Introduction

Flood is a hydrological condition characterized by high peak discharge that could lead to inundation of lands in adjacent areas, affecting people’s lives and livelihoods (Nkeki et al. 2013). Heavy rainfall for long days on the upstream causes most rivers to swell and overflow their courses and sinking the surrounding flat fields (Bhatt et al. 2014). The amount of rainfall, intensity, and duration in the upstream area causes flooding in down streams. Therefore, floods are related to extreme precipitation and spatial features in the catchments (Boudiaf et al. 2020; Trenberth 2008).

Extensive deforestation and farmland expansion, together with urbanization, raise sensitivity to river peak overflow. Land use changes such as urbanization across the catchment area trigger flood occurrences (Barasa and Perera 2018). This reduction of surface roughness could increase overland flow velocity and floodplain flow rates (Hadjimitsis 2010).

Unpredicted and temporal flash floods are related to rivers in the land areas, where heavy rainfall can change them into destructive torrents within a short period (Obeta and Ochege 2018; Ologunorisa 2004). Since the local climate varies significantly in Ethiopia, flooding incidents also vary in time and space (Desalegn and Mulu 2021; Getahun and Gebre 2015). Flash and river floods are common in Ethiopia during the rainy season between June and September (Akola et al. 2019). Although floods in many parts of Ethiopia are a seasonal phenomenon, the magnitude of the current flooding is unprecedented (Desalegn and Mulu 2021; Bishaw 2012). The recent incident in Dire Dawa city was typical of a flash flood. Furthermore, most flood-related disasters in Ethiopia were attributed to rivers that overflow or burst their banks and inundate downstream plane lands. The situation resulted in many human deaths, displacement, suffering, property loss, and crop damage. The most affected areas are Dire Dawa, South Omo zone of Southern Nations Nationalities and People Region, and parts of Amhara, Oromia, Gambella, Somali, Afar, and Tigray regions (Weldegebriel and Amphune 2017). The Dire Dawa city flood event in 2006 has caused severe direct and indirect damages to social, infrastructure, and economic sectors. The monetary or infrastructural damages also caused 256 people to die, 244 missed, and 15,000 people displaced from their dwellings. This problem is more acute in lowland areas under substantial environmental degradation due to frequent drought. For a large part, this is due to heavy rains falling for long days on the upstream highlands (Alemu and Legesse Belachew 2011; Assegid 2007).

Though flood risk occurrence is a combination of natural and anthropogenic factors, it calls for a better understanding of its spatial extent (Danumah et al. 2016). Flooding is a dynamic process and exhibits spatial characteristics because floods occur at a particular location at which various contributing factors for the event exist (Wu et al. 2018). Flood risk is the product of flood damage and the probability of its occurrence (Dennis Mileti 1999). Therefore, a flood risk map is a crucial information to identify the most vulnerable areas to flooding and estimate the number of people affected by floods in a particular area (Eleni et al. 2011). The extent of flood damage depends not only on the flood characteristics, but also on the inundated site’s vulnerability. For the same flood, a more vulnerable area experiences more serious flood damage (De Risi et al. 2020).

The advancement of GIS and remote sensing technologies is enabling scholars to investigate flood hazards and vulnerability. Different remote sensing and GIS approaches have been used to analyze and map flood (Elkhrachy 2015; Ghani et al. 2010; Shaaban et al. 2021; El Bastawesy et al. 2009). Moreover, GIS multi-criteria decision analysis (Feloni et al. 2020), integrated multi-parametric analytic hierarchy process (AHP) and GIS (Ouma and Tateishi 2014), spatial analysis using weighted overly (Desalegn and Mulu 2021; Rimba et al. 2017), frequency ratio and statistical index (Cao et al. 2016), frequency ratio model (Sarkar and Mondal 2020; Rahmati et al. 2016; Tehrany et al. 2014; Lee et al. 2012), AHP and support vector machine (SVM) (Xiong et al. 2019), and morphometric analysis (Ogarekpe et al. 2020) were used for analyzing and mapping flood hazard, vulnerability, and risk. Other common methods such as artificial neural network (Kia et al. 2012), AHP (Chen et al. 2011), fuzzy analytical hierarchy process (FAHP) (Papaioannou et al. 2015), SVM (Tehrany et al. 2015), and decision tree (Tehrany et al. 2013) were also used for flood investigation.

Several parameters were used by many researchers worldwide to understand and analyze flood hazards and vulnerability and map its risk zone. Factors such as elevation (Desalegn and Mulu 2021; Sarkar and Mondal 2020; Rimba et al. 2017; Cao et al. 2016; Niyongabire et al. 2016; Ouma and Tateishi 2014), aspect (Niyongabire et al. 2016), slope (Desalegn and Mulu, 2021; Sarkar and Mondal 2020; Rimba et al. 2017; Cao et al. 2016; Niyongabire et al. 2016; Elkhrachy 2015; Kazakis et al. 2015; Ouma and Tateishi 2014), LULC (Desalegn and Mulu 2021; Sarkar and Mondal 2020; Barasa and Perera 2018; Rimba et al. 2017; Cao et al. 2016; Elkhrachy 2015; Ouma and Tateishi 2014), soil (Desalegn and Mulu 2021; Sarkar and Mondal 2020; Kusmiyarti et al. 2018; Rimba et al. 2017; Cao et al. 2016; Lincoln et al. 2016; Elkhrachy 2015; Ouma and Tateishi 2014), drainage density (Desalegn and Mulu, 2021; Sarkar and Mondal 2020; Rimba et al. 2017; Elkhrachy 2015; Ouma and Tateishi 2014; Singh and Pandey 2021; Pallard et al. 2009), rainfall (Desalegn and Mulu, 2021; Sarkar and Mondal 2020; Rimba et al. 2017; Kazakis et al. 2015; Ouma and Tateishi 2014), standardized precipitation index (SPI) (Cao et al. 2016), road network (Sarkar and Mondal 2020; Ouma and Tateishi 2014), geology (Cao et al. 2016), surface roughness, distance to river, runoff (Elkhrachy 2015), and population density (Sarkar and Mondal 2020) were used for flood analysis.

The vulnerability to flooding in the North Wollo Zone results from the topography of the district, adjacent districts’ watershed direction, the settlement of the households, and poor community management of the watersheds and riverbanks. The climate change-induced excess runoff from these degraded and bare mountainous watershed areas creates powerful erosive forces and hydrodynamic pressures, severely affecting the livelihood of the lowland households. Alawuha river is a persistent river with a significant catchment area (101,988.51 ha). It originates in the western mountainous regions (in the Guba Lafto Woreda) and drains towards Afar Region. An overflow of Alawuha Rivers has a devastating impact on people who live in the watershed. The 2016 flood in the watershed was exceptional because it caused the death of many and the destruction of settlements and infrastructures. Therefore, creating digital terrain maps and flood vulnerability maps of the Alawuha watershed is essential to show the magnitude of flood hazard, vulnerability, and risk zones.

Scholars like Bishaw (2012), Chikoto et al. (2013), and Sena and Michael (2006) conducted a few types of research for flood-related issues in Ethiopia. However, the previous studies did not sufficiently describe the distribution and magnitude of flood-related disasters that happened in the country due to spatiotemporal variation of flooding events. Most of the tasks related to flooding in Ethiopia, specifically in the study area, were based on the administrative boundary. This study aimed to display how land use and land cover change coupled with climate change causes flooding and its effect on socio-economic and biophysical environments. It is also aimed to identify the practical measures of flood risk assessment and vulnerability mapping using GIS and remote sensing application platforms unlike previous flood studies in Ethiopia that use administrative boundaries as unit of analysis; this study is based on watershed unities. The study investigates each criterion in detail at the micro-watershed level to highlight the minimum flood aggravating factors for watershed management practices. The objectives of this study are (1) to analyze LULC change; (2) to map flood hazard, vulnerability, and risk zones in the Alawuha watershed; and (3) to analyze flood return periods and exceedance probability in the study area. Factors such as LULC, soil type, slope, elevation, drainage density, rainfall distribution, distance to streams, and population density are used as main predictive and determinant factors to map analyze flood vulnerability. The study also used rainfall frequency to estimate exceedance probability of flooding and the possible flooding return periods in the watershed. Even though this study applies multiple criteria and recognized indexes, it’s limited in using up-to-date analysis systems like aridity index (AI) and resent machine learning programs. Furthermore, the study will be more accurate and precise through using Copernicus products like sentinel and other open science programs.

Materials and methods

Description of the study area

Alawuha watershed is geographically located between 10°40′ to 12°00′ N and 37°40′ to 40°00′ E (Fig. 1). It has an estimated area of 101,988.51 ha. The altitude of the watershed varies from 800 to 3802 m above mean sea level (amsl).

Fig. 1
figure 1

Study area map

Full size image

The area is part of the north-central plateau which is often extensively eroded and dissected by the deep gorges. The eastern boundary of the district is part of the southwestern block of the Afar depression. The elevation varies from 800 to 3200 m in the western plateau to 800 to 1500 m amsl in the low eastern land of the Afar depression. The southeast monsoon and high-elevation westerly’s winds are highly influential for the distribution of rainfall; the total annual rainfall is 1000 to 1100 mm year−1 (Loakes et al. 2018). According to the temperature data collected by Sirinka Agriculture Research Center, the mean monthly temperature of the study area ranges between 21 and 25 ℃. A bi-modal rainfall is the most dominant in all parts of the watershed. The short rainy season (Belg) occurs between February and May, and the long rainy season (Kiremit) occurs between June and September. In most cases, the highland areas with elevation values vary from 3200 to 3700 m (Dega) which mainly depend on Belg rain. The elevation values range from 1500 to 3200 m (Woyina Dega), and the elevation values vary from 500 to 1500 m; Kolla areas are rain-dependent from October to January (Meher) for crop production (Mengistie and Kidane 2016; Ali 2002).

Types and sources of data

The study used Landsat images (Landsat 5 TM, 7 ETM + , and 8OLI) to extract LULC types of the watershed. To make the appropriate analysis, Landsat 5 TM 1985, Landsat 7 ETM + 2000, and Landsat 8 OLI 2016 images were selected. The rationale used for the selection of these periods was due to significant LULC changes resulting from simultaneous occurrences of drought and flood in the study area. Besides, during these three study periods, a significant change of LULC was observed in preliminary survey analysis mainly in 2000 and 2016. Cloud-free Landsat images were downloaded freely from the US Geological Survey (USGS) portal (http://glovis.usgs.gov). GPS data were used for training sites, ground verification for image classification, and accuracy assessment control points. A minimum of 30 training sites per class is recommended to classify and validate a satellite image (Congalton and Green 2008). The study used a total of 500 GPS points to validate LULC classification.

Rainfall data from 2000 to 2016 was obtained from the Ethiopian National Metrological Agency (NMA) Kombolcha station. It is used to develop a rainfall distribution map and compute the return period of flooding in the watershed. Furthermore, digital elevation model (DEM) with the 30-m horizontal resolution was obtained from the USGS webpage (http://glovis.usgs.gov). The DEM data is used to extract elevation, slope, drainage density, flow direction, stream network, flow accumulation, and watershed map of the study area. The soil data obtained from the Amhara Regional Agricultural Office at a scale of 1:20,000 was also used for flood analysis since it affects the hydrological characteristics. The study area population was obtained from the Central Statistics Agency (CSA), Ethiopia, to develop a flood vulnerability map since it is one of the affected elements by the flood.

Parameter selection

LULC, soil type, elevation, slope, drainage density, rainfall distribution and frequency (amount), distance to streams, and population density parameters were selected to investigate flood hazard, vulnerability, and risk in the study watershed. These parameters were selected according to the literature presented in Table 1 and the nature of the study area and available data.

Table 1 Parameters used for the study
Full size table

Flood factor development

Image classification

The study employed digital image processing techniques such as an atmospheric correction to improve the quality of the images and sub-setting to extract images of the study area. A maximum likelihood supervised algorithm was used to classify the Landsat images. During classification, forest, built-up area, agricultural land, bare land, and riverbank were identified carefully with intensive field observation using local indigenous knowledge to easily interpret and reduce the effects of spatial autocorrelation and redundancy. A confusion matrix was used to validate the image classification and accuracy assessment report for overall accuracy and kappa statistics extracted from the matrix. The images were processed and interpreted in ERDAS Imagine 2015 environment. The LULC type is one of the main factors (Desalegn and Mulu 2021; Sarkar and Mondal 2020; Barasa and Perera 2018; Cao et al. 2016) which is considered in this study because flood vulnerability varies with the variation of surface type and disaster does not equal damage.

Surface analysis

The study used surface analysis to extract slope and elevation thematic layers from DEM using ArcGIS 10.5. The DEM data were preprocessed to remove errors such as sink from the data and enhance its quality. Therefore, fill analysis was applied to fill a hole in the DEM data. The DEM is the most common source of data for elevation and slope (Desalegn and Mulu 2021; Sarkar and Mondal 2020; Rimba et al. 2017; Niyongabire et al. 2016).

Hydrological analysis

Flow direction, flow accumulation, stream network, drainage density, and watershed were developed using hydrological analysis (Desalegn and Mulu 2021; Sarkar and Mondal 2020; Rimba et al. 2017; Elkhrachy 2015). Watershed delineation analysis was done to extract streams from DEM-based watershed snap pour points. The DEM is used as an input rater to extract surface elevation raster (fill DEM). Flow direction raster was extracted from surface raster, and flow accumulation was extracted from flow direction. Stream raster was created from flow accumulation using a conditional standard query language (SQL) in the Raster Calculator tool. The Strahler stream ordering method from raster and flow direction was employed to create stream order (Fig. 2). Then, snap pour points were digitized to identify river outlets. Using digitized snap pour points and flow accumulation, snap pour raster was created. Then, from flow direction and snap pour raster, the watershed was delineated and converted to polygon (Fig. 3). Finally, a buffer analysis was applied in ArcGIS 10.5 to extract the distance of the stream layer. Distance from stream channel is the basic element in flood hazard mapping because the peak flow of the river largely damages the surrounding area of the river.

Fig. 2
figure 2

Stream network extraction model

Full size image
Fig. 3
figure 3

Watershed delineation model

Full size image

Drainage density is the ratio of the total length of all the rivers and streams in a watershed in km and the total surface area of a watershed in km2 (Darama and Seyrek 2016; Horton 1945). Watershed with high drainage density indicates that more precipitation quickly joins streams whereas low density means more precipitation traveled as surface runoff, through flow and baseflow. The drainage density determines water outlets and rock structures. The drainage density analysis provides a numerical measurement of landscape dissection and runoff potential. It can be calculated using Eq. (1):

$$DD= \frac{\sum {L}_{s}}{{A}_{w}}$$
(1)

where \(DD\) is the drainage density, \({L}_{s}\) is the length of streams, and \({A}_{w}\) is the area of watershed.

Surface interpolation

The study employed inverse weight distance (IDW) surface interpolation to develop rainfall distribution thematic layer of the Alawuha watershed from rainfall frequency data. Rainfall distribution was considered as a layer for flood vulnerability analysis because it shows the possible surface runoff water in the area. It is used by some previous studies (Desalegn and Mulu 2021; Sarkar and Mondal 2020; Rimba et al. 2017; Elkhrachy 2015; Kazakis et al. 2015; Ouma and Tateishi 2014).

Population density analysis

Population density layer was also used for flood vulnerability parameter (Sarkar and Mondal 2020) because flood hazard and its vulnerability are very severing in the area of settlement than other open environment. Population density factor was extracted using kernel density analysis in ArcGIS environment since the kernel density is a common nonparametric estimator of variety of variable (Kim and Scott 2012).

Soil types

Haplic Xerosols, Chromic Cambisols, Euthric Cambisols, Vertic Cambisols, Chromic Vertisols, Euthric Regosols, Calcaric Flubisols, Leptosols, and Phaeozems are main the soil types found in the watershed. These soil types were extracted with the watershed boundary and then classified into five classes to evaluate soil susceptibility to flooding. Soil type is also taken as a factor for flood analysis (Desalegn and Mulu 2021; Sarkar and Mondal 2020; Kusmiyarti et al. 2018; Rimba et al. 2017; Lincoln et al. 2016) because it influences the magnitude and direction of flooding.

All layers were standardized to create uniform thematic layers for overlay analysis. Standardized layers were developed based on the literature, expert’s knowledge, the nature of the topography, and the data of the study. Equal weight influence was used to create a final flood hazard map of the watershed since no specific measures are taken to minimize their effect on flooding. In evaluating parameters, sub-criteria were ranked from 1 to 5, where 1, 2, 3, 4, and 5 represent very high, high, moderate, low, and very low respectively (Table 2).

Table 2 Flood hazard parameters and weights
Full size table

Flood analysis

Hazard

A flood hazard is a characteristic occasion that may cause a death toll, injury, or other wellbeing impacts, property harm, loss of people and administrations, social and financial disruptions, or natural harm (ISDR, U. N. 2009). Flood hazard is the function of factors that cause flooding in the watershed such as slope, elevation, soil type, and distance to rivers, drainage density, and rainfall distribution. These factors were standardized to five classes as shown in Table 2. Multi-criteria evaluation (MCE) methods were employed to integrate these datasets to produce a flood hazard map. According to Saini and Kaushik (2012), GIS allows multi-criteria evaluation that the decision-makers use to identify a predefined set of criteria with the overlay process. The flood hazard map can be calculated using the following formula adopted from Kazakis et al. (2015):

$$\mathrm{FH}={\sum }_{i=1}^{n}ri.wi=S.ws+E.wE+L.wl+R.wr+So.wso+D.wd$$
(2)

where \(\mathrm{FH}=Flood\;Hazard;\) \(ri\) = the rank of the parameter in each point; \(wi\) = the weight of each parameter; \(n\) = the number of the criteria; \(S.ws\) = soil and weight of soil; \(E.wE\) = elevation and weight of elevation; \(L.wl\) = land use and weight of land use; \(R.wr\) = rainfall and its weight; \(So.wso\) = soil and its weight; and \(D.wd\) = drainage density and weight.

Multi-factor analyses were performed by using the multi-criteria ranking method to produce the final flood hazard map. Each parameter was ranked and weighted according to the estimated significance for causing flooding and created the final hazard factors map. The weighted overlay result flood hazard map was reclassified into very high, high, moderate, low, and very low.

Vulnerability

The attributes and conditions of a local area, framework, or resource make it defenseless to the harmful impacts of a flood (ISDR, U. N., 2009). Population distribution and LULC thematic layers were used to develop flood vulnerability map of the watershed. A GIS overlay analysis technique was employed in the ArcGIS 10.5 environment to combine the two exposed thematic layers of the study area. The result of the flood vulnerability map was reclassified into very high, high, moderate, low, and very low.

Risk

Flood risk is the blend of the likelihood of an occasion and its adverse results (ISDR, U. N. 2009). A GIS overlay analysis was employed in the ArcGIS 10.5 Raster calculator tool to create a flood risk map of the watershed. A flood risk map is a composite map that developed from flood hazard and vulnerability maps. It can be computed using Eq. 3:

$$\mathrm{Flood}\;\mathrm{risk}=\mathrm{Hazard}\ast\mathrm{Vulnerability}$$
(3)

Finally, flood risk map was classified into very high, high, moderate, low, and very low risk zones.

Rainfall frequency analysis

Rainfall frequency analysis becomes the most appropriate technique for flood managers to describe floods and identify the extreme events. We used the plotting position rank method to build the relationship between the probability of occurrence or return period of certain events and their magnitude. The ranking method with the Weibull formula plotting position is done by ranking from largest to smallest (Zachariev 2016; Weibull 1939). Hence, the largest observation is assigned plotting position \(1/n\) and the smallest \(n/n=1\) for its annual exceedance probability, AEP. The return period, T, of an event, is then the inverse of the AEP. The Weibull formula is stated in Eqs. (4) and (5):

$$AEP=\frac{i}{n+1}$$
(4)
$$T=\frac{n+1}{i}$$
(5)

where \(AEP\) = annual exceedance probability; \(T\) = the return period of an event or frequency; \(i\) = rank of an events; and \(n\) = the number of considered events.

Results

Land use/land cover change for the period 1985–2016

The land use/land cover of the watershed was classified into five classes, namely, agricultural land, forest land, bare land, built-up area, and riverbed for the years 1985, 2000, and 2016 (Fig. 4a–c), and the area for each LULC classes presented (Table 3). In 1985, the majority of the watershed (44.5%) was categorized as forest land, whereas agricultural land, riverbed, built-up land, and bare land covered 26.29, 16.38, 9.74, and 3.07%, respectively. For the year 2000, agricultural land accounted for 29.61% of the watershed, whereas bare land, forestland, riverbed, and built-up areas covered 18.54, 17.97, 17.92, and 15.96%, respectively. In 2016 agricultural lands covered 34.14% of the watershed, whereas built-up areas, riverbed, forest land, and bare land covered 21.72, 21.59, 19.19, and 3.36%, respectively.

Fig. 4
figure 4

LULC maps of the watershed from 1985 (a), 2000 (b), and 2016 (c)

Full size image
Table 3 Trend of LULC changes from 1985 to 2016
Full size table

Net LULC classes’ area changes due to natural and anthropogenic factors for the periods 1985 to 2000, 2000 to 2016, and 1985 to 2016 which were presented in Table 3. Between the years 1985 to 2000, forest area showed decreasing trend by 26.55%, whereas the bare land class and built-up areas increased by 15.47 and 6.22%, respectively. Agricultural land and riverbed also show an increase of 3.32 and 1.54%, respectively. In a similar analysis for the period 1985 to 2016, a significant change in either direction for all land use categories was observed. From 1985 to 2016, forest area decreased by 25.34% primarily due to the expansion of settlement and agricultural land, which increased by 11.98 and 7.84%, respectively. Riverbed and bare lands also showed a significant increase of 5.21 and 0.3%, respectively. Generally, the result has indicated a series of LULC change for the last 30 years (1985–2016). In this regard, the trend has shown more land being used for agriculture and built-up areas. Except for forest land, all other land use categories show expansion for the past 30 years in the watershed and which can cause ecosystem imbalance.

The overall accuracy and kappa statistics of the LULC classification for 1985, 2000, and 2016 were 83.25% and 0.825, 85.13% and 0.843, and 91% and 0.901, respectively (Table 4).

Table 4 Accuracy report of LULC for 1985, 2000, and 2016
Full size table

Flood hazard factor analysis

Flood hazard aggravating factors considered in this study include elevation, slope, drainage density, soil type, proximity to the river, and rainfall distribution for the hazard analysis. Understanding the factors which aggravate flooding is significant to watershed conservation and management practices. These flood aggravating factors of the catchment are analyzed as follows.

Slope and elevation

The lower value of the slope is a flatter terrain, and the higher value of the hill is steeper terrain. Based on their susceptibility to flooding, slope and elevation have been classified into five classes. The result indicates that area less than 1000 m in height was very highly susceptible to flood. Places found in elevation between 1001–1500, 1501–2000, 2001–2500, and > 2501 m were high, moderate, low, and very low sensitivity levels to flood, respectively. Therefore, the result of the study revealed that flood hazard more affects the lowland areas than the highlands. The study also shows that places found in the lowland and smooth surface of downstream of the watershed were very high sensitivity levels of flooding. Such places in the study area have a slope value of less than 1% slope. On the other hand, places found in slope values between 1–3, 3–4, 4–5, and > 5% were highly, moderately, low, and very low affected by flood hazard, respectively (Fig. 5).

Fig. 5
figure 5

Slope and elevation flood susceptibility map

Full size image

Soil type

Nine main soil classes were identified in the watershed based on the soil erosion grouping system (Nachtergaele et al., 2010). These are Chromic Vertisols, Chromic Cambisols, Euthric Cambisols, Euthric Regosols, Calcaric Flubisols, Haplic Xerosols, Leptosols, Phaeozems, and Vertic Cambisols. According to their flood generating capacity, these nine soil types were grouped into five classes. Based on the degree to cause a flood, soils were categorized as very high, high, moderate, low, and very low. Finally, Haplic Xerosols, Chromic Cambisols, Euthric Cambisols, and Vertic Cambisols are categorized as a very high flooding capacity, Chromic Vertisols and Euthric Regosols are categorized as high, Calcaric Flubisols are classified as moderate, Leptosols are classed as a low, and the last Phaeozems are categorized as very low flood generating capacity (Fig. 6).

Fig. 6
figure 6

Soil flood susceptibility map

Full size image

Rainfall and drainage density

The average rainfall distribution map of the study area was classified into five classes. The study shows that areas that receive with higher rainfall amount annually (> 1200 mm/year) were categorized under “very highly” affected by flood hazard and places that obtain rainfall vary from 900 to 1200 mm/year highly affected by flood hazards. The study also stated that places receive rainfall between 700–900, 500–700, and < 500 mm/year were found in moderate, low, and very low flood hazard zone, respectively, as shown in Fig. 7.

Fig. 7
figure 7

Drainage density and rainfall hazard map

Full size image

The drainage density of the catchment is used for calculating the volume of the flood in a given stream. However, all the valleys do not necessarily carry water permanently; rather, they are filled by seasonal flooding. The drainage density thematic layer was classified into five classes. The result of this study indicates that areas with a higher drainage density (> 0.68 km/km2) were found in a very high flood hazard zone, while drainage density varies from 0.54 to 0.68 km/km2 lay in the high flood hazard zone. Moreover, the drainage density varies from 0.35 to 0.54, 0.19 to 0.35, and less than 0.19 km/km2, and places found in these densities were moderate, low, and very low affected by flood hazard, respectively.

Distance to streams

Analysis of distance to streams shows that areas within 100 m distance from streams were very highly exposed to flood hazards. Places found between 100 and 150 m distances from the streams were highly exposed, while the stream distance between 150 and 200 m caused moderate flood hazard. Besides, places lay between the distance of the streams 200–250 and > 250 m were found in low and very low flood hazards, respectively (Fig. 8).

Fig. 8
figure 8

Distance to stream map

Full size image

The GIS weighted overlay analysis result of flood hazard factors shows that 12,773.13 ha of the watershed was a very high (dark blue color) flood hazard zone, while 16,644.69 ha was a highly (light blue) hazard zone. Besides, moderate (light purple), low (olive), and very low (green) flood hazard zones have covered an area of 32,211.49, 24,468.58, and 15,731.78 ha, respectively (Fig. 9).

Fig. 9
figure 9

Flood hazard map

Full size image

Flood vulnerability factor analysis

Population distribution and LULC

According to the North Wollo Zone agricultural office, the total household living in the catchment was 197,410. From those households, based on their location, economic activity, property damage (lose), livestock loss, and injured, about 9440 households were affected by the flooding. The result of the study indicates that 8652 in Woldia town, 181 in Guba Lafto Woreda, 162 in Habiru Woreda, 143 in Gidan woreda, 132 in Raya kobo woreda, 120 in Dawint woreda, 35 in Ewa woreda, and 15 households in Gulina woreda will be affected. The result of the study also shows that agricultural land and built-up areas were very high and high vulnerable to flood whereas riverbed, bare land, and forest cover were moderate, low. and very low vulnerable to flood, respectively (Fig. 10).

Fig. 10
figure 10

Households distribution and LULC map

Full size image

The result of regression statistics confirmed that the agricultural land and flood vulnerability have a strong positive correlation with a confidence level of 94% and p value of 0.06 (Table 5). The correlation of the dependent (flood vulnerability) and independent (agriculture land) variable was 0.96. Built-up and flood vulnerability also have a significant positive correlation with R2 of 0.93 with the confidence level of 91% and p value of 0.09 (Table 6).

Table 5 Correlation statistics of flood vulnerability and agricultural land
Full size table
Table 6 Correlation statistics of flood vulnerability and built-up area
Full size table

The result of the final flood vulnerability analysis indicates that 41,824.12, 17,809.22, 33,042.88, 8815.14, and 497.5 ha of Alawuha watershed were very high, high, moderate, low, and very low vulnerable to flood, respectively. Figure 11 indicates that flame red and orange colors represent very high and high flood-vulnerable area, respectively. Moreover, light green, green, and dark blue colors show moderate, low, and very low flood-vulnerable zones, respectively.

Fig. 11
figure 11

Flood vulnerability map

Full size image

Flood risk analysis

Flood risk is a composite map of flood vulnerability and hazard which indicates the possible impacts that happened due to flood hazards. The final composite flood risk map shown in Fig. 12 indicates that 42,035.66, 22,575.09, 35,658.63, 691.38, and 1027.74 ha of Alawuha watershed was very high, high, moderate, low, and very low to flood risk, respectively. As shown in Fig. 12, red, yellow, olivenite green, light blue, and dark green colors represent very high, high, moderate, low, and very low flood risk zones, respectively. The study revealed that Sanka, Afrikari, and Gedo-ber, Woldia town, Alawuha agricultural land, and the low-lying area of Gulina and Ewa are highly flooded risk areas.

Fig. 12
figure 12

Flood risk map

Full size image

Rainfall frequency

Flood frequency analysis is one of the important studies of flood risk analysis. It is essential to interpret the record of flood events and the amount of precipitation to predict future possibilities of flood occurrences. It helps to determine the return levels of extreme rainfall and to understand how often such rare events might occur in the future. Frequency analysis result of the watershed indicates that the highest rainfall amount (356.4 mm) was observed in 2016 with the return period of 18 years and annual probability exceedance of 0.05 or 5% (Table 7). The amount of rainfall recorded in 2000 was 260.7 mm, and it has an exceedance probability of 7.14 and the return period of 0.14 years.

Table 7 Rainfall frequency analysis from 2000 to 2016
Full size table

Discussion

LULC change and flood

The result of this investigation revealed that built-up, agriculture, and riverbed areas had been increased drastically from 2000 to 2016. Such rapid change of land use/land cover increases flood occurrence in the study area. Particularly, uncontrolled urbanization and extensive agricultural land expansion bring a dramatic change of landscape patterns and types (Li et al. 2018; Jiang and Tian 2010; Fan et al. 2009); this has a powerful effect on floods. Large areas of the watershed are deforested or drained, increasing or decreasing antecedent soil moisture and triggering erosion. Furthermore, an extreme expansion of urban areas and impermeable surfaces such as asphalt and concrete contribute to the change of river morphology and slope instability. Drains and sewers take water directly to the river, which also increases flood risk. Therefore, managing the land use/land cover is a very significant component of mitigating the impact of flood disasters (Barasa and Perera 2018; Getahun and Gebre 2015). Moreover, the current study stated that urban area (built-up) was the most vulnerable land use class to flood because it is the area where most of the people lived, economic interaction processed, and basis of financial mobility. Thus, the finding of the study agrees with previous studies (Lin et al. 2020; Naif Rashed Alrehaili 2021; van der Sande et al. 2003). Agricultural land is also the second vulnerable land use class in the watershed because it is the primary economic activity of the country and the study area and the main source of livelihood of the societies. The study confirmed a strong positive correlation between agriculture and flood vulnerability (R2 = 0.96) and built-up and flood vulnerability (R2 = 0.93).

Land use/land cover changes impact the condition for the change of precipitation into the spillover. Built-up impermeable regions cause the arrangement of a quick overland flow at the expense of the natural retention and subsurface or groundwater flow. There are a few bits of proof that adjustments of land use have impacted the hydrological system of different river basins. These effects can be huge in small basins (Oda et al. 2021; Jones and Grant 1996). The hydrological effects of land use change depend not just on the general changes in land use types yet additionally on their spatial conveyance (Schumann et al. 2000). Besides, the change of agricultural, forest, grassland, and wetlands to urban regions generally accompanies a tremendous expansion in impenetrable surfaces, which can adjust the regular hydrologic condition inside a watershed. It is surely known that the result of this change is ordinarily reflected in an expansion of the volume and pace of surface overflow and diminishes in groundwater re-energize and base flow (Dey and Mishra 2017; Moscrip and Montgomery 1997). This ultimately prompts bigger and more incessant occurrences of local flooding.

Flood and hydrology of a watershed show a close connection with the common land use. As the watershed turns out to be more evolved, it additionally turns out to be hydrologically more dynamic, changing the flood volume, overflow parts just as the beginning of streamflow. Nonetheless, floods that once happened rarely during predevelopment periods have now gotten more continuous and more extreme because of the change of watersheds starting with one land utilize then onto the next. Anthropogenic land use changes cause different hydrologic and geomorphic changes, including adjustments for the size and timing of flood peaks (Ricker et al. 2012; Miller et al. 1993) and in the size and kind of soil erosion (Zhu and Zhu 2011; Costa 1975). Then again, afforestation and the advancement of forest management will impressively expand the water maintenance in the landscapes (Marchi et al. 2018; Loyche Wilkie et al. 2003). Change in land use applies a critical impact on the relations of precipitation overflow and additionally spillover dregs (Li et al. 2019; Fu et al. 2002) and adjust soil and water trouble appropriately (Zheng and Wang 2013; Nemani et al. 2002). Recorded datasets of land use can be utilized to know the area and sorts of land use change as a stage towards examining the effect of such change on floods. Consequently, to moderate the flooding and catastrophe supportable flood, the board approach must be carried out because it has financial advantages like flood relief and ecological advantages like habitat reinstatement. This new worldview requires a coordinated and comprehensive way to deal with flood management, which must be fused in both straightforwardly influenced regions and the whole basin or watershed. The consideration of the uplands and lowlands in the basin normally consolidates a wide scope of land utilizes, including forest, pasture, arable land, and urban regions. Thus, successful implementation of the sustainable flood management approach entails interdisciplinary coordination and participation at all levels of government, across different sectoral policies, and with all concerned stakeholders.

Flood risk indicators

The nature of the terrain and the direction of the slope affect the spatial variation of flooding and its risk. The result stated that steep slope terrain is more risk for flooding particularly when the land cover is altered and degraded. Rogger et al. (2017) found that when hill slopes are modified for agricultural production, there is a changing inflow path, flow velocities, water storage, and flow connectivity and concentration times. Moreover, the study confirmed an extreme agricultural land expansion worsens the risk of flooding mainly in highland area. On the other hand, the risk of flooding in lowland areas is commonly related to river peak flow, surface runoff, baseflow, and regression water from the larger water body (water accumulated areas). For instance, lowland areas of Dembia, Fogera, Dera, and Libo Kemkem districts were damaged due to peak flow water from the river and overflow water from Lake Tana. The study revealed that places found in downstream and lowland areas, as well as topographically steep slope places, have a high risk of flooding.

The type of slope coupled with the types of soil, rainfall amount, and elevation aggravating flood risk. They have impacts on flood generation, frequency, and magnitude of floods. The result of this study indicates that Haplic Xerosols, Chromic Cambisols, Euthric Cambisols, and Vertic Cambisols have very high flooding capacity due to their low moisture availability and high porosity. Rogger et al. (2017) indicate that soil nature affects flow paths, including the type of runoff generation mechanism (overland flow versus subsurface storm flow). This study argues that slopes tend to reduce the amount of water infiltration into the ground; this water can then flow quickly down to rivers as overland flow mainly, steep slopes cause more through flow within the soil. Moreover, elevation is used to determine flood insurance costs for high-risk zones (Samanta et al. 2018). The high potential of drainage density increases the possibility of flood risk since surface runoff and stream water overflow can be raised; particularly, the problem is severe during rainy seasons. Besides, drainage density increases in thickness from low values which will produce significant increases in the flood peaks (Pallard et al. 2009). Therefore, the study revealed that heavy and prolonged rainfall amount raises flood risk because it causes an increase of river water overflow from its channel and brings surface runoff. So, areas that are very close to the river are more at risk than others far from the river. Therefore, household’s property and livelihood are more damaged when they are very close to the river and streams.

Causes of flood and management practices

Owing to its topographic and altitudinal characteristics, flooding is not new to Ethiopia (Shimeles and Woldemichael 2013). They have been occurring at different places and times with varying magnitude. A study also confirmed that flood disasters in South Omo and Dire Dawa city administration were an example of indicating flood occurrence severity in Ethiopia. Even though the cause of flooding will vary depending on the nature of the locality and the interventions that have been taken, intense rainfall is one of the main trigger factors of flooding mainly in the rainy season (Getahun and Gebre 2015).

The study area is characterized by rigid topography which makes it more to flooding especially for the communities around the riverside. The causes of flooding in the study area are improper tillage, slope terrain, deforestation, high seasonal rainfall, and the absence of protection measures. In addition, the increment of flooding is resulted from lack of agroforestry practices, farm terraces, homestead development, residue management, and free grazing. Weak awareness of local people towards soil and water conservation measures is also a contributing factor. The flooding events that happened from 2000 to 2016 have an effect on death, and displacement of peoples, and animals, crop loss, house destruction, topsoil removal, water pollution, and reduction of streamflow farmland fragmented and runoff increased, and more gullies and rills are created. Moreover, the 2016 flood event affected 9,440 households and caused the displacement of households from their location, injured, damage their economy, loss of property, livestock, and agriculture even though the severity of the risk varies spatially and temporally. Unless strong and sustainable flood management practice measures are taken, the study confirms that the area will be affected by flood disasters like the 2016 flood with a return period of 18 years with a 0.05 probability of occurrence.

To minimize the problems of flooding in the study area, various soil and water conservation practices were performed in the watershed including terraces, ditch trenches, contour plaguing, stone bunds, check dams, soil bunds, and others. However, implementations of these soil and water conservation practices do not follow any scientific approaches. According to field observation made in the watershed, the farmland in dry areas has fewer agroforestry practices with predominately covered by cereal crops, but Dega parts of the watershed have relatively more agroforestry practices. Eucalyptus is the most common tree found in the upper watershed. Few farmers in the upper parts of the watershed use local stone bunds to avoid excess runoff. Some farmers use traditional soil bund in their farmland. Most soil and water conservation activities are introduced by the Woreda agriculture office through its regular agricultural extension services and productive safety net programs and other supporting organizations. High cost of construction and labor shortage are challenges of efficient implementation of flood control mechanisms practiced due to the high cost of construction and high labor demanding.

Conclusions

Flood disaster management can be effective only when thorough information is acquired about the expected frequency, character, and magnitude of hazardous events in an area and the vulnerabilities of the people, buildings, infrastructure, and economic activities in potentially dangerous areas. In this investigation, various parameters were identified as flood aggravating factors with different hazard values such as slope, elevation, soil type, drainage density, distance from streams, and rainfall amount. These flood aggravating factors result in flood vulnerability on people, land uses, land cover, and other elements like infrastructures and economic activities. The main causes of flooding in the catchment area are improper tillage, slope terrain, deforestation, high seasonal rainfall, and absence of protection measures. These problems result in a decrease or dry stream flow, farmland fragmentation, increase runoff, creation of gullies and rills, death of cattle, crop and infrastructure destruction, a decline in soil fertility, and pollution of water bodies as well as the death of people. The GIS multi-criteria evaluation of flood risk assessment shows the area under the highly vulnerable in the watershed. Therefore, for those areas, different watershed conservation measures were taken by the stakeholder to minimize the flood’s physical and biological risks in the Alawuha watershed even though their effectiveness is under question.

To effectively adopt various biological and physical measures for flood management in the watershed, the local government authorities are responsible for creating awareness about the high probability of flood occurrences and their risk in the area. The local disaster risk management and food security offices must install flood hazard indicator technology in high flood risk areas to alarm the people who live in the room when flooding occurs. Besides, based on the nature and extent of degradation as well as the purpose of conservation, different soil and water conservation (SWC) measures will be identified. This may include physical or structural measures, biological (vegetative) measures, agronomic (best management practices) measures, or area closure. Physical measures are structures built for soil and water conservation. They are built with the principle to (1) increase the time of concentration of runoff, thereby allowing more of it to infiltrate into the soil; (2) divide a long slope into several short ones and thereby reduce the amount and velocity of surface runoff; (3) reduce the velocity of the surface runoff; and (4) protect against damage due to excessive runoff. Moreover, biological measures of soil and water conservation provide a protective impact through vegetation cover. Vegetation cover has the functions of (i) preventing splash erosion, (ii) reducing the velocity of surface runoff, (iii) facilitating the accumulation of soil particles, (iv) increasing surface roughness which reduces runoff and increases infiltration, and (v) stabilizing soil aggregates through their roots and organic matter which in turn increase infiltration. Thus, watershed conservation practices must be conducted based on the hydrological unit rather than the administrative unit.