping accuracy. The classification accuracy was assessed using a confusion matrix, with 300 points reserved for accuracy assessment. Land use classification was incorporated to distinguish areas with high impervious surfaces, which retain more heat.

Distance to GAs and canopy coverage were the environmental indicators categorized into the exposure component of this study. The inclusion of these indicators was based on extensive literature highlighting their influence on urban heat island (UHI) effects. Studies (Duncan et al., 2019; Galalizadeh et al., 2024) emphasize that urban greenery and tree canopy coverage significantly reduce heat stress by providing shade and enhancing evapotranspiration. Distance to GAs was also specifically chosen as an indicator because the cooling effect of parks varies based on proximity, with research showing that temperature mitigation benefits extend between 50 and 800 m from GAs (Doick et al., 2014; Ghosh and Das, 2018; Jaganmohan et al., 2016; La Rosa, 2014; Lin et al., 2017; Shashua-Bar and Hoffman, 2000). Additionally, “travel method” was introduced as an exposure indicator, recognizing that individuals walking or biking to work have significantly higher exposure to heat than those using private vehicles (Karner et al., 2015). These factors have not been widely incorporated in previous heat risk assessments, making it a novel contribution of this study.

Based on the variation of cooling distance, the “Distance to GAs” layer was generated by creating 0–50 m, 50–120 m, 120–300 m, 300–800 m and more than 800 m buffers around the GAs feature in ArcMap. The canopy map for the region was extracted from Urban Forest Mesh Blocks (DPLH, 2020). This map was categorized into 0%–15%, 15%–30%, 30%–50%, 50%–60% and more than 60%. Additionally, normalized difference vegetation index (NDVI) data were derived from the same satellite imagery using GEE, given the strong negative correlation between LST and NDVI during warm months (Sun and Kafatos, 2007). NDVI values range from −1.0 to +1.0, with negative values indicating water bodies. Lower values (≤0.1) are typically associated with barren surfaces. Moderate NDVI values (0.2–0.5) correspond to sparse vegetation, including shrubs, grasslands, or crops. In contrast, high NDVI values (0.6–0.9) indicate dense vegetation, such as woodland (USDA, 2008).

The “Travel method” data, along with demographic, health and socio-economic information from the Census of Population and Housing at the SA1 geographical unit, is presented in Table 2. Each SA1 typically represents about 400 people, with a range between 200 and 800 people (ABS, 2021). This data set was then integrated with the shapefile of SA1 boundaries in ArcMap, capturing the population count in each unit. The use of SA1-level demographic and socio-economic data provides a fine-grained representation of community characteristics while maintaining confidentiality, thereby enabling precise mapping of population exposure and vulnerability at a neighbourhood scale (Astell-Burt et al., 2014; Wang et al., 2023). The resulting layers were subsequently converted into raster format at a 30 m resolution.

When discussing heat, vulnerability refers to a person’s level of heat sensitivity and capacity, which can depend on several environmental, health and demographic factors (Inostroza et al., 2016; Wu et al., 2022). The sensitivity of vulnerable groups, such as older adults and small children, to physiological and social impacts has been further intensified by heat stress (Abrar et al., 2022; Alonso and Renard, 2020; Gonzalez-Trevizo et al., 2021; Heckert and Rosan, 2016). According to several studies (Çağlak and Matzarakis, 2024; Mallen et al., 2019; Thanvisitthpon, 2023; Wolf and McGregor, 2013), people with lower health conditions and those who are battling chronic diseases are also more sensitive to heat. Moreover, low-income individuals may face challenges affording homes in more expensive areas with abundant GA, resulting in a potentially higher exposure to heat stress (Chuang and Gober, 2015; Mallen et al., 2019; Pereira Barboza et al., 2023; Thanvisitthpon, 2023; Wolf and McGregor, 2013). They may also rent and/or not have access to well-insulated accommodation or cooling in the home (Eugenio Pappalardo et al., 2023). It implies that individuals with lower incomes may be more vulnerable to health-related heat risk if exposed, as they possess a lower capability to cope with and adapt to heat (Adams et al., 2022).

In this study, cardiovascular diseases and chronic respiratory diseases were used as the health indices, as these conditions consistently show the strongest and most immediate increases in morbidity and mortality during hot weather. Multi-city meta-analyses demonstrate significant heat-attributable increases in cardiorespiratory outcomes (hospitalizations, ED attendances and deaths) during heatwaves (e.g. pooled effects for cardiovascular and respiratory endpoints) (Anderson and Bell, 2011; Cheng et al., 2019). Australian syntheses similarly report elevated risk for cardiovascular and respiratory diseases during heat events and increased health-service demand (Amoatey et al., 2025; Mason et al., 2022). Asthma-specific impacts are also documented. For example, childhood asthma ED presentations increase at higher temperatures after controlling for air pollution and seasonality in Brisbane (Xu et al., 2013). On this basis, cardiovascular disease and chronic respiratory disease were selected as core health indicators for vulnerability in our HHRI. This choice aligns with prior work emphasizing conditions where thermoregulation, cardiorespiratory load and airway reactivity make individuals uniquely susceptible to heat stress.

Data analysis in this study included two main stages of normalizing and overlaying of different indicators, displayed in Figure 1, to generate the final risk map.

The first step was taken to normalize the values using two methods of fuzzy membership: user-defined and linear. Fuzzy assesses the likelihood that each pixel is a member of a fuzzy set by evaluating any of a number of fuzzy set membership functions (Akbari et al., 2023). Fuzzy membership functions were chosen because they capture nonlinear relationships between indicator values and heat risk, permit continuous scaling of disparate variables onto a common 0–1 scale and are widely used in urban heat studies (Kumar and Mishra, 2025; Santos et al., 2024). A fuzzy set is a group of objects where each object has a degree of membership, indicated by a membership function that assigns a value between 0 and 1 to each object, with 0 indicating no membership and 1 indicating full membership (Sema et al., 2017).

User-defined function: This function allows for customization based on the specific characteristics of each indicator (Rajaei et al., 2021). It was used to normalize biophysical indicators, such as LST, distance to GAs, NDVI and canopy coverage. The membership function was designed such that the most important classes (such as areas with high LST) were assigned values closer to 1, while the least important or least extreme values (e.g. areas with low LST) were assigned values closer to 0.

Using the same function, “Travel method” was normalized based on the average time spent on each method. People who walked to work covered an average distance of 6.1 km with an average speed of 143.4 cm/s (Bohannon and Williams Andrews, 2011), while bicycle commuters had only a slightly longer average commuting distance (7.6 km) (ABS, 2023) with an average speed of 20 kph for two-wheelers (Western Australian Legislation, 2021). Furthermore, the average time for biking and walking is 23 and 67 min, respectively. Consequently, people who walk to work are more exposed to heat impacts, and as a result, this class obtained the higher value.

Linear function: The linear fuzzy membership function was applied to normalize socio-economic indicator layers. This function is particularly effective for data sets that exhibit a linear relationship between values (Mostafa et al., 2023). In this case, the linear function was used to normalize the vulnerability of populations based on the number of vulnerable individuals in an area. The linear function assigns a membership value of 0 to the minimum value (population) and 1 to the maximum value, with all other values scaled proportionally in between. For example, if a socio-economic indicator such as the percentage of elderly or low-income population increased, its corresponding membership value also increased, indicating higher vulnerability to heat risks. The linear function is mathematically expressed as follows, where min and max are user-defined inputs for the range of values:

As shown in Figure 3, the linear function can have either a positive or negative slope depending on whether the min value is smaller or larger than the max value. When min is less than max, the function generates a positive slope, and when min exceeds max, the function generates a negative slope (Mohebbi Tafreshi et al., 2021).

The final risk layer was generated by overlaying all heat risk indicators of the components: hazards, exposures and vulnerabilities. Because heat health risk assessment is inherently multidisciplinary, requiring expert input from various fields, developing highly specific, expert-driven classification thresholds is complex and beyond the scope of this study. Consequently, an equal weighting scheme was adopted for hazard, exposure and vulnerability layers, as is common in heat-risk literature (Cutter et al., 2003; Koks et al., 2015) when empirical evidence is insufficient to justify differential weights, ensuring transparency and reproducibility in the composite risk surface. Equal weighting is also widely used in operational risk indices and recent heat-risk applications, supporting interpretability and comparability across places and time (Hua et al., 2021; Olivares et al., 2025). Within each component, indicators were normalized (via fuzzy membership) and combined with equal weights to prevent overemphasizing any single factor when robust, population-specific evidence for differential weighting is not available (Papathoma-Köhle et al., 2019). The approach included overlaying the layers to create a final risk layer. This process was carried out using the raster calculator in ArcMap, which uses map algebra expressions to create new images from input rasters. Essentially, map algebra is used to perform mathematical operations on raster data, enabling the creation of new layers based on the input values (Caner and Aydin, 2021).

Subsequently, the pixels of the final risk layer were normalized again to range from 0 to 1 using the fuzzy linear membership function to ensure consistency across all layers. Furthermore, the equal interval classification method was used in ArcMap to classify the final risk map, which divides the normalized risk values into equal-sized ranges. The classification thresholds were determined by segmenting the risk values into five equal intervals, similar to the approach used by Savić et al (2018) and Wang et al (2023). The categories defined were very low risk (0–0.2), low risk (0.2–0.4), moderate risk (0.4–0.6), high risk (0.6–0.8) and very high risk (0.8–1).

This study was designed as an indicator-based, spatial risk assessment within the IPCC framework, integrating hazard, exposure and vulnerability indicators to identify neighbourhood-scale hot spots (SA1) rather than as a retrospective epidemiological analysis of health outcomes. The scope intentionally extends beyond purely health indicators to include outdoor exposure and urban form factors (e.g. canopy, distance to GAs, travel method) to reflect where and how people encounter heat in the built environment. In the absence of publicly available, ethics-cleared SA1-level outcome data sets (e.g. admissions, emergency visits and mortality), and given our use of aggregated, non-personal data, direct outcome-based validation was out of scope for this manuscript. Instead, we prioritized a parsimonious, empirically supported indicator set to avoid unnecessary model complexity while preserving interpretability and policy utility.

A comparison of LST and maximum air temperature from seven Bureau of Meteorology weather stations revealed that in 87% of cases where T-air was 35°C or higher, LST reached 40°C or above (Figure 4). Motivated by this empirical evidence, a threshold of LST greater than or equal to 40°C was proposed as a reliable indicator for identifying instances where the air temperature reaches or exceeds the critical threshold of 35°C. The LST map (Figure 5–1) shows that the study area is predominantly characterized by high LST classes, forming a pronounced heat hazard area that includes the entire coastal plain, running parallel to the coastline.

The resulting heat risk map showed that large areas of Perth experience elevated LSTs on hot summer days, creating a widespread heat risk zone (Figure 5-1). The land use map of the study area was first classified into six classes: built, woodlands, farmland, shrub/grass, water and others (Figure 5-2). Built areas include mainly residential, commercial and industrial uses, while the “others” category encompasses bare or cleared land and mine sites around the city. Comparing land use and LST maps allows for an understanding of temperature variations across various land uses in the study area. Figure 6 shows that more than 90% of the areas classified as “built”, “farmland” and “shrub/grass” fall under the “high” LST class. Interestingly, more than 50% of the areas classified as woodlands are also in this class. The bush/trees in the northern region mainly fall into the high LST class, while they fall into the lowest LST in the eastern part of the study area.

To explore the reasons for the observed disparity in LST of “woodland” cover located in the northern and eastern parts of the study area, the NDVI was examined (Figure 5-3), finding that vegetation with high LST mainly contains sparse vegetation. Notably, part of the east and southeast of the study area, where dense vegetation is dominant, also has the lowest LST. Figure 7 shows that more than 90% of areas with sparse vegetation have high or very high LST. However, only 29% of areas covered by dense trees exhibit high LST. Figures 5–4 and 8 also show that LST is lower in the areas with a higher canopy percentage. The combined indices into a final heat exposure component (Figure 9) demonstrate that the residential areas, especially the regions that do not have a high canopy coverage, have a higher heat risk.

The vulnerability index map, generated by overlaying multiple indicators (Figure 10), reveals that highly vulnerable populations are predominantly concentrated on the outskirts of the residential areas in Perth, where urban and rural characteristics often intermingle. These peripheral zones exhibit higher levels of vulnerability compared to the central urban areas. For example, the proportion of children under five years old is higher, particularly in the northern outskirts. Other vulnerability indicators follow a similar spatial pattern, though with some variations. In the southern and central parts of the residential zones, significant pockets of populations over 65 years old, with low incomes and health problems are also observed. As illustrated in the final vulnerability index map, areas where individuals are affected by more than two high vulnerability indicators exhibit the highest risk of vulnerability. These highly vulnerable SA1s are primarily located in the eastern and northernmost parts of the study area.

The final risk layer revealed that the distribution of overall heat risk at the SA1 level varies across Greater Perth (Figure 11), with high-risk areas distributed sporadically throughout the metropolitan region. The map highlights more visible high-risk classes (more than 0.6) in residential areas, and by moving towards the outskirts of the residential area, the areas with higher risk increase.

For a comprehensive risk management approach in high-risk areas, it is essential to examine all indicators in these areas individually. Figure 12 supports this purpose by illustrating the locations of all indicators with high-value classes (0.6–1). This map reveals that demographic, social, economic and health indicators do not consistently exhibit high values across all high-risk areas. For example, regions with higher populations of individuals aged over 65 and low-income households are distributed throughout the study area, while areas with a high population of children under 5 and longer average travel times are more sparsely located. In contrast, high-value physical and biological indicators are observed in numerous SA1s.

Reviewing both local and national approaches to heatwave risk assessment, such as the Perth metropolitan heatwave impact map and the national heatwave service, reveals that not all criteria have been considered to address heat risks for vulnerable populations. As a result, the latest heat-related risk assessment framework from the IPCC was used in this study to comprehensively analyse risk components and their implications for urban planning and policy interventions. This methodology was selected because it allows for a multi-dimensional analysis of heatwave risk, accommodating complex interactions between diverse socio-economic, environmental and health-related factors. Using spatially explicit data and GIS techniques ensures detailed identification of vulnerable areas, essential for effective urban heat risk management. Beyond the methodological innovations, the research findings have substantial policy implications, providing critical insights for targeted urban planning and heat risk mitigation strategies. Nevertheless, this study acknowledges several limitations that should be addressed in future research to enhance the robustness and applicability of the findings.

Recent research has supported a 35°C threshold for air temperature (Faisal et al., 2021; Loughnan et al., 2012; Mohammad et al., 2022; Williams et al., 2012). However, LST serves as a practical alternative in heat hazard assessments, as existing meteorological stations cannot fully cover the study area (Faisal et al., 2021; Mohammad et al., 2022). The findings of this study challenge the assumption that a threshold of 35 degrees for LST alone serves as a comprehensive indicator of extreme heat conditions. This uncertainty was addressed by integrating LST with in situ air temperature measurements from meteorological stations. This approach enables researchers to calibrate LST-based models using real-world temperature observations, improving the overall reliability of heat risk assessments.

Given the crucial role of vegetation in urban climate regulation, canopy coverage and proximity to GAs were incorporated as key exposure indicators in this study. Trees mitigate urban heat through shading and evapotranspiration, reducing both surface and air temperatures while enhancing thermal comfort (Geneletti and Zardo, 2016; Yenneti et al., 2020). By blocking direct solar radiation, trees limit heat absorption by surfaces, while evapotranspiration cools the surrounding air by dissipating thermal energy and increasing humidity (Chang et al., 2007; Galalizadeh et al., 2024; Zhang et al., 2022). GAs further extend the cooling effect beyond their immediate surroundings, playing a critical role in urban heat mitigation (Balist et al., 2022; Galalizadeh et al., 2024; Zahoor et al., 2022). This effect is evident in Greater Perth, where the northern and northeastern regions, characterized by lower canopy coverage, exhibit higher temperatures compared to the eastern region, which has greater vegetation cover. The effectiveness of GAs in reducing urban heat can be further enhanced when combined with other mitigation strategies such as high-albedo and permeable surfaces, cool roofs, water features and thermally efficient building designs (Bouketta, 2023; Lee et al., 2023).

The vulnerability indicators were selected based on their direct association with susceptibility to heat-related illnesses. Demographic indicators, such as age groups over 65 and under 5, were included because of their heightened physiological sensitivity to extreme temperatures (Wu et al., 2022). Socioeconomic status, particularly low income, was considered a key determinant of adaptive capacity, as lower-income individuals often lack access to cooling infrastructure (Pereira Barboza et al., 2023). Health conditions, including cardiovascular, respiratory and metabolic disorders, were integrated as they exacerbate heat stress responses, increasing hospitalization and mortality risks during heatwaves (Mason et al., 2022).

Various commuting methods can influence individuals’ exposure to different heat levels. For instance, those who opt for walking or cycling instead of using public transportation are more exposed to heat. Conversely, individuals who walk to a bus station and wait for a few minutes there are more susceptible to heat effects than those who use a private car for travel (Karner et al., 2015). Therefore, the travel method index was, to the best of the authors’ knowledge, for the first time, incorporated into heat risk assessment in this study.

By including the indices such as distance to gas, canopy coverage and travel method, the study expanded the scope beyond assessing indoor heat risks, as factors pertinent to outdoor activities, such as walking and biking, were considered. By considering both indoor and outdoor heat exposure, the assessment provides a comprehensive understanding of heat-related risks across diverse settings.

Reducing uncertainties in policy and planning can be achieved by demonstrating the significance of various indicators in the risk assessment process. For this purpose, the spatial distribution of physical and biological indicators was investigated to understand how effective each of them can be in the risk assessment process. These findings have immediate policy relevance as they complement the Western Australian State Hazard Plan for Heatwave and the Department of Health’s Heatwave Impact Strategy at the regional scale. At the national level, they align with Australia’s National Climate Resilience and Adaptation Strategy, while internationally they address calls from the World Health Organization and the IPCC for locally tailored heat health action plans. In doing so, the results are directly connected to both domestic and global adaptation agendas.

The spatial distribution of LST classes, particularly the formation of a distinctive heat risk area in the entire coastal plain, offers valuable insights for risk assessment and urban planning. This information informs decision makers about areas prone to extreme heat, guiding the development of strategies to enhance resilience and sustainability. A crucial implication of this study is that urban planning authorities should integrate LST-based risk assessments into broader city planning initiatives, ensuring that infrastructure development and zoning regulations proactively mitigate heat exposure. This includes implementing tree canopy enhancement programmes, increasing permeable surfaces and prioritizing green space development in high-risk areas. Furthermore, the study’s spatial risk analysis underscores the necessity for policy interventions that consider not only climatic factors but also socio-economic vulnerabilities to create equitable climate adaptation strategies.

The increased vulnerability observed in the outskirts of the residential area, compared to central urban areas, can be attributed to a combination of socio-economic and demographic factors. These disparities include lower income levels and a higher proportion of vulnerable groups, such as the elderly or individuals with pre-existing health conditions, as demonstrated in this research. Lower socio-economic and health status are well-established factors contributing to increased vulnerability, as individuals facing health challenges with limited resources may encounter difficulties in adapting to extreme events (Eugenio Pappalardo et al., 2023; Venter et al., 2020). This underscores the necessity of embedding climate adaptation policies into social equity initiatives, ensuring that infrastructure investments and heat mitigation measures are distributed equitably among economically disadvantaged communities.

By meticulously analysing various indicators, the research enables stakeholders, including urban planners and local authorities, to identify high-risk heat neighbourhoods accurately. This targeted approach facilitates effective intervention strategies in place-based risk planning initiatives. The provision of detailed evidence, potentially in the form of a comprehensive data set, elucidates the specific risk levels associated with each indicator in every area. Armed with this information, authorities can strategically implement preventive, control and mitigation measures tailored to the unique risk profiles of different neighbourhoods.

By embedding these spatial insights into existing policy frameworks, including local municipal heatwave plans, state-level hazard management strategies and national climate adaptation programmes, the study directly addressed the research questions. The highest heat-risk SA1 neighbourhoods were identified on Perth’s urban fringe, and within these areas, socio-demographic vulnerabilities emerged as the dominant drivers of risk. Furthermore, as hypothesized, neighbourhoods farther from green spaces with lower canopy cover experienced higher exposure and composite heat health risk.

The final heat risk map shows that hot conditions are widespread across the coastal plain. High-risk SA1 areas are more common near the urban fringe, where socio-demographic vulnerability is greatest. Councils with SA1s that overlap with these hotspots should be prioritized for targeted interventions.

• The City of Wanneroo and the City of Swan, located on the northern and eastern fringe, should focus on increasing street-tree canopy along residential streets and school routes. They should also install shade and drinking water at bus stops, and trial cool pavements and reflective roofs near local centres. Outreach efforts should focus on elderly and low-income areas identified in the vulnerability map.

• The City of Armadale and the City of Kalamunda, in the southeast and eastern fringe, should connect small parks with canopy corridors and support home retrofits that improve heat safety. These include insulation and reflective roofing. Energy cost support for cooling should be provided in vulnerable areas, and cooling centres should be located near aged-care and health service clusters.

• The City of Kwinana and the City of Rockingham, in the southern coastal plain, should expand shade in school zones and employment areas with high surface temperatures. Neighbourhood centres should include permeable surfaces and water features to reduce heat.

• The City of Joondalup and the City of Stirling, in the northern coastal plain, should increase tree canopy street by street and improve cooling around parks. This is especially important in residential areas with limited shade and longer outdoor travel distances.

To apply this approach at the council level, the used indicator-based approach can be used to identify priority neighbourhoods. After this, block-level audits should be conducted to assess tree canopy, surface cover, transit stops and the presence of aged-care facilities and schools. Project types should then be matched to local needs. In areas with high heat exposure, councils should focus on increasing tree canopy, improving surface permeability and adding shade at key travel points. In areas with high vulnerability, physical cooling measures should be combined with health and energy support programmes, such as subsidized air conditioning and assistance with operating costs.

These locations and measures are important because the coastal plain experiences high LSTs on hot days. Residential areas show more high-risk zones, especially towards the urban fringe. Northern and northeastern areas with less tree cover tend to be hotter than the more vegetated eastern areas, making tree planting and reflective or permeable surfaces especially effective. Vulnerable SA1s are concentrated along the eastern and northern edges, so combining cooling infrastructure with social support is essential for a meaningful impact.

These actions can be included in existing local heat strategies. Councils can align them with the State Heatwave Plan and Department of Health programmes. This allows the indicator results to be translated into specific project lists, such as priority streets, centres and facilities within high-risk SA1s.

In spite of the robustness of the spatial analysis used in this study, certain limitations in data sources must be acknowledged. First, using remotely sensed LST as a proxy for air temperature may bring potential biases, because LST does not always correspond exactly to near-surface air temperatures. LST measurement can be affected by variations in land cover, emissivity and atmospheric conditions, leading to overestimating or underestimating heat exposure in some areas. To mitigate this issue, in situ temperature measurements from meteorological stations were integrated, improving the reliability of heat hazard estimations. While these meteorological stations were strategically selected across urban and rural zones, a network of only seven points may not capture all fine-scale UHI variations. This introduces some uncertainty, suggesting that a denser array of observations or higher resolution modelling would further improve the robustness of the hazard calibration.

Although the land use map provided a clearer representation of cover types, limitations remain in differentiating specific vegetation structures because of the spatial resolution of the satellite imagery used. While Landsat imagery effectively distinguishes broad vegetation categories, it lacks the resolution necessary to separate trees from shrubs and shrubs from grasses at a finer scale. Furthermore, although NDVI has been widely recognized as a key heat risk indicator in numerous studies (García et al., 2023; Pereira Barboza et al., 2023; Rao, Singh, et al., 2021; Vancutsem et al., 2010), its application in heat risk assessment requires reconsideration. This may be because of NDVI being a measure of vegetation cover and reflectance, which is insufficient in explaining changes in LST, particularly in housing areas (Güller and Toy, 2024). Consequently, features in the actual world with similar reflectance properties can be grouped into the same class or category (Aryal et al., 2022). It indicates that using this indicator has two primary limitations: the precision of the generated data and the subsequent derivatives. Thus, relying solely on NDVI may not offer accurate insights into landscapes and their coverage.

Another critical limitation relates to the socio-economic and demographic data sourced from the Australian Bureau of Statistics (ABS) at the SA1 scale. Using SA1-level data may overlook important local differences. For example, a single street segment with high heat exposure or a cluster of heat-sensitive residents, such as those in aged-care facilities, may exist within an SA1 that appears to have only moderate risk. In practice, SA1 hotspots should be used as a first step to identify priority neighbourhoods and councils. These maps should then be combined with detailed local audits of streets, facilities and housing types, along with high-resolution data such as canopy cover and surface materials, before selecting specific interventions. It is not recommended to target or assess interventions at the household or property level using the SA1 indicator-based approach alone. Where possible, future updates should include more detailed data, such as mesh blocks, property records or facility registers, to reduce bias from data aggregation and improve decisions about where to place shade, cooling centres or home energy support.

While SA1 data provide valuable insights into population vulnerability, they may not capture intra-area disparities, particularly in rapidly developing or highly diverse communities. Aggregation bias can lead to overgeneralization, where variations in heat risk within an SA1 are overlooked. Furthermore, the socio-economic and demographic indicators were derived from the most recent census data available at the time of study. Given that such data are updated infrequently (every five years), there is an inherent lag in reflecting current population dynamics. Future assessments should incorporate more up-to-date or real-time data sources where possible to improve the timeliness and accuracy of vulnerability estimates.

Additional indicators such as housing quality, insulation, roof materials and access to cooling systems may influence indoor heat exposure and a household’s ability to adapt. However, detailed and reliable data sets for these factors are not publicly available at the SA1 level for Greater Perth. Existing proxies, such as housing tenure or dwelling type, are limited and closely linked to socio-economic indicators already included in the index. Future assessments should aim to include these housing- and cooling-related indicators, ideally using expert input or data-driven weighting methods.

Comparing mapped heat risk with actual health outcomes, such as hospital visits or heat-related deaths, would improve confidence in the index. These health data are usually protected for privacy and not available at the SA1 level. Future work could explore partial validation using aggregated and de-identified health data at broader scales, such as SA2 or local government areas. This would help assess whether high-risk SA1 clusters align with observed health impacts and support integration with health-focused planning tools used by the Western Australia Department of Health.

Although equal weights are a reasonable starting point, different weighting methods can change how areas are ranked. Sensitivity testing, such as adjusting weights within realistic ranges, can show how stable hotspot patterns are. Future work will compare equal-weight results with expert-informed methods and multi-criteria decision tools such as the analytic hierarchy process. Where possible, these alternative weightings will be checked against independent health data. These steps follow best-practice guidance and respond directly to reviewer feedback.

Recent studies such as Lui et al. (2023) and Kumar and Mishra (2025) have increasingly used advanced modelling techniques such as machine learning and agent-based models to capture complex, nonlinear interactions and behavioural responses to heat-health risks. These approaches offer enhanced predictive capabilities and can simulate dynamic scenarios under varying conditions. However, they often require granular behavioural or micro-climatic data and may obscure causal relationships because of their complexity. In contrast, the indicator-based method adopted in this study prioritizes interpretability and transferability, making it more suitable for practical applications in policy and urban planning, especially where data availability is limited.

By conducting a fine-grained spatial analysis of heat hazards, exposures and vulnerabilities, this study revealed pronounced heat-risk disparities across Greater Perth; notably, peripheral communities face substantially higher heat health risks because of elevated exposure and vulnerability.

By using novel indicators and integrating demographic, socio-economic and environmental dimensions, the study expands the assessment’s scope and provides insights into the differential impact of heatwaves on vulnerable populations. Using remote sensing data and GIS techniques, the research methods offer a rigorous analysis of heatwave risks, contributing to the identification of high-risk areas and the development of tailored adaptation strategies. Additionally, the research significantly advances understanding in the field of vulnerability assessment and directly informs regional policy frameworks such as the Western Australian Heatwave Management Plan and Australia’s National Climate Resilience and Adaptation Strategy, while contributing to global discussions on climate adaptation and sustainable development. Overall, the study highlights the importance of considering both demographic and biophysical indicators in heat risk assessment, thereby enhancing the effectiveness of heatwave management approaches. However, to conclude this research, a few initiatives that would contribute to refining and enhancing spatial heat health risk assessment were determined.

In synthesizing these findings, the study provides a novel contribution to the heat-risk literature by demonstrating the value of integrating diverse indicators within a unified assessment framework. This advance not only deepens scientific understanding of urban heat vulnerability but also offers practical guidance for policymakers, enabling more targeted heat mitigation interventions in the most at-risk communities.

The selection of indicators for this study was guided by an emphasis on health and socio-economic variables deemed most critical in the literature for assessing heat risk, particularly in urban environments. Only variables with strong empirical support and direct relevance to the study objectives were prioritized, ensuring analytical clarity and avoiding unnecessary complexity in the model. However, relying on a single indicator may not always provide an accurate representation of an individual’s situation. For example, income alone does not fully reflect one’s financial capacity. A person with a substantial income may still lack the means to address heat-related challenges because of additional expenses, such as medical costs or family obligations. Future research could expand this framework by incorporating additional health and socio-economic indicators to assess their relative influence on heat risk.

Given the dynamic nature of climate-related impacts, ongoing research could explore including additional biophysical indicators or modifying existing ones to capture the complexities of heat risk better. For example, including cooling indices provided by GAs, such as cooling distance, cooling area, cooling gradient and cooling effect duration, can significantly improve heat health risk assessment by enhancing the ability to identify hotspots. These inform urban planners about the spatial distribution and characteristics of cooling zones and help them develop targeted public health interventions based on a more precise understanding of where and when cooling effects are most critical for vulnerable populations. In this way, the assessment contributes more effectively to the development of strategies for enhancing community resilience to extreme heat events.

The indicator framework can be extended to create separate heat-risk maps for specific priority groups. These include older adults, people with chronic health conditions, low-income households and outdoor workers. For health-care professionals, these maps can support early outreach to high-risk neighbourhoods through targeted communication, heat-health education and planning for medication and hydration. They can also help with operational planning during heat alerts, such as adjusting staffing, preparing ambulance services, extending clinic hours and coordinating welfare checks. After heat events, the maps can support evaluation by comparing emergency department visits or call-outs with areas identified as high risk.

For city planners, subgroup maps can guide infrastructure planning in areas where vulnerable residents are concentrated. This includes increasing tree canopy, adding shade and water at bus stops and school routes and using reflective or cool materials in local centres. They can also help target programmes such as home retrofits for heat safety and energy cost support in low-income or elderly areas. Cooling centres can be located near aged-care and health service clusters. In addition, planners can use these maps to guide development approvals and maintenance, such as requiring shade and permeable surfaces in high-risk zones and prioritizing irrigation and tree care along key walking routes. These subgroup-specific maps help turn broad screening tools into practical action plans for health services and local governments. However, as noted earlier, SA1-level data should be used for neighbourhood-level planning only. Local site checks are needed before making decisions about specific locations.

Although key indicators of heat risk assessment were incorporated in this study, equal weights were applied when overlaying these indicators. While previous studies have assigned weights to certain indicators, their focus has predominantly been on health and socio-economic factors, leaving a critical gap in understanding the contributions of other dimensions to heat risk assessment. Given the diverse nature of the indicators spanning biophysical, socio-economic, demographic and health domains, a comprehensive survey is essential to accurately determine the relative weight of each. Therefore, future research should prioritize developing a holistic weighting framework considering expert judgements and contextual knowledge. This could lead to a more refined and accurate understanding of the factors contributing to heat-related risks, helping communities better prepare for and respond to extreme heat events.