Introduction

Globally, an estimated 1965 million hectares of land is subjected to one or other kind of degradation, out of which soil erosion by water and wind accounts for 1643 million hectares (Oldeman, 1991). According to the studies conducted by Rao (2000), annually, 6000 million tons of soil are lost in India. According to the Ministry of Agriculture (Soil and Water Conservation Division), 175 million hectares of vegetative land are degraded in some form or other. In India, 29.30% of the total geographical area, equivalent to 96.43 million hectares of land, is severely affected by varied land degradation (SAC, 2016). According to Singh et al. (1992), areas such as Shivalik Hills, northwestern Himalayan regions, Western Ghats, and parts of peninsular India are most severely affected by soil erosion, i.e., about 20 Mg/ha/year.

Due to increased environmental degradation with deforestation, the material carried by rivers has risen from 9.30 billion tons to 23 billion tons in a year in the last 50 years. Further, the situation becomes worse, noting that 15 billion tons a year is being contributed by Asian rivers to the ocean. Narayana and Babu (1983) observed that the Indian rivers carry and transport about 1572 million tons of soil to sea. Further, about 480 million tons of soil are deposited into reservoirs, reducing their storage capacity by 1–2%.

There has been a similar increase in the water scarcity problems with much severity in most parts of the world. India receives 4000 BCM (Billion Cubic meter) in form of precipitation (including snowfall and rainfall); out of this, 1869 BCM accounts for the available water. The utilizable water from available surface and groundwater accounts for 690 BCM and 490 BCM, respectively. As per the water stress index given by Falkenmark and Lindh (1976), the countries with less than 1700 cubic meters of annual per capita water availability are considered water-stressed. Similarly, when this per capita water availability goes below 1000 cubic meters, the country is declared water scarce. As per studies of the Central water commission, per capita average annual water availability in India for the year 2010 was 1588 m3/year but by 2025 and 2050, it is going to drop to 1434 m3/year and 1140 m3/year, respectively (Dadhwal et al., 2012).

The watershed is considered the most viable planning unit to conserve these resources (F.A.O, 1985). The rational utilization of land and water resources, ensuring its sustainable and optimum production, and at the same time exerting minimum pressure on natural resources and the environment calls for the concept of Integrated Watershed Management. This watershed-based methodology has been logical since the land and water resources have concordant effects when developed on a watershed basis. Various techniques have evolved for watershed prioritization, out of which quantitative morphometric analysis has given promising results. Morphometry is defined as the quantitative analysis of the earth’s shape, its configuration, and various landforms. It has evolved as an essential tool for identifying and prioritizing highly eroded watersheds (Nautiyal, 1994). The in-situ monitoring of soil erosion for large watersheds is very costly. Hence the geomorphometric analysis is mainly carried out using the geographical information systems approach. The regionalization of the hydrologic models is implemented effectively using these geomorphologic studies. The linear, areal, and relief characteristics have played a significant role in understanding the watershed’s hydrological nature (Chow, 1964; Strahler, 1964). Due to the acceleration of watershed management programs for conservation, development, and beneficial use of natural resources such as soil and water, the demand for timely and updated information on watershed runoff and sediment yield has grown enormously in the last decade (Gajbhiye & Mishra, 2012; Meshram & Sharma, 2017; Mishra et al., 2013).

Several researchers use the conventional approach of compounding different morphometric factors for watershed prioritization and the sustainable conservation and management of the soil and water resources (Abdeta et al., 2020; Chandniha & Kansal, 2017; Nookaratnam et al., 2005; Thakkar & Dhiman, 2007; Waiyasusri & Chotpantarat, 2020). But it has been noticed that morphometric data is inherently multivariate (Mather & Doornkamp, 1970). Therefore, the traditional statistical approaches cannot simultaneously address the similitude between the morphometric parameters. Globally, researchers have started implementing principal component analysis, factor analysis, and hierarchical clustering analysis based on multivariate techniques. They are utilizing the inherent multidimensional data of morphometry to reveal its underlying structure. Bothale et al. (1997) delineated 135 watersheds from the Bajaj Sagar dam catchment of the Mahi basin. The authors used the data reduction technique of principal component analysis (PCA) for reducing the data redundancy. The PCA resulted in 12 components, grouped lately by hierarchical cluster analysis resulting in 50 eco watersheds having similar characteristics. Gopinath et al. (2016) analyzed the watershed management planning activities for the Kuttiyadi river basin, Kerala. The basin is susceptible to flood and inundation and exhibits high runoff. The authors used a multi-criteria decision-making approach to identify the integrated effects of morphometric parameters on soil erosion. Mangan et al. (2019) utilized the correlation matrix to assess the interrelationship between the Nanganji river basin's morphometric parameters situated in Tamil Nadu. The authors implemented factor analysis and grouped the parameters into three factors, i.e., shape, magnitude, and runoff factors. Jhariya et al. (2020) prioritized the Bindra watershed in Chhattisgarh by implementing the analytical hierarchical process (AHP) based on multi-criteria decision analysis. The author delineated 16 watersheds and categorized them into very high, moderate, and low priority zones using various factors such as soil loss, land capability classification, runoff, and sediment yield in a GIS environment. Arefin et al. (2020) delineated seventeen fifth-order and three sixth-order watersheds using SRTM DEM. They utilized 16 morphometric parameters and implemented principal component analysis (PCA) for watershed prioritization. The drainage density, circularity index, elongation ratio, and bifurcation ratio were compounded for final priority. Prieto-Amparán et al. (2019) performed the quantitative multivariate geomorphic characterization of the Conchos River basin, Mexico. They delineated 31 watersheds for prioritization purposes and utilized principal component analysis (PCA), group analysis (GA), and the compounded parameter ranking methodology. The 31 watersheds were grouped into five groups based on their erosion susceptibilities potential.

All the prior studies are based on open-source digital elevation models (DEM) such as shuttle radar terrain mapping (SRTM) and advanced spaceborne thermal emission and reflection radiometer (ASTER) onboard Terra satellite having a spatial resolution of 30 m. These existing datasets are relatively older, and morphometric phenomena are dynamic. Therefore, in this study, interferometrically generated Sentinel 1A/1B based on DEM was used to prioritize the upper Ghaggar watershed, which forms the part of one of the most fragile ecosystems in Lower Shivaliks having highly erodible soils which are lost at an alarming rate of 0 to 20,119 tones/hectare/year (Chauhan et al., 2020a). The principal component analysis (PCA) was used to reduce the parameters into five components or factors. The agglomerative hierarchical clustering technique by implementing Ward’s method was used to create three priority clusters for the upper Ghaggar watershed finally.

Study Area

The study area chosen for the present study is the upper Ghaggar watershed having its confluence with the Medkhali river. It is part of India's fragile ecosystems apart from the Western and Eastern Ghats. The basin area extends from 76.86252 E and 30.61292 N to 77.21258 E and 30.90755 N and covers an area of 559.14 km2 (Fig. 1). It covers Panchkula, Solan & Sirmaur, and Sahibzada Azad Singh Nagar districts of Haryana, Himachal Pradesh, and Punjab. The upper Ghaggar watershed has a hilly topography ranging from 249 to 1869 m. The Ghaggar river originates from the Dagshai village near Shimla, Himachal Pradesh, and is one of the most prominent ephemeral streams of the Lower Shivaliks. The river has tributaries such as the Vedic river Saraswati, Medkhali, Markanda, Tangri, and Chautang. The Ghaggar River, while flowing southwest, demarcates the boundary of Haryana and Punjab in the north. The soils of the upper Ghaggar watershed are classified into three classes, i.e., Eutric Cambisols (54.98%), Eutric Regosols (33.15%), and Eutric Fluvisols (11.88%). The study area's climate is humid subtropical and characterized by hot to sweltering summers and cool to mild winters. The average annual rainfall of the study area is 1260 mm calculated for the period ranging from 1969 to 2016. The Landuse/Landcover change analysis of the upper Ghaggar watershed conducted by Chauhan et al., (2020b) revealed that the built-up, shrubland and barren land have increased by 157.5 ha/year, 131.5 ha/year and 10.77 ha/year, respectively in the last 30 years (1985–2015) while the deciduous forest, evergreen forest and agriculture were reduced at a rate of 167.4 ha/year, 46 ha/year and 66 ha/year, respectively. The temperature of the study area varies from 4° to 37 °C. The study area also exhibits tremendously dry summers accompanied by dust storms. The geological structure of the study area is mainly composed of sandstone and conglomerate rocks.

Fig. 1
figure 1

Study area

Full size image

Dataset and Software Used

All-weather, day and night capability Sentinel 1A of spatial resolution 5 m in range direction and 20 m in azimuth direction of 06.01.2016 and 13.01.2016 (Table 1) was used to generate the DEM (Digital Elevation Model). The preprocessing of the pair of single look complex (SLC) products acquired in interferometric wide (IW) mode was performed in sentinel application platform (SNAP) software (version 6.0). Sentinel images were co-registred to create a stack by utilizing the precise orbit ephemerides (POE) orbit files. One image was used as master and the other as slave, and pixel values of the slave images were moved to align with the master dataset to attain a sub-pixel accuracy to ensure the same range and azimuth was contributed by each ground object. The complex conjugate of the master image was multiplied with the slave image for the generation of the interferogram. The process of terrain observation with progressive scans (TOPS) Deburst and TOPS Merge led to the formation of the seamlessly merged single image file. The speckle filtering algorithms and Goldstein Phase Filtering (Goldstein et al., 1988) incorporated within the SNAP software were used to decrease the speckle, simultaneously maintaining radiometric information. With the implementation of Statistical-cost, Network-flow, Algorithm for Phase Unwrapping (SNAPHU), the resultant phase was generated as a result of unwrapped filtering (Chen & Zebker, 2000). The unwrapped phase was imported using the NEXT European space agency (ESA) synthetic aperture radar (SAR) toolbox (NEST) software, and DEM was prepared from it. The resultant DEM was of 13.96 m resolution against the traditional 30-90 m DEM from SRTM and was updated topographically as compared to SRTM. Its vertical accuracy was 10 m when compared with the spot heights available on the Survey of India (SOI) toposheet.

Table 1 Details of the Sentinel 1 data used
Full size table

Methodology

Watershed Delineation and Drainage Analysis

The Sentinel 1 DEM was further used to generate the streams (Fig. 2) and delineate 92, i.e., sixty-third-order and thirty-two fourth-order sub-watersheds (Fig. 3), using the ArcSWAT 10.5 extension of ArcGIS 10.5. The area of individual 32 fourth-order sub-watersheds ranges from 2.749 (SW-4) to 25.862 km2 (SW-94) while that of individual 60 third-order sub-watersheds ranges from 0.515 (SW-11) to 8.881 km2 (SW-84). The ordering of the extracted streams is based on the nomenclature proposed by Strahler (1964). The linear, areal, and relief morphometric parameters analyzed during this study were calculated based on the formula shown in Table 2.

Fig. 2
figure 2

Drainage map of upper Ghaggar watershed

Full size image
Fig. 3
figure 3

Ninety-two sub-watersheds of upper Ghaggar watershed

Full size image
Table 2 Methods used to calculate the morphometric parameter
Full size table

Morphometric Analysis

The linear aspects extracted for these sub-watersheds are stream order (SOu), stream length, stream number (SNu), mean stream length (MSLu), Mean Stream Length Ratio (MSLRu), Bifurcation ratio (Rb), and Rho Coefficient (ρ). Basin area (A), basin length (Lb), basin perimeter (P), Lemniscate’s value (Lk), form factor (Ff), elongation ratio (Re), ellipticity index (Ic), and circularity ratio (Rc) are the areal aspects extracted for these sub-watersheds. Drainage density (Dd), drainage texture (Dt), stream frequency (Sf), infiltration number (If), and drainage intensity (Di) are the drainage characteristics extracted for the sub-watersheds. The relief characteristics of a basin represent the areal, volume, and altitudinal aspects of the basin landscape. The relief characteristics of morphometric analysis investigated in the study are absolute relief (Ra), dissection index (Dis), relative relief ratio (Rhp), relative relief (Rr), Ruggedness number (Rn), and slope (S).

The MSLu acts as an essential parameter for the calculation of drainage density. The MSLu helps in revealing the bedrock hydrological characteristics and extent of the basin. The MSLRu is an indicator of the geomorphological development stage of a basin. Horton (1945) explained that the Rb value ranges from 2 for flat or rolling basins and may go up to 3–4 for highly dissected or mountainous river basins. Strahler (1964) inferred from his study that the Rb values between 3 to 5 for the watersheds indicate that the basin's geological development is ineffective in altering the drainage patterns. The basins with higher Rb values produce a low but extended peak flow, whereas basins with lower Rb have a sharp peak flow (Strahler, 1964). The ρ helps in determining the relation of drainage composition and physiographic development of a sub-basin. The low value of ρ is indicative of low water storage during flood periods. It has a high erosion effect, while the higher value is indicative of higher hydrologic storage during floods and thus reduces erosion effects at peak discharges. The Lk values drastically control the shape of the drainage basin. The Lk values are one for circular basins, and as its value increases, the shape of the drainage basins becomes more and more elongated. An elongated basin is less effective in runoff discharge than a circular basin (Singh & Singh, 1997). The Ff is an indicative morphometric parameter of the flood-regime of the stream in the case of long and elongated drainage basins (Horton, 1932). The value of Ff varies from 0 for highly elongated basins to unity, i.e., 1 for perfectly circular shaped basins. Re acts as an imperative index for basin shape analysis, and the areas having higher values possess high infiltration capacity and low surface runoff. The Ie provides a close relationship between morphometry and hydrology (Stoddart, 1965). Lower Ie values indicate a quick runoff draining basin because of which the stream channels might swell or overflow, resulting in downstream flooding in case of heavy rainfall. Rc defines the circularity of the basin and is a dimensionless entity. Rc values range from 0 to 1 with zero value given to basins having a highly elongated shape to one value given to circular basins.

Dd acts as a permeability indicator of the drainage basin surface. High Dd is pertinent to the regions having impervious subsurface with weak structure resulting in more runoff. In contrast, low Dd is associated with highly resistant subsurface covered by vegetation and areas with low relief, resulting in lower runoff (Prasad et al., 2008). Singh (1976) defined the drainage texture based on the relative spacing of the streams. Dt is an indicator of the infiltration capacity, rock permeability, and relief aspect. The fine texture is prevalent in areas having low permeability and less resistance to erosion, and in the case of coarse texture areas, it is vice-versa. Sf shows a positive correlation with the drainage density and is, therefore, a similar indicator of infiltration capacity, rock permeability, and basin relief. High Sf values in hilly regions are associated with steep slopes and more significant rainfall, while in plain areas, lesser surface flow and low permeability result in low Sf value. Faniran (1968) defined the If of a drainage basin as the product of Dd and Sf. The higher values of If contribute to lower infiltration rates and result in higher runoff. Di was defined by Faniran (1968) as the ratio of the Sf to the Dd. It is an indicator of the extent to which the denudational agents depress the surface under drainage density and stream frequency effect.

Ra defines the maximum elevation of a basin area. This parameter helps in determining the rate of erosion concerning the recent summit or hilltops of a basin. Higher Ra is associated with high erosional activities. Smith (1935) coined the term Rr to the highest and lowest altitude points of a particular area. Rr plays a vital role in calculating average slope, dissection determination, and assessing the terrain development stages. Melton (1958) gave the Rhp as the ratio of the Rr and P of the area. Rhp ratio acts as an indicator of the relative velocity of the vertical tectonic movements. The lower values of Rhp pertain to the less resistive rocks. Dis is the ratio of Rr to Ra of an area. It acts as a vital parameter for developing an understanding of the degree of dissection and evolution of landform development stages in any given physiographic region.

The values of Dis range from 0, which implies a theoretical value as there is no region in nature that is passive to erosion to 1. Rn measures the structural complexity of the terrain. It is a dimensionless property calculated as a product of Rr and Dd of a given basin area having the same units. Patton and Baker (1976) discussed that the regions having higher ruggedness numbers accompanied with fine drainage texture and minimalistic length of overland flow on steep slopes have the expectancy of potential flash flooding. These morphometric parameters combination may lead to higher flood peaks for an area having a low ruggedness number even for the equivalent of rainfall events. The slope is an invariable relief morphometric parameter related to the infiltration capacity and runoff of an area. It is inversely proportional to the infiltration capacity. The mean values of linear, areal, and relief morphometric parameters for 92 sub-watersheds are given in Online Resource 1 (ESM_1.docx).

Principal Component Analysis of the Morphometric Parameters

The interrelationship between the different morphometric parameters was analyzed using a correlation matrix. To further strengthen the correlation matrix analysis results and give a clearer insight, a data reduction or factor analysis of the above-given parameters was done using the principal component analysis (PCA) in statistical package for the social sciences (SPSS) software version 25. Principal component analysis (PCA) is a multivariate data dimensionality reduction technique, which reduces the data into fewer factors but represents the entire dataset. The factors having an eigenvalue larger than one were used to maintain the significance of the contributing factor, and as a result, only five components were used. A normalized varimax rotation was applied on the resultant matrix to compute the factor loadings between the original geomorphic parameters and the final components/factors. The factor loading governs the correlation's strength, i.e., a higher factor loading value means stronger correlation and vice versa. All the score factors of the 92 sub-watersheds were divided into four factors, i.e., magnitude, relief, drainage composition, and dissection intensity factors depending on the morphometric aspects, which define 87.8% of the total variance.

Watershed Prioritization Using Hierarchical Clustering Analysis

In the study, agglomerative Hierarchical Cluster Analysis using Ward’s method (Ward, 1963) was applied for identifying the natural grouping within a given dataset for watershed prioritization. The more related data come together to form a homogeneous group while the dissimilar data become part of some other contrasting group. Ward’s method (Ward, 1963) is also called the minimum variance method because of its ability to create even-sized compact clusters but is computationally intensive. It offers the advantage of separating the cluster if the variance among the cluster is higher than the defined threshold. The squared Euclidean distance was used as the distance parameter to assess the distance between the respective cluster.

Results and Discussion

Morphometric Analysis

The morphometric analysis of the upper Ghaggar watershed reveals that it is a seventh-order watershed. It consists of 3483 first, 737 s, 160 third, 32 fourth, 9 fifths, 3 sixth, and 1 seventh order stream, suggesting that these stream channels generate a significant surface runoff and sediment yield (Bhat et al., 2019; Obeidat et al., 2021). The stream length of the basin is higher for its first order and decreases with an increase in stream order. Any deviation of this trend is due to inherent variations of topography and slope. The upper Ghaggar watershed's stream lengths for its 1st, 2nd, 3rd, 4th, 5th, 6th, and 7th orders are shown in Table 3. The Upper Ghaggar watershed's mean stream length for its 1st, 2nd, 3rd, 4th, 5th, 6th, and 7th orders are shown in Table 4. Rai et al. (2017) indicated in their study that low values of MSLu are mostly associated with the mountainous environment. The low values of MSLu represent the high erosion potential and younger geomorphological development of the landscape. The general trend of MSLu increases as the stream order increases, but any abnormalities in this suggest changes in the underlying slope and geology, which affects the flow characteristics. The MSLRu of the upper Ghaggar watershed displays an increasing trend from smaller order to higher order indicating their matured geomorphological development stage as against the sub-watersheds displaying abrupt changes between the orders indicating the late youth stage of geomorphological development. The MSLRu of the upper Ghaggar watershed for 1st to 2nd, 2nd to 3rd, 3rd to 4th, 4th to 5th, 5th to 6th, and 6th to 7th, respectively are shown in Table 5. Mean Rb values for the upper Ghaggar watershed for 1st to 2nd, 2nd to 3rd, 3rd to 4th, 4th to 5th, 5th to 6th, and 6th to 7th, respectively are shown in Table 6. The inconsistent values across the orders are due to the strong geological and lithological impact over the basin area. The Rho coefficient value of the upper Ghaggar watershed is 0.56, indicating less water storage and high erosion during the flood.

Table 3 Stream length of upper Ghaggar watershed
Full size table
Table 4 Mean stream length of upper Ghaggar watershed
Full size table
Table 5 Stream length ratio of upper Ghaggar watershed
Full size table
Table 6 Mean bifurcation ratio of upper Ghaggar watershed
Full size table

The Lk, Ff, Re, and Ie values of the upper Ghaggar watershed are 3.20, 0.25, 0.56, and 2.27, respectively, displaying its elongated shape. The higher values of Lk suggest a potentially high erosive nature of the basin. The low Ff represents a flatter peak of flow for a longer duration, which facilitates groundwater percolation. The elongated basins are less effective in runoff discharges because of their high elevation and steep gradient (Rai et al., 2018). The Dd of the upper Ghaggar watershed is 3.15 (km/km2), having a medium drainage density. The upper Ghaggar watershed is having a coarser Dt (0.62), higher Sf (11.59), poor Di (3.94), and higher If (39.04), resulting in a higher runoff.

The average Ra and Rr of the upper Ghaggar watershed are 902.48 m and 221.56 m, respectively. The average Rhp of the upper Ghaggar watershed is 0.11 indicator of less resistive rocks resulting in a higher runoff. The average Dis of the upper Ghaggar watershed is 0.22. A moderately high Rn (0.62) is shown by the upper Ghaggar watershed with a higher potential for flash flooding. The average S value of the upper Ghaggar watershed is 15.84, which indicates a steep slope representing high erosion operable within the basin.

Correlation Analysis of Morphometric Parameters

The interrelationship between the 27 morphometric parameters for the 32 forth order sub-watersheds and 60 third-order sub-watersheds were analyzed to assess the mutual interdependence prevalent among the different parameters shown in Table 7. It is inferred from the correlation analysis that most of the morphometric parameters are positively correlated to each other. Further, there is no single dependent variable, but all the variables influence each other and are interconnected. The correlation matrix depicts that there exists strong (0.8–0.9), Good (0.7–0.8), moderate (0.5–0.7), and also negative correlation. Basin Area (A), being the critical morphometric parameter responsible for controlling the peak flow and the amount of sediment being eroded, strongly correlates with SNu, SLu, MSLu, Rb, P, Lb, and Lk. It is evident that as the area increases, so does the SNu and SLu. Basin area is strongly positively correlated to Lb, so an increase in the area leads to an increase in Lb.

Table 7 Pearson Correlation Matrix for the 25 parameters for sub-watersheds of upper Ghaggar watershed
Full size table

Similarly, the area also defines the watershed's shape since it is inversely proportional to Lk, so watersheds having lower Lk values are more circular and therefore have more area. The Lb shows a strong correlation with SNu, SLu, MSLu, Rb, A, and P. Strahler (1964) further indicated that basins with higher Rb values are more elongated and, therefore, produced a low but extended peak flow. In contrast, basins with lower Rb are more circular and produce a sharp peak flow. The Ff is strongly correlated with the Re. The value of Ff varies from 0 for highly elongated basins to unity, i.e., 1 for perfectly circular shaped basins. It is shown that the sub-watersheds are having slightly elongated shapes having low Ff with a flatter peak of flow for longer duration resulting in groundwater percolation. The Rc shows a good correlation with the relief parameters such as S, while it shows a moderate correlation with Ra, Rr, Rhp, and Rn. Rc attributes to the high to moderate relief and structurally controlled drainage system. The Dd shows a strong correlation with If. Although the inter-correlation matrix was helpful, it induced a larger number of components into fewer components based upon their importance (Siddiqui et al., 2020). Hence, the inter-correlation matrix is subjected to Principal component analysis (PCA) for categorizing the 27 morphometric parameters into 5 principal components.

Multivariate Analysis Using the Principal Component Approach

The PCA resulted in 5 significant components having an eigenvalue greater than one, as shown in Fig. 4. The eigenvalues for PC1, PC2, PC3, PC4, and PC5 are 10.79, 7.48,2.79,1.57 and 1.07, respectively. They account for percentage variance of 39.96, 27.70, 10.32, 5.82, and 3.98, respectively, which account for 87.78% of the total variance explained by 27 morphometric parameters. The component-wise eigenvalues, variance percentage, and cumulative variance percentages are shown in Table 8. The most correlated and loaded variables in a particular component displayed in bold are shown in the varimax rotated component matrix Table 9. The PC1 shows the highest factor loading for positively SNu, SLu, MSLu, A, P, Lb, and Lk, contributing to 39.96% of the total variance and designated as the sub-watershed magnitude factor corresponding to the watershed dimensions. The PC2 shows the highest factor loading for positively correlated Rc, Di, Ra, Rr, Rhp, Dis, Rn, and S, contributing to 27.70% of the total variance and designated as the Relief factor corresponding to the basin relief and slope steepness.

Fig. 4
figure 4

Scree plot illustrating the 5-component solution resulting from PCA

Full size image
Table 8 Component wise explanation of total variance
Full size table
Table 9 Varimax rotated component matrix for 27 variables distributed in five principal components
Full size table

This factor illustrates that the sub-watersheds' hydrological response behavior and higher value or steeper slopes are more prone to runoff and soil losses. The PC3 shows the highest factor loading for positively correlated MSLRu, and ρ contributes to 10.32% of the total variance and can be designated as the Drainage composition factor. This factor illustrates the Channel storage capacity, which controls flood crest intensities downstream and is necessary for flood routing and control (Horton, 1945). The PC4 and PC5 show the highest factor loading for positively correlated Sf and Dt, contributing to 5.82% and 3.98%, respectively of the total variance and designated as the Dissection Intensity factor. The factor scores for the 92 (32 Fourth + 60 Third Order) sub-watersheds of the upper Ghaggar watershed are shown in Table 10. The bold values show the highest among the magnitude, relief, drainage composition, and Dissection Intensity factors for each of the 92 (32 Fourth + 60 Third Order) sub-watersheds. The dominating factor scores prevalent within the 92 sub-watersheds are shown in Fig. 5.

Table 10 Factor scores for 92 (32 fourth and 60 third-order) sub-watersheds of upper Ghaggar watershed
Full size table
Fig. 5
figure 5

Dominating factor scores prevalent within the upper Ghaggar watershed

Full size image

Prioritization Using Hierarchical Clustering Analysis

The clustering analysis resulted in three prominent clusters and is represented as Dendrogram (Fig. 6). In the Dendrogram, the sub-watersheds are plotted on the vertical axis, and the relative cluster difference represented by linkage distances are plotted on the horizontal axis. The Cluster wise mean values of the prominent linear, sub-watershed geometry characteristics, drainage, and relief characteristics, and their respective ranks are mentioned in the brackets are listed in Table 11. The Upper Ghaggar watershed and its sub-watersheds have dendritic drainage patterns, and these kinds of patterns are prominent where the underlying rock structure does not firmly control the stream channels, and channels follow the slope of the terrain. There is a high proportion of first-order streams in Cluster 3 sub-watersheds and indicating the structural weakness prevalent in the sub-watersheds associated with these clusters.

Fig. 6
figure 6

Dendrogram of the clustered sub-watersheds in upper Ghaggar watershed

Full size image
Table 11 Results of morphometric parameters derived based on hierarchical cluster analysis method for the 92 sub-watersheds
Full size table

Cluster Id 3 is having the highest values of Rb, which is indicative of highly dissected sub-watersheds. The higher Rb values correspond to sub-watersheds significantly influenced by drainage network, whereas its low values are associated with sub-watersheds having minimal disturbing drainage patterns structurally less disturbed. The ρ is dependent on hydrologic, geologic, and physiographic factors, which ultimately determine the relation of drainage composition and physiographic development of a sub-watershed. The low value of ρ indicates low water storage during flood periods and has a high erosion effect, while the higher value indicates higher hydrologic storage during floods and thus reduces erosion effects at peak discharges. Hence, the lowest ρ values are in sub-watersheds associated with Cluster 1.

The shape of sub-watersheds is also a prominent factor in governing the prioritization since it controls the water flow rate to the main channel. In hierarchical cluster analysis, Lk, Ff, Re, Ie, and Rc were used. The sub-watersheds belonging to cluster 3 have low Ff, Re, Ie, and Rc are shown in Table 11, which implies that they have an elongated shape. According to Singh and Singh (1997), an elongated watershed is less effective in runoff discharge than the circular watershed. These elongated sub-watersheds have a flatter peak of flow for a longer duration, which results in high infiltration and low surface runoff. The sub-watersheds belonging to Cluster 1 have relatively high Ff, Re, Ie, and Rc in Table 11, indicating its circular shape, which further implies high peak flows and shorter lag time than elongated sub-watersheds. Circular sub-watersheds are at higher risk of erosion and simultaneously higher sediment load since the runoff is contributed by the entire area, putting these sub-watersheds at higher risk of flash floods.

The morphometric parameters such as Dd, Sf, Dt, Di, and If indicate the landscape's extent being dissected by the stream network. These help in the prediction of sub-watershed surface runoff, its sediment yield, and terrain dissection magnitude. The drainage texture of the 3 clusters varies from moderate to coarser. However, according to the area's geology, the entire watershed’s 60% area is under Shale, Sandstone, and Limestone characterizing its low permeability and its erosion-prone nature. High values of Dd, Sf, and If are present in Cluster 3, whereas low values prevail in Cluster 1, as shown in Table 11. The high values of Cluster 3 are characterized by steep slopes, high relief, weak and impervious subsurface resulting in high surface runoff. The low values of Cluster 1 are characterized by a gentle slope, low relief, resistance and permeable subsurface, resulting in less surface runoff and providing excellent groundwater recharge sites.

The relief characteristics define the terrain's erosional properties, landform, development of drainage network, and overland flow. In the study, morphometric parameters such as Ra, Rr, Rhp, Dis, Rn, and S were used to reveal the watershed’s erosion potential and runoff. The Ra and Rr of the sub-watersheds coming under Cluster 1 are 1321.11 m indicating its high erosional capacity to move water and sediments down the slope. The Rhp ranges from 0.37 to 3.22, with sub-watersheds coming under Cluster 3 having low values due to its low slopes and resistant bedrock, whereas Cluster 2 and Cluster 1 have relatively higher values (Table 11). Higher values of Rhp characterize steep slopes, high erosion intensity, and higher sediment transport. Therefore, the sub-watersheds having higher Rhp having peak discharges resulting in higher erosion. The mean values of Dis, Rn, and S of the sub-watersheds falling in different clusters range from 0.16 to 0.25, 0.30 to 0.88, and 7.92 to 22.67, respectively, as shown in Table 11. Higher Dis, Rn, and S values correspond to sub-watersheds falling under Cluster 1, while low values of Dis, Rn, and S correspond to sub-watersheds falling under Cluster 3. The higher values of Dis, Rn, and S indicate steeper slopes, higher dissection, less overland flow due to lower time of concentration resulting in higher possibilities of flash floods and more erosion than sub-watersheds having lower Dis, Rn, and S.

The average values of all the Cluster's priorities 1, 2 and 3 are 1.77, 2.07, and 2.16, respectively. The spatial distribution of the priority-wise sub-watersheds is shown in Fig. 7. The Cluster 1 sub-watersheds are with the highest priority, followed by Cluster 2, and finally, the least priority sub-watersheds falling under Cluster 3.

Fig. 7
figure 7

Final prioritized sub-watersheds of upper Ghaggar watershed

Full size image

Conclusions

The upper Ghaggar watershed, forming the part of the lower Shivaliks, is the most fragile ecosystem due to its highly erodible soil structures. Therefore, prioritization at the sub-watershed scale can help the decision/policy makers to address the soil erosion problems eminent within a specific region of the upper Ghaggar watershed. The area is also prone to flooding during the monsoon season, hence water conservation structures such as check dams, gabion walls, etc., can be constructed in the downstream of the basin which will reduce the potential of flash floods and will increase the groundwater recharge and utilization of surface water for irrigation. In the present study, watershed prioritization of the erosion susceptible areas has been achieved by synergistically fusing the remote sensing and geographical information system techniques backed with statistical techniques such as correlation analysis, principal component analysis, and hierarchical clustering analysis. Twenty-seven linear, areal, and relief morphometric parameters were chosen for ninety-two sub-watersheds, where thirty-two are fourth-order, and sixty are third-order. The Principal Component Analysis applied on the inter-correlation matric resulted in five significant components representing the entire dataset. These five principal components were grouped into four factors, i.e., magnitude factor (PC1), relief factor (PC2), drainage composition factor (PC3), and Dissection intensity factor (PC4 & PC5). The Upper Ghaggar watershed is primarily controlled by magnitude and relief factors since these together explain 67.66% of the total variance. Because of the larger number of sub-watersheds involved in the present study, hierarchical clustering techniques based on Ward’s algorithm resulted into three prime clusters, i.e., high, medium, and least priority showing considerable spatial variability. Out of ninety-two, 29, 46, and 17 sub-watersheds fall in high, medium, and least priorities clusters. The results of this study provide vital information to the planners and decision-makers regarding drainage morphometry utilizing the geospatial techniques, which can act as a base criterion for implementing soil erosion conservation strategies (such as planting vegetation, hillshade terracing) and assessment of surface/groundwater recharge potential and flash flood mitigation (by constructing water harvesting structures such as check dams, gabion structures, etc.).