The Intergovernmental Panel on Climate Change (IPCC) report [1
] highlights the projected increases in heatwave frequency, intensity and duration, and resulting deaths both globally and in India. Heatwave events have caused massive deaths in the past; the most famous among them are the European 2003 and Russian 2010 heatwaves, where tens of thousands died [2
]. India has experienced several heatwaves, and most recently, just in the past two years, thousands have reportedly died [5
]. Research has documented an increase in cardiovascular [6
], respiratory [7
], and all-cause [6
] mortality along with increases in ambulance calls and admissions [8
] resulting from heatwave exposure. While most of the evidence is from North America and Europe, there is an emerging body of evidence from developing countries, including India [10
], where heat wave deaths may currently be underestimated [11
At the same time, heat-related deaths are preventable and prevention programs have been shown to be extremely cost effective [12
]. Population adaptation [13
] along with preparedness measures have reduced mortality. Indeed, several cities and countries around the world have adopted heatwave preparedness plans [13
]. However, in India, this effort is limited to only a few cities [14
]. A broader preparedness strategy is particularly important given the large population, difficult local conditions, and limited adaptive capacity.
Health vulnerability can be conceptualized as complex and multidimensional [15
]. Vulnerability encompasses individual biophysical characteristics, as well as population-level socio-economic-environmental characteristics. These population measures have typically included measures of age, income, discrimination, social isolation, vegetation, and health characteristics [16
]. Incorporating multidimensional data can present a more comprehensive characterization of vulnerability. Given the considerable intra-country variations in these measures that exist, it is prudent to use this characterized vulnerability to identify communities in need of prioritized and focused interventions. Heat Vulnerability Indices (HVI) have been found to be a useful screening tool for targeting heat risk interventions [16
Several international studies have explored vulnerabilities at the national, county, and city levels [17
], but none have comprehensively examined India. Additionally, most of this work has been performed in the context of urban settings, while the majority of the Indian population resides in rural areas. To our knowledge only one Indian study looked at agricultural vulnerability using census data from 2001 [21
]. In fact, a recent review of heat vulnerability indices points out that that the majority of studies have been performed in Europe and the United States and recommends further study in other countries and regions to account for local context [16
]. We, therefore, aim to create and map an integrated district level heat vulnerability index for India that can be used to identify the most heat-vulnerable districts in the country.
demonstrates the sources and descriptive characteristics of the 17 raw variables included. Some of these variables showed interesting variations, for example, on average only 42.3% of households had drinking water inside their premises but it ranged from 2.4% to 93.8% across districts. Similarly TV ownership, immunization status, having nearby health sub-centers and vegetation also showed such marked variations.
Variable correlations showed many of the variables highly correlated with each other at the 0.001 significant level. (not shown here).
shows the PCA results with Varimax rotation. The factors have been reduced to four dimensions. These correspond to demographic, socio-economic, vegetation, and health systems. The PCA led to four factors with primary loadings, these appeared to be (1) demographic; (2) socio-economic; (3) environmental; and (4) health factors. Demographic loadings were constituted of extremes of age, socio-economic loadings were driven by household amenities, environmental loadings were contributed by the VF and NDVI scores and health was driven by availability of health facilities nearby.
Of the 17 variables that we started with, these four groups of factors were able to account for 78% of the total variation. The scree plot of eigenvalues showed a clean break at four components (Figure 1
). This was also in agreement of the Kaiser criteria. The Kaiser–Meyer–Olkin (KMO) test shows adequacy of our analysis (>0.50).
The HVI calculated as a sum of the four individual factors for each district ranged from −11.8 to 9.4, it had a mean close to 0 and SD of 3.5. Figure 2
maps the vulnerability across the country. Spatial clustering of these “hot spots” is observed in central India. These districts have poor socio-economic and development indicators and appear to be high on the heat vulnerability index.
We classified these categories based on the SD scores as “very high” (>2SD), “high” (1–2SD), “high normal” (0–1SD), “low normal” (−1–0SD), “low” (−2–1SD), and “very low” (<−2SD). We chose this SD based classification instead of equal categorization to better represent the variation. Table 3
shows the number of districts according to HVI standard deviations.
Ten districts had an HVI score of “very high” (>2SD), most of them in central India in the states of Madhya Pradesh and Chhattisgarh (Table 4
). Twenty districts had an HVI score of “very low”; most of them were in the relatively developed states of Kerala and Goa and union territories of Lakshadweep.
This study provides a relative ranking of heat wave vulnerability for all districts in India. Although much is known about factors that contribute to vulnerability from other settings [16
], there has been minimal research conducted within India on heat-related vulnerability. By coupling this knowledge with local context and using methods previously applied in other settings [17
], we created an index that describes relative variation in heat-related vulnerability across all of India. This index can be used by planners, policy makers, and disaster mitigation experts to target climate adaptation efforts.
Similar to the findings of other international studies [17
], our index too identified demographic, socio-economic, environmental, and health system factors. However, there are important differences in the choice of initial variables making this index useful to the Indian and developing country settings.
The high and very high HVI districts were in the central part of the country. With a higher tribal population, these states have been at the lower end of various health, education, economic and population growth indicators. They are referred to as the Empowered Action Group (EAG) states and often targeted for focused interventions. These land-locked, high HVI districts in the North and Central Indian plains are classically known as the “heat belt”.
While the use of air conditioning has been observed to have the greatest impact in reducing heat wave deaths in the US [36
] it is unlikely to be a solution for India at least in the short term because of lack of a reliable and continuous power supply, the high cost and low penetration of air conditioning.
Suitable local adaptation strategies therefore need to be considered. These may include a range of measures, some of which have been discussed in the literature, such as public messaging (Radio, TV), mobile phone-based text messages, automated phone calls, and amber alerts; to others such as traditional adaptation practices of staying indoors, wearing comfortable clothes, and diets. These are often visible in terms of the housing design and construction material used. Simple design features such as having shaded windows and underground water storage tanks can be helpful. Use of insulator housing materials similarly can be an effective method of prevention. Having access to drinking water within housing premises and indoor toilets could be important. We chose several household amenities not just to proxy for income but also for their protective role.
For risk management, it would make sense to observe whether these identified areas of high vulnerability are also the same as those with higher temperatures and humidity. Similarly, had district level heat wave death data been available, we could have used it to validate this index. Our index does show high (>0.70) and significant correlations with literacy rates, low income status, TV ownership, having toilets and drinking water and open defecation practices. These could be seen as starting points framing local adaptation strategies. These correlations highlight the importance of interventions against other associated diseases such as gastrointestinal diseases in children and water-borne illnesses etc. There is a moderate correlation of 0.42 between HVI and average summer land surface temperatures (from satellite data) suggesting a relationship between higher temperatures and heat vulnerability. The index also shows moderate negative correlation (–0.46, p < 0.001) with urbanization signifying a possible greater vulnerability threat in rural areas. Since the majority of Indians reside in rural areas, this could have important implications. Outdoor workers have been identified as being at a greater risk during heatwaves. In rural settings, agricultural practices in different regions of India may also have diverse vulnerability patterns.
Some limitations of this study also arise from availability of data. Cardiovascular and/or respiratory diseases are more closely related to heat vulnerability but prevalence of chronic diseases at the district level was not available. Similarly, there was no pan India district level data on social isolation or electricity. For the three DLHS variables, we had missing data for the state of Nagaland. We used state averages instead. However, since Nagaland is a small state which has not reported heatwave deaths, we believe this substitution is unlikely to have major effects. In calculating HVI, we assumed a linear combination of factors with no weighting as a good first assumption. Inclusion of temperature as an exposure variable could have been helpful but because temperatures vary at country-wide levels in a thermally diverse India, it would serve to bias the index in favor of places with higher normal temperatures. Our approach is in line with established methods [17
] for large areas. Also, district level temperature data is only collected for a small fraction of the total 640 districts. In view of still building evidence base from Indian temperature-mortality studies we have cited western literature and some Asian and Indian studies identifying vulnerability factors. This approach has also been demonstrated previously in the air pollution literature [38
In many prior heat vulnerability studies, rural areas have been overlooked, but may have high vulnerability, and this may be especially important in India given the importance of poverty, and agricultural livelihoods in mediating the temperature mortality relationship. However, since the district level data includes both rural and urban areas; by aggregating them we may have missed the differences between these patterns of vulnerability [39
]. Also, since we only had district level data, if such data was available at the finer block (Taluka) level, we would have a better identification of vulnerable areas. Similarly, intra-city vulnerability patterns would have been interesting to observe given availability of urban ward level data. Given data availability, future work could also identify areas using the Koeppen climate classification assuming that the warm, dry, arid, and humid areas are more vulnerable. As research continues, we may identify more complex relationships and therefore we could conceive of other heat vulnerability indices with non-linear relationships and differential weights.
Despite the above discussed limitations there are several strengths to this work. This is the first study to look at heat vulnerability across India. It provides a preliminary screening to target heat-health and climate adaptation efforts. This methodology can be used to for further investigations into vulnerability.