GIS-based application of hydrological modelling in the Borkena Catchment, Ethiopia

Knowledge about rainfall–runoff and streamflow is used for several purposes to understand the hydrological and hydraulic characteristics of water bodies or flood forecasting and responses or future water availability. However, most of the time, neither the precipitation nor the streamflow data found are accurate or sufficient enough. In order to overcome this barrier, physical models or hydrological models are used (Makki Vijayaprakash 2021).

Models are the primary way to predict the values of various system performance indicators in hydrologic studies and models of different types provide a means of quantitative extrapolation or prediction that will be helpful in decision-making (Beven et al. 2012; Zhang 2012). A number of hydrologic models have been developed which estimate the peak discharges and the runoff hydrograph for a given rainfall distribution. The applicability and performance of these hydrological models depend on factors such as a mathematical representation of processes occurring, structural complexity of the model, and reliability of the model predictions for the available data, geographic location, climatic conditions, and area of interest, physiographic characteristics, computational skill level, cost, and others. However, the ultimate aim of prediction using models must be to improve decision-making about a hydrological problem, whether that be in water resources planning, flood protection, licencing of abstractions, or other areas (Tian et al. 2013; Sahu et al. 2023).

Recently, floods have become increasingly common and unpredictable throughout the world due to human interference with the natural environment and the impacts of global climate change (Lastoria 2008). Communities were exposed to flood risks to a greater extent as a result of urbanization and changing demographic characteristics in the river floodplain zones. This type of tragedy is being experienced in developing nations, especially those in Africa, as a result of deforestation and a lack of knowledge about land use strategies (Action Aid International 2006; Salami et al. 2017). On the other hand, developed countries with significantly more built-up areas and infrastructures in place are subjected to huge economic losses if a flood is not managed properly (Ashley & Ashley 2007; Hong 2013). Therefore, sufficient and reliable flood prediction and proper design of flood control structures as mitigation measures are the best solutions to challenge this type of disaster. However, this requires a proper understanding of the hydrological and hydro-dynamic processes of the river and the associated catchment characteristics (Booij 2002).

The Hydraulic Engineering Center Hydrologic Modelling System (HEC-HMS) hydrologic model is among the most applicable ones. It has been chosen to estimate and simulate the flow in the hydrological response units of the area (Khaddor & Alaoui 2014; Gebre 2015; Tassew et al. 2019). The main distinction between this model and other hydrological models is that it takes into account the basin's features and changes in the underlying surface elements. As a result, both domestically and internationally, flood research has extensively utilized this model (Zheng et al. 2013; Hu et al. 2017; Tassew et al. 2019). For example, using HEC-HMS, Yucel (2015) examined how various precipitation products, including those from a rain gauge, radar, satellite, and mesoscale Numerical Weather Prediction (NWP) model, responded to the flash flood event. Sardoii et al. (2012) used the HEC-HMS and Geographic Information System to simulate the rainfall–runoff process in the Amirkabir watershed. The aim was to compare different loss methods (Green and Ampt, Soil Conservation Service (SCS), and initial and constant) in the HEC-HMS Model and found that Green and Ampt is the best loss method in the Amirkabir Dam Watershed, Iran. For the Gilgel Abay Catchment in the Upper Blue Nile Basin of Ethiopia, Tassew et al. (2019) simulated surface runoff using the HEC-HMS model. Event-based simulation was conducted and SCS-Curve Number (CN), SCS unit hydrograph, and Muskingum routing methods were selected for each component of the runoff process as runoff depth, direct runoff, and channel routing, respectively. Better results (R² = 0.84, Nash–Sutcliffe Efficiency (NSE) = 75%) were obtained between the simulated and observed peak discharges during the calibration period. Zelelew & Melesse (2018) assessed the applicability of the Hydrological Modelling Software (HEC-HMS) for a simulation of runoff at the Abay River Basin. Their study showed a good fit (R² = 0.70, 0.83) and (NSE = 64.7, 82.8%) between the observed and simulated values of the peak discharge for the selected loss (SCS-CN) and transform (SCS unit Hydrograph) methods. Among the different models provided in the HEC-HMS to estimate rainfall–runoff, the method developed by the Natural Resources Conservation Service (NRCS) of the United States Department of Agriculture (USDA), CN is one of the most popular methods for estimating the volume and peak rates of surface runoff (Fan et al. 2013).

The lack of data for predicting the streamflow of the basin for planning, designing, and carrying out multiple national development projects in the area is one of the key issues facing the Borkena Catchment. Moreover, settlers near the Borkena River are highly affected by flooding. Within the study area, Kombolcha, Harbu, and Kemsie towns are vulnerable to floods. For example, during the 2006 floods, many buildings were destroyed and people were affected in Kemsie and its surroundings, and in 2009 many people were also displaced from their homes in Kombolcha town. As a result, managing water resources effectively is challenging, and there are no historical data to forecast runoff generation for sustainable water resource planning.

The objective of this research was to evaluate the applicability of the semi-distributed hydrological model HEC-HMS in the Borkena Catchment. This was accomplished by estimating the discharge of the river in the catchment using this hydrological model. Achieving this objective is very important for the construction of certain water-related structures such as dams, spillways, dikes, reservoirs, etc. in the study area. Planners and decision-makers can use this model as a method to predict floods in the selected catchment. Governmental or non-governmental organizations can use flood prediction methods without observing the discharge at any point of the river in the catchment using this model. Therefore, this research is crucial to understand and estimate future floods and indirectly save lives and resources lost due to floods in the selected region.

The soil of the Borkena Watershed is characterized by five major dominant soil types: Chromic Vertisols, Eutric Cambisols, Rock Surfaces, Lithosols, and Eutric Regosols (Figure 2(b)). The flat plains of the Borkena Watershed are dominated by Eutric Cambisols and Eutric Regosols which have a dominant textural class of clay and clay loam, respectively. Lithosols and Chromic Vertisols are the dominant soil types found in the mountains and hills of the watershed.

The Borkena Catchment is classified as urban, cultivated, forest, shrub, grassland, and water body in land use and land cover (Figure 2(c)). The classification was done using the supervised land use/cover classification technique. Resource-intensive economic activities often precipitate environmental degradation. Furthermore, population growth contributed to the expansion of cultivated lands and rapid urbanization of the catchment area.

The daily discharge of the study area is collected from the Ministry of Water Irrigation and Electricity of Ethiopia (MoWIE). Unlike the daily precipitation, the daily discharge has continuously recorded data for the considered stations to represent the study area. The gauge is located at the Borkena River outlet close to the main road from Dessie to Addis Ababa. The monthly flow of the Borkena River (1989–2002) is seen in Figure 3(b).

Data, such as topographic maps, soil, land use/land cover, Digital Elevation Model (30 m × 30 m), and meteorological and hydrological data are collected from different organizations (Table 1). The field investigations were made to visualize the study area, observe hydro-geological features, and check and confirm the secondary data collected, which are inputs for the conceptual model.

The HEC-HMS watershed model has been evaluated to simulate the runoff process in the Borkena Catchment. The model is a highly graphical user interface model which requires the construction of three built-in modelling components: The basin model, meteorological model, and control specification. In order to simulate floods with a 100-year return period, the HEC-HMS model was run for 14 years of meteorological and hydrological data. At the watershed's outlet point, peak flows, time to peak, and flow volumes were estimated. The model inputs are daily rainfall, daily observed flow, the digital elevation model, and catchment characteristics (slope, soil type, land use/cover) of the area. To determine the effective rainfall and hydrological process loss, the method of the CN developed by the NRCS of the United States was used. The NRCS-CN method accounts for most of the characteristics of the basins to produce flow, such as soil type, land use, hydrologic conditions, and antecedent moisture (Mishra & Singh 2004). The model considers different sub-watersheds within the catchment. These sub-watersheds provide an efficient way to discretize large watersheds where simulation at the field scale may not be computationally feasible. The model generates three sub-basins, namely Kutaber (41.13 km²), Dessie (89.42 km²), and Kombolecha (95.50 km²) in the Borkena River catchment. It has also seven river reaches, seven junctions, and one sink.

A runoff will only occur when P > 0.2 S.

Calibration is the estimation and adjustment of model parameters and constants to improve the agreement between model output and a dataset (Rykiel 1996). Calibration uses observed hydro-meteorological data in a systematic search for parameters that yield the best fit of the computed results to the observed runoff. The goal of this comparison is to judge how well the model fits the real hydrologic system (Feldman 2000).

The HEC-HMS hydrological model has been calibrated manually and automatically to optimize and gain the correct possible option fit (Gebre 2015). The model was calibrated based on rainfall and observed discharge data of 1989–1997. The rainfall of the three stations (Kombolcha, Dessie, and Kutaber) was analyzed using the Thiessen Polygon method to generate the areal rainfall. The meteorological model in HEC-HMS is responsible for the analysis of the rainfall data. SCS CN loss, SCS unit hydrograph transform, and recession base flow methods were used in this study. The estimation/calibration and verification/validation of the HEC-HMS model is carried out by comparing the daily simulated runoff with the observed streamflow at the outlet of the sub-catchments.

Model validation is the process of testing the model's ability to simulate observed data other than those used for the calibration within acceptable accuracy. During this process, calibrated model parameter values are kept constant. The quantitative measure of the match is again the degree of variation between computed and observed hydrographs (Yassin 2009). The HEC-HMS model is validated by data on daily rainfall and observed discharges for five years from 1998 to 2002. Similar to the calibration, the three rainfall stations were analyzed using the Theissen Polygon method in order to compute the areal rainfall.

The model performance is evaluated by four objective functions: NSE, Percent Bias (PBIAS), Root Mean Square Error (RMSE), and Coefficient of Determination (R²).

The results and discussions are offered in three main sub-parts: calculating the peak discharge and runoff volume, validating and calibrating the model, and assessing the model's performance parameters. In order to develop more precise and potent models, the hydrologic community has invested a lot of time and energy in understanding current hydrological systems in recent years (Pechlivanidis et al. 2011). Based on the amount of rainfall in the watershed, the HEC-HMS model may produce several forms of information.

According to the results of the simulated flow, a large portion of the rainwater that falls in the watershed is converted into direct runoff. Along with the state of the soil and land cover, the watershed's physical characteristics play a vital role. The time of peak discharge 10 August 1996 for computed result and observed flow indicate that the time to peak flow is perfectly simulated for calibration. However, the time of peak discharge on 24 July 2002 for the computed result and the time of peak discharge on 19 July 2001 for observed flow indicate that the simulated time to peak flow was different for the validation period.

The HEC-HMS model used observed precipitation and streamflow data from 1989 to 1997 for calibration, the calibrated model was further verified by observed data from 1998 to 2002 without changing model parameters. Model performances were assessed based on four criteria and the results are shown in Table 3.

The performance measures provide further insights into the accuracy of the simulated flow rates. The NSE values during calibration and validation were 0.65 and 0.70, respectively. These values indicate that the simulated flow rates during validation were closer to the observed flow rates than during calibration. An NSE value of 1 indicates perfect agreement between observed and simulated flow rates, while a value of 0 indicates no agreement. Therefore, the NSE values in this case indicate reasonable agreement between observed and simulated flow rates.

The RMSE values during calibration and validation were 0.6 and 0.5, respectively. These values indicate that the average difference between observed and simulated flow rates was relatively small. The PBIAS values during calibration and validation were 12.84 and 12.16%, respectively. These values indicate a slight bias toward underestimating the simulated flow rates.

Finally, R² values during calibration and validation were 0.69 and 0.71, respectively. These values indicate that the simulated flow rates were moderately correlated with the observed flow rates. In conclusion, the simulation results suggest that the simulated flow rates were generally lower than the observed flow rates, with a slight bias toward underestimation. However, the NSE, RMSE, PBIAS, and R² performance measures indicate reasonable agreement between observed and simulated flow rates, particularly during validation.

There could be several factors contributing to the underestimation of simulated flow rates. Some possible factors include inaccurate input data, inappropriate model structure, poor calibration, uncertainty in the observed data, and limitations of the simulation model. Overall, the underestimation of simulated flow rates could be due to a combination of these factors and further analysis and refinement of the simulation model may be necessary to improve the accuracy of the simulated flow rates.

Hydrologic models and GIS integration demonstrated that it is possible to create a trustworthy flood assessment for basins using digital data, satellite imaging, and a strong network of field observations. In this study, the Borkena Watershed has been analyzed and a simulation model based on the past rainfall event of this area has been run, and information has been generated. The simulated peak discharge of 122.6 m³/s, the time to peak on 10 August 1996, and a runoff volume of 12,424.8 mm of the outlet points were obtained from the model output, which are very useful for flash flood analysis. The model is well simulated with the daily streamflow at the outlet point of the watershed (Kombolecha reach). However, there is a slight under-prediction of the peak flows and overestimation of the runoff volumes. The model statistical performance evaluation results showed a very good and acceptable accuracy on the daily time scale.

The simulated streamflow results were compared with monitored values at each change of parameters. Muskingum K, CN, and initial abstraction (I_a) were the most sensitive parameters for the simulation of runoff. In conclusion, the application of the HEC-HMS model shows the ability to simulate the peak flow and runoff volume at the Borkena Catchment.

We appreciate the editor and the two anonymous reviewers for their constructive feedback.

Data cannot be made publicly available; readers should contact the corresponding author for details.

The authors declare there is no conflict.