1. Introduction
Responsible for providing approximately 80% of total annual rainfall in major regions of South Asia, the South Asian monsoon (SAM) is the primary freshwater source for more than 22% of the world’s population inhabiting this region (Turner and Annamalai 2012). As is the case with any monsoon circulation, that of South Asia develops due to the differences in the specific heat capacity of continent and ocean (Webster et al. 1998). Other features of the monsoon process, including the westerly jet that characterizes the flow of air onto land, the impact of the topographic barrier comprising the Himalayan Mountains and the Tibetan Plateau (Wu et al. 2007), and additional local features including land–sea or land–lake contrasts and locally complex and steep orography also affect the strength and variability of the monsoon. Since all of these regional influences are poorly represented in global climate models (GCMs), these models have limited ability to simulate the monsoon phenomenon (Turner et al. 2011; S. K. Mishra et al. 2018; Anand et al. 2018). To adequately investigate the nature of expected future climate change over South Asia, it is therefore necessary to conduct climate simulations, which take into account both small-scale features contributing to local climate processes and future changes in the large-scale circulation.
To this end, increasingly numerous and technically demanding climate simulations continue to be conducted over South Asia using state-of-the-art regional climate models (RCMs) running at high horizontal spatial resolution (10–30 km), employed to dynamically downscale global-scale future climate projections derived from global coupled models. A number of recent studies of this kind have suggested that the monsoon circulation of this region may already be recognized as changing over recent decades due to the increasing concentration of atmospheric greenhouse gases (Roxy et al. 2015; A. K. Mishra et al. 2018). The direction and extent of the future changes of the SAM will depend on the rise of Indian Ocean sea surface temperatures (SSTs) relative to the warming over the subcontinent itself. Recent assessments of the monsoonal changes indicate that rising SSTs in the Indian Ocean have led to a weakening of the moisture-bearing winds that blow from the ocean to the land (Roxy et al. 2014), while increased land surface warming has countered this influence upon monsoon strength (Roxy 2017). It has also been suggested on the basis of earlier GCM-based simulations that the strength of the southwesterly summer monsoon winds was projected to decrease (Meehl et al. 2007). On the other hand, the increased SST in the Indian Ocean produces greater evaporation, an increased moisture source for the monsoon. Higher air temperature also means a larger moisture-holding capacity, and a consequent increase in monsoon rainfall (Meehl and Arblaster 2003; May 2004). Many studies have also reported observed changes in the precipitation pattern over India, including an increase in heavy precipitation and a decrease in low or moderate rainfall events (Mishra and Liu 2014; A. K. Mishra et al. 2018). To investigate the expected future changes in both average and extreme precipitation events for both the middle and the end of the current century, our current study analyzed a physics-based mini ensemble of simulations driven by a single GCM projection based upon the use of the National Center for Atmospheric Research Community Earth System Model, version 1 (CESM1).
Another purpose of this study is to compare the projected changes in the extremes of precipitation intensity with the changes in average precipitation, and more important to investigate the validity of the thermodynamical constraint based on the Clausius–Clapeyron (CC) relation (Clapeyron 1834; Clausius 1850) as a means of understanding the increase of not only mean precipitation but also its extremes. If the atmospheric lapse rate remains unchanged, global precipitation and total atmospheric water vapor concentration are expected to increase at the CC reference rate of approximately 7% increase per °C of surface warming (Wentz et al. 2007). Using observations and model simulations, the column integrated water vapor (Santer et al. 2007; Allan et al. 2014) and the global-mean net precipitation (Held and Soden 2006) are found to increase at a rate very close to this thermodynamics-based expectation. However, some other studies have suggested that model projected changes in global-mean precipitation do not scale with CC (Allen and Ingram 2002; Stephens and Ellis 2008). Moreover, studies have decomposed the causes of average precipitation change into dynamic and thermodynamic components, the former related to atmospheric circulation changes and the latter related to atmospheric moisture content changes due to temperature change. Analysis of the large-scale hydrological cycle projected by 15 GCMs found that the change in net precipitation is largely accounted for by the rise in specific humidity that accompanies atmospheric warming (Seager et al. 2010). Specifically in the tropics, Romps (2011) analyzed cloud-resolving simulations with different CO2 concentrations and local precipitation fluxes and found them to obey CC scaling. However, assessment of tropical precipitation anomalies under global warming in GCMs has demonstrated that dynamic feedback can also substantially affect precipitation anomalies within lower atmospheric convergence zones (Chou et al. 2009), although the thermodynamic component is found to remain dominant in large-scale averages.
Likewise, some studies have also analyzed changes in precipitation extremes in a way similar to average precipitation. Model simulations (May 2004; Rupa Kumar et al. 2006) have demonstrated that both the frequency and the intensity of precipitation extremes are projected to increase over India, especially over western and central India. The argument that changes in extreme precipitation may roughly follow the CC relationship and increases at about 7% per degree of warming has been supported by many studies (Allen and Ingram 2002; Kharin et al. 2013). Unlike the local average precipitation, which is constrained by the energy and water vapor budget (Muller and O’Gorman 2011), precipitation extremes are likely to occur when effectively all the moisture in an air volume condenses into precipitation, implying that the intensity of precipitation extremes should depend strongly upon moisture availability and thus its change should closely follow the CC relationship. In the tropics, observations have shown that the increase in local temperatures plays a major role in the rising frequency of extreme rainfall events (Wang et al. 2017). The rate of increase may be even larger than the extratropical rate in the tropics since a large amount of latent heat is released during rainfall events, which may significantly strengthen the updrafts where precipitation occurs (Trenberth 1999). Utilizing the Hadley Centre coupled model, Turner and Slingo (2009a) found that the increases of the monsoon rainfall extremes over India primarily result from the rise of air temperature and associated increases in specific humidity and that these do also approximately follow the CC relationship. Nevertheless, the assessment of such model produced data in the IPCC Fourth Assessment Report shows that changes of the spatial distribution and magnitude of daily monsoon rainfall extremes over South Asia have large uncertainties and depend strongly upon the choice of models and convective parameterization schemes (Turner and Slingo 2009b). In the current study, the projected changes in the extremes of precipitation intensity will be compared with the changes in average precipitation, and more importantly the sensitivity to temperature changes of both will be analyzed.
The current work involves a further use of the dynamical downscaling pipeline first introduced in Gula and Peltier (2012) and subsequently further employed in d’Orgeville et al. (2014), Erler and Peltier (2016), and Peltier et al. (2018), although all of these analyses were dedicated to understanding the expected precipitation changes over eastern and western North America. Using an ensemble of high-resolution RCM projections, Erler and Peltier (2016), in particular, analyzed precipitation changes in western Canada, where the Rocky Mountains play an important role, and found that although winter precipitation over western Canada was projected to increase significantly toward the end of the century, changes in summer precipitation were found to be smaller than in winter and associated with large model uncertainty. Also, d’Orgeville et al. (2014) and Peltier et al. (2018) employed an ensemble of RCM projections for the analysis of changes in summer precipitation extremes in the Great Lakes region of North America. Both find a general precipitation increase and that heavy rainfall changes are projected to be larger than or equal to the thermodynamic expectation of 7% per degree of surface warming. The same dynamical downscaling pipeline has been verified in the context of the SAM by Huo and Peltier (2019, hereinafter HP2019), although their analysis was based upon use of an older version of WRF (V3.4.1) than the one to be employed herein (V3.9.1.1). Moreover, they used a smaller physics ensemble and discussed the mean climate and the precipitation extremes over India in the instrumental era only, thus leaving the global warming impact on the SAM system undiscussed.
The goal of this study is therefore to analyze the possible impact of climate change on SAM precipitation extremes, based on high-resolution regional climate simulations. The next section (section 2) provides a detailed description of the experimental design and datasets used for validation, a reprise of the validation step being necessary here as we will be employing a newer version of the WRF regional climate model and a larger WRF ensemble than in HP2019. Section 3 reviews and updates the detailed analyses recently published in that paper of the ability of our downscaling pipeline to fully reconcile instrumental observations, as well as presenting projected changes in the seasonal average precipitation over different subregions of the Indian landmass. Section 4 discusses results from the extreme value analysis, and a summary of our conclusions concerning expected future changes in the SAM precipitation under global warming is offered in section 5.
2. Models, simulations, and climate change projections
The RCM employed for dynamical downscaling in the current work is the WRF Model (Skamarock and Klemp 2008), version 3.9.1.1. The first four distinct WRF physics configurations to be employed in the current study are the same as those previously employed by HP2019 in their analyses of instrumental era data, but an older version of the WRF model (version 3.4.1) was employed in this previous paper. We have also increased the ensemble size by adding two more ensemble members in the current study (Table 1). The major physics parameterizations to be employed include the Mellor–Yamada–Nakanishi–Niino level-2.5 planetary boundary layer scheme (Nakanishi and Niino 2009); the Rapid Radiative Transfer Model for GCMs radiation scheme (Iacono et al. 2008); the Kain–Fritsch scheme (Kain 2004), the Grell-3 ensemble scheme (Grell and Dévényi 2002), the Tiedtke scheme (Tiedtke 1989) or the Betts–Miller–Janjić (BMJ) scheme (Janjić 1994) for cumulus parameterization; the Noah land surface model (Noah LSM; Tewari et al. 2004) or the Noah LSM with multiparameterization options (Noah-MP; Niu et al. 2011) for land surface parameterization; and the single-moment 6-class (WSM6) scheme (Hong and Lim 2006) or the Morrison two-moment (Morrison et al. 2009) scheme for microphysics parameterization. Table 1 summarizes the specific six different mixtures of physics parameterizations employed to define the different members of the physics ensemble, which provides the means whereby we may study the impact of different physics parameterization schemes and estimate the uncertainty associated with future projections of the strength of the regional warming process: the first ensemble member can be compared with the second to investigate the effect of land surface model; the first, third, fourth, and fifth ensemble members can be compared for the impact of the choice of cumulus scheme; and comparison between the first and sixth ensemble members can be used for studying the impact of different microphysics schemes.
Simulations of the physics ensemble and their selected parameterization.
This study applies a downscaling pipeline based upon use of a two-step nesting procedure (see section S1 in the online supplemental material for further information), details of which have been most recently fully discussed in Peltier et al. (2018). The positioning of the two nested domains is illustrated in Fig. 1: the first nested outermost domain covers most of Asia and some of the surrounding ocean at 30-km resolution with 224 grid points in the east–west and north–south directions, respectively, and the inner domain covers most of the Indian subcontinent south of the Himalayas at 10-km resolution with 249 grid points in the east–west direction and 285 grid points in the north–south direction.
The configuration of the outer regional climate model (WRF) domain, which encompasses most of Asia, as well as the inner domain, situated over the Indian subcontinent. In the outer domain the resolution of WRF is 30 km, and in the inner domain the resolution is 10 km. The shading shows the JJAS average temperature (K).
Citation: Journal of Climate 33, 6; 10.1175/JCLI-D-19-0268.1
The single GCM simulation that is employed to drive the dynamical downscaling experiments has been produced using the Community Earth System Model, version 1.0.4, in its default coupled configuration and was run at the default horizontal resolution of 1° (Gent et al. 2011). This CESM1 simulation includes a historical period (1870–2005) and a projection period (2006–2100). During the historical period, the GCM is driven by a combination of anthropogenic and natural forcings, whereas during the projection period the global simulation follows the representative concentration pathway 8.5 (RCP8.5) forcing scenario. The CESM1 simulation has been dynamically downscaled for 15 years for a historical period (1979–94), a future period in the mid-twenty-first century (2045–59), and an additional future period for the end of the century (2085–99).
Owing to a scarcity of individual observational records of sufficient length, it is difficult to detect changes in precipitation extremes, even with decadal return periods, let alone those with longer return periods, which are expected to be significantly more extreme. To overcome this problem in our relatively small ensemble, we have employed a pooling technique to combine data from different meteorological stations (or model grid points near station locations) based on the similarity of their climate. It is clearly inappropriate to combine the data over the entire domain due to the heterogeneity of the climate over the Indian subcontinent. An objective way in which we may define such pooling regions is to apply a clustering algorithm based on average climatic characteristics of the stations. By using pooled data from station clusters for the extreme value analysis (EVA) in the historical period, we have previously demonstrated (HP2019) that this technique results in good fits to the observational data and statistically meaningful results. We will therefore employ similar clusters as identified in that work to analyze data for the projection periods in the current discussion (see Fig. 2), where these clusters, originally identified through application of the k-means clustering algorithm correspond to locations from central India (blue), from southern India (green), and from the west coast area (red) respectively.
Station locations with topography. The symbol color indicates station cluster association clusters 1 (blue), 2 (green), and 3 (red). Stations with more than 300-m elevation error in WRF have black triangular symbols and were excluded from the analysis. The topography of the outer WRF domain is shown in the background (30-km resolution).
Citation: Journal of Climate 33, 6; 10.1175/JCLI-D-19-0268.1
Station-based measurements of daily precipitation from the Global Historical Climatology Network (GHCN) dataset (Menne et al. 2012) were used for validation. Taking data availability and the long-term trend caused by the global warming process into consideration, only observations from 1950 to 2010 were used in this study. Additionally, records from stations separated by less than 10 km have been merged, so that correlations between stations should be small. Furthermore, to mitigate disagreement between model simulations and station observations due to inconsistent topography, 43 stations with elevation error larger than 300 m in the inner WRF domain have been discarded. The stations used for this analysis are shown in Fig. 2. Note that the analyses to be presented herein make use of data from many more stations than those that were analyzed in HP2019 in that in the previous paper the criteria being employed on station selection were overly stringent. Clusters 1, 2, and 3 in the current analysis consist of 899, 757, and 140 stations respectively, whereas in that previous paper there were only 41, 14, and 3 stations in clusters 1, 2, and 3 respectively (Fig. 2).
3. Validation and future projections of average precipitation
A detailed discussion of the climate simulation over South Asia was presented in HP2019 for the historical physics ensemble. Although some of its conclusions may be repeated here when necessary, this study extends the results to the mid- and end-of-twenty-first-century periods.
Note that the numerical biases and changes reported in this section are averages over the land surface area of the inner WRF domain. The seasonal averages of the historical simulations of the six physics configurations are validated against the Climatic Research Unit (CRU; Harris et al. 2014) time series datasets, version 3.20, for 2-m air temperature in Fig. 3 and total precipitation in Fig. 4. Their differences were computed after the reprojection of the CRU dataset to the native grid of each model.
(top left) JJAS average of temperature from the CRU 1980–94 dataset. Also shown are the JJAS averaged biases of the CESM1 global simulation and six simulations that constitute the WRF historical physics ensemble (K).
Citation: Journal of Climate 33, 6; 10.1175/JCLI-D-19-0268.1
(top left) JJAS average daily precipitation from the CRU 1980–94 dataset. Also shown are JJAS averaged biases of the CESM1 global simulation and six simulations that constitute the WRF historical physics ensemble (mm day−1).
Citation: Journal of Climate 33, 6; 10.1175/JCLI-D-19-0268.1
a. Validation against historical observations
The June–September (JJAS) temperature bias of the CESM1 simulation and the six WRF physics ensemble members against the CRU observations is shown in Fig. 3. The distribution of the observed JJAS mean temperature over South Asia is also shown in Fig. 3, top left. Temperature biases in JJAS are shown to largely depend on the models and model configurations. CESM1 has a cold bias over the high topography in the southwestern part of the subcontinent and a warm bias in the north, which is somewhat inherited by all the WRF ensemble members. Beyond that, CESM1 and the third WRF ensemble member both have a cold bias of around 1°C when integrated over the land surface in the WRF inner domain, while the first and fourth WRF ensemble members both show a warm bias of around 0.7°C (see Table S1 in the online supplemental material). Among all ensemble members, the smaller average biases of the fifth ensemble member using the BMJ cumulus scheme (0.05°C) and the sixth ensemble member using the Morrison microphysics scheme (0.16°C) are a consequence of opposite and compensating biases in central and southern India. The second ensemble member using the Noah-MP land surface scheme has the third smallest average bias (−0.33°C) and the smallest average root-mean-square error (1.14°C) over the surface of the subcontinent and thus can be considered to have the best representation of JJAS temperature. Table S1 shows more details of the average biases of different regions over India based on the station clustering method introduced in section 4. In comparison with the results obtained using WRF V3.4 in HP2019, the WRF ensemble produced by WRF V3.9 generally has a superior representation of summer temperature over India (see section S2 in the online supplemental material for further discussion).
The distribution of average JJAS total precipitation in observations and the differences of the six WRF ensemble members with the observational dataset are shown in Fig. 4. The JJAS precipitation bias of CESM1 is also shown. The predominant orographic forcing of the Western Ghats and a large-scale gradient away from the west coast are evident in the spatial distribution of the Indian summer monsoon precipitation. This results in very high precipitation rates along the windward side of the coastal mountain ranges and much less precipitation in their rain shadows over the interior subcontinent. However, the representation of orographic precipitation is strongly dependent on model resolution: the rain shadow over the interior of the subcontinent is almost nonexistent in CESM1, resulting in a large overestimation in the southeast. The distribution of biases in extreme precipitation (99th percentile) produced by CESM1 is similar to the average daily precipitation in terms of spatial characteristics, but is characterized by larger magnitudes (section S3 and Fig. S1 in the online supplemental material). This underestimation of the rain shadow effect is greatly improved in WRF at 10-km resolution. Five of the ensemble members are characterized by a small or moderate overestimation of total precipitation averaged spatially over the surface of the subcontinent, which is smaller than 0.6 mm day−1 for the first and second WRF ensemble members, around 1 mm day−1 for the fifth and sixth ensemble members, and around 2 mm day−1 for the third ensemble members (Table S2 in the online supplemental material). Only the fourth ensemble member using the Tiedtke cumulus scheme features an overall dry bias (−1.3 mm day−1). Summer precipitation differences are mainly due to the different cumulus parameterization schemes employed: the third ensemble member with the Grell-3 scheme and the fifth ensemble member with the BMJ scheme tend to have excessive precipitation in JJAS while the Tiedtke cumulus scheme in the fourth WRF ensemble member generally has a tendency to produce somewhat less monsoon precipitation; the Kain–Fritsch scheme used in the first and second ensemble members produces a result closer to the observations (Fig. 4). The spatial patterns of these JJAS precipitation biases for the Kain–Fritsch cumulus scheme (the first and second ensemble members) and the Grell-3 scheme (the third ensemble member) are characterized by a large overestimation in front of the rain barrier west of the Western Ghats and an excess in the northeast of the Indian subcontinent, which agrees with the bias patterns found by Mukhopadhyay et al. (2010) using the Kain–Fritsch cumulus scheme in an older version of WRF (V2.2). They also found that the BMJ convective parameterization scheme is able to produce a better mean monsoon pattern in the WRF model, which is different from our finding. For the fifth ensemble member using the BMJ cumulus scheme, the overestimation along the Western Ghats is reduced, but there is a larger wet bias over southern India. Conversely, for the fourth ensemble member with the Tiedtke scheme, all these precipitation excesses are muted, whereas there is a lack of precipitation in general. The choice of microphysics schemes and land surface schemes can also affect the magnitude of the precipitation biases: both the second ensemble member with the Noah-MP land surface scheme and the sixth ensemble member with the Morrison microphysics scheme produce a slightly larger moist bias when compared with the first ensemble member, mainly over the northeast of the Indian subcontinent. Despite a few small-scale defects, the first ensemble member with the Kain–Fritsch cumulus scheme can be considered to best reproduce the SAM precipitation as a whole. The distribution of summer precipitation biases of the WRF ensemble produced by WRF V3.9 is generally similar to the results obtained using WRF V3.4 in HP2019. Further discussion is provided in section S4 of the supplemental material.
Construction of a Hovmöller diagram that integrates the spatial and temporal analyses allows us to visualize the motion of the intertropical convergence zone and monsoon rainfall during the event. In the following analysis, a Hovmöller diagram is constructed by zonally averaging the data between 70° and 90°E. The analysis is performed only over land to exclude the coastal ocean precipitation and to take advantage of the CRU dataset that is only available over land. Figure 5 shows the Hovmöller diagrams for the 15-yr mean annual cycle of precipitation for the CRU datasets in 1980–94, and the WRF historical and two projection simulations of the first ensemble member. For all plots, there is a distinct northward motion of the monsoon precipitation from May to August, and then a southward motion from August to November. Overall, the latitudinal motion of the SAM is well captured by WRF, but the intensity of precipitation is overestimated in the WRF historical simulation (top right) during the monsoon period in southern India, which is in agreement with the wet bias in the JJAS average precipitation distribution.
Hovmöller diagrams for the 15-yr average daily precipitation (mm day−1) between 70° and 90°E on land only for (top left) the CRU 1980–94 dataset and (top right) historical and (bottom left) mid- and (bottom right) end-of-twenty-first-century simulations of the first WRF ensemble member.
Citation: Journal of Climate 33, 6; 10.1175/JCLI-D-19-0268.1
b. Precipitation changes in South Asia
A major objective of this study is to document the projected changes of the SAM precipitation, particularly the extremity of extreme precipitation events and the return time of events of a given extremity. Specifically, the analysis we have performed provides answers to the following questions: 1) Will the SAM precipitation increase or decrease by midcentury and at the end of the century? 2) Will the tail of the distribution of monsoon precipitation extremes fatten? 3) Will the changes of extreme precipitation intensity scale with the change in the surface temperature at the rate expected by the CC relation?
To begin to answer these important questions, the JJAS precipitation changes delivered by the future projections of the physics ensemble by midcentury and at the end of the century are presented in Fig. 6. Obviously, all WRF physics ensemble members for all time horizons and the CESM1 simulation project an overall increase in rainfall, which is a general characteristic of CMIP5 projections (Menon et al. 2013) and consistent with the previous dynamically downscaled projections by Bal et al. (2016) at 25-km resolution. The local trends clearly vary from place to place, with regional patches of decreased precipitation projected to be surrounded by large areas of increased precipitation. Moreover, comparing results from different WRF physics ensemble members, regional changes within the WRF inner domain are evident and constitute a climate change signal independent of the physics configuration. Specifically, by the mid-twenty-first century (Fig. 6, top), areas with a drying trend concentrate in the northern part of the inner domain for all physics configurations. This is consistent with results of some previous studies (Jayasankar et al. 2018; Sandeep et al. 2018), although the areas with the projected drying trend in the present study are more restricted geographically. For the first, second, fifth, and sixth ensemble members, regions with decreased JJAS precipitation mainly lie in the northwest corner of the domain. From the Student’s t test, these drying trends are statistically significant at the 90% confidence level. Actually, this region of precipitation decline also exists in the global CESM1 simulation, which is somewhat muted relative to the WRF high-resolution results and is not statistically significant. The second–fourth ensemble members all have small areas with decreased JJAS precipitation in the mountain ranges in central India (the Vindhya Range), although the decreasing trends in these three small regions are not statistically significant. In contrast, the west coast and southern India are characterized by a consistent increase in JJAS precipitation by mid-twenty-first century for all simulations. The increase in absolute precipitation over southern India is consistently smaller than that over the west coast, although the amount of the increase in JJAS precipitation depends on the simulation. This spatial divide by the mid-twenty-first century between the north and the south is also clear at the end of the century (Fig. 6, bottom). The first, second, and sixth physics ensemble members project increased precipitation everywhere at the end of the twenty-first century, while the third, fourth, and fifth ensemble members still have regions in the north characterized by a drying trend, which is, however, statistically insignificant at the 90% confidence level. Moreover, averaged over the land surface area of the inner WRF domain, the projected change in JJAS precipitation produced by the fourth ensemble member is around 2 mm day−1 by the end of the century and is the smallest among all the ensemble members, while the third ensemble member produced the largest precipitation increase by the end of the twenty-first century (4 mm day−1). Thus, despite these small-scale variations, the JJAS precipitation over India is generally projected to increase and this result is robust across all members of the WRF physics ensemble, but the amount of the increase in monsoon precipitation depends on the individual ensemble member.
Future JJAS average daily precipitation changes (mm day−1) from future experiments at (top),(top middle) midcentury and (bottom middle),(bottom) the end of the century for the six physics ensemble members and the CESM1 simulation.
Citation: Journal of Climate 33, 6; 10.1175/JCLI-D-19-0268.1
4. Station clustering and the mean annual cycle
a. Station clustering
To investigate the shape of the tail of the daily rainfall distribution, EVA, which involves fitting a specific distribution to a sample of extreme events, is applied. A huge sample size is necessary for estimating the distribution parameters, especially at the extreme ends of the distribution. For instance, a record that is very much longer than 100 years would be desirable for the estimation of the magnitude of a 100-yr return period event. Nevertheless, there are very few meteorological stations with records of such length. Thus, aggregating observations from similar stations is necessary to increase the effective size of the record. Precipitation over the Indian subcontinent is characterized by large spatial variability, so the distribution parameters vary greatly, making it impossible to combine all stations. To define station clusters based on the similarity of the station’s climatological seasonal cycle, each station was first represented by a vector comprised of its climatological annual cycle of monthly mean precipitation (12 values per station) and then the k-means clustering algorithm was employed. Figure 2 presents the station locations and the clusters to which they belong. Three clusters in total were defined and used for this analysis; the number of clusters is decided on the basis of experimentation. The number of clusters is somewhat arbitrary, but certain types of clusters occur robustly, regardless of the number of clusters: the first division is between stations at the west coast and inland stations, which differ greatly in their summer precipitation amounts; the next division that emerges through objective application of the k-means algorithm is between the stations in northern and central India and the stations in southern India, which differ significantly in their time of peak precipitation. Note that this clustering algorithm does not take precipitation extremes into consideration. Also note that clustering was performed only on the observed station data; model data were not employed by the algorithm in determining the makeup of the individual clusters. At each station location, the closest model grid point is assigned to the same cluster as the station. The total number of stations used herein is larger than that in HP2019 because a more stringent method of station selection was employed in that paper and only stations with more than 50 years of data were employed. However, in the current analysis, all stations in the inner WRF domain (except those with elevation error larger than 300 m) are used, leading to a significantly increased number of data points that constrain the fit parameters. The panel headers in Figs. 7–9 (as well as in Figs. 12 and 13, described below) provide the number of stations and observational data points in each cluster for the corresponding season; the number of data points for the WRF simulations can always be calculated as the number of stations in each cluster multiplied by 15, which is the simulation length in years.
Seasonal cycle of average daily precipitation for station clusters 1, 2, and 3 for the first WRF ensemble member. The error bars show the standard error of the climatological mean (σ/n1/2). The number of stations (indicated with #) and the number of samples (yr) for each cluster are indicated above each plot.
Citation: Journal of Climate 33, 6; 10.1175/JCLI-D-19-0268.1
Distribution of JJAS (clusters 1 and 3) and SON (for only cluster 2) maxima of average daily precipitation in the three station clusters for observations (black) and results from the WRF outer (blue) and inner (red) domains of the historical simulation. The dashed lines are fits that were bias corrected so that the mean and variance of the model distribution matches those of the observed distribution. The number of stations (indicated with #) and the number of samples (yr) for each cluster and season are indicated above each plot.
Citation: Journal of Climate 33, 6; 10.1175/JCLI-D-19-0268.1
Distribution of JJAS (clusters 1 and 3) and SON (for only cluster 2) maxima of daily precipitation in the three station clusters for observations (black) and results from the WRF historical (blue), midcentury (green), and end-of-century (red) simulations. The dashed lines are fits that were bias corrected so that the mean and variance of the model distribution matches those of the observed distribution. The number of stations (indicated with #) and the number of samples (yr) for each cluster and season are indicated above each plot.
Citation: Journal of Climate 33, 6; 10.1175/JCLI-D-19-0268.1
b. The average seasonal cycle
The climatological seasonal cycles of monthly mean precipitation for station clusters 1, 2, and 3 (northern and central India, southern India, and the west coast) are displayed in Fig. 7. Observational data are shown as black curves with error bars, and downscaled data from the outer and inner domains of the first WRF ensemble member are shown as solid and dashed lines respectively: blue lines for the historical period, green lines for the midcentury period, and red lines for the end-of-century period. The error bars and bands are based on the standard error of the mean (SEM), which is related to the sample standard deviation s by SEM = s/n1/2, where n is the number of samples (the number of years in this case) in each station cluster.
In terms of monthly mean total precipitation (Fig. 7), summer monsoon precipitation at the west coast and in northern and central India is underestimated (quite substantially for the west coast), whereas it is overestimated for southern India when compared to station observations. It is interesting to note that in southern India the total precipitation amount and the precipitation change in autumn [September–November (SON)] are larger than those in JJAS, whereas both are the largest in summer for the other two clusters. In general, the first WRF ensemble member reproduces the seasonal cycle of precipitation well; only the station cluster at the west coast has a significant (low) bias in summer. All clusters are characterized by a significant increase in mean monsoon precipitation intensity. At the end of the century, average daily precipitation is projected to increase by approximately 40% in JJAS for cluster 1, 30% for cluster 3, and 45% for cluster 2 in SON.
5. Analysis of precipitation extremes
a. Extreme value analysis and bias correction
b. Distributions of precipitation extremes
In Fig. 8, PDFs of seasonal maxima of daily precipitation are shown for the outer (blue; 30-km resolution) and inner (red; 10-km resolution) domains of the first WRF ensemble member in the validation period (1979–94). In Fig. 9, similar results are shown but for only the inner (10-km resolution) domain of the first WRF ensemble member in the validation period and two projection periods (2045–59 and 2085–99). Dashed lines represent the bias-corrected model distributions. Also shown are GEV distributions fitted to the observational precipitation maxima (black). The mean of the observational distributions appears to be significantly greater than that of the model distributions before bias correction; the magnitude of simulated extremes is around 70% of the observed extremes in all three clusters. This is certainly associated with the fact that station observations are point values, while model output must be interpreted as representing gridcell averages. The reduction in extreme precipitation intensities due to spatial aggregation has been investigated by Eggert et al. (2015), using a radar composite over Germany and comparing extremes at different scales. They found strong reduction for convective precipitation extremes (30% stronger than for stratiform events), which may partially explain the reduction in the magnitude of monsoon precipitation extremes, which is dominated by convective precipitation.
The model output apparently fits the observational distribution the best in cluster 1 (northern and central India), a region characterized by modest variations in topographic relief. The quality of fit in southern India in autumn is much lower, and in this region a clear dependence on resolution is apparent. In other words, owing to the more rugged terrain and long and complex coastline in southern India, very high spatial resolution is desirable and data from the inner WRF domain can achieve a better fit. This is also true for the model output in cluster 3 at the west coast. These results suggest that summer precipitation is parameterized somewhat adequately in the model, and the improvement due to higher resolution in relatively flat regions seems to be small. On the contrary, higher resolution is superior for clusters 2 and 3, in which topography is more complex.
The projected PDFs for the two future time periods (2045–59 and 2085–99) are shown in Fig. 9. The simulated distributions for the projection periods have been bias corrected using the same method as for the historical distributions so that they can be usefully compared. Evidently, there is an increase in the distribution means of precipitation extremes in all three clusters. This is consistent with the projected increase in high to extreme rainfall by IPCC AR5 climate models (Jayasankar et al. 2015) and the recent observed increasing trend in heavy precipitation over India (Goswami et al. 2006; A. K. Mishra et al. 2018). Comparing the results for different time horizons, the increase in the extreme precipitation intensity is maximized at the end of the century when the strongest warming is occurring. The changes in the distribution means of precipitation extremes are smaller than the changes in the average precipitation: roughly 30% increase in JJAS extremes in northern and central India, 20% at the west coast, and 40% increase in fall in southern India by the end of the century.
c. Validity of the CC scaling as a reference to detect changes in extreme precipitation
Considering the changes in the distribution means of precipitation extremes and the fact that the temperature increase in JJAS at the west coast is about 3°C at the end of the century (Fig. S2 in the online supplemental material), this implies a 7% °C−1 increase in summer at the west coast—consistent with the CC-based expectation.
However, the projection of the future extreme precipitation increase is larger than the Clausius–Clapeyron reference value for clusters 1 and 2. Since a large portion of rain originates from the midtroposphere, where warming is larger than surface warming in JJAS for cluster 1 and in SON for cluster 2 when most extreme precipitation events occur (Fig. 10; ΔT), there is therefore an artifact if local surface temperatures are used to assess the predictive capacity of the CC relation for extreme precipitation intensity changes. In cluster 1 (northern and central India), using the temperature increase at about 4 km instead of surface warming accounts for most of the additional increase in precipitation extremes. Nevertheless, there is still about 25% of the additional increase in precipitation extremes over southern India (cluster 2) that cannot be explained by this effect even after midtropospheric warming is considered. In fact, moisture flux from the Arabian Sea is projected to increase and there will be strong moisture convergence over southern India in autumn (Fig. 11), which may also contribute to the further extreme precipitation increase. Instrumental observations have already shown that the rise in widespread extremes in the most recent 20 years is associated with increased moisture transport from the Arabian Sea, via low-level monsoon westerlies (A. K. Mishra et al. 2018).
The vertical profiles of future changes in temperature multiplied by 7% °C−1 (ΔT × 7% °C−1) in JJAS (clusters 1 and 3) and SON (for only cluster 2).
Citation: Journal of Climate 33, 6; 10.1175/JCLI-D-19-0268.1
Color shading shows (left) the historical water vapor convergence and the (center) mid- and (right) end-of-twenty-first-century change at (top) 250 and (bottom) 850 hPa in SON [kg (m3 s)−1]. Arrows are the historical (in the left column) and future changes (center and right columns) in water vapor flux.
Citation: Journal of Climate 33, 6; 10.1175/JCLI-D-19-0268.1
d. Extreme quantiles
Figure 12 shows the cumulative density functions of daily precipitation intensities of extreme events with decadal return periods, which illustrates the tail behavior of the distributions. As in Fig. 8, the model results after bias correction are shown as dashed lines. Because the most extreme tails of these distributions are constrained by very few rare events, the sampling uncertainty is larger than for the bulk of the distribution. The return periods for events with a return period of 20 and 50 years in the bias-corrected historical simulation are also listed in Table 2. In cluster 2, the tail behavior of historical and future projections is significantly separated in fall with a larger decreasing rate in return periods for rarer events. In JJAS in northern and central India and at the west coast, the return periods decrease by a factor of approximately 2.5 at the end of the twenty-first century. It is interesting to note that in JJAS at the west coast, the separation between mid-twenty-first century and end-twenty-first century is weak and there is even a small increase in return periods.
Extreme quantiles of maxima of daily precipitation in JJAS (clusters 1 and 3) and SON (for only cluster 2) for GHCN observations (black), WRF historical simulation (blue), and projections in midcentury (green) and at the end of the century (red). The dashed curves are the model distributions that have been bias corrected to match the mean and variance of the observed distribution (equivalent to the dashed lines in Fig. 8). The number of stations (indicated with #) and the number of samples (yr) for each cluster–season are indicated above each plot.
Citation: Journal of Climate 33, 6; 10.1175/JCLI-D-19-0268.1
The return periods (yr) of extreme daily precipitation events, which have a return period of 20 or 50 yr in the historical simulations (bias corrected, for comparison with the observed return periods).
e. Sensitivity to physics parameterizations
To test the robustness of the results, data from the small WRF physics ensemble using different cumulus, land surface, and microphysics parameterizations were also employed in EVA. The distributions of seasonal maxima of daily precipitation based on the other five WRF ensemble members are shown in Fig. 13. The figure is analogous to Fig. 9, except that the historical and projected simulations as well as the observations have been combined and, for clarity, only bias-corrected model distributions are shown.
Distribution of JJAS (clusters 1 and 3) and SON (for only cluster 2) maxima of daily precipitation in the three station clusters (from top to bottom) from the second to sixth WRF physics ensemble members with validation against observations for the historical period (blue), midcentury (green), and the end of the century (red). The dashed lines are fits that were bias corrected so that the mean and variance of the model distribution matches those of the observed distribution. The number of stations (indicated with #) and the number of samples (yr) for each cluster and season are indicated above each plot.
Citation: Journal of Climate 33, 6; 10.1175/JCLI-D-19-0268.1
As is the case of the first WRF ensemble member, the fit of the second ensemble member to observations is quite good after bias correction. In fact, its quality of fit is slightly better than for the first ensemble member in clusters 1 and 3, and is almost the same as in cluster 2. Since the only difference between these two ensemble members lies in the land surface scheme, results reproduced by the more complex Noah-MP scheme can thus capture the full distribution of precipitation more accurately than the default and simple Noah LSM, even if these two land surface schemes produce very similar monthly climatology. However, the second ensemble member still appears to have a relatively large mismatch with observations in cluster 2.
It is evident that significant differences between the third ensemble member and observations occur in clusters 1 and 3 during JJAS. The differences occur despite bias correction and are most prominent in northern and central India, where the distribution is too narrow and the tail is too thin, implying that the most extreme events have lower probability than the observed distribution; at the west coast, on the other hand, the distribution is slightly too wide and the tail is too fat. Since the only difference from the first ensemble member lies in the cumulus scheme, the differences in the distribution shape can only be attributed to the Grell-3 scheme used in the third ensemble member. It should be noted here that JJAS mean precipitation modeled by the Grell-3 scheme also shows larger moist bias in northern and central India and at the west coast than the first WRF ensemble member. In southern India, on the contrary, the third ensemble member shows the smallest dry bias in summer mean precipitation (Fig. 4) and its extreme precipitation distribution also fits the observations the best among the six ensemble members before bias correction.
For the fourth ensemble member, bias correction greatly improves the quality of fit in clusters 2 and 3, implying that the biases produced by the Tiedtke scheme are well covered and corrected by adjusting the mean and variance of the distribution. The quality of fit is actually slightly better than the first ensemble member in clusters 1 and 2 after bias correction. However, departures from the observations are most prominent at the west coast (cluster 3), where the distribution is too narrow and the tail is too thin. This mismatch is consistent with the large dry bias in JJAS mean precipitation (Fig. 4), again implying that the Tiedtke scheme used in the fourth ensemble member is the least active and has an overall tendency to produce less mean precipitation and less extreme events than observations.
The fifth ensemble member also manages to achieve a quite good fit to observations after bias correction, although it does not fit the observed distribution of cluster 1 as well as the first ensemble member. However, the quality of fit is actually slightly better than the first ensemble member in clusters 2 and 3, although the average precipitation produced by the BMJ cumulus scheme is characterized by a larger moist bias over southern India (2 mm day−1; Fig. 4).
The quality of fit of the sixth ensemble member is very similar to that of the first ensemble member in all three clusters. Since the only difference between these two ensemble members lies in the microphysics scheme, this implies that the microphysics scheme does not play as important a role as the cumulus scheme in simulating precipitation extremes. In fact, in comparing the first ensemble member using the WSM6 microphysics scheme and the sixth ensemble member using the Morrison scheme, we see that they also produce similar JJAS climatology (Fig. 4) and average precipitation changes (Fig. 6).
Nevertheless, despite the mismatch with observations, all the WRF ensemble members feature an increase in the distribution means of precipitation extremes and a fattening of the distribution tails in all three clusters. In other words, the climate change responses in precipitation extremes are essentially the same in all WRF ensemble members, so the climate change signal can be regarded as robust. However, there are still uncertainties in the projected changes in extreme precipitation, especially over southern India. Table 3 shows the changes in the distribution means of precipitation extremes by the end of the twenty-first century over the three Indian regions based on the bias-corrected results from the WRF inner domain. By the end of the twenty-first century, the largest increase is projected by the second ensemble member using the Noah-MP land surface scheme (46%), while the third ensemble member projects the smallest increase rate (23%). There is less uncertainty in extreme precipitation increase over northern and central India by the end of the century: only the third ensemble member projects an increase in rate by over 40%, while the increase rates projected by the remaining five ensemble members are all between 31% and 38%. Along the west coast, four of the ensemble members (the first, second, fourth, and sixth) agree that the projected increase is around 20% by the the end of the twenty-first century, while the Grell and BMJ cumulus scheme used in the remaining two ensemble members project the extreme precipitation to increase by 34% and 40% respectively.
The changes in the distribution means of precipitation extremes by the end of the twenty-first century over the three Indian regions on the basis of the bias-corrected results from the WRF inner domain.
6. Summary and conclusions
An initial physics-based mini ensemble of WRF dynamical downscaling analyses of climate simulations for the Indian subcontinent (HP2019) has been considerably expanded to include climate change projections for both midcentury and the end of the century. An analysis of precipitation extremes has also been presented, based on a small physics ensemble that was dynamically downscaled to 10-km resolution in six different configurations. Validation against station observations and projections by the end of the twenty-first century under the RCP8.5 scenario were discussed. Stations were combined into three clusters based on their similarity of mean climatological seasonal cycle of monthly mean precipitation, and data from stations in the same cluster were pooled for the extreme value analysis. WRF is able to reproduce the observed distribution of summer and fall precipitation extremes after linear bias correction in all three clusters.
The goal of this work has been to investigate future precipitation change in the Indian summer monsoon region. We find that in our domain of interest, South Asia, independently of the physics configuration and time horizon, the average monsoon precipitation will increase in the future, and also there is a fattening of the tail of the distribution of monsoon precipitation extremes. Both total precipitation and precipitation increase peak in JJAS for northern and central India (cluster 1) and the west coast (cluster 3), and in SON for southern India (cluster 2), leading to an increase in flood risk. Extreme precipitation intensity is projected to increase by about 30% for cluster 1, 20% for cluster 3 in summer, and about 40% for cluster 2 in fall.
The projected increase in extreme precipitation intensity is larger than the CC reference value of 7% increase per degree Celsius of surface warming over most of India. Larger midtropospheric warming than surface warming is the primary factor responsible for this additional increase in rate. Dynamical effects related to increased moisture flux from the Arabian Sea are also likely to exacerbate precipitation extremes.
Details of the physics parameterization (precipitation microphysics and land model) appear to play only a secondary role in determining the future changes. The more complex Noah-MP land surface model reduces the JJAS temperature overestimation and captures the distribution of extreme precipitation more accurately than the default and simple Noah LSM. Among all the ensemble members, the Kain–Fritsch cumulus scheme can be considered to best reproduce the average SAM precipitation as a whole. As compared with the Kain–Fritsch cumulus scheme, the Grell-3 scheme used in the third ensemble member tends to produce a large moist bias along the west coast and in the northeast of the domain in JJAS; the BMJ scheme of the fifth ensemble member produces a larger moist bias over southern India, while the Tiedtke scheme in the fourth ensemble member has a general dry bias. However, the Grell-3 scheme appears to be the best in representing the average monsoon precipitation and extreme precipitation distribution over southern India. The Morrison microphysics scheme produces climatology and precipitation changes that are very similar to those of the WSM6 scheme used in the first ensemble member, implying that the choice of microphysics does not impact the SAM precipitation simulation as greatly as does the cumulus scheme.
Despite different observational biases in average precipitation and precipitation extremes, the physics ensemble presented herein provides a very coherent projection of the future increases both in average and extremes. Specifically in terms of the spatial pattern of projection, the absolute precipitation increase over southern India is consistently smaller than that over northern and central India and the west coast, although the amount of the increase in JJAS precipitation differs among ensemble members, suggesting uncertainty in the projection of changes in the monsoon precipitation intensity. The uncertainty in extreme precipitation increase is larger over regions with more complex topography (southern India and the west coast) than over northern and central India. Moreover, most of the ensemble members agree well with each other in terms of the increase in rate while only one (two) of the ensemble members projects much larger increase in extreme precipitation over northern and central India (the west coast). Ensemble averaging with the assignment of weights to different ensemble members based on model evaluation is one of the methods to address such uncertainty and can be used in future analyses for regional-scale impact assessment (Ghosh and Mujumdar 2009). It must also be noted that the quantitative estimates also have large uncertainties associated with them due to limited number of high-resolution regional model simulations from a single RCM.
Furthermore, because of the nonlinear dynamical processes intrinsic to the atmosphere, there are variations between independent GCM integrations under the same forcing scenario, which are commonly associated with the influence of natural variability (Flato et al. 2013). In terms of climate change in a particular region (e.g., South Asia), the uncertainty arising from natural climate variability on regional scales has been estimated to be as large as model uncertainty, and is larger than forcing uncertainty for mid-twenty-first century projections (Deser et al. 2012). However, since our WRF physics ensemble has been forced by a single global simulation, further ensembles designed to more fully capture the variability associated with initial conditions from the GCM will be needed to confirm the robustness of this result and to characterize the influence of natural variability.
Moreover, it will also be desirable to extend this study with an ensemble of simulations with forcing produced by simulations based upon the use of different GCMs and under different forcing scenarios, which are representative of the range of expected changes in the region of interest so as to further assist in estimating and reducing the uncertainties.
Acknowledgments
The simulations presented in this paper were performed at the SciNet High Performance Computing facility at the University of Toronto, which is a component of the Compute Canada HPC platform. The authors thank Dr. A. Erler for assistance with the initial setup of the modeling chain and the SciNet team for assistance during the setup and operation of WRF. The research of author Peltier at the University of Toronto is supported by NSERC Discovery Grant 9627. The CRU TS3.10 and GHCN daily observational datasets employed in this study are publicly available online (http://www.cru.uea.ac.uk/cru/data/hrg/ and https://www.ncdc.noaa.gov/ghcnd-data-access, respectively).
REFERENCES
Allan, R. P., C. L. Liu, M. Zahn, D. A. Lavers, E. Koukouvagias, and A. Bodas-Salcedo, 2014: Physically consistent responses of the global atmospheric hydrological cycle in models and observations. Surv. Geophys., 35, 533–552, https://doi.org/10.1007/s10712-012-9213-z.
Allen, M. R., and W. J. Ingram, 2002: Constraints on future changes in climate and the hydrologic cycle. Nature, 419, 228–232, https://doi.org/10.1038/nature01092.
Anand, A., S. K. Mishra, S. Sahany, M. Bhowmick, J. S. Rawat, and S. K. Dah, 2018: Indian summer monsoon simulations: Usefulness of increasing horizontal resolution, manual tuning, and semiautomatic tuning in reducing present-day model biases. Sci. Rep., 8, 3522, https://doi.org/10.1038/s41598-018-21865-1.
Bal, P. K., A. Ramachandran, K. Palanivelu, P. Thirumurugan, R. Geetha, and B. Bhaskaran, 2016: Climate change projections over India by a downscaling approach using PRECIS. Asia-Pac. J. Atmos. Sci., 52, 353–369, https://doi.org/10.1007/s13143-016-0004-1.
Chou, C., J. D. Neelin, C.-A. Chen, and J.-Y. Tu, 2009: Evaluating the “rich-get-richer” mechanism in tropical precipitation change under global warming. J. Climate, 22, 1982–2005, https://doi.org/10.1175/2008JCLI2471.1.
Clapeyron, M. C., 1834: Mémoire sur la puissance motrice de la chaleur (Thesis on the motive power of heat). J. Ec. Polytech., 23, 153–190.
Clausius, R., 1850: Ueber die bewegende Kraft der Wärme und die Gesetze, welche sich daraus für die Wärmelehre selbst ableiten lassen (On the motive power of heat and the laws that can be deduced therefrom regarding the theory of heat). Ann. Phys., 155, 500–524, https://doi.org/10.1002/ANDP.18501550403.
Deser, C., A. Phillips, V. Bourdette, and H. Teng, 2012: Uncertainty in climate change projections: The role of internal variability. Climate Dyn., 38, 527–546, https://doi.org/10.1007/s00382-010-0977-x.
d’Orgeville, M., W. R. Peltier, A. R. Erler, and J. Gula, 2014: Climate change impacts on Great Lakes Basin precipitation extremes. J. Geophys. Res. Atmos., 119, 10 799–10 812, https://doi.org/10.1002/2014JD021855.
Eggert, B., P. Berg, J. Haerter, D. Jacob, and C. Moseley, 2015: Temporal and spatial scaling impacts on extreme precipitation. Atmos. Chem. Phys., 15, 5957–5971, https://doi.org/10.5194/acp-15-5957-2015.
Erler, A. R., and W. R. Peltier, 2016: Projected changes in precipitation extremes for western Canada based on high-resolution regional climate simulations. J. Climate, 29, 8841–8863, https://doi.org/10.1175/JCLI-D-15-0530.1.
Flato, G., and Coauthors, 2013: Evaluation of climate models. Climate Change 2013: The Physical Science Basis, T. F. Stocker et al., Eds., Cambridge University Press, 741–866.
Gent, P. R., and Coauthors, 2011: The Community Climate System Model version 4. J. Climate, 24, 4973–4991, https://doi.org/10.1175/2011JCLI4083.1.
Ghosh, S., and P. P. Mujumdar, 2009: Climate change impact assessment: Uncertainty modeling with imprecise probability. J. Geophys. Res., 114, D18113, https://doi.org/10.1029/2008JD011648.
Goswami, B. N., V. Venugopal, D. Sengupta, M. S. Madhusoodanan, and P. K. Xavier, 2006: Increasing trend of extreme rain events over India in a warming environment. Science, 314, 1442–1445, https://doi.org/10.1126/science.1132027.
Grell, G. A., and D. Dévényi, 2002: A generalized approach to parameterizing convection combining ensemble and data assimilation techniques. Geophys. Res. Lett., 29, 1693, https://doi.org/10.1029/2002GL015311.
Gula, J., and W. R. Peltier, 2012: Dynamical downscaling over the Great Lakes Basin of North America using the WRF regional climate model: The impact of the Great Lakes system on regional greenhouse warming. J. Climate, 25, 7723–7742, https://doi.org/10.1175/JCLI-D-11-00388.1.
Harris, I., P. D. Jones, T. J. Osborn, and D. H. Lister, 2014: Updated high-resolution grids of monthly climatic observations—The CRU TS3.10 dataset. Int. J. Climatol., 34, 623–642, https://doi.org/10.1002/joc.3711.
Hawkins, E., T. M. Osborne, C. K. Ho, and A. J. Challinor, 2013: Calibration and bias correction of climate projections for crop modelling: An idealised case study over Europe. Agric. For. Meteor., 170, 19–31, https://doi.org/10.1016/j.agrformet.2012.04.007.
Held, I. M., and B. J. Soden, 2006: Robust responses of the hydrological cycle to global warming. J. Climate, 19, 5686–5699, https://doi.org/10.1175/JCLI3990.1.
Hong, S.-Y., and J.-O. J. Lim, 2006: The WRF single-moment 6-class microphysics scheme (WSM6). J. Korean Meteor. Soc., 42, 129–151.
Huo, Y., and W. R. Peltier, 2019: Dynamically downscaled climate simulations of the Indian monsoon in the instrumental era: Physics parameterization impacts and precipitation extremes. J. Appl. Meteor. Climatol., 58, 831–852, https://doi.org/10.1175/JAMC-D-18-0226.1.
Iacono, M. J., J. S. Delamere, E. J. Mlawer, M. W. Shephard, S. A. Clough, and W. D. Collins, 2008: Radiative forcing by long-lived greenhouse gases: Calculations with the AER radiative transfer models. J. Geophys. Res., 113, D13103, https://doi.org/10.1029/2008JD009944.
Janjić, Z. I., 1994: The step-mountain eta coordinate model: Further developments of the convection, viscous sublayer, and turbulence closure schemes. Mon. Wea. Rev., 122, 927–945, https://doi.org/10.1175/1520-0493(1994)122<0927:TSMECM>2.0.CO;2.
Jayasankar, C. B., S. Surendran, and K. Rajendran, 2015: Robust signals of future projections of Indian summer monsoon rainfall by IPCC AR5 climate models: Role of seasonal cycle and interannual variability. Geophys. Res. Lett., 42, 3513–3520, https://doi.org/10.1002/2015GL063659.
Jayasankar, C. B., K. Rajendran, and S. Surendran, 2018: Monsoon climate change projection for the orographic west coast of India using high-resolution nested dynamical downscaling model. J. Geophys. Res. Atmos., 123, 7821–7838, https://doi.org/10.1029/2018JD028677.
Kain, J. S., 2004: The Kain–Fritsch convective parameterization: An update. J. Appl. Meteor., 43, 170–181, https://doi.org/10.1175/1520-0450(2004)043<0170:TKCPAU>2.0.CO;2.
Katz, R. W., 2010: Statistics of extremes in climate change. Climatic Change, 100, 71–76, https://doi.org/10.1007/s10584-010-9834-5.
Katz, R. W., M. B. Parlange, and P. Naveau, 2002: Statistics of extremes in hydrology. Adv. Water Resour., 25, 1287–1304, https://doi.org/10.1016/S0309-1708(02)00056-8.
Kharin, V. V., F. W. Zwiers, X. Zhang, and M. Wehner, 2013: Changes in temperature and precipitation extremes in the CMIP5 ensemble. Climatic Change, 119, 345–357, https://doi.org/10.1007/s10584-013-0705-8.
May, W., 2004: Simulation of the variability and extremes of daily rainfall during the Indian summer monsoon for present and future times in a global time-slice experiment. Climate Dyn., 22, 183–204, https://doi.org/10.1007/s00382-003-0373-x.
Meehl, G. A., and J. M. Arblaster, 2003: Mechanisms for projected future changes in South Asian monsoon precipitation. Climate Dyn., 21, 659–675, https://doi.org/10.1007/s00382-003-0343-3.
Meehl, G. A., and Coauthors, 2007: Global climate projections. Climate Change 2007: The Physical Science Basis, S. Solomon et al., Eds., Cambridge University Press, 747–845.
Menne, M. J., I. Durre, R. S. Vose, B. E. Gleason, and T. G. Houston, 2012: An overview of the Global Historical Climatology Network-daily database. J. Atmos. Oceanic Technol., 29, 897–910, https://doi.org/10.1175/JTECH-D-11-00103.1.
Menon, A., A. Levermann, J. Schewe, J. Lehmann, and K. Frieler, 2013: Consistent increase in Indian monsoon rainfall and its variability across CMIP-5 models. Earth Syst. Dyn., 4, 287–300, https://doi.org/10.5194/esd-4-287-2013.
Mishra, A. K., and S. C. Liu, 2014: Changes in precipitation pattern and risk of drought over India in the context of global warming. J. Geophys. Res. Atmos., 119, 7833–7841, https://doi.org/10.1002/2014JD021471.
Mishra, A. K., V. Nagaraju, M. Rafiq, and S. Chandra, 2018: Evidence of links between regional climate change and precipitation extremes over India. Weather, 74, 281–221, https://doi.org/10.1002/WEA.3259.
Mishra, S. K., S. Sahany, P. Salunke, I.-S. Kang, and S. Jain, 2018: Fidelity of CMIP5 multi-model mean in assessing Indian monsoon simulations. npj Climate Atmos. Sci., 1, 39, https://doi.org/10.1038/s41612-018-0049-1.
Morrison, H., G. Thompson, and V. Tatarskii, 2009: Impact of cloud microphysics on the development of trailing stratiform precipitation in a simulated squall line: Comparison of one- and two-moment schemes. Mon. Wea. Rev., 137, 991–1007, https://doi.org/10.1175/2008MWR2556.1.
Mukhopadhyay, P., S. Taraphdar, B. N. Goswami, and K. Krishnakumar, 2010: Indian summer monsoon precipitation climatology in a high-resolution regional climate model: Impacts of convective parameterization on systematic biases. Wea. Forecasting, 25, 369–387, https://doi.org/10.1175/2009WAF2222320.1.
Muller, C. J., and P. A. O’Gorman, 2011: An energetic perspective on the regional response of precipitation to climate change. Nat. Climate Change, 1, 266–271, https://doi.org/10.1038/nclimate1169.
Nakanishi, M., and H. Niino, 2009: Development of an improved turbulence closure model for the atmospheric boundary layer. J. Meteor. Soc. Japan, 87, 895–912, https://doi.org/10.2151/jmsj.87.895.
Niu, G.-Y., and Coauthors, 2011: The community Noah land surface model with multiparameterization options (Noah-MP): 1. Model description and evaluation with local-scale measurements. J. Geophys. Res., 116, D12109, https://doi.org/10.1029/2010JD015139.
Peltier, W. R., M. d’Orgeville, A. R. Erler, and F. Xie, 2018: Uncertainty in future summer precipitation in the Laurentian Great Lakes Basin: Dynamical downscaling and the influence of continental-scale processes on regional climate change. J. Climate, 31, 2651–2673, https://doi.org/10.1175/JCLI-D-17-0416.1.
Romps, D. M., 2011: Response of tropical precipitation to global warming. J. Atmos. Sci., 68, 123–138, https://doi.org/10.1175/2010JAS3542.1.
Roxy, M. K., 2017: Land warming revives monsoon. Nat. Climate Change, 7, 549–550, https://doi.org/10.1038/nclimate3356.
Roxy, M. K., K. Ritika, P. Terray, and S. Masson, 2014: The curious case of Indian Ocean warming. J. Climate, 27, 8501–8509, https://doi.org/10.1175/JCLI-D-14-00471.1.
Roxy, M. K., K. Ritika, P. Terray, R. Murtugudde, K. Ashok, and B. N. Goswami, 2015: Drying of Indian subcontinent by rapid Indian Ocean warming and a weakening land–sea thermal gradient. Nat. Commun., 6, 7423, https://doi.org/10.1038/ncomms8423.
Rupa Kumar, K., A. K. Sahai, K. Krishna Kumar, S. K. Patwardhan, P. K. Mishra, J. V. Revadekar, K. Kamala, and G. B. Pant, 2006: High resolution climate change scenarios for India for the 21st century. Curr. Sci., 90, 334–345, https://www.jstor.org/stable/24091867.
Sandeep, S., R. S. Ajayamohan, W. R. Boos, T. P. Sabin, and V. Praveen, 2018: Decline and poleward shift in Indian summer monsoon synoptic activity in a warming climate. Proc. Natl. Acad. Sci. USA, 115, 2681–2686, https://doi.org/10.1073/pnas.1709031115.
Santer, B. D., and Coauthors, 2007: Identification of human-induced changes in atmospheric moisture content. Proc. Natl. Acad. Sci. USA, 104, 15 248–15 253, https://doi.org/10.1073/pnas.0702872104.
Seager, R., N. Naik, and G. A. Vecchi, 2010: Thermodynamic and dynamic mechanisms for large-scale changes in the hydrological cycle in response to global warming. J. Climate, 23, 4651–4668, https://doi.org/10.1175/2010JCLI3655.1.
Skamarock, W. C., and J. B. Klemp, 2008: A time-split nonhydrostatic atmospheric model for weather research and forecasting applications. J. Comput. Phys., 227, 3465–3485, https://doi.org/10.1016/j.jcp.2007.01.037.
Stephens, G. L., and T. D. Ellis, 2008: Controls of global-mean precipitation increases in global warming GCM experiments. J. Climate, 21, 6141–6155, https://doi.org/10.1175/2008JCLI2144.1.
Tewari, M., and Coauthors, 2004: Implementation and verification of the unified Noah land surface model in the WRF model. 20th Conf. on Weather Analysis and Forecasting/16th Conf. on Numerical Weather Prediction, Seattle, WA, Amer. Meteor. Soc., 14.2a, https://ams.confex.com/ams/84Annual/techprogram/paper_69061.htm.
Tiedtke, M., 1989: A comprehensive mass flux scheme for cumulus parameterization in large-scale models. Mon. Wea. Rev., 117, 1779–1800, https://doi.org/10.1175/1520-0493(1989)117<1779:ACMFSF>2.0.CO;2.
Trenberth, K. E., 1999: Conceptual framework for changes of extremes of the hydrological cycle with climate change. Climatic Change, 42, 327–339, https://doi.org/10.1023/A:1005488920935.
Turner, A. G., and J. M. Slingo, 2009a: Subseasonal extremes of precipitation and active-break cycles of the Indian summer monsoon in a climate change scenario. Quart. J. Roy. Meteor. Soc., 135, 549–567, https://doi.org/10.1002/qj.401.
Turner, A. G., and J. M. Slingo, 2009b: Uncertainties in future projections of extreme precipitation in the Indian monsoon regions. Atmos. Sci. Lett., 10, 152–158, https://doi.org/10.1002/ASL.223.
Turner, A. G., and H. Annamalai, 2012: Climate change and the South Asian summer monsoon. Nat. Climate Change, 2, 587–595, https://doi.org/10.1038/nclimate1495.
Turner, A. G., K. R. Sperber, J. Slingo, G. Meehl, C. R. Mechoso, M. Kimoto, and A. Giannini, 2011: Modelling monsoons: Understanding and predicting current and future behavior. The Global Monsoon System: Research and Forecast, 2nd ed. C.-P. Chang et. al., Eds., World Scientific, 421–454.
Wang, G., D. Wang, K. E. Trenberth, A. Erfanian, M. Yu, M. G. Bosilovich, and D. T. Parr, 2017: The peak structure and future changes of the relationships between extreme precipitation and temperature. Nat. Climate Change, 7, 268–274, https://doi.org/10.1038/nclimate3239.
Webster, P. J., V. O. Magaña, T. N. Palmer, J. Shukla, R. A. Tomas, M. Yanai, and T. Yasunari, 1998: Monsoons: Processes, predictability, and the prospects for prediction. J. Geophys. Res., 103, 14 451–14 510, https://doi.org/10.1029/97JC02719.
Wentz, F. J., L. Ricciardulli, K. Hilburn, and C. Mears, 2007: How much more rain will global warming bring? Science, 317, 233–235, https://doi.org/10.1126/science.1140746.
Wu, G., and Coauthors, 2007: The influence of mechanical and thermal forcing by the Tibetan Plateau on Asian climate. J. Hydrometeor., 8, 770–789, https://doi.org/10.1175/JHM609.1.
