Global mean temperature has increased by 0.87°C from the preindustrial period to the present day (Jia et al., 2019), and it is expected to continue to increase, in total 2.0–4.9°C by 2,100 (Raftery et al., 2017). The temperature increase leads to more frequent and intense extreme heat events which cause adverse heat-related impacts that will also result in human fatalities and economic damages. Heat stress has been identified as one of the direct and negative impacts of climate change which can increase morbidity and mortality, and reduce work productivity worldwide (Kjellstrom et al., 2009; García-Herrera et al., 2010; Singh et al., 2015). A previous study has reported that a 0.5-°C increase of the global mean temperature results in a global 38% increase of heat-related exposure (Sun et al., 2019). Furthermore, under a 1.5-°C global warming scenario and the consequent additional heat stress, the world is expected to experience a US$2,400 billion loss and 2.2% reduction of total working hours by 2030, with agriculture and construction as the most affected sectors (Kjellstrom et al., 2019).

In terms of human health, urban areas are considered to be at higher risk due to the so-called urban heat island (UHI) effect, which means that urbanized areas tend to have a higher temperature than their rural surroundings (Loughnan et al., 2012). A recent study estimated that urbanization contributes to about 30–50% of the heat stress index increase during the summer season (Luo and Lau, 2018; Wang et al., 2019). Moreover, considering higher density of population groups that are more likely to suffer from heatstroke (i.e., elderly, impoverished people, and outdoor workers), the UHI effect makes urban areas more vulnerable during heat waves. High temperature may also have adverse effects on sleep quality and appetite (Zheng et al., 2019). At the same time, these adverse effects contribute to depression, heart disease, type II diabetes, stunting, and dementia (Rifkin et al., 2018). Thus, in order to prevent population’s health deterioration, identifying adaptation strategies that allow quality sleep is essential (Rifkin et al., 2018). Since the global urban population is projected to increase by 68% by 2050 (UN DESA, 2019), the urgency of tackling the heat-related risks in urban areas will become even a higher priority than at present.

Anticipating the effects of climate and urban demographic changes, many cities have been developing adaptation strategies toward a higher resilience (Boeckmann and Rohn, 2014). The adaptation to climate change usually includes hazard assessment, as well as infrastructure and population vulnerability components, in order to develop plans to alleviate the adverse effects of changing climate (IPCC, 2014). Common adaptation options against prolonged heat waves include electric fan utilization, active monitoring programs, and warning systems (e.g., heat stroke–related educational campaigns and information distribution campaigns through mass media). However, there are less explored measures that involve infrastructural changes such as urban greening or air conditioner installations (Boeckmann and Rohn, 2014). Some adaptation strategies against heat stress may become a part of the climate feedback loop. For example, cooling down residences or working spaces can cause an increase in household consumption of fossil fuel–generated energy, which results in higher CO₂ emission into the atmosphere and consequently increasing the greenhouse effect. (Arifwidodo et al., 2019). On the other hand, urban greening and heat wave warnings are examples of the so-called low-regret strategies which are characterized by a relatively low cost and provide relatively large benefits under the predicted future climates (Boeckmann and Rohn, 2014).

The Republic of Indonesia is an example of a rapidly developing country that is at risk of seeing large numbers of heatstroke cases caused by the UHI effect since 67% of its population is expected to be living in cities by 2035 (Jones, 2010). In addition, observations show that the number of hot days and warm nights has an increasing trend between the years 1960 and 2003 (Alexander et al., 2006). In a survey conducted among Bandung residents, almost 50% of the respondents indicated that they have experienced difficulties during their commuting or at work due to hot weather (Arifwidodo et al., 2019). Those difficulties and other similar situations are likely to become more intensive according to the climate change projections for this region. The output of climate models that corresponds to the representative concentration pathway (RCP) that projects the largest increase of global mean temperature by the end of the 21st century (i.e., the RCP 8.5 scenario) shows projections in Indonesia with three heat waves in the period from 2020 to 2052, which can get as frequent as every two years during the period 2068–2,100 (Russo et al., 2014). As the result of the climate change, the number of heat-related mortality for elderly people in the Jakarta metropolitan area will increase by 12∼15-fold in the 2050s under the RCP 8.5 scenarios (Varquez et al., 2020). Climate change may also force the urban population to resettle in the country (Zander et al., 2019).

As presented above, the climatic and demographic changes are practically unavoidable, and their combined effect is likely going to pose a major stress to urban populations and growing economies. Thus, the decision-making process involved in heat-related adaptation planning requires a detailed impact assessment with consideration of the future climatic scenarios (i.e., RCPs), demographic changes, and urban spatial distribution.

In this study, we aimed to evaluate the magnitude and distribution of heat in relation to human health in three cities in Indonesia, namely, Surabaya, Medan, and Denpasar, at present and for the period 2018–2042 under climate change scenarios, and to determine the role of urban greening in the urban heat reduction.

To assess the impact of heat strain on human health, we utilized the Universal Thermal Climate Index (UTCI), which is regarded as one of the most comprehensive indices to determine the changes in human thermal perception in outdoor spaces as stated by European COST (Cooperation in Science and Technical Development) Action 730 (Blazejczyk et al., 2012). The UTCI index is exceptionally useful in evaluating the physiological response to extreme heat events due to the fact that the input data required to compute the index include not only meteorological variables but also non-meteorological variables (e.g., metabolic rate). The index is expressed in units of temperature, and its classification is divided into several stress levels, allowing its application in different climate types, seasons, and scales (Blazejczyk et al., 2012; Mcgregor, 2012; Pantavou et al., 2018). Even though the UTCI has been used in studies conducted in China (Ge et al., 2016), Poland (Milewski, 2013), Iran (Nassiri et al., 2017; Zare et al., 2018), Canada, and Republic Korea (Park et al., 2014) to evaluate temperature stresses, there is no relevant report of its application in tropical regions.

The main purpose of the study was to investigate the spatial distribution of heat hazard through the applied UTCI index in selected cities, under projections of land-cover change and various climate change scenarios. In the current study, the spatial pattern of heat hazard varied depending on the location, indicating highest degree within central districts and lower degree in the suburbs. Higher degrees of heat hazard in the central districts clearly indicated that the UHI phenomenon was taking place. These results are the indication of UHI effect, where higher levels of heat hazard were found in the main urban center. The large contributors to the heat radiation are built-up and paved areas (Lamarino et al., 2011), and these types of area categories were found in this study to have a higher ratio among land use types (Supplementary Figure S1).

Four levels of the heat hazard have been identified in the three cities, namely, no thermal stress, moderate thermal stress, strong thermal stress, and very strong thermal stress (Figure 5). Very strong thermal stress was recorded as the highest level of UTCI in Medan and Surabaya and strong thermal stress in Denpasar during the baseline period (Figures 5A,C,E). However, mean values indicated that Surabaya has the highest spatial average of UTCI (33.2 ± 4.4°C), followed by Denpasar (30.5 ± 2.7°C) and Medan (29.4 ± 4.5°C). UTCI values classified as very strong and strong heat stress ranges can be observed mainly in the downtown areas, revealing the UHI effect (Lamarino et al., 2011). The result was consistent with build-up and bare soil land cover patterns. These areas have higher UTCI degrees than other land cover types, especially the vegetation land cover.

Future trends in the three cities are connected with the land cover change, demonstrating linkages between rapid land conversion into built-up areas (Figure 6) and the increase of UTCI degree (Figures 5B,D,F). A strong increase of heat hazard in the near future is predicted to occur, driven by urban development especially in the cities of Medan and Denpasar (Figures 5B,D). This can be attributed to the urban surface having lower albedo, higher surface roughness, and lower surface wind speed, which cause an increase in heating effect in the area. Also, anthropogenic heat released by the building contributes to the increasing heating effect. In contrast, Surabaya presented very different spatial patterns of heat hazard, and it has a decreasing trend of UTCI in the near future under LC scenario (Figure 5F). The change of heat hazard from no thermal stress into moderate and strong thermal stress was observed (Figure 5). In the future projection, we determined that the area with the strong heat stress will expand in both cities, especially in Denpasar, where the very strong thermal stress range was observed under the LC scenario. On the other hand, the bare land and built-up areas in Surabaya decreased significantly in southwestern part of the city and partially in the central part, reducing the average UTCI degree for 0.84°C in the future (Supplementary Figure S1). The bare land in Surabaya was converted into green park in the southwestern and central part of the city, from 1995 to 2016 (Figures 2, 7). The increasing vegetation coverage was a factor in decreasing the urban heat stress. In future, it is projected that Surabaya will reach 26.3% of vegetation coverage, which will contribute to reducing the heat hazard level in the city (Figure 6C, 7). These findings are consistent with analysis conducted in previous studies, stating that vegetation is able to lower the air temperature by 0.47°C and lower the thermal environment for 10°C (Wu et al., 2016). The positive mechanisms of vegetation coverage are primarily attributed to its shading effects, reduced absorbed radiation quantity by the surface, and slowering the intensity of the evapotranspiration process. These mechanisms are effectively cooling the air temperature (Mcgregor, 2012). In terms of shading, trees with dense canopy are able to enhance the outdoor thermal comfort and to reduce areas of very strong heat stress (Wu et al., 2016). Open green spaces covered by grass do not provide the same role as trees with dense canopy, but they have the ability to lower the near surface temperature than bare land/build-up areas (Wu et al., 2016), and they can be used as a recreational place during evening time with lower temperature and Sun radiation.

Adding climate scenarios for future projection significantly increases the heat hazard status in all three cities (Figures 8, 9). The result stated that three selected cities will enlarge the area of the strong and very strong heat stress status under climate change scenario. The area of very strong thermal stress will cover 0.5, 10, and 22% of Denpasar, Medan, and Surabaya, respectively, under the ensemble mean CC scenario (Figure 10), whereas under the extreme climate scenario, the coverage of the very strong heat stress will be 24.3, 28, and 67% of Denpasar, Medan, and Surabaya cities, respectively. Higher thermal stress (i.e., the strong and very strong heat stress) will be observed in central parts of these cities (Figures 8A–D, 9A–D). Consistently, our results showed that the coverage area of no thermal stress status decreased significantly in three cities in 2030. Under the CC scenario, the area of very strong heat stress expansion in Denpasar can be explained by the rapidly growing tourism industry of Bali Island where the city is located. On the other hand, in order to attract and protect tourists, there is a need for urban heat reduction in the future. Opposite to Denpasar which is a smaller city but with growing heat hazard, Surabaya is a larger city but having the opposite trend of heat hazard under the LC scenario (Figures 5E,F, 8E,F, 9E,F, 10).

This study also addressed a number of methodological challenges and advanced heat evaluation. The selected CMIP5 model itself does not count the LC change in their future projection; thus, this study included it as one of important variables. However, conducting the future LC scenarios for each specific city requires clear images of time series satellite data and raw master plan geographical data. Thus, conducting future projection of urban thermal stress for both the LC and the CC scenarios allows producing science-based guidelines on adaptation strategies, setting clear targets on urban greening in regional master plans. Identifying UTCI stress categories can also help to define appropriate preventive measures against the heat stroke (Table 2). We recommend setting such measures according to current and projected UTCI-based heat stress levels, for adaptation strategies.

Urban greening is one of the low-regret actions that provide relatively large benefits with lower cost, under predicted future climate conditions. It can also provide other regulating, provision, and cultural ecosystem services, such as air purification, flood control, recreational space, CO₂ sequestration, or food production (Farugia et al., 2013; Sioen et al., 2017; Helen and Gasparatos, 2020). Thus, considering urban greening should be a part of any development plans. In Indonesia, national regulation of urban greening follows the Law No. 26 of 2007 which states that at least 30% of urban area shall be allocated for open green spaces. This law also applies to provincial/district spatial planning (known as RTRW). The open green spaces include conservation areas, city forests, parks and green belts, reservoir areas, and commentaries(Surabaya Municipal Government, 2016).

In RTRW 2014–2034 master plan of Surabaya city, the municipal government has set a quantitative target for open green spaces, stating that the open green spaces must cover about one-third of the total urban area, of which 20% must be public space and 10% a private space. This policy was aligned with national regulation, the Law No. 26 of 2007, and efficiently implemented and enforced with the ratio of public open green space of 20.75% in 2015 (Surabaya Municipal Government, 2016). However, to achieve the national target, the private sector engagement is necessary. Surabaya municipal government has started a program called “Green Building Initiative” that provides tax reduction incentives for building meeting criteria for green building (Surabaya Municipal Government, 2016). Despite the current development planning, Surabaya municipal government did not include the “climate change adaptation,” per se however addressed the urban heat stress through urban greening, which can be defined as one of adaptation strategies at the local level.

In the case of Denpasar, the current ratio of open green space (including agricultural land) in Denpasar was estimated to be 36% (Denpasar Government, 2016). Bali provincial government has committed that additional 5% of urban area will be allocated for open green space on the Bali Island (including the city of Denpasar) (BAPPEDA Bali, 2019). This decision was made to control tourism-related urban sprawl and to reduce urban heat through various stakeholders. In this study, we demonstrated that adding regulatory measures to increase green spaces in Denpasar is necessary, as the heat stress is estimated to be very strong under the future projection.

Among the three studied cities, Medan has the lowest ratio of urban green spaces (10% as of 2015) (according to Medan Spatial Planning Agency, 2015), and this is lagging behind the national target set by the Law No. 26 of 2007. Furthermore, almost 90% of residential roof tops are made of aluminum and cement sheets that drastically decrease the urban albedo (BPS Medan, 2018). A local Act No. 522.4/1,553. X/IX 2013 that requires greening enacted in Medan was enacted, but it has not been translated into clear technical guidelines, resulting with no proper implementation.

Urban climate adaptation measures have not yet been embedded into both national and regional policies and planning documents including Jakarta as a benchmark of mega cities in Indonesia. Thus, urban climate adaptation is experiencing a difficult path to find its operationalization and implementation, leaving it as a responsibility of provincial governments or city authorities. Therefore, we expect this assessment to be beneficial in the context of local-level risk management for urban heat hazard. This is a necessary step to support local adaptation planning to heat-related stress and reducing impact on human health.

The UTCI-based urban heat stress assessment performed in this study under climate change and land-cover change scenarios in three cities, namely, Medan, Surabaya, and Denpasar, included in the National Adaptation Plan of Indonesian government demonstrated that under future projections, Medan is the most vulnerable city in terms of increasing the UTCI level, followed by Denpasar and Surabaya. These results demonstrated that the highest level of UTCI identified in the three cities was classified as very strong thermal stress due to the contribution of climate change and urban development. However, this can vary, depending on the city, climate, and land use planning. Under the land-cover scenario, we found that urban greening in Surabaya city was a factor reducing the area of high heat stress level. Visualizing the spatial distribution of UTCI as one of human-centric assessments of outdoor thermal comfort can facilitate planning the adaptive measures deployment, especially to areas of strong and very strong thermal risk. Urban greening, green building incentives, rising awareness of the heat-related risk, and health protective measures are among the most important adaptation measures that can be more efficiently implemented after UTCI analysis. These approaches also give the opportunity to study about the linkage among climate condition, urban structures, and human physiological response. Thus, delivering to policy-makers and to regional and local governments, the science-based evidence on the actual and projected heat risk can support national, regional, and local planning necessary adaptation strategies. It can protect human health and reduce economic damages under climate change. However, our assessment also had limitations on the sample number of collected satellite images in three pilot sites, and this could have been a source of uncertainties in the results. Thus, considering a slightly higher cloud covering percentage threshold of Landsat satellite images or using the daily Level 3 LST product from the moderate resolution imaging spectroradiometer (MODIS) could be an option to increase the sample number of satellite images. Moreover, other spatial data distributions are needed for other potentially relevant indicators of the UTCI as the hazard index.