1) Land use and land cover and Soil Map

Landsat 8 ETM+ with a 30 m resolution for Path/Row: 145/043 and 145/044

retrieved on 6 March 2015 was used. The images were retrieved from the USGS’ EarthExplorer site (https://earthexplorer.usgs.gov/). Landsat 8 ETM+ satellite images with optical bands with the standard false-color combinations were used to prepare the land use and land cover map of the basin, for which subsequent ground truth verification was carried out through extensive field visits. For the land use land cover, supervised classification using maximum likelihood classifier was applied. Land use and land cover spatial data were reclassified into SWAT land cover/crop types.

Soil toposheets with sheet numbers of 1, 2, 4, and 5 prepared by the Indian Council of Agricultural Research-National Bureau of Soil Survey & Land Use Planning (ICAR-NBSS & LUP) on a scale of 1:250,000 and printed on a scale of 1:500,000 were used to prepare soil map of the watershed. The soil map was linked with the user soil database. A user lookup table was created that identifies the SWAT code for the different categories of soil and land use a land cover on the map as per the required format.

2) Aster Global Digital Elevation Model Version 002 (ASTER GDEM v2)

The digital elevation model of 30 m spatial resolution was downloaded from the EarthExplorer website (https://earthexplorer.usgs.gov/) and used to delineate the watershed and to analyze the drainage patterns of the land surface terrain.

Daily streamflow, precipitation, air temperature (maximum and minimum), relative humidity, average wind speed and solar radiation from the Bina basin were used for the period 1989-1996. These data were collected from Madhya Pradesh State Data Center (MPSDC), Bhopal. The daily weather data and weather generator location (wgnloc) were prepared into a separate excel sheet and converted into .dbf format using Microsoft access before imported into the model setup. The model was set up with a two-year warmup period. Model calibration was conducted using data from 1991 to 1993 while data from 1994 to 1996 was used for validation.

All spatial data inputs (DEM, land use the land cover map, and soil map) were projected to Universal Transverse Mercator (UTM) 43 North and World Geodetic System (WGS) 84 datum and were resampled to a 30-pixel size using Bilinearresampling technique in ArcGIS 10.5.1. The Soil and Water Assessment Tool (Arc SWAT 2012.10.4.21) interfaced with Arc GIS 10.5.1. The software was used to derive catchment characteristics (e.g., watershed boundary, drainage area, slope, flow path, etc.) solely based on the spatial data inputs [13]. Watershed delineation and spatial arrangement of basin elements (e.g. sub-basin, reach segments and point sources) were defined [14]. The stream drainage lines were created using threshold stream cells of 348,395. The most popular setup was the sub-basin configuration, where the basin is divided into sub-basin and further sub-divided into hydrologic response units (HRUs) [15]. The minimum

threshold area of 5/5/5 [%] for land use/soil class/slope over the sub-basin area was used. The land use, soil and slopes percentage areas covering below the minimum threshold area were excluded, and then the remaining area was redefined so that 100% of the sub-basin area could be used in the simulation.

HRUs represent the smallest unit areas within the watershed with similar soil, topography, and land-use class [16]. In this study, HRUs definition was done based on eight classes of soil and eight classes of land use and land cover categories, and multiple slope discretization with three slope classes [<15%, 15% - 30% and >30%].

Land use and land cover map were reclassified into SWAT land cover/plant types [17]. Land use and land cover (LULC) of the basin was classified into eight classes and the final land use classes were decided to be assigned as, agriculture land-generic, barren land, current fallow, forest-deciduous, forest-evergreen, sandy area, urban area, and water body (Figure 2). Similarly, the basin’s soil was categorized into eight classes (Figure 3).

The Soil Conservation Service Curve Numbers (SCS-CN) were determined based on the USDA National Engineering Handbook [18] [19] [20]. The SCS-CN is a function of the soil permeability, land use, and antecedent soil water conditions. The SCS-CN method is an approach that is used in rainfall-runoff modeling to compute direct runoff. This method assumes an initial abstraction (I_a) before ponding, which is related to the SCS-CN. SCS-CN defines three antecedent moisture conditions: I—dry (wilting point), II—average moisture and III—wet (field capacity) [21]. The SCS-CN method, in SWAT, relates runoff to soil type, land use, and management. The SCS-CN method is based on the principle of water balance and two fundamental assumptions [22].

The first assumption is that the ratio of direct runoff to potential maximum runoff is equal to the ratio of infiltration to potential maximum retention. The second assumption states that the initial abstraction is proportional to the

potential maximum retention. The water balance equation and the two assumptions are expressed mathematically [23] :

P = I a + F + Q (1)

Q P − I a = F / S (2)

I a = λ S (3)

where P is the total precipitation (mm), Ia is the initial abstraction before runoff (mm), F is the cumulative infiltration after runoff begins (mm), Q is direct runoff (mm), S is the potential maximum retention (mm), and λ is the initial abstraction coefficient. The combination of Equations (1) and (2) leads to the popular form of the original SCS-CN method [24] :

Q = ( P − I a ) 2 ( P − I a ) + S       when   P > I a Q = 0 ,     when   P ≤ I a (4)

S = 25400 C N − 254 (5)

where the CN is a dimensionless variable, ranging from 0 to 100 and it depends on land use, hydrological soil group, hydrologic conditions, and antecedent moisture conditions [25]. This increases accuracy and gives a much better physical description of the water balance. The hydrologic cycle as simulated by SWAT is based on the water balance equation [26] :

S W t = S W o + ( R d a y − Q s u r f − E a − W s e e p − Q g w ) (6)

where SWt is soil H₂O content (mm) at time t in days, SWo is the initial soil H₂O content (mm), R_day is amount of rainfall on day i (mm), Q_surf is the amount of surface runoff on day i (mm), E_a is the amount of evapotranspiration on day i (mm), W_seep is the amount of percolation and bypass exiting the soil profile bottom on day i (mm), Q_gw is the amount of return flow on day i (mm).

The SWAT model was run both on a daily and monthly timesteps. A two-year model warm-up period (1989 and 1990) was used. Model sensitivity analysis, model calibration and validation were done using the SWAT-CUP tool. Eighteen parameters were considered and tested for the model parameterization and sensitivity analysis. The model uncertainties have been tested and analyzed using the SUFI-2 uncertainty analysis procedure in SWAT-CUP [27] [28].

The parameters were related to stream-flow assessment and include viz. r_CN2.mgt (curve number), v__ALPHA_BF.gw (base flow alfa factor), v__GW_DELAY.gw (groundwater delay time), v__GWQMN.gw (threshold depth of water in shallow aquifer required for return flow), v__GW_REVAP.gw (groundwater ‘revap’ coefficient), v__ESCO.hru (soil evaporation compensation factor), v__CH_N2.rte (manning roughness for the main channel), v__CH_K2.rte (effective hydraulic conductivity in main conductivity), r__SOL_ AWC.sol (soil available water capacity), r__SOL_K.sol (soil hydraulic conductivity) v_RCHRG_DP.gw (Deep aquifer percolation fraction), r_SOL_BD.sol (Moist bulk density), r_SOL_Z.sol (Depth from the soil surface to bottom of the layer), r_SLSUBBSN.hru (Average slope length), r_OV_N.hru (Manning's "n" value for overland flow), CANMX.hru (Maximum canopy storage), v_EPCO.hru (Plant uptake compensation factor), v_SURLAG.bsn (Surface runoff lag time) have been considered for model parameterization and calibration and validation process.

A variety of verification criteria that could be used for the evaluation of models were proposed by the World Meteorological Organization (WMO) and other investigators [29]. Model evaluation was conducted using selected statistical evaluation metrics. The following model evaluation techniques were chosen to check the performance of the SWAT model. Moriasi et al. [30] recommended the use of the coefficient of determination (R²) together with the Nash-Sutcliffe model efficiency coefficient (NSE) to evaluate the performance of the SWAT model. The R² (Equation (7)) value is a measure of the strength of the linear correlation between the predicted and observed values. The NSE (Equation (8)) is a measure of the predictive power of the model and is the most frequently used method for hydrological applications [31].

R 2 = ∑ i = 1 n ( O i − O ) ( P i − P ¯ ) ∑ i = 1 n ( O i − O ¯ ) 2 ∑ i = 1 n ( P i − P ¯ ) 2 (7)

NSE = 1 − ∑ i = 1 n ( O i − P i ) 2 ∑ i = 1 n ( O i − O ¯ ) 2 (8)

where O_i is i^th observed streamflow; O ¯ is mean observed streamflow; P_i is i^th predicted streamflow and; P ¯ is mean predicted streamflow values and, n is the total number of observations.

An NSE value of 1 indicates a perfect match between simulated and observed data. A value of 1 for the R² also indicates a perfect linear correlation between simulated and observed data. In addition, Percent bias (PBIAS, Equation (9)), which measures the average tendency of the simulated data to be larger or smaller than their observed counterparts, was used in this study. The optimal value of PBIAS is 0.0, which indicates accurate model simulation. Positive PBIAS values indicate model underestimation bias, and negative values indicate model overestimation bias [30].

PBIAS = ∑ i = 1 n ( O i − P i ) ∗ ( 100 ) ∑ i = 1 n   O i (9)

The global sensitivity of streamflow parameters has been estimated by calculating multiple regression system, which regresses the Latin hypercube generated parameters against the objective function values. The t-stat and p-value are two factors commonly used to evaluate the sensitivity of model parameters in SWAT-CUP. The t-stat provides a measure of sensitivity as its absolute values go larger while the p-values determine the significance of the sensitivity magnitudes with close to zero value as more significant [32].

The elevation of the basin ranges from 380 - 710 m. Among the land use and soil type classes, Forest-Deciduous and Clay Soil-Moderately Well Drained-Deep (CS-MWD-D) were dominant in the catchment, respectively (Table 1). Most (98%) of the catchment area has a flat to the moderate slope (0% - 15%). The catchment was divided into four sub-basins and classified into 68 HRUs (Table 2). The HRUs of this basin have been classified into different classes mainly based on land use land cover, soil type, and slope.

The catchment has an average CN of 83.3 (Table 3). Higher CN indicates greater run-off potential. Curve number is governed by land use, hydrological soil group, hydrologic conditions, and antecedent moisture conditions which depend on the average slope of the basin.

The SWAT model has over 30 parameters. Arnold et al. [33] categorized SWAT model parameters by the process such as surface runoff, baseflow, sediment and

nutrient and pesticide using the report of input parameters in SWAT model calibration for 64 selected watershed studies.

In this study, following a comprehensive literature review, 18 parameters were selected for model simulation on daily and monthly timesteps. The parameters primarily represented the channel, runoff and soil processes. The initial value ranges used for these selected parameters are shown in Table 4. It was observed that using the fitted parameters and their appropriate initial range had a significant effect on the streamflow simulation process.

There are mainly two approaches to analyze the sensitivity of model parameters: local sensitivity analysis and global sensitivity analysis. The local sensitivity analysis is a one-at-a-time (OAT) technique that analyses the impact of a single parameter at a time, keeping the other parameters fixed [9]. The global sensitivity of model parameters has been estimated by calculating the multiple regression system, which regresses the Latin hypercube generated parameters against the objective function values [32]. In the present study, the most sensitive parameters observed after global sensitivity analysis for daily and monthly calibration in SUFI-2 are shown in Table 5 and Table 6, respectively. Results showed that r_SOL_BD.sol (moist bulk density), v__ALPHA_BF.gw (base flow alfa factor) and v__CH_N2.rte (Manning roughness for the main channel) for a daily basis and r__SOL_AWC.sol (soil available water capacity), r_SOL_Z.sol (Depth from the soil surface to bottom of the layer) and r_CN2.mgt (curve number) for monthly simulations were found the most sensitive model parameters. It was experienced that the streamflow simulations process was not affected by parameters that are relatively insensitive compared to sensitive parameters and changes in their range had not caused significant changes in the model result.

V = Replaced by value, R = (1 + multiply by value (%)) and A = added on value.

These results were supported by various authors i.e. Singh et al. [34] calibrated SWAT model for Tungabhadra River and found CH_K2, SOL_K, CN2, ALPHA_BF, ALPHA_BNK as most sensitive parameters. Setegn et al. [35] simulated streamflow using the SWAT model in the Lake Tana Basin, in their study, they have evaluated the relative sensitivity of the Nineteen parameters and found that soil evaporation compensation factor (ESCO), initial SCS Curve Number II value (CN2) and base flow alpha-factor (Alpha_Bf) [days] were the most sensitive parameters. Himanshu et al. [36] indicated that a total of 27 sensitive parameters were considered collectively for runoff and sediment, and their rank was determined according to sensitivity to the output. Sensitivity analysis shows that curve number (CN2) and effective hydraulic conductivity (Ch_K2) are the most sensitive model parameters for both runoff and sediment yield computations. Soil evaporation compensation factor (Esco), an available water capacity of soil layer (Sol_Awc), depth from the soil surface to bottom of (Sol_Z) are relatively more sensitive to runoff but less to sediment. Hosseini et al. [10] applied the SWAT model for the runoff estimation in a Taleghan basin and found that the Baseflow alpha factors (ALPHA_BF) followed by Snowfall temperature (SFTMP) and Groundwater delay time (GW_DELAY) are more sensitive parameters.

Overall, the SWAT model performed “satisfactorily” during daily simulations while during the monthly simulation the model performed “very good”. The PBIAS for both daily and monthly time periods was in the acceptable range; with 2.2% and 18% for calibration and 4.5% and 3.9% for validation, respectively [30]. The coefficients of determination (R²) of calibration for the daily and monthly data were 0.66 and 0.96, respectively. The R² value of both daily and monthly timescales shows there is a good correlation between the observed and simulated flows [30]. However, it was clear that the model’s performance significantly improved with monthly simulations. Similarly, NSE values for monthly simulations both during calibration and validation showed significant improvements compared to respective daily simulations. These are related studies that could support our results: Jain and Sharma. [37] found that the SWAT model could be employed for simulation of runoff and sediment yield behavior of the Vamsadhara river basin. Srinivas G and Naik [38] reported that the SWAT model gave good correlation during daily simulation results and a very good correlation for monthly time series at the Musi river basin. Jain et al. [39] reported that the SWAT model was calibrated and validated with daily streamflow data and the results were indicated a good simulation of streamflow of the Himalayan mountain basin.

The R² and NS coefficient are two important statistical analyses for the evaluation of the results. According to Santhi et al. [40], when R² equals 1, the regression equation model considered as a perfect fit model, while an R² value less than 0.5 (near to zero), suggests that the model is not suitable. The strength of the model calibration and uncertainty procedure was also analyzed using the R-factor. The R-factor values were estimated as 0.22 and 0.52 for the daily and monthly simulations, respectively (Table 7). The best model parameters and their value ranges for both daily and monthly model simulations are presented in Table 8 and Table 9. In addition, observed and simulated time series daily and monthly streamflow were plotted for visual comparison to explore how the model performs during peak and low flows (Figures 4-11).