Intensification and spatial shifts of heatwave hotspot across India under climate and ENSO influences with health risk assessment

Abstract

Heatwaves (HWs) in India have intensified significantly over the past two decades, with projections of further increase in frequency and duration. While global warming mitigation efforts are advancing, urgent local adaptation measures are required to minimize societal impacts. However, understanding regional climate risks has been limited. This study aims to address this gap by examining the intensification and spatial shifts of HW hotspots across India under the influence of climate change and the El Niño-Southern Oscillation (ENSO), along with a health risk assessment. A comprehensive archive of HWs using reanalysis-based daily maximum temperatures (Tmax) data for hot weather season has been constructed across India from 1981 to 2020. From 1981 to 2020, India’s average summer Tmax increased by 1.0 ± 0.12 °C, primarily driven by global warming, resulting in a higher frequency, duration, and extent of extreme HWs. Notably, the impact of El Niño has amplified this trend, with average summer Tmax increasing by 1.03 ± 0.44 °C, compared to a 0.77 ± 0.20 °C increase during Neutral ENSO conditions from the 1981–2000 period to the 2001–2020 period. A HW hotspot index (HHI) is introduced to identify the most vulnerable HW-prone regions, evaluated using satellite-based NDVI and LST data, indicating around 1.5 times increase in the spatial extent of HW hotspots between 1981–2000 and 2001–2020. Based on the HW health risk analysis, the central plains, the southeastern coast, parts of western India, and Gangetic West Bengal, along with certain areas in the central Indo-Gangetic Plain (IGP) region, were identified as the most at risk under the current scenario (2011–2020). This information enables informed decision-making and the implementation of measures for adaptation and mitigation.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-026-38289-x.

Introduction

Global temperatures have increased since the Industrial era, and a corresponding increase in the incidence of extreme heatwave (HW) events is occurring worldwide. Summertime episodes with extremely high surface air temperatures lasting for several days are often referred to as HW¹. The IPCC’s fourth assessment report, released in 2007, defines HWs as a continuous period of abnormally and uncomfortably hot weather², while the fifth assessment report, released in 2014, reported an increase in the number of hot days over the past decades³. The sixth assessment report⁴ of the IPCC, released in phases between 2021 and 2023, indicates that global warming has significantly increased the frequency, duration, and severity of HWs (and marine HWs) globally. This shows that over the past 15 years, the intensity of HWs has escalated from minimal recognition to an alarming level. This increase emphasizes the impacts of climate change on forest fires, ecosystems, agriculture, marine biota, snow cover, human and animal health^5–11. Projections indicate a concerning rise in intensity, frequency, and duration of HWs, which is expected to amplify their detrimental effects, especially in the most vulnerable regions^12–14. South Asia is among the most vulnerable regions to climate extremes, with many areas already experiencing extreme heat stress¹⁵. Therefore, it is essential to develop appropriate tools for identifying hazardous regions and associated risks to assess potential impacts and plan effective adaptation and mitigation efforts.

A review of HWs across South Asian countries reveals that the region has documented some of the warmest temperatures ever recorded¹⁶. Notably, from 2015 to 2025, each year ranked among the 10 warmest years on record, with the trend worsening yearly. India has the longest record of data in South Asia, and reports indicate that the frequency of hot days and hot nights has increased since 1960^17–19. From 1970 to 2019, India experienced over 706 HWs²⁰, with the number increasing from approximately 3 HWs per year in the 1970s to around 25 per year in the 2010s and 2020s. HWs in India have increased in frequency, total duration, and maximum duration²¹, with significant changes in their spatial extent compared to the base period 1950–1980^22,23. The frequency and intensity of HWs in India are projected to increase in the future²¹.

Currently, there is no universally accepted index for defining a HW^24,25; however, HWs are typically identified either based on absolute or percentile thresholds when the only temperature is used²⁴. The Indian Meteorological Department defines HWs based on climatological thresholds and deviation from these norms, which include absolute thresholds for both hilly and plain regions²⁶. However, several studies have employed percentile thresholds to define HWs over India^27–30. Additionally, various approaches have been reported in the literature that utilize both temperature-based indices and those that incorporate additional meteorological factors to quantify HWs^16,31–33. Recently, a “HW proneness index” was introduced to identify regions prone to HWs by integrating three HW attributes, namely the maximum annual HW magnitude, the mean HW magnitude and the annual frequency of HWs³⁴. The “HW Magnitude Index daily” (HWMId) integrates both intensity and duration. A “Heat-wave-Intensity-Duration-Frequency” (HWIDF) curve is employed to establish the relationship between the intensity, duration, and frequency of HW incidents³⁵. Each of these approaches has its advantages and limitations³³; however, they often overlook critical aspects of HWs, such as their changes in patterns over time. Since this study focuses on identifying HW hotspots, monitoring shifts in their patterns over time, and assessing health risk associated with HWs, we employed the HW Hotspot Index (HHI). This index integrates Mean Maximum Temperature Anomaly, the Number of Hot Days, and considers HW duration, intensity, and frequency. By doing so, we capture both deviations from climatology and the persistence of extreme heat. This allows us to illustrate the temporal evolution and spatial patterns of HWs, which is the primary objective of this study. To enhance the HHI’s reliability, we also compare this index with independent satellite-measured Land Surface Temperature (LST) and Normalized Difference Vegetation Index (NDVI) data.

Systematic efforts have been made to understand the primary mechanisms behind HWs in India. Key factors include large-scale atmospheric patterns, such as the persistent subtropical high and quasi-stationary Rossby waves in the mid-latitudes^10,36,37. Additionally, global warming has had a significant impact on the total number of HWs and HW days, resulting in an increase in their frequency, intensity, and duration³⁸. These patterns drive the summer season temperatures through land-atmosphere feedback³⁹, while factors such as soil moisture and clear skies further amplify their effects²⁷. Additionally, natural climate variability, including changes in large-scale climate patterns of the ocean and atmosphere, such as the El Niño-Southern Oscillation (ENSO), the Indian Ocean Dipole (IOD), and the North Atlantic Oscillation (NAO), can also contribute to the occurrence of HWs. HWs in India are linked to climate patterns⁴⁰, with the maximum number of HW days found to be preceded by warm ENSO years^41–44. Although previous studies have improved our understanding of the impacts of climate change and climate variability on HW characteristics over India, the trends of HW hotspots and their spatial shifts remain unexplored. This study addresses this gap by quantifying the intensity of HW hotspots and explicitly examining the changing patterns of these hotspots in India from 1981 to 2020. Understanding these patterns is crucial for enhancing our predictions about HWs and mitigating their potential impacts. Given the ongoing impacts of climate change, it is imperative to assess the risks of heat-related illnesses and heat stress, especially for vulnerable populations and in regions with limited access to cooling infrastructure. Therefore, this study investigates the health risks associated with heat exposure, which has significant implications for mitigating its detrimental effects. Overall, this study will significantly contribute to our understanding of HW patterns and their implications for public health.

Results

Evolution of HWs over the past 40 years

Spells of hot weather occasionally occur over certain parts of India during the pre-monsoon season. The time series of mean maximum air temperature (Tmax) for each hot weather season over India from 1981 to 2020 is presented in Fig. 1a, highlighting trends in Tmax across the decades. The average Tmax across India from 1981 to 2020 shows a significant warming trend of 0.26 ± 0.03 °C per decade at 90% confidence interval. Given the high mortality rates from HWs in recent decades^32,33, the warming trend in Tmax for the periods 1981–2000 and 2001–2020 is analyzed. To supplement, the last two decades (i.e., 2001–2020) have been the hottest recorded, with the previous year marking the warmest year ever documented. The warming trend remained consistent during these two periods (Fig. 1a). However, the Tmax increased by 0.5 °C (0.50 ± 0.18 °C) by 2000 and by 1.0 °C (1.04 ± 0.12 °C) by 2020 compared to 1981 levels, due to anthropogenic warming. India’s climate change assessment report indicates that the average temperature in India has increased by approximately 0.7 °C from 1901 to 2018⁴³. This suggests that Tmax is increasing more rapidly during the pre-monsoon season in India, resulting in the most intense HWs recorded in recent decades.

Fig. 1.

Observed changes in daily maximum temperatures averaged over the hot weather season from 1981 to 2020 time period over India. (a) Temporal variation of annual mean maximum temperature during the hot weather season from 1981 to 2020 over India. Yellow shading indicates the standard error. Trends are estimated at 90% significance level. Spatial distribution of trends in maximum temperatures for hot weather season during (b) 1981–2000 and (c) 2001–2020 over India. Hatches indicate trends at 90% significance level. These maps were generated in ArcGIS Pro 3.5.3 (https://www.esri.com/en-us/home ). The India shapefile is from https://onlinemaps.surveyofindia.gov.in/.

Fig. 5.

The maps showing heatwave hazard, exposure and vulnerability indices and composite Heatwave Health Risk index for India. (a) Heatwave hazard levels, (b) human exposure levels, (c) vulnerability levels, and (d) risk levels for the time period 2011–2020. These maps were generated in ArcGIS Pro 3.5.3 (https://www.esri.com/en-us/home). The India shapefile is from https://onlinemaps.surveyofindia.gov.in/.

The spatial distribution of trends in average Tmax during the hot weather season for the periods 1981–2000 and 2001–2020 over the Indian subcontinent is illustrated in Fig. 1b and c, respectively. Hatches in these figures indicate locations where trends are statistically significant at a 90% confidence level. These figures indicate that the warming trends in India are unevenly distributed, with variations between the two periods. Jammu and Kashmir are experiencing a significant increase in Tmax. However, the current Tmax values for the region are low, indicating that current conditions may not pose a serious health risk to humans. However, this could have a severe impact on fragile mountain ecosystems. Between 1981 and 2000 (Fig. 1b), two main regions experienced significant increase in Tmax. One region was in north-western and central India, while the other was along the coast of Andhra Pradesh. In these regions, the average warming trend reached up to 0.06 °C per year. Whereas during the period from 2001 to 2020 (Fig. 1c), this trend intensified, with a statistically significant warming of up to 0.08 °C per year, particularly in western, south-central, and peninsular India. This marks a sharp rise in regional Tmax compared to the 1981–2000 period. The increasing trend of Tmax in southern India is concerning, as these regions have become more vulnerable to HWs in recent years⁴⁴.

Overall, the spatial distribution of Tmax trends indicates that the regions from northwest to north-central and extending to the south-central parts of India are the hottest compared to the rest of the Indian subcontinent. As a result, the frequency and duration of HWs have also increased over the past four decades. Between 1981 and 2000, regions in northwest India, central India, and coastal Andhra Pradesh experienced an average of 2.5–5.5 HW days per year (see Supplementary Fig. S1). However, over the last two decades (2001–2020), the frequency of HWs has increased, with 3.5–8.5 days per year, particularly in northern, western, south-central, central-east, and peninsular India. Similarly, trends in HW duration indicate that from 1981 to 2000, central and eastern India, including coastal Andhra Pradesh, experienced prolonged HWs lasting 2.5-4.0 days per year (see Supplementary Fig. S1). In contrast, over the last two decades (i.e., 2001–2020), the duration of HWs increased to 3–5 days per year, with significant intensification observed in northern, western, south-central, central-eastern, and peninsular India. This increase in frequency and duration over the last two decades (2001–2020) indicates a trend towards more severe and prolonged HWs compared to the previous decades (1981–2000). To better understand these aspects, we examined the changes in HW parameters at a decadal scale from the 1981–2020 time period.

Observed trends in the spatial distribution of HW parameters on a decadal scale

The statistics for HW parameters, including frequency, duration, intensity, and anomalies in Tmax and the number of hot days, are presented in Fig. 2 on a decadal scale. This data reveals significant changes in HWs across India over the past 40 years.

Fig. 2.

Decadal changes in the Heatwave metrics from 1981–2020. The panels represent (a) average heatwave duration (days per year), (b) average HW frequency (days per year), (c) Average heatwave intensity (°C), (d) average maximum temperature (Tmax) anomaly (°C), and (e) Average number of warm days across different decades (1981–1990, 1991–2000, 2001–2010, and 2011–2020). These thematic layers are used for heatwave hotspot Index (HHI) estimation. These maps were generated in ArcGIS Pro 3.5.3 (https://www.esri.com/en-us/home ). The India shapefile is from https://onlinemaps.surveyofindia.gov.in/.

During 1981–1990: During this decade (Fig. 2, first row), the duration of HWs varied from 1–4 days per year, with the longest events occurring in central and northern India. The frequency of HWs was highest in the central, northern, and eastern regions, ranging from 1–6 days per year, particularly in central India and parts of northern and eastern India. The intensity of these HWs peaked in central and northern regions, with temperatures reaching as high as 47 °C. Anomalies in Tmax indicated significant deviations, especially in the extreme peninsular region. The highest number of hot days was recorded in northern and central India.

During 1991–2000: During this decade (Fig. 2, second row), the duration of HWs varied from 1–4 days per year, mainly affecting central, northern, and western India, as well as in newly affected areas of western India, the eastern coast, and peninsular India. HW frequency also increased, with central, northern, and western regions experiencing 4–6 HW days per year, while the eastern coast and peninsular India had 2–4 days per year. Additionally, HW intensity intensified, especially in northwestern and northern regions, where temperatures exceeded 47 °C. Anomalies in Tmax showed significant positive deviations in the Gangetic plains and central India, while other regions experienced negative deviations. The number of hot days increased in central and western India, expanding into northern regions.

During 2001–2010: This decade (Fig. 2, third row) indicated intense HWs, with increased spatial extent and longer durations. The duration of HWs has increased to 2–4 days per year across most regions, with the longest events occurring in eastern and central India, where they last 4–6 days annually. The frequency of HWs has also risen, especially in the eastern, eastern coastal, and peninsular regions, with some areas experiencing HWs for 6–8 days a year. Additionally, HW intensity has intensified, with Tmax exceeding 47 °C, particularly in northern, central, and peninsular India. There have been significant positive deviations in Tmax anomalies, especially in the northwestern, central, northern, and eastern regions, with anomalies reaching up to 0.6 °C in parts of north and northwest India. The number of hot days has increased in most areas, particularly expanding in the northwestern and adjacent central regions.

During 2011–2020: The recent decade (Fig. 2, fourth row) exhibits a significant intensification and wider occurrence of HW characteristics compared to previous decades, with broader spatial coverage across India, capturing attention and prompting consideration. The average duration of HW days has increased by 2–4 days per year, affecting most of the country. The longest HWs, now lasting 4–10 days annually, are concentrated in large parts of peninsular India, particularly in coastal Andhra Pradesh, where they can extend up to 8–10 days each year. The frequency of HW occurrences has sharply increased, with the eastern, northwestern-central, and peninsular regions reporting 8–10 days of HWs per year. Extreme northern India is also experiencing a higher frequency of these events. The intensity of HW incidents has reached unprecedented levels, with temperatures exceeding 47–50 °C in the northwestern and central regions; most of the country is now experiencing temperatures in the range of 44–50 °C. Anomalies in Tmax have shown significant deviations in parts of peninsular India, as well as in western and adjoining central and eastern regions, with increases of 0.8–1 °C. HW duration has also increased significantly in western and central India. HWs have started impacting regions that were previously less affected, including significant areas of the south-central and southern peninsula.

These analyses highlight the spatial patterns of HW origins and their spread during India’s hot weather season over the past four decades. From 1981 to 2020, there has been a noticeable increase in the area and regional spread of hot weather conditions, extending diagonally from northwest India to north and central India, with some influence observed in south India in the recent decade (Fig. 2). Most areas in and around northwest India are dry and semi-dry. This climatic condition increases the frequency of HWs in these regions, underscoring the connection between extreme heat and parched soils⁴⁵ and corroborating recent studies^46,47. Consequently, the frequency of HWs is higher in northwest India (desert region), compared to other parts of the country. However, in recent decades (i.e., from 2001 to 2020), HWs in northern and central India have become more intense but last for a shorter duration than those in southern India (Fig. 2). This increase in the severity of extreme heat events is likely driven by regional anthropogenic changes⁴⁸.

Intensity of HWs during ENSO events over the last four decades

There is clear evidence of a change in the patterns of HW metrics. Climate change, primarily driven by greenhouse gas emissions, has resulted in an increase in the intensity, duration, and frequency of extreme HWs. Additionally, natural climate variability, such as changes in large-scale atmospheric patterns like the ENSO, can also contribute to the occurrence of HWs. Previous studies have documented the influence of major climate modes, including ENSO, North Atlantic Oscillation (NAO), and Pacific Decadal Oscillation (PDO), on HW variability over India^28,49,50. But this study measures the extent to which El Niño contributes to the anthropogenic background warming during the periods of 1981–2000 and 2001–2020.

The ENSO primarily involves the interaction between the ocean and atmosphere in the tropical Pacific; however, it is known to significantly impact the global climate system by releasing heat into the atmosphere and altering atmospheric circulation patterns. In this study, we examine the Niño 3.4 index as an ENSO indicator to assess its impact on Tmax in India. The time series of mean Tmax and mean Niño 3.4 index for the hot weather season of each year from 1981 to 2020, depicted in Fig. 3a, reveals the dynamic interplay between natural climate variability associated with ENSO and its possible modulation of Tmax over India. Over the four decades from 1981 to 2020, there were instances where the Tmax distinctly increased during the El Niño years, and in some cases, the upward trend persisted following these preceding El Niño years. This highlights the intricate links between ENSO phenomena and regional temperature patterns.

Fig. 3.

ENSO modulation of maximum temperature (Tmax) variability and distributional shifts over India. (a) Interannual variability of the April–May Niño 3.4 index (Red bars: El Niño, Blue bars: La Niña, Grey bars: Neutral years) together with the corresponding annual maximum Tmax (black line) over India. Dashed horizontal lines indicate the El Niño and La Niña thresholds and between these dashed lines indicates the Neutral condition. (b–d) Probability density functions of Tmax during the hot weather season for (b) El Niño, (c) Neutral, and (d) La Niña conditions, shown separately for the periods 1981–2000 (magenta) and 2001–2020 (cyan). Legends denote the peak Tmax values for each period, illustrating a consistent rightward shift of Tmax distributions in the recent decades across all ENSO phases. These plots were generated in Python (Jupyter Notebook v3.12.3).

The analysis focused on three distinct categories: Warm, Neutral, and Cool phases of ENSO, which are classified based on the ENSO index (see methods section). The Probability Density Function (PDF) for Tmax versus frequency shows a significant increase in Tmax across the Indian region during the two critical periods for three categories, as illustrated in Fig. 3b–d. The data unveil the most likely temperatures and illustrate how the peaks and distributions shift (i.e., revealing extremes and skewed PDFs) during the varying phases of ENSO.

During the warm phase of the ENSO (known as El Niño), the peak Tmax value increased from 38.5 °C in the 1981–2000 period to around 40.0 °C in the 2001–2020 period, along with a positive skewness in the Tmax distribution. The shift of the peak Tmax distribution to the right (by around 1.5 °C), along with this positive skewness, indicates a greater frequency and intensity of extreme hot weather events, such as HWs, during the El Niño conditions. In contrast, during the Neutral phase of the ENSO (referred to as Normal), the peak Tmax increased from 38.1 °C in the 1981–2000 period to 39.0 °C in the 2001–2020 period, also accompanied by positive skewness in the Tmax distribution. The shift of the peak Tmax distribution to the right (by around 0.9 °C), along with moderate positive skewness, indicates a significant frequency and intensity of extreme hot weather conditions, including HWs, during Neutral conditions. In contrast, during the cool phase of ENSO (known as La Niña), the peak Tmax decreased slightly from 39.11 °C in the 1981–2000 period to 38.86 °C in the 2001–2020 period, exhibiting a slight negative skewness in the Tmax distribution. The decrease in the peak Tmax (by around 0.25 °C), together with the slight negative skewness, indicating the frequency and intensity of high-temperature events remaining largely unchanged between the two periods.

To further quantify these changes, seasonal area-averaged Tmax statistics were examined across three categories (see Supplementary Table S1 for respective statistics). In “Neutral” years, the average Tmax increased from 32.91 ± 0.15 °C during the period of 1981–2000 to 33.68 ± 0.13 °C from 2001–2020. This reflects a significant increase of 0.77 ± 0.20 °C, underscoring an important trend in the warming trajectory of our climate. In contrast, during La Niña years, the average Tmax experienced a slight decrease from 33.72 ± 0.09 °C in 1981–2000 to 33.13 ± 0.10 °C in 2001–2020. This indicates a reduction of -0.59 ± 0.13 °C, showcasing the net impact of the climate cooling trend alongside the cooler phase of the ENSO. In El Niño years, however, the average Tmax has experienced a significant increase from 32.89 ± 0.28 °C between 1981 and 2000 to 33.92 ± 0.34 °C from 2001 to 2020. This shift represents a significant warming of 1.03 ± 0.44 °C, underscoring the combined effects of climate warming trends and the warm phase of ENSO. These findings highlight the profound impacts that these natural climate patterns have on our planet.

During the period from 1981 to 2000, the average Tmax was insignificantly about 0.02 °C higher under “Neutral” conditions compared to “El Niño” conditions. During this time, El Niño affected an area of approximately 316,000 km², while the “Neutral” area was around 324,000 km² (see Supplementary Table S1). However, between 2001 and 2020, we observed a significant shift: the average Tmax increased to approximately 0.24 ± 0.36 °C higher during “El Niño” events compared to their “Neutral” counterparts. Additionally, El Niño impacted around 2,410,000 km², while the “Neutral” area was about 122,000 km² (see Supplementary Table S1). This striking comparison underscores the significant climatic differences tied to the ENSO phenomenon. An analysis of the statistics indicates that El Niño conditions have increased the average Tmax during Neutral hot weather in India by approximately 0.26 ± 0.48 °C, comparing the periods 1981–2000 and 2001–2020. This amplification of Tmax caused by El Niño significantly exacerbates HW conditions across India. The two-sample Student’s t-test shows that the average maximum temperature (Tmax) has increased significantly from 1981–2000 to 2001–2020 for all ENSO phases (Supplementary Table S2). For El Niño and Neutral years, the calculated t-values are higher than the 95% confidence threshold, indicating a strong and consistent warming. Although the change during La Niña years is also statistically significant, the warming effect is smaller than that seen during El Niño and Neutral conditions. Overall, these results suggest that ENSO phases influence how strongly seasonal temperature extremes respond to long-term warming, with El Niño and Neutral years showing a larger increase in Tmax.

To assess the impact of ENSO on the frequency and intensity of regional HWs in India, Tmax recorded during HWs in both “El Niño” and “Neutral” years are analyzed for the periods 1981–2000 and 2001–2020. The HW frequency increased from 1-6 days per year in 1981–2000 to 4–12 days per year in 2001–2020, and a similar pattern was observed for HW duration, which exhibited a marked increase in northwestern, central-eastern, and peninsular India during 2001–2020 (see Supplementary Fig. S2 a, b). “El Niño” events tended to drive long-duration and high-intensity HWs with a higher frequency. Further, the spatial distribution of temperature differences particularly during the HW days between “Neutral” and El Niño events for the periods 1981–2000 and 2001–2020 (see Supplementary Fig. S2 c, d) show three distinct regions of strongest warming during 2001–2020, which are Peninsular and Eastern India, where Tmax increased by over 0.9–1.2 °C, followed by western-central India with an increase of 0.3–1.2 °C. In contrast, during the earlier period (i.e., 1981–2000), the warming was largely confined to Southern (0.3–0.9 °C), followed by Western-Central, Western, and Eastern Coastal regions (0–0.6 °C). These results indicate that although global warming has led to an overall increase in HW intensity, El Niño events have imposed an additional warming effect in recent decades (2001–2020), further amplifying extreme heat conditions. Much of this change has occurred over the last two decades, indicating that the warming trend accelerated over the 2001–2020 period. This reflects the increasing vulnerability of the region to extreme heat. Moreover, this suggests that we can expect a further increase in HW days under continued global warming.

HW hotspots across India and its migration over the last 4 decades

A HW hotspot is a region whose climate is particularly sensitive to global warming. The identification and characterization of climate response hotspots provide crucial insights into regional climate change processes. To unveil HW hotspots, we developed the HHI to identify the most exposed regions in India by integrating five key attributes of HWs as illustrated in Fig. 2, viz., frequency, duration, intensity, Tmax anomalies, and number of hot days (see Methods section). The HHI values range from 1 to 5, with a higher number indicating greater intensity.

To evaluate the effectiveness of HHI in delineating the hotspots, we compared this index with satellite-based data sets for LST and NDVI for the 2011–2020 period. We specifically analysed the ratio between NDVI and LST due to their inverse relationship, meaning areas with higher NDVI tend to have lower LST and vice versa. This NDVI/LST ratio is a powerful tool in remote sensing for assessing the interactions between vegetation, land cover, and surface temperature, particularly strong during the hot weather summer season^51,52. The estimated NDVI/LST ratio and LST values were grouped according to HHI classes and are illustrated in Fig. 4a through a box plot graph representing the hot weather seasons from 2011 to 2020. The figure includes key statistical indicators, such as the mean, median, upper limit, lower limit, and standard deviation for both the NDVI/LST ratio and LST values across the five HHI indices. The correlation coefficient between HHI and NDVI/LST is -0.83, indicating a strong negative linear correlation between variables (see Supplementary Fig. S3). The figure clearly shows that lower HHI values correspond to higher NDVI/LST ratios and lowest LST values, as areas with dense vegetation (high NDVI) contribute to cooling through evapotranspiration. Conversely, high HHI values are associated with low NDVI/LST ratios and high LST values, as the sparse vegetation or bare soil (low NDVI) experiences elevated temperatures. Further, ANOVA analysis indicates that each HHI class has a distinct NDVI/LST level, NDVI/LST decreases systematically as HHI increases from 1 to 5, and no two classes have the same mean (see Supplementary Table S3). Thus, the observed nature of increasing HHI as the NDVI/LST ratio decreases and LST rises reinforces the reliability of using this index for identifying HW hotspots. Moreover, for HHI = 5, the LST values 48.72 ± 3.93 °C during “El Niño” years for 2011–2020 (see Supplementary Table S4). Similar distinctions are observed across all HHI classes during the 2011–2020 period (see Supplementary Table S5), further supporting the validity of using this index.

Fig. 4.

Validation and spatial distribution of the Heatwave Hotspot Index (HHI). (a) Boxplots of NDVI/LST (blue, left axis) and LST (red, right axis) across HHI classes (1–5), showing decreasing NDVI/LST and increasing LST with higher HHI. (b–c) Spatial distribution of HHI depicting HW hotspots (i.e., shaded hatching denoting consistently identified hotspot regions during 1981–2000 and 2001–2020). These maps were generated in ArcGIS Pro 3.5.3 (https://www.esri.com/en-us/home). The India shapefile is from https://onlinemaps.surveyofindia.gov.in/.

Figure 4b and c illustrate the spatial extent of the HHI over India for two time periods: 1981–2000 and 2001–2020, respectively. The HHI values ranged from 1 to 3 during the 1981–2000 period, while the values increased to a range of 5 from 2001 to 2020, indicating an enhancement in the HW hotspots. Between 1981 and 2000, HW hotspots (i.e., HHI = 3) were primarily concentrated in the Gangetic plains, central India, coastal Andhra Pradesh, and parts of western India. In contrast, the period from 2001 to 2020 saw a significant expansion of HW hotspot coverage (i.e., HHI > 3), reaching into northwestern, central, and eastern regions, as well as large areas of peninsular India. Analysis of the HHI on a decadal scale shows a clear trend of spatial expansion of HW hotspots throughout each decade (see Supplementary Figure S4). Notably, extensive areas in peninsular, central, and western India, along with some pockets in eastern India, have emerged as new hotspots in recent years. These regions are becoming ever more susceptible to HWs. Potential contributing factors to these regional changes include natural climatic variations as well as anthropogenic influences, such as deforestation and urbanization. Projections indicate a two- to three-fold increase in the frequency and intensity of HWs across India by the mid-21st century compared to the 1981–2010 baseline³⁴. There is approximately 1.52 times increase in the spatial extent of HW hotspots, rising from 1,190,000 km² in 1981–2000 to 1,810,000 km² in 2001–2020 (see Supplementary Table S6). This enhancement and migration of HW hotspots present challenges in addressing the growing impacts of climate change and developing strategies to strengthen resilience in these vulnerable regions. Therefore, we further examined the health risk assessment related to HWs.

HW health-risk across India under intensifying climate change

India experienced its longest recorded HWs in 2024, with cities warming twice as fast as the rest of the country⁵³. Extreme heat has already begun to strain public health systems, push power demand to record highs, and deplete water resources, significantly impacting daily life. To enhance heat resilience in India, understanding heat risk assessment is a crucial step. In this study, we mapped HW health risks across India based on the IPCC framework⁵⁴, which defines risk as a combination of hazard, exposure, and vulnerability. We calculated each of these metrics and created gridded spatial layers for each identified metric (refer to the Data and Methods section for more details). The HW hazard was estimated using the HHI parameter. The HW exposure index was generated by combining gridded population data with population density data from India. The vulnerability index is estimated as a function of sensitivity and coping capacity. For sensitivity, we considered factors such as children under six years old, the elderly population, agricultural workers, and construction labourers, as these groups are particularly sensitive to extreme heat. For coping capacity, the factors include natural resources (measured by the greenness Index, or NDVI, and the distribution of inland water bodies), the number of medical centres per capita, and literacy rates. These details, along with the selected indicators and their assigned weights for estimating risk, are provided in the Data and Methods section. The period from 2011 to 2020 was selected for risk evaluation, as it aligns with the most recent Indian population census conducted in 2011, which is the latest available dataset. Although population projections indicate an increase of around 15% by 2021(as per Ministry of Health and Family Welfare (MoHFW)), the use of 2011 census data do not affect the relative spatial patterns of risk, as population growth has been broadly uniform across states. The estimated risk is compared with state-wise records of mortality due to heat and sunstroke (Supplementary Figure S5), revealing a correlation of 0.8. The minor discrepancies between the two parameters are expected, given the lack of comprehensive morbidity data (i.e., daily health facility reporting and reported emergency visits). However, considering the strong agreement between the HW health risk and mortality, we can conclude that the risk estimate provides a valuable insight into identifying the regions most at HW risk.

Figure 5a–d present the individual findings from all three indices, viz., hazard, exposure and vulnerability, providing a detailed assessment of the factors that contribute to overall health risk associated with HWs. Over the last decade (i.e., 2011–2020), nearly 60% of districts experienced HWs. Areas with high population density, such as Andhra Pradesh, Telangana, Maharashtra, Central-lower Indo Gangetic Plain (IGP), and many parts of Rajasthan and Gujarat, face the highest exposure to extreme heat. With nearly 50% of India’s population expected to live in urban areas by 2050, this situation poses a serious threat to public health. Districts in Maharashtra, Western Madhya Pradesh, Central to lower IGP, and parts of Andhra Pradesh, Telangana, and Jharkhand are particularly vulnerable to extreme heat. This is due to the significant proportion of the population being elderly persons (60 + years) and young children (under 5 years), along with a high prevalence of working individuals, all of whom are more susceptible to the impacts of extreme heat. In contrast, districts such as those in Rajasthan, Gujarat, Jharkhand, and some in Madhya Pradesh, which experience increased heat extremes but demonstrate low vulnerability, benefit from higher green cover and other adaptative factors. These factors improve their adaptive capacity, allowing communities cope more effectively with extreme heat. However, this does not mean that these districts are free of heat-related risk; rather, their risks are relatively lower compared to other districts. By aggregating these factors, we have identified five states and Union Territories have the highest HW health risk.

The regions at risk include districts of central plains, the southeastern coast, some parts of western India and Gangetic West Bengal, along with some areas in the central Indo-Gangetic Plain (IGP) region, indicating the increased health risk due to HWs associated with global warming. The current findings provide valuable insights for developing future local adaptive planning aimed at mitigating HW risks and protecting vulnerable populations.

Summary and discussion

This study offers a comprehensive assessment of the evolution of HWs in India during the hot weather season from 1981 to 2020. The analysis indicates a consistent and concerning rise in the average Tmax at a rate of 0.26 ± 0.03 °C per decade. A notable increase in Tmax of 1.04 ± 0.12 °C has been observed in recent years (2020s) compared to the 1981 period. Spatial trends indicate an expansion of warming across central, western, and peninsular India, with significant increases in HW frequency (~1.5 fold increase), duration (~1.2 fold increase), and intensity (~3 fold increase) between 1981–2000 and 2001–2020. The area affected by HWs has also grown substantially, with newly vulnerable regions emerging in central, eastern, and southern India. This confirms a shift in the spatial distribution of HWs and the intensification of hot extremes.

El Niño conditions have significantly intensified warming across India from 2001–2020, resulted in more intense and prolonged HWs. These findings underscore the compounding impact of global warming and climate variability (particularly El Niño events) in exacerbating HWs across India, posing significant risks to public health and the economy. Apart from the large-scale drivers, HW intensity and spatial distribution are further influenced by regional-scale factors, including land-atmosphere interactions and aerosol direct effects, which require further investigation.

The evaluation shows that the HHI is consistent with satellite-based products, including the NDVI/LST ratio and LST values. Over the past four decades, HW hotspots have expanded from the Gangetic plains, central India, and coastal Andhra Pradesh to large parts of peninsular, western, and eastern India, with around 1.5 times increase in spatial extent from 1981–2000 to 2001–2020. Previously less affected regions, such as peninsular and eastern India, have now become the hottest hotspots.

When combined with HW hazard, socio-demographic vulnerabilities, and exposure, health risks were found to be highest in the central plains, southeastern coast, Gangetic West Bengal, and sections of the central IGP. Our analysis shows that both climatic drivers and socio-demographic conditions are amplifying the risk of HWs across India. Nearly 60% of districts experienced HWs during 2011–2020, with Maharashtra, Andhra Pradesh, Telangana, western Madhya Pradesh, and Jharkhand among the most affected due to large shares of elderly persons, young children, and outdoor workers. By contrast, regions such as Rajasthan, Gujarat, and parts of Madhya Pradesh and Jharkhand, though hazard-prone, showed relatively lower vulnerability due to greater coping capacity. Together, these results suggest that the intensification of HWs in the future may pose a growing threat to human health. The findings provide critical evidence to inform targeted adaptation, including region-specific heat action plans, improved healthcare infrastructure, and green infrastructure planning to reduce population exposure.

Data and methodology

The study utilised daily maximum temperature (Tmax) data over India obtained from the ERA5 (European Centre for Medium-Range Weather Forecasts Reanalysis version 5)⁵⁵, which has a spatial resolution of 0.25°. This data is employed to investigate the spatiotemporal evolution of HW hotspots in India over the period from 1981 to 2020. The ERA5 reanalysis is found to be more accurate than other recent reanalysis products⁵⁶ and is reported to better replicate the observed spatial patterns of extreme events and their changes across various regions of India⁵⁷. The focus of the study is on the hot weather season, specifically April and May, as the summer season in India begins in March and extends until mid-June. However, the most intense heat typically occurs throughout April and May in many regions of India.

ENSO events

The Oceanic Niño Index (i.e., sea surface temperature departures from average) in the Niño 3.4 region, commonly referred to as the Niño 3.4 index, is used for assessing ENSO. The data is classified according to the Niño 3.4 index anomaly for the April-May season, which categorizes years as warm ENSO years (Niño 3.4 anomaly > + 0.5 °C), cool ENSO years (Niño 3.4 anomaly < − 0.5 °C), and Neutral ENSO years (− 0.5 °C ≤ Niño 3.4 anomaly ≤ + 0.5 °C). During the 1981–2000 period, there were 6 warm ENSO years, 6 cool ENSO years, and 8 Neutral ENSO years. In contrast, from 2000–2020, there were 3 warm ENSO years, 2 cool ENSO years, and 15 Neutral ENSO years as per the above-mentioned criteria.

Identification of HWs and HW parameters

A heatwave (HW) is defined as a period during which the daily maximum temperature (Tmax) surpasses the 90th percentile for three consecutive days. This percentile is calculated based on a moving 15-day window of Tmax over the study period from 1981 to 2020. From this, five specific metrics for HWs are estimated, viz. (1) HW duration (length of the longest HWs, measured in number of days per year), (2) HW frequency (total number of days that contribute to HWs throughout the year), (3) HW intensity (Tmax observed during HWs each year), (4) Mean Tmax Anomaly (the difference from the mean) and (5) Number of Hot Days (the total number of days in which Tmax exceeds 40 °C, regardless of whether the days are consecutive, for each year). The metrics are established indicators of the impacts of HWs on human health and the environment^58,59. For each of the metrics for April and May from 1981 to 2020, calculations were performed, and gridded spatial layers were created for each identified metric.

Identification of HWs hotspots

To systematically identify regions affected by HWs, we developed a Heatwave Hotspot Index (HHI) using the Weighted Overlay Technique (WOT) within a Geographic Information System (GIS) framework, by integrating the five HW metrics: HW frequency, HW duration, HW intensity, Mean Tmax Anomaly, and the number of hot days. The HHI is used to examine the temporal evolution and spatial patterns of the HW hotspot across India.

The methodology employs the weighted overlay technique (WOT), a multi-criteria evaluation method that utilizes a spatial decision support tool to rank and identify optimal locations or pixels according to predefined criteria^58,60–63. As per the proposed HHI methodology (flowchart is provided in Supplementary Figure S6), each HW metric was computed at each grid cell and converted into a spatial raster layer. To ensure comparability across metrics with different physical units, each raster was normalized using a min–max classification approach and reclassified into five ordinal severity classes (1–5), representing very low to very high HW severity. For a given metric k, classification thresholds were derived from the spatial distribution of that HW metric within each decade using equal interval breaks between the observed minimum (X_min) and maximum (X_max) values:

where Δ = (X_max − X_min)/5 and C_ik represents the standardized severity score (1–5) of HW metric k at pixel i (an example of the classification scheme and scoring ranges for each HW metric is provided in Supplementary Table S7).

Weight determination using the multi-influence factor method: The relative importance of each HW metric was quantified using the Multi-Influence Factor (MIF) method. Pairwise interactions among the five HW metrics were evaluated based on expert judgment and literature^58–64 and assigned influence scores of 1.0 (major influence), 0.5 (minor influence), or 0.0 (no influence). These interactions were organized into a square influence matrix (Supplementary Table S8).

The total influence score for each HW metric was calculated as:

where I_kj denotes the influence of HW metric k on metric j. The normalized weight for each HW metric was obtained as:

Based on this approach, the final weights were: 0.25 for HW duration, 0.20 each for HW frequency, HW intensity, and mean Tmax anomaly, and 0.15 for the number of hot days. The complete influence matrix and final normalized weights are provided in Supplementary Tables S7 and S8.

Weighted overlay technique and HW hotspot identification: The HHI for each pixel i was calculated using a weighted linear combination.

where W_k is the normalized weight of HW metric k and C_ik is the standardized severity score (1–5) of HW metric k at pixel i.

The resulting HHI class values range from 1 to 5 and were categorized into five levels of increasing cumulative heat stress. Grid cells with HHI ≥ 3 were classified as HW hotspots (as in previous decades the HHI classes were between 1–3 followed by 2–5 in the recent decade). Finally, the hotspot layers were converted into vector shapefiles to facilitate comparison and interpretation over decadal periods.

Evaluation of HHI with satellite-based LST and NDVI data

For evaluation of the HHI, daily MODIS Aqua land surface temperature (MYD11A1) and NDVI (MYD09GA) products were processed in the Google Earth Engine platform to generate the seasonal (April-May) mean of each year between 2011–2020. In addition to LST and NDVI, the NDVI/LST ratio was derived. Spatial and statistical comparisons were conducted between HHI and LST, and between HHI and NDVI/LST, to assess whether HHI distinctly classifies the HW hotspots.

HW health risk assessment

This study implemented the IPCC’s conceptual framework for assessing HW health risk assessment, where risk is determined by the factors of hazard, exposure, and vulnerability⁵⁴.

Risk index: Risk refers to the potential for quantified losses (such as life, property, etc.) resulting from the interaction of hazards, exposure, and vulnerability. The mathematical relationship is:

Hazard refers to “the potential occurrence of a natural or human-induced physical event or trend or physical impact that may cause loss of life, injury, or other health impacts, as well as damage and loss to property, infrastructure, livelihoods, service provision, ecosystems, and environmental resources⁵⁴. In the context of this study, the Hazard Index is calculated as a composite score based on different HW hazard characteristics.

where H_{indicator, j} are normalized indicators of HW hazard magnitude, frequency, etc., and w_i are their respective weights. The HW hazard was estimated using the HHI parameter. The results were then normalized to a range of 0 to 1 index-value range and categorized into a five-level qualitative scale.

Exposure refers to “the presence of people, livelihoods, species or ecosystems, environmental functions, services, and resources, infrastructure, or economic, social, or cultural assets in places and settings that could be adversely affected⁵⁴. The Exposure Index refers to the inventory of “elements-at-risk” (people, property, infrastructure, and ecosystems) located in a hazardous area. An exposure index quantifies the extent to which these elements are present in a given area. Mathematically, it can be defined as a measure of the amount or value of assets in a specific location prone to a hazard:

where P_l are assets located in the hazard zone can be a simple count of population or the number of buildings, or a more complex calculation of economic value and w_l are their respective weights.

In the context of this study, we estimated the HW exposure index by combining gridded population data and population density data over India (See Supplementary Table S9). We normalize these values to the same 0 to 1 range and categorize them into the same five-level qualitative scale, assuming that the index increases as population density rises.

Vulnerability refers to “the propensity or predisposition to be adversely affected and encompasses a variety of concepts and elements, including sensitivity or susceptibility to harm and lack of capacity to cope and adapt⁵⁴. Therefore, vulnerability is a function of both sensitivity and coping capacity. The Vulnerability Index measures the susceptibility of exposed elements to damage and their capacity to cope with or recover from a hazardous event. It often combines various socio-economic, physical, and environmental factors. A common approach for a Vulnerability Index is a weighted aggregation of normalized indicators:

Where S_j are normalized metrics (e.g., income level, education, building quality, access to healthcare) and W_j are the weights reflecting their relative importance. Similarly, C_k are normalized metrics and W_k are the weights reflecting their relative importance. In terms of sensitivity, the factors considered are children under six years old, the elderly population, agricultural workers, and other workers broadly encompassing individuals engaged in non-agricultural and non-household activities such as construction, mining, manufacturing, transport, salespersons, etc., as these groups are more sensitive to heat (see Supplementary Table S9). For coping capacity, the factors include natural resources (measured by the greenness index NDVI and inland water bodies), the number of medical centers per capita, and literacy rates (see Supplementary Table S9). Relative weights for these factors are estimated (as discussed below and see Supplementary Table S9) and used to aggregate the indicators, resulting in two components of vulnerability: the sensitivity index and the capacity index. The heat vulnerability index is calculated as the ratio of the sensitivity index to the capacity index, reflecting the inverse relationship between capacity and vulnerability. Finally, the vulnerability index values are normalized to a range of 0 to 1. This method of combining various components and indicators using relative weights aligns with the IPCC’s explanation of vulnerability indexes. According to the IPCC, “a climate vulnerability index is typically derived by combining, with or without weighting, several indicators assumed to represent vulnerability⁵⁴.

In this study, we employed an objective weighing technique known as the entropy-based method to quantify the significance of indicators objectively by assessing the variation in their numerical values across different units. This method quantifies the relative importance of indicators by evaluating the difference in their numerical values⁶⁵. Unlike subjective weighting techniques, entropy-based weights are generally considered more reliable and accurate, as they minimize human biases in the weighting process⁶⁶. For the entropy method, all indicators were first normalized using:

where I_jd is the normalized value of the indicator for each district; x is the original value of the indicator; x_min and x_max are the minimum and maximum value of each indicator, respectively. Here, d represents the number of districts and j represents the serial number of the indicator.

Contributing Factors (CF_j) of each indicator were then calculated as:

The entropy (є_j) of each indicator was obtained as: please correct c as є  in the below equation and keep the left side expression same as in our original manuscript  

where n is the total number of districts.

And the final weight (W_j) of each indicator is:  please correct c as є  in the below equation and keep the right side expression same as in our original manuscript  

Where j ranges from 1, 2,….,m.

The weight obtained from this approach is included in the Supplementary file (see Supplementary Table S9). The HW health risk index is then determined by multiplying the derived indices of three risk components: hazard, exposure, and vulnerability.

Statistical methods

The non-parametric Mann-Kendall test^67,68 was employed to assess the presence of monotonic trend and to estimate the trends in time series data for near-surface maximum temperature (Tmax) and HW parameters at both the pixel level and the regional level. Additionally, Sen’s slope method⁶⁹ was utilized to quantify the magnitude of these trends. All trend analyses are evaluated for statistical significance at the 90% confidence level (p < 0.1). In this study, we employed a similar approach to that previously used in one of our studies⁷⁰ for handling the autocorrelation in time series data.

The Probability Density Function (PDF) for Tmax was analyzed for three ENSO classes, viz., warm, cool and Neutral phases of ENSO for two time periods (i.e., 1981–2000 and 2000–2020). The PDF of Tmax was generated using the non-parametric kernel density estimate (KDE) method with a kernel bandwidth of 0.5, capturing the distribution of temperature values with precision. The shifts in mean, skewness, and mean ± standard deviation for three ENSO class PDFs are evaluated for statistical comparison. Two-sample t-tests were conducted to examine whether the changes between the two time periods are statistically significant.

Spearman’s correlation coefficient method was employed to quantify the statistical validity between HHI and NDVI/LST as well as between HW health risk and heat-related mortality. Additionally, an ANOVA (Analysis of Variance) analysis is conducted for HHI and NDVI/LST to determine whether the classes are distinct.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

BP and SB conceptualized and designed the study. MVR helped in the conceptualization of the heatwave health Risk. SB analyzed the data, methods, prepared figures and wrote the first draft of the manuscript. BP and MVR supervised the study, methodology, reviewed and edited the manuscript. All authors contributed to interpreting the results and preparing the final manuscript.

Funding

This research did not receive any funding.

Data availability

All the data supporting the findings of this study are available publicly. The daily temperature data for this study is obtained from ERA5 reanalysis, which is available online (https://cds.climate.copernicus.eu/datasets/reanalysis-era5-pressure-levels?tab=overview). The LST data used in the study is available online (https://developers.google.com/earth-engine/datasets/catalog/MODIS_061_MYD11A1). The NDVI data used in the study is available online (https://developers.google.com/earth-engine/datasets/catalog/MODIS_MYD09GA_006_NDVI). The Indian census data used in the study is available online (https://censusindia.gov.in/census.website/data/population-finder). The State-wise details due to Heat/Sun-stroke data is taken from National Crime Record Bureau statistics, which is available online (https://www.data.gov.in/resource/stateut-wise-details-deaths-due-heatsun-stroke-2013-2022).

Declarations

Competing interests

The authors declare no competing interests.