Two decades of nighttime surface urban heat island intensity analysis over nine major populated cities of India and implications for heat stress

Warmer global climate and urban heat islands (UHIs) interact, by exacerbating heatwaves and increasing the extreme heat days in cities. The implications of added heat stress in urban environments due to intensifying surface UHIs (SUHIs) is of utmost concern. Seasonal, annual and decadal nighttime SUHI intensities (SUHIIs), from 2001 to 2020, for nine major populated cities of India are analyzed. This includes five megacities- Delhi, Mumbai, Kolkata, Bangalore, and Chennai, and four incipient megacities- Hyderabad, Ahmedabad, Surat, and Pune. The key role of increasing urbanization (pre- and post-2010) in expansion and intensification of nighttime SUHIs in India is highlighted. For all cities either pre-monsoon (MAM) or winter (December-February; DJF) seasons show the strongest SUHII development. During the 2001–2010, and the 2011–2020 decade, a nighttime SUHII maxima of respectively (i) 2.1°C and 2.5°C for Delhi, (ii) 1.3°C and 1.5°C for Mumbai, (iii) 1.3°C and 1.5°C for Kolkata, (iv) 0.6°C and 1.0°C Bangalore, (v) 1.7°C and 1.9°C for Chennai, (vi) 1.8°C and 2.3°C for Hyderabad, (vii) 2.8°C and 3.1°C for Ahmedabad, (viii) 1.9°C and 2.4°C for Surat, and (ix) 0.8°C and 1.3°C for Pune is noted. Further, all incipient megacities showed a mean annual growth rate of nighttime SUHII of over 0.007°C/year, substantially greater than in the megacities. High SUHII magnitudes, greater growth rates of SUHII, and huge populations, severely compounds the vulnerability of Indian cities to excessive heat exposure risk, especially during MAM heatwaves. Lastly, the implications of nighttime SUHII findings from the present study, on the increase in heat stress, the loss of labor productivity and the rise in heat-related mortality rate is emphasized. The study recommends implementation of city-specific action plans to mitigate the heat stressed urban environment. Targeted use of cooling strategies in localized hotspots within the urban areas where high intensity SUHIs are likely to form is also suggested.


. Introduction
The global population crossed 8 billion in the end of the year 2022 and according to the latest United Nations (UN) publication on the "World Population Prospects, " the global population is expected to increase by another half a billion by 2030 (United Nations (UN), 2022). In 2022, just two countries-India and China accounted for over 2.8 billion people, and in 2023, India with a projected population of over 1.6 billion, is expected to overthrow China to become the most populous country in the world (United Nations (UN), 2022). At present, India has five megacities Jain .
(i.e., cities with population >10 million) viz., Delhi (28.5 million), Mumbai (19.9 million), Kolkata (14.6 million), Bangalore (11.4 million), and Chennai (10.4 million) (United Nations (UN), 2018a). In addition to these five megacities, India also has four rapidly growing incipient megacities viz., Hyderabad (9.4 million), Ahmedabad (7.6 million), Surat (6.5 million), and Pune (6.2 million). Studies point that recently turned megacities-Bangalore and Chennai, followed the urbanization patterns of older Indian megacities, such as, Delhi and Mumbai remarkably well (Taubenböck et al., 2009(Taubenböck et al., , 2012Jain et al., 2016;Ghosh and Das, 2017). Similar trends in urban growth are also expected to be followed by the other incipient megacities of India, throughout the next decade (Taubenböck et al., 2010;United Nations (UN), 2018a). The sheer enormous magnitude of population in just these nine Indian cities-Delhi, Mumbai, Kolkata, Bangalore, Chennai, Hyderabad, Ahmedabad, Surat, and Pune, can be judged from the fact that they inhabited a total of 1.1 billion people in 2018. This roughly amounted to one-seventh of the world's entire population, and further these nine Indian cities are projected to inhabit over 1.5 billion people by the year 2030 (United Nations (UN), 2018a).  Jain, 2022). In the two decades from 1999 to 2018, Eckstein et al. (2019) found records of over 12,000 extreme event occurrences globally, which resulted in the direct loss of lives of about 500,000 people, and financial damages amounting to approximately USD 3.5 trillion.
Studies have found a significant increase in frequency, severity and intensity of climate extremes e.g., droughts, heatwaves, wildfires, hurricanes, and floods; which likely will exacerbate life and property losses in the future, along with causing food shortages and an increase in disease causing pathogens (Robinson, 2001;Black et al., 2004;Li et al., 2009;Dai, 2011;Intergovernmental Panel on Climate Change (IPCC), 2012Mishra et al., 2017;Hawcroft et al., 2018;Bisht et al., 2019;Hasegawa et al., 2021;Jain, 2021aJain, , 2022Jain et al., 2021;Climate Transparency, 2022). Unlike the other climate extreme events, heatwave caused fatalities are not immediately obvious, but lead the weather-related cause of deaths globally (Hajat and Kosatky, 2010;Peterson et al., 2013;World Health Organization (WHO), 2020). Heatwave occurrences are of particular interest in cities, as the urbanized (or built-up) areas in cities are, typically, warmer by 1-4 • C than their nearby rural surroundings, a phenomenon extensively documented as the urban heat island (UHI) effect (Oke, 1973;Voogt and Oke, 2003;Parker, 2010;Peng et al., 2012;Nuruzzaman, 2015;Shastri et al., 2017;Zhou and Chen, 2018;Anjos et al., 2020). In 2015, an excess of 175 million people, globally, suffered from heatwave exposure compared to the average number of people exposed to heatwaves during previous years (World Health Organization (WHO), 2020). Moreover, the 95 th percentile global heatwave exposure is projected (RCP8.5-SSP3 scenario) to be 18-37 times more during 2071-2100 than it was century ago   (Liu et al., 2017). Both, inland Indian cities, such as, Delhi, Ahmedabad, and Hyderabad, and coastal Indian cities, such as, Mumbai, Kolkata, and Chennai are severely impacted by heatwaves. The infamous 2015 Indian subcontinent heatwave killed over 2,500 and 7,00 people in India and Pakistan, respectively (Wehner et al., 2016). Moreover, Mishra et al. (2017) stresses that a 2 • C increase in global temperatures will cause a tremendous increase in occurrence of severe heatwaves, such as the 2015 Indian subcontinent heatwave, by almost 15 times by the next 30 years and by an astounding 92 times by the end of this century. Against the backdrop of continuously rising global temperatures and increasing extreme heat events, the heat stress in cities is further compounded by the spatial expansion and increasing intensity of UHIs, especially during the night (Intergovernmental Panel on Climate Change (IPCC), Jain, 2021bJain, , 2022Shreevastava et al., 2021).
Traditionally, the UHI intensity (UHII) is computed as the difference in the 2m air temperature, measured at two separate weather stations, each located at a fair distance from each other, such that they can distinctly be characterized as "urban" or "rural, " and yet fall within the same geographical and climatic set-up (Oke, 1973(Oke, , 1997Arnfield, 2003;Voogt and Oke, 2003). However, with the availability of freely accessible remote sensing and satellite data, it has become easier for urban meteorologists and urban climate scientists to study both, the daytime and the nighttime UHIs across the world (Peng et al., 2012;Zhou et al., 2014;Shastri et al., 2017;Liu et al., 2021;Siddiqui et al., 2021). Since satellite based UHI estimations use land surface temperature (LST) instead of 2 m air temperature, the term surface UHI (or SUHI) is used instead of UHI. A major benefit of SUHI over UHI is that satellite based LST data is two-dimensional (gridded data) while weather station based 2 m air temperature is onedimensional (point data). Thus, SUHI assessments come in handy when questions overlooking-(1) mapping of SUHI footprint and spatial expansion of SUHI over time in a city, (2) identification of microclimate SUHI hotspots within the urban area, (3) seasonal, annual, and decadal changes in the intensity of SUHI, and (4) comparative SUHI analysis (devoid of instrument and methodology biases) with other cities arise. In most studies, it has been found that the SUHII is strongest during clear nights, during days of calm or no wind, and in the summer season (Oke, 1981(Oke, , 1997Arnfield, 2003;Voogt and Oke, 2003;Parker, 2010;Peng et al., 2012;Nuruzzaman, 2015;Santamouris, 2016;Anjos et al., 2020). Nighttime LST is more sensitive toward urbanization led thermal heating than daytime LST and is preferred for SUHI estimations (Siddiqui et al., 2021).
A fairly long-term record of over two decades of Moderate Resolution Imaging Spectroradiometer (MODIS) nighttime LST data has come into fruition recently. From a climate science perspective, a long-term LST record is crucial for analyzing changing trends in SUHII for a city, along with its larger impact on urban climate, heat stress and heat-related mortality. At a global scale Peng et al. (2012) used 5-year MODIS LST data to assess seasonal and diurnal SUHII for 419 big cities, while Chakraborty and Lee (2019) used 15 years of MODIS LST to calculate diurnal, seasonal, and annual SUHII for more than 9500 urban clusters. Through satellite LST observations (2002)(2003)(2004)(2005)(2006)(2007)(2008)(2009)(2010)(2011)(2012)(2013)(2014)(2015)(2016)(2017)(2018)(2019)(2020)(2021), Liu et al. (2022) found mean surface warming trend per decade for over 2000 city clusters in the world, of which the highest was noted in Asian megacities, heavily contributed due to rapid urbanization. Variability in seasonal and diurnal UHI, for 101 Asian and Australian cities, was explored by Santamouris (2015), and the relationship between city size and UHII was analyzed. Using a regional climate model and MODIS LST data on cities in the US, Sarangi et al. (2021) noted how Jain .
/frsc. . the UHI effect dominates the diurnal variations in urban heat stress intensity, is at peak during the nighttime, and that the heat stress exposure during nighttime intensified by 3-5 hours per day. Raj et al. (2020) found increasing nighttime SUHIIs in 44 major cities of India using satellite temperature data from 2000 to 2017. Kumar et al. (2017) in their study over 89 Indian cities highlight that nighttime SUHI is strongly correlated to the percentage of imperviousness in the urban areas. Shastri et al. (2017) calculated daytime and nighttime SUHII during two seasons-pre-monsoon (MAM) and winter (DJF), for 84 urban locations in India, using MODIS LST data from 2003 to 2013. However, the study was based on city clustering algorithm, which relies on population estimates, rather than impervious area estimates, to delineate urban and rural boundaries; the latter being a better criterion, especially in a populous country like India. In this regard, the present paper is the first to assess 20 years (2001-2020) of nighttime SUHIs, during all four seasons-winter (December-February; DJF), premonsoon (March-May; MAM), monsoon (June-August; JJA), and post-monsoon (September-November; SON), over all the nine major populated cities of India i.e., Delhi, Mumbai, Kolkata, Bangalore, Chennai, Hyderabad, Ahmedabad, Surat, and Pune. Moreover, the study provides an insight into the SUHII variability and trends at a decadal, annual, and at a seasonal time scale for the nine Indian cities. The paper also discusses the far-reaching implications of nighttime SUHI expansion and the increasing severity of SUHII, in worsening the heat stress and discomfort of city dwellers of India. Ramifications of increasing SUHIIs during pre-monsoon (or summer; MAM season) in synergy with increasing global temperatures and increasing heatwaves on the human mortality rate are also discussed. Lastly, a call to action via targeted SUHI mitigation strategies and development of city-specific action plans focused to reduce heat stress are suggested.

. . Selection of cities
Past (2000), contemporary (2018), and future (2030) population estimates (United Nations (UN), 2018a) were used to identify nine major populated cities of India. In 2018, five cities, viz., Delhi (28.5 million), Mumbai (19.9 million), Kolkata (14.6 million), Bangalore (11.4 million), and Chennai (10.4 million) had a population of over 10 million, while four cities, viz., Hyderabad (9.4 million), Ahmedabad (7.6 million), Surat (6.5 million), and Pune (6.2 million) had a population of over 5 million. The population statistics, including the annual rate of population change between 2000-18 and 2018-2030 for these cities are provided in Table 1. It is estimated that by 2030, Hyderabad and Ahmedabad would likely grow to become megacities, while the population of two cities-Surat and Pune will likely increase to 9.7 million and 8.4 million, respectively, almost within reach of the megacity status (Table 1 The two main reasons for choosing these particular nine Indian cities are-(1) their massive population size (>5 million), which inherently increases their vulnerability to loss of lives in the wake of extreme events or climate disasters, and (2) their status as centers of national, financial, technological, scientific, and cultural importance, which inherently increases their vulnerability to a greater loss of gross domestic product (GDP) in such events.

. . Delineation of urban-rural extents
Global Man-made Impervious Surface (GMIS) dataset available for the year 2010, at 1 km resolution , was accessed from the Socioeconomic Data and Applications Center (SEDAC) (https://sedac.ciesin.columbia.edu/data/set/ulandsat-gmis-v1). The GMIS 2010 dataset is derived from Global Land Survey (GLS) Landsat image archive and maps subpixel urban imperviousness using the "Human Built-up And Settlement Extent" (HBASE) approach. The GMIS dataset limitations include possibility of incorrect removal of small areas of urban pixels by HBASE, and some pixel misclassification to "NoData" category where cloud/ shadow areas and Landsat 7 ETM+ SLC off image data gaps exist (Brown de Colstoun et al., 2017). Even with these limitations, the GMIS and Jain .
/frsc. .  For each city shown, the upper panels represent nighttime LST for the decade, and the lower panels represent nighttime LST for the -decade. SUHI formation during the night in the cities is clearly discernable in regions of greater impervious area percentage. Impervious area for the year (Source: Wang et al., ) is classified into four categories-( ) -% as "rural", ( ) -% as "peri-urban," ( ) -% as "low-density urban," and ( ) -% as "high-density urban" in the present study. The figures are generated with ArcMap ver. . . HBASE dataset for the year 2010 was found to have good (0.91) kappa coefficient over Europe (Zhang et al., 2020), and is considered a valuable global impervious area dataset for many urban based research studies (Yoo et al., 2019;Cho et al., 2020;Melancon et al., 2021;Shen and Zhao, 2021).
At the same time, Kim and Brown (2021) highlight that there still remains a general lack of protocol for classification criteria required when estimating SUHI. This is due to the fact that there is no clear international definition of what is considered as "urban, " and the administrative boundary of a city may not accurately represent the Jain .
/frsc. . urban extent (especially due to creation of new satellite towns beyond the administrative limits of the city). However, many studies suggest that the use of impervious area percentage is a good delineator of urban extent and can be used to efficiently quantify SUHI (Li et al., 2018). Within the GMIS 2010 dataset, the class values range from 0 to 100, each representing the percentage of impervious area within a 1 km grid, and is used as a primary dataset in the present study for delineating the urban and rural extents. The GMIS 2010 dataset was qualitatively cross-referenced for urban area extents with the Bhuvan state-wise land use land cover dataset (2011-12; https://bhuvan-app1.nrsc.gov.in/thematic/ thematic/index.php) and with the Google Earth imageries for the nine Indian cities for the year 2010. A review of ancillary information such as city maps, master plans and available town planning documents were also performed so that a good reference for urban and rural extents could be established. Based on these careful qualitative evaluations on the urbanization patterns in India, the GMIS data over the entire Indian landmass was classified into four categories-(1) all grids having impervious area < 10% as "rural, " (2) grids having impervious area between 10 and 24% as "peri-urban, " (3) grids having impervious area between 25 and 49% as "low-density urban, " and (4) grids having impervious area > 50% as "high-density urban." Furthermore, it was established that for all the selected nine Indian cities, a 100 km bounding box (region of interest) was sufficiently (1) large enough to contain the entirety of the "high-density urban, " "low-density urban, " and "peri-urban" areas, (2) large enough to sufficiently include the surrounding "rural" areas for SUHII estimation, and at the same time, and (3) small enough to limit the topography related local weather changes as well as a temperature bias in SUHII. The centroid of each of the 100 km bounding box was superimposed over the city center points. The city center point vector data was prepared in ArcGIS ver.10.6, from the existing geographical coordinates (latitude and longitude) information for each city (Figure 1). The location of the bounding box for each city was determined by fulfilling the above conditions, whilst also ensuring that they do not overlap with nearby cities. In case of coastal cities, the area beyond the Indian landmass i.e., the Arabian Sea or the Bay of Bengal, was not considered in the SUHII estimations.

. . Estimation of SUHII
Quality checked monthly night-time MODIS LST product at 0.05 • spatial resolution (MOD11C3v006; https://giovanni.gsfc.nasa. gov/giovanni/) was downloaded for the years 2001-2020 over the Indian landmass. For each year, the data was averaged seasonally for DJF, MAM, JJA, and SON. Seasonal night LST for each year was subset (clipped) for the nine city domains (100 km bounding boxes) and further for the four urban-rural classes viz., "rural, " "peri-urban, " "low-density urban, " and "high-density urban" within each city as elaborated in Section 2.2.
In order to estimate SUHII, the LST of pixels assigned class as "high-density urban" were subtracted from the mean LST in pixels assigned class as "rural." Classes "peri-urban" and "low-density urban" were disregarded in their entirety for the computation of SUHII. Both these classes have a mixed proportion of urban-rural elements in varying degrees; an urban bias in the "low-density urban" class and a rural bias in the "peri-urban" class, and therefore the inclusion of these two "buffer zone" classes would have corrupted the true SUHII estimates (Voogt and Oke, 2003).
. Results and discussion . . Spatio-temporal nighttime SUHI footprint Land-air interactions play a major role in urban climate, especially in governing urban heating (Jain et al., 2017;Jain, 2021bJain, , 2022. Further, nighttime LST is found to be a better indicator of SUHII than daytime LST due to its higher sensitivity toward urbanization (Siddiqui et al., 2021). Figures 2-4 shows the seasonal nighttime LST for the 2001-2010 and 2011-2020 decades, over the nine major populated cities of India. The formation of a distinct nighttime SUHI in regions of greater impervious area percentage is clearly discernable from this figure. Formation of nighttime SUHI is found to relate remarkably well to the city size, and is primarily evident over the "high density urban" class i.e., pixels having over 50% of impervious area.
Of all the nine Indian cities, Delhi, which is both-the largest populated city of India, and the largest urbanized city (area under "high density urban" class), shows the largest nighttime SUHI footprint (Figure 2A On the other hand, incipient megacities with a greater city size, such as, Hyderabad ( Figure 3C) and Ahmedabad ( Figure 4A), show a substantially larger nighttime SUHI footprint, than a small-sized incipient megacity, such as, Surat ( Figure 4B). SUHI is found to weaken over the "peri-urban" and "low-density urban" classes. This is most prominently observed in Pune ( Figure 4C), an incipient megacity with a city size large enough to be comparable to some of the megacities, and yet most of its area falling in either the "peri-urban" or the "low-density urban" class. The most reasonable explanation for the observed SUHI weakening over these mixed urban-rural classes is the greater abundance of thermal cooling classes e.g., forests and waterbodies than in the "high density urban" areas (Arnfield, 2003;Anjos et al., 2020;Jain, 2021b). It should be noted that certain "rural" class land use land covers, such as, barren lands, rocky terrain, and uncultivated agricultural lands can, at times, emit higher a proportion of thermal radiation during the night, and as a result show an even higher LST than the city (Jain et al., 2017). Few isolated rural areas nearby to the cities of Delhi, Kolkata, and Hyderabad do show higher nighttime LST in the present study (Figures 2, 3). Such scenarios are heavily dependent on the intrinsic thermal properties e.g., specific heat capacity, thermal conductivity, and emissivity (Chen et al., 2006;Xian and Crane, 2006;Bowler et al., 2010;Hua et al., 2017) of the materials found in such areas, which lead to a higher nighttime LST.
For all the cities in the present study a seasonal variation in the footprint of nighttime SUHI is noted. From Figures 2-4, the most intense formation of nighttime SUHI is noted during pre-monsoon (MAM) and winter (DJF) season. Similar results of a seasonal maxima during MAM and DJF have been highlighted in a previous nighttime  SUHII study over India by Shastri et al. (2017). For most cites, nighttime SUHI during the post-monsoon (SON) season is not observed to be as intense as either during DJF or during MAM season. Further, all cities show that the nighttime SUHI formation is weakest during the monsoon season i.e., JJA. The monsoon season is when the cities of India experience maximum precipitation, along with strong convective winds and lush vegetation; all factors that mitigate SUHI formation. Based on the mean seasonal SUHI for the two decades, Ahmedabad is observed as the only city to show the existence of an urban cool island, evident during the JJA season of 2001-2010 ( Figure 4A). The significant impact of increasing urbanization (preand post-2010 scenarios) on the nighttime SUHI is also evident from Figures 2-4. Compared to the 2001-2010 decade, the nighttime SUHI size during the 2011-2020 decade is found to have significantly grown for all the nine major cities of India. This nighttime SUHI expansion is observed during all the seasons, however, most strongly evident during MAM. The pattern of radial outward growth of nighttime SUHI is consistent with the patterns of urban sprawl witnessed over these Indian cities (Taubenböck et al., 2009(Taubenböck et al., , 2012Jain et al., 2016;Ghosh and Das, 2017).

. . Seasonal, annual, and decadal nighttime SUHII trends
The characterization of SUHII variability and trends is of foremost importance in order to understand the processes related to heat stress in humans (Anjos et al., 2020). The season of the highest intensity of nighttime SUHI, for each city, in each decade, is highlighted in the present study. For all the nine Indian cities, either DJF or MAM season shows the prevalence of the strongest nighttime SUHII. During the 2001-2010 and the 2011-2020 decades, a nighttime SUHII magnitude of respectively (i) 2.1 • C and 2.5 • C (in MAM) for Delhi, (ii) 1.3 • C and 1.5 • C (in DJF) for Mumbai, (iii) 1.3 • C and 1.5 • C (in DJF) for Kolkata, (iv) 0.6 • C and 1.0 • C (for both DJF and MAM) for Bangalore, (v) 1.7 • C and 1.9 • C (in DJF) for Chennai, (vi) 1.8 • C and 2.3 • C (in DJF) for Hyderabad, (vii) 2.8 • C and 3.1 • C (in DJF) for Ahmedabad, (viii) 1.9 • C and 2.4 • C (in DJF) for Surat, and (ix) 0.8 • C and 1.3 • C (in MAM) for Pune (Table 2) is found to be the strongest.
The mean annual nighttime SUHII from 2003 to 2008 over majority (95%) of the 419 big (population > 1 million) cities of the  world was found to range between 0 and 2 • C (Peng et al., 2012). In a separate study over India, the MAM and DJF nighttime SUHII, from 2003 to 2013, for 84 urban locations was noted to range between 0-1 • C and 0-1.5 • C, respectively (Shastri et al., 2017). Figure 5 depicts the 20-year annual trend of the nighttime SUHII for all the nine Indian cities. For each year, the mean nighttime SUHII for the seasons DJF, MAM, JJA and SON are shown. The present study finds that the mean decadal nighttime SUHII from 2001 to 2020 in most of the nine (population > 5 million) cities of India ranged between 0-2.5 • C (Table 2) (Table 2). Interestingly, a substantially higher decadal increase of 0.2-0.6 • C in DJF, 0.4-0.5 • C in MAM, 0.4-0.7 • C in JJA, and 0.2-0.5 • C in SON, is noted in the four incipient megacities of India. The mean annual growth rate of nighttime SUHII from 2001 to 2020 was found to be (i) 0.0057 • C/year and for Delhi, (ii) 0.0026 • C/year for Mumbai, (iii) 0.0042 • C/year for Kolkata, (iv) 0.0069 • C/year for Bangalore, (v) 0.0047 • C/year for Chennai, (vi) 0.0093 • C/year for Hyderabad, (vii) 0.0073 • C/year for Ahmedabad, (viii) 0.0076 • C/year for Surat, and (ix) 0.0105 • C/year for Pune ( Figure 5). It is interesting to note that every incipient megacity had a mean annual growth rate of nighttime SUHII of over 0.007 • C/year, while all megacities remained below this threshold rate. One reason for this larger SUHII growth rate in the incipient megacities could be due to the faster urban growth rate reported in these cities (Taubenböck et al., 2009). This could lead to a significantly higher urban heating rate, a higher urban LST, a stronger SUHII, and significant increase in the heat stress.

. . Implications for heat stress in Indian cities
A warmer global climate, and UHIs in the cities, interact by exacerbating heatwaves and increasing extreme heat days during the summer season (Kim and Brown, 2021;Jain, 2022). In cities, the extremely high day or night air temperature due to this synergistic effect can directly cause an increase in the heat-related morbidity and mortality rates (Santamouris, 2016;Kumar et al., 2017;Kim and Brown, 2021) Indian cities in particular have been found to be more affected by heatwaves, than the rural areas due to the UHI (or SUHI) development, and in addition, the heatwaves during MAM have also increased significantly in India (Shastri et al., 2017). Some of the most notable severe heatwave events in India have occurred in 2002 (April-May), 2015 (May-June), 2016 (April-May), 2019 (May-June), and 2022 (March-May), where the maximum recorded temperature during each heatwave event ranged from 49 to 51 • C. Delhi and Ahmedabad are two prominent cities in northern India that face severely harsh MAM season, with daytime (maximum) and nighttime (minimum) air temperatures often crossing 40 • C and 25 • C, respectively. In such a case, a 2.1-2.8 • C and a 2.5-3.1 • C mean nighttime SUHII during MAM, observed in the last two decades, in Delhi and Ahmedabad, respectively (Table 2) is extremely concerning; more so during the period of severe heat events in summer. Similarly concerning is the nighttime SUHII maxima of over 1.5 • C witnessed in other Indian cities, such as, Chennai, Hyderabad, and Surat. This compounds the heat stress faced in cities, especially during episodes of severe heatwaves during summers, and leads to higher incidences of heat-related fatalities, such as, deaths due to heat strokes.
As per a recent report by Climate Transparency (2022), at a warming of 1.5 • C, about 142 million more Indians, each year, are projected to be exposed to heatwaves than the average exposure in 1986-2006, and about 2.8 times greater exposure at a warming of Jain .
/frsc. .   Figure 6A). Along with the air temperature anomalies, the reported (annual) number of heat stroke deaths (solid bars; secondary y-axis; Figure 6A) in India in these 50 years are also shown. The earlier 25-year period from 1971 to 1995, mostly shows the prevalence of a negative air temperature anomaly for both, the minimum, and the maximum air temperature. However, the latter period from 1996-2020, generally consistent with the onset of rapid urbanization in Indian cities (Taubenböck et al., 2009(Taubenböck et al., , 2012 shows notably higher minimum and maximum air temperature anomalies (+0.5 to +1.0 • C; Figure 6A). Moreover, in the figure, black arrows are used to mark the occurrence of severe heatwave events (e.g., 1995, 1998, 2003, 2010, 2012, 2015, 2016, and 2019) in India in the recent years.
A congruence of increasingly positive air temperature (minimum and maximum) anomalies, with the increasing occurrences of severe heatwave events and increased heat stroke deaths (reported) in India is highlighted. As a caution, is important to note, that the actual number of heat stroke deaths, and other heat-related morbidities in India, is difficult to estimate, even for individual cities. Moreover, is widely accepted that the heat stroke deaths are much (even many times) higher than the official numbers reported in the National Crime Records Bureau (National Crime Records Bureau (NCRB), 2022) Accidental Deaths & Suicides in India (ADSI) yearly reports, and the official heat-related morbidities data is entirely unavailable (Hajat and Kosatky, 2010;Tran et al., 2013;Azhar et al., 2014;The Times of India, 2015;Swain et al., 2019). Even still, a significant sharp increase in the heat stroke deaths (annually reported) in India is observed, when the annual mean air temperature anomaly (with respect to 1980-2010 baseline) of +0.2 • C and above, occurs over the Indian landmass ( Figure 6B). The impact of heat-stress, including, excess deaths due to heat strokes, would be compounded in the Indian megacities and incipient megacities, wherein, the present study shows the existence of high mean nighttime SUHIIs (up to 3.1 • C), high increases in mean decadal nighttime SUHIIs (up to 0.4 • C), and high mean annual growth rates of nighttime SUHIIs (up to 0.0105 • C/year) in the period from 2001 to 2020. Furthermore, studies show that for every 1 • C increase in ambient temperature, the mortality rate has been observed to increase by 1-3%, and above a critical heat threshold the mortality rates rise exponentially (Hajat and Kosatky, 2010). The mortality rate in the Mediterranean increased by 3.12% per 1 • C as the ambient temperature increased above the heat threshold (Santamouris, 2016). In case of Delhi (India), the mortality rate during a 1991-1994 study increased by 3.94% per 1 • C increase in ambient temperature above 29 • C (Hajat and Kosatky, 2010). In 2018 and 2019, India observed the largest increase in heat related mortalities in the world (Romanello et al., 2021). Moreover, in 2021 an estimated loss of 167 billion potential labor hours and USD 159 billion in GDP due to extreme heat caused labor capacity reduction was reported (Climate Transparency, 2022). Increasingly warmer decades, increasing mean annual growth rate of SUHIIs, and increasing frequency of prolonged severe heatwaves, will continue to enlarge the gap between the ambient air temperature and the heat threshold temperature megacities like Delhi and incipient megacities like Ahmedabad. This puts tens of millions of city-dwellers in megacities and incipient megacities of India at a grave risk of heatrelated morbidity and mortality in the near future.

. Conclusion and recommendations
From a heat stress perspective, the triple whammy of unabated globally rising temperatures, increasing UHIIs (or SUHIIs), and increasingly frequent prolonged heatwaves of extreme severity during pre-monsoon (MAM) season is of utmost concern in cities of India (World Health Organization (WHO), 2020; India Meteorological Department (IMD), 2021a; Jain, 2021bJain, , 2022Jain et al., 2021). A number of factors such as urban area (e.g., impervious surface area, built-up density, building heights, etc.), city morphology (e.g., sky view factor, canyon width, building material properties, presence of thermal cooling classes, etc.), meteorological conditions (e.g., wind speed, wind direction, boundary layer height, precipitation, etc.), anthropogenic heat fluxes (vehicular emissions, air-conditioning heat output, industrial emissions, human metabolism, etc.) affect the ambient temperature, SUHI formation, and ultimately heat stress in cities. In the nine highest populated cities of India, the key role of increasing urbanization (pre-and post-2010) in nighttime SUHI expansion and intensification is highlighted in the present study. Two seasons, MAM and DJF show the strongest prevalence of nighttime SUHII in Indian cities, a period coinciding with increasing number of severe heatwaves in India.
The study also highlights that the incipient megacities of India showed a substantially greater mean annual growth rate of nighttime SUHII than the megacities where this rate remained below 0.007 • C/year. In the last two decades, a mean nighttime SUHII of 2.1-2.8 • C was witnessed during MAM in Delhi, a megacity with 28.5 million people (and a projected population of 39 million in 2030) and a mean nighttime SUHII of 2.5-3.1 • C was witnessed during MAM in Ahmedabad, an incipient megacity showing the highest SUHII out of all Indian cities. Such situation is extremely concerning in the context of increasing heat stress, heat-related morbidity and mortality. An example is the 2010 (May) Ahmedabad heatwave which led to over 1,340 excess deaths (Azhar et al., 2014). A nighttime SUHII maxima of over 1.5 • C, witnessed in other Indian cities, such as, Chennai, Hyderabad, and Surat, is also similarly concerning. In general, the intensities of nighttime SUHIs witnessed around the year in all nine cities of India are substantially high to cause significantly lowered thermal comfort and quality of life of the people.
In order to reduce the heat-related mortality in cities of India, key high intensity SUHI hotspots need to be identified. These hotpots of excessive heating within the cities should be of utmost focus for implementing targeted strategies aiming to reduce heat stress. In a previous local-scale line transect study over Delhi, the presence of thermal cooling classes along densely built-up urban areas was found to significantly reduce LST (Jain et al., 2017). Studies show that a strategic increase in inclusion of blue-green infrastructure (i.e., "blue" infrastructure e.g., rivers, ponds, watersheds, tanks, other waterbodies, etc. and "green" infrastructure e.g., forests, parks, gardens, etc.) significantly reduces ambient temperature, lowers UHII (or SUHII), and mitigates heat stress (Nuruzzaman, 2015;Hua et al., 2017;Jain, 2021a;Sanusi and Jalil, 2021). Urban greening strategies such as vertical gardens, rooftop and terrace gardens, community forests, urban agriculture are aesthetically pleasant and also provide thermal comfort in heat-stressed urban environments (Anjos et al., 2020;Jain, 2021b). In addition, inclusion of bluegreen infrastructure also leads to greater energy savings due to lowered electricity power (e.g., air-conditioning) demands, while simultaneously decreasing the anthropogenic heat flux (Akbari et al., 1997). Other strategies such as cool roofs, cool pavements, green buildings, and water spray systems have also shown great potential in significantly reducing heat stress during summer extreme heat events (Fahed et al., 2020;Khare et al., 2021;Sanusi and Jalil, 2021).
It is recommended that all the major populated cities of India should formulate city-specific action plans incorporating the mitigation of UHIs (or SUHIs) and extreme heat events under a warmer future climate. Currently, Ahmedabad is the only city of India with a sophisticated and strategic action plan, called the "Ahmedabad Heat Action Plan 2019, " formulated in response to high mortality rates witnessed during frequently occurring extreme heat events. Targeted city-specific mitigation and action plans have a key role in combating the increasing heat stress faced in Indian cities. Policy focus should be on identification of localized hotspots of strongest UHI (or SUHI) formation within the cities and where the high intensity UHIIs (or SUHIIs) are likely to develop (Kim and Brown, 2021;Sanusi and Jalil, 2021). By taking urgent action through implementation of targeted cooling strategies and heat action plans,