As one of the integral components of the global climate system, the Indian monsoon provides water to the densely populated region of South Asia. About 80 % of the annual precipitation over India occurs during the summer period, supplying water to the crops during the prime agricultural season . The crop yields (especially rice, which is dominant in the region) are highly sensitive to the monsoon rainfall variability . As agriculture contributes to about 20 % of the gross domestic product , the monsoon's rainfall also has an effect on India's economy . Therefore, there is an inextricable link between the Indian summer monsoon and the health as well as the socioeconomic wellbeing of people. Thus, understanding the response of the Indian summer monsoon and its interannual variability to different global warming scenarios is critical for designing management strategies of water resources and agricultural policies in the future.

In order to understand future projections about potential changes in the monsoon rainfall, it is crucial to understand historic changes and their underlying forcings. In this context, it is important to distinguish between external and internal drivers. External drivers can be of natural (insolation changes due to changes in orbital parameters, volcanic aerosols) and anthropogenic (greenhouse gases – GHGs, aerosols, land use change) origin, whereas internal drivers refer to variations based on the interaction within the climate system (air, sea, sea ice, land) . While orbital forcing mainly shaped the changes of monsoon rainfall on multi-millennial paleoclimatic timescales, anthropogenic forcings competed during the 20th century, and since the early 21st century, GHGs have been dominating as an external forcing .

Multi-millennial paleorecords indicate strong changes both in the Indian and East Asian summer monsoons. These paleoclimatic changes have been revealed by, e.g., oxygen isotope analysis from different caves in Asia for the past several thousand years , analyzing marine sediment records for the Neogene and Quaternary , as well as other methods . Most studies link the paleoclimatic changes of monsoon rainfall predominantly to solar insolation variations in the Northern Hemisphere affecting the Intertropical Convergence Zone (ITCZ) position due to orbital forcing changes .

Especially to explain abrupt nonlinear monsoon transitions as observed in the Holocene in the Tibetan Plateau, gradual insolation changes are not sufficient, and thus internal feedback mechanisms seem to be at play . The moisture-advection feedback might be such an internal mechanism that is able to provoke abrupt transitions and might be responsible for the abrupt Tibetan Plateau transitions in the Holocene . Other amplifying effects might have occurred due to water vapor and cloud feedback .

Observations of the Indian summer monsoon in central India have revealed a decreasing rainfall trend in the second half of the 20th century . Since orbital forcing is playing a negligible role in external forcing in the current centuries , the competing effects of external anthropogenic forcings dominate these long-term trends: anthropogenic forcings are firstly the effect of GHGs and secondly the effect of sulfate aerosols and land-surface changes . The weakening trend of the Indian monsoon is associated with the GHG-induced warming of the Indian Ocean sea surface and the fact that the concurrent warming over the Indian subcontinent was dampened due to aerosols and land-cover changes . The dampening effect over land results from the steep rise of anthropogenic emissions including sulfate aerosols in India and neighboring regions as well as enormous changes in land cover since the 1950s due to the strong expansion of industry and the population growth . The Indian Ocean warming has been linked to anomalies in the lower and upper troposphere due to enhanced latent heat uplift resulting from convection over the ocean . The warming of the Indian Ocean could intensify the convection over the ocean which is compensated by the subsidence of air masses over land. By preventing the convection over the subcontinent, the Hadley cell is modulated in such a way that a drying trend over the region is introduced . Another significant aspect contributing to the rainfall decrease is discussed to be the narrowing of the ITCZ and, correspondingly, the decrease of the associated belt of intense rainfall .

The declining trend has been reversed in various datasets since the beginning of the 21st century, except in the Indian Meteorological Department dataset where a stabilization was captured . The revival of the rainfall in central and northern India might be explained by the emerging land warming due to GHG emissions dominating over the effect of sulfate aerosols and land-cover change. The compensating effect of aerosols in particular has declined and is expected to further decline due to policy interventions . The decreased ocean evaporation in the Arabian Sea leads to a decrease of moisture transport to India and thus fewer low clouds, which results in a warming of the Indian subcontinent. The rising land-surface temperature increases the meridional temperature gradient in the lower troposphere, enhancing the Hadley circulation and summer monsoon rainfall . Therefore, the magnitude of future monsoon rainfall may depend on where temperature rises faster – on the sea surface or land masses . Since this goes back to the competing influence of GHGs and aerosol forcing over land, the task of modeling the future monsoon rainfall coincides with projecting the magnitude of the different forcing mechanisms and capturing the monsoon's sensitivity to it.

Within the latest studies using global coupled models, there is a widespread consensus that the Indian monsoon rainfall will increase due to climate change in the 21st century . This trend is found for various Coupled Model Intercomparison Project phase 5 (CMIP5) models , the multi-model mean , the mean of only the four best models or the model with the best deep convection scheme . Under Representative Concentration Pathway 8.5 (RCP8.5), CMIP5 models project a median increase in Indian monsoon rainfall of 2.3 % K-1 . Also under RCP8.5, the amount of rainfall over India is projected to increase by 18.7 % by the end of the 21st century compared to 1961–1999 . This trend is expected to be the consequence of the warming of the Indian Ocean enhancing atmospheric moisture content and thus moisture flux convergence arising from changes in moisture which generally follow the Clausius–Clapeyron relation . This so-called thermodynamic effect dominates over the dynamic effect, which refers to weaker winds and a reduced monsoon circulation due to a weakened tropical overturning circulation and an expected decrease of rainfall . Besides, the interannual variability is projected to increase in most models under the strongly forced scenarios as well as in models with good performance in capturing the mean seasonal cycle in the present climate .

The capabilities of climate models in simulating the Indian monsoon have improved over time, such that more accurate projections can be expected from the latest update of the climate models in CMIP6. Several studies found a broad range of improvements between CMIP3 and CMIP5 in simulating the 20th century monsoon , though one study, based on a small subset of models, however, disagrees . found a significant improvement between CMIP5 and CMIP6 in simulating the Indian summer monsoon rainfall for the period 1951–2005. Models in CMIP5 still struggled with various issues including displaying the decrease in rainfall in the second half of the 20th century and capturing observed trends in the extremes and seasonality indices . With the new generation, models' capacities in capturing the spatiotemporal pattern of Indian summer monsoon, especially in the Western Ghats and the northeast foothills of Himalaya mountains, have undergone significant progress . While global coupled models in CMIP5 failed to capture the influence of topography, land-surface feedback and land use change due to their coarse spatial resolution, the general higher resolution in CMIP6 contributes to an improved simulation of Indian monsoon dynamics . Further improvements have been achieved by updating deep convective schemes, modifying parameterization on microphysical scale, integrating indirect effects of aerosols in cloud formation and advancing ocean-ice models .

Here, we aim to update the CMIP projections for the Indian monsoon rainfall and its interannual variability for the 21st century by using 32 models of the latest climate model generation. For this purpose, we use the Shared Socioeconomic Pathways (SSPs) and possible corresponding forcing levels as a scenario framework . Section  gives a brief overview of the data used and processed. In Sect. , we evaluate the participating models according to their capacity of modeling the Indian summer monsoon in historic periods. Section  presents the results of the mean summer monsoon precipitation, while Sect.  focuses on the long-term trend of interannual variability. The results are discussed in Sect. .

In this study, we investigate the mean Indian summer monsoon rainfall and its interannual variability under four different scenarios using 32 global climate models that participated in CMIP6. The four scenarios (SSP1-2.6, SSP2-4.5, SSP3-7.0 and SSP5-8.5) are based on different socioeconomic scenarios and combined with the resultant forcing level . The models are chosen according to their data availability for the historic period (1850–2015) and the future period (2015–2100) under SSP5-8.5 in the Scenario Model Intercomparison Project (ScenarioMIP) . For each model, for consistency, we use one ensemble member (if available: r1i1p1f1) even if more are available. An overview of modeling centers and data availability for the different scenarios is given in Table . The short names of the models used in this study can be found in Table . We select the land area with latitude 6–36∘ N and longitude 67.5–98∘ E, comprising India and neighboring regions. The land area is obtained by using land–sea masks for each model that are based on the percentage of the grid cells occupied by land (see Fig.  for each model). The resolution strongly differs between the models ranging over land from about 100 to 500 km (see Table ). Mean rainfall is obtained by averaging the monthly rainfall data from June–September over the region of interest. For comparison of models to observational data, we use precipitation over land from global reanalysis data at 0.5∘ spatial resolution. The WFDE5 (Forcing Data methodology applied to ERA5 reanalysis data) dataset is used for the period 1985–2015, while for the pre-satellite era period 1900–1930, we use the GSWP3 (Global Soil Wetness Project Phase 3) dataset . Both datasets are based on Global Precipitation Climatology Centre (GPCC) monthly precipitation rates . For calculating the change in interannual variability, we apply the singular spectrum analysis method with a window size of 20 years to extract the nonlinear trend.

In this study, the long-term trend of the Indian summer monsoon and its variability have been analyzed based on the latest global coupled model simulations under the SSP scenarios. Our approach addresses the question of whether the results from earlier studies can be confirmed or need to be adapted in their sign or magnitude.

By comparing the CMIP6 projection results with the WFDE5 reanalysis data, we classified some models as probably more capable of simulating a realistic representation of the monsoon rainfall. The share of models that capture the reference rainfall within twice the standard deviation has slightly increased in CMIP6 (16 out of 32) in comparison to the precursor models in CMIP5 (9 out of 20) . But it has to be noted that the validation period and the used reanalysis data differ between and this study. The observation of quantitatively measurable improvement between CMIP5 and CMIP6 coincides with the results of . While all the models that were out of the 2 standard deviation range underestimated the mean in CMIP5, thus revealing a very clear general tendency of underestimation, the 16 models outside of the range in CMIP6 partly underestimated (13 models) and party overestimated (3 models) the observed mean in 1985–2015. Modeling centers whose models underestimated the rainfall within 2 standard deviations in our study mostly underestimated the rainfall already in CMIP5. Some models with realistic patterns in CMIP6 are updates from CMIP5 that already revealed a pattern relatively similar to reanalysis data, e.g., NorESM2-MM. As in CMIP5, models with the tendency to underestimate the rainfall in the evaluation period are also mostly not capable of capturing the spatial rainfall pattern in CMIP6. But there are also various models that improved their capacity in capturing the Indian monsoon, such as the models from Centre National de Recherches Métérologiques (CNRM-CM6-1, CNRM-CM6-1-HR, CNRM-ESM2-1). This observed inconsistency among models in improving their spatial representation of the Indian monsoon was already noted by . Besides, the capacity of capturing the rainfall pattern over the Western Ghats has improved, which also coincides with the results of .

The CMIP6 models project a robust intensification of the Indian summer monsoon rainfall under climate change. All of the 32 available models exceed the envelope of baseline variability from 1850–2015 until 2100 under SSP5-8.5, while just 17 out of 20 exceeded the natural variability threshold under RCP8.5 in a previous study based on CMIP5 . Additionally, we calculated the average multi-model trend of projected change in mean rainfall by the end of the 21st century. As some modeling centers provide several models and some of them are based on overlapping model components, the models cannot be regarded as independent from each other see, e.g.,. The results have to be interpreted against this background. The average multi-model trend found in CMIP6 with an increase of +24.3 % by 2100 seems stronger in comparison to CMIP5 .

found an increase of 18.7 % in RCP8.5 by 2099 compared to the period 1961–1990 in CMIP5 models. But because of the used time periods as well as the different study area of India without adjacent regions, this study is not directly comparable to ours. An intensification of the Indian monsoon rainfall has also been found in other studies using CMIP5 . There is a widespread agreement that a reason for the intensification of the South Asian monsoon rainfall is an increase in moisture flux convergence . This enhanced thermodynamic effect dominates over the dynamic effect which refers to the decreasing monsoon circulation. quantified the increase of the thermodynamic component of the moisture budget for the Indian monsoon with about 0.7 mm d-1 and the decrease of the dynamic component with 0.4 mm d-1 using nine CMIP5 models in RCP8.5 determining the positive sign of the change in monsoon rainfall .

We found that the monsoon rainfall is linearly dependent on the GMT. This is not in contradiction with the observed decline in monsoon rainfall during the second half of the 20th century: while between the 1950s and 1970s, approximately, high aerosol loadings led to subdued warming and a weakened land–sea thermal gradient, greenhouse-gas-induced warming has dominated since then and is the dominant forcing in the 21st century projections. The projected increase in rainfall is 0.33 mm d-1 per kelvin of global warming. The agreement between models and the independence of the scenario is remarkable. The median dependence of relative change in precipitation on GMT taking into account all models has increased from 3.2 % K-1 in CMIP5 to 5.3 % K-1 in CMIP6. Considering only the models with a more realistic representation of the monsoon, the increase is even more noticeable from 2.3 % K-1 in CMIP5 to 6.1 % K-1 in CMIP6. It also has to be mentioned that the range of projected sensitivities has decreased remarkably from 1 % K-1–19 % K-1 in CMIP5 to 2 % K-1–13 % K-1 in the latest generation of climate models; i.e., the uncertainty in hydrological sensitivity has decreased with the model updates. Similar tendencies have been found for the equilibrium climate sensitivity in CMIP6 . Which of the updated processes between CMIP5 and CMIP6 described by dominate in causing the increased sensitivity of the monsoon to global warming needs further investigation.

The increase in rainfall is projected to contribute to the precipitation in the Himalaya region, the northeast Bay of Bengal and the northwest coast of India. These regions coincide to a large extent with the existing monsoon rainfall pattern, leading to a “wet-regions-get-wetter” pattern during JJAS monsoon rainfall. The distribution of regions with projected increasing precipitation in CMIP6 confirms the projection of previous studies using CMIP5 models . Furthermore, the increasing pattern is shared by a larger percentage of available models in CMIP6 compared to CMIP5. But our projection of increased rainfall over the Western Ghats does not coincide with the study of projecting a decrease in this region. By focusing on high-resolution models with the best deep convection scheme, their study reveals decreasing precipitation on the southwest coast of India, which is only captured by one-third of the CMIP6 models in our study, including the CNRM-CM6-1-HR model. A finer resolution seems to be necessary to capture this trend, which is not given for all CMIP6 models.

From the 32 available models, 28 models project an increase in interannual variability. This result is not directly comparable to the study of since the removal of the trend in our study has a relevant influence on the results. Without the removal of the trend, i.e., following the method of , all 32 models project an increase in interannual variability, which shows that the signal has become clearer in comparison to the results in CMIP5 models. The projected increase in interannual variability coincides with other studies . A dominant role in shaping the interannual variability is taken by the El Niño–Southern Oscillation (ENSO) . As El Niño events typically coincide with dry monsoon years and La Niña years are often accompanied by strong monsoon rainfall , changes in the emergence of these events have a relevant impact on the Indian summer monsoon. applied spectral analysis and found a shortening of the spectral periods of ENSO which might lead to a shift in the relationship of ENSO and monsoon rainfall.

We used 32 CMIP6 models to analyze the Indian summer monsoon's response to climate change. In order to identify models with a good representation of the Indian monsoon, we compared the models' simulations in the past to WFDE5 reanalysis data. We found that there are 16 out of 32 models that are able to capture the monsoon rainfall within twice the standard deviation in the period 1985–2015. This is a slight increase compared to CMIP5. The models outside that range in CMIP6 still have a tendency to underestimate the amount of precipitation in this period. This was already observed in CMIP5, where all of the models out of the range underestimated the rainfall. In our analysis, we focused on the models with the more realistic representation of the Indian monsoon. We found that all models show an increase in mean summer monsoon rainfall under SSP5-8.5 and SSP3-7.0 by the end of the 21st century. An increase also was found in SSP2-4.5 and SSP1-2.6 by all models apart from two models in SSP2-4.5 and one model in SSP1-2.6. Under SSP5-8.5, the models exceed the envelope of the baseline's variability on average in 2045. An multi-model mean increase of rainfall of 24.3 % is projected under SSP5-8.5 and of +18.6 % in SSP3-7.0, +11.9 % in SSP2-4.5 and +9.7 % in SSP1-2.6. The majority of models project that the increase will contribute to the precipitation especially in the Himalaya region, the northeast of the Bay of Bengal and to the west coast of India. Besides, the simulation ensemble indicates a linear dependence of rainfall on global mean temperature independent of the SSP in the 21st century; the multi-model mean for JJAS projects an increase of 0.33 mm d-1 and 5.3 % per kelvin of global warming. Furthermore, under SSP5-8.5, a majority (28 out of 32 models) project an increase in interannual variability by the end of the 21st century after removing the trend with singular spectrum analysis.

We have seen in this study that low-resolution models did not capture the spatial pattern of the monsoon rainfall in historic periods well. Small-scale topography and its atmosphere feedback influence the rainfall to a relevant extent. Thus, the ongoing effort to improve the resolution of the individual CMIP models should be continued. Since other rainfall features such as extremes and the variability of rainfall on a subseasonal scale are beyond the scope of this study, they need to be analyzed in further studies due to their high relevance, e.g., for high-risk flooding events.

The projected increase in summer monsoon rainfall in combination with the projected long-term increase in interannual variability will be accompanied by an increased number of extremely wet years and potentially more high-rainfall events . While crops need water especially in the initial growing period, high-rainfall events during other growing states can harm the plants . Thus, the projected development might have serious consequences for the agriculture in India and neighboring regions. Since the change differs from the decreasing tendency in the second half of the 20th century, the development of adaptation strategies for the 21st century is required.