Prioritising sub-watersheds using morphometric analysis, principal component analysis, and land use/land cover analysis in the Kinnerasani River basin, India

Geomorphometric is a quantitative land surface study science that uses statistical, mathematical, and image analysis techniques to assess the morphology, hydrology, ecology, and other features of any geographical location (Obi et al. 2002). The assessment and quantitative examination of the planet's surface layout, structure, and size of features is known as morphology (Clarke 1966; Agarwal 1998). According to multiple morphometric studies, drainage basin morphology reflects various geomorphological and geological processes across time, which is a well-acknowledged morphometric principle (Strahler 1952; Miller 1953).

The drainage divide is the drainage catchment or basin's physical boundary. All locations above the elevation of the outlet and inside the drainage divide that separates adjacent watersheds are included in the watershed area (Banerjee et al. 2015). A catchment's classification is determined by its size, drainage, form, and land use pattern. There are many types of watersheds, including mini watersheds, micro watersheds, milli watersheds, sub-watersheds (SWs), and macro watersheds (Singh 1994). Morphometric analysis is beneficial for watershed studies because it demonstrates the connection between numerous characteristics of a river catchment, such as the order of a stream and the length of the stream (Sreedevi et al. 2005). In addition, watershed management has raised awareness of the planet's natural resources, including water and soil (Redvan & Mustafa 2021).

The size, structure, gradient, drainage density, and other catchment features can be linked to different hydrological measurements (Singh et al. 2021). As a result, morphometric studies reveal details concerning the production of a wide range of ground-level processes (Singh 1992), which can be effectively depicted through the assessment of the relief feature, the shape or aerial feature, and the linear feature (Nautiyal 1994). Using traditional techniques, the drainage features of numerous river catchments and sub-basins around the world have been investigated (Horton 1945; Strahler 1957, 1964). Geomorphic parameters and morphometric aspects can be used to assess the drainage system's surface runoff and flow intensity (Ozdemir & Bird 2009).

For the morphometric study, geographic information system (GIS) and remote sensing (RS) techniques can be utilised to measure the linear feature, areal feature, and relief feature of a watershed (Bogale 2021; Khan et al. 2021). In comparison to the traditional method (Horton 1945), the incorporation of GIS approaches is very appropriate for morphometric evaluation. Numerous morphometric investigations have been conducted in numerous places around the globe, as well as in India's several river basins (Chalam et al. 1996; Chaudhary & Sharma 1998), and they have all come to the same conclusion: GISs are an effective approach for investigating catchment morphology (Prakash et al. 2019). Many watershed features have been investigated using the traditional approach, which is time-consuming and inconvenient (Strahler 1954, 1957). The morphometric assessments of natural drainage and their drainage network assessments may be done more precisely and inexpensively with the advancement of RS and GIS (Arabameri et al. 2020).

To highlight the catchment's natural qualities related to the primary issue of interest, geomorphometric criteria are mostly used in prioritising analysis. According to Malik & Bhat (2014), geomorphometric parameters could be used with other thematic maps to identify the optimal region for soil and water conservation and watershed prioritisation. Additional thematic maps that focus on managerial features, which are human-made, are provided. It has been considered a management factor for prioritising because the change in land use has an impact on the hydrological process, particularly in the acceleration of soil erosion (Javed et al. 2009). Geomorphometric and land use/land cover (LULC) data were combined by Puno & Puno (2019) to identify watersheds for conservation in the Philippines. Geomorphometric and land use/landcover data were utilised in a study by Javed et al. (2011) that looked at prioritising SWs using RS and GIS techniques.

Previous research has mainly used a standard compound value for the analysis method of prioritising, which is calculated by averaging the starting ranks of importance for all parameters (Aher et al. 2014). However, principal component analysis (PCA) has been utilised to inform a few investigations (Farhan et al. 2017; Meshram & Sharma 2018). The authors (Sharma et al. 2015; Arefn et al. 2020) proposed reducing the dimension of morphometric parameters based on the PCA and finding variables that largely account for the variance exhibited in a variety of metrics. Examples of multivariate statistical methods that can be used to pinpoint the underlying components or variables that largely account for system fluctuations include factor analysis (Shrestha & Kazama 2007). They are created to condense a large number of components into a condensed set of features while maintaining the linkages between the actual data.

The regionalisation of the hydrologic models is made easier by recent geomorphologic research. The study of the geomorphologic properties of such catchments becomes significantly more relevant because the majority of catchments are either ungauged or lack appropriate data (Sharma et al. 2009). Due to the acceleration of watershed management programmes for the protection, development, and good use of all-natural resources, including soil and water, the demand for precise information on watershed runoff and sediment output has expanded significantly over the past few decades (Mishra et al. 2013).

The Kinnerasani watershed was chosen for this study because no previous investigations have been conducted. For this study's SW prioritisation, morphometric analysis, PCA and LULC analysis for soil, water conservation, and natural resource management were taken into consideration. Furthermore, the study uses these three methods to identify the SWs related to the common priority.

PCA was utilised to evaluate one of the important morphometric characteristics for prioritising catchments based on characteristics that are highly correlated with components. Using statistical programme for the social sciences (SPSS) version 22.0 software, the 18 morphometric characteristics were reduced to 5 important components in current research. The rotated component matrix reveals that each component considers one highly correlated characteristic. After utilising the PCA approach to obtain five parameters, the following step is to rank each SW feature. The following stage is to determine the Cp value. The SWs were classified into three categories based on their Cp values: high, medium, and low.

Using 2020 land cover Sentinel-2 imagery from the Environmental Systems Research Institute (ESRI), the LULC mapping was done at the SW level. The LULC categories were determined based on the ESRI land cover. Based on a common criterion that applies to each SW, the LULC categories are taken into consideration for prioritising SWs. The Cp value should be calculated next. Based on the SWs' Cp values, three categories, high, medium, and low, were created.

Each SW of the Kinnerasani is classified into three categories for examination and analysis: linear, relief, and shape features.

The stream order, stream length, RHO coefficient, and other linear aspects of catchment morphometric analysis are discussed here.

The number of stream segments in a single order is calculated individually and is referred to as the ‘stream number of that order’ (Horton 1945). As the order of the stream gets higher, the stream's number decreases. The order of the streams and the stream number in the respective order have a negative connection.

According to Schumm (1956), it is the ratio of the overall number of stream segments of one order to the next maximum order in a river catchment, and it is associated with the arrangement of branches of a river system. SW12 has a maximum value in this study, whereas SW24 has a minimum value.

The overall length of every order's distinct stream segments is called the order's stream length. It is computed by classifying the overall distance of every stream in a specific order by the stream number in that order to get the average distance of a stream in each order (Horton 1945). In the present research, SW14 has a maximum stream length and SW16 has a minimum stream length.

It is determined by multiplying the overall length of order's stream by the overall number of segments in the order (Strahler 1964). In the present research, SW2 has a maximum L_sm and SW10 has a minimum L_sm.

It is the ratio of the average stream length of the current order to that of the next smaller order. In the current study, SW7 has a maximum stream length ratio and SW24 has a minimum stream length ratio.

In order to arrive at a more accurate R_b, Strahler (1954) used a weighted average ratio of bifurcation, which was calculated by multiplying the R_b for every successive set of patterns by the overall number of streams occupied in the ratio. In the present research, SW12 has a maximum mean bifurcation ratio and SW13 has a minimum mean bifurcation ratio.

The overall amount of passage segments of all stream patterns in each unit area is called stream frequency. In the present research, SW3 has a maximum stream frequency and SW24 has a minimum stream frequency.

According to Horton (1945), it is described as the length of streams in each unit area. The five classifications of drainage densities are: very coarse (is less than 2), coarse (is between 2 and 4), moderate (is between 4 and 6), fine (is between 6 and 8), and very fine (is greater than 8) (Strahler 1964). In the present research, SW3 has a maximum D_d and SW24 has a minimum D_d.

It is the overall amount of stream segments of all orders in a catchment to the catchment's perimeter. SW3 has the maximum value, whereas SW24 has the minimum value in this present research.

According to Schumm (1956), the maximum result of the L_o indicates maximum surface runoff, and the minimum result of the L_o shows minimum surface runoff. In the present research, SW24 has a maximum L_o and SW3 has a minimum L_o.

It is a significant measure that links D_d to the physiographic improvement of a catchment, making it easier to assess the drainage network's storage capacity and, as a result, a predictor of the watershed's eventual degree of drainage development (Horton 1945). In the present research, SW12 has a maximum RHO coefficient and SW7 has a minimum RHO coefficient.

According to Faniran (1968), it is described as the ratio of F_s to D_d. D_i is the symbol for it. In the present research, SW3 has a maximum drainage intensity and SW24 has a minimum drainage intensity.

According to Faniran (1968), it is the product of F_s and D_d. I_f is the symbol for it. In the present research, SW3 has a maximum I_f and SW24 has a minimum I_f.

It was first proposed by Schumm (1956). It is the reverse of D_d. In the present research, SW24 has a maximum C_cm and SW3 has a minimum C_cm.

The relief features of the relief, the relief ratio, the relative relief, and the roughness number have all been determined.

It is the difference in elevation between catchment's upper and lower points (outlet). In the present research, SW21 has a maximum relief and SW24 has a minimum relief.

It is the ratio of catchment's overall relief to its longest dimension that is similar to the major drainage line (Schumm 1956). In the present research, SW4 has a maximum relief ratio and SW18 has a minimum relief ratio.

From the maximum level on the catchment perimeter to the stream's mouth, the maximum basin relief was achieved (Melton 1957). In the present research, SW4 has a maximum R_hp and SW14 has a minimum R_hp.

It refers to the overall region estimated on a horizontal plane that contributes overland flow to the canal segment of the provided order, which contains all lower-order branches. It includes the form factor, circularity ratio, and elongation ratio.

According to Miller (1953), it is the ratio of catchment's region to the region of a circle with the same circumference as the catchment's perimeter. In the present research, SW12 has a maximum R_c and SW3 has a minimum R_c.

It is the ratio of the diameter in a circle of the same region as the catchment to the catchment's maximum length (Schumm 1956). In the present research, SW16 has a maximum R_e and SW14 has a minimum R_e.

It is the ratio of the catchment region to the square of the catchment distance. According to Horton (1932), the intensity of the flow of a catchment in a specific area is represented by this factor. The basin would be extended as the F_f value decreases. Maximum peak flows of a shorter span occur in a catchment with high form factors, whereas minimum peak flows of a longer span occur in extended catchments with low form factors. In the present research, SW16 has a maximum F_f and SW14 has a minimum F_f.

According to Chorley et al. (1957), the gradient of the catchment is determined by the Lemniscate value. In the present research, SW19 has a maximum K and SW16 has a minimum K.

The most essential quantitative morphometric features for this research are identified and utilised. The three types of morphometric features (linear, relief, and shape) have been used to rank highly vulnerable SWs since they are associated with surface overflow and the possibility of soil erosion either directly or indirectly (Nookaratnam et al. 2005; Javed et al. 2009). The most erodible soil in a basin is indicated by the most significant value of the relief and linear features. As a result, the SW with the highest value receives a rank of 1 and so on. On the other hand, the most erodible soil in a basin is indicated by the lowest value of the shape features. As a result, the SW with the lowest value receives a rank of 1 and so on. For example, the Cp value would be 12.17 if all the SW1 ranks were totalled up and divided by the 18 features. Other SWs have undergone the same process.

A correlation matrix is generated using the SPSS version 22.0 software to determine the inter-correlations between the geomorphic features. The correlation matrix of the 18 features reveals that the perfectly positive (correlation coefficient +1 or −1) correlation occurs between C_cm and L_o. The strong correlations (correlation coefficient more than 0.90) occur between D_d and F_s; between L_o and D_d; between D_i and F_s; I_f with F_s, D_d and D_i; between C_cm and D_d; between F_f and R_e; K with F_f and R_e. A good correlation (correlation coefficient more than 0.75) occurs between R_hp and R_h; D_t with F_s and D_d; L_o with F_s and D_t; D_i with D_d, D_t and L_o; I_f with D_t and L_o; C_cm with F_s, D_t, D_i, and I_f. Moderately correlated parameters (correlation coefficient more than 0.60) occur between P and R_bm; R_n with B_h and R_h. At this point, dividing the features into components is quite challenging and assigning physical significance. Therefore, the correlation matrix has been subjected to the PCA in the next phase.

PCA is a statistical tool for identifying hidden factors that explain the pattern of correlations within a set of observable variables while maintaining true initial data. Using SPSS version 22, correlation analysis was used to evaluate the relationship between each morphometric feature and the others. Blue and red cells with values of 1 and −1, respectively, represent the strongest positive and negative correlation between two features in the provided correlogram, as shown in Table 4. A component loading matrix is used in PCA to express quantitatively how closely the component values relate to the original morphometric features. Each component has a certain parameter assigned to it, and these weights are referred to as loadings on each component. In addition to comprehending the mechanisms that lead to the observed connections between the selected variables, PCA offers a reduced data matrix called the component score (or weightings) matrix. The component loading matrix shows how the initial 18 morphometric features were reduced to 5 significant components, and it occasionally takes into account the interactions between the rotated components and the original values. These connections are shown in terms of the percentage they each contributed to the variation in the starting set of data. Furthermore, it is evident that each component has stronger correlations with some parameters than others, depending on which parameters are thought to be the most beneficial. The component loading matrix is shown in Table 5. The first five components, whose combined eigenvalues are more than 1 and account for 90.205% of the total variance in the original data, are clearly significant. Component 1 is substantially connected with stream frequency, component 2 is highly correlated with form factor, component 3 is highly correlated with the ruggedness number, component 4 is highly correlated with mean bifurcation ratio, and component 5 is highly correlated with circulatory ratio, according to the rotated component matrix (Table 6) (Meshram & Sharma 2017). Stream frequency, form factor, ruggedness number, mean bifurcation ratio, and circulatory ratio are the essential morphometric characteristics derived from PCA. As a result, these factors are used to prioritise the Kinnerasani catchment's 24 SWs.

Table 4

Correlogram

View Large

Trees are defined as any notable collection of tall (15 feet or higher) dense vegetation, typically with a closed or dense canopy; examples include wooded vegetation, collections of tall, dense vegetation within savannas, plantations, swamps, or mangroves (dense or tall vegetation with ephemeral water or a canopy too thick to detect water beneath). Any notable collection of tall (15 ft or higher) dense vegetation, usually with a closed or dense canopy; examples include wooded vegetation, collections of tall, dense vegetation within savannas, plantations, and mangroves. SWs with a maximum percentage (%) of trees have been provided with a minimum priority, whereas those with a minimum % of trees have been provided with a maximum priority. In the present research, SW2 has the maximum % of trees, whereas the minimum % of trees is found in SW24.

The definition of grass is ‘open areas covered in homogenous grasses with little to no taller vegetation; wild cereals and grasses without obvious human plotting (i.e., not a plotted field)’. Examples include natural meadows and fields with sparse to no tree cover, open savanna with few to no trees, parks, golf courses, lawns, and pastures. Open spaces with uniform grass cover and little to no higher plants. SWs with a maximum % of grass have been provided with a minimum priority, whereas those with a minimum % of grass have been provided with a maximum priority. In the present research, SW12 has the maximum % of grass, whereas the minimum % of grass is found in SW24.

Cereals, grasses, and crops not at tree height that have been planted or plotted by humans include corn, wheat, soy, and fallow areas of structured land. SWs with a maximum % of crops have been provided with a minimum priority, while those with a minimum % of crops have been provided with a maximum priority. In the present research, SW21 has the maximum % of crops, whereas the minimum % of crops is found in SW1.

Scrub/shrub is defined as a mixture of small groups of plants or a single plant scattered across a terrain with exposed rock or dirt; thick woodlands with visible gaps that are clearly not taller than trees; savannas with very scant grasses, trees, or other vegetation; and areas with a moderate to sparse cover of bushes, and tufts of grass. A landscape with single scattered plants, small groups of plants, and exposed dirt or rock. SWs with a maximum % of scrub have been provided with a minimum priority, while those with a minimum % of scrub have been provided with a maximum priority. In the present research, SW22 has the maximum % of scrub, while the minimum % of scrub is found in SW21.

To determine the common SWs falling under each priority, the results of the three methods, such as morphometric analysis, PCA, and LULC analysis, have been compared. Three methods identify five SWs as common SWs with a medium priority: SW4, SW6, SW13, SW15, and SW21. SW10 is a common SW that is low priority on the other side. The other 18 SWs exhibit a slight difference in their priority under the three methods. Table 9 shows the common priority among the three methods.

The morphometric analysis, PCA, and LULC analysis determined using RS and GIS methodologies provided researchers with a good understanding of the development of a catchment and its response to hydrologic conditions, allowing for more effective natural resource management strategies in the Kinnerasani River basin. In the present research, 18 morphometric features, 5 PCA features, and 4 LULC features have been derived and scientifically investigated. The SW3, SW9, and SW12 SWs are of high priority according to the morphometric analysis-based prioritisation approach. The outcomes of the PCA-based prioritisation place a high priority on the SW2 and SW3 SWs. The results of the LULC-based prioritisation place a priority on the SW17, SW19, SW23, and SW24 SWs. The common SWs within each priority according to three different methodologies are SW4, SW6, SW10, SW13, SW15, and SW21. In order to stop additional soil degradation, it is also crucial to implement the proper soil erosion management techniques in high-priority SWs. The study results here suggest a helpful tool to define areas (high priority) for planning the methods to prevent soil erosion and encourage soil conservation. Depending on the appropriate location (high priority) and design criteria, this may involve both physical and biological solutions, including building bunds, check dams, providing vegetative and stone barriers, and planting multipurpose tree species. In addition, the study helps in protecting the existing natural resources and helps water resource managers and policymakers make better decisions in a field where data are scarce. This information can be utilised to design, execute, and adapt the best SW-level planning and management techniques.

The authors would like to thank the anonymous reviewers for their instructive comments, which helped to improve this paper.

P.R.S. conceptualized the whole article, developed the methodology, involved in software, conducted data curation, and wrote the original draft. A.M. supervised, visualised and investigated the article and wrote the review and edited the article.

There was no funding for this project.

All relevant data are included in the paper or its Supplementary Information.

The authors declare there is no conflict.