Spatial synchrony, temporal clustering and dominant driver of streamflow droughts in Peninsular India

Understanding the spatio-temporal evolution of streamflow droughts and their relationship with potential causative processes is critical for effective drought management. This study assesses spatial synchrony and temporal clustering of streamflow droughts in six major river basins of Peninsular India. The importance of baseflow, rainfall deficits, soil moisture deficits and high temperatures in triggering streamflow droughts is also investigated to identify the dominant driver during the period 1981–2015. Spatial synchrony of streamflow droughts is investigated using multivariate Ripley’s K function and temporal clustering is evaluated using univariate Ripley’s K at various timescales. The interactions of streamflow droughts with potential causative processes are investigated using event coincidence analysis. At regional scale, streamflow droughts in peninsular catchments show strong spatial synchrony even at longer timescales. However, at basin scale, droughts in the catchments show strong spatial synchrony only at smaller timescales, behave independently of each other and achieve asynchrony with time, especially at longer timescales. Streamflow droughts show the strongest temporal clustering at smaller timescales and the strength of clustering decreases after a 3 year timescale. Rainfall deficits primarily control streamflow droughts in Peninsular India at a range of lags, except just before the onset of drought events where baseflow dominates. In addition, trigger coincidence rates of baseflow are lower than rainfall deficits but higher than soil moisture deficits and high temperatures at longer time lags.


Introduction
Streamflow droughts are driven by a variety of hydrometeorological processes, last for longer durations than other types of droughts (Van Loon and Van Lanen 2012, Barker et al 2016, Tijdeman et al 2022) and spread over larger spatial domains (Brunner and Chartier-Rescan 2024).Water scarcity begins with rainfall deficits, further develops into a soil moisture drought due to enhanced evapotranspiration and the situation is worsened into surface water deficits with time.Meteorological droughts have been the focus of most of the previous studies in India (Das et al 2016, Pai et al 2017, Sharma and Mujumdar 2017, Adarsh et al 2019, Udmale et al 2020), despite the implications of meteorological drought propagation to streamflow droughts (Shah and Mishra 2020).Drought propagation and climatecatchment-soil control on streamflow droughts are examined in India (Bhardwaj et al 2020, Ganguli et al 2022).However, spatio-temporal evolution and causative mechanisms of streamflow droughts have not yet been explored in India.
Spatial synchrony (Gaupp et al 2020, Singh et al 2021, Mahto and Mishra 2023, Mondal et al 2023) and temporal clustering of droughts (Brunner and Stahl 2023) have recently received attention due to the amplified risk on water resources management, agriculture production and the global economy.Spatial concurrence as well as the consecutive occurrence of streamflow droughts can cause severe water scarcity and crop failures in India.Therefore, investigating the strength of streamflow drought across space and time in major river basins is crucial for better management of water resource systems.The concepts of spatial compounding where multiple locations are simultaneously affected and temporal compounding where extreme events occur in succession have recently been introduced to advance research on compound events.However, guidelines on how to analyze these events are still lacking (Hao et al 2018, Bevacqua et al 2021).This study identifies new tools to analyze the compounding of streamflow droughts across space and time.
Physical processes responsible for the initiation and propagation of droughts are not well understood (Van Loon and Van Lanen 2012, Apurv and Cai 2020) due to complex interactions across multiple temporal and spatial scales.The interactions of hydrological droughts with climate and catchment characteristics, soil properties, catchment storage and release states have been investigated in various studies (Van Loon and Laaha 2015, Konapala and Mishra 2020, Ganguli et al 2022).In contrast to the catchment-based perspective, Brunner et al (2022) used an event-based perspective to understand the interactions of streamflow droughts of different severity with precipitation deficits, temperature and snow-water-equivalent anomalies.The event-based approach is more robust in assessing the importance of drivers since it extracts events of all magnitudes using a variable threshold approach.This approach identifies severe and moderate drought events within a catchment.Unprecedented droughts in South India and recent water scarcity (Ghosh and Srinivasan 2016, Mishra et al 2021) require detailed investigation to explicitly describe the interactions of streamflow droughts with potential governing hydrometerological processes.Therefore, an event-based approach is used in this study to identify the dominant driver of streamflow droughts in Peninsular India.
Flow regulations are found to affect the timing, magnitude and frequency of downstream low and high flows (Magilligan and Nislow 2005, Wang et al 2017, Volpi et al 2018, Brunner 2021).The natural flow regime of Indian catchments has been significantly altered due to the construction of dams.Streamflow is regulated for various purposes such as irrigation, hydropower, water supply and flood control.The effect of flow regulations is assessed in this study to understand the impact of human interventions on drought duration and severity.Understanding the extent to which reservoir regulations affect drought characteristics is crucial for designing better reservoir operation rules in Peninsular catchments.
An assessment of streamflow droughts across a range of spatio-temporal scales and their linkages with plausible causative hydrometeorological processes can provide useful insight for improving the scientific understanding of this phenomenon in Indian catchments.Specifically, we investigate (1) the spatial synchrony of streamflow droughts at multiple catchments to understand when the droughts in catchments synchronize, at which temporal windows they behave independently and when they become completely asynchronous, (2) the occurrence of drought events in close temporal succession to understand the strength of temporal clustering and (3) the association of streamflow droughts with hydrometeorological processes to identify the dominant driver in Peninsular India.

Study area
The locations of 70 catchments (including nested catchments) in the study area are shown in figure 1(a).Mahanadi, Narmada, Tapi, Godavari, Krishna and Cauvery are six major river basins of Peninsular India.The west flowing rivers Narmada and Tapi join the Arabian sea and four other east flowing rivers that drain into the Bay of Bengal.The largest river basin of Peninsular India is Godavari (312 812 km 2 ) and the smallest river basin is Tapi (65 145 km 2 ).The longest river is Godavari (1465 km) and the shortest river is Tapi (724 km).The elevation varies between a minimum of 1 m to a maximum of 937 m (figure S1(a)).Spatial variation of Aridity Index (AI) defined as the ratio of mean annual precipitation to mean annual potential evapotranspiration is shown in figure S1(b).According to the United Nations Environment Programme climate classification scheme (UNEP 1997) peninsular catchments have semi-arid (AI 0.2-0.5),dry sub-humid (AI 0.5-0.65)and humid (AI > 0.65) climate conditions.

Hydrometeorological data
Daily streamflow time series for 70 catchments of Peninsular India for a period of 35 years from 1981-2015 are obtained from the India-Water Resources Information System (India-WRIS).Methods for synthesizing missing streamflow records (Beauchamp et al 1989, Elshorbagy et al 2000) are used to fill the gaps in the streamflow time series.Fixed interval smoothing constructed via structural time series model is used to fill in the missing values.The smoothing method uses the slope information from two previous values to the missing value and two values following the missing values for smooth interpolation.Streamflow records for a few stations are available for a longer time.However, a common data length is required to investigate the spatial synchrony of streamflow droughts.Therefore, a common period of 35 years is selected for 70 catchments in this study.Catchments are delineated using a digital elevation model obtained from the Shuttle Radar Topographic Mission at 30 m spatial resolution.A high-resolution daily rainfall data set on a grid size of 0.25 where b i is the baseflow, α is the recession constant, Q i is the discharge for time step i and BFI max is the maximum baseflow index modelled by the algorithm.Eckhardt (2005) suggested a BFI max value of 0.25 for the catchments with hard rock aquifers.Peninsular catchments are underlain by hard rock aquifers and this value has been used in recent studies to compute baseflow in the study area (George and Sekhar 2022, Sharma and Mujumdar 2024).Therefore, BFI max = 0.25 is used in the Eckhardt filter.The master recession curve (MRC) is used to compute the recession constant as per the procedure given in the WMO manual on low-flow estimation and prediction (WMO 2008).The start of recession is marked below the 70th percentile flow (Q 70 ) threshold at least two days after the peak flood discharge.The segment length is computed and the MRC is obtained by plotting pairs of Q t−1 and Q t .The slope of the MRC provides the estimate of recession constant α (figure S2).

Extraction of drought events and spatial extent
The standardized streamflow index (SSI), a probability-index-based approach is used to characterize the anomalies in streamflow at an averaging period of 1 month.The SSI expresses streamflow as a non-exceedance probability and is calculated using the same procedure as the standardized precipitation index (Mckee et al 1993).The use of a variable threshold-level approach has increased in recent studies (Van Loon and Laaha 2015, Brunner et al 2021, 2023, Brunner and Stahl 2023).However, the characteristics of SSI-1 drought events extracted using a limit value of −0.84 are found to be close to what is being identified using the monthly variable threshold method (Sutanto and Van Lanen 2021).A limit value of SSI − 1 < −0.84 offers a fair comparison with the commonly used 20th percentile threshold approaches.The SSI has been used to study the spatio-temporal patterns of streamflow droughts (Shukla and Wood 2008, Teutschbein et al 2022).
Both the SSI and empirical streamflow percentiles are comparable over space and time (Tijdeman et al 2020).Therefore, the use of SSI for an averaging period of 1 month with a limit value of −0.84 to identify droughts is justified.The spatial extent of the streamflow drought events is determined at monthly scale as the percentage of catchments affected by drought during a specific month.The number of catchments affected by droughts are counted and divided by the total number of catchments.A time series of monthly spatial extent of droughts is created for the period 1981-2015 and a non-parametric Mann-Kendall test is applied to detect the presence of statistically significant trends (Mann 1945).Major streamflow drought events are identified based on the largest spatial extent affected during a specific month of a year.This preliminary investigation is important to understand the behavior of streamflow droughts across space.

Influence of flow regulations on drought characteristics
The natural flow regime of Indian catchments is significantly altered due to the construction of reservoirs.The effect of flow regulations is assessed using a 'pre-post-disturbance' approach to understand the impact of human interventions on drought duration and severity.Stream networks are delineated for six river basins and locations of dams are marked as per the information from the National Register of Large Dams (CWC 2019).Streamflow gauges, which lie downstream on the same streamflow network and have a good length of records for pre and postdisturbance periods, are identified.
Drought characteristics are computed based on the run theory (Yevjevich 1967), also referred to as the threshold level method.Drought duration is the length of consecutive time series when the SSI is below the threshold value −0.84.Drought severity is defined as the sum of the SSI below the threshold.Streamflow records are divided into the undisturbed and disturbed periods.Drought duration and drought severity are computed for the two periods.The length of streamflow records varies between a maximum of 49 years  to a minimum of 35 years  for comparison of drought characteristics.Changes in the drought characteristics are estimated as (C D − C U ) /C U , where C D and C U are the mean characteristics after the disturbance and before the disturbance, respectively.

Spatial synchrony of streamflow droughts
Multivariate Ripley's K-function (Doss 1989), a modification of Ripley's K (Ripley 1977) is used to describe the spatial characteristics of drought events in multiple catchments.It tests whether the occurrence of drought events in the catchments of a basin co-occur more than what is expected purely by chance.Ripley's K in its original form is widely used to model spatial point processes (i.e.data on the location of events).The modification in the form of multivariate Ripley's K is used to model the timing of events (temporal point process) in this study without any assumption about spatial independence of droughts.The applications of multivariate Ripley's K include testing the spatial synchrony of occurrence of forest fires (Gavin et al 2006); long-term interactions between climate variability and wildfires (Schoennagel et al 2007, Gartner et al 2012, Rother and Grissino-Mayer 2014) and spatial synchrony of drought-induced plague in Europe (Yue and Lee 2020).This approach is suitable for temporal point processes, avoids loss of information from merging events into bins from time series of event frequency and does not assume any periodic structure in the data (Gavin et al 2006).It preserves the firstorder properties of each site in each randomization and only addresses the dependence between the two records (Wiegand and Moloney 2004).For the bivariate case, K-function gives the number of events at site Y occurring within ±t temporal window of each event at site X, scaled by a factor T/n X n Y , where T is the length of record, and n X and n Y are the number of events at sites X and Y, respectively.Drought events between the two sites X and Y show synchrony for K > 2t, independence if values of K are close to 2t and asynchrony for K < 2t within a time window t.
Gavin et al (2006) added two modifications to the empirical K-function presented in Doss (1989) to make it suitable for temporal point processes.An edge correction was used to make the K-function unbiased: where X i and Y j are times of events, I is the identity function and w X i , Y j is a mirror edge correction.This edge correction causes a difference in values of K based on whether the distances are measured from a site X to Y or from site Y to X.A second modification was made to produce an unbiased estimate KXY (t) by the weighted averaging of KXY (t) and KYX (t) , Ripley's K-function is suitable after these two modifications for examining whether events at one site occur closer to the events at another site irrespective of direction.The K-function is easier to interpret if its mean and variance are stabilized over the time window t.Therefore, the results for spatial synchrony of droughts are presented using the L function LXY (t) = KXY (t) /2 − t in this study.We constructed the 95% confidence envelops for LXY (t) from 1000 randomizations of the records and interpreted the results using these envelopes.Streamflow droughts at two sites show statistically significant synchrony in a temporal window t if LXY (t) falls above the upper confidence band, show independence if LXY (t) falls within the confidence bands and shows statistically significant asynchrony if the LXY (t) value is less and falls below the lower confidence band (figure S3).Multivariate extension of the K-function compares events at one site to the aggregated events at all other sites, and weighted average is computed as for the bivariate case.

Temporal clustering of streamflow droughts
The clustering of droughts across time is quantified at various timescales using univariate Ripley's K.A few recent studies used this function to estimate the temporal clustering of extremes at different timescales (Barton et al 2016, Tuel and Martius 2021, Brunner and Stahl 2023).A binary time series of event occurrences is created where the presence of the event is indicated by 1 and its absence is indicated by 0. Ripley's K function is: where λ is the density and E is the expected value of events.K (t) gives the average number of events within a time window 't' around a randomly chosen event (Dixon 2002).The term λ −1 is dropped for simplicity since it does not affect the results as long as the same estimator is applied to the empirical and simulated data (Barton et al 2016).The significance of clustering is tested using a bootstrap experiment by simulating 1000 binary time series from a homogeneous Poisson process using the same mean occurrence rate of drought events as observed in the original binary time series.Ripley's K function is computed for the 1000 simulated binary time series and sampling distribution of the function is derived.The critical value is computed at the 95th quantile of the sampling distribution.If the observed Ripley's K is greater than the critical value, the time series has significant temporal clustering.The procedure is illustrated for a randomly selected catchment in figure S4.

Quantifying statistical interdependence of droughts on physical processes
Event coincidence analysis (ECA) provides a framework to quantify the strength, directionality and time lag of statistical interdependency between two event series (Donges et al 2011, 2016, Siegmund et al 2017).
ECA is used to test for possible causal influence based on the hypothesis that baseflow, rainfall deficits, soil moisture deficits and high temperatures are triggering the streamflow droughts.The hypothesis is tested separately for all the pairs of variables with streamflow droughts and the results are used to determine the variable that has a higher influence on the streamflow droughts in Peninsular India.Figure S5 demonstrates the use of ECA for investing the triggering effect of causative mechanisms on streamflow droughts.
Coincidences of extreme events are counted in two event series X and Y and the strength of their statistical interrelationship is quantified using a measure called trigger coincidence rate (r t ).It measures the fraction of Y events, which are followed by X events, to test whether the extreme Y event has a triggering effect on the extreme X event.The trigger coincidence rate is defined as follows: (5) where F (•) is the Heaviside function, which conveys information on whether the Y event has a triggering effect on the X event or not based on the values 1 and 0 at any time t i .I [0,∆T] investigates the triggering effect, τ is the time lag between X and Y and ∆T is the coincidence interval.N X and N Y are the number of X and Y events, respectively.The values of r t are 0 in the case of the complete absence of a triggering effect and 1 if X events succeed all the Y events.The statistical significance of the coincidence rates is tested using the Poissonian approximation described in Siegmund et al (2017) by testing the null hypothesis that the number of coincidences is explained by two independent series of randomly distributed events.The null hypothesis is rejected if the p-value is smaller than the defined significance level α.

Spatio-temporal evolution of streamflow droughts
Percentage spatial extent affected by streamflow droughts is extracted to understand the strength of droughts across space at monthly scale.Spatial extent varies from 0% to a maximum of ∼83% and shows a statistically significant increasing trend during the period 1981-2015 (figure 1(b)).An alternative way of determining the spatial extent is by considering the actual area of catchments that is not found to alter the main conclusion of the study.Non-nested catchments are separated based on the criterion that there is no intersection with any upstream catchment to understand the effect of information contained in the downstream catchment (figure S6(a)).Nestedness of catchments does not alter the main finding of increasing trend in the spatial extent of streamflow droughts (figure S6(b)).Major spatially extensive streamflow drought events are observed in monsoon months (figure S7).
It is important to quantify the influence of flow regulations on drought characteristics since the river flows are largely altered in Peninsular catchments.Total 31 dams are considered and 25 downstream streamgauges, which have longer flow records, are identified on the same flow lines (figure 1(c)).In the case of two dams upstream of a streamgauge, the impact is evaluated first for the structure that came first, and the year of construction of the old structure decides the length of records for the new structure downstream of the older one.The predisturbance period of the new dam begins with the year of construction of the older dam lying upstream and ends at its own year of construction.Flow regulations have negatively impacted the drought characteristics in Peninsular India.Streamflow droughts are longer and more severe in the post-disturbance period (figure 1(d)).
The synchrony of streamflow droughts at spatial scale is tested by considering all catchments of Peninsular India at once (regional scale analysis) and separately in each river basin (basin scale analysis).A series recording the time of occurrence of drought events is created for each catchment.A horizontal line on the y-axis represents a catchment and black circles indicate the month of occurrence of drought events in that catchment (figure 2(a)).Signatures of spatial synchrony of droughts across Peninsular catchments are visible before 100 months and after 210 months in figure 2(a).Multivariate Ripley's K function will explicitly describe the synchronous behavior.The entire record series is divided into two halves at 210 months and K functions are computed for each half separately to detect changes in synchrony over time.Multivariate Ripley's K identifies the temporal windows of synchrony, independence and asynchrony of streamflow droughts.The results are presented in terms of L-function for easy visual inspection (figures 2(b) and (c)).Statistically significant spatial synchronies for streamflow droughts across 70 catchments of Peninsular India in the first half length of records are observed up to a temporal window of 84 months, behave independently of each other after that and become asynchronous after ∼140 months (∼12 years) (figure 2(b)).The second half (1998)(1999)(2000)(2001)(2002)(2003)(2004)(2005)(2006)(2007)(2008)(2009)(2010)(2011)(2012)(2013)(2014)(2015) shows statistically significant synchronization of streamflow droughts throughout the study period (figure 2(c)).Regional scale analysis shows strong spatial synchronization of streamflow droughts between Peninsular catchments at longer timescales.
At basin scale, streamflow droughts in the Narmada river basin synchronize strongly across space for 72 months, behave independently after that (figure 3  in the Godavari catchments, remains independent for the temporal window from 13.5-18 years and becomes asynchronous after 18 years (figure 3(d)).
The Krishna river basin shows strong spatial synchrony for the longest timescale of 168 months, i.e. 14 years, drought events in the catchments behave

Interactions of streamflow droughts with hydrometeorological processes
The relationship of streamflow droughts with baseflow, rainfall deficits, soil moisture deficits and high temperatures is investigated with trigger coincidence rate r t at significance level α = 0.05 (figure 5).Lag parameter τ is varied between 1-3 months to make sure that causative mechanisms occur before the streamflow drought event.Three months represent a typical seasonal scale that can capture the effect of slow development of drought due to the deficit conditions in the previous season.Droughts generally develop over longer timescales and, therefore, it is important to consider longer lags.The coincidence interval ∆T is also varied between 1-3 months to capture the drought event occurring a few months after the occurrence of a causative mechanism.Extreme events for all the variables are extracted using the fixed threshold approach at monthly scale.Deficits are extracted for baseflow, soil moisture and rainfall below the 20th percentile threshold and extremes of high temperatures are extracted above the 80th percentile threshold.The dominant driving mechanism is identified based on the higher rate of trigger coincidences.Soil moisture deficits and high temperatures show low triggering effects on streamflow droughts in most of the Peninsular catchments (figure 5).The trigger coincidence rates are high for baseflow compared to the other three variables at a smaller time lag of 1 month, which indicates the strong influence of groundwater storage just before the occurrence of a streamflow drought event.We observe that rainfall deficits have the highest triggering effect on streamflow droughts at longer lags.In addition, baseflow shows higher trigger coincidence rates compared to soil moisture deficits and high temperatures.Similar results are obtained for streamflow droughts defined below the 15th percentile threshold (figure S11).

Discussion
Our results show an increase in the spatial extent of streamflow droughts in Peninsular India, which means that these droughts are synchronizing more in space and are not limited to the boundaries of a catchment or river basin.Spatially extensive streamflow droughts can cause synchronous crop failures and make water management more challenging since the upstream catchments or neighbouring water-abundant catchments can no longer provide water supply to the downstream catchments.
The impact of flow regulations on droughts is assessed by comparing the characteristics before and after the disturbance.Drought duration as well as severity have increased after the construction of dams in Peninsular catchments.However, human influences do not have a consistent effect on streamflow drought characteristics.Inconsistency of the effects might be due to different reservoir operation rules and priorities that might change over time.Tijdeman et al (2020) found that the same human influence may intensify or mitigate streamflow drought characteristics, i.e. the effect of human influence can change over time.More detailed information about the type and degree of all human influences, such as groundwater abstractions, reservoir operations and other water demands, can more accurately quantify the human influences in the catchments.
Our findings clearly show strong statistically significant spatial synchrony of streamflow droughts for longer timescales at regional scale.On the other hand, basin scale analysis reveals that droughts in the catchments achieve asynchrony with time, especially at longer timescales.Consequently, the conclusion on spatial synchronization of droughts is highly dependent on the scale of the analysis.Further research is needed to investigate the causes of spatial synchrony of streamflow droughts.Identifying the hydrological processes and catchment characteristics that control the spatial synchrony of streamflow droughts at different scales presents a fascinating intellectual challenge for hydrologists.
Temporal clustering of streamflow droughts was recently analyzed using a quasi-global data set based on the hypothesis that a catchment with less storage has a higher likelihood of showing temporal clustering (Brunner and Stahl 2023).However, the study does not include Indian catchments.Peninsular catchments are rainfall driven and receive most of the storage during monsoon months.Temporal clustering of streamflow droughts is expected due to the recurrent failure of monsoon, persistent soil moisture deficits and low baseflow contributions due to lower groundwater levels.Streamflow droughts show strongest clustering at smaller timescales and the strength of clustering decreases after a 3 year timescale.The presence of clustering across ∼11% catchments at a longer timescale of 10 years indicates the decadal drought clustering in a few catchments of Peninsular India.
This study extends our knowledge of spatiotemporal evolution and the underlying process controls of streamflow droughts.The study finds rainfall deficits as the dominant driver of streamflow droughts in Peninsular India.In addition, the triggering effect of baseflow is considerably higher than soil moisture deficits and high temperatures.This can be linked to the overuse of groundwater resources in India.Groundwater is depleting at an alarming rate in India (Devineni et al 2022, Bhattarai et al 2023).Some regions are consuming groundwater faster than it is naturally replenished (Rodell et al 2009).Drought in a catchment typically begins with accumulated rainfall deficits, propagates through different segments of a hydrological system and results in other types of droughts (Mishra and Singh 2010).The complex interaction of baseflow and accumulated rainfall deficits cannot be ignored since baseflow is often characterized by the slow response of a catchment to rainfall (Lee and Ajami 2023, Mao et al 2023).ECA is performed with individual drivers in this study.However, droughts can arise through different combinations of causative mechanisms.A possible extension of the present work is to investigate the combined effect of causative mechanisms on streamflow droughts.Multivariate extreme value models are suitable for studying the combined effect of physical processes leading to an extremal phenomenon (Sharma and Mujumdar 2022).Drought generating mechanisms other than the four considered in this study can also be included.

Conclusion
Changes in the spatial extent of streamflow droughts, influence of reservoir flow regulations on drought characteristics, temporal variation of spatial synchrony of droughts, temporal clustering of droughts and interactions of streamflow droughts with potential hydrometeorological processes are investigated in Peninsular Indian catchments.The key findings of the study are as follows: 1) An increase in the spatial extent of streamflow droughts is observed in Peninsular India for the period 1981-2015.2) Streamflow droughts are longer and more severe downstream of the dams in Peninsular India.However, this effect might change over time.3) Results on spatial synchronization of droughts are highly dependent on the scale of the analysis.At regional scale, strong spatial synchrony is observed even at longer timescales.However, at basin scale, droughts in the catchments show strong spatial synchrony only at smaller timescales, behave independently of each other and achieve asynchrony with time, especially at longer timescales.4) Temporal clustering is observed in all Peninsular catchments because the storage of catchments is mostly rainfall driven confirming that precipitation-driven catchments are more prone to temporal clustering of streamflow droughts.5) Accumulated rainfall deficits primarily control streamflow droughts in Peninsular India except just before the onset of a drought event where baseflow dominates.We also observe that the triggering effect of baseflow is higher compared to soil moisture deficits and high temperatures.
These investigations are useful for understanding the strength of streamflow droughts and improving the scientific understanding of drought generation processes.Results of the study can assist in better management of water resource systems for drought mitigation at catchment as well as basin scale.

Figure 1 .
Figure 1.(a) Locations of 70 streamflow gauges with catchment boundaries in six major river basins of Peninsular India, (b) percentage of catchments affected by streamflow droughts for a period of 35 years from 1981-2015, (c) locations of reservoirs and streamflow gauges lying downstream on the same flow lines and (d) impact of reservoir flow regulations on drought duration and severity.Horizontal lines of boxplots represent the minimum, 25th percentile, 50th percentile, 75th percentile and maximum values of drought characteristics.

Figure 2 .
Figure 2. (a) Month of occurrence of drought events marked with black circles in rows for Peninsular catchments, (b) and (c) multivariate Ripley's K across all Peninsular catchments is shown by a black dashed line and a 95% confidence band is shown by a sky-blue polygon.Streamflow droughts of Peninsular India are synchronizing across space in two temporal windows <84 months and >210 months.

Figure 3 .
Figure 3. Multivariate Ripley's K function to describe spatial synchrony across catchments in six major river basins of Peninsular India.Function estimates are shown by a black dashed line and the 95% confidence band is shown by a sky-blue polygon.Drought events in the Mahanadi river basin synchronize less in space and occur independently after 3 years, whereas the Godavari and Krishna river basins show statistically significant synchrony across space for longer durations of ∼14 years.

Figure 4 .
Figure 4. (a) Univariate Ripley's K for three randomly selected catchments at different timescales of 1 , 3 and 6 months, 1, 3 and 6 years.Statistically significant observed estimates at significance level 0.05 are marked by red circles and the confidence intervals of simulated estimates are indicated by sky-blue polygons, (d) statistically significant Ripley's K values at annual scale, insignificant temporal clustering is marked with grey points and (e) percentage of catchments showing temporal clustering at various time windows.Strength of temporal clustering is strongest at small timescales, indicating the occurrence of consecutive streamflow droughts in close temporal succession.

Figure 5 .
Figure 5. Triggering effect of (a) baseflow, (b) soil moisture deficits, (c) rainfall deficits and (d) high temperatures on streamflow droughts.Trigger coincidence rates are high for rainfall deficits compared to other causative mechanisms at lags greater than 1 month.Triggering effect of baseflow is higher than soil moisture deficits and high temperatures.
• (Pai et al 2014) and daily maximum temperature data set on a grid size of 1 • (Srivastava et al 2009) are obtained from the India Meteorological Department.A monthly soil moisture data set at a depth of 1.6 m (Fan and van den Dool 2004) on a grid size of 0.5 • is obtained from the NOAA Climate Prediction Center.
Brunner M I and Stahl K 2023 Temporal hydrological drought clustering varies with climate and land-surface processes Environ.Res.Lett.18 034011 Brunner M I, Swain D L, Gilleland E and Wood A W 2021 Increasing importance of temperature as a contributor to the spatial extent of streamflow drought Environ.Res.Lett.Water 10 16-21 Konapala G and Mishra A 2020 Quantifying climate and catchment control on hydrological drought in the continental United States Water Resour.Res.56 1-25 Lee S and Ajami H 2023 Comprehensive assessment of baseflow responses to long-term meteorological droughts across the United States J. Hydrol.626 1-10 Magilligan F J and Nislow K H 2005 Changes in hydrologic regime by dams Geomorphology 71 61-78 Mahto S S and Mishra V 2023 Increasing risk of simultaneous occurrence of flash drought in major global croplands Environ.Res.Lett.18 044044 Mann H B 1945 Nonparametric tests against trend Econ.Soc. 13 245-59 Mao R, Shi A, Song J, Xu W, Tang B and Li B 2023 Response of the runoff process to meteorological drought: baseflow index as an important indicator J. Environ.Manage.345 118843 Mckee T B, Doesken N J and Kleist J 1993 The relationship of drought frequency and duration to time scales Eighth Conf. on Applied Climatology (Boston, Massachusetts) (American Meteorological Society pp 179-84 Mishra A K and Singh V P 2010 A review of drought concepts J. Hydrol.391 202-16 Mishra V, Thirumalai K, Jain S and Aadhar S 2021 Unprecedented drought in South India and recent water scarcity Environ.Res.Lett.16 054007 Mondal S, Mishra K A, Leung R and Cook B 2023 Global droughts connected by linkages between drought hubs Nat.Commun.14 1-13 Pai D S, Guhathakurta P, Kulkarni A and Rajeevan M N 2017 Variability of meteorological droughts over India Observed Climate Variability and Change over the Indian Region ed M N Rajeevan and S Nayak (Springer) pp 73-87 Pai D S, Sridhar L, Rajeevan M, Sreejith O P, Satbhai N S and Mukhopadyay B 2014 Development of a new high spatial resolution (0.25 • ×0.25 • ) long period (1901-2010) daily gridded rainfall data set over India and its comparison with existing data sets over the region Mausam 65 1-18 Ripley B D 1977 Modelling spatial patterns J. R. Stat.Soc.B 39 172-92 Rodell M, Velicogna I and Famiglietti J S 2009 Satellite-based estimates of groundwater depletion in India Nature 460 999-1002 Rother M T and Grissino-Mayer H D 2014 Climatic influences on fire regimes in ponderosa pine forests of the Zuni Mountains, NM, USA For.Ecol.Manage.322 69-77 Schoennagel T, Veblen T T, Kulakowski D and Holz A 2007 Multidecadal climate variability and climate interactions affect subalpine fire occurrence, Western Colorado (USA) Ecology 88 2891-902 Shah D and Mishra V 2020 Drought onset and termination in India J. Geophys.Res.Atmos.125 e2020JD032871 Sharma S and Mujumdar P P 2022 Modeling concurrent hydroclimatic extremes with parametric multivariate extreme value models Water Resour.Res.58 e2021WR031519 Sharma S and Mujumdar P P 2024 Baseflow significantly contributes to river floods in Peninsular India Sci.Rep. 14 1-10 Sharma S and Mujumdar P 2017 Increasing frequency and spatial extent of concurrent meteorological droughts and heatwaves in India Sci.Rep. 7 15582 Shukla S and Wood A W 2008 Use of a standardized runoff index for characterizing hydrologic drought Geophys.Res.Lett.35 1-7 Siegmund J F, Siegmund N and Donner R V 2017 CoinCalc-a new R package for quantifying simultaneities of event series Comput.Geosci.98 64-72 Singh J, Ashfaq M, Skinner C B, Anderson W B and Singh D 2021 Amplified risk of spatially compounding droughts during co-occurrences of modes of natural ocean variability npj Clim.Atmos.Sci. 4 7 Srivastava A K, Rajeevan M and Kshirsagar S R 2009 Development of a high resolution daily gridded temperature data set (1969-2005) for the Indian region Atmos.Sci.Lett. 10 249-54 Sutanto S J and Van Lanen H A J 2021 Streamflow drought: implication of drought definitions and its application for drought forecasting Hydrol.Earth Syst.Sci. 25 3991-4023 Teutschbein C, Quesada Montano B, Todorović A and Grabs T 2022 Streamflow droughts in Sweden: spatiotemporal