Evaluating the Performance of HEC-HMS and SWAT Hydrological Models in Simulating the Rainfall-Runoff Process for Data Scarce Region of Ethiopian Rift Valley Lake Basin

A number of physically-based and distributed watershed models have been developed to model the hydrology of the watershed. For a specific watershed, selecting the most suitable hydrological model is necessary to obtain good simulated results. In this study, two hydrologic models, Soil and Water Assessment Tool (SWAT) and Hydrological Engineering Centre-The Hydrologic Modeling System (HEC-HMS), were applied to predict streamflow in Katar River basin, Ethiopia. The performances of these two models were compared in order to select the right model for the study basin. Both models were calibrated and validated with stream flow data of 11 years (1990-2000) and 7 years (2001-2007) respectively. Nash-Sutcliffe Error (NSE) and Coefficient of Determination (R 2 ) were used to evaluate efficiency of the models. The results of calibration and validation indicated that, for river basin Katar, both models could simulate fairly well the streamflow. SWAT gave the model performance with the R 2 > 0.78 and NSE > 0.67; and the HEC-HMS model provided the model performance with the R 2 > 0.87 and NSE > 0.73. Hence, the simulated streamflow given by the HEC-HMS model is more satisfactory than that provided by the SWAT model.

the Canagagigue Watershed simulated by MIKESHE was more accurate than that simulated by SWAT and APEX model, for both calibration and validation periods. In Virginia, Polecat Creek watershed, SWAT and HSPF were evaluated and reported as both models are able to simulate effectively the streamflow [23].
From these studies we can concluded that the models' performances are very site specific, and no one model is superior under all hydrologic conditions. Therefore, a complete understanding of comparative model performance requires applications under different hydrologic conditions and watershed scales. As reported by [6], SWAT has been successfully applied to simulate streamflow in different basins of the country Ethiopia and HEC-HEM has been also successfully applied for hydrological studies in different Ethiopian basins [14]. Therefore, the objective of the present study is to compare and assess the suitability of those two widely-used watershed simulation models, namely HEC-HMS and SWAT, for simulating the hydrology of a Katar River Basin, the Central Rift Valley (CRV) Lake Basin of Ethiopia.

Description of Study Area
The study area, Katar River Basin is sub catchment of Ethiopian CRV Basin. The Katar River and its tributaries start from the eastern parts of mountains Chilalo, Galema and Kakka of Arsi Zone and drains to Lake Ziway. Topographically, the Katar River Basin shows variation with altitude ranging from around 1617m near Abura (at gauging Station) to about 4211 m above mean sea level on the high volcanic ridges along the eastern basin. Geographically, the basin lies between 7˚21'34" to 8˚9'55" North latitudes and 38˚53'57" to 39˚24'46" East longitudes and has a total surface area of 3354 km 2 ( Figure 1).
Regards to climate, the River basin is characterized by dry to sub-humid climate with mean annual precipitation of 650 to 1200 mm and temperature of the basin varies between 15˚C and 25˚C [24]. The rainfall pattern of the basin is largely influenced by the annual oscillation of the inter-tropical convergence zone. As the result [25] has classified the basin in three main seasons namely warm period (small rainy season) which extends from March to May, wet summers or longest rain season (Jun to September) and dry period (October-February).  [26]. These are Andosols, Cambisols, Fluvisols Leptosols, Luvisols and Vertisol.
In the basin, there is dynamic land use change [26] [27] [28] and this has Open Journal of Modern Hydrology extended to cultivate the marginal lands and for poor land management practices [26]. Inside the basin, land degradation, soil erosion and declining of soil fertility are seen due to cultivation of more marginal lands and improper land management systems [4] [26].

Meteorological Data
The meteorological data included daily precipitation, maximum and minimum temperature, daily wind speed, daily sunshine hours and daily relative humidity, and they were obtained from eight meteorological stations available within and nearby the study area. The collected data was appropriately adjusted for inconsistency, corrected for errors, extended for insufficient, and filled for missing. Daily data of 31 years (1987-2019) were collected for the study.

Stream Flow
For both SWAT and HEC-HMS models, discharge data were also required for calibration and validation of streamflow. On the Katar River Basin, Abura is the terminal gauging station and the runoff data of the station was collected from the Ethiopian Ministry of Water, Irrigation and Electricity.

Soil Data
In the year 2010, Ethiopian Rift Valley Lake Basin Master Plan study was con-ducted and, in that study, the soil samples were collected from all soil units of the basin. In Ethiopia Water Works Design and Supervision Enterprise soil analyses laboratory, all physical and chemical parameters of those soils were tested. Hence, from Ethiopia Ministry of Water Irrigation and Electricity (MoWIE), the analyzed soil laboratory results and the soil maps of the basin were collected. The soil properties required to set up the models namely soil texture, soil saturated hydraulic conductivity, grain size percentage, bulk density, texture class, soil available water and others were obtained from the analyzed laboratory results.

DEM
Digital elevation model is the basic input data for any hydrological modelling study, including SWAT and HEC-HMS. A 30 m × 30 m digital elevation model (DEM) was obtained from shuttle radar topography mission (SRTM) of https://earthexplorer.usgs.gov/. From the collected DEM ( Figure 2), the relevant basic information of the basin like physiographic characteristics of the catchment, including elevation and slope is extract in ArcGIS program.

Land Use Land Cover
To develop the SWAT and HEC-HMS model, one of the basic input data is land use cover map of the study area. For model calibration and validation, researchers are using a different years metrological and stream flow data with a fixed land use land cover map. Practically, the land use land cover pattern of the study area will also changes with season. Unless we assume a constant land use type inside the study area, using a single land use map for model calibration and validation may be erroneous. In this study, the models are going to be calibrated and validated for the years 1990 to 2000 and 2001 to 2007 respectively. Hence the stream flow, metrological and Land use land cover data should be prepared for each calibration and validation periods. To do so, the land use land cover map of the study basin is prepared for the years 1990 and 2001 from Land sat images by downloading from United States Geological Survey website (https:earthexplorer.usgs.gov) ( Figure 3). The date from where the image was downloaded is given in Table 1 and they are cloud free and almost similar in months of the date.
The image is available in the form of GCS_WGS_1984 raster form with 30 m × 30 m resolutions. Preprocessing such as layer stacking, mosaic king and band color combination are carried out in order to Ortho-rectify the images. The images are process using GIS software. The developed land use and land cover image  For generated land used land cover (LULC) map of the study basin, the accuracy assessment was done by comparing the classification product with the reference data, which accurately reflects the true land cover. The accuracy assessment reflects the difference between the classified data and the referenced data. The most common way is to represent the classification accuracy of remotely sensed data Kappa coefficient [29]. The Kappa coefficient was calculated according to the formula given by [29]: where "r" is the number of rows in the matrix, xii is the number of observations in row i and column i, xi+ and x + i are the marginal totals of row i and column i, respectively, and N is the total number of observations. After the critical evaluations given above, the exact study area's land use land cover was clipped from projected land use map. After reclassification, the classified land use raster map was converted to land use shape file maps using raster to polygon function. The developed land use land cover map was merged with soil data for curve number generation and extraction of area coverage for each class.
The LU/LC map of the study area was coded to the SWAT four letter codes and linked to the SWAT land use database. Hence, after preparing the look up table the land use types were made compatible with the input required by the model.

SWAT Model
The Soil and Water Assessment Tool (SWAT) is a physical based model used to estimate the runoff, sediment and chemical yields in gauged and un-gauged basins [30]. The hydrologic cycle of a basin simulated by SWAT is based on the following water balance equation: where: SW t is final soil water content (mm), SW 0 is initial soil water content (mm), t is time (days), R day is amount of precipitation (mm), Q surf is amount of surface runoff (mm), E a is amount of evapotranspiration (mm), W seep is amount of water entering the vadose zone from the soil profile (mm) and Q gw is amount of return flow (mm).
In SWAT, the surface runoff from daily rainfall is estimated using a modified SCS curve number method, which estimates the amount of runoff based on local land use, soil type, and antecedent moisture conditions [31]. The surface runoff component of the water balance is determined from the SCS method as: where, I a = 0.2S and S = 25.4 (1000/CN − 10); hence the amount of surface runoff is equated as: where: I is initial abstraction (mm), S is relation parameter (mm) and CN is curve number. which are unique combinations of land use, soil type and slope ( Figure 4).

HEC-HMS Mode
HEC-HMS is hydrologic modeling software developed by the US Army Corps of Engineers-Hydrologic Engineering Center (HEC) and is designed to simulate the rainfall-runoff processes in a wide range of geographic areas such as large river basin water supply and flood hydrology to small urban and natural watershed runoff [14]. HEC-HMS model setup consists of a basin model, meteorological model, control specifications, and input data (time series data). Hence to prepare the input files for HEC-HMS, the DEM of the study area were processed Open Journal of Modern Hydrology using the geographic information system (GIS) interface of the HECGeoHMS model in an ArcGIS programme. Terrain pre-processing and basin processing tools were used to generate basin characteristic parameters and input files for HEC-HMS including a stream network, subbasin boundaries, and the connectivity of various hydrologic elements. The basin model and basin features which derived from HEC-GeoHMS were taken as a background map file and imported to HEC-HMS ( Figure 5). In the meteorological model, the Priestley-Taylor evapotranspiration method was used for hydrological simulation and the sub basins areal precipitation was estimated by Thiessen polygon method with respect to the centroid point of each sub-basin.
The system encompasses runoff volume by computing the volume of water that is intercepted, evaporated, infiltrated, stored, and subtracting it from the total precipitation. In HEC-HMS model, around eleven kinds of loss estimation methods are embedded in the programme. In this study, Soil Curve Number (SCS Curve Number) method was used.

Model Evaluation and Statistical Analyses
The Nash-Sutcliffe efficiency (NSE) and coefficient of determination (R 2 ) were used as statistical indices to assess the model performance. R 2 (the coefficient of determination) indicates the degree of linear relationship between simulated and observed data. A R 2 value close to one indicates a better performance. However, it is very sensitive to extremely high values. The Nash-Sutcliffe efficiency (NSE) is one of the most commonly used criteria [32]. This is a normalized statistic, which can be used to determine the goodness of fit. The NSE ranges from −∞ to 1, with 1 indicating a perfect match [33].

Sensitivity Analysis
Out of the 16 parameters, only few parameters that are most sensitive for flows were determined based on p-test and t-stat. A t-stat determines the relative significance of each parameter and ranks the parameter based on the absolute values. To obtain optimal fitting with the measured data, calibration was conducted manually and automatically by SUFI-2 program. The most sensitive parameters identified during calibration are SCS runoff curve number (CN2.mgt), effective hydraulic conductivity in main channel (mm/hr) (CH_K2.rte), Saturated hydraulic conductivity (mm/hr) (SOL_K.sol), Ground water "revap" coefficient (GW_REVAP.gw), available water capacity of the soil layer (mm H 2 O/mm soil) (SOL_AWC.sol), surface runoff lag time (SURLAG.bsn), slope of watershed (HRU_SLP.hru), base flow alpha factor (days) (ALPHA_BF.gw) and groundwater delay (days) (GW_DELAY.gw). The sensitivity result shows the SCS runoff curve number (CN2.mgt) as a major critical parameter for SWAT model. For HEC-HMS model, SCS (unit hydrograph lag time, curve number, initial abstraction), Muskingum (k and x) and sub reaches as a major critica parameter with different sensitivity ranks ( Table 2).

HEC-HMS Model Simulation Result
The calibration and validation of the HEC-HMS model for the Katar River Basin were carried out by comparing the simulated streamflow with the observed flow at main gauging station (Abura station).
The graphical representation of measured and simulated flows matched flows matched well for both calibration and validation periods ( Figure 6 and Figure  7). This shows that the model produced a similar trend between observed and simulated streamflow during the calibration and validation periods. However, total runoff volume and peak flow were slightly overestimated. This difference might be due to the routing coefficients or simulating at an hourly time step. Based on the calibration and validation results, the model performance has been evaluated in terms of the statistical indicators (Table 3).    As recommended by [33], the high R 2 and NSE values indicate the very good correlation and agreement between the observed and simulated values. Based with Reference [33] rated vales, the model performance statistics determined for the model HE-HMS (Table 3) is good to very good and hence the model HEC-HMS is a suitable model for hydrological related studies for the river basin Katar.

SWAT Model Simulation Result
The calibration and validation of the SWAT model for the Katar River basin were also conducted by comparing the simulated streamflow with the observed flow at the Abura gauging station. The plots of observed and simulated daily flow are shown in Figure 9 and Figure 10.
The calibration and validation greatly improved the agreement between the measured and simulated daily discharges. The graphical representation of observed and predicted monthly stream flows matched well for both calibration and validation periods (Figure 9 and Figure 10). The goodness-of-fit test statistics for calibration and validation periods are sown in Table 4. As shown in Table 4, all the numerical model performance indicators designated for the model performance evaluation are concordant with NSE and R 2 values ranging 0.69 to 0.67 and 0.8 to 0.78 respectively, for the calibration and validation periods.     SWAT developers recommend an acceptable calibration for hydrology at a R 2 > 0.6 and NSE > 0.5 [34]. Similarly, its performance is good to very good compared with the statistical performance value recommended by Reference [33]. Based with numerical model performance measures, SWAT model had accurately simulated the measured stream flow. Therefore, similar to that of HEC-HMS model, SWAT is also a suitable model for the study area.

Comparison of SWAT and HEC-HMS Models in Simulating the Rainfall-Runoff Process
To compare the model performance in producing streamflow, experiments based on the well calibrated HEC-HMS and SWAT models were conducted.
These results are shown for daily simulation period of the year 1990-2007 ( Figures 6-10). The statistical indices (NSE and R 2 ) are presented in Table 3 and   Table 3 and In similar manner, [2] were tested three Hydrological Distributed Watershed Models MIKE-SHE, APEX and SWAT at Grand River Basin, Canada. As the result, the simulated flows generated by the three models were quite similar and closely match the observed flow, for the calibration period. For validation, they reported as MIKE SHE predicted the streamflow slightly better than either SWAT or APEX. Reference [22] were also compared the performance of the SWAT and XAJ models, and showed that both models performed well in the Xixian River Basin, with a NSE > 0.69 and R 2 > 0.72 for both calibration and validation periods. SWAT and HEC-HMS model were compared for their applicability on Central Highlands of Vietnam by [3] and reported that as both models could simulate fairly well the streamflow for the study area.

Conclusion
This study evaluates the performances of two hydrologic models, namely, HEC-HMS and SWAT, to find the suitable model for hydrological modeling in Katar river basin, Ethiopia. The two models require almost the same data input and model parameters. Both models were calibrated using the observed daily streamflow at Abura gauging station for the period of 1990 to 2000 and validated for a period of 2001 to 2007. The results of calibration and validation indicated that, for River Basin Katar, both models could simulate fairly well the streamflow. SWAT gave the model performance with the R 2 > 0.78 and NSE > 0.67; and the HEC-HMS model provided the model performance with the R 2 > 0.87 and NSE > 0.73 for the calibration and validation periods. In general, the simulated streamflow given by the HEC-HMS model is more satisfactory than that provided by the SWAT model.