2. Temperature Changes in India

The mean temperature over India has warmed from the mid-twentieth century (Fig. 2.1), with an increased rate of warming of 0.15 °C, 0.15 °C and 0.13 °C per decade for the annual mean, maximum and minimum temperatures, respectively, between 1986 and 2015 (Table 2.1).

The warming is not uniform across the seasons, with considerably more warming in the pre-monsoon season (March–May; MAM) than in other seasons. The rate of warming of 0.26 °C, 0.29 °C and 0.20 °C per decade for the pre-monsoon season mean, maximum and minimum temperatures, respectively, between 1986 and 2015 are relatively higher than that for the respective annual values (see Fig. 2.1 and Table 2.1). These changes in temperatures are statistically significant at the 90% confidence level. This behaviour of temperature changes is broadly consistent with the long-term variations and trends among the various global and regional gridded temperature estimates derived with different approaches (see Fig. 2.2 and Table 2.2).

These long-term changes in surface air temperature over India during the twentieth century also broadly agree with earlier assessments (e.g. Rupa Kumar et al. 1994; Sen Roy and Balling 2005; Kothawale and Rupa Kumar 2005; Srivastava et al. 2009; Kothawale et al. 2010a; Jain and Kumar 2012; Rai et al. 2012; Vinnarasi et al. 2017; Kothawale et al. 2016; Kulkarni et al. 2017; Srivastava et al. 2017, 2019). Indian annual mean land surface air temperatures have warmed by 0.6 °C century⁻¹ between 1901 and 2018 (Srivastava et al. 2019). All India mean annual tropospheric temperature measured by radiosonde stations also showed an increasing trend from the surface to 500 hPa during the period 1971–2015 (Kothawale and Singh 2017). A similar warming trend is revealed by dendroclimatic studies over the eastern Himalaya including Sikkim and Bhutan in recent decades (Krusic et al. 2015; Yadava et al. 2015; Borgaonkar et al. 2018). Further details on the long-term changes of temperature over India based on paleoclimatic proxies are provided in Box 2.1.

The observed warming is also unevenly distributed across India (Fig. 2.4). The largest increase in the annual mean temperature of more than 0.2 °C per decade are observed in some areas of north India between 1986 and 2015. The warming is much weaker in the southern peninsula, with mean temperature increase in some parts of the west coast lesser than 0.1 °C per decade. The winter warming is limited to peninsular India. The pre-monsoon season shows predominant warming of more than 0.5 °C per decade over north India. The summer monsoon season warming is confined to the eastern parts of the Indo-Gangetic plains and adjoining central India. The post-monsoon season warming pattern is similar to the pre-monsoon season, but with smaller magnitude and more uniformly distributed across the country than for other seasons. These estimates of warming across India based on simple linear trends are found to be, in general, similar to the earlier assessments of the temperature trends derived using non-stationary approach (Vinnarasi et al. 2017).

The all India averaged annual mean temperature increases due to greenhouse gas forcing outweighs the observed decrease in solar radiation (solar dimming; Padma Kumari et al. 2007). The radiative forcing is more effective in altering the strength of hydrological cycle than thermal forcing due to changes in the greenhouse gases (Padma Kumari and Goswami 2010; Soni et al. 2012; Padma Kumari et al. 2013).

The amount of water vapour in the atmosphere is expected to increase at a rate governed by the Clausius–Clapeyron (~7% °C⁻¹) under conditions of warming (Willett et al. 2007; Boucher et al. 2013). The all India averaged annual and seasonal mean specific humidity and relative humidity are found to be increasing significantly for both gridded observations (HadISDH, Willet et al. 2014) and reanalysis (ERA-Interim, Dee et al. 2011) datasets during the period 1979–2015 (Fig. 2.5 and Table 2.3).

The estimated magnitude of the annual specific humidity trend is similar for both datasets and is found to be comparable with the earlier assessment using IMD station data (Mukhopadhyay et al. 2017). The significant increasing trend in specific humidity assessed during the pre-monsoon season is consistent with the largest surface warming trend found for this season (see Table 2.1). Past studies had also reported a rise in the moisture content of the atmosphere associated with warming over the Indian region (Krishnan et al. 2016; Mukhopadhyay et al. 2017). This increased water vapour under conditions of regional warming may lead to significant positive feedback on human-induced climate change, as water vapour is the most important contributor to the natural greenhouse effect (Willett et al. 2007; Boucher et al. 2013).

The surface air temperature changes over India between 1956 and 2005 are attributed to anthropogenic forcing mostly by greenhouse gases and partially offset by other anthropogenic forcings including aerosols and land use land cover change (Dileepkumar et al. 2018). The observed changes in maximum temperature during the post-monsoon and minimum temperature during the pre-monsoon and monsoon seasons in South India during 1950–2005 are assessed to be detectably different from natural internal climate variability, and these temperature changes are attributed with confidence to climate change induced by anthropogenic effects (Sonali et al. 2018). These assessments are based on detection and attribution studies using observational datasets and the multiple atmosphere–ocean coupled general circulation model (AOGCM) outputs of historical simulation experiments conducted in the fifth phase of the Coupled Model Intercomparison Project (CMIP5; Taylor et al. 2012), from which human and natural causes of climate change could be identified and quantified.

The all India averaged annual frequency of warm days and nights have increased, and cold days and nights have decreased since 1951 (Fig. 2.6 and Table 2.4). These extreme temperature indices are defined from daily temperatures as days when daily maximum (daytime) and minimum (nighttime) temperatures are above the 90th (warm) or below the 10th (cold) percentile (see more details in Box 2.2).

The significant annual increase of warm days (about 7.4 days per decade) is found to be higher than that of warm nights (about 3.1 days per decade) during the period 1951–2015. The magnitude of the significant annual decrease of cold days (about −1.4 days per decade) is weaker than that of the annual increase in the frequency of warm extremes for this long-term period. The seasonal frequency of warm days and warm nights also increased significantly since 1951, except for warm nights in the pre-monsoon season (Table 2.4). The increase in frequency of warm days during monsoon (about 2.4 days per decade) and winter (about 2.2 days per decade) seasons contribute largely to the highest annual increase in the number of warm days over India since 1951. The recent 30-year period (1986–2015) shows a significant acceleration in the all India averaged annual increase of warm days (about 9.9 days per decade) and warm nights (about 7.7 days per decade) and decrease of cold nights (about −6.9 days per decade). In recent decades, the significant increase of winter season warm days (about 3.3 days per decade) largely contributes to the accelerated annual increase in number of warm days over India since 1986. The significant decreases in the frequency of cold nights during monsoon (about −3.0 days per decade) and pre-monsoon (about −2.6 days per decade) seasons contribute largely to the accelerated annual decrease in number of cold nights over India in this recent period (Table 2.4). The significant increase of pre-monsoon season warm days (about 2.5 days per decade) assessed for the recent period 1986–2015 is consistent with the earlier assessments of a gradual increasing trend in all India averaged warm day frequency during 1970–2005 period (Kothawale et al. 2010b; Revadekar et al. 2012). However, the decreasing trend in cold day frequency reported in these past studies is not significant between 1986 and 2015 (see Table 2.4).

The all India averaged annual intensity of warmest day shows a significant increasing trend (about 0.07 °C per decade) between 1951 and 2015 (Fig. 2.6 and Table 2.5). The recent 30-year period (1986–2015) shows significant acceleration for the annual increase in the intensity of warmest day (about 0.21 °C per decade) and warmest night (about 0.12 °C per decade), and decrease in the intensity of coldest night (about 0.13 °C per decade). The significant increase in the intensity of warmest day during pre-monsoon (about 0.29 °C per decade) and winter (about 0.26 °C per decade) seasons contribute largely to the accelerated annual increase in the intensity of the warmest day over India in the recent period (Table 2.5). The annual increase in the intensity of the warmest night is dominated by the significant increases in the winter season (about 0.17 °C per decade). The significant decrease in the intensity of coldest night during the pre-monsoon (about 0.28 °C per decade) and monsoon (about 0.15 °C per decade) seasons contribute to the accelerated annual decrease in the intensity of the coldest night over India during the recent period 1986–2015.

Significant increasing (decreasing) trends in heatwaves (cold waves) are observed during the hot (cold) weather season over most parts of India (Rohini et al. 2016, 2019; Ratnam et al. 2016; Pai et al. 2017). These periods containing consecutive extremely hot days (cold nights) are defined when departure in daily maximum (minimum) temperature exceeds (are below) the objectively defined threshold value (Pai et al. 2017). The observed frequency, total duration and maximum duration of heat waves during the hot summer months (April–June) are increasing over central and north-western parts of India (Rohini et al. 2016). The increase in the number of intensive heat waves between March and June in India over a recent-past decade was attributed to the presence of an upper-level cyclonic anomaly over the west of North Africa and a cooling anomaly in the Pacific (Ratnam et al. 2016). A significant decadal variation was observed in the frequency, spatial coverage and area of the maximum frequency of heat (cold) wave events over India (Pai et al. 2017). The variability of heat waves over India was found to be influenced by both the tropical Indian Ocean and central Pacific sea surface temperature anomalies. A noticeable increase (decrease) in the frequency of heatwave days was observed during the El Nino (La Nina) events. It is also assessed that the spatial extent affected by concurrent meteorological droughts and heatwaves is increasing across India during the period 1981–2010 relative to the base period 1951–1980 (Sharma and Mujumdar 2017).

The CORDEX South Asia multi-RCM ensemble mean projected long-term (2070–2099) annual warming exceeds 4 °C over most parts of India except the southern peninsula, relative to the reference period 1976–2005 under the high (RCP8.5) emission scenario, with relatively higher change exceeding 5 °C projected in the semi-arid north-west and north India (Fig. 2.7).

The geographical patterns of long-term change remain below 2 °C relative to the reference period under the low (RCP2.6) emission scenario over most parts of India. The projections of mid-term (2040–2069) change in these CORDEX South Asia multi-RCM ensemble mean indicate modest sensitivity to alternate RCP scenarios over the Indian land area. The projected mid-term annual warming exceed 1 °C over most parts of the country, with higher warming exceeding 2 °C projected in the north-west and north India under the medium (RCP4.5) emission scenario (Fig. 2.7). The summer monsoon temperature projections by a small subset of the CORDEX South Asia RCMs had indicated mean warming of more than 1.5 °C over the central and northern parts of India for the period 2031–2060 under this medium emission scenario (Sanjay et al. 2017a). The CORDEX South Asia multi-RCMs had provided relatively better confidence than their driving CMIP5 AOGCMs in projecting the magnitude of seasonal warming for the hilly sub-region within the Karakoram and north-western Himalaya, with a higher projected change of 5.4 °C during winter than of 4.9 °C during summer monsoon season by the end of the twenty-first century under the high (RCP8.5) emissions scenario (Sanjay et al. 2017b). The CORDEX South Asia RCMs and their ensemble had projected statistically significant strong rate of warming (0.03–0.09 °C per year) across all seasons and RCPs over the Indian Himalayan region (Dimri et al. 2018a). The seasonal response to warming with respect to elevation was found to be substantial with December–January season followed by October–November showing the highest rate of warming at higher elevation sites such as the western Himalayas and northern part of central Himalayas.

The earlier assessment of temperature projections for India using the CMIP5 multi-model and multi-scenario ensemble had also suggested that generally in future the northern part of the country will experience higher warming compared to the southern peninsula (Chaturvedi et al. 2012). It was assessed that the areas in the Himalayas and Kashmir will be particularly subject to large warming to the tune of 8 °C in RCP8.5 by 2099 relative to the pre-industrial baseline (the 1880s). It was concluded that this assessment of a broad range of temperature projections for India, ranging from 1 to 8 °C during the period 1880–2099 under different RCP scenarios indicated that these regional climate change projections were associated with a range of limitations and uncertainties—driven mainly by the climate model and future scenario uncertainties (Chaturvedi et al. 2012). Also, an earlier study using multiple CMIP5 model outputs together with a single model ensemble assessed that for temperature in most regions within India the component of uncertainty due to model spread tends to be larger than that arising due to natural internal variability, and tends to grow with time (Singh and AchutaRao 2018).

A consistent and robust feature across the downscaled CORDEX South Asia RCMs is a continuation of warming over India in the twenty-first century for all the RCP scenarios (Fig. 2.8). The CORDEX South Asia historical RCM simulations capture the observed interannual variations and the warming trend reasonably well. The all India averaged annual surface air temperature increases are similar for all the RCP scenarios during the first decade after 2005. The warming rate depends more on the specified greenhouse gas concentration pathway at longer time scales, particularly after about 2050. The multi-RCM ensemble mean under RCP2.6 scenario stays around 1.5 °C above 1976–2005 levels throughout the twenty-first century, clearly demonstrating the potential of mitigation policies. The ensemble mean annual India warming exceeds 2 °C within the twenty-first century under RCP4.5, and the warming exceeds 4 °C by the end of the twenty-first century under RCP8.5 scenario. The spread in the minimum to maximum range in the projected warming among the CORDEX South Asia RCMs for each RCP scenario (shown as shading in Fig. 2.8) provides a simple, but crude, measure of uncertainty.

A reliable quantitative estimate of all India averaged annual surface air temperature changes and the associated uncertainty range of future temperature projections for India are obtained by incorporating individual model performance and model convergence criteria within a reliability ensemble averaging (REA) methodology (see more details in Box 2.4).

The projected REA estimates of warming for the mid-term (2040–2069) ranges between 1.39 °C under RCP2.6 scenario to 2.70 °C under RCP8.5 scenario (see Table 2.7). The natural variability in the observed all India annual mean surface air temperature is 0.347 °C (see more details in Box 2.4) while the REA based temperature increases are well above this natural variability estimate. The uncertainty range defined by the root-mean square difference varies in the mid-term between 0.17 and 0.37 °C for the three RCP scenarios, with the RCP4.5 mid-term warming of 1.92 °C indicating the maximum uncertainty of 15.6% (see Table 2.7). The proper weighting of individual CORDEX South Asia RCMs based on their present-day performance by the REA method has resulted in the REA warming under RCP2.6 scenario to be 1.33 °C above 1976–2005 levels till the end of the twenty-first century, which is slightly higher than that defined by multi-RCM ensemble mean shown in Fig. 2.8. However, this estimate of annual warming well below 2 °C by the end of the twenty-first century for RCP2.6 is found to be associated with the highest uncertainty of 18% among all the RCP scenarios. The REA estimates of long-term (2070-2099) warming are 2.44 ± 0.41 °C and 4.44 ± 0.45 °C under RCP4.5 and RCP8.5 scenarios, respectively (see Table 2.7). The assessment of 4.44 °C warming by the end of the twenty-first century under RCP8.5 scenario is highly reliable as it is associated with the lowest uncertainty (of 10.1%) among the three RCP scenarios. These CORDEX South Asia multi-RCM ensemble-based temperature projections for India are generally in line with earlier estimates. The CMIP5 ensemble based on 18 models had projected warming of 1.5 °C, 2.8 °C and 4.3 °C under the RCP2.6, RCP4.5 and RCP8.5 scenarios, respectively, for the 30 year average of 2071–2100 relative to the 1961–1990 baseline (Chaturvedi et al. 2012).

The forced signal of warming occurs not only in the annual mean of daily mean surface air temperature but also in the annual means of daily maximum and daily minimum temperatures (see Tables 2.8 and 2.9). The CORDEX South Asia REA estimate of warming for the two 30 year future periods are lower (higher) for the annual means of daily maximum (minimum) temperature than the respective warming assessed for the annual mean of daily mean temperature under all three RCP scenarios. The REA changes for annual minimum temperature of 4.71 ± 0.35 °C (see Table 2.9) is more pronounced than that of 4.10 ± 0.45 °C and 4.44 ± 0.45 °C increases estimated for all India annual maximum (see Table 2.8) and mean (see Table 2.7) temperatures respectively by the end of the twenty-first century under the high (RCP8.5) emission scenario. The assessment of 4.71 °C warming for annual mean of daily minimum surface air temperature by the end of the twenty-first century under RCP8.5 scenario is highly reliable as it is associated with the lowest uncertainty (of 7.4%) among not only the three RCP scenarios for this variable but also for the annual mean and maximum statistic shown in Tables 2.7 and 2.8. This finding illustrates that the dynamically downscaled CORDEX South Asia RCMs based regional climate projections do certainly bring more confidence to future temperature projections for India than the regional climate change information provided by the statistically downscaled NEX-GDDP dataset.

These CORDEX South Asia multi-RCMs had also projected statistically significant higher warming rate (0.23–0.52 °C/decade) for both minimum and maximum air temperatures over the Indian Himalayan region under RCP4.5 and RCP8.5 scenarios (Dimri et al. 2018b).

The CORDEX South Asia multi-RCMs project that all India averaged annual frequency of warm nights and warm days will increase from about 10% in the reference base period (1976–2005) to 80% and 65%, respectively, by the end of the twenty-first century under the high (RCP8.5) emission scenario (Fig. 2.9). The future changes in the percentile indices based on minimum temperature (warm nights and cold nights) are more pronounced than those based on maximum temperature (warm days and cold days). The downscaled future temperature projections under the high (RCP8.5) emission scenario also indicate that by the end of the twenty-first century there will be virtually no cold nights and cold days over India as defined for the reference base period (1976–2005). The spread among the RCMs (shading in Fig. 2.9) generally becomes smaller as the projection approaches the zero exceedance rates as more models simulate fewer cold nights and cold days. The largest decreases in cold nights and largest increases in warm nights projected over India are typical for tropical regions that are characterized by small day-to-day temperature variability so that changes in mean temperature are associated with comparatively larger changes in exceedance rates below the 10th and above the 90th percentiles.

The CORDEX South Asia multi-RCMs project robust increase (decrease) in the all India averaged annual intensity of warm (cold) temperature extremes by the end of twenty-first century, with the magnitude of the changes increasing with increased anthropogenic forcing (Fig. 2.9). The coldest night of the year warms (about 5.5 °C) more than the warmest day (about 4.7 °C) over India by the end of the twenty-first century relative to the reference base period (1976–2005) under the high (RCP8.5) emission scenario. This tendency is consistent with the assessment that the increases in the frequency of warm nights are greater than increases in the frequency of warm days.

The CORDEX South Asia multi-RCM ensemble simulate about one heatwave event with an average total duration of about 5 days per summer season (April to June) over India during the historical period 1976–2005 (Fig. 2.10). These heatwave characteristics are identified based on the 90th percentile threshold of daily maximum air temperature from each CORDEX model outputs after applying a quantile mapping bias-correction following Lafon et al. (2013), which was based on an empirical distribution correction method (Wood et al. 2004). These findings compare well with the heatwave climatology estimated using the CMIP5 multi-model ensemble and IMD observations over north-west India for the present period (Rohini et al. 2019).

The CORDEX multi-RCM ensemble mean project that all India averaged frequency of summer heatwaves will increase to about 2.5 events per season by the mid-twenty-first century (2040–2069), with a further slight rise to about 3.0 events by the end-twenty-first century (2070–2099) under the medium (RCP4.5) emission scenario (Fig. 2.10). The average total duration of summer heatwaves is projected to increase to about 15 and 18 days per season during the mid- and end-twenty-first century respectively under this future scenario. The projected increase in these all India averaged summer heatwave characteristics are similar to the assessment based on CMIP5 multi-model ensemble over north-west India in the period 2020–2064 under the RCP4.5 scenario (Rohini et al. 2019). The CORDEX ensemble mean projects that India averaged frequencies of summer heatwaves will increase to about 3.0 and 3.5 events per season during the mid- and end-twenty-first century, respectively under the high (RCP8.5) emission scenario (Fig. 2.10). The future rise in heatwave frequencies under this high emission scenario is marginally higher than the increase in the number of summer events under the RCP4.5 scenario for the corresponding periods. The average total duration of summer heatwaves in India under the RCP8.5 scenario is assessed to be substantially higher than that under RCP4.5 scenario, with about 25 and 35 heatwave days per season during the mid- and end-twenty-first century respectively (Fig. 2.10). The projected CORDEX ensemble mean change in the frequency of heatwaves for the mid- and end-twenty-first century under RCP8.5 scenario relative to the historical reference period (1976–2005) are higher over the north-west region (more than 3 days per summer season) compared to other parts of India (Fig. 2.11). The CORDEX ensemble members consistently project a substantial increase in the average total duration of heat waves of more than 40 days per summer season relative to the historical reference period over north-west India by end of the twenty-first century under this high emission scenario.

These findings are consistent with results based on the CMIP5 multi-model ensemble assessments that the heatwave frequency, duration, intensity and areal coverage over India will substantially increase during the twenty-first century (Murari et al. 2015; Im et al. 2017; Mishra et al. 2017; Russo et al. 2017; Dosio et al. 2018; Rohini et al. 2019). The CMIP5 ensemble had projected that the future warm season (March to June) severe heat waves in India (based on the IMD criteria) will be more intense, with longer durations and will likely occur at a higher frequency and earlier in the year (Murari et al. 2015). The southern India, currently not influenced by heatwaves, is expected to be severely affected by the end of the twenty-first century. Rohini et al. (2016) reported that in near future more frequent and long lasting heatwave events are projected to affect the Indian sub continent in response to the warming of the tropical Indian Ocean (Roxy et al. 2014, 2015) and the increasing frequency of extreme El Nino events (Cai et al. 2015). A CMIP5 multi-model ensemble-based assessment projected 40 times rise in the frequency of severe heat waves over India than in the present climate by the end of the twenty-first century under the RCP8.5 scenario with a heatwave index that captures both duration and magnitude estimated using CMIP5 model projections of all India averaged daily maximum air temperature without bias-correction (Mishra et al. 2017). This study also assessed using a global earth system model ensemble that the projected frequency of severe heatwaves by the end of the twenty-first century will be about 2.5 times lesser (than for RCP8.5 scenario) if the global mean temperature is limited to the low-warming scenario of 2.0 °C above pre-industrial conditions. However an assessment of the projected population exposure in India due to heat stress (Im et al. 2017) differed considerably from the findings of Mishra et al. (2017) based on a heatwave index without considering the effect of humidity in heatwave estimations.

The projected future heatwaves based on the extremes of wet-bulb temperature, which includes the effect of humidity, are found to be concentrated around densely populated agricultural regions of the Ganges and Indus river basins (Im et al. 2017). The human populations in these regions highly vulnerable to heat stress are assessed to experience maximum daily wet-bulb temperatures exceeding 31 °C by 2100 under RCP4.5 scenario, considered dangerous levels for most humans (Im et al. 2017). These extreme wet-bulb temperatures are likely to approach and, in few locations, exceed 35 °C by the 2070s under RCP8.5 scenario, considered as an upper limit on human survivability (Coffel et al. 2018). The annual probability of occurrence of heat waves with magnitude greater than the one in Russia in 2010 (the most severe of the present era) at 1.5 and 2 °C global warming above pre-industrial levels is assessed using CMIP5 ensemble to be different than zero in few parts of India when measured by means of an heatwave index, which takes into account both temperature and relative humidity (Russo et al. 2017). The yearly probability of occurrence of a heatwave at 4° global warming with magnitude greater than this most severe event in the present period will be greater than 10% in many parts of India, with the southern peninsula expected to experience such type of humid heatwaves with an annual probability greater than 50%, corresponding to an average return period of two years (Russo et al. 2017).