Integrating Satellite Rainfall Estimates with Hydrological Water Balance Model: Rainfall-Runoff Modeling in Awash River Basin, Ethiopia

: Hydrologic models play an indispensable role in managing the scarce water resources of a region, and in developing countries, the availability and distribution of data are challenging. This research aimed to integrate and compare the satellite rainfall products, namely, Tropical Rainfall Measuring Mission (TRMM 3B43v7) and Precipitation Estimation from Remotely Sensed Information using Artiﬁcial Neural Networks-Climate Data Record (PERSIANN-CDR), with a GR2M hydrological water balance model over a diversiﬁed terrain of the Awash River Basin in Ethiopia. Nash–Sutcliffe efﬁciency ( NSE ), percent bias ( PBIAS ), coefﬁcient of determination ( R 2 ), and root mean square error ( RMSE ) and Pearson correlation coefﬁcient ( PCC ) were used to evaluate the satellite rainfall products and hydrologic model performances of the basin. The satellite rainfall estimations of both products showed a higher PCC (above 0.86) with areal observed rainfall in the Uplands, the Western highlands, and the Lower sub-basins. However, it was weakly associated in the Upper valley and the Eastern catchments of the basin ranging from 0.45 to 0.65. The ﬁndings of the assimilated satellite rainfall products with the GR2M model exhibited that 80% of the calibrated and 60% of the validated watersheds in a basin had lower magnitude of PBIAS (< ± 10), which resulted in better accuracy in ﬂow simulation. The poor performance with higher PBIAS ( ≥± 25) of the GR2M model was observed only in the Melka Kuntire (TRMM 3B43v7 and PERSIANN-CDR), Mojo (PERSIANN-CDR), Metehara (in all rainfall data sets), and Kessem (TRMM 3B43v7) watersheds. Therefore, integrating these satellite rainfall data, particularly in the data-scarce basin, with hydrological data, generally appeared to be useful. However, validation with the ground observed data is required for effective water resources planning and management in a basin. Furthermore, it is recommended to make bias corrections for watersheds with poorlyww performing satellite rainfall products of higher PBIAS before assimilating with the hydrologic model. have the of


Introduction
To address water resource planning and management problems, different rainfallrunoff models are used to understand the hydrological processes in a basin. However, it is necessary to test whether a specific model fits a particular basin [1,2]. The applicability of these models is mainly constrained by the type and availability of input data in specific basins. The availability and distribution of ground-based rainfall-runoff data in African river basins are sparse [3]. This makes hydrological studies difficult in a basin where gauging stations are poorly distributed, particularly in the river basins of Ethiopia [4,5]. The Awash River Basin (ARB) has a complex landscape, varied climatic conditions, and an uneven distribution of hydrometeorological stations [6,7]. Rapid population growth, settlement, expansion of agricultural activities, upstream soil erosion, and pollutants in the basin threaten the freshwater resource availability of the ARB [8][9][10].
There are policy challenges on implementation of integrated water resource management (IWRM) principles in ARB. Adey et al. [11] have made an in-depth insight on IWRM policies and practices on a basin. They explained that there is a considerable disagreement on IWRM principles and the approach followed in a river basin. This has resulted in poor water management practices and scarcity of freshwater resource in a basin among different water users. Therefore, different water resource management and planning tools need to be synchronized to improve the implementation of IWRM on a river basin.
Various studies worldwide have used conceptual lumped hydrologic models to estimate the regional water availability for an ungauged basin [12][13][14]. Others have also used this type of model to assess the climatic impacts on different hydrological conditions [4,5,15,16]. Among different hydrologic models, the GR2M water balance model is tested in ARB.
The GR2M water balance model is similarly categorized as a global conceptual rainfallrunoff model. It is a monthly lumped hydrological model characterized by its parsimonious and low-level complexity. Furthermore, it is mainly focused on prominent features of the rainfall-flow transformation. This model is valuable for managing water resources, reservoir simulation, and drought predictions. In addition, Coron et al. [17] explained that GR lumped hydrological models are suitable for flood forecasting and impact assessment on climate change.
In Ethiopia, different studies have been conducted using hydrologic models over various catchments or at the river basin scale. Tadesse and Dai [28] predicted sedimentation in reservoirs by combining catchment-based (Soil and Water Assessment Tool (SWAT)) and stream-based (Hydrologic Engineering Center-River Analysis System (HEC-RAS)) models to estimate the sediment load reaching the Koka reservoir in the Upper Awash Basin, Ethiopia. Furthermore, Setegn et al. [29] and Mekonnen et al. [30] tested the selected catchment of the Upper Blue Nile using the SWAT model. Uhlenbrook et al. [31] analyzed the catchment behavior of the Upper Blue Nile catchment using Hydrologiska Byråns Vattenbalansavdelning (HBV) modeling. In the same basin, Abdo et al. [32] assessed the climate change influences on the hydrology of Gilgel Abay watershed using the HBV model.
Hydrometeorological data play a significant role for hydrologic modeling purposes to manage the water resources in a basin [33]. The rainfall gauging station distribution map of Ethiopia shows that the concentration of gauging stations is relatively high in the Uplands and Western highlands of the river basin, but the stations are sparsely distributed in the Upper valley, Middle valley, Eastern catchment, and Lower basin [34]. The adequacy and quality of recorded hydrometeorological data in a basin remain a challenge in hydrology and water resource-related studies in the ARB [35]. Therefore, testing and integrating satellite rainfall products with the available flow data greatly improves the applicability of any hydrological model that associates the rainfall-runoff relations, and also offers an alternative to ground-based rainfall estimates in an area where no records of observed rainfall are available [20,36,37].
To date, various studies related to satellite rainfall products over Ethiopia's river basins have been conducted, and the abilities of the products to detect rain events have been tested [3,6,35,38,39]. However, integrating these satellite rainfall products with the GR2M water balance model has never been tested in the Ethiopian river basin. Furthermore, research on integrating satellite rainfall products with hydrological models to study the rainfall-runoff process at a large river basin scale is scarce. This study provides insights on the rainfall-runoff modeling using different satellite rainfall (Tropical Rainfall Measuring Mission (TRMM) 3B43 and Precipitation Estimation from Remotely Sensed Information using Artificial Neural Networks-Climate Data Record (PERSIANN-CDR)) products with a GR2M hydrologic water balance model in the water-stressed ARB of Ethiopia.

Study Area
The Awash River is one of the largest rivers within Ethiopian territories. It is located at 7 • 53 N-12 • N and 37 • 57 E-43 • 25 E and covers an area of 116,373 km 2 . The elevation in the river basin ranges from 240 to 4187 m above sea level (m a.s.l.) (Figure 1). Based on the hydrologic conditions of the ARB, it has been classified into seven sub-basins, which are the Uplands (Holeta, Melka Kuntire, Hombole, and Mojo areas), the Upper valley (Sire and Metehara areas), the Western highlands (Kessem Kebena areas), the Middle valley (Adaitu area), the Lower valley, and Lower plain. These Lower valley and Lower plain is considered as Lower basin (Tendaho sub-basin). The Western highlands has a major contribution to the surface flow of the river basin [40,41]. water balance model has never been tested in the Ethiopian river basin. Furthermore, research on integrating satellite rainfall products with hydrological models to study the rainfall-runoff process at a large river basin scale is scarce. This study provides insights on the rainfall-runoff modeling using different satellite rainfall (Tropical Rainfall Measuring Mission (TRMM) 3B43 and Precipitation Estimation from Remotely Sensed Information using Artificial Neural Networks-Climate Data Record (PERSIANN-CDR)) products with a GR2M hydrologic water balance model in the water-stressed ARB of Ethiopia.

Study Area
The Awash River is one of the largest rivers within Ethiopian territories. It is located at 7°53′ N-12° N and 37°57′ E-43°25′ E and covers an area of 116,373 km 2 . The elevation in the river basin ranges from 240 to 4187 m above sea level (m a.s.l.) (Figure 1). Based on the hydrologic conditions of the ARB, it has been classified into seven sub-basins, which are the Uplands (Holeta, Melka Kuntire, Hombole, and Mojo areas), the Upper valley (Sire and Metehara areas), the Western highlands (Kessem Kebena areas), the Middle valley (Adaitu area), the Lower valley, and Lower plain. These Lower valley and Lower plain is considered as Lower basin (Tendaho sub-basin). The Western highlands has a major contribution to the surface flow of the river basin [40,41]. The average annual rainfall of the ARB varies from 160 (in Asaita of the Lower plain sub-basin) to 1600 mm at Ankober (the Western highland sub-basin). Additionally, mean annual temperature of the ARB ranges from 20.8 to 29 °C at Koka (in the Upland subbasin) and Dubti (in the Lower valley sub-basin), respectively [42]. The major land use The average annual rainfall of the ARB varies from 160 (in Asaita of the Lower plain sub-basin) to 1600 mm at Ankober (the Western highland sub-basin). Additionally, mean annual temperature of the ARB ranges from 20.8 to 29 • C at Koka (in the Upland sub-basin) and Dubti (in the Lower valley sub-basin), respectively [42]. The major land use land cover proportionate of the basin includes agricultural land (51.39%), grassland (29.79%), and shrublands (8.11%), respectively [40]. A climatic data summary of the ARB is provided in Table 1.

Data Sources
The delineation of the sub-basins for selected stations and the whole river basin was performed using a 90 m resolution digital elevation model of Shuttle Radar Topographic Mission (SRTM) downloaded from the United States Geological Survey (USGS) website (https://earthexplorer.usgs.gov/ accessed on 10 January 2020).
The daily rainfall data (41 stations) and other climatic parameters (daily minimum and maximum temperature, wind speed, relative humidity, sunshine hour duration) for 10 subbasins were obtained from the national meteorology agency of Ethiopia. The satellite rainfall products (TRMM 3B43v7 and PERSIANN-CDR) with spatial resolution of 0.25 • × 0.25 • (~27.8 × 27.8 km) were retrieved from the data center of US National Aeronautics and Space Administration (NASA, http://giovanni.gsfc.nasa.gov/giovanni/ accessed on 5 March 2020) and Center for Hydrometeorology and Remote sensing (http://chrsdata.eng.uci.edu/ accessed on 15 March 2020) for the length of records described in Table 1, respectively. These daily data were used to compute the monthly and annual climatic parameters depending on the needs of the analyses and in accordance with the available monthly discharge data of a river in a basin.

Preliminary Data Analysis
The missing observed rainfall data for long-term daily average values of various years were infilled if the missing data were less than 20% of the total. In addition, some missing rainfall data were replaced from the corresponding stations that had similar hydrometeorological characteristics. However, data with continuous missing values were systematically ignored from the analysis. There were no missing data for monthly satellite rainfall products of the sub-basins.
A Tukey fence method was used to test the outliers that can affect the detection of inhomogeneity of rainfall data series [43,44]. The rainfall data range is explained below.
where Q 1 and Q 3 are the upper and lower quartile points, respectively, 1.5 refers to the standard deviation from the mean, and IQR are the interquartile ranges. The consistency of the observed rainfall data was analyzed using double-mass curve techniques [43,45] for only 10 meteorological stations with river flow data. A Theissen polygon method was used to convert point rainfall to areal rainfall for specific sub-basins that had river flow records. For comparison, mean multi-annual isohyetal rainfall maps over the entire basin using the observed and satellite rainfall were analyzed using kriging techniques to compare the patterns and ranges of rainfall. Similarly, the monthly potential evapotranspiration (PET) for the study period, for 10 river gauging stations, were performed using the Penman-Monteith (CROPWAT 8.0 software) and Blaney-Criddle methods depending on the climatic data availability.
For this study, the monthly river flow data (discharge, m 3 /s) were obtained from the Global Runoff Data Center (GRDC, http://www.bafg.de/GRDC/ accessed on 17 December 2019) and Ministry of Water, Irrigation, and Electricity of Ethiopia (MoWIE).
ArcMap 10.1 and ArcSWAT were used for delineating and extracting the river networks, determining the basin area, and other basin characteristics. In addition, an elevation map, soil water holding capacity maps, and interpolation of point rainfall data using kriging were analyzed using Arc Map 10.1. MATLAB R2020a and Microsoft Excel 2016 were used for statistical analysis and graph development.

Soil Water Holding Capacity (SWHC)
The SWHC represents a soil moisture reserve in a soil which is being utilized for growing of vegetations in water deficit periods when rainfall does not meet the crop evapotranspiration demands [46,47]. The soil map of the river basin was prepared using FAO soil database as information. These 14 soil groups have a distinct proportionate of soil textural classes. Therefore, field capacity (FC) and permanent wilting point (PWP) ranges of values of a dominant texture of soil types were extracted from Allen et al. [48]. In addition, depth of the root zone of dominant crops in various parts of the river basin was used as input to compute the total available water (TAW or SWHC). This information was later used as an input to estimate the initial filling rate of the two tanks assumed in the GR2M conceptual model. Figure 2 shows the soil map of the ARB that is dominated by 14 soil groups. The textural classes of each soil group were identified from harmonized world soil database (HWSD) documents [49].
where TAW is total available water, which is equal to SWHC, Z r is the depth of root zone (m), θ FC and θ PWP is the soil moisture at field capacity and wilting point (m 3 m −3 ), respectively. r 2021, 13, x FOR PEER REVIEW 5 of 18 over the entire basin using the observed and satellite rainfall were analyzed using kriging techniques to compare the patterns and ranges of rainfall. Similarly, the monthly potential evapotranspiration (PET) for the study period, for 10 river gauging stations, were performed using the Penman-Monteith (CROPWAT 8.0 software) and Blaney-Criddle methods depending on the climatic data availability. For this study, the monthly river flow data (discharge, m 3 /s) were obtained from the Global Runoff Data Center (GRDC, http://www.bafg.de/GRDC/ accessed on 17 December 2019) and Ministry of Water, Irrigation, and Electricity of Ethiopia (MoWIE).
ArcMap 10.1 and ArcSWAT were used for delineating and extracting the river networks, determining the basin area, and other basin characteristics. In addition, an elevation map, soil water holding capacity maps, and interpolation of point rainfall data using kriging were analyzed using Arc Map 10.1. MATLAB R2020a and Microsoft Excel 2016 were used for statistical analysis and graph development.

Soil Water Holding Capacity (SWHC)
The SWHC represents a soil moisture reserve in a soil which is being utilized for growing of vegetations in water deficit periods when rainfall does not meet the crop evapotranspiration demands [46,47]. The soil map of the river basin was prepared using FAO soil database as information. These 14 soil groups have a distinct proportionate of soil textural classes. Therefore, field capacity (FC) and permanent wilting point (PWP) ranges of values of a dominant texture of soil types were extracted from Allen et al. [48]. In addition, depth of the root zone of dominant crops in various parts of the river basin was used as input to compute the total available water (TAW or SWHC). This information was later used as an input to estimate the initial filling rate of the two tanks assumed in the GR2M conceptual model. Figure 2 shows the soil map of the ARB that is dominated by 14 soil groups. The textural classes of each soil group were identified from harmonized world soil database (HWSD) documents [49].

GR2M Hydrological Modeling
The GR2M is a spatially lumped hydrologic model with two-parameter estimation from a monthly time-step input data of a given basin [19,50,51]. The general concept, scheme, and detailed formula of the GR2M model are explained in [50]. Here, this GR2M hydrologic model was used to calibrate and validate in ARB using different rainfall data sets.

Satellite Rainfall and Hydrological Model Evaluation Criteria
The data records from each sub-basin were categorized into two different periods-for calibration and simulation of flows. These periods varied depending on the length of available data for each sub-basin. The first year of simulated discharge, which considered a warmup period, was not used in computations of performance evaluation of the model.
where Q obs,i is the ith observed value, Q sim,i is the ith simulated value, Q obs is the mean observed value, and n is the total number of observations, Q sim is the mean simulated value, P Gauge i and P Sat i are annual or monthly on-site observed rainfall (gauged) and satellite rainfall estimates, STDEV obs is the observed standard deviation. The ratings of the evaluation criteria were performed as shown in Table 2.

Comparison of Satellite Rainfall with Observed Rainfall
The isohyetal rainfall maps using ordinary kriging methods were developed ( Figure 3). In addition, comparisons of the satellite (TRMM 3B43v7 and PERSIANN-CDR) and observed rainfall data were performed using statistical descriptors such as PCC and RMSE (Figure 4).

Comparison of Satellite Rainfall with Observed Rainfall
The isohyetal rainfall maps using ordinary kriging methods were developed ( Figure  3). In addition, comparisons of the satellite (TRMM 3B43v7 and PERSIANN-CDR) and observed rainfall data were performed using statistical descriptors such as PCC and RMSE (Figure 4). Isohyetal rainfall (Figure 3) shows that the observed rainfall captured a wider range of rainfall amounts on a yearly basis, with a minimum of 242 in the Lower Awash basin and a maximum of 1536 mm in the Western highland catchment (Table 3). However, the Isohyetal rainfall (Figure 3) shows that the observed rainfall captured a wider range of rainfall amounts on a yearly basis, with a minimum of 242 in the Lower Awash basin and a maximum of 1536 mm in the Western highland catchment (Table 3). However, the satellite rainfall estimates in a basin captured nearly the same minimum rainfall amount per year for both products, but showed a discrepancy in the maximum ranges of rainfall records in TRMM 3B43v7 (1185 mm) and PERSIANN-CDR (1457 mm) ( Table 3). The isohyetal pattern of rainfall showed a resemblance between the observed and PERSIANN-CDR data, but it was completely different in the TRMM 3B43v7 product. This variation might be due to the nature of the product produced, the elevation, and rainfall regime of the basin. Both satellite rainfall estimations showed a higher PCC with areal observed rainfall in the Uplands, the Western highlands, and the Lower sub-basins. However, it was The isohyetal pattern of rainfall showed a resemblance between the observed and PERSIANN-CDR data, but it was completely different in the TRMM 3B43v7 product. This variation might be due to the nature of the product produced, the elevation, and rainfall regime of the basin.
Both satellite rainfall estimations showed a higher PCC with areal observed rainfall in the Uplands, the Western highlands, and the Lower sub-basins. However, it was weakly associated in the Upper valley and the Eastern catchments of the basin (Figure 4). A higher RMSE was noted in the Upper valley area of the basin for both satellite rainfall data, and it extended into the Eastern catchment while using PERSIANN-CDR. The long-term annual PERSIANN-CDR rainfall with station elevation showed a decreasing trend, particularly in the highest elevation areas (2250-2800 m). The elevations of the selected river gauging stations were located below 2250 m. Therefore, it was possible to apply the two-satellite data depending on the location and altitude of the basin.

SWHC/TAW
Using the soil groups and textural information, the TAW in the soil of various ranges, such as minimum, maximum, and mean values, were computed. The TAW maps for the soil group in ARB are shown in Figure 5.
two-satellite data depending on the location and altitude of the basin.

SWHC/TAW
Using the soil groups and textural information, the TAW in the soil of various ranges, such as minimum, maximum, and mean values, were computed. The TAW maps for the soil group in ARB are shown in Figure 5. The Uplands, the Western highland catchments, and the Eastern catchments were dominated by higher TAW in the root zone ( Figure 5). In contrast, less available water was observed in the Middle Awash and Lower Awash sub-basins.

GR2M Hydrological Modeling
The areal ground rainfall observation stations (GROS) generated using the Thiessen polygon for the sub-basins were compared with the areal satellite rainfall data of individual stations in the ARB. The comparison was based on the different statistical evaluation criteria discussed below. The Uplands, the Western highland catchments, and the Eastern catchments were dominated by higher TAW in the root zone ( Figure 5). In contrast, less available water was observed in the Middle Awash and Lower Awash sub-basins.

GR2M Hydrological Modeling
The areal ground rainfall observation stations (GROS) generated using the Thiessen polygon for the sub-basins were compared with the areal satellite rainfall data of individual stations in the ARB. The comparison was based on the different statistical evaluation criteria discussed below.

Relationship of Mean PET, Streamflow vs. Rainfall
The mean annual PET trends in the basin showed a 54 mm drop for every increment of 50 mm rainfall in the basin, and PET was negatively correlated (−1.07) with mean rainfall in the basin. In addition, other climatic factors may influence the PET of the ARB. The mean annual rainfall trends exhibited a high coefficient of determination (R 2 = 0.81) with a mean streamflow rise of 14 mm for every 50 mm increment of rainfall ( Figure 6).

Relationship of Mean PET, Streamflow vs. Rainfall
The mean annual PET trends in the basin showed a 54 mm drop for every increment of 50 mm rainfall in the basin, and PET was negatively correlated (−1.07) with mean rainfall in the basin. In addition, other climatic factors may influence the PET of the ARB. The mean annual rainfall trends exhibited a high coefficient of determination (R 2 = 0.81) with a mean streamflow rise of 14 mm for every 50 mm increment of rainfall ( Figure 6). showed that model performance capabilities were "good" or higher. Furthermore, 80% and 17% of the calibrated PBIAS for gauged and satellite rainfall (TRMM 3B43v7 and PER-SIANN-CDR) data showed a performance rating of "very good" and "good," respectively. The low-magnitude PBIAS in 80% of the calibrated watersheds indicates an accurate model simulation. The satellite rainfall data in the Metehara watershed exhibited a result of "satisfactory" compared to other watersheds in the basin (Table 4).  showed that model performance capabilities were "good" or higher. Furthermore, 80% and 17% of the calibrated PBIAS for gauged and satellite rainfall (TRMM 3B43v7 and PERSIANN-CDR) data showed a performance rating of "very good" and "good," respectively. The low-magnitude PBIAS in 80% of the calibrated watersheds indicates an accurate model simulation. The satellite rainfall data in the Metehara watershed exhibited a result of "satisfactory" compared to other watersheds in the basin ( Table 4).
The model validation results indicate that 53% and 27% of the validated watersheds exhibited performance ratings of a model as "good" and "very good" when using NSE and RSR as evaluation criteria. Despite this, 20% were identified as unsatisfactory for integration with the GR2M model ( Table 5).
Part of the Upper valley (Metehara watershed) and Middle valley (Sire watershed) sub-basins did not perform well with the GR2M model and scored an NSE of less than 50. However, the degree of collinearity between simulated and measured data was in the range of 0.58 to 0.64 in both watersheds, respectively.
A higher percentage of bias (PBIAS) was identified while validating the observed flow of the rivers using satellite rainfalls of some watersheds in a basin (Table 5) Table 5).
The rainfall-streamflow hydrograph using the observed and the satellite rainfall products of the ARB are shown in Figures 7-9.

Discussion
Previously, different authors made intensive studies on rainfall-runoff modeling in Ethiopian river basins using various hydrologic models. These models can be distributed, semi-distributed, or lumped hydrologic models. As stated in the introduction section, SWAT, HBV, HEC-RAS, and other hydrologic models were tested and commonly recommended to use in different river basins of Ethiopia. The choice of these models may depend on the availability of the data or input requirement, topographic nature, climatic situations, soil type, land use, morphological characteristics of a watershed, and other related factors. The finding of this GR2M lumped water balance model provides better performance in ARB, except for some watersheds that require correction of the satellite rainfall data with higher PBIAS. The orographic effect in mountainous areas, rainfall regime and nature of the satellite rainfall products might affect the quality of satellite rainfall estimates and contributes to the variation in estimates of the product which later has an impact on rainfall-runoff modeling. Therefore, careful observation in the trends and amounts of satellite rainfall estimations versus elevation must be given due attention. Dinku et al. [35] suggested the variation in satellite rainfall estimation could be minimized through local calibration of satellite algorithms with remotely retrieved data and merging with ground-observed data as one technique. In general, evaluating this two-parameter water balance model in the ARB provides good alternative options for water resource planning in data-scarce regions of Ethiopia.

Conclusions and Recommendations
In this study, two satellite rainfall products and the observed rainfall from gauging stations were integrated with the GR2M hydrological water balance model over the complex and diverse terrain of the ARB in Ethiopia. The major findings of this study are as follows: • Both satellite rainfall estimations showed a higher PCC with areal observed rainfall in the Uplands, the Western highlands, and the Lower sub-basins. However, the correlations in the Upper valley and the Eastern catchments of the basin were weak. Therefore, the satellite rainfall data in a watershed that are weakly associated need to be bias-corrected to improve the performance for integration with the GR2M model.

•
The findings of the assimilated satellite rainfall products with GR2M model exhibited that 80% of the calibrated and 60% of the validated watersheds in a basin had lower magnitude of PBIAS (<±10), which resulted in better accuracy in flow simulation. • The hydrologic model validation results revealed that 80% of the validated watersheds using observed and satellite rainfall data sets showed a "good" or higher performance rating when the NSE and RSR were used as evaluation criteria. In contrast, 20% of them were unsatisfactory for integration with the GR2M model.

•
The PBIAS results showed that the majority of the rainfall dataset of the watershed (73%) underestimated the simulated flow while integrating with the GR2M model. A higher PBIAS value, indicating unsatisfactory results, was observed only in the Melka Kuntire (TRMM 3B43v7 and PERSIANN-CDR), Mojo (PERSIANN-CDR), Metehara (all rainfall data set), and Kessem (TRMM 3B43v7) watersheds.

•
Integrating these satellite rainfall data, particularly in data-scarce basins, with hydrological data generally appears to be useful in the environmentally diversified climate and topography of ARB. However, this requires validation with the groundobserved data.

•
In general, this conceptual lumped model displayed better performance in majority of the ARB parts and is recommended to be tested in other river basins of Ethiopia for effective water resource planning and management.