Observed Evidence for Steep Rise in the Extreme Flow of Western Himalayan Rivers
Abstract
We present the first observational evidence of changing patterns of extreme streamflows at multiple locations across Western Himalayan (WH) rivers. We find that the frequency of extreme flow events during the period 1980–2003 has doubled with a statistically significant increasing trend in annual maximum streamflow. We postulate that this streamflow change is due to the increased precipitation extremes occurring during both the summer monsoon and the winter seasons. We further found a stepwise increase of “chi-gradient” in Bhagirathi and Sutlej Rivers, indicative of a landscape that facilitates rapid flow of water generating severe floods downstream. Our results highlight the severity of hydroclimatic changes underway in the WH region and the critical need for a hydro-infrastructure for flood forecasting to benefit more than 17 million inhabitants and for ecosystem services to mitigate increasing flood hazards.
Key Points
- First evidence from observations shows that occurrences of extreme streamflow in Western Himalayan rivers have doubled during 1980–2003
- Increasing trends of rainfall extremes and the geomorphology of the rivers have driven the dramatic increase in extreme streamflow
- There is a critical need for a dedicated, high-resolution, real-time flood forecasting system for the Western Himalayan region
Plain Language Summary
Flooding in the Western Himalayan rivers has caused colossal loss of life and property in the recent decades. We present a first analysis of changing patterns of observed peak streamflow in Western Himalayan rivers at multiple locations, using high-quality in situ data. With a high degree of confidence, we conclude that the occurrence of extreme flow events during the period of 1980–2003 has doubled with an increasing trend in annual maximum streamflow. Our analysis shows that this streamflow change is due to the increased precipitation extremes during both the summer monsoon and the winter seasons. We also found that the profiles of two Western Himalayan rivers (Bhagirathi and Sutlej) exhibit steep slopes in a stepwise manner, indicative of a landscape that rapidly responds to extreme precipitation by contributing to severe floods. Our results contextualize the severity of hydroclimatic changes underway in the WH region. Results also highlight the critical need for a hydro-infrastructure and an operational flood forecasting system to benefit more than 17 million inhabitants and for ecosystem services to mitigate increasing flood hazards.
1 Introduction
Western Himalayan (WH) rivers have been affected by changing climate (Barnett et al., 2005; Cyranoski, 2005; Immerzeel et al., 2010; Lutz et al., 2014), and hydroclimate model projections suggest a continued increase in river flows through the 21st century (Immerzeel et al., 2011, 2013). Colossal loss of life, property, and infrastructure has been witnessed in recent decades owing to flooding (Hewitt, 1968; NIDM, 2014; Shroder et al., 1998). The severity of impacts associated with extreme events is often a result of multiple geophysical systems working in tandem, from the atmospheric processes responsible for extreme precipitation (Vellore et al., 2016) to the geomorphic factors (Whipple & Tucker, 2002) that contribute toward consequent streamflow.
The meteorology of the WH region is governed by the southwest monsoon (Simpson, 1921; Webster et al., 1998) during summer (June to September, JJAS). During Indian winter (December to February, DJF), westerly disturbances (WDs; Dimri et al., 2015; Pisharoty & Desai, 1956) embedded in the subtropical westerly jet dominate the climatology.
The WH region is a mountainous area susceptible to widespread landslides because the landscape quickly responds to extreme streamflow events (Devrani et al., 2015; Seeber & Gornitz, 1983). The rivers in this region originate from the highest elevated sources in the world and rapidly descend to valleys and flatlands. These perennial rivers are fed by the melt water of Himalayan snow and glaciers in the warm season and by rainfall during JJAS. While enhanced melt water in a warmer future climate from Asian water towers (Immerzeel et al., 2013) is a cause of concern, an urgent and immediate threat is posed by raging runoff resulting from enhanced extreme rainfall events during both the summer monsoon (Goswami et al., 2006; Krishnamurthy et al., 2009) and the winter (Guhathakurta & Rajeevan, 2008; Revadekar et al., 2011; Shekhar et al., 2017). Consequently, heavy rain and floods put towns and settlements along the floodplains of WH rivers at imminent risk.
Increase in precipitation intensity is one of the most notable among warming-induced hydrological changes. The increase in the intensity and the frequency of heavy precipitation events is noted for the Indian region (Goswami et al., 2006). For reasons outlined above, the WH region is increasingly vulnerable to flash flooding-related natural disasters. Examples of recent major disasters associated with the Himalayan rivers are floods in Pakistan (2010) that affected over 20 million people and resulted in 1,800 fatalities (Warraich, 2011), floods in Uttarakhand (2013) that claimed 6,000 lives (Tayal, 2015), and floods in Jammu and Kashmir (2014).
It is, therefore, critical to evolve an improved understanding of the cascade of geophysical systems that result in extreme flood situations. To the best of our knowledge, because of the difficulty in obtaining high-quality data sets, the analysis of observed streamflow data of WH Rivers to understand the occurrences and trends of extremes has not been performed. Therefore, the objective of this study is to present the first results using quality-assured, streamflow measurements. We extend this analysis by studying the associated meteorological and geomorphological factors using remotely sensed and reanalysis data.
2 Data and Methods
We use daily station data for the period 1980–2003 at four river gauging stations of Upper Ganges—Mandakini, Alaknanda, Uttarkashi, and Deoprayag—to study the trend of summer (JJAS) maximum streamflow and a similar data set for two stations of Sutlej—Rampur and Suni—to study the trend of winter (DJF) maximum streamflow. As per the Indian government's data dissemination policy for international rivers, the data for the four Ganges stations has been scaled by the same integer factor for all four stations. The scaling allows for the variation to be shown without disclosing the magnitudes. The trend analysis of rainfall extremes was performed using 0.25° resolution gridded rainfall data from the India Meteorological Department (available from 1901). The pattern and trend of atmospheric variables was investigated using reanalysis data sets from the European Centre for Medium-Range Weather Forecasts (ECMWF) atmospheric reanalysis for the 20th century (ERA-20C). The ERA-20C relies on its advanced data assimilation capability to constrain model estimates close to observations. The historical meteorological data record for the WH region is sparse and thus influences the accuracy of ERA (and other reanalysis) products.
The WD activity in the WH region was determined using an objective procedure based on Madhura et al. (2015). After chronologically arranging the 500 hPa daily geopotential data for DJF season in the period 1900–2010 (111 years), we subtract the zonal (or latitudinal) mean to prepare the data for identification of features associated with baroclinic disturbances. A Lanczos high-pass filter (Duchon, 1979) was applied to retain variations with a time scale smaller than 10 days and remove low-frequency synoptic-scale variations like Madden Julian Oscillation and El Niño Southern Oscillation. Finally, empirical orthogonal function (EOF)/principal component (PC) analysis was applied to the daily high-frequency geopotential anomalies to assess the dominant mode of variability during DJF.
(1)3 Results and Discussions
3.1 Doubling of Extreme Flow Events
We first investigate the frequency and temporal trend of observed extreme streamflow events for the period 1980–2003 for four prominent river gauging stations in the Upper Ganges basin and two in the Sutlej basin (Figures 1 and 2b). Figure 1 shows the scaled summer maxima of streamflow records over the Ganges basin (Figures 1b–1e) along with the winter maxima in the Sutlej basin (part of the Indus basin; Figures 1f and 1g). The short period (1980–2003) selected for our analysis was governed by the availability of high-quality data and is considered representative of the spatiotemporal changes in the region. Other than analyzing trends, we also divide the period of data availability into two 12-year periods (1980–1991 and 1992–2003) to compare the differences between the two periods. This reduced sample size, governed by the short length of available streamflow data, is a limitation of the current study.
We additionally analyze the association of extreme precipitation to flooding for Ganges and Indus basins with summer and winter precipitation, respectively. We consider the average of seasonal maximum streamflow (JJAS for Ganges basin and DJF for Sutlej basin) at a gauging station for 1980–2003 as the threshold for defining extremes at that station. The division of stations between the two seasons was done with the recognition that the Himalayas play a unique role in the generation of precipitation by “blocking” monsoon circulations in summer (JJAS) and extratropic winds during winter (DJF). The former results in rainfall that often lead to floods in the Ganges basin during the summer monsoon, while the latter results in extreme precipitation due to WDs during the winter season, mostly in the Indus basin (Dimri, 2006).
Results indicate a remarkable increase (doubling) in the number of occurrences of extreme flows during 1992–2003 as compared to the period 1980–1991 (Figure 2b). The threshold of mean seasonal maxima during 1980–2003 corresponds to greater than 99th percentile of daily discharge for each Ganges station. For the Sutlej stations, this threshold corresponds to the 93rd percentile for Suni station and 88th percentile for Rampur station. Figure S1 shows how the occurrences of extreme flow change with respect to the threshold. Increase in the extreme threshold value for these two stations leads to an increase in the ratio of occurrences in 1992–2003 to those in 1980–1991. This is a robust finding as the results are consistent across all stations and across multiple percentile thresholds for defining streamflow extremes from daily discharge measurements (Figure S1).
Next, we compute the trend of seasonal maxima across the stations (Figures 1b–1g). Four out of six stations have increasing trends of extreme streamflow at the statistical significance level of 5%. The two stations Rampur and Uttarkashi show statistically significant increasing trend at the 12% level. Such consistent changes are indicative of an alarming increase in extreme streamflow occurrences for the WH rivers. The spatial variability in the slope and the statistical significance of results could partially be attributed to anthropogenic influence of the 17 million population within the basins.
3.2 Intensification of Extreme Rainfall Events and Their Impact on Stream Discharge
Increases in river flow of Himalayan basins are typically attributed to the increase in glacier and snow melt (Immerzeel et al., 2013). We additionally examine the association between the increase in extreme flow and precipitation. Figure 3 shows the changes in precipitation patterns for Ganges (JJAS) and Sutlej basins (DJF). The precipitation and runoff regimes of the upper Ganges basin are monsoon dominated (Immerzeel et al., 2013). Figure 3a shows the seasonal maxima (JJAS) time series of zonally averaged rainfall (average over all longitudes within the WH region for each latitude) in the period 1980–2003. The time series shows a positive slope of the linear fit line, but we do not find any statistically significant trend. Over the longer period 1901–2013, the same metric shows a statistically significant positive trend (Figure S2a). We further plot the composite spatial map of precipitation anomaly (mm/day) corresponding to the day and the previous day for which the seasonal maximum streamflow was observed for the Ganges basin stations. These plots are presented separately for the events occurring during the two periods 1980–1991 and 1992–2003 (Figures 3b–3e). We find increased values of composites of extremes with a greater spatial extent during the recent period as compared to the past. This indicates that the increase in flooding during summer monsoon over the Ganges basin is associated with the increase in extreme monsoon precipitation.
Winter (DJF) season precipitation constitutes 40–45% of the annual precipitation in the Indus basin (Madhura et al., 2015). Observations indicate a significant increasing trend in the winter season maxima of zonally averaged rainfall over the WH region since 1901 (Figure S2b). However, we do not find any statistically significant trend in the shorter period of 1980–2003 (Figure 3f). We follow the same methodology as used in Figures 3b–3e for assessing precipitation changes associated with extreme winter discharge observed at the Sutlej stations. Results lead to similar conclusions as obtained for the summer season, that is, increased flooding events occurs a day after the increased extreme precipitation (Figures 3g–3j). This again highlights the strong association between the flooding and precipitation extremes over the WH region.
3.3 Meteorological Aspects of Extreme Flow Events
We analyzed the occurrences of extreme rainfall events considering a similar threshold (average of annual maxima) for the region. We find a similar doubling of extreme rainfall, including over the Ganges basin (Figure 2a). We reviewed the reanalysis variables to infer the change in large-scale dynamics surrounding the summer and winter extremes (Figures S3 through S9). Figure S3 shows the anomalies of geopotential height and relative vorticity at 500 hPa for the same day and the day before of summertime streamflow extremes, averaged for the periods 1980–1991 and 1992–2003. We find that the WH region is under a relatively lower pressure regime in the latter period (comparing Figures S3a and S3b with Figures S3c and S3d). Increased positive vorticity anomalies in the latter period around the western flank of the WH domain are also noted (comparing Figures S3e and S3f with Figures S3g and S3h). This indicates the presence of a stronger trough with high positive vorticity which would ultimately contribute to an environment that is more conducive to regional instability. Specifically, for the upper Ganges basin, the reduction in geopotential heights is less pronounced, and there is a slight indication of increased negative vorticity anomalies. The 500 hPa level was selected based on the analysis undertaken in Madhura et al. (2015) that captured the increased baroclinic instability over the WH.
A slightly increasing tendency for divergence at upper levels and reduction in divergence (increase in convergence) at the lower level is observed during summer monsoon season over the WH region (Figures S4b, S4c, S5b, and S5c). Total precipitable water does not show any strong trend (Figures S4a and S5a). We find a weak increasing trend of seasonal maxima of convective available potential energy (CAPE) with a Sen's slope magnitude of 1.3518 and decreasing trend of seasonal maxima of convective inhibition (CIN) with a magnitude of −0.648 in the recent period 1980–2003 (Figure S6). The trends appear to be clearer and statistically significant for divergence as compared to precipitable water, CAPE, and CIN, which suggests the role of changing large-scale dynamics rather than local, thermodynamic processes. The role of intensification of the dynamic circulation and the processes that control rainfall over the study area needs to be explored in future work. These indicators consistently highlight a change in the behavior of the forcings that synergistically favor extreme precipitation occurrence over the region. Coupled with a projected increase in glacier melt in a future warmer climate, this positive tendency noted in the meteorological setting to produce heavy rains can have dire consequences for the floodplains of the WH rivers vulnerable to severe risk of flash floods.
The nature of background atmospheric conditions for winter extremes was also assessed. The dominant mechanism that drives the precipitation over the WH region during the winter season occurs in association with synoptic weather systems that propagate eastward from the Mediterranean region toward South Asia. These extratropical-like disturbances, known as WDs, are embedded in subtropical westerlies. They often produce significant rainfall over WH by extending down to lower atmospheric levels over the northern Indian latitudes (Dimri et al., 2015; Pisharoty & Desai, 1956). Figure S7 shows the geopotential height anomaly at 500 hPa and the anomaly of relative vorticity at 500 hPa corresponding to maximum winter discharge recorded at the stream-gauge stations during both periods. A change is noted in the midlevel atmospheric patterns preceding the extreme flood discharge days, with the WH region under a relatively lower pressure regime and higher positive vorticity in recent years. The upper-level (200 hPa) atmospheric patterns of antecedent geopotential height anomaly (Figure S8) also show similar changes in the recent period.
To analyze these extremes further, we use a quantitative index to assess the trend in WDs. We performed this analysis using data for both longer (1901–2010) and shorter periods (1980–2003). The spatial structure of the first EOF (EOF1) of daily, high-pass (<10-day) filtered, geopotential anomalies at 500 hPa (Figure S9a) shows a negative anomaly centered over WH (near 78°E, 32°N), alongside a positive anomaly located to the west (60°E, 33°N). This EOF1 spatial pattern explains 62.5% and 64.5% of the total variance for 1901–2010 and 1980–2003, respectively. This dipole spatial pattern in the geopotential anomaly field is typically indicative of WDs in the WH region (Dimri, 2006; Hunt et al., 2018). Accordingly, the associated PC time series (PC1) can be used to infer the activity of WDs. Since the analysis was applied on all available daily geopotential data for the DJF season (1900–2010), the longer PC1 time series consists of 9,990 temporal points (90 days per DJF season × 111 years; leap-year days removed). We computed the standard deviation of this time series for the 90 days of each DJF season to assess the temporal trend in its variability. The annual standard deviation of PC1 exhibits a positive trend for the 1900–2010 period, indicating enhanced activity of WDs over the region (Figure S9b). For the shorter period, the linear fit shows a negative slope (not shown). However, the slope values are not statistically significant at the 90% level. Notably, the increase in WD activity is consistent with other studies that have used different data sets (e.g., Madhura et al., 2015). The consistency and reproducibility of the EOF/PC patterns in different reanalysis data sets provides robust support for interpreting the increased activity of WDs. This increase along with the meteorological setting that has evolved in the WH highlights a growing risk for increased extreme precipitation that has already been witnessed in recent years.
3.4 Susceptibility of Regional Geomorphic Conditions to Flash Floods
The WH rivers traverse through several steep, active faults (e.g., the Main Central Thrust; Seeber & Gornitz, 1983) to reach the plains from their elevated sources. Such increases in channel gradients via tectonic forcing increase the unit stream power resulting in higher incision rates. Extreme geomorphic events are known to accentuate the long profile of river channels as tectonic forcing alters the channel gradient, further contributing to increase in incision rates, causing a feedback loop. Steeper channel gradients enhance soil erosion and sediment transport capacity of a river in the upper parts, resulting in a higher differential of sediment transport capacity as the rivers approach the plains. Consequently, the sedimentation along flood plains is also enhanced. This self-sustained potential for channel incision as well as sedimentation imparts increasing destructive potential to the WH rivers.
Past studies have employed channel steepness to approximate long-term erosion rates (Kirby & Whipple, 2012). Recently, channel steepness analysis has also been shown to match the distribution of incision (and sedimentation) along the channel during an extreme event. Devrani et al. (2015) showed that the spatial pattern of channel steepness before the flooding event in Uttarakhand region (June 2013) was consistent with the topographic modification along the Mandakini River due to the event. The channel steepness technique can be effectively employed to quantify the flash flood potential of WH river profiles and help identify the potential for the impacts of extreme flood events in the future. We analyze the long river profiles of Sutlej (in Indus basin) and Bhagirathi (in Upper Ganges basin) in the WH region using the χ analysis approach (Royden et al., 2000), which expresses tectonic-climate forcing through topography. Here, χ refers to a transformed coordinate, which has the dimensions of length (Perron & Royden, 2013). A χ-elevation space for a channel that has undergone stepwise increases in uplift rates will show two or more segments appearing convexly in the χ plot. Channel elevations plotted against χ or Mχ (gradient of transformed profile in χ-elevation space) allow comparisons between channel segments with different drainage areas.
We follow the methodology presented in Mudd et al. (2014) to extract χ and Mχ (Equation 1) after determining the best-fit value of m/n (Figure S10). We focus on the primary channels of Sutlej and Bhagirathi Rivers. Figures 4a and 4c show the transformed profiles, indicating a stepwise increase in uplift rates. The headwaters have relatively low Mχ values of less than 3 (Figures 4b and 4d), but the downstream steepening in this upper stretch of the rivers is visible as they descend, and multiple stretches with high Mχ values (increased risk of the incision) followed by significantly horizontal stretches (increased risk of sedimentation) are seen. This demonstrates a distinct geomorphic signal of the flash flood potential of the WH rivers.
4 Conclusion
We find observational evidence of extreme flood occurrences more than doubling in the WH rivers, which is significant for the locale in terms of human and economic losses. Earlier studies highlight an increase in the extreme precipitation over the same region (Madhura et al., 2015); however, this is the first study to cement this finding with in situ observed streamflow data. The analysis concludes that increasing trends in regional extreme rainfall during both summer monsoon and winter drive the notable increase in extreme streamflow. The WH region has also seen increased occurrence of cloud burst events in recent times (Dimri et al., 2017). The weaker observed trend in winter precipitation extremes is heavily influenced by a period of anomalously intense extreme events during 1950–1960 (Figure 3f), but the analysis of changing meteorological conditions (Figures S7–S9) further backs the conclusions of the study. This increased streamflow is due not only to the increased precipitation or snowmelt but also to regional geomorphic conditions that causes extreme precipitation to result in severe floods. The flooding during 2013 over the Uttarakhand region (upper Ganges basin) is one such example. χ profile of the headwaters of Sutlej River indicates higher potential to generate severe floods. The lack of severe flooding-related disasters in this region in recent years can be explained by the observed increase in precipitation extremes upstream of all upper Ganges stations but not over the Sutlej stations (Figure 2a), but the increasing tendency of westerly activity in winter season coupled with changing meteorological conditions presents a severe threat. Mapping of locations along WH river profiles for high risk of incision and/or aggradation to predict the geomorphic signal of extreme events is a potential area of future research.
Study findings, when considered in the context of ground reality where currently there is no operational, real-time, regional flood forecasting system, emphasize the urgent need of the same for the 17 million population of this region and downstream locations. Though there exists an operational rainfall forecast system in India, a regionally dedicated, high-resolution system is needed, especially considering the geomorphology of the region. Further, floodplain zoning of WH river plains to regulate construction activities, considering the intensification of extreme floods, should be implemented.
Conflict of Interest
The authors declare no competing interests.
Acknowledgments
The authors sincerely acknowledge the Ministry of Environment, Forest, and Climate Change for financial support. D. N. acknowledges US NSF, USDA NIFA Hatch Grant, and the Indian Monsoon Mission project for enabling participation in this research. We are grateful to the two anonymous reviewers who positively contributed in improving this manuscript.
Open Research
Data Availability Statement
The reanalysis data used in the present analysis are freely available online from https://apps.ecmwf.int/datasets/data/interim-full-daily/levtype=sfc/. The streamflow data were obtained from the Central Water Commission (CWC, Ministry of Jal Shakti, Government of India) for research purposes only. As per the data sharing policy of the Government of India for the Ganga basin, the flow data cannot be shared without their written consent. The rainfall data may be obtained from the India Metrological Department (available online at http://www.imdpune.gov.in/Clim_Pred_LRF_New/Grided_Data_Download.html). The Cartosat-I stereo data are also freely available from National Remote Sensing Centre, India, at https://bhuvan-app3.nrsc.gov.in/data/download/index.php?c=s&s=C1&p=cdv2.
