Land-use and land-cover change and its impact on flood hazard occurrence in Wabi Shebele River Basin of Ethiopia

The hydrological cycle is the phenomenon of the water recycling system on the earth, which is in the oceans, atmosphere, land surface, biosphere, soil, and groundwater systems. Water contained in the sea and the land surface can evaporate into the air and move up into the atmosphere directly and through vegetation as a process of evaporation and transpiration, which occurs through condensation, clouds, and rain as a process of precipitation (Yulianto et al. 2020). Rainfall that falls to the ground surface becomes runoff, and some water will be infiltrated into the ground and become groundwater. The surface of the earth on which the hydrological cycle process occurs continuously is a ‘watershed’. It is an area bounded by surface topography and drainage or river patterns. Rainfall that accumulates in the watershed will flow through drainage or river to an outlet on the earth's surface (Marshall 2014; Yulianto et al. 2020). The characteristics and conditions of watersheds can be explained by conditions of landforms on the surface of the earth that include the nature of landforms, genesis, processes, and material composition.

Land-use/land-cover (LULC) changes can impact the hydrological conditions, such as land surface coefficient, discharge, and hydrographic characteristics and affect the related runoff and infiltration characteristics and hydrological patterns in a watershed. Land use means the use to which the land is being put or the utilization of land devoted to human activities and land cover is the physical surface of the earth (Tali & Kanth 2011). Land cover is continually molded and transformed by land-use changes; for example, when a forest is converted to pasture or cropland. Land-use change is the proximate cause of land-cover change.

The rapid increase in the human population and the activities they undertake have consequences for LULC in terms of fulfilling their needs, both socially and economically. The water resources development in the current and future conditions is sensitive to LULC changes and the intensification of human activities (Liu et al. 2017; Kundzewicz et al. 2018; Sutfin & Wohl 2019). Different human activities exert an influence on the hydrological cycle and water resources. LULC in watersheds exhibit differences in hydrologic runoff response, which can be linked to flood events. Increased runoff because of LULC changes can affect the frequency of flooding, base flow, and annual average flow in such a way as to alter the hydrological cycle (Sutfin & Wohl 2019; Yulianto et al. 2020). The types of LULC by themselves exhibit differences in hydrologic runoff response, which can be linked to flood events. Loss of land floor cover, thinner forest canopies, grasslands, and reduced infiltration of rainfall resulted in a rapid hydrologic response and increased flood magnitudes and frequency (Sutfin & Wohl 2019).

In the last 40 years, a significant deforestation process occurred in East African countries. For instance, Hailemariam et al. (2016) analyzed that a large area of forest land decreased due to farmland and urban settlement expansion in the Bale Mountain Eco-Region of Ethiopia from 1985 to 2015. Similarly, over-exploitation of resources due to an increase in population and demand for food supply has caused land degradation in western Kenya (Githui et al. 2009). However, deforestation and land development for agriculture has not necessarily led to an equal increase in food production but have often led to land erosion in the upstream area and triggered heavy floods in the downstream areas as presented in Vandaele & Poesen (1995) and Adnan (2010). In the Wabi Shebele basin, the extent of shrublands indicates significant increasing trends, while grassland and cultivated area showed decreasing trends from 1984 to 2004 (MoWR 2004).

Flooding is one of the most frequently occurring types of natural disaster in the Wabi Shebele basin of Ethiopia (Admassu et al. 2010). In the river basin, the magnitudes and frequency of flood events in middle watersheds indicate an increasing trend in recent decades (Wudineh et al. 2021). In mountainous zones, increased potential for intense convective storms, increased cultivated land, and more highly confined river valleys result in more rapid runoff response to precipitation and cause flood risks. Flood events can cause problems such as the inundation of settlements, damage to infrastructure, disruption to community activities, health problems, and loss of life, and can create economic losses.

Understanding LULC changes in the Wabi Shebele basin is important in the decisions of land planning in the region since it can reflect the dimensions, potential impacts, and interactions of the relationship between human activity and the environment. Thus, this study aimed to analyze land use/land cover change and relate those changes to flood occurrences in terms of magnitude and frequency of flood in the Wabi Shebele basin.

The Wabi Shebele River Basin (WSRB) is a transboundary basin between Ethiopia and the Republic of Somalia in the horn of Africa. It originates from the Bale Mountain ranges of the Galama and Ahmar of Ethiopia, about 4,000 m above sea level, and drains a portion of Somalia before draining to the Indian Ocean. About 72% of the catchment (202,220 km²) is lying in Ethiopia. In this study, the Wabi Shebele basin is used to represent the catchment that is lying in Ethiopia within 4° N 45′ to 9° 45′ N latitude and 38° 45′ E to 45° 45′ E longitude (Figure 1). The watershed is divided into three geographical areas: the upper valley, which is characterized by a mountainous area with abrupt valleys; the middle valley, which is wider and rainy; and the lower valley, which is an arid lowland area. The areal distribution of rainfall varies from 271 mm at the lower arid portion (Gode) to 1,320 mm in the upstream highlands of the basin (Seru), out of which a major portion is discharged to streams as runoff. The spatial variability of temperature is significant with maximum and minimum values of 27.1 and 12.6 °C, respectively. While having the largest area coverage, the basin water yield is only 0.43 l/s/km², which is exceptionally low relative to other basins in Ethiopia (MoWR 2003; Awass 2009).

Soil and water assessment tool (SWAT) is a physically based distributed hydrological model (Arnold et al. 1998; Neitsch et al. 2005; Abbaspour et al. 2007) that was developed by the United States Department of Agricultural Research Service to simulate the impact of land management practices on hydrology and water quality under complex watersheds with heterogeneous soil and land use conditions. In recent decades, it has been widely used for water cycle simulation and water resources management, especially for the analysis of streamflow variation under climate change and LULCC (Adamu 2014; Guo et al. 2016; Näschen et al. 2019). In addition, SWAT can also be used to predict the impact of future climate on the evolution of water resources and streamflow under different preset scenarios of climate (Schulze 2000; Camici et al. 2014; Gaur et al. 2020). The future climate conditions can be obtained from the general circulation models (GCMs) developed by the Intergovernmental Panel on Climate Change (IPCC) (Schulze 2000; Gaur et al. 2020), thus the effects of different future climate conditions can be further discussed for different drainage basins.

The SWAT model proceeds in two steps to simulate the hydrological cycle (Guo et al. 2016): runoff generation and its confluence in the river channels. To generate runoff, a watershed is firstly divided into several sub-basins, each of which is composed of one to several hydrological response units (HRUs) that consist of homogeneous land use, topographical, and soil characteristics. Threshold values for land use, soil types, and slope are set up to remove the insignificant land use, soil type, and slope in each sub-basin, thereby avoiding the generation of many HRUs. Next, the river network connects the discharge produced in sub-basins based on the water balance equation and water flows through the river channels and toward the basin outlet (Neitsch et al. 2005; Guo et al. 2016). The surface streamflow was calculated by a modified Soil Conservation Service (SCS) curve number method (USA_SCS 1972). The potential evapotranspiration and channel routing were estimated and simulated by the Penman–Monteith method and a variable storage method, respectively.

The optimum parameters of the SWAT model can be determined by sensitivity analysis, which assesses the sensitivity between a parameter and other parameters in different areas. Based on parameters available for water production identified by Arnold et al. (2012) and preliminary identification in the SWAT model, SWAT-CUP global sensitivity analysis is conducted to identify the most sensitive parameters for watersheds. The p-value and t-statistic are used to eliminate nonsensitive parameters from the calibration process. The higher the absolute value of the t-stat and the smaller the value of the p-value, the more sensitive is the parameter (Abbaspour et al. 2007; Moreira et al. 2018).

In this study the separation method is applied to extreme high flows since it focuses on the impact of LULC change on flood occurrence. Six extreme high flow indices (flood indices) are extracted from simulated streamflow under different climate change and LULC change conditions and the impact level of climate change and LULC change are analyzed in each index. These flood indices are: annual maximum discharge (AMAX), peak over the threshold (third quartile) frequency (POTF), peak over the threshold (third quartile) magnitude, seasonal maximum discharge for winter (SMW), seasonal maximum discharge for spring (SMSp), and seasonal maximum discharge for summer (SMSu) are used to define extreme high discharges.

In extreme value analysis ensuring the independence of samples is an initial task. In this study, the time interval approach is used to ensure the independence of flow discharges. Time intervals between 5 and 14 days between successive peaks; 5 days for catchments <10,000 km² and 14 days for catchments ≥10,000 km², are used. This approach is reported as a strong flood-frequency estimations approach (e.g., Malamud & Turcotte 2006; Keast & Ellison 2013).

In data-sparse watersheds, like Wabi Shebele River Basin, developing a representative hydrological model (e.g., in generating the observed streamflow) is challenging but it is a prerequisite to accurately assess variabilities in extreme flows. In this study, we used a combination of datasets to calibrate and validate the hydrological model. In addition to the field-based ground stations, some weather data like relative humidity, solar ratio, and wind speed data from Climate Forecast System Reanalysis (CFSR) climate data to improve the hydrological model accuracy during both calibration and validation.

In detail, the Wabi Shebele River Basin was divided into 14 sub-basins and 311 HRUs. To define HRUs, threshold values of 10% were chosen for land use/soil type and slope, respectively. Meteorological data from the period of 1988–2000 and land use data from 1997 were used for the calibration and validation of the SWAT model with 3 years warming period. Seven hydrologic stations were selected for calibration and validation in the basin. To evaluate the model performance, three parameters have been used, namely R² and NSE and P-bias, and are estimated as presented in Table 3.

The sensitivity of the discharge to the model parameters was checked through global sensitivity analysis performed using SUFI-2 of SWAT-CUP out of 20 model parameters identified for streamflow prediction in literature (Neitsch et al. 2005; Narsimlu et al. 2015; Moreira et al. 2018). Based on the results obtained from the global sensitivity analysis, the first six parameters categorized as especially important and important were found to be sensitive (p-value < 0.05) in Wabi Shebele River Basin (Table 4). Griensven et al. (2006) classified parameters regarding their sensitivity based on their increasing hierarchical position of the parameters. They categorized parameters as especially important (1st), important (2nd–6th), slightly important (7th–14th), and not important (15th–20th). From Table 4, it is evident that there is a spatial distribution of parameters contributing to water production at three parts of the Wabi Shebele basin watersheds: at the northwestern highland, middle to the northeastern basin, and lower downstream of the basin. In the northwestern highland of the basin surface parameters (SLSUBBSN.hru, HRU_SLP.hru, CANMX.hru) and groundwater parameters (i.e., GWQMN.gw, RCHRG_DP.gw and GWREVAP.gw) are the most significant parameters in water production, in the middle and northeastern part of the basin soil parameters (i.e., ESCO.hru, SOL_K sol and SOL_Z.sol) and surface parameters (i.e., SLSUBBSN.hru and HRU_SLP.hru) are the most significant parameters and in the lower downstream part of the basin surface parameters (i.e., SLSUBBSN.hru, HRU_SLP.hru, CANMX.hru) and soil parameters (i.e., SOL_K sol and SOL_Z.sol) are the most significant water production parameters identified in this study.

The effects of LULC change on streamflow were distinguished by simulations of multi-year daily catchment stream flows using 1986, 1997, and 2016 land covers, respectively, as presented in Table 5 and Figure 4. We used two different conditions to see the effects of land use and cover changes on streamflow: condition one, stream flow change under LULC change at small period of 12 years between 1986 and 1997; and condition two, stream flow change under LULC change prolonged period of 31 years in between 1986 and 2016.

In a prolonged period of 31 years of LULC observation from 1986 to 2016 (Figure 3(b) and 3(d)), LULC was dominated by woodland, shrubland, cropland, grassland, bare land, and forest in the entire basin. Increases in LULC during this period occurred in shrubland, cropland, grassland, and barren land whereas forest and woodland are LULC which show decreases in this period. The basic change in this condition is observed in shrubland coverage. Shrubland indicates increases in a prolonged period from 1986 to 2016 while it showed decreases during a small period between 1986 and 1997. But other LULCs follow similar trends as condition one.

Out of the 10 LULC types analyzed in the study area, 3 of them, namely cropland, grassland, and, bare land, showed growth in both conditions. The coverage of cropland was increased by 47.56% in condition one and by 48.63% in condition two. Contrarily, the coverage of forest and woodland were decreased by 86.37 and 97.39%, respectively, between 1986 and 1997 (Figure 3(a)). Between 1986 and 2016, large areas of forest (49.14%) and woodland (14.76%) were converted to other LULC types (Figure 3(b)). Similarly, an IWMI (2015) study report indicates forest degradation in the western upper Wabi Shebele basin by an average annual deforestation rate of 0.25%. The recent data analysis in Bale Eco-Region (i.e., upper Wabi Shebele basin) indicates a reduction in forest area (forest, woodlands, Erica Forest) of about 2.3% between 2010 and 2014 (IWMI 2015). It is known that deforestation and forest degradation in Ethiopia are for the expansion of subsistence agriculture and grazing lands implies that the land use and land cover change over the middle and northern parts of the Wabi Shebele basin from the forest, woodland, grassland, and shrubland to agricultural land (cropland) is one of the causes of streamflow increases. The reduction of forest cover amplifies flood events, as more rainfall directly turns into runoff instead of being slowed down or buffered by forests (Bradshaw et al. 2007; IWMI 2015). In the northeastern downstream of the Wabi Shebele basin (Erer and Wabi at Gode watersheds) simulated streamflow did not show significant changes with land use change in both conditions as presented in Table 5 and Figure 4. Flood indices calculated from simulated daily streamflow under different land use and cover map indicate comparable situations with annual average flow variations in the basin. In the middle and northwestern basin flood indices like AMAX, SMW, MSp, SMSu, and volume of discharge indicate increasing as forest coverage indicates decreasing trends (Table 6).

The influence level of LULC change and climate change on the stream flow of the Wabi Shebele basin is estimated using the separation method. As presented in Table 7, the response of the streamflow to climate change is higher than that of LULC change in the Wabi Shebele basin. However, LULC change also has a significant impact on watersheds like Wabi at Lege Hida, Wabi at Dodola, Maribo and Robe. Annual maximum discharge (AMAX) decreases in watersheds where forest and shrubland coverage increases in the study period. For instance, in Wabi at Lege Hida and Erer watersheds, the magnitude of floods decreases while the coverage of forest increases in condition one. In north western watersheds like Wabi at Dodola, Maribo, Robe, and Wabi at Lege Hida flood discharge estimated using LULC of 2016 is greater than flood estimate using LULC of 1986 by 3.91, 2.33, 1.92, and 128.66 m³/s and as a result, flood magnitude increases by 0.18, 1.83, 0.57, and 0.44% in watersheds, respectively. In Wabi at Gode watershed flood magnitude under condition one is greater than flood magnitude in condition two by a value of 1,285.18 m³/s, which is contributed by both climate change and LULC, accounting for 105.12 and 5.12%, respectively. The results indicated that climate change is the major factor influencing the streamflow and flood values in Wabi Shebele River Basin in the period from 1980 to 2010, which is like the conclusion drawn by Akola et al. (2018).

Land-use and land-cover (LULC) changes can impact hydrological conditions such as land surface coefficient, discharge, and hydrographic characteristics and can affect the related runoff and infiltration characteristics and hydrological patterns in a watershed. The impact of land use and land cover on flood occurrence is investigated. To identify the most sensitive parameters to water production, the hydrological model is performed in this study. Furthermore, which share LULC and/or climatic changes cause the increase of flooding in the Wabi Shebele basin is answered partly by this study.

In Wabi Shebele basin parameters contributing to water production are spatially distributed over the basin at three different parts: at northwestern highland, surface and groundwater parameters (i.e., SLSUBBSN.hru, HRU_SLP.hru, CANMX.hru, GWQMN.gw, RCHRG_DP.gw, and GWREVAP.gw); at the north eastern and southern lowland parts of the basin, soil and surface parameters (i.e., ESCO.hru, SOL_K sol and SOL_Z.sol, SLSUBBSN.hru and HRU_SLP.hru) are the most sensitive parameters identified for water production. These parameters are directly and indirectly based on land use and land cover in the watersheds. For instance, CANMX.hru is maximum canopy storage (mm H₂0) which can significantly affect infiltration, surface runoff, and evapotranspiration. The influence the canopy exerts on these processes is a function of the density of plant cover and the morphology of the plant species (Neitsch et al. 2005).

Out of the 10 LULC types analyzed in the study area, three of them, namely cropland, grassland, and bare land, showed growth in both short and long period conditions. The coverage of cropland was increased by 47.56% in condition one and by 48.63% in condition two. Contrarily, the coverage of forest and woodland decreased by 86.37 and 97.39%, respectively, between 1986 and 1997. Between 1986 and 2016, large areas of forest (49.14%) and woodland (14.76%) were converted to other LULC types mainly to cropland and shrub land. Streamflow simulated under condition one indicates that increases in those watersheds show significant cropland increases and forest coverage decreases particularly in Wabi at Dodola, Maribo, Robe, and Erer watersheds. In the upper Wabi Shebele basin at Dodola watershed alone, annual streamflow increased by 7% when LULC changed from 1986 to 1997 land use map. Flood indices (i.e., AMAX, SMW, SMSp, and SMSu) calculated from simulated daily streamflow under different land use and land cover map indicates an increase in watersheds located in the middle and north western part of the basin where most of the forest coverage decreases.

The influence level of LULC change and climate change on streamflow analyzed using the separation method indicates that climate change is the major factor influencing the streamflow and flood values in Wabi Shebele River Basin. However, LULC change also has a significant impact in middle and upper watersheds like Wabi at Lege Hida, Wabi at Dodola, Maribo, and Robe. In north western watersheds like Wabi at Dodola, Maribo, Robe, and Wabi at Lege Hida flood discharge estimated using LULC of 2016 is greater than flood estimate using LULC of 1986 by 3.91, 2.33, 1.92, and 128.66 m³/s and as a result, flood magnitude increases by 0.18, 1.83, 0.57, and 0.44% in watersheds, respectively.

Separation was a useful approach to differentiate the impact level of LULC change from other potential drivers like climate change. Finally, I recommend a better determination of historical land-use changes in high-resolution hydrological rainfall–runoff models and validation of special hydrological parameters, which are capable of reproducing those land-use changes and their impact on flood conditions.

All relevant data are included in the paper or its Supplementary Information.

The authors declare there is no conflict.