A framework to incorporate spatiotemporal variability of rainfall extremes in summer monsoon declaration in India

The Indian summer monsoon rainfall is a lifeline for agricultural activities and the socio-economic development of more than 1 billion people. All-India averaged summer monsoon rainfall has about 10% variability from its long-term mean. A departure of all-India averaged precipitation within ±10% is declared a normal summer monsoon. Using the long-term (1901–2021) gridded rainfall observations, we highlight the limitations in the current approach to the declaration of the normal summer monsoon, which ignores the role of spatiotemporal variability of rainfall. Dry and wet extremes within the same monsoon season can lead to a normal monsoon. Moreover, different parts of the country face drought and wet extremes, while the summer monsoon can be declared normal. Considering the profound implications of dry and wet extremes on agricultural activities, we propose a novel framework to account for the rainfall variability in the declaration of the summer monsoon. The proposed framework accounts for the temporal variability through a combined severity coverage index, while spatial variability using a clustering approach. Based on the existing framework, we find that 84 years were declared normal in the last 121 years (1901–1921). However, 13 years (out of 84) were not normal based on the new framework due to dry and wet extremes occurring at different times and in different regions. The new framework of summer monsoon declaration can account for the occurrence of extremes and their implications for agriculture and water management.


Introduction
Indian summer monsoon rainfall is the lifeline for millions of people's water availability and agricultural activities. India receives about 80% of the total annual rainfall during the four months (June-September) of the summer monsoon season. The summer monsoon season overlaps with one of the major crop-growing seasons (Kharif), and monsoon rain is a prominent water source for supporting agricultural activities , Gadgil and Kumar 2006, Prasanna 2014. Failure of the summer monsoon leads to meteorological droughts hampering water availability and crop growth (Mishra 2020). A substantial precipitation deficit in the summer monsoon season leads to agricultural and hydrological droughts affecting crop growth, food production, and reservoir storage. Prolonged summer monsoon rainfall breaks affect groundwater storage, with additional irrigation water requirements being only met by the groundwater pumping (Mukherjee et al 2015, Asoka et al 2017. The linkage between summer monsoon rainfall deficit and groundwater depletion in India has been well established (Asoka et al 2017. For instance, groundwater abstraction for irrigation increases substantially . On the other hand, surplus rainfall during the monsoon season increases the risk of flooding and challenges reservoir operations (Nagesh Kumar et al 2009, Nanditha and. Therefore, both deficit and surplus summer monsoon rainfall cause detrimental impacts.
Several parts of the country, including the Indo-Gangetic Plain, western Ghats, and northeastern India, have experienced a significant decline in the summer monsoon rainfall over the last few decades (Krishnan et al 2016, Asoka et al 2018, Singh et al 2019, Mishra 2020. The declining trends in the summer monsoon rainfall in these regions caused droughts affecting water resources and flood production (Mishra 2020). On the other hand, parts of western and central India have experienced an increase in summer monsoon rainfall (Asoka et al 2018). The increased summer monsoon precipitation in these regions is primarily due to extreme rainfall, which contributes to flooding (Roxy et al 2017, Mukherjee et al 2018. Additionally, the variability of rainfall extremes during the monsoon season has increased (Ghosh et al 2011), causing dry and wet spell frequency (Ghosh et al 2011. The India Meteorological Department (IMD) monitors the progress and state of the summer monsoon every year. At the end of the monsoon season, IMD declares the overall condition of the summer monsoon as drought (precipitation deficit), surplus, and normal (https://mausam.imd.gov.in/imd_latest/ contents/monsoon_activity.php). Since the longterm variation in the summer monsoon rainfall is about ± 10% from its mean, a threshold of 10% is used to declare drought or surplus monsoon . For instance, if the rainfall deficit at the end of the season is more than 10%, then the year is declared a drought. On the other hand, if the summer monsoon rainfall exceeds 10%, the monsoon is surplus. All-India averaged summer monsoon rainfall anomaly within ±10% is considered normal. This declaration of the summer monsoon based on rainfall departure or more than 10% (one standard deviation) from its long-term mean is based on the Normal distribution of summer monsoon rainfall anomalies. Therefore, about 67% of years fall in the Normal category of the monsoon. Moreover, the definition of the normal monsoon is based on the statistical sense and not necessarily based on the impact of the rainfall variability during the summer monsoon season or in different parts of India. Therefore, this declaration of different states of the summer monsoon is based on all-India averaged rainfall and does not account for the spatial and temporal variability (i.e. spatial and intra-seasonal rainfall variability). As expected, even during the normal monsoon years, there can be considerable spatial variability in rainfall in the country. IMD provides rainfall anomalies at the district level to report the spatial variability in the summer monsoon rainfall. Similarly, IMD also develops weekly and monthly rainfall totals at district and sub-divisional levels along with the time series of active and break spells to account for temporal variability in summer monsoon rainfall. All the information related to summer monsoon rainfall and its spatial and temporal variability prepared by IMD is available in the public domain. However, the declaration of the summer monsoon 'Normal' based on the standardized departure of all-India averaged rainfall from its long-term mean at the end of the season mean does not account for the anomalous nature of rainfall due to spatiotemporal variability.
Both dry and wet extremes can occur during the same monsoon season (Singh et al , 2019, which can be declared as normal based on the all-India averaged rainfall anomalies. The fundamental issue with the existing approach is that drought and extreme wet periods within the same monsoon season can have profound implications. However, the declaration of the normal monsoon does not account for that (Rajeevan et al 2010). Similarly, drought and wet extremes can occur in different parts of the country during the same monsoon season with detrimental impacts. However, the current approach to the monsoon declaration does not account for the spatial and temporal variability of the summer monsoon rainfall. In addition, drought and wet extremes have affected different parts of the country in the last few decades (Mishra et al 2016, Roxy et al 2017. For instance, Mishra et al (2016) reported a decline in the summer monsoon rainfall over the Indo-Gangetic Plain while the monsoon rainfall has increased over western India (Asoka et al 2018). Roxy et al (2017) reported a three-fold rise in widespread extreme precipitation events during the monsoon season over central India. Therefore, the observed changes in the total rainfall and rainfall extremes during the summer monsoon season indicate the prominent role of spatial variability during the season. Similarly, Sahany et al (2018) found a decline in the wet season duration and later occurrence of peak rainfall in the monsoon season. Intraseasonal variability of the summer monsoon rainfall is driven by active and break spells (Krishnamurthy and Shukla 2000, Rajeevan et al 2010, Pai et al 2009, 2016. Spatial and temporal variability in the summer monsoon rainfall and application of the indices have been widely discussed in previous studies (Kulkarni et al 1992, Dwivedi et al 2019. Notwithstanding the importance of rainfall variability, the declaration of the monsoon does not account for these spatiotemporal features. Here we provide a coherent framework to account for the temporal and spatial variability and hydrologic extremes in the monsoon declaration. We contrast and demonstrate the applicability of this new approach with an existing IMD approach at an all-India scale using a long-term (1901-2021) precipitation. We do not aim to diagnose the mechanism of anomalous summer monsoon rainfall. Rather, we use the existing criteria that defines surplus, deficit, and normal monsoon based on the long-term rainfall anomalies. We, then, propose a framework to incorporate the spatial and temporal variability of summer monsoon rainfall in the declaration. If the summer monsoon rainfall is deficit/surplus, a considerable part of the country experiences drought/extreme wet conditions. However, during the normal monsoon years, a significant part of the country may face both dry and wet extremes, which is not accounted in the declaration.

Data and methods
We obtained daily gridded precipitation at 0.25 • spatial resolution, which was developed using more than 6900 observational stations across India . The high-resolution gridded precipitation  is an updated version of the dataset available at a relatively coarser resolution (1 • , Rajeevan et al 2006). The gridded precipitation captures spatial and temporal variability of the summer monsoon in India, especially the high rainfall in the coremonsoon region and low rainfall in the semi-arid and arid areas of western India. The high rainfall in the foothills of the Himalayas and the Western Ghats are well resolved in the gridded dataset. The dataset has been widely used for hydrometeorological studies over India (Prakash et al 2016, Shah and Mishra 2020, Mishra et al 2021b. We used daily precipitation to estimate weekly rainfall during the summer monsoon season (June-September). We identified 18 weeks during the monsoon season, and each week's precipitation total was estimated. Moreover, we calculated total precipitation during the summer monsoon season to assess the all-India averaged rainfall anomaly against the long-term (1901-2021) mean. All-India averaged precipitation anomaly less than −10% were identified as the monsoon season droughts. On the other hand, the years with rainfall anomalies of more than 10% of the long-term mean were categorized as surplus monsoon years. Therefore, normal, surplus, or deficit monsoon declaration is based on the departure of more than one standard deviation. One standard deviation of the long-term summer monsoon rainfall is about 10% of the long-term mean. We estimated precipitation anomalies for each week during the summer monsoon season to examine the temporal variability of the monsoon. Weekly precipitation anomalies were estimated against the long-term (1901-2021) mean of rainfall for the corresponding weeks.
We estimated a standardized precipitation index (SPI, Mckee et al 1993) to examine the monsoon season drought in India and the selected regions. Since our focus is mainly on characterizing meteorological states, the role of temperature/potential evapotranspiration was not considered. We used the Standardized Drought Analysis Toolbox (Farahmand and AghaKouchak 2015) to estimate weekly and monthly SPI. Instead of a standard probability distribution, we used the empirical probability to derive nonparametric SPI using the Gringorten plotting position (Gringorten 1963, Farahmand andAghaKouchak 2015). We used SPI to examine the drought conditions during the summer monsoon season. For instance, a four month SPI at the end of September was used to analyze drought during the summer monsoon season. Similarly, a weekly SPI was used to investigate the drought/wet conditions for the corresponding week. Drought/wet categories were defined using the range of SPI as described in Svoboda et al (2002).
We obtained sea surface temperature (SST) data available at 2 • spatial, and monthly temporal resolution from Extended Reconstructed SST v5 (Huang et al 2017(Huang et al , 2018. Using monthly SST, the SST departure field was estimated after subtracting the global mean monthly SST from each grid as in Mishra et al (2012). We calculated SST anomalies for the summer monsoon (June-September) season. Geopotential height and wind (u, v) at 850 hPa were obtained from ERA-20C reanalysis for 1900-2010 (Poli et al 2016) and ERA5 (Hersbach et al 2020) for the 2011-2021 periods. We constructed climatological and anomaly fields using geopotential height and wind datasets for the selected years.
The impacts of drought or prolonged wet periods are often measured using the severity and area affected. Drought can be categorized into: Abnormally dry (D1), moderate drought (D2), severe drought (D3), extreme drought (D4), and exceptional drought (D5). The area under different drought categories can be linked with the drought impacts. For instance, more severe consequences can be expected if a large area is under exceptional drought. As drought involves several aspects (intensity, duration, and coverage), it is difficult to compare the severity of two drought events in the same region (Martin et al 2020). Drought Severity Coverage Index (DSCI, Akyuz 2017, Johnson et al 2020 was designed to provide an integrated measure considering the area under different drought categories for each week. A weight of 1-5 was assigned to the drought area under each category (D1-D5), multiplied by the drought area (%) under different categories: Similarly, we estimated the Wet Severity Coverage Index (WSCI) using the area under different wet categories (W1-W5) as: DSCI and WSCI were estimated for 18 weeks during the summer monsoon season june-september (JJAS). Active and break spells and associated dry and wet extremes within the summer monsoon season are driven by low-pressure systems and cyclonic storms, which are short-time scale systems (Ajayamohan et al 2010, Revadekar et al 2016, Tomas et al 2021. Ajaymohan et al (2010) showed that extreme rainfall events during the monsoon season are strongly associated with synoptic disturbances. However, at a seasonal time scale, extreme rainfall events are independent of seasonal mean rainfall. We incorporated temporal variability of dry and wet extremes during the monsoon season using cumulative DSCI and WSCI for the 18 week period. Cumulative DSCI and WSCI at the end of the monsoon season represent the overall drought and surplus/wet summer monsoon conditions, respectively (figure S1). We estimated Combined Severity Coverage Index (CSCI) as the sum of cumulative DSCI and WSCI for the summer monsoon season. Therefore, CSCI accounts for the spatial-temporal variability of the dry and wet conditions during the monsoon season. The departure of more than one standard deviation of CSCI can be considered as the monsoon season with high temporal variability in the combined dry and wet extremes. Hence, the years with the CSCI of more than one standard deviation can be considered ones that experience extremes during the summer monsoon season regardless of the normal monsoon declaration.
We identified homogeneous precipitation clusters in India (Shah and Mishra 2020) to account for the spatial variability of rainfall during the summer monsoon season (figure S2). A novel spatial clustering algorithm that uses traditional interpoint distance metric (Singh and O'Gorman 2014), ensuring the minimum size of each cluster was used (Sanderson et al 2019). We identified eight clusters with similar drought characteristics in India (Shah and Mishra 2020). A detailed description of clustering can be found in Sanderson et al (2019) and Shah and Mishra (2020). Clusters that experienced drought (SPI < −0.8) and wet conditions (SPI > 0.8) for each summer monsoon season were identified. During the summer monsoon, clusters that experienced extreme wet conditions (SPI > 1.5) were also identified. Dry extremes are often widespread, while wet extremes are localized. To avoid the localized impact of extreme rainfall in a cluster, we considered SPI threshold exceeding 1.5 for wet extremes in a cluster. On the other hand, a cluster was considered affected by dry extremes if the area averaged SPI was less than −0.8.

The role of temporal variability
First, we examined the temporal variability of India's 2021 summer monsoon season rainfall. The temporal variability of weekly rainfall shows that India experienced both dry and wet extremes during the summer monsoon season ( figure 1(a)). As the all-India averaged precipitation anomaly fell within ±10% (one standard deviation), the 2021 summer monsoon was declared normal by the IMD . The weekly precipitation anomaly (%) show wet condition in the early summer monsoon season, followed by the two long rainfall breaks during July and August ( figure 1(a)). Extreme precipitation in several parts of the country created wet conditions at the end of the summer monsoon season ( figure 1(a)). The temporal variability of rainfall shows extreme dry and wet conditions in the same monsoon season. We estimated weekly DSCI and WSCI for the summer monsoon season of 2021, which show high DSCI during July and August while high WSCI during June and September. Several parts of the country, especially central India, experienced extreme to exceptional category drought during the fifth and tenth week of the 2021 summer monsoon season. In contrast, a large part of the country experienced extreme wet conditions during the 17th and 18th weeks of the monsoon season. Both extreme dry and wet conditions posed challenges associated with drought and floods affecting water resources, agriculture, and the socio-economic well-being of the people. Notwithstanding the severe implications, the 2021 summer monsoon normal declaration based on the statistical framework raises questions on the methodology associated with the declaration that does not account for intra-seasonal rainfall variability.
Next, we contrast the spatial-temporal variability of rainfall during the driest and wettest summer monsoon based on the all-India averaged precipitation anomalies from 1901 to 2021 (figure S3). The driest summer monsoon occurred in 1972, with around 22.6% precipitation deficit from the long-term mean. On the other hand, the wettest summer monsoon was recorded in 1917, with 21.5% surplus rainfall. Most of India witnessed drought during the summer monsoon of 1972, which was more prominent in the central region ( figure S3(a)). For instance, a large part of Maharashtra experienced an exceptional drought in 1972 ( figure S3(a)). The driest summer monsoon experienced most of the weeks dry with precipitation anomalies around −70% ( figure S3(b)). Only a few weeks witnessed positive precipitation anomalies during the summer monsoon season. The temporal variability of precipitation deficit was reflected in the cumulative DSCI, which was considerably higher during the 7-10th weeks of the monsoon season ( figure S3(c)). Monsoon weeks with a moderate rainfall surplus resulted in moderate cumulative WSCI during 1972 ( figure S3(d)). A large part of India witnessed a surplus summer monsoon during 1917, with western India being exceptionally wet ( figure S3(e)). Except for the three weeks, the entire monsoon season experienced a surplus monsoon, which was more prominent during the end of the season ( figure S3(f)). Temporal variability of the extreme wet season for the 1917 summer monsoon is reflected in the cumulative DSCI and WSCI (figures S3(g) and (h)).
We examined the linkage between precipitation anomalies during the summer monsoon and cumulative DSCI and WSCI for the 1901-2021 period (figures 2, S4 and S5). Cumulative DSCI and WSCI capture the temporal variability of both dry and wet extremes during the monsoon (figures S1, S4, and S5). During 121 years, 23 experienced more than a 10% deficit in the summer monsoon season precipitation and were declared drought years. During the last 121 years, the top five worst drought years based on the failure of summer monsoon rainfall  2(a) and (b)). All the drought years witnessed higher cumulative DSCI indicating the prolonged monsoon breaks that put a large part of the country under drier conditions (figures 2(c), (d) and S1). While the driest summer monsoon of 1972 also had the highest cumulative DSCI (1957.6), the cumulative DSCI showed a negative relationship (r = −0.88) with the monsoon season precipitation anomaly. For instance, the overall rainfall deficit during 2002 (−20.6%) is higher than in 2018 (−13.7%). 2018 has a higher cumulative DSCI than 2002, indicating the importance of the temporal variability of the summer monsoon rainfall. Similarly, cumulative WSCI is strongly linked (r = 0.89) with precipitation anomaly during the summer monsoon season (figure S1). Our results show that cumulative DSCI and WSCI capture the temporal variability of summer monsoon season precipitation and area affected by drought and wet extremes. In contrast, the declaration of the summer monsoon based on the all-India total summer monsoon rainfall does not account for the rainfall variability.
We combined the cumulative DSCI and WSCI for the summer monsoon season to obtain the temporal development of the CSCI over the 1901-2021 period (figures 2(g) and (h)). A high value of CSCI indicates a robust temporal variability of dry and wet extremes during the monsoon season. High CSCI during the summer monsoon season can be due to high cumulative DSCI or WSCI indicating the dominance of dry or wet extremes. On the other hand, relatively higher cumulative DSCI and WSCI can also contribute to high CSCI in case of domination of both dry and wet extremes during the same monsoon season, as in the summer monsoon of 2021. When drought and wet extremes occur at different times during the monsoon season, precipitation totals over the entire season may not capture the influence of the intraseasonal variability. In these cases, the summer monsoon is likely to be declared normal based on the statistical framework followed by IMD, notwithstanding the profound implications of dry and wet extremes. Seven years were reported as drought from the top ten CSCI years (figures 2(g) and (h)). The other three years (2005, 2006, and 1991) were normal monsoon years. Therefore, our results show that during these three years, the country most likely witnessed drought and wet extremes at different times during the monsoon season. Overall, we demonstrate that high temporal variability during the summer monsoon can cause both dry and wet extremes, which are not usually considered in the declaration of the normal summer monsoon based on the current framework.

The role of spatial variability
Next, we examined the role of spatial variability of the summer monsoon rainfall on the monsoon declaration (figure 3). For instance, if different regions experience dry and wet extremes, all-India averaged precipitation during the summer monsoon season can be normal based on the all-India rainfall anomaly at the end of the season. Both dry and wet extremes can affect the affected regions, which is not reflected if the monsoon is declared normal. We divided the country into eight clusters (figure S2) based on the longterm precipitation (Shah and Mishra 2020) to understand the regional variability of dry and wet extremes during the summer monsoon season. Both drought and surplus monsoon years exhibited negative and positive 4 month SPI at the end of the monsoon season ( figure 3(a)). Out of 121 years , 84 years were declared normal monsoon. Several clusters experienced drought (SPI < −0.8) during the normal monsoon in India ( figure 3(b)). All eight clusters witnessed drought in one year ( figure 3(b), table S1). Moreover, four regional clusters (C1, C6, C7, and C8) experienced drought in 12 or more years during the 84 normal monsoon years. We identified years in which clusters experienced drought, and the summer monsoon was declared normal. Our results show that a few clusters experienced extreme to exceptional category droughts, which may have posed severe water management and agriculture challenges. For instance, during the normal monsoon years, two or more clusters experienced drought 22 times out of the total 84 years. Eleven times three or more clusters experienced drought during the normal monsoon years. Similarly, different parts of the country witness extreme wet conditions during the summer monsoon, which may not reflect the normal monsoon declaration for the 2021 monsoon season.
We identified five declared normal years during which four or more clusters were under drought (SPI < −0.8) (figure 4). The summer monsoon was declared normal in 1925 (precipitation anomaly = −3.1%), 1928 (−8.6%), 1939 (−5.5%), 1974 (−8.4%), and 2017 (−5.6%) during the last 120 years (figure 4). During the monsoon season of 1925, a large part of western India was affected by drought, while the central-eastern part received a rainfall surplus. Out of eight, four clusters (C1, C3, C5, and C6) witnessed drought, while two (C2 and C7) witnessed wet conditions (figures 4(a) and (b)). One cluster (C7) experienced extreme wet conditions as the four month SPI at the end of the monsoon season exceeded 1.5 ( figure 4(b)). Drought and wet conditions in different parts of the country contributed to an overall monsoon rainfall anomaly of −3.1%. North-central and parts of peninsular India were under drought, while northeastern and a few other regions were moderately wet during the 1928 monsoon (figures 4(c) and (d)). Four clusters were under drought, while none experienced wet conditions during the summer monsoon of 1928 ( figure 4(d)). We also found considerable spatial variability in precipitation during the monsoon season of 1939 (figures 4(e) and (f)). For instance, western India experienced extreme drought while the eastern part received surplus monsoon rainfall in 1939 (figure 4(e)). Four clusters (C1, C2, C4, and C5) were under drought, while one (C8) experienced extreme wet conditions (SPI > 1.5).
Similarly, in 1974, western and part of eastern India experienced extreme drought during the summer monsoon season (figures 4(g) and (h)). On the other hand, northeastern India received surplus monsoon rainfall in 1974. Six clusters (C1-C5 and C7) were under drought, while three (C1, C5, and C7) experienced extreme drought ( figure 4(h)). On the other hand, one cluster (C8) received surplus monsoon rainfall (SPI > 2.5). The exceptional wet conditions in the northeastern region contributed to reducing the all-India averaged deficit in the summer rainfall, and consequently, the whole monsoon season was declared normal. For example, the case of the 2017 summer monsoon was declared normal given that the precipitation deficit was less than 10%, while at the regional scale, different clusters experienced hydrologic extremes (figures 4(i) and (j)). The northcentral part was affected by drought, while western India received surplus summer monsoon rainfall (figure 4(i)). Four clusters witnessed drought. At the same time, one (C1) experienced wet conditions during the monsoon of 2017 (figure 4(j)). We examined the temporal variability of the normal monsoons (1925, 1928, 1939, 1974, and 2017), which had four or more clusters under drought (figure S6). All the five years experienced dry and wet monsoon weeks exhibited by the weekly precipitation anomalies. Weekly DSCI and WSCI agree with the precipitation anomalies (figure S6). Overall, drought and wet conditions in different parts of the country during the same monsoon season can lead to the declaration of the normal summer monsoon.

Framework to incorporate spatial and temporal variability
The current methodology to declare the summer monsoon normal is based solely on the all-India averaged precipitation anomaly. As one standard deviation of India's long-term summer monsoon rainfall is about 10% of its mean, the departure beyond 10% is declared as drought or surplus monsoon. Therefore, if the all-India averaged summer monsoon precipitation anomaly is within ±10%, the monsoon is declared normal regardless of its temporal and spatial variability. However, the method of declaring the summer monsoon solely based on a statistical framework has issues as it does not account for the temporal and spatial variability of the summer monsoon precipitation in India. Both temporal and spatial variations in rainfall can considerably affect water availability, agriculture, and infrastructure. Our results show that the temporal variability of the summer monsoon rainfall can lead to drought and wet extremes during the same season, impacting various sectors. On the other hand, droughts and floods can occur in different parts of the country and deficit, and surplus summer monsoon rainfall anomalies can cancel each other out, leading to a normal monsoon.
Accounting for spatial and temporal variability of the summer monsoon rainfall is essential to examining the implications of drought and wet extremes in India. We estimated cumulative DSCI and WSCI that measure drought and wet extremes during the summer monsoon season. The CSCI integrates dry and wet extremes during the 18 weeks of the summer monsoon season. Thus, CSCI measures the temporal variability of the summer monsoon season precipitation. Therefore, the departure of CSCI with more than one standard deviation can be considered a basis for identifying the summer monsoon season with high temporal variability. Any year with less than 10% departure from its long-term mean should have CSCI departure less than or equal to one standard deviation to be considered normal from the long-term mean and temporal variability perspective. As there have been several years since the summer monsoon was declared normal, several clusters experienced drought and wet extremes. To account for the spatial variability of the summer monsoon season, we used a 4 month SPI for different regional clusters at the end of September. Out of the total of eight clusters, if more than three clusters experience drought (SPI < −0.8) or extreme wet (SPI > 1.5) conditions during the summer monsoon season, the year can then not be declared as the normal monsoon. Therefore, we recommend three-step criteria to examine if the summer monsoon rainfall was normal or not, including (1) all-India averaged summer monsoon rainfall departure should be within ±10%, (2) standardized departure of CSCI should be less than 1, and (3) not more than three clusters should be under drought or extreme wet conditions.
To further illustrate this aspect, we identified 13 years declared normal based on the all-India averaged summer monsoon rainfall (table S2), with summer rainfall anomalies within ±10% of their long-term mean. Out of these 13 summer monsoon seasons, eight (1962, 1971, 1984, 1991, 2001, 2005, 2006, and 2008) summer monsoon had more than one standard deviation of CSCI, which indicate the high intra-seasonal rainfall variability leading to severely dry and wet conditions. We estimated the area under drought and wet conditions each week during the summer monsoon season (figure S7). We find drought and wet conditions affected the country during the summer monsoon season. Therefore, drought and wet extremes occurred, which might have implications for water resources and agriculture. Similarly, five (1925, 1928, 1939, 1974, and 2017) out of these thirteen years had high spatial variability of the summer monsoon season. More than three clusters witnessed drought or extreme wet conditions in all five years. We estimated anomalies of geopotential height (850 hPa), wind (850 hPa), and SST during the summer monsoon season for these 13 years (figures S8-S10, table S3) to examine the anomalous patterns in SST and atmospheric fields. While these anomaly fields do not represent intra-seasonal (or temporal) variability in rainfall during the summer monsoon season, these exhibit abnormal conditions during the monsoon season. During the summer monsoon season, westerly winds transport moisture from the Indian ocean to land due to low pressure/ geopotential height in the northern and central parts of the country as shown by climatological fields constructed for the 71 normal years during 1901-2021 (figure S8). We find that several years (among 13) experienced positive SST anomalies over the Pacific Ocean (figures S9 and S10). Warmer SST anomalies over the Pacific are linked with a reduction in the summer monsoon rainfall over India (Kumar et al 1999, Mishra et al 2012. In a few years (1971, 1974, 1962, and 2001), cool SST anomalies were observed over the Pacific Ocean during the summer monsoon season. Atmospheric anomalies of geopotential height and wind do not show a consistent pattern for all the selected years, which is expected (figures S9 and S10) as these can be influenced by both temporal and spatial variability in the monsoon rainfall. However, positive geopotential anomalies and anticyclonic wind patterns can be seen for the years that witnessed drought in some parts of the country (figures S9 and S10). As anomalies are based on the seasonal mean, these may not capture the temporal variability of the summer monsoon rainfall reflected by the cumulative CSCI. Moreover, drought and wet extremes during these years did not affect the entire country. Therefore, unlike in the previous studies (Webster andYang 1992, Wang et al 2008), relatively stronger monsoon circulation anomalies were not observed. Overall, incorporating spatial and temporal variability of dry and wet extremes in the monsoon declaration can be essential for understanding the impacts of extremes in India.

Summary and conclusions
India experiences droughts and floods during the summer monsoon season, impacting water availability, agricultural activities, and the socio-economic well-being of millions of people. Moreover, drought and wet extremes can occur during the same monsoon season. However, the temporal and spatial variability and occurrence of drought and wet extremes are not accounted for in the normal summer monsoon declaration. Out of 121 years, 84 summer monsoon seasons were declared normal. On the other hand, 23 years were drought years, and 14 years experienced surplus monsoon. The drought years 1972, 1918, and 2002 experienced the highest precipitation deficit during the summer monsoon. The 1917 summer monsoon was the wettest, with 21.6% more rainfall than its long-term mean. Drought and surplus years affected a large part of the country. However, a profound temporal and spatial variability in rainfall and associated extremes was noted during the normal monsoon years. For instance, in the summer monsoon of 2021, the initial and final phases of the summer monsoon were wetter than normal, while a large part of the country was affected by drought during the mid-season. However, the 2021 summer monsoon was declared normal based on rainfall anomaly.
We accounted for the temporal and spatial variability of the summer monsoon using a CSCI and clustering approach. The occurrence of wet and drought extremes during different times of the summer monsoon season was estimated using the cumulative WSCI and DSCI, respectively. Similarly, we identified eight homogenous clusters based on the longterm summer monsoon rainfall. Out of 84 normal monsoon years, we found that 13 had high temporal and spatial variability that was not accounted for as the declaration was based on the rainfall anomaly only. During the five years, more than three (out of a total eight) clusters experienced drought or extreme wet conditions. Moreover, in the remaining eight years, the standardized departure of CSCI was more than one, indicating high temporal variability of rainfall during the summer monsoon season. Around 1000 people lost their lives due to extreme precipitation and floods in India and considerable damage to agriculture in the 2017 monsoon, which was declared normal (DWE 2017). Similarly, around 2051 people lost their lives due to floods and heavy rain during the summer monsoon season of 2005 (DWE 2005). Gujarat and Maharashtra states were the worst affected. However, the summer monsoon was declared normal, with a surplus precipitation of 9.6%. Part of the northeastern region witnessed moderate to severe drought during the summer monsoon of 2005. During the normal summer monsoon of 2008, 1800 people lost their lives due to floods (DWE 2008). Therefore, our findings highlight that spatial and temporal variability in mean and extreme rainfall must be incorporated in the declaration of the summer monsoon.
The proposed framework can be used to examine the spatial and temporal variability of extremes during the monsoon season for any year, regardless of whether a year is normal. For instance, cumulative WSCI can provide the role of temporal variability of wet extremes for a surplus monsoon year. At the same time, the number of clusters with positive SPI (SPI greater than +1.5) can highlight the regions that witnessed wet extremes during the summer monsoon season. On the other hand, cumulative DSCI can highlight the role of the temporal variability of dry extremes during a deficit monsoon year. The clusters with a deficit (SPI less than −0.8) can help us identify the regions affected by drought.
Based on the findings, we conclude the following: • The existing declaration of the summer monsoon rainfall condition is based on the all-India averaged rainfall at the end of the season. All India precipitation anomalies above 10% and less than −10% represent the drought and surplus monsoon season. The normal summer monsoon is declared if the all-India average rainfall anomaly is within ±10% from its long-term mean. • Prolonged active and break spells during the monsoon season led to wet and drought extremes, which can occur during the same season. Occurrences of drought and wet extremes at different times can still lead to a normal monsoon at the end of the season, notwithstanding the profound implications. The combined severity and coverage index (CSCI) incorporate temporal variability in the summer monsoon rainfall and associated extremes. • Spatial variability of rainfall and drought/wet extremes is considered using the clustering approach. Any monsoon season to be declared normal should have two-thirds or more clusters free from drought or extreme wet conditions. • Considering the all-India averaged precipitation anomaly, CSCI, and clustering approach, we identified 13 years that should not have been declared normal out of 84 normal monsoons during the last 121 years. The proposed framework considers the spatial and temporal occurrence of drought and wet extremes and can be essential for impact assessment.

Data availability statement
The data that support the findings of this study are available upon reasonable request from the authors.