Hydromorphological shifts in the Yamuna River: Post-barrage impacts and water management implications

Barrages impact sediment transport, disrupt fish migration, and alter habitat conditions (Agarwal & Narain 1997; Misra 2010; Sharma & Kansal 2011; Joshi et al. 2016; Sharma et al. 2017; Yadav et al. 2023). For instance, the Farakka barrage significantly changes the Ganges River's flow (Gain & Giupponi 2014). Recognizing these impacts is crucial for formulating strategies to maintain a sustainable water supply amidst growing demand and climate change (Jain 2012; Shahid et al. 2018).

The present work focuses on studying the impact of barrages in India's upper segment of the Yamuna River, crucial for water supply in North India. The section of the river is interrupted by a major barrage, the Tajewala barrage, replaced by the Hathnikund barrage in 1999. An analysis of the planform characteristics of the Yamuna River stretch would provide us with a great deal of information on the morphological alterations in the river basin.

The oldest commonly used method for assessing hydrological alterations was developed by Richter et al. (1996), which demonstrated the use of the ‘range of variability approach’ (RVA). The method computed the 25th and 75th percentiles of 33 hydrological parameters (indicators of hydrological alterations, IHAs) used in the RVA based on the pre-dam and post-dam discharge records. Although this technique is widely used, it has some likely impediments. Specifically, in the RVA technique, the values beyond the interquartile range are neglected, thus underestimating the overall alteration value (Olden & Poff 2003; Mathews & Richter 2007). The River Flow Health Index (RFHI) method, developed by Mohanty and Tare in 2022, is a powerful tool to quantify the hydrological allegations on a 0–1 scale. The main advantage of this method is that it constructs the confidence intervals with which the results are reported, thereby addressing the uncertainties associated with large hydrologic data.

Due to the limited research on quantifying changes in different components of flow regime and planform caused by barrages (Shih et al. 2022), as well as the lack of study on the effect of construction works on river planform (Lawrence 2001; Bellmore et al. 2017), our objective is to investigate (i) the significant changes in the flow regime brought by decommissioning and constructing barrages using the RFHI method and (ii) the resultant alterations in channel planform, using remote sensing imagery. This methodology is significant as it provides a structured framework for evaluating the impacts of barrages on river hydrology and morphology, aiding in decision-making processes vital for ecological sustainability and effective water resource management.

Yamuna, a major tributary of the river Ganga, originates from the Yamunotri glacier at 6,320 m above mean sea level. It has a total length of around 1,376 km and a catchment area of 366,000 km². In the basin, annual rainfall averages around 1,200 mm, most falling during the monsoon season (June to September). The stream travels through the Shivalik mountain range in the Indian states of Himachal Pradesh and Uttarakhand. It runs past the well-known Sikh shrine Poanta Sahib after leaving Dak Pathar. After passing through Poanta Sahib, the river enters the Yamuna Nagar district of Haryana and reaches Hathnikund, where it is once again diverted for irrigation into the Western Yamuna Canal and the Eastern Yamuna Canal. The Hathnikund barrage was constructed in June 1999 for irrigation purposes. It replaced the Tajewala barrage, built in 1873, and is no longer used. It is located 3 km upstream. The Hathnikund barrage on the Yamuna River primarily provides water to the Indian states of Haryana and Delhi. The barrage diverts water into the western and eastern Yamuna canals, significantly impacting the water supply to these densely populated and agriculturally intensive areas. The socio-economic imperatives for constructing the Hathnikund barrage included improved water distribution efficiency, better flood management, and enhanced capacity for water storage.

The geological formation of the study area in the Yamuna basin consists of alluvium of the Quaternary period of the Cenozoic era (Singh 2012). The soils in the study area have been classified as brownhill soils (Devi 1992), formed from the weathering of granite, gneiss, and schist.

Numerous hydrological stations are located along the Yamuna River and its main tributaries. These stations are operated by the Central Water Commission (CWC), a department of the Ministry of Jal Shakti, Government of India. Discharge characteristics such as river level and velocity are monitored once daily (Jain et al. 2007) or three times daily (CWC 2017). The discharge is then determined by multiplying the cross-sectional area derived from the river stage by the velocity. Data on daily water discharge are collected from the CWC maintained at Kalanaur gauging station with coordinates 30°04′10″ E, 77°21′52″ N for 40 years (1976–2015).

Satellite images encompassing several Landsat satellites are obtained from the United States Geological Survey (https://earthexplorer.usgs.gov/) for the year 1993, prior to the start of construction of the Hathnikund barrage, 1999 during the construction of the barrage, and different periods spanning 19 years after the barrage is operational. Details of the remote sensing images obtained are tabulated in Table 1.

Mohanty & Tare (2022) proposed a powerful tool to estimate the hydrological alterations induced by anthropogenic interventions. The approach for calculating the RFHI comprises four steps:

• (a) Segregation of flow data into different periods;

• (b) Determining significant hydrological parameters;

• (c) Evaluation of flow alterations; and

• (d) Quantifying the health of the river flows during altered and intermediate periods.

The flowchart outlining the methodology for computing RFHI is presented in Figure 2.

Reference conditions (near-natural) refer to the flow state before any anthropogenic intervention on the studied river. Here, the river has already been altered; hence, we do not have any reference period. Altered conditions are considered to have occurred after the barrage across the river became operational. We have two altered periods, i.e., 1976–1993 during the operation of the Tajewala barrage and 2001–2015 during the operation of the Hathnikund barrage. The period of construction (1995–2000) of the Hathnikund barrage in the Yamuna River is considered the intermediate condition.

Mohanty & Tare (2022) developed a powerful methodology to evaluate the hydrological alterations that account for the uncertainties in hydrologic modeling. The RFHI method includes 171 hydrologically relevant parameters published by different authors (Supplementary Table S1). These parameters are divided into seven components of the flow regime (Richter et al. 1996; Magilligan & Nislow 2005):

• (i) Magnitude (n, number of parameters = 76): The amount of water flowing past a specific location in a time.

• (ii) Variability (n = 52): The traits that demonstrate how the flow changes throughout time.

• (iii) Duration (n = 12): The amount of time corresponding to a certain flow condition.

• (iv) Frequency (n = 12): The number of times a certain amount of flow happens over a given time interval.

• (v) Timing (n = 50): The prediction of a particular magnitude of flows.

• (vi) Rate of change (n = 6): It describes how quickly flow changes from one value to another.

• (vii) Other (n = 17): This category includes the flow regime's distribution, shape, and range.

The hydrological parameters are estimated using a mix of computer programs developed in the MATLAB programming language from the daily flows (Henriksen et al. 2006) for different periods in the river. The value of each parameter is calculated for each water year of the study period.

Principal component analysis (PCA) is employed to identify key subsets of hydrological parameters that represent natural flow patterns at specific locations, reducing data redundancy. However, traditional PCA methods often overlook the uncertainties inherent in hydrological data (Babamoradi et al. 2013). To mitigate this issue, a bootstrapping approach is integrated with PCA (Erfon & Tibshirani 1986). This method enhances the analysis by accounting for variability and uncertainty, using random sampling with replacement to generate multiple data samples.

The methodology comprises several steps:

• (a) PCA on Original Dataset: Initially, PCA is applied to the hydrological data to diminish its dimensionality. This step isolates principal components that capture critical patterns in the flow regime. However, this traditional PCA might not fully account for data uncertainties.

• (b) Bootstrapping for Eigenvalue/Eigenvector Analysis: Bootstrapping is paired with PCA to address this limitation. Multiple sample sets are created from the original data, and PCA is performed on each set. This repetition, perhaps hundreds of times, aids in understanding the data's variability and uncertainty, thus strengthening the analysis.

• (c) Variance Calculation with Confidence Intervals (CIs): This involves calculating the variance explained by the top eigenvalues, their median percentages, and 95% CIs. The CIs gauge the reliability of the eigenvalues, reflecting their potential variability across samples.

• (d) Differentiation of Eigenvalues Based on CIs: Eigenvalues are differentiated based on their CIs. Non-overlapping intervals indicate distinct eigenvalues, whereas overlapping intervals suggest similarities. This step is crucial for identifying consistent data patterns despite inherent variability.

• (e) Identification of Significant Hydrological Parameters: The analysis is refined by examining eigenvector loadings to pinpoint significant hydrological parameters in each component. This step emphasizes the flow regime's key features and validates their significance across different samples, adding a layer of uncertainty analysis.

• (f) Projection of Original Data onto Feature Vectors: The original data is mapped onto the identified feature vectors (principal components). This projection aids in understanding the temporal variation of these components while considering data uncertainty.

• (g) Statistical Comparison Across Periods: Utilizing a two-sample Kolmogorov–Smirnov test (Kroll et al. 2015), the principal components are compared across different periods. This comparison detects significant changes in the flow regime, offering insights into its stability and variability.

• (h) River Flow Health Index: The RFHI is formulated using the geometric mean of the D-statistics from the Kolmogorov–Smirnov test. This index, reflecting the cumulative probability distribution difference between principal component (PC) sets, indicates the flow regime's alteration degree. A D-value near 1 implies significant changes, while a value close to 0 indicates minimal changes. This index is a comprehensive measure of the river flow's overall health and variability, encapsulating the impact of uncertainties.

This methodology effectively utilizes bootstrapping with PCA to address uncertainties inherent in hydrological data. It ensures that the identified patterns and changes in the flow regime are statistically reliable and truly reflect the underlying variability in the data. The overall index is computed at Kalanaur station in Yamuna using the above approach.

For dam operators, this metric is valuable in decision-making processes. By understanding the degree of change in the flow regime (as indicated by the D-value), operators can assess the impact of their activities on river health. A higher D-value would signal a need for more careful management strategies to minimize hydrological alterations, whereas a lower value would suggest that current practices are not significantly impacting the flow regime.

Morphological alterations are evaluated using a geographic information system (GIS) using ArcMap 10.4. The geomorphic units are classified using the procedure outlined in Wheaton et al. (2015).

Planform parameters (Table 2) are calculated from the satellite-generated maps. The framework for image processing and geomorphic mapping is outlined in Figure 2. To find the sinuosity, the length of the central line (centerline of the stream) and the straight line connecting the endpoints of the reach are calculated. The ratio between central line length (Lcmax) and straight length (Lr) is recorded as sinuosity (Friend & Sinha 1993). All mid-channel and main channel lengths are generated for braid-channel ratio calculation. Then, the braid-channel ratio is calculated by dividing the lengths of all mid-channel lengths (Lctot) by the main channel length (Lcmax) for the reach (Friend & Sinha 1993). Perpendiculars are drawn to the central line of the stream to generate the average channel width. The average width of these perpendiculars in the reach is the average channel width. In multi-channel reach, the width of the bars is excluded to compute channel width. The area of all the bars is generated and recorded as bar area in aggregate. The channel belt area, excluding the bar area, is reported as the channel area. The number of bars per kilometer of river reach determines bar density. The study considered different periods: the first altered period, 1993, during construction (1999), and the second altered period, 2003, 2008, 2013, and 2013.

The results of the analysis for each group of hydrological parameters are as follows:

Group 1: Magnitude: The first eigenvalue in this group is highly significant, contributing to 41.07% of the total variance in the first PC within the magnitude category (Figure 3(a)). The top five variables that exert the most influence on this primary axis are high peak flow (MH 26), high flow volume (MH 22 and MH 23), low flow index (ML 15), and baseflow index (ML 19).

Group 2: Variability: This group's first eigenvalue is also substantial, accounting for 30.51% of the overall variance (Figure 3(b)). The five most influential factors contributing to this component are variability across monthly flow (MA 36, MA 38, and MA 29), variability across maximum monthly flows (MH 13), and standard deviation of the percentiles of the logs of the entire flow record divided by the mean of percentiles of the logs (MA 4).

Group 3: Duration: The first PC in this group explains 35.97% of the total variance (Figure 3(c)). It is primarily influenced by parameters related to high flow duration (DH 20, DH 15, DH 18, and DH 19) and high flood pulse count (FH 4).

Group 4: Frequency: Within the frequency component, the first PC, selected for its significance, contributes 37.72% of the variance (Figure 3(d)). It is mainly associated with high flood pulse count (FH 1), flood frequency (FH 7, FH 8, and FH 9), and low flood pulse count (FL 1).

Group 5: Timing: In the timing component, the first PC accounts for 26.94% of the variance (Figure 3(e)). Significant parameters in this component include mean maximum October (MH 10) and January (MH 1) monthly flows, mean minimum November (ML 11) and May (ML 5) monthly flows, and mean monthly flow in October (MA 21).

Group 6: Rate of change: The main parameters in the rate of change component capture 42.35% of the variance and are related to reversals (RA 8), no. day rises (RA 5), fall rate (RA 3), change of flow (RA 7), and rise rate (RA 1).

Group 7: Other: In the other component, the first PC is highly significant, explaining 61.55% of the variance (Figure 3). It is primarily linked to the skewness in daily flows (MA 5), spread in daily flows (MA 9 and MA 10), and skewness in monthly flows (MA 40).

Combining the identification of principal components and the selection of a representative subset of parameters with the physical and biological understanding of the targeted streamflow regimes can enhance analysis (Olden & Poff 2003). To support the choice of a specific subset, it is essential to demonstrate the ecological significance of these parameters. The comprehensive analysis of combining bootstrapping with PCA, as outlined in this section, forms the foundation for the next valuation phase.

During the barrage construction at the Kalanaur gauging station, significant alterations were observed in various flow regime components. Table 3 summarizes the observed alterations in each component.

These results indicate that the most significant alteration was in the frequency component (0.500), suggesting a notable change in the occurrence rate of specific flow events. Conversely, the timing and rate of change components experienced the least alteration, each with a value of 0.278. The overall alteration value, which represents an aggregate measure of all components, was calculated to be 0.379, denoting a moderate change in the flow regime due to the barrage construction.

This section deals with comparing hydrological alterations prior to the construction and after the operation of the Hathnikund barrage (Table 4).

The flow regime at Kalanaur, influenced by the Hathnikund barrage, displayed moderate alterations across various components. The magnitude of flow changed with an alteration value of 0.378, reflecting moderate changes in water volume. Flow variability saw a lower alteration value of 0.267, indicating minor inconsistencies in flow. Duration changes were also moderate, with a value of 0.378. Frequency alterations were less pronounced at 0.222, suggesting minimal impact on the occurrence of flow conditions. The timing of flow events was moderately altered at 0.311. The most significant change was in the rate of flow alterations, marked by a value of 0.500, pointing to a substantial increase in the speed of flow changes. Other unspecified flow regime changes were also moderate at 0.311. Overall, the cumulative impact of the barrage on the flow regime was moderate, with a total alteration value of 0.328.

The planform changes are studied in three stages: prior to the construction (1993), during construction works of the barrages (1999), and different periods after the operation of the Hathnikund barrage (Table 5).

The multivariate analysis underscores the complexity of river flow characteristics across different hydrological groups, revealing their critical ecological roles. Group 1 emphasizes the significance of peak flow magnitudes in shaping river ecosystems, highlighting their dominant impact on ecological integrity (Poff et al. 1997). Group 2 identifies flow variability as essential for maintaining aquatic biodiversity (Richter et al. 1996), while Group 3 focuses on the ecological importance of prolonged high flows for nutrient cycling and habitat sustainability (Junk et al. 1989). In Group 4, the frequency of floods is linked to impacts on riverine landscapes and biota, influencing sediment transport and species diversity (Tockner & Ward 1999). Group 5's analysis of timing stresses the importance of seasonal flow variations for triggering critical life-cycle events in riverine species (Thoms & Sheldon 2000). Group 6 explores how rapid changes in river flows can disrupt ecological balances, affecting physical and biological river processes (Kuriqi et al. 2021). Lastly, the ‘Other’ group examines flow skewness, a less studied but significant factor in predicting the extremity and predictability of flow events, impacting flood risk management and ecological resilience (Doyle 2005).

The RFHI offers a method to assess the impact of human interventions on river systems, highlighting significant modifications in flow regime components due to the construction and subsequent operation of the Hathnikund barrage across the Yamuna River. The flow magnitude was altered during construction by draining water stored behind the Tajewala barrage, using a rapid-release approach (Stroud 2012). The variability component was most affected during construction as the structures were not used for water abstraction, irrigation, or domestic use, leading to significant alterations. Additionally, the frequency component saw a substantial change, with a rating of 0.500, indicating a notable shift in the occurrence rate of specific flow events. This finding aligns with research by Lee et al. (2023), who reported similar frequency alterations in river systems following infrastructural developments.

The moderate overall alteration values (0.379 during construction and 0.328 post-operation) suggest that significant changes do not constitute extreme modifications to the flow regime. However, moderate changes can accumulate over time, leading to substantial ecological and geomorphological impacts (Poff et al. 1997). These alterations were largely due to the increased discharge capacity of the Eastern Yamuna Canal from 800 cusecs in 1,830 to 4,000 cusecs to address water scarcity (Irrigation and Water Resources Department, Uttar Pradesh). Additionally, the construction of Powerhouse D at the Western Yamuna Canal hydroelectric station in 2004 in Haryana was aimed at addressing power shortages (Haryana Power Generation Corporation Limited). Increased urbanization in Haryana and Uttar Pradesh has led to a significant gap between water availability and demand, resulting in extensive water extraction from Hathnikund during lean seasons to meet agricultural needs (Kumar et al. 2019). Post-operation, the rate of change component showed the most notable alteration (0.500), highlighting risks of rapid flow changes to ecological balance and sediment dynamics (Acreman et al. 2014). Moderate changes in duration (0.378) and timing (0.311) indicate shifts in seasonal river patterns. Maintaining natural flow regimes is important, as frequency deviations can disrupt the reproductive cycles of aquatic life and habitat integrity (Richter et al. 1996) and impact species dependent on seasonal flows (Lytle & Poff 2004). The reduced variability (0.267) and moderate changes in magnitude (0.378) could influence sediment deposition patterns, potentially leading to habitat degradation for riverine species (Chen & Olden 2020). The significant rate of change could disrupt ecological balances, affecting species reliant on stable flow conditions for breeding and feeding (Iqbal & Dutta 2022).

The observed morphological changes in the Yamuna River from 1993 to 2018, detailed through planform parameters such as sinuosity, braid-channel ratio, channel width, channel area, bar area, and bar density, reflect significant influence from barrage construction and operation. The initial increase in sinuosity and braid-channel ratio suggests enhanced meandering and bifurcation, likely due to changes in flow velocity and sediment transport facilitated by barrage construction (Gurnell et al. 2016). Over time, a reduction in these parameters indicates a stabilization effect from the barrages, aligning with findings that link river engineering to reduced river fragmentation (Surian & Rinaldi 2003). Fluctuations in bar area and density highlight changes in sediment deposition patterns, initially increased by sediment accumulation upstream of barrages, then decreased due to regulated flows and sediment capture by the barrage (Major et al. 2012). Such alterations in sediment dynamics can significantly impact river morphology, influencing channel form and habitat structures (Magilligan & Nislow 2005).

The decommissioning of the Tajewala barrage introduces substantial sediment loads back into the river system, potentially leading to channel aggradation and increased braiding (Pearson et al. 2011; Major et al. 2012; East et al. 2015). This reintroduction of sediments can lead to significant geomorphological changes downstream, including altered sediment regimes and increased deposition (Doyle et al. 2003; Cheng & Granata 2007; Riggsbee et al. 2007; Major et al. 2012; Wilcox et al. 2014).

The observed drying of the river channel following barrage operations poses serious ecological risks, disrupting river continuity and affecting aquatic habitats (Palmer et al. 2009; Grill et al. 2015). Such conditions can lead to a cascade of ecological consequences, from altered sediment transport to the loss of habitats essential for aquatic species (Kumar et al. 2019).

Examinations of other barrages like Farakka reveal similar hydrological and ecological disruption patterns, including altered flow patterns, increased salinity, and impacts on aquatic and riparian ecosystems (Gain & Giupponi 2014). These findings emphasize the need for comprehensive river basin management strategies that consider human needs and ecological sustainability. The RFHI and the Environmental Flows Assessment will help examine historical flow data, current ecological conditions, and the life history requirements of important species (Richter et al. 1996; Poff et al. 1997).

The method presented has limitations. For example, in our case study, variations in extreme events within the Rate of change, Frequency, and Other flow regime categories are challenging to interpret, possibly influenced by climate changes (Sillmann et al. 2017) or data insufficiency. Additionally, our results were derived by equally weighting all flow regime categories, but adjusting weights could better capture critical extreme events tied to specific processes.

Mishra et al. (2007) observed a decline in the catch of IMCs from 2002 to 2004, which they attributed to water withdrawals from the Hathnikund barrage and changes in physical habitat. Specifically, the percentage of IMC and large catfish catches dropped significantly from a high of 5.4% (1997–1998) to 2.45% (2003–2004) and from 10.5% (1997–1998) to 6.53% (2003–2004), respectively. This decrease coincides with the start of increased water diversion around 2000–2001, suggesting a direct link to habitat alterations from the barrage operations. Although mahseer populations showed stability, likely due to fishery interventions like seed induction, the barrage threatens their long-term sustainability, which impedes migration necessary for breeding (Mathur et al. 2022).

Barrages often disrupt sediment transport and hydrological flows, crucial for maintaining aquatic habitats (Graf 2006). For example, changes in flow regimes can severely impact fish migration patterns and spawning sites, directly affecting biodiversity (Bhatt et al. 2017). Furthermore, alterations in river morphology, such as changes in channel width and depth due to barrage operation, can lead to habitat loss for various aquatic and riparian species, altering the ecosystem structure and function (Bunn & Arthington 2002). The Hathnikund barrage could similarly influence the Yamuna River's ecological balance, affecting its biota and ability to support the surrounding biodiversity. By examining these impacts, the paper could offer insights into sustainable water management that balances human needs with ecological preservation (Poff et al. 1997; Palmer et al. 2009).

The study on the hydromorphological shifts of the Yamuna River due to barrage construction provides significant insights that can guide future river management and policymaking:

• 1. The RFHI revealed moderate alterations in flow frequency and rate of change due to barrage operations. This demonstrates the barrages' significant role in altering the natural hydrological regime, which can have lasting impacts on river health and downstream water availability.

• 2. Analysis through satellite imagery confirmed that barrages have contributed to changes in river sinuosity, braid-channel ratio, and other planform parameters. Such physical transformations suggest that barrage operations are a major driver of river morphology changes, impacting sediment transport and river stability.

• 3. The alterations in flow and morphology due to barrages likely lead to significant ecological impacts, including changes in habitat quality and biodiversity. The study underscores the importance of maintaining ecological integrity through sustainable flow management practices.

• 4. The findings advocate for an integrated approach to river basin management that balances developmental needs with ecological sustainability.

The financial support for this study was received from cGanga, NMCG, Department of Water Resources, River Development & Ganga Rejuvenation, Ministry of Jal Shakti, India.

M.M.: Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Validation, Visualization, Writing – original draft, review, and editing. V.T.: Conceptualization, Funding acquisition, Supervision, Writing – review and editing.

Data cannot be made publicly available; readers should contact the corresponding author for details.

The authors declare there is no conflict.