Modeling Earth Systems and Environment https://doi.org/10.1007/s40808-018-0439-8

ORIGINAL ARTICLE

Modelling hydrological response under climate change scenarios using SWAT model: the case of Ilala watershed, Northern Ethiopia Henok Shiferaw1 · Amdom Gebremedhin1 · Tesfay Gebretsadkan3,4 · Amanuel Zenebe1,2 Received: 8 December 2017 / Accepted: 5 March 2018 © Springer International Publishing AG, part of Springer Nature 2018

Abstract This study evaluates surface runoff generation under climate change scenarios for Ilala watershed in Northern highlands of Ethiopia. The climate change scenarios were analyzed using delta based statistical downscaling approach of RCPs 4.5 and 8.5 in R software packages. Hydrological response to climate changes were evaluated using the Soil and Water Assessment Tool model. The Soil Water Analysis Calibration and Uncertainty Program of Sequential Uncertainty fitting version 2 algorithm was also used to compute the uncertainty analysis, calibration and validation process. The results show that the minimum and maximum temperature increases for the future of 1.7 and 4.7 °C respectively. However, the rainfall doesn’t show any significant increase or decrease trend in the study area. The 95% prediction uncertainty brackets the average values of observation by 71 and 74% during the calibration and validation processes, respectively. Similarly, R-factor equals to 0.5 and 0.6 during calibration and validation periods. The simulated and observed hydrographs of the total river yield showed a good agreement during calibration (NSE = 0.51, R ­ 2 = 0.54) and validation (NSE = 0.54, ­R2 = 0.63). From the total rainfall received only 6.2% portion of the rainfall was changed into surface runoff. The rainfall-runoff relationship was strongly correlated with ­R2 = 0.97. Moreover, there had been also high evapotranspiration (ET) loss in the watershed; almost 75% of the total rainfall was lost as ET and 7.8% as ground water recharge. Due to an increase trend in temperature and evaporation loss for the future, the surface runoff also declined from 1.74% in RCP4.5 near-term to 0.36% in RCP8.5 end-term periods. This implies, proper planning and implementation of appropriate water management strategies is needed for sustainable water resources management in the region. Keywords  Evapotranspiration · Runoff · SWAT​ · SWAT-CUP · RCP · Watershed

Introduction Ethiopia often referred as the water tower of East-Africa, characterized by mountainous topography (Awulachew et al. 2009). The rainfall-runoff processes on the mountainous slopes are the sources of the surface water for much of the * Henok Shiferaw [email protected] 1



Institute of Climate and Society, Mekelle University, POB 231, Mekelle, Ethiopia

2



Department of Land Resources Management and Environmental Protection, Mekelle University, Mekelle, Ethiopia

3

Tigray Agricultural Research Institute (TARI), Mekelle, Ethiopia

4

UNESCO-IHE Institute for Water Education, Delft, The Netherlands



country. Based on this, the government has implemented different land and water management practices in the highlands to potentially use the resource and minimize degradation problems (Ministry Water Resources 2002). Land degradation is a common occurrence in most highlands of the country (Hurni et al. 2005; Taddese 2001). As mentioned by Hurni et al. (2005) and Nyssen et al. (2007) the problem is mainly attributed to population growth, climate change and lack of effective land and water management practices in the country. Climate change is believed to have led to the changes in global patterns of precipitation, thereby changing the global water cycle and causing the redistribution of water resources in time and space (Milly et al. 2005). Presently, there are scientific evidences indicating that the average temperature of the earth’s surface is increasing due to greenhouse gas emission and other natural and anthropogenic factors. According to the 2014 report of the Intergovernmental Panel on Climate Change (IPCC), the

13

Vol.:(0123456789)



mean annual global surface temperature has increased by 0.3–0.6 °C since the late nineteenth century and it is anticipated to further increase by 1.0–2 °C over the next 100 years (IPCC 2014). Climate change affects the function and operation of existing water infrastructures including hydropower, structural, drainage and irrigation systems as well as water management practices (Eromo et al. 2016). Impact of climate change on water resources, particularly on surface runoff is a key and current research agenda at global level (IPCC 2014; Pandey et al. 2016; Yin et al. 2016). There are many studies at global level focused on impact of climate change on water resources, particularly on surface runoff (Yin et al. 2016). For example, Raneesh and Santosh (2011) assessed the effect of climate change on streamflow and vegetative growth in a humid tropical watershed of India. Similarly, Eromo et al. (2016) investigated the impact of climate change on surface hydrological process in Omo-Gibe river basin of Ethiopia and reported that surface water decreases in terms of mean monthly discharge in the dry season and increases in the wet season. The percentage change in future seasonal and annual hydrological variables was shown increasing trends. However, in the semi-arid northern Ethiopia, only few studies are available which mainly focused on hydrological modelling at different watershed scales. For example, Ashenafi (2014) reported that land use and climate change are affecting water resources in Geba catchment, northern Ethiopia. The report from this study shows that the stream flow decreased by 10% during the wet season and by 30% during the dry season. According to Abebe (2014) as a result of climate change, surface runoff in Suluh watershed of the Geba catchment increased by 4.6% and base flow and deep percolation were reduced by 10 and 7.4% respectively. Hydrological modeling tools are very useful to investigate the impact of climate change on the hydrology of a given watershed (Pandey et al. 2016). Today, the use of hydrological modeling for water resources planning and management is becoming increasingly popular in various research studies. Among many others, spatially and temporally distributed or semi-distributed hydrological models such as SWAT have important applications for discovering the relationships between the climate of the watershed and hydrological process (Mango et al. 2011). Such models are able to explicitly represent the spatial variability of land surface characteristics such as elevation, slope, vegetation, land use, soil and climate (Abebe 2014). Given its spatial and temporal capability of SWAT, different researchers are using SWAT to model watershed at different scales. For instance, Kumar and Singh (2016) used SWAT-VSA to model the surface runoff in the Himalayan landscapes. He concluded that paddy croplands followed by scrub, maize and forest cover as most contributing areas of surface runoff generation in the watershed.

13

Modeling Earth Systems and Environment

Similarly, Gyamfi et al. (2017) also applied SWAT model for groundwater recharge in a large scale basin in Olifants basin, South Africa. The findings indicated that groundwater recharge declined by 10.37 mm (30.3%) and 2.34 mm (9.82%) during the periods 2000–2007 and 2007–2013 respectively over the study area. Makwana and Tiwar (2017), also carried out stream flow modeling using SWAT and neural networks (NNs). They found that the SWAT model provides a better description of water balance of the watershed, whereas NN models present the surface runoff at the outlet without any explicit consideration of different components of the hydrologic cycle. Furthermore, Raneesh and Santosh (2011) also used SWAT for evaluating the effect of climate change on streamflow and vegetative growth in a humid tropical watershed of India. However, in Ethiopia, only few studies are available which lay emphasis on hydrological modelling at intermediate scale of a watershed. These studies demonstrate the capability of SWAT model to simulate runoff at different scales of watershed in different parts of the country. Understanding the impact of climate change on the hydrological processes at watershed level is crucial for water and land resource management in order to put appropriate adaptation and mitigation measures (Gebrekristos 2015; Ashenafi 2014). Hence, this study was aimed at investigation on the impacts of climate change on water resources at intermediate watershed scale which is helpful to put appropriate amelioration measures, proper plans, and policy measures to have sustainable water resources management. Therefore, the main objective of this study was to determine and simulate the impact of climate change on surface runoff in Ilala watershed of Northern Ethiopia. The specific objectives were to (1) assess and model climate change scenario over the study area (2) determine surface runoff generation and quantify the evapotranspiration of the study area, and (3) evaluate possible effects of climate change on surface runoff for the future.

Description of the study area Ilala watershed (Fig. 1) is located at 13°23´ to 13°31´30″ in North and 39°27´ to 39°31´48″ in East of Tigray region, Northern Ethiopia. The watershed covers an area of 215 km2 and has an altitude ranging from 1964–2680 meter above sea level (m.a.s.l). The climate of Ethiopia influenced by the Indian and Atlantic Oceans. As the area is characterized by a mono modal rainfall type the long rainy season (summer) lasts from June to September and locally called "kirmt”. It receives rainfall only from June to September. The rainfall distribution during this period varies between 240 and 398 mm with a peak rainfall in August. The “Kirmit” season contributes about 83% of the annual rainfall, while about 17% of the annual rainfall comes from May and October.

Modeling Earth Systems and Environment

Fig. 1  Location map of Ilala watershed

The average maximum and minimum temperature is 28 and 11 °C respectively. The highest maximum temperature is observed in May (29 °C) and June (30 °C). The minimum temperature record was observed in December (8 °C) and January (7 °C) and both months were dry and free of substantial rain.

Land use/land cover Land use and land cover is an important factor affecting different processes of a watershed, such as surface runoff, infiltration, recharge and evapotranspiration. Following the basic principles of the land use/land cover classification system, the study area was classified into eight classes: cultivated land (AGRL)—33%, built-up area (URBN)—24%, bush land (RNGE)—15%, bare land (BARL)—8%, shrubland (RNGB)—16%, grass land (PAST)—3.5%, forest-land (FRST)—(0.1%) water body (WATER)—0.5% as shown in Fig. 2. Built-up area and cultivated land were the most dominant types of land cover types in the study area.

Soil and geology  The response of runoff to a rainfall event depends on the nature and conditions of the underlying soils. The most dominant soil type in the watershed is Calcaric Cambisol (110 km2), Calcic Vertisol (95 km2) and Eutic Vertsol

(10 km2) (Ministry of Water Resources 1997). The SWAT model requires soil property data such as the texture, chemical composition, physical properties, and available moisture content, hydraulic conductivity, bulk density and organic carbon content of different layers. The soil information was collected from Tekeze River Basin Master Plan Project of Ethiopia (TRBMPP). The geology of Ilala catchment is highly disturbed due to volcanic eruptions. The catchment is mainly dominated by limestone-shale-marl intercalation, limestone, dolerite, sills and quaternary sediments (Geological map of Ethiopia 1996).

Materials and methods Data collection Climate and hydrological data Daily climate data (1980–2009) of the Mekelle gauging station (rainfall, air temperature, wind speed, humidity, and sunshine hour) were collected from Ethiopia National Metrological Agency (NMA). For each data variables, two main activities have been carried out in order to validate and screen the reliability of input data. The first activity was selection of good representative data years for both rainfall and temperature that have long term

13



Modeling Earth Systems and Environment

Fig. 2  land use and land cover

record. Finally, data analysis and interpretations has been conducted using Excel spread sheets to sort out outliers. The screened climatic data were used as an input to the hydrological model. In addition, 12 years (1990–2002) of observed hydrological river discharge were obtained from Ethiopia Ministry of Water, irrigation and Electric city for the purpose of model calibration and validation. The period from 1990 to 1996 was used for calibration whereas; the period from 1997 to 2002 was used for validation of the SWAT model.

13

Spatial data The digital elevation model (DEM), soil map, and land use map were used by the SWAT model to delineate the hydrological response units. The 30 m DEM was obtained from United State Geological Survey Database (USGS) (http:// glovi​s.usgs.gov/). The land use and land cover of Ilala watershed was prepared from Landsat imaginary products (http://glovis​ .usgs.gov/) with spatial resolution of 30 m. The images were downloaded in a dry season January month to

Modeling Earth Systems and Environment

minimize errors of haze and cloud. ERDAS imagine 2014 was used to process and classify the image. The soil physical properties (e.g. bulk density, available water capacity, hydraulic conductivity, saturated hydraulic conductivity, particle-size distribution) were taken from TRBMPP. Data quality Before starting any data analysis, quality of the data collected (missing data, consistency, and outliers) was checked using double mass curve analysis method. The Double mass curve is a simple, visual and practical method, and it is widely used in the study of the consistency and long-term trend test of hydro-meteorological data. Moreover, the Agricultural Modern Era Retrospective Analysis for Research and Application (AgMERRA) with 0.25° × 0.25° resolution of bias-corrected reanalysis data were used for filling the missing values of the observed climate datasets.

three time periods. As the CMPI5 have 20 GCMs in total we selected only 5 GCMs for further analysis which have low errors between the observed and the baseline data (Table 1). Using multi-GCMs for climate modelling helps to minimize the degree of uncertainties than using a single GCM. Making a conclusion about the effect of climate change on the catchment hydrology using a single GCM may not give a clear representation of the future changes. High uncertainty is expected with climate change impact studies if the simulation is a result of a single GCM (IPCC 2014; Mango et al. 2011).

Hydrological modelling

Method

The Soil and Water Assessment Tool (SWAT) (Arnold 1998), was used for simulating flows and evapotranspiration of the study area. SWAT is one of the most widely used watershed modeling tools, applied extensively in a broad range of water quantity and quality problems worldwide (Gassman et al. 2014). The details of SWAT model description are here below explained.

Climate change modelling scenarios

Description of SWAT model

To generate the future climate change of the study area, delta based statistical downscaling approach of the phase Five Coupled Model Intercomparison (CMIP5) was used in R software version 3.4.2. The general circulation model (GCM) was undergone in the recent two representative concentration pathways (RCPs) 4.5 and 8.5 emission scenarios. The Agricultural Model Intercomparison and Improvement Project (AgMIP) guidelines and scripts were used to run the model in “R” software. Detailed descriptions of RCPs and their equivalent representation for atmospheric C ­ O2 concentrations are available in Wayne (2013) and Ruane et al. (2015). The time period was classified according to AgMIP protocol as; 1980–2009 (baseline period), 2018–2039 (nearterm), 2040–2069 (midterm) and 2070–2099 (end-term) periods respectively (Rosenzweig et al. 2013; Thomson et al. 2011). Hence, changes in rainfall, minimum temperature and maximum temperature for future were projected for the

The hydrological model used in this study is a continuous time model that operates on a daily/sub-daily time step. It is physically based and can operate on large basins for long periods of time (Arnold et al. 1998; Neitsch et al. 2005). The hydrological cycle in SWAT model is operated based on the water balance equation as shown below (Eq. 1);

SWt = SW0 +

t ∑

(Rday − Qsur − Ea − Wdeep − Qgw )

(1)

i

where: ­SWt, is the final soil water content in mm, ­SW0, is the initial soil water content on the day (mm), t is the time days, ­Rday is the amount of precipitation in a day (mm), ­Qsurf is the amount of surface runoff in a day (mm), Ea is the amount of evapotranspiration on the day in mm, ­Wdeep is the amount of water entering the vadose zone from the soil profile in a day (mm), ­Qgw is the amount of return flow in a day (mm).

Table 1  Detail descriptions of the selected GCMs of the CMIP5 SN Model code GCMs

Institutional reference

Horizontal resolution

1 2 3 4

E I K O

CCSM4 GFDL-ESM2M HadGEM2-ES MIROC5

5

Q

MPI-ESM-LR

US National Center for Atmospheric Research (NCAR) ∼ 0.9° × 1.25° NOAA/Geophysical Fluid Dynamic Laboratory (GFDL) (Modular Ocean Model) ∼ 2.0° × 2.5° UK Meteorological Office—Hadley Centre (all Earth system components) 1.25° × 1.875° ∼1.4° × ∼ 1.4° University of Tokyo, Japanese National Institute for Environmental Studies (NIES) and Japan Agency for Marine-Earth Science and Technology (JAMSTEC) Max Planck Institute (MPI) for Meteorology (low resolution) ∼1.9° × 1.875°

13



Modeling Earth Systems and Environment

The model uses the concept of infiltration excess runoff mechanism. It assumes the runoff occurs whenever the rainfall intensity is greater than the rate of infiltration (Neitsch et al. 2005). This process is very important in areas where significant soil crusting and/or surface sealing occurs during storm events, in irrigated fields, in urban areas and more generally during very high rainfall intensity storms. For estimation of surface runoff, SWAT uses two methods based on the above assumption. The soil conservation curve number method and Green and Ampt infiltration method (Green 1911). For this particular research, the soil conservation services (SCS) curve number was employed. Because of its capability to use daily input data the SCS is commonly used. The SCS curve number is described in Eq. (2) as follows:

Qsurf =

(Rday − 0.2)2 (Rday + 0.8S)

(2)

where: ­Qsurf is the accumulated runoff or rainfall excess (mm), ­Rday is the rainfall depth for the day (mm), S is the retention parameter (mm). The retention parameter is defined by Eq. (2) with curve number (CN), as shown in Eq. (3) below: ( ) 100 S = 25.4 − 10 (3) CN SWAT calculates the peak runoff rate with a modified rational method. In rational method, it is assumed that a precipitation of intensity I begins at time t = 0 and continues indefinitely, the rate of runoff will increase until the time of concentration, t = tconc. The modified rational method is mathematically expressed as in Eq. (4) below:

qpeak =

𝛼tc ∗ Qsurf ∗ A 3.6 ∗ tconc

(4)

where, qpeak is the peak runoff rate (­ m3/s), αtc is the fraction of daily precipitation that occurs during the time of concentration, Qsurf is the surface runoff (mm), Area is the sub-basin area ­(km2), tconc is the time of concentration (hr), and 3.6 is a conversion factor. SWAT model provides three methods for estimating potential evapotranspiration: Penman-Monteith, Priestly-Taylor and Hargreaves methods (Nietsch et al. 2005). The three methods included in SWAT vary in the amount of required inputs (Neitsch et al. 2005). For this study Hargreaves method was employed as it required relatively a limited data.

Model setup SWAT model requires intensive data including topography, soil, and land use and weather data as input. To capture the spatial and temporal variations of the watershed, it is

13

necessary to delineate the watershed into smaller sized sub basin areas where the variables can be considered homogenous. Furthermore, the digital elevation model (DEM) was used to divide the watershed into several hydrological response units and to predict the location of the stream. The Land use and soil maps were used to define the characteristics of land cover and soil properties of the watershed, respectively. Using the combination of the land use, DEM, soil and slopes, the hydrological response units were developed to simulate the basin characteristics. Furthermore, a weather generator model from statistical data summarized over long-term monthly average series was developed in order to fill missing values and to generate the other climatic parameters (wind speed, sunshine hours and solar radiation).

Sensitivity analysis, calibration and validation of the model Sensitivity analysis is crucial in modeling as it helps to understand the rate of change in the outputs of the model as a result of changes in model inputs (Gassman et al. 2007). Knowing which parameter of the model is more sensitive to the given inputs can help to determine parameters values which is important to have more accurate values in calibration of model and thus to better understand the characteristics of hydrological processes in a given watershed. In this study, Latin Hypercube One-factor-At-a-Time (LH-OAT) was employed. The LH-OAT sensitivity analysis combines the strength of global and local sensitivity analysis methods (Van Griensven et al. 2006). After sensitivity analysis, model calibrations were done by selecting the most sensible parameters of the model. This was done by checking results against the observations at the watershed outlet to ensure similar response over time which involves comparing the model outputs, generated with the recorded stream flows. In this process, model parameters varied until recorded flow patterns were accurately simulated.

Uncertainty analysis Uncertainty analysis was performed after sensitivity analysis using SWAT-CUPv.2012 (Soil Water Analysis Calibration and Uncertainty Program) software. SWAT-CUP SUFI-2 (Sequential Uncertainty fitting version 2) was used for uncertainty and calibration, validation process (Abbaspour et al. 2004). The degree of uncertainties was measured as the P-factor, which is the percentage of observed data related by the 95PPU (95% prediction uncertainty). The 95PPU is calculated at the 2.5 and 97.5% levels of cumulative distribution of the output variables. Another measure quantifying the strength of uncertainty analysis was the R-factor, which is the average thickness

Modeling Earth Systems and Environment

of the 95PPU band ( r̄  ) divided by the standard deviation of the measured data as described in Eqs. (5) and (6): n

r=

1∑ M (y ti , 97.5% − yM ti , 2.5%) n t

(5)

i

r − factor =

p − factor 𝜎obs

Evaluation of model performance The performance of SWAT model was evaluated using both qualitative and quantitative measures such as graphs and statistical indices based on simulated and observed values. In this study, the performance of the model was evaluated using both hydrograph comparisons, statistical indices; the Nash–Sutcliffe simulation efficiency (ENS) and the coefficient of determination ­(R2). The details of those statistical indices (NSE and ­R2) are well documented in Neitsch et al. (2005), Green (1911), Gassman et al. (2007) and Moriasi et al. (2007a, b):

� ∑� ̄ m )(Qs − Q ̄ s) 2 (Qm − Q

(7)

i

2

̄ i (Qmi − Qm )

∑

(8)

where: ­Qm is the measured discharge, Qs is the simulated ̄ m is the average measured discharge and Q ̄ s is discharge, Q the average simulated discharge.

(6)

where; y M ti , 97.5% and yM ti , 2.5% represents the upper and lower boundaries of the 95PPU, and 𝜎obs is the standard deviation of the measured data.

R2 = ∑

∑ (Qm − Qs )2 NSE = 1 − ∑ i ̄ 2 i (Qm − Qm )

2

̄ i (Qs − Qs )

Table 2  Changes in maximum temperatures (°C) compared to the baseline across the five GCMs in RCP4.5 and RCP8.5

Table 3  Changes in mean minimum temperature (°C) compared to the baseline across the five GCMs in RCP4.5 and RCP8.5

Result and discussion Climate change modelling scenarios Future temperatures have generally increased with the time period in both RCP’s across all the GCMs. Many studies also clearly indicated that minimum and maximum temperature are expected to increase in the future (Araya et al. 2015b; Ashenafi 2014). The highest minimum and maximum temperature were simulated during the end-term period under RCP8.5 (Tables 2, 3). The highest temperature was recorded in “HadGEMs-ES”model (+ 5.7 and + 6.3 °C) in RCP8.5 for both minimum and maximum temperature respectively (Table 4). The lowest temperature also recorded in CCSM4 model (+ 1.3 °C), GFDL-ESM2M model (1.4 °C), MIROC5 (1.4 °C) model, in RCP 4.5 near-term periods respectively. Similarly, highest mean annual rainfall was simulated during the end term period under RCP8.5 with the “MIROC5” (+ 42.5%), whereas the lowest mean annual rainfall was simulated during the near-term period under RCP4.5

T_Max GCMs

RCP4.5 Near-term

RCP4.5 Mid-term

RCP4.5 End-term

RCP8.5 Near-term

RCP8.5 Mid-term

RCP8.5 End-term

CCSM4 GFDL-ESM2M HadGEMs-ES MIROC5 MPI-ESM-MR Average

1.3 1.5 3.1 1.6 2.1 1.7

1.5 1.6 2.7 1.5 2.2 1.9

1.7 2.4 3.6 1.8 2.5 2.4

1.9 2.5 3.8 1.3 2.7 2.6

2.1 2.6 3.5 1.4 2.9 2.5

3.7 3.9 5.7 2.5 4.8 4.1

GCMs

RCP4.5 Near-term

RCP4.5 Mid-term

RCP4.5 End-term

RCP8.5 Near-term

RCP8.5 Mid-term

RCP8.5 End-term

CCSM4 GFDL-ESM2M HadGEMs-ES MIROC5 MPI-ESM-MR Average

1.8 1.4 3.0 1.4 2.6 2.3

1.7 1.5 3.1 1.6 2.4 2.1

2.0 2.2 3.9 1.9 2.8 2.5

2.0 2.8 4.9 1.9 2.8 2.9

2.3 2.7 4.1 1.8 3.1 2.8

4.0 4.0 6.3 3.1 5.3 4.5

T_min

13



Modeling Earth Systems and Environment

Table 4  Changes in mean annual rainfall (%) compared to the baseline across the five GCMs in RCP4.5 and RCP8.5 Rain (%) GCMs

RCP4.5 Near-term

RCP4.5 Mid-term

RCP4.5 End-term

RCP8.5 Near-term

RCP8.5 Mid-term

RCP8.5 End-term

CCSM4 GFDL-ESM2M HadGEMs-ES MIROC5 MPI-ESM-MR Average

7.7 4.6 − 6.2 15.7 − 13.6 1.1

7.2 4.8 − 9.2 13.0 − 13.2 0.5

8.5 − 1.2 − 3.2 18.4 − 14.1 1.7

4.8 2.7 8.7 32.0 − 13.2 6.6

4.1 1.2 1.7 33.0 − 12.2 5.6

7.5 − 9.7 15.7 42.5 − 18.2 7.6

Table 5  Statistical outputs of SUFI-2 algorithm

Calibration Validation

Variable

P-factor (%)

R-factor

R2

NSE

Flow Flow

71 74

0.5 0.6

0.54  0.63

0.51  0.54

in GFDL-ESM2M (− 1.2%) (Table 4). The difference among the simulated climate outputs could be mainly due to the basic modeling structures and parameterization of the GCMs (Araya et al. 2017). It is assumed that the climate impacts of multi-model predictions could help to explore the magnitude of changes and the likely occurrence of events together with reasonable uncertainty (Araya et al. 2015a).

Hydrological modelling Sensitivity and uncertainty analysis Twenty-seven hydrological parameters were tested for sensitivity analysis for the simulation of the stream flow in the study area. From the total 27 hydrological flow parameters, 4 parameters were found to be sensitive. Curve number (CN2), Base flow alpha factor (ALPHA_BF), Groundwater delay (GW_DELAY) and Threshold depth of water in the shallow aquifer required for return flow to occur (GWQMN.gw) have shown a relatively higher sensitivity as described in Table 5. The analysis was conducted for the entire period of calibration and validation periods. SUFI-2 is given several iterations to get the acceptable results. Each of iterations provides the suggested values for the new parameters to be used in the next iteration. Finally, it provides the acceptable result with the values of the Nash–Sutcliffe, Coefficient of determination and other statistical parameters. The p-factor, which is the percentage of observations bracketed by the 95% prediction uncertainty (95PPU), brackets 71, 74% of the observation during calibration and validation periods. R-factor equals to 0.5 and 0.6. The result categorized within the acceptable range values as recommended by Abbaspour et al. (2004).

13

Based on these recommendations, the performance of SWAT model for the study area was good during the calibration period with NSE > 0.50. The uncertainty analysis indicated parameter effectiveness  of  SCS curve number for moisture condition II (CN2.mgt), base-flow alpha factor for bank storage (ALPHA_BNK), Groundwater delay (GW_DELAY) and Threshold depth of water in the shallow aquifer required for return flow to occur (GWQMN.gw) played an important roles in the calibration and validation of SWAT model. Table 5 explains the computed P-factor, R-factor, ­R2 and NSE values of the analysis.

Calibration and validation surface runoff using SUFI‑2 algorithm In this study, the comparisons between the observed and simulated daily values of flow of the watershed were done using different evaluation criteria. Those were, comparing annual water balance, statistical index Nash Sutcliffe efficiency (ENS), coefficient of determination (­ R2), and graphical comparison of hydrograph shape. The optimum parameter values obtained during calibration period were used to validate the model with independent observed data. The model was calibrated using SWAT-CUP SUFI-2 algorism and the comparison between the observed and simulated stream flow showed a reasonable good agreement (Fig. 3). As shown below in Table 6 four main hydrological parameters (CN2.mgt, ALPHA_BF, GW_DELAY.gw and GWQMN.gw) were selected which create significant variation in the calibration. Accordingly, by adjusting the values of parameters, the hydrograph of calibration was fitted as shown in the graph (Fig. 3). The result shows that the timing and shape of the runoff hydrograph were well predicted. The low flow and high flows (peak flow) are well represented as well. The performance of the model was also valued using statistical indicators and based on the R ­ 2 and ENS. In the 2 calibration period ­R was 0.55 and ENS = 0.51, while for the validation period was 0.63 and 0.54 (Fig. 4). These values can be considered as satisfactory as suggested by Moriasi et al. (2007a, b).

Modeling Earth Systems and Environment Fig. 3  Calibration period of observed vs simulated Surface runoff of Ilala watershed

Table 6  Maximum and the parameters and fitted values after calibration

Parameters

Descriptions

Fitted value

Min_value

Max_value

CN2.mgt ALPHA_BF.gw GW_DELAY.gw GWQMN.gw

SCS curve number for moisture condition II Base flow alpha factor Groundwater delay (days) Treshold depth of water in the shallow aquifer required for return flow to occur (mm)

− 0.2 0.65 303 1.9

− 0.2 0 30 0

0.2 1 450 2

Fig. 4  Validation period of observed vs simulated Surface runoff of Ilala watershed

The Figs.  3 and 4 shows that the mean monthly observed and simulated discharge of the calibration and validation periods. It is known that the main rainfall season of the study area is between June and September and the higher surface runoff also occurs during these months. The coefficient determination for the calibration and validation periods was 0.54 and 0.63. The higher value of R ­ 2 implies that the simulated discharge agreed well with the observed discharge. The detailed statistics of simulated and observed values are presented in Table 6. The rainfall in the area comes with great intensity over a short period of time and concentrated in a few months. Hence, much water is falling on the ground very rapidly and gets little time to percolate-in and most of the water flows as runoff into nearby rivers and lakes.

Evapotranspiration The highest evapotranspiration were recorded in April and March with a value of 184.3 and 181.8 mm per month respectively (Fig. 5). In contrast, the lowest evapotranspiration were recorded in July and August 69 and 63 mm per month. The evapotranspiration loss was considerable in the study area. Almost 75% of the total rainfall received was changed into evapotranspiration.

Rainfall‑ runoff relationships The surface runoff has direct relation with the rainfall as shown in Fig.  6. The higher the rainfall, the higher the surface runoff. Rainfall and runoff are responsible factors

13



Modeling Earth Systems and Environment

Fig. 5  Evapotranspiration trend of the study area

for the detachment, transport, and deposition of sediment particles. At the beginning of the rainy season, surface runoff increased rapidly, with a peak in June in most of the year. The average peak rainfall in the area was in July and the average peak discharge was in August. Both the simulated and observed flows indicated that during July and August, there was a highest intensity of rainfall that contributed to high surface runoff. The runoff was highly correlated with rainfall ­(R2 = 0.97) as shown in Fig. 7.

Water balance of the watershed The annual average rainfall and other hydrological components were compared for each year of the calibration and validation periods. The main water balance components of

Fig. 6  Rainfall runoff relationships

13

the watershed include: the total amount of rainfall falling on the watershed, actual evapotranspiration and the net amount of water that leaves the watershed and contributes to stream flow in the reach (water yield). The water yield during validation period was higher than the calibration period. This could be because the rainfall amount during the validation period was relatively higher. Table 7 showed that water balance components of Ilala watershed during the simulation period. From these components only, 6.2% of the total rainfall received in the study area was changed into surface runoff, while the majority of the component around 75% has contributed to evapotranspiration. Ground water recharge contributed about 7.8% of the total water balance. This high evapotranspiration loss could be due to temperature fluctuations and dry wind weather condition of the watershed (Tesfaye et al. 2017).

Modeling Earth Systems and Environment

Impact on future hydrological response In the mean ensemble model result in RCP4.5, the change in surface runoff ranged from 1.75 to 0.74%, whereas, in RCP8.5 from 0.76 to 0.36% as shown in Table 8. The minimum and maximum monthly variation change in surface runoff volume is 0.36% in RCP8.5 end-term and RCP4.5 in near-term (1.74%) respectively as shown in Table 8. The overall increase in temperature and rainfall results in reduction in surface runoff in the watershed. The result of the

study has a similarity with findings of other researches like (Tesfaye et al. 2017; Ashenafi 2014). As many studies indicated that increasing water abstractions, particularly in the semi-arid catchments of the basin, might have caused the decline of stream flow during dry and small rainy seasons (Gebremicael et al. 2016; Alemayehu et al. 2009; Nyssen et al. 2010). Hence, this could be due to surface and shallow groundwater development and abstraction for irrigation have significantly increased since the mid-2000s, after implementation of intensive catchment management programmes as

Fig. 7  Correlation of rainfall runoff relationships

Table 7  Water balance of Ilala watershed

Period

Rainfall (mm)

ET (mm)

Qsurf (mm)

Qlat (mm)

GWQ (mm)

Water yield (mm)

SW (mm)

PERC (mm)

TLosses (mm)

Calibration % Validation %

511 100 560 100

400 78 407 72.0

28 5.5 35 6.25

8 1.6 15 2.7

38 7.4 46 8.2

82 16.0 103 18.4

10 1.9 11 1.9

11 2.1 13 2.3

2 4

ET actual evapotranspiration, Qsurf surface runoff, Qlat lateral flow, GWQ ground water contribution to stream flow, SW soil water content, PERC water that percolates past the root zone, TLosses water lost from tributary channels Table 8  Future runoff under change in climate

Runoff (%) GCMs

RCP4.5 RCP4.5 MidNear-term term

RCP4.5 RCP8.5 NearEnd-term term

RCP8.5 RCP8.5 End-term Mid-term

CCSM4 GFDL-ESM2M HadGEMs-ES MIROC5 MPI-ESM-MR Average

1.2 4.8 1.6 0.3 0.8 1.74

0.6 0.9 1.3 0.3 0.6 0.74

0.14 0.3 1.21 0.5 0.5 0.53

0.3 0.42 1.5 1.23 1.3 0.95

0.5 1.3 1.2 0.4 0.4 0.76

0.21 0.4 0.21 0.7 0.3 0.364

13



explained in Gebremicael et al. (2016). Moreover, climate change has also a considerable effect on declining the surface runoff (Abebe 2014; Ashenafi 2014; Tesfaye et  al. 2017). In most of the models except HadGEMs-ES they showed both increasing and decreasing trends in surface runoff in all RCPs and time periods. However, the overall average trend showed that a decreased trend in surface runoff for the future in all time segments and RCPs from near-term to end-term periods respectively. This could be potentially attributed to the effect of climate change (Gebremicael et al. 2017). The other reason could be that the total amount of rainfall received in the study area has a relatively higher amount of water converted to ground water recharge (7.8%) than the surface runoff (6.2%). As reported by Nyssen et al. (2010); Gebremicael et al. (2017) there is an increasing trend in water abstractions particularly in the semi-arid catchments of the basins, and this might have caused to decline of surface runoff during dry and small rainy seasons. Moreover, due to the increment of the temperature there was high evaporation loss in the study area. This all could contribute to reduction in surface runoff in the study area.

Conclusions Minimum and maximum temperature values show a considerable variation in the future in the study area. The minimum temperature shows an increase trend by 2.3 °C in near-term and 4.5 °C in end-term periods. Similarly, maximum temperature increase from 1.7 to 7.6 °C in near-term and end-term time periods. However, the rainfall did not show a systematic increase or decrease trend in the study area. The SWAT was successfully used to simulate the hydrological dynamics of Ilala watershed. Sensitivity analysis was done to select the most sensitive parameters for further calibration processes. SWAT-CUP was also used to calibrate and validate the model. The calibration parameters of SWAT were selected based on sensitivity analysis of model results. The graphical comparison of simulated and observed time series of the monthly flows has shown good match of the hydrographs. The calibration results of the flows, showed that Nash–Sutcliffe efficiency (NSE) of (0.51 and 0.54) and coefficient of determination (­ R2) (0.54, 0.63) for the calibration and validation periods respectively. The total average simulated water yield was 188 mm/year. The water yield was higher in the validation period than the calibration period, which could be attributed to higher rainfall during the validation period. In line with this, 6.2% of the total rainfall received in the area was changed into surface runoff. The surface runoff is directly associated with rainfall. The rainfall- runoff relationships was strongly correlated with ­R2 = 0.97. In addition, there had been high evapotranspiration loss in the

13

Modeling Earth Systems and Environment

watershed as almost 75% of the total rainfall received were lost as evapotranspiration and only 7.8% as ground water recharge. For all investigated periods the average surface runoff for future will decrease under climate change scenarios with a value of 1.74–0.36%. There will be a considerable effect of climate change on surface runoff. Thus, the findings could be useful for water managers’ decisions’ and policymakers. It gives also a direction mechanism for adaptation and mitigation measures. Moreover, it helps to implement appropriate watershed management activities and to ensure sustainable water resources management in the watershed and in other agro-ecologically similar watersheds. Acknowledgements  The research was supported by the Open Society Foundation-Africa Climate Change Adaptation Initiative (OSFACCAI) project at  Institute of Climate and Society of Mekelle University (MU-ICS). The authors also would like to thank the Ethiopia National Metrological Agency (NMA), and Ethiopian Ministry of Water, Irrigation and Electricity for providing meteorological and hydrological data of the study area. The authors express sincere appreciation to Agricultural Model Intercomparison and Improvement Project (AgMIP) for their initiation and development of the technical scripts for climate modelling.

Compliance with ethical standards  Conflict of interest  The authors declare that they have no conflict of interest.

References Abbaspour KC, Johnson CA, van Genuchten MT (2004) Estimating uncertain flow and transport parameters using a sequential uncertainty fitting procedure 3:1340–1352 Abebe BA (2014) Modeling the effect of climate and land use change on the water resources in Northern Ethiopia: the case of Suluh River Basin, p 23 Alemayehu F, Taha N, Nyssen J et al (2009) The impacts of watershed management on land use and land cover dynamics in Eastern Tigray (Ethiopia). Resour Conserv Recycl 53:192–198. https​://doi. org/10.1016/j.resco​nrec.2008.11.007 Araya A, Girma A, Demelash T et al (2015a) Assessing impacts of climate change on tef (Eragrostis tef) productivity in Debrezeit area. Ethiopia 4:39–48 Araya A, Hoogenboom G, Luedeling E et al (2015b) Assessment of maize growth and yield using crop models under present and future climate in southwestern Ethiopia. Agric For Meteorol 214– 215:252–265. https​://doi.org/10.1016/j.agrfo​rmet.2015.08.259 Araya A, Kisekka I, Girma A et al (2017) The challenges and opportunities for wheat production under future climate in Northern Ethiopia. J Agric Sci 155:379–393. https:​ //doi.org/10.1017/S0021​ 85961​60004​60 Arnold JG (1998) The GIS tool chosen was the geographical 34:91–101 Arnold JG, Srinivasan R, Muttiah RS, Williams JR (1998) Large area hydrologic modeling and assesment part I: model development. JAWRA J Am Water Resour Assoc 34:73–89. https​://doi. org/10.1111/j.1752-1688.1998.tb059​61.x Ashenafi AA (2014) Modeling hydrological responses to changes in land cover and climate in Geba River Basin, Northern Ethiopia. Ph.D. Thesis, Freie Univ Berlin, Ger 187

Modeling Earth Systems and Environment Awulachew SB, Erkossa T, Smakhtin V, Fernando A (2009) Improved water and land management in the ethiopian highlands: its impact on downstream stakeholders dependent on the Blue Nile Eromo S, Adane C, Santosh A, Pingale M (2016) Assessment of the impact of climate change on surface hydrological processes using SWAT: a case study of Omo-Gibe river basin, Ethiopia. Model Earth Syst Environ 2:1–15. https​://doi.org/10.1007/s4080​ 8-016-0257-9 Gassman WP, Reyes MR, Green CH, Arnold JG (2007) The soil and water assessment tool: historical development, applications, and future research directions. Trans ASABE 50(4):1211–1250 Gassman PW, Sadeghi AM, Srinivasan R (2014) Applications of the SWAT model special section: overview and insights. J Environ Qual 43:1. https​://doi.org/10.2134/jeq20​13.11.0466 Gebrekristos ST (2015) Understanding catchment processes and hydrological modelling in the Abay/Upper Blue Nile basin, Ethiopia Gebremicael TG, Mohamed YA, Betrie GD et al (2013) Trend analysis of runoff and sediment fluxes in the Upper Blue Nile basin: a combined analysis of statistical tests, physically-based models and landuse maps. J Hydrol 482:57–68. https​://doi.org/10.1016/j. jhydr​ol.2012.12.023 Gebremicael TG, Mohamed YA, van der Zaag P, Hagos EY (2016) Temporal and spatial changes of rainfall and streamflow in the Upper Tekeze–Atbara River Basin, Ethiopia. Hydrol Earth Syst Sci Discuss 0:1–29. https​://doi.org/10.5194/hess-2016-318 Gebremicael TG, Mohamed YA, Zaag PV, Hagos EY (2017) Temporal and spatial changes of rainfall and streamflow in the Upper Tekezē-Atbara river basin, Ethiopia. Hydrol Earth Syst Sci 21:2127–2142. https​://doi.org/10.5194/hess-21-2127-2017 Geological map of Ethiopia (1996) Geology of Ethiopia Green AGA (1911) Studies on soil physics. Agric Sci 4:1 (1–24) Gyamfi C, Ndambuki M, Anornu GK, Gislar Edgar K (2017) Groundwater recharge modelling in a large scale basin: an example using the SWAT hydrologic model. Model Earth Syst Environ 3:8-017-0383 Hurni H, Tato K, Zeleke G (2005) The implications of changes in population, land use, and land management for surface runoff in the upper Nile Basin area of Ethiopia. Mt Res Dev 25:147–154 IPCC (2014) Summary for policymakers Kumar S, Singh S (2016) Modelling spatially distributed surface runoff generation using SWAT-VSA: a case study in a watershed of the north-west Himalayan landscape. Model Earth Syst Environ 2:1–11. https​://doi.org/10.1007/s4080​8-016-0249-9 Makwana JJ, Tiwar MK (2017) Hydrological stream flow modelling using soil and water assessment tool (SWAT) and neural networks (NNs) for the Limkheda watershed, Gujarat, India. Model Earth Syst Environ 3:635–645 Mango LM, Melesse AM, McClain ME et al (2011) Land use and climate change impacts on the hydrology of the upper Mara River Basin, Kenya: results of a modeling study to support better resource management. Hydrol Earth Syst Sci 15:2245–2258. https​ ://doi.org/10.5194/hess-15-2245-2011 Milly PCD, Dunne KA, Vecchia AV (2005) Global pattern of trends in streamflow and water availability in a changing climate. Nature 438:347–350 Ministry of Water Resources (1997) Tekeze river basin integrated development master plan project Ministry Water Resources M (2002) Water sector developemnt program main report of Ethiopia. Addis Ababa, Ethiopia Moriasi DN, Arnold JG, Van Liew MW et al (2007a) Model evaluation guidelines for systematic quantification of accuracy in watershed

simulations. Trans ASABE 50:885–900. https​://doi.org/10.13031​ /2013.23153​ Moriasi DN, Arnold JG, Van Liew MW et al (2007b) Model evaluation guidelines for systematic quantification of accuracy in watershed simulations. 50:885–900 Neitsch SL, Arnold JG, Kiniry JR (2005) SWAT theoretical documentation (version 2005). Grassland, Soil and Water ResearchLaboratory, Agricultural Research Service, Temple, TX. pp 95–116 Nyssen J, Poesen J, Gebremichael D et al (2007) Interdisciplinary onsite evaluation of stone bunds to control soil erosion on cropland in Northern Ethiopia. 94:151–163. https​://doi.org/10.1016/j.still​ .2006.07.011 Nyssen J, Clymans W, Descheemaeker K et al (2010) Impact of soil and water conservation measures on catchment hydrological response—a case in north Ethiopia. Hydrol Process 24:1880– 1895. https​://doi.org/10.1002/hyp.7628 Pandey A, Himanshu SK, Mishra SK, Singh VP (2016) Physically based soil erosion and sediment yield models revisited. Catena 147:595–620. https​://doi.org/10.1016/j.caten​a.2016.08.002 Raneesh KY, Santosh GT (2011) A study on the impact of climate change on streamflow at the watershed scale in the humid tropics. Hydrol Sci 56:946–965 Rosenzweig C, Jones JW, Hatfield JL et al (2013) Agricultural and forest meteorology the agricultural model intercomparison and improvement project (AgMIP): protocols and pilot studies. Agric For Meteorol 170:166–182. https ​ : //doi.org/10.1016/j.agrfo​ rmet.2012.09.011 Ruane AC, Goldberg R, Chryssanthacopoulos J (2015) Agricultural and forest meteorology climate forcing datasets for agricultural modeling: merged products for gap-filling and historical climate series estimation. Agric For Meteorol 200:233–248. https​://doi. org/10.1016/j.agrfo​rmet.2014.09.016 Taddese G (2001) Land degradation: a challenge to Ethiopia. Environ Manag 27:815–824. https​://doi.org/10.1007/s0026​70010​190 Tesfaye S, Birhane E, Leijnse T, Zee SEATM., Van Der (2017) Science of the total environment climatic controls of ecohydrological responses in the highlands of northern Ethiopia. Sci Total Environ 609:77–91. https​://doi.org/10.1016/j.scito​tenv.2017.07.138 Thomson AM, Calvin KV, Smith SJ et al (2011) RCP4.5: a pathway for stabilization of radiative forcing by 2100. Clim Change 109:77– 94. https​://doi.org/10.1007/s1058​4-011-0151-4 van Griensven A, Meixner T, Grunwald S et al (2006) A global sensitivity analysis tool for the parameters of multi-variable catchment models. J Hydrol 324:10–23. https​://doi.org/10.1016/j.jhydr​ ol.2005.09.008 Vancampenhout K, Nyssen J, Gebremichael D et al (2006) Stone bunds for soil conservation in the northern Ethiopian highlands: impacts on soil fertility and crop yield. Soil Tillage Res 90:1–15. https​:// doi.org/10.1016/j.still​.2005.08.004 Wayne GP (2013) The Beginner’s guide to representative concentration pathways (RCPs). Skept Sciece 1.0:1–24 Welde K, Gebremariam B (2017) International soil and water conservation research e ff ect of land use land cover dynamics on hydrological response of watershed: case study of Tekeze Dam watershed, northern Ethiopia. Int Soil Water Conserv Res 5:1–16. https​://doi.org/10.1016/j.iswcr​.2017.03.002 Yin J, He F, Xiong Y, Qiu G (2016) Effect of land use/land cover and climate changes on surface runoff in a semi-humid and semi-arid transition zone in Northwest China. Hydrol Earth Syst Sci Discuss. https​://doi.org/10.5194/hess-2016-212

13