Introduction

Due to demographic expansion, the areal extent land of use/land cover has been changed widely and it may disturb ecological balance (Yin et al. 2011; Hassan et al. 2016). In water resource planning, management, and hydrological modelling of a drainage basin, evaluating the relation between rainfall and runoff is crucial (Welde and Gebremariam 2017; Gayathri and Jayalakshmi 2018). For rainfall-runoff modelling, the Soil Conservation Service Curve Number (SCS-CN) method is an important and basic concept (Satheeshkumar et al. 2017; Singh and Satapathy 2017). The United States Department of Agriculture (USDA) established the SCS-CN system, which is also known as the NRCS-CN method. It is adequately defined in Sect. 4 of the National Engineering Handbook (SCS 1956, 1964, 1971, 1985, 1993). The procedure is straightforward, systematic, and reliable. The methodology’s reliance on a single parameter known as CN distinguishes it from other approaches (Rawat and Singh 2017). Climate parameters, as well as basin features such as soil group, LULC, soil texture, soil permeability, land slope, and antecedent moisture conditions, are all taken into account in this method (Peng et al. 2012; Satheeshkumar et al. 2017; Gayathri and Jayalakshmi 2018). The NRCS-CN method is most commonly used to calculate the depth of surface runoff from a small agricultural watershed during a rainfall-runoff event. This method makes use of simple empirical formulas as well as widely available curves and tables (Shadeed and Almasri 2010; Nayak et al. 2012; Rawat and Singh 2017). This is a method which considers land use for computing runoff depth due to occurrence of rainfall event over a catchment. The NRCS-CN approach became the most common for runoff estimation due to the extensive availability of data required as input and the method’s simplicity. The NRCS-CN approach was originally designed to estimate event runoff in a small agricultural watershed, but it can be used in a variety of climates. The NRCS-CN approach can be used to determine the effects of LULC changes on the watershed’s runoff generation process (Shadeed and Almasri 2010). The fundamental flaw in this technique is that it does not take into consideration rainfall intensity or its variation over time. It does not take into account the impact of nearby soil moisture conditions on runoff creation (Soulis et al. 2009).

The standard method of determining curve numbers using widely available curves and tables is a time-consuming and tedious operation. To get around this limitation, we can use a combination of remote sensing and GIS techniques to figure out the curve number (Dewan and Yamaguchi 2009; Yu et al. 2010; Shadeed and Almasri 2010; Mahmoud 2014; Chatterjee et al. 2016; Pandey and Stuti 2017; Gayathri and Jayalakshmi 2018; Siddi et al. 2018; Matomela et al. 2019). GIS can provide required spatial information for this method for runoff calculation. Many watershed models such as SWAT, EPIC, AGNPS, and CREAMS use the NRCS-CN approach for runoff estimating during rainfall-runoff modelling because of the low data requirements and simplicity of the method (Siddi et al. 2018). The NRCS-CN approach created a curve number that represents the amount of runoff potential of a watershed for a specific rainfall event. LULC, soil type, soil texture, and antecedent soil moisture data can all be used to determine a watershed’s CN (Akinyemi 2017; Singh and Satapathy 2017; Ara and Zakwan 2018). The NRCS-CN method was originally developed for event runoff estimation of small agricultural watershed but can be applied under different climatic conditions. However, different researchers are also using it for larger river basins under different hydro-geological conditions. Napoli et al. (2014) applied this NRCS-CN method for the area of about 885 km2 in central Italy for the purpose of runoff estimation and identified potential rainwater harvesting sites by taking the advantage of GIS technique. Pandey and Stuti (2017) used geospatial technique for estimation of runoff with SCS-CN method in upper South Koel River basin of Jharkhand (India) for the subtropical area of about 772 km2. Singh and Satapathy (2017) used SCS-CN method along with remote sensing and GIS techniques for surface runoff estimation of Kuakhai River basin of India with total area of 453.5 km2. Kumar et al. (2021) used this method for Sind River basin having an area of 26,207.02 km2. Sishah (2021) applied this SCS-CN method for rainfall runoff estimation for a wash river basin of Ethiopia with an area of about 112,000 km2 and correlation between predicted and observed runoff was found to be about 0.92. Koyna River basin is a sub-region comes under upper Krishna basin region of Maharashtra which has experienced unprecedented flood situation which affected social, economic, and environmental conditions of the region (Kale 2009). Due to high water table, changed cropping pattern, unauthorized construction, silt and sand deposition on the river bed, meandering of the river courses, etc., are responsible for increasing runoff/flood magnitude of the study area. Therefore, it becomes very important to predict the runoff/flood potential of the Koyna River (tributary of the Krishna River). NRCS-CN method can be used to estimate runoff potential of the study area using rainfall data for preparing against flood hazard. Therefore, in this study, the applicability of the NRCS-CN method is tested for runoff prediction purpose. By considering all these things, the following goals were set for this research: (1) to detect LULC dynamics for the Koyna River basin of India; (2) to quantify its impact on runoff generation; and (3) to test the applicability of the modified NRCS-CN method for runoff estimation of the Koyna River basin of India using the geospatial techniques.

Material and methods

Study area

The Koyna River basin spans 1917 km2 area in Western Maharashtra, India. The study region spans the latitudes of 17° 7′ 55″ N to 17° 57′ 50″ N as well as the longitudes of 73° 33′ 15″ E to 74° 11′ 10″ E (Fig. 1). During the monsoon season, about 88% of the rain falls. The climate in the basin is subtropical. The basin’s altitude ranges from 534 to 1437 m. This study area comes under Survey of India (SOI) toposheets (1:50,000 scale) 47G/9, 47G/11, 47G/12, 47G/13, 47G/14, 47G/15, 47G/16, 47 K/13, and 47 K/14. The majority of the basin is prone to runoff and soil erosion due to steep sloping conditions and variable topography (Bajirao and Kumar 2021; Bajirao et al. 2021a, b). The study area is subjected to a wide range of climatic and geographical variables. The basin is elongated one and parallel drainage pattern is the dominant drainage pattern in the basin. The study area comes under moderate roughness and unevenness. The study area represents a moderate risk of flood hazard, soil erosion, or mass movement (Bajirao and Kumar 2021). The soil of the upstream area is of light laterite type while black cotton soil is found at the central and downstream area of the study area (Bajirao et al. 2021a). The majority of the people in this basin rely on agriculture to meet their daily needs. The entire basin area is covered by agricultural land, bare land, open forest land, built-up land, dense forest land, and water body.

Fig. 1
figure 1

Location of the study area

Full size image

Data collection

Daily rainfall and runoff data during the monsoon season (1st June to 30th September) were obtained from the Maharashtra state agriculture department and the Central Water Commission (CWC), New Delhi, respectively, for a period of 13 years (from 1999 to 2011). The USGS website http://earthexplorer.usgs.gov/ availed the satellite data which was collected for delineating the basin boundary, preparing the LULC map, and creating slope maps. The toposheets of the Survey of India (SOI) were obtained from India’s Nakshe webpage. The FAO’s global soil map data was downloaded and utilized to extract the study area’s soil map classes.

Quantification of impact of spatio-temporal variability of LULC on runoff generation using modified NRCS-CN method

Surface runoff estimation is necessary for determining the catchment area or different areas of the river basin’s surface water yield potential as well as for planning and managing watershed development projects such as flood prevention work design and water balancing studies. This data is also necessary for prioritizing the development of soil and water conservation treatment systems to promote groundwater recharge and reduce runoff and soil erosion losses. In this study, the spatio-temporal variability of LULC occurring within a 13-year time frame was observed to evaluate the impact of LULC change on runoff generation. The variations in runoff generation from the study area due to spatio-temporal variability of LULC were quantified using the CN values of the year 1999 and 2011 separately.

Preparation of LULC map

The study area was classified using cloud-free satellite images obtained by the Landsat 7 satellite, Enhanced Thematic Mapper (ETM + with path/row 147/48) on November 14, 1999, and November 15, 2011. The ArcGIS 10.2.2 software was used to perform the LULC classification analysis. True colour composite (TCC) and standard false colour composite (FCC) were created for both images following the layer stacking technique. The primary LULC classification was done using the iso cluster unsupervised classification approach. Supervised classification with maximum likelihood method was used to produce the final LULC map of the study area. Visual interpretation techniques based on colour, shape, size, texture, tone, pattern, and association of different LULC phenomena were used in supervised classification for both years. In order to update supervised classification and assess accuracy, high-resolution google map images and Survey of India (SOI) toposheets were also employed. The LULC classes like dense forest, open forest, agriculture, bare soil, built-up, and water body were prepared for both years 1999 and 2011.

Detection of LULC change in percent

To detect spatio-temporal variability of LULC from the year 1999 to 2011, the satellite images were classified for different LULC phenomena. The percentage of individual LULC change that has been occurred from the year 1999 to the year 2011 was detected by using the following formula,

$$\mathrm{LULC change}\left(\mathrm{\%}\right)=\frac{\mathrm{Final LULC area}-\mathrm{Initial LULC area}}{\mathrm{Initial LULC area}}\times 100$$
(1)

where initial LULC area is the area of the selected LULC class in the year 1999 and the final LULC area is the area of the selected LULC class in the year 2011.

Detection of NDVI change

The NDVI index is used to determine the greenness and health of plants (density). The NDVI scale runs from − 1 to + 1 theoretically. The negative value represents the presence of rocks, water, and cloud. Positive values represent the degree of biomass health or biomass density. As the biomass increases, NDVI also increases. The dense forest represents higher NDVI values followed by open forest, sparse agriculture, bare soil, built-up land, and water body. The NDVI map of the study area was created using the Landsat 7 satellite’s Near Infrared (NIR/Band 4) and Red (R/Band 3) bands. NDVI can be expressed as:

$$NDVI=\frac{NIR-R}{NIR+R}$$
(2)

The spatio-temporal variability of LULC is responsible to change NDVI spatially and temporally. As per Bajirao et al. (2018), the natural vegetation condition was divided into four classes namely full vegetation coverage (FVC, 1 ≥ NDVI ≥ 0.9), high vegetation coverage (HVC, 0.9 > NDVI ≥ 0.5), medium vegetation coverage (MVC, 0.5 > NDVI ≥ 0.26), and low vegetation coverage (LVC, 0.26 > NDVI ≥  − 1).

NRCS-CN method for runoff estimation

The water balance approach is used in this method, which is based on two key assumptions (Satheeshkumar et al. 2017). The first is that the ratio of actual infiltration/retention (F) to potential maximum retention (S) equals the real amount of actual runoff (Q) divided by the maximum possible runoff (P − Ia). The second assumption is that the initial abstraction represents a percentage of the maximum possible retention. The empirical relation is stated mathematically as (USDA-SCS 1986):

$$\frac{F}{S}=\frac{Q}{P-{I}_{\mathrm{a}}}$$
(3)

where Ia is the initial abstraction (mm).

The amount of actual infiltration (retention) is defined as:

$$F=P-Q-{I}_{\mathrm{a}}$$
(4)

Combining Eq. (3) and (4), it reduces to:

$$Q=\frac{{(P-{I}_{\mathrm{a}})}^{2}}{(P-{I}_{\mathrm{a}}+S)}$$
(5)

As defined by NRCS, the initial abstraction consists of losses due to interception, infiltration, evaporation, and depression storage that occurs before the runoff begins. Initial abstraction is highly variable but its value highly depends on soil type and land cover. According to SCS (1956), the linear relation between Ia and S is:

$${I}_{\mathrm{a}}=\mathrm{\lambda S}$$
(6)

where λ is the initial abstraction ratio whose values range between 0 and 0.3 and these values have been used in many studies. By combining Eqs. (5) and (6), the expression for Q becomes:

$$Q=\frac{{(P-\lambda S)}^{2}}{(P+(1-\lambda )S)} (\mathrm{for }P >\lambda S)$$
(7)

Under Indian conditions, λ is taken as 0.1 and 0.3 with some constraints of soil type and AMC (Ministry of Agriculture, Govt. of India, Handbook of hydrology 1972). Equation (7) can now be rewritten as follows:

$$Q=\frac{{(P-0.1\mathrm{S})}^{2}}{(P+0.9\mathrm{S})} (\mathrm{for }P >0.1\mathrm{S})$$
(8)

Equation (8) is only valid for black soil under AMC-II and AMC-III conditions and

$$Q=\frac{{(P-0.3\mathrm{S})}^{2}}{(P+0.7\mathrm{S})}(\mathrm{for }P >0.3\mathrm{S})$$
(9)

The Eq. (9) is valid for black soil under AMC-I and all the other types of soils with AMC-I, AMC-II, and AMC-III conditions.

The soil type of the study area is black soil. Hence, Eq. (8) was used under AMC-II and AMC-III conditions while Eq. (9) was used under AMC-I condition. Equations (8) and (9) indicate the form of rainfall-runoff relationship used in the CN method. These equations are used to estimate event runoff depth when the rainfall and S is known. S varies as per antecedent soil moisture condition (AMC) and other variables. S can be obtained from curve number (CN) by using the relationship as:

$$CN=\frac{25400}{254+S}$$
(10)

S varies between zero and infinity whereas CN is a dimensionless number that varies between 0 and 100. A CN of 100 denotes a completely impermeable watershed with no retention, meaning that all rainfall is converted to runoff. A CN of 0 denotes an inverse extreme state in which the watershed abstracts all rainfall without generating any runoff, regardless of rainfall depth. The CN can be chosen based on empirical data. The SCS has created standard curve number value tables depending on catchment LULC conditions and HSGs. These values were taken from the SCS user manual (USDA-SCS 1986) and used in this research.

Hydrologic soil group

Soil textural classes were determined based on the relative proportion of clay, silt, and sand present in the soil. HSG stands for the standard SCS soil classifications, which are classified from HSG A to HSG D by SCS scientists. The hydrologic soil group (HSG) represents the soil’s permeability and surface runoff potential. Low, moderately low, moderately high, and high runoff rates are represented by the soils of HSGs A, B, C, and D, respectively. With the use of SPAW (Soil–Plant-Air–Water) computer software developed by the USDA-ARS, the soil of the study area was divided into distinct HSGs based on soil texture and minimum infiltration rate. Two types of soil texture classes such as clay and clay loam were observed in the study area. The basin’s maximum area was found to be beneath HSG-D, indicating a strong potential for runoff generation due to the low infiltration rate of soil. The area covered by HSG-C and HSG-D was found to be 324.26 km2 and 1592.7 km2, respectively.

Antecedent moisture condition

The amount of runoff is significantly affected by antecedent moisture condition (AMC). Because of its importance, SCS has created a guideline for adjusting a CN according to AMC based on total rainfall in the last five days leading up to a storm occurrence. The SCS-CN approach uses three types of AMC: AMC-II for normal, AMC-I for dry, and AMC-III for wet conditions (Satheeshkumar et al. 2017).

$${CN}_{\mathrm{I}}=\frac{ {\mathrm{CN}}_{\mathrm{II}}}{(2.281-0.01281{\mathrm{CN}}_{\mathrm{II}})}$$
(11)

Equation (11) was used to convert the CN from normal to dry condition (AMC-II to AMC-I).

$${CN}_{III}=\frac{ {\mathrm{CN}}_{\mathrm{II}}}{(0.427+0.00573{\mathrm{CN}}_{\mathrm{II}})}$$
(12)

Equation (12) was used to convert the CN from normal to wet condition (AMC-II to AMC-III), where CNI, CNII, and CNIII are the CN for dry, normal, and wet condition, respectively.

A composite CN must be established for a catchment with varying LULC and soil types by weighting the CN values with their respective sub-areas associated with it based on LULC and HSGs. As a result, the weighted curve number (CNW) was estimated as follows:

$${CN}_{\mathrm{W}}=\frac{ {CN}_{1}{A}_{1}+{CN}_{2}{A}_{2}+\dots +{CN}_{\mathrm{n}}{A}_{\mathrm{n}}}{\mathrm{A}}$$
(13)

where CNn is a CN for 1 to any n number, An is the sub-area for 1 to any n number, and A is the total area.

Modified CNII for slope adjustment

The conventional NRCS-CN method does not consider slope as it is based on an average slope of 5%. However, the land slope represents the important parameter for the movement of water over the surface. Sharpley and Williams (1990) developed slope adjusted curve number (iIIα) considering the importance of slope and it can be expressed in the form as:

$${CN}_{\mathrm{II\alpha }}=\frac{1}{3}\left({CN}_{\mathrm{III}}-{CN}_{\mathrm{II}}\right)\left(1-2 {\mathrm{e}}^{-13.68\mathrm{\alpha }}\right)+{CN}_{\mathrm{II}}$$
(14)

where CNIIα is the slope adjusted CNII and α is the slope in m/m. However, Huang et al. (2006) indicated the limited application of this approach in the field and he had developed a simplified slope modified CNII (CNIIα) equation for slope gradient ranges from 0.14 (14% slope) to 1.4 (140% slope) which is expressed as:

$${CN}_{\mathrm{II\alpha }}={CN}_{\mathrm{II}}\frac{322.79+15.63 (\mathrm{\alpha })}{\mathrm{\alpha }+323.52}$$
(15)

Equation (15) was used in this study for correction of CNII values due to slope variation from 5%. Here, the conventional NRCS-CN method was modified with slope-corrected CN values to improve the prediction performance of conventional NRCS-CN method.

Methodology adopted for runoff computation using modified NRCS-CN method

HSGs map, slope map, LULC map, and selection of AMC to compute daily runoff were the key input data required for the modified NRCS-CN technique. A flowchart showing the approach adopted for computing the runoff is shown in Fig. 2. Initially, all of the necessary data was collected from various sources. The intersect tool in the ArcGIS 10.2.2 software was used to create each LULC and its accompanying HSG polygon. The area of each LULC-HSG polygon was determined and assigned a CN value based on LULC and HSG. The LULC-HSG polygons determined the weighted CN for the entire study area. Finally, the weighted CN was utilized to compute the runoff generated from the entire study area.

Fig. 2
figure 2

Flowchart of runoff computation using modified NRCS-CN method

Full size image

Model verification

The modified NRCS-CN method was verified using some selected low, medium, and high magnitude of runoff events. Total 35 observed and estimated runoff events were used for that purpose. The modified NRCS-CN model was tested by comparing the estimated and observed values of runoff based on time series and scatter plot.

Results and discussion

LULC change detection

The LULC maps showing spatio-temporal variability of different themes from the year 1999 to 2011 are shown in Fig. 3a and b, respectively. Different LULC classes and their areal coverage are presented in Table 1. As presented in Table 1, dense forest area has been decreased by 27.6 km2 (14.72%) from 187.44 km2 (9.78%) to 159.84 km2 (8.34%) which is responsible to reduce initial abstraction losses and groundwater recharge. Open forest area has been increased by 43.69 km2 (14.62%) from 298.91 km2 (15.59%) to 342.60 km2 (17.87%) which is responsible to increase the surface runoff. The agricultural area has been decreased by 335 km2 (31.84%) from 1052.13 km2 (54.88%) to 717.13 km2 (37.41%) which shows the negative effect as the productive land converted into a non-productive one and also responsible to reduce initial abstraction losses and groundwater recharge. All these effects are responsible for increasing bare soil area. Bare soil area has been increased by 287.25 km2 (140.51%) from 204.44 km2 (10.66%) to 491.69 km2 (25.65%) which shows the negative effect as the productive agricultural land converted into non-productive bare land which is also responsible to increase surface runoff and soil loss. The built-up area has been increased by 42.6 km2 (54.90%) from 77.60 km2 (4.05%) to 120.20 km2 (6.27%) which shows the negative effect as the productive agricultural land converted into non-productive built-up land which is also responsible to increase surface runoff. Water body area has been decreased by 10.94 km2 (11.32%) from 96.62 km2 (5.04%) to 85.68 km2 (4.47%) which is responsible to reduce the surface runoff formation as the water body has the highest curve number. The maximum change in the spatial extent was observed in the case of bare soil followed by built-up land, agricultural land, dense forest, open forest, and water body over the time interval of 13 years (Table 1).

Fig. 3
figure 3

LULC map of the Koyna River basin for the year a 1999 b 2011

Full size image
Table 1 Areal extent and temporal variation (1999 to 2011) of different LULC classes in the Koyna River basin
Full size table

NDVI change detection

The decrease in the spatial extent of dense forest or healthy (dense) vegetation is reflected through decreased NDVI from the year 1999 to 2011 (Figs. 4a and b). During the year 1999, the area under LVC, MVC, and HVC was observed to be 697.68, 1081.93, and 137.56 km2, respectively. During the year 2011, the area under LVC, MVC, and HVC was observed to be 999.30, 914.54, and 3.33 km2, respectively. Vegetation coverage has a great impact on runoff generation from the basin. If the vegetation is denser, then abstraction loss is more which is responsible for low runoff generation. In this study, it was found that the area under LVC has been increased which is responsible to reduce abstraction losses and increases runoff loss. Similarly, it was found that medium and high vegetative areas had been decreased significantly which is responsible to increase the runoff amount and reduce the groundwater recharge. As a result, the shift from dense vegetative forest to non-productive low vegetative bare soil and built-up land which indicates the negative effect on natural resource management and development of study area.

Fig. 4
figure 4

Spatial extent of different vegetation coverage grade during year a 1999 b 2011

Full size image

Runoff estimation using modified NRCS-CN method

Curve number analysis

The spatial extent of different LULC-HSG classes is presented (Table 2) and also shown (Figs. 5a and b) for the years 1999 and 2011, respectively. The curve number (CN) values selected under AMC-I (CNI), AMC-II (CNII), and AMC-III (CNIII) using SCS guidelines for different LULC-HSG combinations are presented (Table 2) for both years 1999 and 2011. The weighted curve number (CNW) and potential maximum retention (S) for the whole study area corresponding to different AMC conditions are presented (Table 2) for both the years 1999 and 2011, respectively.

Table 2 Spatial coverage of each LULC-HSG class and its associated curve number under AMC-I, AMC-II, and AMC-III condition for the year 1999 and 2011
Full size table
Fig. 5
figure 5

Spatial distribution of different LULC and its associated HSG classes for the year a 1999 b 2011

Full size image

Slope of the basin

The degree of slope of any watershed directly affects non-linear and complex behaviour of different hydrological processes. Leveled or gentle slope is favourable for runoff infiltration while steep slope is responsible for low infiltration and thus, responsible to increases runoff a drainage basin (Ahmed et al. 2010). Runoff and soil loss are more prevalent in watersheds with steep slopes, which is responsible for increased erodibility of soil and runoff (Bajirao and Kumar 2021). The slope map of the Koyna River basin shows that the basin’s greatest area is covered by a steep slope (Fig. 6).

Fig. 6
figure 6

Slope map of the Koyna River basin

Full size image

Slope correction for curve number (CN α )

The maps for curve number (CNIIα) values corrected for slope under normal condition (AMC-II) for the study area are shown for the years 1999 and 2011, respectively (Figs. 7a and b). The slope-corrected values for dry (CNIα) and wet (CNIIIα) conditions were determined using slope correction formula and these values are also presented for both the year 1999 and 2011 (Table 3). There is an inverse relationship between CN and S. The weighted curve number (CNwα) values increase from dry (AMC-I) to wet (AMC-III) condition. Similarly, in the inverse manner, S values decrease from dry (AMC-I) to wet (AMC-III) condition. From the year 1999 to 2011, during the time interval of 13 years, the CNwα values under dry, normal, and wet condition were observed to be increased from 66.55 to 69.54, 80.65 to 82.34, and 90.23 to 91.05, respectively (Table 3). Similarly, over 13 years of interval, S values were observed to be decreased from 127.65 to 111.26 mm, 60.94 to 54.48 mm, and 27.5 to 24.97 mm for dry, normal, and wet condition, respectively. This change in the values of CNwα and S is responsible to reduce surface infiltration or groundwater recharge and increases the potential of surface runoff generation from the study area. The higher values of the CN indicate the greater runoff potential of the respective area and hence, indicated the suitability of sites for rainwater harvesting.

Fig. 7
figure 7

Slope-corrected curve number (CNIIα) map under normal (AMC-II) condition for the year a 1999 b 2011

Full size image
Table 3 Slope-corrected curve number (CNwα) and potential maximum retention (S) under different AMC condition for the year 1999 and 2011
Full size table

Estimation of potential surface runoff using modified NRCS-CN method

The seasonal runoff values in terms of depth were determined by making the cumulative sum of daily runoff depth over a season for the different years from the year 1999 to 2011. The seasonal runoff values in terms of volume were determined by considering seasonal runoff depth and runoff contributing area. The proportion of runoff generated from the total rainfall in a monsoon season in terms of depth is also determined as the runoff coefficient for each of the years. The estimated runoff volumes using curve number values of years 1999 and 2011 indicate that seasonal rainfall varies from 1253.99 to 3405.48 mm with an average value of 1923.82 mm during 13 year’s time interval (Table 4). Similarly, seasonal runoff depth varies from 354.23 to 2011.7 mm and 375.24 to 2078.56 mm with the CN values of the year 1999 and 2011, respectively. The average runoff volume in the monsoon season was observed to be 1628.22 Mm3 and 1700.68 Mm3 using the CN values of the year 1999 and 2011, respectively. The runoff coefficient varies from 0.28 to 0.59 and 0.29 to 0.61 for the years 1999 and 2011, respectively. The average runoff coefficient was observed to be 0.41 for the CN values of the year 1999 and 0.43 for the CN values of the year 2011. Hence, it was observed that the average runoff coefficient has been increased from 0.41 to 0.43 during 13-year interval which is responsible to increase the runoff generation and decrease the groundwater recharge.

Table 4 Estimated seasonal runoff values using CN values of the year 1999 and 2011 for the Koyna River basin
Full size table

Impact of spatio-temporal variability of LULC on runoff generation

The impact of spatio-temporal variability of LULC on direct runoff generation was assessed using curve number values of the year 1999 and 2011 separately. The increments in seasonal runoff generation were observed to be varied in between 3.32 to 6.36% for different years (Table 5). The maximum increment in runoff generation was found to be in the year 2001 (6.36%) followed by the year 2003 (5.93%). The average of increased surface runoff generation over 13 years (1999–2011) was observed to be 4.85%. The increase in surface runoff generation is due to an increase in CN values over 13 years (1999 to 2011) duration. CN value increased due to spatio-temporal variability of LULC classes. Dense forest area has been reduced and open forest area has been increased which is responsible to reduce infiltration. Built-up and bare soil areas also increased from the year 1999 to 2011 which have the characteristics of low infiltration and high runoff generation. Agricultural land has been decreased over 13 year’s time interval which has a negative effect on infiltration. All the spatio-temporal changes of different LULC phenomena were observed to be responsible to increase the surface runoff generation from the Koyna River basin.

Table 5 Impact of spatio-temporal variability of LULC on surface runoff generation in Koyna River basin
Full size table

Correlation between observed rainfall and estimated runoff

A linear regression between these two variables was used to find the association between observed rainfall and estimated surface runoff. The straight-line linear regression between rainfall and estimated runoff was found to be

$$Y=0.861 X -12.78$$
(16)

where Y is the estimated surface runoff in mm and X is the observed rainfall in mm. Observed rainfall and estimated runoff were found to be closely associated with a coefficient of (R2) of 0.91 (Fig. 8). For regression analysis, a total of 35 rainfall-runoff events were used in this study. The conventional data required for computing discharge using different methods are hardly available at the watershed level for non-gauging watershed. Hence, this regression equation can be used to estimate runoff when only rainfall as an input parameter is known.

Fig. 8
figure 8

Correlation between observed rainfall and estimated runoff

Full size image

Verification of modified NRCS-CN method

The performance of the modified NRCS-CN method for estimation of surface runoff was verified from the comparison of graphical plots by making visual observations in between observed and estimated runoff values. In this study, it was observed that time series and scatter plot of observed and estimated surface runoff using modified NRCS-CN method are approximately in close agreement with each other (Figs. 9, 10). The coefficient of determination (R2) between estimated and observed runoff was observed to be 0.77 which showed its good predictive performance. This method closely models low and medium runoff events. However, this method slightly under-estimating high magnitude of runoff event as depicted in time series and scatter plot.

Fig. 9
figure 9

Time series plot of observed and estimated runoff with modified NRCS-CN method

Full size image
Fig. 10
figure 10

Scatter plot of observed and estimated runoff with modified NRCS-CN method

Full size image

Conclusion

This study revealed that the negative change in the spatial extent of different LULC has been taken place over 13 years duration. In this study, it was observed that the average runoff coefficient has been increased from 0.41 to 0.43 during 13-year interval which is responsible to increase the runoff from the study area and decreases the runoff water available for recharging of the groundwater aquifer. In this study, it was observed that dense forest area decreased by 14.72%, open forest area increased by 14.62%, agricultural area decreased by 31.84%, bare soil area increased by 140.51%, built-up area increased by 54.9%, water body area has been decreased by 11.32% over the time interval of 13 years. The spatio-temporal variability of LULC significantly increased the runoff generation from the study area which would be responsible to reduce groundwater infiltration and increase soil and runoff loss. All the effects are responsible to cause ecological imbalance. The increment in runoff generation is responsible to increase the frequency of flood, drought hazards, etc. The decrease in the NDVI indicates the need for the implementation of forest protective measures and tree plantation programs for sustainable development of the Koyna River basin. The validity of the modified NRCS-CN method was evaluated by considering estimated runoff and observed runoff events. R2 value between estimated runoff and observed runoff events was found to be 0.77 which shows its good predictive performance. Modified NRCS-CN method could be easily employed to detect spatio-temporal variability and its impacts on runoff generation using the geospatial technique. The higher values of the curve number indicate the greater runoff potential of the respective area and hence, indicated the suitability of sites for rainwater harvesting. The future study could be focused to select the location of different water harvesting structures based on stream order, lithology, the permeability of selected sites, etc.

Recommendation

It is highly recommended that this analysis should be carried out by dividing entire Koyna River basin into different sub-basin scale. In this study, FAO soil data has been used but it is also necessary to carry out field survey to identify the soil type of the study area. It is also important to carry out field survey to validating land use/land cover classes to identify drawbacks of remote sensing and GIS techniques if any. This methodology should be applied to other climatic and geological conditions to evaluate its validity.