*2.3. SWAT Model Description*

The SWAT model is a semi-distributed, watershed scale, eco–hydrological model that deals with land–soil–water–plant systems [27]. This model has been tested for a wide variety of watershed and environmental conditions worldwide [46–56]. ArcSWAT jointly developed by USDA Agricultural Research Service (USDA-ARS) and Texas A&M AgriLife Research, was used to spatially link multiple model input data, such as watershed topography (digital elevation model (DEM)), soil, land use, land management, and climatic data. During watershed delineation, the entire watershed was divided into different sub-basins. Then, each sub-basin was discretized into a series of hydrologic response units (HRUs) as the smallest computation unit of a SWAT model, which were characterized by homogeneous soil, land use, and slope combinations. The daily climate input data for defined locations were spatially related to the different sub-basins of the model using a nearest neighbor GIS algorithm. The simulated sediment yield for each HRU was then aggregated and processed to sub-basin level results on a daily time step resolution. The surface runoff computation was performed using modified Soil Conservation Service–Curve Number (SCS–CN) method [57]. Sediment yield levels from each HRU were estimated using the Modified Universal Soil Loss Equation (MUSLE) [58] written as a mass balance equation as follows:

$$\text{SY} = 11.8 \left( \text{Q}\_{\text{surf}} \times \text{q}\_{\text{peak}} \times \text{area}\_{\text{hru}} \right)^{0.56} . \text{K}\_{\text{USLE}.} \text{C}\_{\text{USLE}.} \text{P}\_{\text{USLE}.} \text{LS}\_{\text{USLE}.} \text{CFRG} \text{ }, \tag{1}$$

where SY is the sediment yield (t), Qsurf is the surface runoff (mm ha−1), qpeak is the peak discharge (m3 s<sup>−</sup>1), and areahru is the area of the hydrological response unit (ha). KUSLE (0.013 (t.m2.hr)/(m3.t.cm)), CUSLE, PUSLE, and LSUSLE are the USLE parameters. CFRG is the coarse fragment factor.

The sediment transport capacity of the stream channel is a direct function of the channel peak velocity, which is used in the SWAT model, as shown in Equation (2):

$$\mathbf{T}\_{\text{ch}} = \mathbf{a} \mathbf{v}^{\text{b}},\tag{2}$$

where *T*ch (t m<sup>−</sup>3) is the transport capacity of a channel, *v* (m s<sup>−</sup>1) is the channel peak velocity, and α and b are constant coefficients.

The channel peak velocity was calculated using Manning's formula in a reach segment, as presented in Equation (3):

$$w = \frac{1}{n} R\_{\rm ch} \, ^{2/3}S\_{\rm ch} \, ^{1/2}\prime \,\tag{3}$$

where *n* is Manning's roughness coefficient, *Rch* (m) is hydraulic radius, and *Sch* (m m<sup>−</sup>1) is the channel bed slope.

Channel aggradation (*Sedagg*) and channel degradation (*Seddeg*) in tons were computed in the channel segment using the criteria presented in Equations (4) and (5):

$$\text{if } \text{sed}\_i > \text{T}\_{\text{ch}} : \text{Sed}\_{\text{h}\text{gg}} = (\text{sed}\_i - \text{T}\_{\text{ch}}) \times V\_{\text{ch}} \text{ & } \text{Sed}\_{\text{deg}} = 0,\tag{4}$$

$$\text{if } T\_{\text{ch}} > \text{sed}\_i: \quad \text{Sed}\_{\text{deg}} = (T\_{\text{ch}} - \text{sed}\_i) \times V\_{\text{ch}} \times K\_{\text{ch}} \times \mathbb{C}\_{\text{ch}} \quad \text{\& } \text{Sed}\_{\text{agg}} = 0,\tag{5}$$

where *sedi* (t m−3) is the initial concentration of sediment, *Cch* is the channel cover factor, *Kch* is the channel erodibility factor, and *Vch* (m3) is the channel segment water volume.

*Sedout* (t) is the total sediment transported out of the channel segment, which was computed using Equation (6):

$$\text{Sed}\_{\text{out}} = \left(\text{sed}\_{i} + \text{Sed}\_{\text{dcy}} - \text{Sed}\_{\text{ag}\text{g}}\right) \times \frac{V\_{\text{out}}}{V\_{\text{ch}}},\tag{6}$$

where *Vout* (m3) is the volume of water leaving the channel segment at each time step, *sedi* (t) is the sediment inflow concentration at each time step, *Sedagg* (t) is the channel aggradation, *Seddeg* (t) is channel degradation at each time step, and *Vch* (m3) is volume of channel segment water at each time step.

Soil erosion is a direct function of the slope length and steepness, and it increases due to increases in shear stress. Thus, a major influence of the slope on erosion appears to be exerted through its impact on runoff velocity, and the sediment transport capacity of runoff increases with the increasing flow velocity.

#### *2.4. SWAT Model Input and Setup*

The requisite spatial data (DEM, land use, and soil data) and temporal data (rainfall and temperature) were prepared for the SWAT model setup. A physical topographical survey of the watersheds was conducted using a GPS. The DEM of each watershed was generated using point-source elevation data in a geographic information system by applying the inverse distance weighting (IDW) method, as shown in Figure 4. The winter wheat land use classification was used according to cropping practice, and the soil type was sandy loam for all small watersheds based upon the soil textural analysis. The daily precipitation and temperature data were collected from the SAWCRI Chakwal for six years from January 2010 to April 2015.

After the preparation of the requisite data file for model input, ArcSWAT9.3 was used to automatically delineate sub-watersheds and to generate a stream network based on the DEM. An appropriate database of sub-basin parameters and a comprehensive topographic report of the watersheds were generated. The sub-watersheds topographic report was rechecked for area, slope, location of outlet, and soil textural class according to the physical characteristics to make appropriate database changes. SWAT coding conventions were used to reclassify the land use and soil maps into HRUs based on the unique land use, soil class, and slope class in the overlaying section.

The weather station location and lookup tables of daily precipitation and temperature (maximum and minimum) data were loaded to link them with the required files. First, the model was simulated for each watershed with validated parameters adopted from Hussain et al. [14] without the consideration of the conservation structures, and then interventions of the soil and water conservation structures were made by modifying the parameters for surface runoff and sediment yield. The setup of model for each watershed with and without the consideration of the conservation structures is shown as the right and left of Figure 4, respectively.

The locations of each soil and water conservation structure were marked and used for the correct delineations of sub-basins. The demarcated sub-basins indicated the boundary of the agriculture fields, while the structures were the outlet of each field in model setup when the conservation structures were considered. The ideal factors that describe the effect of stone bunds are the USLE support practice factor (P-factor), the curve number, and the average slope length for the sub-basin (SLSUBBSN).

The SLSSUBSN value was modified by editing the HRU (.hru) input table, whereas the P-factor and curve number values were modified by editing the Management (.mgt) input table. Three more parameters were modified, namely the average slope steepness (HRU\_SLP) of the HRU input tables and two basin parameters (SPCON and SPEXP) representing the general watershed attributes in the Basin (.bsn) input files. SPCON and SPEXP are linear and exponential channel sediment routing factors, respectively, that affect the movement and separation of sediment fractions in the channel and were used to calculate the maximum amount of sediment re-entrained during channel sediment routing.

**Figure 4.** *Cont*.

(**d**)

**Figure 4.** *Cont*.

**Figure 4.** Topographic maps of selected small watersheds in the Chakwal and Attock Districts for model application: (**a**) Khokar Bala watershed, (**b**) Khandoya watershed, (**c**) Khaliq Gully watershed, (**d**) Ashraf Gully watershed, (**e**) Chak Khushi watershed, (**f**) Dhoke Dhamal watershed, and (**g**) Dhoke Hafiz Abad watershed.

#### *2.5. Model Calibration and Validation—Reference to the Previous Study*

In this study, the calibrated and validated model was adopted from a previous study [14]. The calibrated parameters were directly used during the simulation of the SWAT model without the consideration of the soil and water conservation structures. Hussain et al. [14] successfully performed the calibration of soil erosion parameters in small watersheds of the Dhrabi River Catchment. In this study, the Catchment-25 parameters were selected, as shown in Table 2 [14]. Catchment-25, having an area of 2.0 ha, is an agricultural watershed consisting of deep gullies, and its average land slope is 10.5%. It has well-defined boundaries and wide gully beds that mimic the full representation of the other study watersheds. The detailed description of Catchment-25 and SWAT model calibration and validation procedure and performance can be seen in the study of Hussain et al. [14].

The SAWCRI collected the surface runoff and sediment yield data at the outlet of Catchment-25. The experimental setup for data collection is shown in Figure 5. The automatic rain gauge and water level recorder were installed for rainfall and runoff depth measurements. The runoff discharge measurement was done using a sharp crested rectangular weir. The settling basin was used for sediment collection. The stilling basin was 3 m wide, 4 m long, and 65 cm deep at the weir and 15 cm deep upstream, in order to trap coarse sediment as bed load, while the suspended load was collected separately in 20 liter plastic buckets covered with a plastic sheet.

The total sediment yield of the catchment for a particular event is the sum of the bed load and suspended sediment. Coarser sediments were trapped in the stilling basin during the runoff event. After each runoff event, the standing water from the stilling basin was drained off through the drainpipe, and the wet sediments were collected and weighed. A composite sample of the wet bed load was obtained after mixing six-to-seven sub-samples collected throughout the stilling basin and oven dried to determine the moisture contents. The moisture contents were deducted from the wet weight to determine the dry weight of the sediment. Finer sediments in the runoff water passing the weir were sampled using vertical sampling tubes with holes. Following the runoff events, the samples present in the container were collected and analyzed. The total suspended sediment loss from the catchment was obtained by multiplying the sediment concentration in the bucket with the runoff volume passing over the weir.

**Figure 5.** The experimental setup for runoff and sediment yield [14].
