Heat wave exposure in India in current, 1.5 °C, and 2.0 °C worlds

We estimate characteristics of heatwaves using the method of Russo et al (2015). The Russo et al (2015) method considers a variable Tmax threshold compared to the absolute values of Tmax used by the IMD, and it considers both magnitude and duration of heat waves, which enable us to quantify characteristics of heatwaves more relevant to their impacts. The heat wave magnitude index daily (HWMId) provides a basis to understand severity (e.g. magnitude) and duration of heat waves in any given region. The HWMId is based on the maximum magnitude of heatwaves in each year, where a heatwave is defined as a period with Tmax above a daily threshold for three or more consecutive days. The daily Tmax threshold for each calendar day is defined as the 90th percentile of Tmax centered on a 31 day window, estimated by considering all daily Tmax data of the 15 days before and after for a particular day over a given reference period (e.g. 1971−2000). For each year, the last and first 15 days of the previous and next years, respectively, were considered to avoid inhomogeneity in estimation of heatwave magnitude. However, in future, methods based on bootstrapping (Zhang et al 2005) can also be applied to deal with the inhomogeneity in heatwave estimations. Following a more conventional approach, we considered a fixed reference period and did not apply any correction, which may lead to some exaggeration in the projected estimates as reported in Sippel et al (2015). The overall magnitude of a heatwave event is estimated as the sum of the daily magnitudes within a heatwave, where the daily magnitude (M_d) is given by:

$$\begin{equation} M_{\rm d} = \left\{ {\begin{array}{l@{}l@{}l@{}}{\frac{{T_{\rm d} \,\, - T_{25{\rm p}} }}{{T_{75{\rm p}} \,\, - T_{25{\rm p}} }},} & \, \,{T_{\rm d} > T_{25{\rm p}} } \\ {0,} & {\,{\rm otherwise}} \\ \end{array}} \right. \end{equation}$$

where T_d is the maximum daily temperature on day d of the heatwave, T_75p and T_25p are respectively the 75th and 25th percentile values of the annual Tmax time series, estimated over the reference period (1971−2000). By using the site-specific threshold information, this (normalized) formulation has the advantage that the estimated M_d can be compared across the locations with different climate conditions. Notably the denominator of daily heatwave magnitude (M_d) represents the inter-quartile range (IQR) of the N reference year annual Tmax values. So that if the M_d value on a day d is 3, then it means that the Tmax anomaly on that day with respect to T_25p is three times the IQR. More details on the estimation of frequency, magnitude, and duration of heatwaves can be obtained from Russo et al (2015). Additionally, 3 day annual maximum temperature (3 day AMT) estimated from the daily Tmax data was considered as an alternative heatwave measure (Meehl and Tebaldi 2004).

We estimate observed heatwaves using Tmax data for each 1° grid in India for the period of 1951−2015. Similarly, heatwaves are also estimated for the data obtained from the CMIP5 and CESM simulations at the same resolution. Since CMIP5 models have different native resolutions, their output is first gridded to the common grid at 1° spatial resolution using the bilinear interpolation scheme. For each CMIP5 model, the period of 1971−2000 is considered as the reference period to estimate heat waves in India under the Hist and HistNat scenarios (table S2).

We evaluated simulations from CMIP5 models and CESM runs against the observed heatwave frequency and 3 day AMT (supplemental section S1). Our analysis showed that both CMIP5 and CESM runs underestimate the observed frequency of heatwaves and trend in the 3 day AMT (figure S1). The Hist simulations more closely match the observed frequency than do the HistNat simulations; because no bias correction has been applied this suggests, but does not confirm, the existence of an anthropogenic contribution in increased heatwave frequency in India (figure S1). We find that the CESM-LENS ensemble (also run under Hist conditions) captures observed patterns of frequency of heatwaves better than CMIP5 models over India for the period of 1980−2005 (figures S1(d)−(e)).

The global dataset of gridded population scenarios, available at 0.5°, is from the Center of Global Environmental Research (www.cger.nies.go.jp/gcp/population-and-gdp.html). The dataset was developed for the three Shared Socioeconomic Pathways (Moss et al 2010, van Vuuren and Carter 2014, O'Neill et al 2017) (SSPs: SSP1, SSP2, and SSP3). The data is aggregated to the 1° resolution of the Tmax data. We estimate the MPEHWd by multiplying the product of frequency and duration (in days) of severe (above magnitude 16) heat waves with the population data from each SSPs. To estimate projected MPEHWd for all-India, MPEHWd values from all the grid cells are aggregated for each CMIP5 model and each CESM-LENS members from 20th century scenario coupled with RCP8.5, and low warming (1.5 °C: CESM-1.5C and 2.0 °C: CESM-2C) scenarios.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

We start by characterising recent heatwaves as monitored through gridded daily station measurements of maximum daily temperature from the IMD (www.imd.gov.in), see section 2) during the 1951−2015 period, following the methodology described by Russo et al (2015) (see section 2), considering the heatwave with the highest magnitude in each year. Most heatwaves in India have occurred in the month of May, and nearly all of them have occurred between April and July (table S3). There has been a substantial increase in the frequency of heatwaves (F-HW) during the period of 1951−2015 (figure 1(a)). Based on the heatwave magnitude (an indicator of heatwave severity), the five most severe heatwaves out of the top ten heatwaves in the entire record (1951−2015) occurred after 1990. The top three heatwaves occurred in 1998, 1995, and 2012, with a magnitude of 17.92, 16.92, and 11.14, respectively (figure 1(a)). Since these heatwaves are estimated based on all-India average Tmax, the 2015 heatwave that had a remarkable impact does not feature in the top three events. The 1998 heatwave was a wide-spread event (figure 1(b)) while the heatwave during 2015 was mainly centered in the east-central region of India (figure 1(c)). In terms of heatwave mortality, the 2015 event ranked second (after 1998) indicating that localized heat waves can also be detrimental (Rohini et al 2016, Ratnam et al 2016, Pattanaik et al 2016). These trends and rankings are robust against use of the 3 day AMT, an alternative measure of heatwaves (Meehl and Tebaldi 2004) (figure S2). However, use of a temperature-based definition neglects other meteorological properties of heatwaves that may be important for heat stress among the human population. Heat stress measures based on temperature and humidity provide a relative measure of discomfort (Bishop-Williams et al 2015) that can be a better indicator than the indices based only on temperature. Sherwood and Huber (2010) reported that peak heat stress (estimated using wet-bulb temperature) above 35 °C for an extended period can cause hyperthermia in human and other mammals. The 2015 event was characterized by both high temperature and high humidity and caused substantial human mortality (Wehner et al 2016 Matthews et al 2017), which highlights the importance of humidity in heat stress estimation. Our analysis is limited to heatwaves based on Tmax as long-term gridded relative humidity observations are unavailable for India.

During the past few years, international climate policy has been framed in terms of avoidance of a specified global temperature level (Horowitz 2016, Marotzke et al 2017). A required action then is the determination of what an acceptable temperature limit might be, through assessment of associated risks; the two candidates in most discussions are 1.5 °C and 2.0 °C limits in rise in global warming. In order to assess potential heatwave related risks in India for these global temperature scenarios, we define a 'severe' heatwave as having a magnitude larger than 16 that being the mean magnitude of the top three heatwaves observed during the historical record (1995, 1998, and 2012).We use simulations (11 each) of future climate change from CESM run under scenarios which reach and maintain 1.5 °C (CESM-1.5C) and 2.0 °C (CESM-2C) by the end of the century. Heatwave characteristics are compared from these low warming scenarios against the CESM-RCP8.5 simulations as well as simulations from 27 CMIP5 models run under the RCP8.5 (CMIP5-RCP8.5) scenario of substantial sustained warming through the century.

We estimate the ensemble mean change in the frequency of severe heatwaves from the CMIP5 (RCP8.5) and CESM (RCP8.5, 1.5 °C, and 2.0 °C) datasets for the mid-(2021−2050) and end-century (2071−2100) considering the reference period of 1971–2000 (figure 2). Usage of 30 year intervals is a balance between having a reasonably large sample of heatwaves and having that sample be representative of a single climate era. Under the two low warming scenarios of 1.5 °C and 2.0 °C from the CESM, the frequency of severe heat waves in India are substantially less than that of RCP8.5 during the mid and end of 21st century (figures 2(e)−(i)). For instance, both the CMIP5 and CESM-RCP8.5 ensembles project an increase from 1971−2000 of 3−9 severe heatwaves (per 30 year period) by the mid-21st century under the RCP8.5 scenario (figure 2(a) and (c)). A larger increase (18−30 events in the 30 year period) in the frequency is projected by the end of the 21st century (figure 2(b) and (d)). Similarly, projected duration of severe heatwaves over India is far shorter in the low warming scenarios than that of the RCP8.5 (figure 2(j), figure S4). These results demonstrate substantial effects of climate change mitigation on the characteristics of severe heatwaves over India.

Based on the 27 CMIP5 models and CESM simulations, the frequency of severe heatwaves is estimated for a moving 30 year window starting from 1980 until the end of the 21st century (figure 3(a)). The CESM-2C low warming scenario projects about an eight-fold increase in the frequency of severe heatwaves by the mid-21st century (2021−2050) from the current (1986−2015) period, rising to about 30 times by the end of the century (figure 3(a) and table S6). The frequency of severe heatwaves is projected to increase more than 40 times (0.5 in the current and 22.5 in the end of 21st century) by the end-21st century under the RCP8.5 scenario of CMIP5 and CESM datasets (figure 3(a) and table S4, S5). Failure to limit the global mean temperature to 2.0 °C from the pre-industrial level will result in an increase of 2.5 times (8.7 in CESM-2C and 22 in CESM-RCP8.5) more severe heatwaves by the end of the century (table S5, S6, figure 3(a)). With the mitigation of global mean temperature rise, the area of India that experiences severe heat waves can be substantially reduced under the 1.5 °C or 2.0 °C scenarios in comparison to RCP8.5 (figure 3(b)). Under the 2.0 °C scenario (CESM-2C), the duration of severe heatwaves is projected to rise about 3 and 5 times of the current climate (3.6, 11.6, and 16.1 in CESM-2C in the current, mid, and end of 21st century, respectively) in the mid and end of the 21st century, which would otherwise be even longer (3.2 (current), 12.7 (mid), and 22 (end)) in CESM-RCP8.5 (figure 3(c), table S9, S10).

With severe heatwaves projected to occur at least once in every three years for all regions of the country under the future temperature limits, the question arises how badly an increased occurrence of severe heatwaves will potentially impact the human population of India, along with the question of the degree to which measures might need to be taken to alleviate such impacts.

In order to understand the potential for impacts, we estimate the projected MPEHWd using the frequency and duration of severe heatwaves (above magnitude 16) along with the population data from the SSPs (figure S3) (Moss et al 2010, O'Neill et al 2017). The measure assumes full exposure of the resident human population, with no shelter or other means of limiting direct exposure. We estimate projected MPEHWd for each 1° × 1° longitude–latitude grid cell by multiplying the product of heatwave frequency and mean duration with the mean population projected in the SSP datasets for each 30 year period.

To estimate projected MPEHWd for all of India, MPEHWd values from all the grid cells are aggregated for each year. All the three SSPs provide similar population projections until the mid-21st century (figure S3), which means that the spread in projected MPEHWd is largely driven by the frequency and duration of severe heatwaves rather than population (figure 3(d)), with duration being the dominant driver. We use SSP1 and SSP3 to estimate MPEHWd for the low warming (1.5 °C and 2.0 °C) and RCP8.5 scenarios, respectively. SSP3 assumes a fragmented world following varied regional social, political, and economic pathways. This may be considered difficult to reconcile with the international collaborative effort that would be required in order to keep the global temperature from exceeding 1.5 °C (Riahi et al 2017).

However, we consider it here on the grounds that what applies as a general rule globally does not necessarily need to apply for India itself (notwithstanding India's outsized contribution to world population), and that having a population scenario that spans a larger range will allow a more expanded study of the relation between heatwaves, national population, and MPEHWd.

Since frequency, duration, and areal coverage of severe heatwaves are projected to increase substantially under a warming climate (figures 3(a)−(c)), the projected maximum population exposed to heatwave days (MPEHWd) is also projected to increase significantly (figure 3(d)). Under the 2.0 °C low warming target, MPEHWd is projected to increase by 15 (1921 and 28804 in the current and mid periods under CESM-2C) and 92 (current: 1921 and end 178004) times of the current level (1986−2015), respectively by the mid and end of the 21st century (figure 3(d)). Keeping the global mean temperature to 1.5 °C by the end of the 21st century provides benefits in comparison to the 2.0 °C temperature target. For instance, under the 1.5 °C temperature target (CESM-1.5C), MPEHWd is projected to increase only 18 (current: 1921 and end: 41560) times the current level, instead of 92 times under the 2 °C scenario. Furthermore, benefits of climate change mitigation on heatwave exposure in India can be better understood when we compare exposure levels of the low warming targets against the RCP8.5. Both CMIP5-RCP8.5 and CESM-RCP8.5 datasets project an increase of 18−20 times in MPEHWd by the mid-21st century (from the current levels), which may increase more than 200 times by the end of the 21st century (figure 3(d), and table S12, S13). Moreover, we find that limiting global mean temperature will result in substantially larger benefits in terms of moderating MPEHWd in comparison to limits in growth of the Indian population (figure 4 and table S16, S17). Additionally, our results show that benefits of a feasible slowing of population growth as encapsulated by a comparison between SSP1 vs SSP3 population scenarios will have an effect that is restricted to the late 21st century (figures 4(b) and (c) and table S16, S17).