Late twentieth century rapid increase in high Asian seasonal snow and glacier-derived streamflow tracked by tree rings of the upper Indus River basin

Given the reported increasing trends in high Asian streamflow and rapidly increasing water demand in the Indian subcontinent, it is necessary to understand the long‐term changes and mechanisms of snow- and glacier-melt-driven streamflow in this area. Thus, we have developed a June–July streamflow reconstruction for the upper Indus River watershed located in northern Pakistan. This reconstruction used a temperature-sensitive tree-ring width chronology of Pinus wallichiana, and explained 40.9% of the actual June–July streamflow variance during the common period 1970–2008. The high level of streamflow (1990–2017) exceeds that of any other time and is concurrent with the impact of recent climate warming that has resulted in accelerated glacier retreats across high Asia. The streamflow reconstruction indicated a pronounced reduction in streamflow in the upper Indus River basin during solar minima (Maunder, Dalton, and Damon). Shorter periods (years) of low streamflow in the reconstruction corresponded to major volcanic eruptions. Extreme low and high streamflows were also linked with sea surface temperature. The streamflow reconstruction also provides a long-term context for recent high Asian streamflow variability resulting from seasonal snow and glaciers that is critically needed for water resources management and assessment.


Introduction
As a result of fast-growing populations and socioeconomic development, demand for water resources in South Asian countries is increasing rapidly, and water disputes are escalating in the South Asian subcontinent (Zawahri 2009, Qiu 2016, Li et al 2016, Zhang 2016, Reddy et al 2017, Vinke et al 2017, Biemans et al 2019. The uncertainty of the impact of climate change and human activities is one of the most important environmental challenges facing water supply assessment, with far-reaching implications on regional sustainable development (Wu et al 2017, Kundzewicz et al 2018. Thus, many studies have focused on the availability of water resources in various climate scenarios (Gosling and Arnell 2016, Tramblay et al 2018, Zhai et al 2020, Liu et al 2020. Although the South Asian monsoon generally brings abundant rainfall, the water resources of the Indian subcontinent, which is one of the most densely populated areas in the world, are unevenly distributed due to monsoon failure; as a result, South Asian countries are currently facing differing degrees of water scarcity (Cook et al 2010, Chen et al 2017, Mishra et al 2019, Zhai et al 2020. Increased meltwater from high Asian glaciers resulting from global warming has raised doubts about the capacity of freshwater supplies to meet the growing water demands of South Asia (Bolch 2017, Gao et al 2019, Pan et al 2019. South Asian socio-economic development has been greatly affected, either directly or indirectly, by the climate-driven changes in water resources. Thus, long-term river streamflow records are needed for proper water resource planning. Nevertheless, the brevity of the span of instrumental streamflow records available in South Asia has greatly hampered the development of appropriate water resource management policies and severely limited our understanding of South Asian water resources from a long-term perspective.
Moisture-sensitive tree-ring width series from high Asian river basins provide reliable highresolution streamflow records dating back several centuries or even millennia (Yuan et al 2007, Gou et al 2010, Liu et al 2010, Yang et al 2012, Cook et al 2013, Singh and Yadav 2013, Shah et al 2014, Xiao et al 2017, Panyushkina et al 2018, Rao et al 2018, Chen et al 2019a, 2019b. Tree-ring width chronologies of conifers from the Indus River basin, which is one of the most important irrigation water sources in South Asia, have been used to develop streamflow reconstructions (Cook et al 2013, Rao et al 2018. By using a network of tree-ring sites from the upper Indus basin, including a large number of climate-sensitive tree-ring width series, precipitationand temperature-related mixed streamflow data have been reconstructed, and thus can effectively reveal the history of the Indus River in terms of water resources (Rao et al 2018). Both the South Asian summer monsoon and glacial changes have significant effects on the streamflow of the Indus River (Koppes et al 2015, Mukhopadhyay and Khan 2015, Minallah and Ivanov 2019, but the effects of these factors on the streamflow have not been quantitatively established. In this study, we develop a new tree-ring width chronology of the Himalayan white pine (Pinus wallichiana) from an upper-treeline site in Northern Pakistan, and apply the reliable period of this chronology to reconstruct June-July streamflow variations in the Indus River since 1350 CE. We use this reconstruction to investigate high Asian streamflow variability resulting from seasonal snow and glacier ice. We also explore the streamflow variations in relation to natural forcings. In particular, we focus on the relationship between streamflow changes and the historical process in South Asia.

Study area
The sampling site (AST, 74 • 48 ′ E, 35 • 20 ′ N, 3450-3500 m a.s.l.) is located in arid and semi-arid regions in Northern Pakistan (figure 1), lies in the path of the westerly and South Asian monsoon is often affected by water vapour transport anomalies (Latif et al 2017, Bibi et al 2020. These characteristics make the region sensitive to climate change (e.g. Anjum et al 2019). Precipitation may exceed 1000 mm at the windward slopes near the upper treeline ; the region surrounding Astore (and the sampling site from which tree-ring cores were collected) was actually more similar to the sub-humid plateau monsoon regions based on variations in precipitation and temperature. As a result, high-altitude areas in the regions are characterised by cold, snowy winters and warm, wet summers, and covered by widespread glaciers and a wide variety of coniferous forests. Mean annual temperature and precipitation values equal approximately 486.2 mm and 9.9 • C, respectively (figure 2(A)). Average May-June precipitation and temperature range from 7.9 to 150.1 mm and from 16.7 • C to 21.8 • C, respectively. The Indus River valleys in the study area are characterised by arid and semi-arid climates (with 139 mm of precipitation in Gilgit). Seasonal distributions of streamflow and precipitation differ somewhat, and streamflow increase rapidly from June to August because of the influence of snow/glacier meltwater (figure 2(B)). High correlations among the instrumental streamflow records indicated that the streamflow of different rivers at the upper Indus River basin was responding to common factors (figure 2(C)).

Tree-ring data
One tree-ring site (AST) was sampled in the Astore region (figure 1). Increment cores were collected from living Himalayan white pine trees at breast height, using 10 mm-diameter increment borers. In total, 75 increment cores were collected from 37 trees from the AST site at the upper treeline near the glaciers in the Astore region. The biophysical environment implies that the growth of Himalayan white pine is limited by temperature at the timberline in the Karakoram (Asad et al 2017). All increment cores were mounted and polished with 400 grit sandpaper; annual ring widths were measured to the nearest 0.01 mm using a LINTAB measuring system. We used the COFECHA program to assess the cross-matching quality of all tree-ring width series (Holmes 1983). Next, we standardised all individual tree-ring width   (Indus River, 1962-1996, Kachura (Indus River, 1970, Yogo (Shyok River, 1973 and Alam Bridge (Gilgit River, 1969-1998 series, using the ARSTAN program, by conservative de-trending methods (negative exponential and straight-line curve fits) to remove non-climatic trends due to tree age and tree size while minimizing the removal of the climatic variance (Cook 1985). The detrended series were combined into the site chronology using a biweight robust mean (Cook 1985). We chose to use the standard version of the tree-ring width chronology (AST), ranging from low-to highfrequency common signals, which includes environmental and climatic signals. A minimum sample depth (number of trees ⩾ 3) was adopted to ensure that strong climate or hydrological signals were based on years when the expressed population signal (EPS) was higher than 0.85 (Wigley et al 1984). The AST chronology used in the streamflow reconstruction below was therefore truncated prior to 1350 CE, based on the threshold values.

Instrumental data and statistical methods
Monthly instrumental climate data from 1972 to 2017, included monthly mean temperature and total monthly precipitation, were obtained for the Astore climate station (74 • 20 ′ E, 35 • 55 ′ N, 1454 m a.s.l.) from the Pakistan meteorological department (figure 2(A)). Monthly mean streamflow data of the upper Indus River were obtained from the Kachura hydrological station, located at 75 • 25 ′ E, 35 • 27 ′ N, 2341 m a.s.l. (table 1). The streamflow records date from 1970 to 2008. Figure 2(B) shows the average monthly streamflow from 1970 to 2008 at the Kachura hydrologic station. Other instrumental streamflow data were also obtained from three hydrological stations, Partab Bridge (Indus River, 1962-1996, Yogo (Shyok River, 1973, and Alam Bridge (Gilgit River, 1969-1998, showing the existence of some strong signals common to several hydrological stations (figure 2(C)). Bootstrapped correlation analysis was performed for initial indication of the relationship between tree growth and monthly streamflow/climate using the DENDROCLIM2002 program (Biondi and Waikul 2004), for which data were available without gaps. As seasonally averaged climate and streamflow is more representative than just one single month (Fritts 1976), we also screened the AST chronology in simple correlation analysis (Pearson's correlation) with the seasonal climate and streamflow subsets to find the most appropriate

1972-2017
Note: MS is the mean sensitivity; SD is the standard deviation; AC1 is the autocorrelation order 1; SNR is the signal-to-noise ratio; VFE is the variance in first eigenvector; EPS is the expressed population signal.
seasonal predictand for the streamflow reconstruction. To examine the lagged effects of prior-year climate/streamflow on subsequent ring formation, the analysis window was extended from previous July to September of the current year. Considering the importance of understanding historical summer streamflow in Indus River and the highest correlation coefficient, standard tree-ring methods were used to reconstruct June-July streamflow for the upper Indus River basin (Fritts 1976). The linear regression model between the predictors (the AST chronology) and the predictand (streamflow) was then developed, retrodicting streamflow data during the pre-instrumental period. Because the instrumental streamflow record is not long enough to be divided into the verification and calibration periods, we used the 'leave-one-out' method (Michaelsen 1987) to assess the statistical fidelity of our reconstruction equation. The statistics include the Pearson's correlation coefficient, reduction of error, sign test and product mean test (Fritts 1976). To verify whether the reconstruction is subjected to over-fitting due to trend distortion, we also calibrated the first differences (year-to-year changes) of the tree-ring series with actual streamflow data. A temperature-sensitive streamflow reconstruction of Kara Darya River in the Pamir-Alai Mountains, Kyrgyzstan , provides a reference to validate our streamflow reconstruction.
In order to establish whether our streamflow reconstruction exhibited links with large-scale temperature and snow cover, we correlated our streamflow reconstruction with the June-July snow cover dataset (1966-2017) (https://climate.rutgers.edu/ snowcover) and HadCRUT4/HadSST4 (Cowtan and Way 2014) June-July temperature (1960−2017). Finally, in order to investigate teleconnections of regional streamflow to remote oceans, the two composite sea surface temperature (SST) anomaly maps of the 10 highest and the 10 lowest streamflow years during the period 1948-2017 were created using the gridded SST dataset (Smith and Reynolds 2003) to indicate the different spatial SST pattern.
In order to indicate the periodicities of our streamflow reconstruction, we performed a multitaper method (MTM) analysis (Mann and Lees 1996) with a 5 × 3 π taper and a red noise background. MTM is an good method for investigating periodicities of the time sequence because it requires very few a priori assumptions concerning the structure of the time sequence, and provides a robust average values for separating the signal and noise components of the time sequence. To reveal possible influence of solar activity on regional streamflow, we compared with the streamflow reconstruction, using solar activity reconstruction (Muscheler et al 2007), sunspot numbers (www.sidc.be/silso/datafiles) and Northern Hemisphere summer temperature reconstruction (Wilson et al 2016, Guillet et al 2017. To determine the impact of volcanic-induced cooling on the high Asian seasonal snow and glacierderived streamflow, we also applied a superposed epoch analysis (SEA, Haurwitz and Brier 1981). A total of 48 primarily volcanic eruption events with high volcanic eruption indexes (VEI ⩾ 5) prior to the 1990s was downloaded from the Smithsonian Institution (http://volcano.si.edu/search_eruption.cfm). In SEA each year in a list of primarily volcanic eruption events is taken as the zero window year. Streamflow values for the volcanic event years and for windows of years, in this case 6 years before and 4 years after the volcanic event years, are expressed as departures from the average values for the 11 years in each case. The departures for all the 11 years windows are averaged and superposed. The Monte Carlo simulation technique was used to evaluate the statistical significance of autocorrelation of the streamflow reconstruction with the random sampling method (Adams et al 2003). SEA was conducted using the EVENT software (version 6.02P) (www.ltrr.arizona.edu/software.html).

Results
Comparison between the AST chronology and monthly streamflow indicate the existence of significant positive correlations during early summer (from June to July) ( figure 3(A)). Further analysis showed that the tree-ring width chronology is more significantly correlated with temperature, and conversely, that there is no significant correlation with precipitation ( figure 3(B)). A significant correlation (r = 0.47, P < 0.01) was found between the AST chronology and the mean temperature from the previous October to the current July; the highest correlation (r = 0.64, P < 0.01) occurred between the tree-ring width chronology and the June-July period, implying that the temperature-sensitive tree-ring width chronology could be used as an indicator of temperaturedominated ice-snow meltwater signals (Starheim et al 2013.
The AST chronology explains 40.9% of the instrumental June-July streamflow variance during the period 1972-2008. Thus, the AST chronology was used in the regression model to reconstruct the June-July streamflow of the upper Indus River. The final connection between this streamflow as a transfer function of tree-ring widths designed by regression analysis is shown below: where Y is the June-July streamflow of the upper Indus River (m 3 s −1) and X is the AST chronology (non-dimensional values).  For the instrumental period  of the final streamflow reconstruction, the adjusted r 2 value was 0.396; the correlation between the initial differences of the tree-ring series and instrumental streamflow was 0.62 (P < 0.01) (figures 4(A) and (B)). The standard error of the estimate was 439.8; the F value was 25.65. The reduction of error (0.35) was strongly positive, using leave-one-out method validation, indicating the reconstructed equation was stable. For additional verification, the product means test statistics (5.49) and sign test (9-/30 + ) were both found to be significant at a 99% confidence level. These tests indicated the validity of the regression model and can be used to show the June-July streamflow variations of the upper Indus River during the period 1350-2017 CE.
The June-July streamflow reconstruction and its 21 year low-pass filtered values for the upper Indus River in the northern Pakistan are shown in figure 4(C). The streamflow reconstruction included a considerable number of low-frequency signals over the past 668 years. The long-term average of the streamflow reconstruction was 2716.0 m 3 s −1 , with a standard deviation of 454.4 m 3 s −1 . The streamflow reconstruction indicated relatively high streamflow from the mid-fourteenth to mid-sixteenth centuries, followed by two centuries, centred around 1650, of relatively low streamflow and pronounced high streamflow from 1750 to 1800. Streamflow over the past 200 years has generally shown a slow upward trend, with some short-term fluctuation in periods such as the 1810s through the 1850s, the 1870 through the 1890s, and the 1960s through the 1980s. The reconstructed streamflow showed an accelerating upward trend during the recent warming period.
The list of the highest and lowest streamflows reconstructed for the upper Indus River watershed since 1350 CE showed that seven of the ten lowest streamflows occurred during the seventeenth and nineteenth centuries, with the two lowest values in 1647 and 1651. Three of the ten highest streamflows occurred during the last 10 years, particularly in 2016-17. The composite map of the 10 highest streamflow years were characterised by a pattern of Tropical Eastern Pacific and mid-high latitudes SST above the average (1981-2010, 2000), resembling the pattern of El Niño years. During the 10 lowest streamflow years, the opposite pattern occurred (figures 5(A) and (B)). As shown in figure 5(C), some significant negative areas of correlation with the Rutgers snow cover dataset (https://climate.rutgers.edu/ snowcover) were found in the Karakoram region. Conversely, the streamflow reconstruction was significantly positively correlated with June-July temperatures in high Asia ( figure 5(D)).
The MTM of spectral analysis revealed 334 year (99%), 51.2 year (99%), 5.1 year (95%), 3.3 year (99%), and 2.2 year (99%) cycles in the reconstructed streamflow data for the upper Indus River basin. Correlations between this study and August-September Kara Darya River streamflow reconstruction , computed over the 1411-2016 common period, equalled 0.21 (P < 0.01), increasing to 0.32 following 40 year smoothing. Analysis of correlations between our streamflow reconstruction and solar activity revealed no systematic connection; this is likely related to the complex forcing data and regional hydroclimatic variation. Detailed comparison, however, revealed some primarily low streamflow periods (relatively low temperatures) following prominent solar minima (Maunder, Dalton and Damon) with an observed sunspot number sequence (Stuiver and Figure 6. Comparison among streamflow reconstruction of the upper Indus River (this study), solar activity (Muscheler et al 2007), sunspot numbers (www.sidc.be/silso/datafiles) and Northern Hemisphere summer temperature reconstruction (Wilson et al 2016, Guillet et al 2017. All series were smoothed with a 21 year low-pass filter to emphasize long-term fluctuations. Red boxes represent low streamflow conditions during during solar minimum periods, and blue box represent out-of-phase relationship between streamflow and sunspot number during the recent warm period. Braziunas 1989; figure 6). Mean streamflow during the three solar minimum periods was 6.4% lower than the long-term average of the streamflow reconstruction during the period 1350-2017. Figure 7 shows the SEA results based on a list of 48 volcano years. A statistically significant (P < 0.01) reduction in June-July streamflow began in the same year as the eruption and lasted for 2-4 more years. We estimated mean peak streamflow decline due to these large volcanic eruptions at 134-141 m 3 s −1 .

Tree growth/streamflow relationships
Most previous dendroclimatic reconstructions of river streamflows have been based on precipitation/ moisture-sensitive tree-ring-width series (Yuan et al 2007, Akkemik et al 2008, Gou et al 2010, Liu et al 2010, Urrutia et al 2011, Yang et al 2012, Singh and Yadav 2013, Shah et al 2014, Woodhouse and Pederson 2018, Chen et al 2019a, 2019b. Due to multiple streamflow contributions caused by the complex mountain terrain of Karakorum, the development of streamflow reconstructions for the upper Indus River basin requires a dense multi-species tree-ring network which is related to both precipitation and temperature, and the highest positive correlation between tree-ring widths and monthly streamflow were found in the current growing season (May-September) (Cook et al 2013, Rao et al 2018. The dissimilar responses of tree rings to streamflow indicate that multi-species tree-ring networks from different environments can be used to capture multiple-season streamflow signals (Cook et al 2013), whereas site chronology captures mostly the link between hydrological information and a single climatic element (the present study). Due to the scarcity of summer precipitation, the streamflow in this period originates mainly from winter precipitation and glaciers ( figure 2(A)). These streamflow reconstructions provided us with accurate hydrological information for the upper Indus River basin (Cook et al 2013, Rao et al 2018; however, this information cannot, due to the positive linkages between precipitation and tree-ring-width series, reflect temperature-dominated ice-snow meltwater signals (Rao et al 2018). Hydrologic information concerning temperature-related ice-snow meltwater is of great significance for revealing the mechanism governing high Asian Alpine glacier water resources in the context of global warming (Smith and Bookhagen 2018, Armstrong et al 2019, Farinotti et al 2020, especially for the Indian subcontinent, with its severe water shortage.
In the past decade, some warm-season streamflow reconstructions have been developed using temperature-sensitive tree-ring sequences (Hart et al 2010, Starheim et al 2013. We extended previously indicated complex linkages between tree-ring widths and climate/streamflow; additionally, although the AST chronology has no significant correlation with precipitation, we confirmed its relationship with ice-snow meltwater based on high sensitivity to temperature, including SST (figures 5(A) and (B)). From these relationships, we developed our reconstruction model of the upper Indus River June-July streamflow. Our models demonstrate that temperature-sensitive treering sequences can be used to reconstruct warmseason streamflow in high Asia. Although responses of tree rings to streamflow and climate are inconsistent, a close correlation (r = 0.47, P < 0.01) between this study and that of Rao et al (2018) was found for the period 1394-2005 CE. The significant negative correlation with the Rutgers snow cover dataset (Kunkel et al 2016) also indicates that the streamflow reconstruction may reflect temperature-dominated ice-snow meltwater, and an increase in June-July streamflow was often accompanied by a reduction in snow cover. However, due to the complex geographical environment of high Asia and the influence of the Asian summer monsoon, linkage between temperature-sensitive tree-ring series and streamflow appears to exist in the western part of high Asia, where summer rainfall is scarce.  figure 6, close synchronism exists between periods of significantly low streamflow and solar minima (Maunder, Dalton, and Damon). Our streamflow reconstruction exhibits a downward trend from the late Mediaeval Warm Period to the Little Ice Age (Lamb 1965). In particular, streamflow decreased by about 16.1% under the cold conditions during the Maunder Minimum , implying that the output of high Asian Alpine glaciers in terms of water output was governed by solar activity during past centuries, notwithstanding the dramatic changes in this relationship in the recent warm period. In addition to providing tree-ring evidence for decreases in meltwater due to low temperature during the three solar minima, we also found that the solar maximum was synchronised with the high streamflow stage, with a lag of 20-40 years. This lag effect may be linked with unexplained variance in tree rings and other forcing factors, such as volcanic eruptions. The out-of-phase relationship since the 1980s likely refers to anthropogenic warming (Cook et al 2016, Santer et  our study further confirmed that volcanic-induced cooling also led to a corresponding reduction in meltwater in subsequent months. The combination of the two effects may lead to a reduction in the annual streamflow output of high Asia. For the upper Indus River basin, the most pronounced volcanic radiative forcing arose from a series of large volcanic eruptions between the 1800s and 1810s (Sigurdsson andCarey 1989, Oppenheimer 2003), including the 1815 eruption of Tambora (Chenoweth 2001, Raible et al 2016. This low streamflow period likely resulted in a volcanic-induced cooling effect, along with lower temperatures due to a low level of solar activity in the Dalton Minimum (Russell et al 2010). Large volcanic eruptions also correspond to the delayed emergence of high streamflow during the sunspot maximum, including Huaynaputina (1600, VEI = 6; De Silva and Zielinski 1998), Laki (1783, VEI = 5; Thordarson and Self 2003), and Tambora (1815, VEI = 7; Raible et al 2016). Overall, low-frequency signals of streamflow reconstruction appear to mimic solar activity; moreover, the timing of large volcanic eruptions matches certain low streamflow events, possibly exacerbating Indian subcontinent water scarcity. Interestingly, large volcanic eruptions may delay the appearance of high streamflow during solar maxima, and the irregular nature of large-scale volcanic eruptions may have led to some errors in the streamflow project. Recent anthropogenic warming has not only further diminished the influence of natural forcing during the modern warm period (Park et al 2018, Marvel et al 2019, but also changed the high Asian water cycle.

Interaction between streamflow and human activities
Although summer streamflow in this region and in Central Asia are in close agreement, drought and low streamflow events occurred in Central Asia from the late fifteenth to the early sixteenth century (Opała-Owczarek and Niedźwiedź 2019, Zhang et al 2020). This dry condition forced Central Asian peoples to migrate outwards, while the neighbouring Indian subcontinent, including the Indus basin, was relatively humid and attracted conquerors from Central Asia (Yadava et al 2016). It is interesting to note the low streamflow phase during the Maunder Minimum (1645-1715). Famine and the economic hardships caused by bad climate during the Maunder Minimum forced the Mughal emperors to undertake the conquests to relieve the Empire's ruling crisis (Panhwar 2004, Uberoi 2012, Parwez and Khan 2017 The Mughal Empire, although it reached its maximum extent under Aurangzeb, was slowly weakened by prolonged low streamflow and frequent warfare (Truschke 2017). Despite high streamflows during the eighteenth century due to relatively high temperatures (Yadav et al 2011), the Indian subcontinent was conquered completely only after the earlynineteenth-century low streamflow (Clingingsmith and Williamson 2008). We must admit that streamflow is not the decisive factor in the historical process; nevertheless, it has affected regional social and economic development to a certain extent. Based on our tree-ring record, streamflow, unusually, has increased by 18.3% over the past 20 years. Streamflow in the Indian subcontinent may have benefitted from recent anthropogenic warming (Lutz et al 2014, Armstrong et al 2019; however, this increase was based on the rapid melting of glaciers, and this area is marked by great uncertainty (Luo et al 2018, Biemans et al 2019. Correlations between this study and temperature under the CMIP6 SSP-585 scenario (CESM2, Eyring et al 2019), computed over the 1850-2017 common period are 0.36, and increase to 0.59 after 21 year smoothing. Since the streamflow is mainly controlled by temperature, according to the linear model, the averaged streamflow will risen by more than 35% during the period 2018-2100.
However, if anthropogenic warming continues and glaciers disappear, a significant reduction in streamflow (Kraaijenbrink et al 2017, Pritchard 2019) may occur; this may have a negative impact on the Indian subcontinent (Biemans et al 2019), which is densely populated and depends heavily on snow and ice melt for irrigation agriculture.

Conclusions
In this paper, a new tree-ring chronology of Pinus wallichiana was developed from the upper Indus River basin in Northern Pakistan. The tree-ring chronology is sensitive to October-July temperature variations, and has a strong association with streamflow changes. Based on this temperature-sensitive chronology, we have presented a well-calibrated and verified June-July streamflow reconstruction of the upper Indus River basin. This streamflow reconstruction placed the unusual and unprecedented most recent upward trend of snow-and glacier-melt-driven streamflow since the 1990s in a long-term context and enabled evaluation of potential impacts of water resources on historical societal changes on the Indian subcontinent. This streamflow reconstruction also enabled an assessment of the possible effect of solar activity and large volcanic eruptions on the variability of high Asian streamflow from seasonal snow and glacier ice.

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