Abstract

Conducting a flood frequency analysis and mapping of the inundated area in rivers are important for river flow modeling. The main purpose of this research is to estimate the peak flow, model the inundated area using HEC-RAS, and conduct an analytical hierarchy process for the upper Baro Akobo basin in Ethiopia. The inundation area and river depth for 25, 50 and 100 years are considered while contemplating several factors which contribute to flooding. The downstream of the basin has experienced numerous floods that occurred in 2006, 2007, 2010, 2011, and 2012. Flood frequency analysis with stream flow data from 1990–2009 at the Baro-Gambella gauging station was carried out to estimate the expected peak floods of the watershed. The analysis was conducted using the Gumbel, Normal, and Log Pearson Type III distribution methods. The peak floods with return periods of 25, 50, and 100 years with a minimum statistical value calculated using the Normal distribution method resulted in 1739.586 m³/s, 1820.872 m³/s, and 1893.974 m³/s, respectively. The HEC-RAS model results indicated that the flood inundation areas under different land use changes for 25-, 50-, and 100-year return periods were 446.2 km² (annual crop cover), 404.4 km² (built area cover), 323.3 km² (flooded vegetation), and 93.58 km² (forest area), respectively, whereas the inundation depth ranged from 0–2.6 m, 0–2.9 m, and 0–3.2 m depth at the upstream and downstream of the river, respectively. The outcome of this study could be used to reduce temporal and permanent flood risk.

1 Introduction

Flooding is a devastating problem in the world which has numerous impacts including loss of life and properties. In developing countries like Ethiopia, the magnitude is massive due to poor river and watershed management practices. A flood is defined as the excess flow of water that exceeds the transportation capacity of a stream channel, lakes, ponds, reservoirs, drainage system, dam, or other water bodies causing water to overflow outside of the water body. This may occur when a river or stream changes its natural course or artificial banks due to heavy rainfall and dam failure. Floods are the most destructive natural catastrophes resulting in injuries and deaths, extensive infrastructure damage, and social problems (Rosser et al. 2017). Flood destruction has increased in recent years due to climate and environmental changes (Kundzewicz et al. 2014; Mangini et al. 2018; Mostofi Zadeh et al. 2020). Its effect in poor countries is more terrifying due to the lack of adequate disaster response and economic resources. According to Asadi and Boostani (2013), flooding happens when river levels rise, which can be influenced by high discharges, water backing up and rising bed levels, and expanding human interference in highland areas. Between the months of June and August, Ethiopia suffers its wettest months, resulting in severe flooding in some parts of the country (Getahun and Gebre 2015). The Gambella region of Ethiopia also suffers a lot, especially when rainfall is at its highest.

The topography of the area contributes to the flooding (Romali 2018; Son and Jeong 2019). As a result, more water accumulates in the downstream region than in the upstream, increasing the risk of flooding. The focus of this research is to prepare a flood inundation map and models using HEC-RAS and EasyFit software to conduct a flood frequency analysis (Ardıçlıoğlu and Kuriqi 2019). Numerous studies (Brunner 2016; Hou et al. 2018; Hou et al. 2021; Karamma and Pallu 2018; Praveen et al. 2020; Romali 2018; Son and Jeong 2019; Wijayanti et al. 2017; Yang et al. 2021) use various modeling approaches to minimize flood hazards. Among those modeling approaches, HEC-RAS and the analytical hierarchy process have been considered for this study, since they focus on detailed terrain analysis in river simulation using multiple criteria when evaluating alternative approaches (Ardıçlıoğlu and Kuriqi 2019; Romali 2018; Praveen et al. 2020; Vijayan et al. 2021). According to the most recent UNICEF report, flooding has affected nearly 53,000 households in both the Gambella and Oromia regional states in August and September of 2022. This demonstrates how serious flooding has become across the country. The Gambella region is one of Ethiopia's regional states that has been notably impacted by floods because most of the population live near the Baro-Akobo River’s bank, exposing them to frequent flooding. Heavy precipitation in the upstream highland rivers such as the Birbir and Geba Rivers is causing the Baro-Akobo basin to overflow. This overflow of water causes flooding in the Gambella region and affects the community and infrastructure. As a result, flooding is the most severe hazard in the downstream area of Gambella region.

This research aimed to minimize the flooding problem in and around the Baro Akobo basin using river simulation models for a return period of 25, 50 and 100 years, including several flood-influencing factors (DEM, aspect, curvature, rainfall, slope, TRI, flow accumulation, flow direction, SPI, NDVI, population density, distance from the river, TWI, drainage density, LULC, soil type, and STI). The research assesses the causes and risks of flooding in the Upper Baro-Akobo basin of Ethiopia. This study also aims to identify the root causes of the vulnerability of the Gambella people. More importantly, the results of the study are being investigated to contribute significantly to the development of sustainable mitigation measures to minimize the impact of floods and associated risks through an integrated study of flood causes and risk assessments.

2 Materials and methods

2.1 Study area

The Upper Baro-Akobo watershed is part of the Baro-Akobo basin in the south-western part of Ethiopia, situated 766 km away from Addis Ababa, the capital city of Ethiopia. The basin lies between 6° 0' 0'' N and 9° 23' 0'' N latitude and between 34° 25'  0" E and 37° 20' 0'' E longitude, with elevations ranging from 390 m to 3266 m above mean sea level (amsl) and has a total catchment area of 20,562.2 km² (Figure 1). The undulating topography is characterized by mountains, slopes, plateaus, plains, and river basins. The Upper Baro-Akobo basin was formed by the two rivers, Birbir and Gebba, joining east of Metu in the Illubabor zone of the Oromia region. It flows through Ethiopia's Gambella region's lowland area and moves towards the Republic of South Sudan, where it is called Sobat, before joining the White Nile River. The major rivers in the basin are Baro, Alwero, Gilo, and Akobo.

Figure 1 Location map of the study area: a) Basins in Ethiopia, b) Baro-Akobo basin, c) Upper Baro-Akobo watershed.

2.2 Materials

The materials used for mapping the flood inundation area and flood frequency analysis for the upper Baro-Akobo basin, Ethiopia are HEC-RAS, EasyFit software, Digital Elevation Model (DEM), ArcGIS, Landsat image, and Google Earth.

Spatial and temporal data collection and analysis

To perform this study, several types of data were collected from various organizations and websites. The data collected for this research was hydrological data, meteorological data, and spatial data such as elevation, aspect, curvature, slope, TRI, flow accumulation, flow direction, SPI, population density, distance from river, TWI, drainage density, soil type, LU/LC, NDVI, and STI. Table 1 describes the different types of data, their sources, and their possible uses.

Table 1 Data types, sources, and purpose.

Factors	Sources	Record length	Purpose
DEM (12.5*12.5m)	Downloaded from https://asf.alaska.edu/on	–	Reclassify the elevation, aspect, curvature, slope, TRI, flow accumulation, flow direction, SPI, population density, distance from river, TWI, drainage density, and STI maps.
Soil type	FAO soil classification	–	Prepare a reclassified soil map of the study area.
Rainfall	National Metrological Agency	1988-2020	Prepare reclassified rainfall map of the study area
LULC	Landsat-8 OLI satellite imagery derived from USGS	1991/2000/2021	Detect change analysis on LULC class of the study area
NDVI	Landsat-8 OLI satellite imagery obtained from USGS	1991/2000/2021	Prepare reclassified vegetation coverage map of the study area
Stream flow (Appendix B) and HEC-Geo RAS data	MOWIE and DEM	1991-2019	Use as an input in the HEC-RAS model to generate a flood inundation map for the study area.

Models used to gather data and description

The models used for collecting and analyzing the data have been selected based on their capability to solve the problems and achieve the specific objectives of the study. The study used three models to produce adequate and reliable data. These are the multi-criteria decision analysis evaluation (MCDA), ArcGIS, and HEC-RAS models.

Multi-criteria decision analysis evaluation (MCDA)

MCDA has the potential to reduce costs and time (Kontos et al. 2005). There are many MCDA methodologies available to solve complex decision problems with multiple criteria (Saaty 1990), but in this study the analytical hierarchy process (AHP) was selected because it is the most widely used comparison method among the others and is simple and easy to use to postulate multi-criteria decision problems. The main goal of AHP-based multi-criteria decision making is to rank and prioritize factors.

Analytical hierarchy process (AHP)

AHP is a good technique for handling complicated multi-criteria decision issues, and its application in analyzing diverse geo-hazard problems, such as floods and landslides, has been recognized by researchers, practitioners, and decision-makers. AHP is a method of dividing a problem into multiple sub-problems that are easier to comprehend and subjectively measure (Saaty 1990).

In this study, the analytical hierarchy process was used for factor weighing by using the nine points pairwise comparison scale (Saaty 1990). In AHP, a pairwise comparison provides the weight of each causative factor along with a consistency ratio that indicates the level of subjectivity in the factor comparison. It consists of four major steps; problem definition, performing pairwise comparison, calculating the maximum eigenvalues and eigenvectors, calculating the consistency ratio to check whether the matrix is consistent or not, and summing-up the relative importance weights to determine the final weight for decision alternatives (Ogato et al. 2020; Ouma and Tateishi 2014).

The matrix's effort relative to a uniform is estimated in Equation 1, even when diagonal members are equal to 1.

(1)

Where:

P	=	the probability of each factor in the matrix, and
i, j, and n	=	row, column, and the number of factors, respectively.

The standard comparison matrix of the pair is generated using the weighted average method, which calculates the comparative materials of even pairs.

The final flood hazard map of appropriateness is ready for comparison and interpretation if the consistency ratio (CR) value is less than 0.1, and the data is acceptable. If the value exceeds 0.1, then the comparison matrix reveals a discrepancy, and the results are readjusted using the principles of Saaty (1990).

Consistency check

The consistency of the assessment was evaluated and confirmed using the Consistency Index (CI) and Consistency Ratio (CR) (Equation 2 and Equation 3, respectively).

	(2)
	(3)
	(4)

Where:

CI	=	Consistency Index,
n	=	number of parameters,
CR	=	Consistency Ratio,
RI	=	Random Index using Saaty (1990),
i	=	every element in matrix,
Xi, j	=	decision parameters,
Wi, j	=	Analytical Hierarchy Process for weight, and
δmax	=	Max Eigen values of the normalized comparison matrix, and the comparison matrix of the flood influence factors were prepared accordingly (Table 2 and Table 3).

Table 2 Random inconsistency index.

n	1	2	3	4	5	6	7	8	9	10
R.I.	0	0	0.58	0.9	1.12	1.24	1.32	1.41	1.45	1.49

Table 3 Continuous rating scale (Taherdoost 2017).

Intensity	Definition	Explanation
1	Equal importance	Two elements contribute in equally to the objective
3	Moderate importance	Experience and judgment slightly favor one element over another
5	Strong importance	Experience and judgment strongly favor one element over another
7	Very strong importance	One element is favored very strongly over another, it dominance is demonstrated in practice
9	Extreme importance	The evidence favoring one element over another is the highest possible order of affirmation

Whereas 2, 4, 6 and 8 can be used to express intermediate values (equal to moderate, moderate to strong, strong to very strong, and very strong to extreme, respectively).

Flood hazard parameter inputs

The parameters for this study were mostly selected based on their theoretical relevance to flood hazards. There is no strong agreement on which parameters should be used in flood susceptibility assessments (Tehrany et al. 2014). A total of seventeen flood parameter inputs were considered to assess the flood hazards in the Upper Baro-Akobo watershed. DEM, aspect, curvature, rainfall, slope, TRI, flow accumulation, flow direction, SPI, NDVI, population density, distance from river, TWI, drainage density, LULC, soil type, and STI are a few examples.

2.3 Methodology

The general methodology of the study varied depending on the approaches and data used. It also presented the flood frequency analysis by Gumbel, Normal, and Log Pearson Type III distribution, and analysis of the inundation area under different land use changes and depths (Kuriqi and Ardiçlioǧlu 2018). These are categorized into the following two main types: stream flow and HEC-geoRAS data (Geometric data) (Figure 2).

Figure 2 Flowchart of the general methodological framework of the study.

Flood frequency analysis

Daily stream flow data from 1990 to 2009 was collected from the Ministry of Water, Irrigation, and Energy and was used to estimate flood frequency analysis (Getahun and Gebre 2015; Bhagat 2017). The discharge at the Baro-Gambella gauging station was computed for 25-, 50-, and 100-year return periods (Issam et al. 2019). The three flood frequency distributions used were selected to compute the discharges from various return periods and to calculate probability estimates for the likelihood of extreme events (Gumbel distribution, Normal distribution, and Log Pearson Type III distribution).

Flood frequency analysis using Gumbel distribution method

Gumbel distribution is one of the most widely utilized distributions to calculate flood frequency. It is one of the techniques for forecasting extreme hydrological occurrences using maximum discharge data. The discharge is computed using probability distribution functions for various return periods. The expected discharge can be calculated using Equation 5.

(5)

The frequency factor, the standard deviation, and reduced variate are calculated using the following formulae (Equations 6–8). 

	(6)
	(7)
	(8)

Where:

XT	=	discharge for different return period,
	=	mean discharge of the river,
Kt	=	frequency factor,
x	=	discharge of the river,
T	=	return period,
δn-1	=	standard deviation,
Sn	=	reduced standard deviation,
	=	reduced mean,
yt	=	reduced variable, and
N	=	sample size, respectively.

Flood frequency analysis using Normal distribution

Normal distribution is one of the most important distributions in statistical hydrology. It is used to fit empirical distributions with a skewness coefficient close to zero. The magnitude X_T of a hydrologic event, the frequency factor for Normal distribution, and the skewness coefficient are given by the following formula (Equations 9–11).

(9)

 Where:

XT	=	discharge for different return period,
	=	mean discharge of the river, and
	=	standard deviation.

(10)

Where:

Z	=	standard normal variable, and
w	=	value of an intermediate variable.

(11)

Where:

p	=	number of parameters used in fitting the proposed distribution, and
T	=	return period.

Flood frequency analysis using Log Pearson Type III distribution

Log-Pearson Type III distribution is a statistical technique for fitting frequency distribution data to predict the peak flood for a river at a given site. By transforming the data into logarithmic form to the base 10, distribution can be estimated and transformed data were analyzed (Phien and Ajirajah 1984). This is helpful for flood modeling and designing the structures to protect against the largest expected event. For this reason, it is customary to perform the flood frequency analysis using stream flow data for 25-, 50-, and 100-year return periods. However, the Log-Pearson Type III distribution can be constructed using the maximum values for mean daily discharge data (Equations 12–13).

(12)

 Where:

Y	=	mean daily discharge data, and
X	=	flood discharge.

(13)

The frequency factor for Log-Pearson Type III distribution is given by Equations 14 and 15.

(14)

(15)

Where:

Cs	=	coefficient of skewness,
Z	=	normal deviate,
k	=	standardized variant,
	=	standard logarithmic deviation, and
Y	=	mean daily discharge data.

Goodness-of-fit tests

Three different types of goodness-of-fit tests were conducted in this study: Kolmogorov-Smirnov, Anderson-Darling, and Chi-squared.

Kolmogorov-Smirnov test (D)

This test was used to compare an observed sample distribution and theoretical distribution. It was determined using Equation 16.

(16)

Where:

D	=	Kolmogorov-Smirnov test, and
F0(x) and Ft(x)	=	Observed cumulative and theoretical frequency distribution, respectively.

Anderson-Darling test (AD)

This test was used to test if a data came from a population with a specific distribution. It is a modification of the Kolmogorov-Smirnov (K-S) test and gives more weight to the tails than the K-S test (Equation 17). 

(17)

Where:

AD	=	Anderson-Darling Test,
n	=	the sample size,
F(x)	=	CDF for the specified distribution, and
i	=	the ith sample, calculated when arranged in ascending order.

Chi-squared test

This test is also used to compare the observed value to the expected value. It was determined by Equation 18. 

(18)

Where:

Xx2	=	Chi-squared,
Qi	=	Observed value, and
Ei	=	Experimental value.

2.4 Mapping flood inundation areas

HEC-RAS

The Hydrological Engineering Center River Analysis System (HEC-RAS) was developed by the US Army Corps of Engineers Hydrological Engineering Center (USACE) and this model is used for flood mapping. HEC-GeoRAS is a GIS extension consisting of a set of methods, tools, and utilities for preparing GIS river geometry data for input to HEC-RAS and creating the final flood map (Azhar et al. 2018). Triangular Irregular Network (TIN), DEM, and land cover are important input data for the HEC-GeoRAS model for preparing flow geometry (Ardıçlıoğlu et al. 2022). Land use was analyzed to determine the Manning (n) value for each cross-section. HEC-RAS uses river shape and river flow data to create a water table along the river. HEC-RAS offers a user-friendly graphical user interface for data management and the presentation of model results, enabling the numerical solution of the complete one-dimensional Saint-Venant equations and the determination of flow characteristics in unstable situations (Siqueira et al. 2016). The Saint-Venant equations are represented by the conservation of mass and momentum. According to the conservation of mass (Equation 19), for a volume control (USACE), the rate of change in storage must be equal to the net rate of flow into the volume.

(19)

Where:

AT	=	flow area of cross-section,
t	=	time,
Q	=	flow entering the control volume,
x	=	distance along the channel flow, and
ql	=	lateral inflow per unit length.

Conversely, the net rate of momentum entering the volume, plus the sum of external forces acting on the volume control, is equal to the rate of accumulation of momentum (Equation 16), which is also known as the dynamic wave equation (Equation 20).

(20)

Where:

Q	=	flow rate,
V	=	velocity along x-direction,
t	=	time,
x	=	distance along the channel flow,
z	=	water surface elevation,
A	=	area of the cross-section,
g	=	acceleration due to gravity, and
Sf	=	friction slope.

Model performance

The model is calibrated by altering the input parameters and then comparing the simulation’s output runoff values to actual runoff measurements until the goal function is achieved.

Nash-Sutcliffe efficiency (ENS)

Nash-Sutcliffe efficiency helps to judge the fit for the outcome of the model and actual measured hydrograph shapes. The effectiveness of the model is determined by ENS. ENS can have a value between one and negative, with one being the optimal value. Values between 0.80 and 0.90 indicate that the model performs well, while values between 0.90 and 1 indicate that the model performs extraordinarily well (Girmay et al. 2021).

Coefficient of determination (R²)

R² reflects the model approach to recreate the observed value through a given time and for a given time step. R² values vary from 1.0 (best) to 0.0 (Jehanzaib et al. 2020).

Percentage of bias (PBIAS)

The predisposition of an anticipated threshold that is higher/smaller than the measured value is assessed using percentage of bias (PBIAS) (Tufa and Sime 2021). The absolute value of PBIAS should be as low as possible for a well-performing model.

Ratio of root mean square error to observation standard deviation (RSR)

Ratio of root mean square error to observation standard deviation serves as an error index indicator (Assamnew and Tsidu 2020). RSR has a value between zero and one, with the lower value, closer to zero, suggesting better model representation, and one indicating poor model performance (Kwakye 2022).

3 Results and discussion

3.1 Flood frequency analysis

Goodness-of-fit test analysis

Different tests of goodness-of-fit have been employed for the selection of best fit distribution to estimate the discharges for 25-, 50-, and 100-year return periods. In this study, the goodness-of-fit tests used were Kolmogorov-Smirnov, Anderson-Darling, and Chi-squared. The selection of the appropriate flood frequency distribution was based on the comparisons of goodness-of-fit test results. Since goodness-of-fit represents the distance between data and fitted distributions, the distribution with the minimum statistical value has been considered as best fitted with the respective data. (Table 4).

Table 4 Baro River at Gambella gauging station Goodness-of-fit test.

Distribution	K-S at a0.05 = 0.29408			A-D at a0.05 = 2.5018			Chi-squared at a0.05 = 5.9915 and 3.8415 for Gumbel Max and Normal, respectively
	Statistic	Rejected	Rank	Statistic	Rejected	Rank	Statistic		Rejected	Rank
Gumbel Max	0.17307	No	1	0.80699	No	1	1.7161		No	2
Normal	0.24284	No	3	1.4783	No	2	0.14096		No	1
Log Pearson type III	0.17977	No	2	4.506	Yes	3	N/A

Table 4 explains that Gumbel’s distribution method offers a good agreement with the yearly discharge data at the outlet when used for both Kolmogorov-Smirnov test (K-S) and Anderson-Darling test (A-D), whereas when using the Chi-squared test, Normal distribution provides a better fit than the other distributions. The hypothesis in Log Pearson Type III has been rejected because the Anderson-Darling statistic value (4.506) has a value greater than the critical value (2.5018).

The rest of the distributions were not rejected because the Kolmogorov-Smirnov, Anderson-Darling and Chi-squared statistics values were less than the critical values. When comparing the three goodness-of-fit tests, it was observed that the Normal distribution provides a good fit for the selected discharge at the outlet when compared to the goodness-of-fit test results using the minimum statistics (Table 4) best-fitted with the data.

Estimation of expected discharges using the best fit distribution

The expected discharges for the return periods of 25, 50, and 100 years were estimated using the best-fit distribution from the goodness-of-fit test results. Normal distribution was selected as a good fit for selected discharge data because it has the minimum statistical value in the Chi-squared test since it expresses the overall best fit in the outlet (Table 5) (Haile et al. 2013; Hu et al. 2022).

Table 5 Normal distribution method for expected discharges in the watershed.

Return periods (y)	p	w	KT	Mean	Standard deviation	XT (m3/s)
25	0.04	2.537272482	1.751076531	1270	268.17	1739.586
50	0.02	2.797149623	2.054188589	1270	268.17	1820.872
100	0.01	3.034854259	2.326785333	1270	268.17	1893.974

Where p is probability of exceedance, and w is intermediate variable. When, p > 0.5, 1-p is substituted, and K_T value computed gives a negative sign. The expected value of rainfall/discharge is 'X_T', at return period T.

Multi-criteria analysis (AHP)

AHP is a means of breaking down the problem into a hierarchy of sub-problems that can be more simply understood and subjectively analyzed. (Taherdoost 2017) In this study, an AHP Excel template (version 2018-09-15) was used for the pair-wise comparisons and estimation of the weights and CR value (Youssef and Hegab 2019). The pairwise comparison matrix developed was used in the AHP to estimate the weight of each factor in the selected criteria. A number scale is used to determine how many times one factor is more important than the other factors after comparison (Taherdoost 2017). The objective of this section is to develop a flood hazard map of the study area with the factors used being rainfall, land use, distance from river, DEM, slope, TRI, TWI, soil type, NDVI, and population density.

Pairwise comparison

The matrix was created and estimated using flood influencing factors based on the Eigenvector principle and the weighted values. Each flood criterion influencing factors in the Upper Baro-Akobo basin are compared to one another, where one or two factors needs to have more influence than the others, basing on the causes and flood risk. If factor A has the same influence as factor B, the pair receives an index of 1. If factor A has more influence than factor B, the pair receives an index from 2 to 9. After the construction of a pairwise comparison, the matrix is normalized so that each column can be added up to one. This is followed by the mean across the rows being computed to give the priority vector or the criteria weights (W).

Tables 6 and 7 show an example of a normalized matrix and weights of flood hazard influence. The CR value for the flood hazard factors is 7.2%, which is less than 10%, indicating an acceptable consistency level (Taherdoost 2017). The overall weights of each factor for flood hazard influence are presented in Table 6 and Table 7. The weights of the flood hazard factors are rainfall (0.1567), land use (0.1501), distance from river (0.1284), DEM (0.1062), slope (0.0976), TRI (0.0928), TWI (0.0928), soil type (0.063), NDVI (0.0588) and population density (0.0537) (Table 8). From the calculated weights of all flood influencing factors, it can be seen that rainfall has the maximum weight, followed by land use, and distance to the river. This indicated that rainfall, land use, and distance to the river have a higher influence and contribution to flooding events in the study area than the other factors.

Table 6 Pairwise comparison matrix of flood influence factors.

Sl. No.	Land use	DEM	Slope	Distance	Soil type	NDVI	RF	Pd	TRI	TWI
Land use	1	3	1	1	3	5	1	3	1	1
DEM	0.33	1	1	1	2	3	1	3	1	1
Slope	1	1	1	1	3	1	0.5	1	1	1
Distance	1	1	1	1	3	2	2	3	1	1
Soil type	0.33	0.5	0.33	0.33	1	1	0.33	3	1	1
NDVI	0.2	0.33	1	0.5	1	1	0.2	1	1	1
RF	1	1	2	0.5	3	5	1	9	1	1
Pd	0.33	0.33	1	0.33	0.33	1	0.11	1	1	1
TRI	1	1	1	1	1	1	1	1	1	1
TWI	1	1	1	1	1	1	1	1	1	1
Sum (col)	7.2	10.167	10.333	7.667	18.333	21	8.144	26	10	10

Table 7 Normalized matrix for flood hazard factors.

Factors	Land use	DEM	Slope	Distance	Soil type	NDVI	RF	Pd	TRI	TWI
Land use	0.139	0.295	0.097	0.130	0.164	0.238	0.123	0.115	0.1	0.1
DEM	0.046	0.098	0.097	0.130	0.109	0.143	0.123	0.115	0.1	0.1
Slope	0.139	0.098	0.097	0.130	0.164	0.048	0.061	0.038	0.1	0.1
Distance	0.139	0.098	0.097	0.130	0.164	0.095	0.246	0.115	0.1	0.1
Soil type	0.046	0.049	0.032	0.043	0.055	0.048	0.041	0.115	0.1	0.1
NDVI	0.028	0.033	0.097	0.065	0.055	0.048	0.025	0.038	0.1	0.1
RF	0.139	0.098	0.194	0.065	0.164	0.238	0.123	0.346	0.1	0.1
Pd	0.046	0.033	0.097	0.043	0.018	0.048	0.014	0.038	0.1	0.1
TRI	0.139	0.098	0.097	0.130	0.055	0.048	0.123	0.038	0.1	0.1
TWI	0.139	0.098	0.097	0.130	0.055	0.048	0.123	0.038	0.1	0.1
Sum	1	1	1	1	1	1	1	1	1	1

Table 8 Weights of flood hazard factors.

Factors	Weights	Weight (%)
Land use	0.1501	15.01
DEM	0.1062	10.62
Slope	0.0976	9.76
Distance	0.1284	12.84
Soil type	0.063	6.30
NDVI	0.0588	5.88
RF	0.1567	15.67
Pd	0.0537	5.37
TRI	0.0928	9.28
TWI	0.0928	9.28
Total	1	100.00

3.2 Flood hazard mapping

Flood hazard mapping is used to determine the region’s possibility of a flood occurrence for a certain return period (Trinh and Molkenthin 2021). The results of the AHP method (pair-wise comparison) for every subclass of the ten thematic maps and their corresponding weights were assigned. The flood hazard map was developed by integrating ten thematic maps of the study area and is shown in Figure 3. Based on the results, the flood hazard map was divided into five classes: very low, low, medium, high, and very. It indicated that very low covers 0.2% (32.2 km²), low covers 26.1% (5369.5 km²), medium covers the maximum area of 51.1% (10515.3 km²), high covers 21.6% (4442.4 km²), and very high covers 1% (202.9 km²) area of the watershed (Table 9).

Figure 3 Flood hazard map of the Upper Baro-Akobo watershed.

Table 9 Hazard map area percentages for the Upper Baro-Akobo watershed.

Hazard class	Area (km2)	Area (%)
Very low	32.2	0.2
Low	369.5	26.1
Medium	10515.3	51.1
High	4442.4	21.6
Very high	202.9	1.00
Total	20562.2	100

During the wet season, the runoff volume is much lower as the river levels decline, and the flows are smallest from October to May. According to Figure 4, river flow increases from June to September (main rainy season) and decreases from October to May. During this period (June to September), rivers in the region carry 75% to 85% of the total annual discharge. The maximum flow occurs in September, and the minimum flow occurs in March.

Figure 4 Mean monthly flow of Baro-Akobo basin.

3.3 Flood inundation areas mapping

Analysis of flood inundation areas under land use change

As presented in Table 10 and Figure 5, flooded vegetation, annual crop, built area, and forest, both in the upstream and downstream of the upper Baro-Akobo watershed, were exposed to overflow by 25-, 50-, and 100-year return periods. The annual crop was found to be the most vulnerable to flooding, followed by the built area, flooded vegetation, and forest. The flood inundation areas under different land use changes for 25-, 50-, and 100-year return periods, respectively were 390.95 km², 422.76 km², and 453.97 km², in which annual crop covers 446.2 km², built area 404.4 km², flooded vegetation 323.3 km², and forest area 93.58 km², respectively (Table 7). Figure 6 shows the flood inundation scenarios, depth under different land use classifications. The most affected areas by flood inundation are annual crops, build area, vegetation, and forest. The annual crop is the most affected area, and forests are less affected as compared to others.

Table 10 LULC inundation area (A) for 25-, 50- and 100-year return periods.

LULC	25-year		50-year		100-year		Total
	A (km2)	A (%)	A (km2)	A (%)	A (km2)	A (%)	Sum (km2)
Flooded vegetation	99.9	25.55	109.3	25.85	114.3	25.18	323.5
Annual crop	138.7	35.48	151.8	35.91	155.7	34.30	446.2
Built area	127.4	32.59	132.7	31.39	144.3	31.78	404.4
Forest	24.95	6.38	28.96	6.85	39.67	8.74	93.58
Total	390.95	100	422.76	100	453.97	100	1267.68

Figure 5 Return periods of flood inundation areas for different land uses and depths,  (a) and (b) for 25 years, (c) and (d) for 50 years, and (e) and (f) under 100 years.

Analyses of factors influencing flooding

To assess the flood hazard map for the study area, the selection of the most flood-influencing factors was very important. A total of seventeen flood influencing factors were considered to assess the flood hazards in the study area. Their data was either freely accessible online or obtained through spatial analysis of existing datasets and satellite images. The data was mostly processed using multi-criteria analysis (weighted overlay method) integrated with ArcGIS. Following the processing of all the datasets individually, they were combined using the weighted overlay technique in ArcGIS. Thematic maps in Figure 6 show the spatial distributions of these factors.

Figure 6 Flood influencing factors: (a) DEM, (b) aspect, (c) curvature, (d) rainfall, (e) slope, (f) TRI, (g) flow accumulation, (h) flow direction, (i) SPI, (j) NDVI, (k) population density, (l) distance from river, (m) TWI, (n) drainage density, (o) LULU, (p) soil type, and (q) STI.

Analysis of flood inundation areas under different depths

The inundation depth for 25-, 50-, and 100-year return periods varied from 0–2.6, 0–2.9, and 0–3.2 m at the upstream and downstream of the river, respectively (Table 11). The flood plain begins to move water downstream mostly through a shorter path than the main waterway as the depth increases. The area under the water increased from 1.9 to 3.2 m, increasing the flooding severity. This demonstrates that the high risk of flood increases during a 100-year return period. Due to the difference in elevation and increasing slope, water flows directly from upstream to downstream as the rain intensifies. This results in more flood water accumulating in the downstream region than the upstream, increasing the risk of flooding.

Table 11 Flood depths for 25-, 50-, and 100-year return periods.

Return period (year)	Flood depth range (m)
25	0-0.5, 0.6-1.3, 1.4-1.8, 1.9-2.1, 2.2-2.6
50	0-0.5, 0.6-0.9, 1.0-1.4, 1.5-1.7, 1.8-2.9
100	0-0.5, 0.6-1.1, 1.2-1.4, 1.5-1.8, 1.9-3.2

3.4 Validation of HEC-RAS model

In order to validate the generated flood hazard map, historical records related to the past floods were compared with the developed maps (Sabri and Yeganeh 2014). The flood inundation mapping was also validated by comparing the model results with the observed stream flow data. The predicted performance of the HEC-RAS model was evaluated using Nash-Sutcliffe efficiency (NSE = 0.97) and Coefficient of determination (R² = 0.79). Depending on the values of the NSE and Coefficient of determination, the results range from good to very good (Table 12). Hence, there is a close agreement between observed and predicted data at the Baro-Akobo Gambella gauging station (Figure 7) verifying that the model outputs are also fitted to the known data.

Table 12 Goodness-of-fit model results (Wang et al. 2017).

S.N.	Goodness-of-fit	NSE	R2
1	Very good	0.75 < NSE < 1	R2  ³ 0.85
2	Good	0.65< NSE < 0.75	0.75< R2 £ 0.85
3	Satisfactory	0.5< NSE < 0.65	0.6< R2 £ 0.75
4	Unsatisfactory	NSE < 0.5	R2 < 0.6

Figure 7 Comparison of observed and predicted data of Upper Baro-Akobo river basin.

4 Conclusion

Floods have been a major problem in the Upper Baro-Akobo basin, particularly in the downstream areas, leading to economic losses and deaths. Flood inundation mapping under different land uses and depths, as well as a flood frequency analysis, were estimated for peak flood using a HEC-RAS model and EasyFit software. The respective analyses were calculated using the Normal distribution method because it has a minimum statistical value when compared with other distributions. The affected areas were modeled using HEC-RAS for various return periods, which depicts that annual crop cover areas as maximum, in comparison to built-up areas, flooded vegetation, and forest. Flood inundation mapping under different land uses shows that annual crops were most exposed to flooding, followed by the built-up areas, flooded vegetation, and forest. This indicated that more flood water gathered in the downstream region than in the upstream, increasing the risk of flooding as the depth increases.

Depending on the values of the Nash-Sutcliffe efficiency (NSE = 0.97) and Coefficient of determination (R² = 0.79), the HEC-RAS model gives an acceptable result between an observed and predicted water level. The flood hazard map with respect to LULC and depth considerations, showed that about 21.6% and 1% of the total study area have a high to very high risk of flooding, respectively. The land's topography, the amount of water drainage, and the depth and form of the river stream all had an impact on this disparity. Because of the surge in water arriving from the feeding areas and the drop in ground level, the towns near the entrance of the basin were severely damaged by the severity of the flood. The government should implement flood protection measures in places which are considered to be high to very high risks, and similar related studies should consider such research findings as an input for their study.

Declaration of interests

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

References

1. Ardıçlıoğlu, M., and A. Kuriqi. 2019. “Calibration of channel roughness in intermittent rivers using HEC-RAS model: Case of Sarimsakli Creek, Turkey.” SN Applied Sciences 1, 1–9.
2. Ardıçlıoğlu, M., A.M.W.M. Hadi, E. Periku, and A. Kuriqi. 2022. “Experimental and numerical investigation of bridge configuration effect on hydraulic regime.” International Journal of Civil Engineering 20 (8), 981–991.
3. Asadi, A., and F. Boostani. 2013. "Application of HEC-HMS for Flood Forecasting in Kabkian Basin and Delibajak Subbasin in Iran." IOSR Journal of Engineering, 03 (09), 10–16. https://doi.org/10.9790/3021-03931016
4. Assamnew, A.D., and G.M. Tsidu. 2020. "The performance of regional climate models driven by various general circulation models in reproducing observed rainfall over East Africa." Theoretical and Applied Climatology 142 (3–4).
   https://doi.org/10.1007/s00704-020-03357-3
5. Azhar, H., S. Mohammed, and L.A. Mohammed. "Simulation of Floods in Delhi Segment of River Yamuna Using HEC-RAS." American Journal of Water Resources 2018, 6, 162–168.
6. Bhagat, N. 2017. "Flood frequency analysis using Gumbel's distribution method: A case study of Lower Mahi Basin, India." Journal of Water Resources and Ocean Science 6 (4), 51–54. https://doi.org/10.11648/j.wros.20170604.11
7. Brunner, G.W. 2016. “HEC-RAS, River Analysis System 2D Modeling User’s Manual Version 5.0.” Hydrologic Engineering Center, Institute for Water Resources, US Army Corps of Engineers, California. https://www.hec.usace.army.mil/software/hec-ras/documentation/HEC-RAS 5.0 2D Modeling Users Manual.pdf
8. Getahun, Y.S., and S.L. Gebre. 2015. "Flood Hazard Assessment and Mapping of Flood Inundation Area of the Awash River Basin in Ethiopia using GIS and HEC-GeoRAS/HEC-RAS Model." Journal of Civil & Environmental Engineering 05, 179.
   https://doi.org/10.4172/2165-784x.1000179
9. Girmay, G., A. Moges, and A. Muluneh. 2021. "Assessment of Current and Future Climate Change Impact on Soil Loss Rate of Agewmariam Catchment, Northern Ethiopia." Air Soil and Water Research 14.
   https://doi.org/10.1177/1178622121995847
10. Haile, A.T., K. Kusters, and N. Wagesho. 2013. "Loss and damage from flooding in the Gambela region, Ethiopia." International Journal of Global Warming 5 (4), 483–497.
    https://doi.org/10.1504/IJGW.2013.057290
11. Hou, J., H. Han, Z. Li, K. Guo, and Y. Qin. 2018. “Effects of Morphological Change on Fluvial Flood Patterns Evaluated by a Hydro-Geomorphological Model.” Journal of Hydroinformatics 20 (3), 633–44. https://doi.org/10.2166/hydro.2018.142
12. Hou, J., X. Li, G. Bai, X. Wang, Z. Zhang, L. Yang, Y. Du, Y. Ma, D. Fu, and X. Zhang. 2021. “A Deep Learning Technique Based Flood Propagation Experiment.” Journal of Flood Risk Management 14 (3), 1–13. https://doi.org/10.1111/jfr3.12718
13. Hu, R., Z-R. Tang, R., E.Q. Wu, Q. Mo, R. Yang, and J. Li. 2022. "RDC-SAL: Refine distance compensating with quantum scale-aware learning for crowd counting and localization." Applied Intelligence 52 (12), 14336–14348.
14. Issam, B., A. Khadzhidi, L. Kravchenko, Y. Tsarev, L. Groshev, and O. Polushkin. 2019. "Flood risk management in Allala River (Algeria) using flood frequency analysis and hydraulic modeling." E3S Web of Conferences 135.
    https://doi.org/10.1051/e3sconf/201913501093
15. Jehanzaib, M., M.N. Sattar, J.H. Lee, and T.W. Kim. 2020. "Investigating the effect of climate change on drought propagation from meteorological to hydrological drought using multi-model ensemble projections." Stochastic Environmental Research and Risk Assessment 34, 7–21. https://doi.org/10.1007/s00477-019-01760-5
16. Karamma, R., and M.S. Pallu. 2018. “Comparison of Model Hidrograf Synthetic Units (HSS) with the Model of Hidrograf Observations on DAS Jeneberang Gowa Regency, Indonesia.” International Journal of Innovative Science and Research Technology 3 (2), 617–23.

17. Kontos, T.D., D.P. Komilis, and C.P. Halvadakis. 2005. "Siting MSW landfills with a spatial multiple criteria analysis methodology." Waste Management 25 (8), 818–832. 
    https://doi.org/10.1016/j.wasman.2005.04.002 

18. Kundzewicz, Z.W., S. Kanae, S.I. Seneviratne, J. Handmer, N. Nicholls, P. Peduzzi, R. Mechler, et al. 2014. "Le risque d’inondation et les perspectives de changement climatique mondial et régional." Hydrological Sciences Journal 59 (1), 1–28.
    https://doi.org/10.1080/02626667.2013.857411
19. Kuriqi, A., and M. Ardiçlioǧlu, 2018. "Investigation of hydraulic regime at middle part of the Loire River in context of floods and low flow events." Pollack Periodica 13 (1), 145–156.

20. Kwakye, S.O. 2022. "Quantification of the hydrological consequences of climate change in a typical West African catchment using flow duration curves." Journal of Water and Climate Change 13 (1), 26–42. https://doi.org/10.2166/wcc.2021.147

21. Mangini, W., A. Viglione, J. Hall, Y. Hundecha, S. Ceola, A. Montanari, M. Rogger, et al. 2018. "Detection of trends in magnitude and frequency of flood peaks across Europe." Hydrological Sciences Journal 63 (4), 493–512.
    https://doi.org/10.1080/02626667.2018.1444766
22. Mostofi Zadeh, S., D.H. Burn, and N. O’Brien. 2020. ”Detection of trends in flood magnitude and frequency in Canada." Journal of Hydrology: Regional Studies 28, 100673. https://doi.org/10.1016/j.ejrh.2020.100673

23. Ogato, G.S., A. Bantider, K. Abebe, and D. Geneletti. 2020. "Geographic information system (GIS)-based multicriteria analysis of flooding hazard and risk in Ambo Town and its watershed, West Shoa zone, Oromia Regional State, Ethiopia." Journal of Hydrology: Regional Studies 27, 100659. https://doi.org/10.1016/j.ejrh.2019.100659

24. Ouma, Y.O., and R. Tateishi. 2014. "Urban flood vulnerability and risk mapping using integrated multi-parametric AHP and GIS: Methodological overview and case study assessment. Water 6 (6), 1515–1545. 

25. Phien, H.N., and T.J. Ajirajah. 1984. "Applications of the Log Pearson Type-3 distribution in hydrology." Journal of Hydrology 73 (3–4), 359–372.
    https://doi.org/10.1016/0022-1694(84)90008-8
26. Praveen, S.S., V.S. Babu, A. Vijayan, and S. Babu. 2020. “Numerical Modelling of Flood Inundation and Run-Up along the Banks of Pamba River, South West Coast of India.” Journal of Advanced Research in Alternative Energy, Environment and Ecology 07 (01), 8–14.
27. Romali, N.S. 2018. “Application of HEC-RAS and Arc GIS for Floodplain Mapping in Segamat Town, Malaysia.”  GEOMATE Journal 15 (47), 7–13.
    https://doi.org/10.21660/2018.43.3656
28. Rosser, J.F., D.G. Leibovici, and M.J. Jackson. 2017. "Rapid flood inundation mapping using social media, remote sensing and topographic data." Natural Hazards 87 (1), 103–120. https://doi.org/10.1007/s11069-017-2755-0
29. Saaty, T.L. 1990. "How to make a decision: The analytic hierarchy process." European Journal of Operational Research 48 (1), 9–26. 
    https://doi.org/10.1016/0377-2217(90)90057-I
30. Sabri, S., and N. Yeganeh. 2014. "Flood vulnerability assessment in Iskandar Malaysia using multi-criteria evaluation and fuzzy logic." Research Journal of Applied Sciences, Engineering and Technology 8, 1794–1806.
31. Son, K.I., and W. Jeong. 2019. "Numerical simulation of flood inundation in a small-scale coastal urban area due to intense rainfall and poor inner drainage." Water 11 (11), 2269. https://doi.org/10.3390/w11112269
32. Siqueira, V.A., M.V. Sorribas, J.M. Bravo, W. Collischonn, A.M.V. Lisboa, and G.G.V. Trinidad. 2016. "Real-time updating of HEC-RAS model for streamflow forecasting using an optimization algorithm." Revista Brasileira de Recursos Hidricos 21 (4), 855–870.
33. Taherdoost, H. 2017. "Decision Making Using the Analytic Hierarchy Process (AHP); A Step by Step Approach." International Journal of Economics and Management Systems 2. Available at SSRN: https://ssrn.com/abstract=3224206

34. Tehrany, M.S., M-J. Lee, B. Pradhan, M.N. Jebur, and S. Lee. 2014. "Flood susceptibility mapping using integrated bivariate and multivariate statistical models." Environmental Earth Sciences 72, 4001–4015.

35. Trinh, M.X., and F. Molkenthin. 2021. "Flood hazard mapping for data-scarce and ungauged coastal river basins using advanced hydrodynamic models, high temporal-spatial resolution remote sensing precipitation data, and satellite imageries." Natural Hazards 109, 441–469. https://doi.org/10.1007/s11069-021-04843-1
36. Tufa, F.G., and C.H. Sime. 2021. "Stream flow modeling using SWAT model and the model performance evaluation in Toba sub-watershed, Ethiopia." Modeling Earth System and Environment 7, 2653–2665. https://doi.org/10.1007/s40808-020-01039-7
37. Vijayan, A., S.S. Praveen, V.S. Babu, and S. Babu. 2021. “Simulation Studies of Flood Along the Periyar River Basin, Central Kerala, India.” Journal of Advanced Research in Geo Sciences & Remote Sensing, 8 (2), 1–7.
38. Wang, Y., Z. Liu, and H. Liao. 2017. "Improving the performance of GIS polygon overlay computation with MapReduce for spatial big data processing Improving the performance of GIS polygon overlay computation with MapReduce for spatial big data processing." Cluster Computing. https://doi.org/10.1007/s10586-015-0428-x
39. Wijayanti, P., X. Zhu, P. Hellegers, Y. Budiyono, and E.C. van Ierland. 2017. “Estimation of River Flood Damages in Jakarta, Indonesia.” Natural Hazards 86 (3), 1059–79. 
    https://doi.org/10.1007/s11069-016-2730-1
40. Yang, Q., T. Liu, J. Zhai, and X. Wang. 2021. “Numerical Investigation of a Flash Flood Process That Occurred in Zhongdu River, Sichuan, China.” Frontiers in Earth Science 9 (July), 1–11. https://doi.org/10.3389/feart.2021.686925
41. Youssef, A.M., and M.A. Hegab. 2019. “Flood-hazard assessment modeling using multicriteria analysis and GIS: a case study-Ras Gharib Area, Egypt.” In Spatial modeling in GIS and R for earth and environmental sciences (pp. 229–257), Elsevier.
    https://doi.org/10.1016/B978-0-12-815226-3.00010-7