Modeling Future Hydrological Responses Through Parameter Optimization and Climate Change Scenarios in Dirima Watershed, Ethiopia

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 19

Modeling Earth Systems and Environment

https://doi.org/10.1007/s40808-023-01817-z

ORIGINAL ARTICLE

Modeling future hydrological responses through parameter


optimization and climate change scenarios in Dirima Watershed,
Ethiopia
Simir B. Atanaw1   · Fasikaw A. Zimale2 · Tenalem Ayenew3 · Gebiaw T. Ayele4

Received: 25 May 2023 / Accepted: 11 June 2023


© The Author(s), under exclusive licence to Springer Nature Switzerland AG 2023

Abstract
Purpose  Hydrological modeling is an important tool for estimating hydrological responses, not only for current conditions
but also for future scenarios by optimizing hydrological parameters. In cases where direct measurements are difficult to
obtain, modeling can be used to fill in the gaps and provide vital information for water resources planning and management.
However, it is important to note that hydrological models need to be calibrated and validated to ensure that the parameter
values are optimized for the specific watershed being studied.
Materials and methods  The Xinanjiang (XAJ) model have been used to estimate the potential hydrological responses under
current and future scenarios. In this study, an attempt was made to optimize the parameters for the Dirima watershed with a
total area of 162 k­ m2 using DEoptim algorithms. In this study, the calibrated parameters were used to simulate the watersheds’
hydrological response for baseline (1996–2009) and three climate change scenarios (SSP2-4.5, SSP3-7.0, and SSP5-8.5)
from a CMIP6 multimodel ensemble (2015–2100).
Results and discussion  The study shows that the baseline period of performance measure values is RMSE = 12, NSE = 0.76,
PBIAS = 10.5% and ­R2 = 0.78, and RMSE = 3.65, NSE = 0.85, PBIAS = 9.9%, and ­R2 = 0.85 indicate that the model has per-
formed well in simulating the streamflow in both the calibration and validation periods, respectively. Lower RMSE, higher
NSE, and lower PBIAS indicate better model performance and suggest that the model has performed well in all these aspects.
In the future (2015–2100) the average Tmax 1.59 , 1.93  and 2.48 °C, and Tmin will raise by 1.83 , 2.33  and 2.85 °C under
SSP2-4.5, SSP3-7.0, and SSP5-8.5, respectively. Additionally, the mean annual future hydrological responses in the upper
soil layer, including evaporation, soil moisture, and runoff will be 1.72, 1.71, and 1.79 mm, 0.63, 0.62, 0.64 mm, and 1.07,
0.97, and 1.18 mm under SSP2-4.5, SSP3-7.0, and SSP5-8.5. The future projections show an increase in average maximum
and minimum temperatures under different SSP scenarios.
Conclusion  The calibrated parameters of the XAJ model were critical to assess the future hydrological responses of the
Dirima watershed under the three SSP scenarios. The precipitation and streamflow values are consistently lower than those
of ET and temperature across all SSP scenarios. This indicates that future water availability is likely to be under pressure
and calls for appropriate strategic measures to create a more resilient water resource system.

Keywords  Differential evolution algorithm · Hydrological responses · Xinanjiang (XAJ) model · Climate Change
modeling · CMIP6

Introduction non-point sources of pollution, and agricultural activities


take place, all of which have both direct and indirect impacts
Water resource systems are recently facing numerous chal- on the system (Tundu et al. 2018). So, understanding the
lenges due to global climate change and rapid population key hydrological processes using a hydrological model is a
growth. In a water resource and watershed system, various potential approach for improving the effectiveness of adopted
processes such as floods, erosion, sedimentation, point and management plans and policies, used to estimate streamflow
for ungauged watersheds, because the model provides the
current and future state of hydrological responses on the
Extended author information available on the last page of the article temporal and spatial scale. By calibrating the parameters

13
Vol.:(0123456789)
Modeling Earth Systems and Environment

specifically for future scenarios, we can enhance the reliabil- (NWSRFS) have also been used to quantitatively simulate
ity of hydrological models in estimating how water systems flow and water quality in watersheds (Soohong et al., 2022).
will respond to changing climate conditions. Various vari- However, these models need a high temporal and geographic
ables affect the fluctuation of hydrological response in the resolution of topography, precipitation, evapotranspiration,
watershed. These include geographical differences (Siswanto and other inputs, all of which have a direct influence on
and Sule 2019), soil characteristics (Patil et al. 2012), plant the model output. Additionally, these models have a limited
cover and density, mean humidity, soil water depth, precipi- capacity to capture the hydrological response of the upper,
tation, and other meteorological variables (Guo et al. 2020). lower, and deep layers of the soil. Conceptual models such
For example, the relationship between SM and precipitation as TANK, HBV, GR4J, and XAJ employ basic mathemati-
is critical, with the direction of the correlation (positive or cal elements to simulate the hydrological response approxi-
negative) and the amplitude varying regionally (i.e. weak mately (Qi et al. 2022; Wang et al. 2021). Recently, gravity
or strong) (Sehler et al. 2019), and concluded that in areas recovery and climate experiment (GRACE) and climate data
with limited vegetative cover, precipitation, and SM have the were used to evaluate the terrestrial water storage (TWS)
strongest correlations, whereas forests and vegetated areas main components and long-term drought monitoring. Oth-
have weaker correlation coefficients. man et al. (2022) evaluate the spatiotemporal mass vari-
Assessing the impacts of climate change on water ations caused by groundwater changes, Gonçalvès (2013)
resources, often related to some aspect of water availabil- estimates groundwater recharge over Sahara aquifers through
ity, is a key variable that motivates scientific research and water storage from GRACE and soil moisture from global
public policies. Thus, we assessed the future hydrological land data assimilation system (GLADS), and the result
responses in the Dirima watershed, which is of paramount shows that recharge variation was correlated at one year lag
importance for understanding the water balance. The hydro- period. Mohamed and Gonçalvès (2021) investigate the vari-
logical response was analyzed considering three climate ation of groundwater recharge reconstruction and reveal that
change scenarios within the coupled model intercompari- the lag period for groundwater recharge reaching the water
son project phase 6 (CMIP6). The Dirima watershed is one table is nearly 45 to 100 years, with anthropogenic effects
of the main water resources for the urban and rural areas being the primary controlling factor, accounting for nearly
of Koladiba and the city of Gondar. Limited studies have 99% of the variance in comparison to rainfall. Zhang et al.
been conducted on the factors that shape the resilience and (2022) investigated the use of soil moisture observations
vulnerability of rural households in the face of recurring from the soil moisture active passive (SMAP) satellite mis-
floods (Weldegebriel and Amphune 2017), and we did not sion to inform simulated carbon fluxes in the international
find any study related to current and future hydrological terrestrial ecosystem model LPJ-wsl includes land surface
responses using baseline and climate models. As Mounir processes.
et al. (2023) conducted by using statistical analysis and the Hydrological modeling is a valuable tool for estimat-
soil and water assessment tool (SWAT) hydrological model ing hydrological responses for current and future scenarios
to predict drier and hotter climatic conditions in the region by optimising hydrological parameters. By simulating the
of northern Morocco (Mounir et al. 2023). The HEC-HMS hydrological processes of a watershed through a hydrologi-
model was adopted to model the rainfall-runoff relation of cal model, it is possible to predict the potential effects of
the Upper Sabarmati River basin (Verma et al. 2022). There- climate change, changes in land use, and water management
fore, hydrological modeling and remote sensing tools are strategies on the water balance of the watershed (Dau et al.
critically essential to understanding hydrological processes, 2021). It is common to use hydrological models, such as
as well as supporting long-term planning of water use and the XAJ model, to simulate the watershed’s hydrological
irrigation application (Jakada et al. 2019). Hydrological response under various climate change scenarios. The XAJ
modeling can be divided into three types: deterministic model is typically applied with precipitation and potential
physically distributed models (white box models), concep- evapotranspiration as primary input data. This model was
tual hydrological models (grey box models), and data-driven selected based on the following main reasons: (1) The model
or empirical models (black box models) (Abdulkareem et al. has been successfully and widely applied in similar humid
2018; Jajarmizadeh et al. 2012). and semi-humid regions of China and other areas such as
Various hydrological responses of the hydrological pro- the USA (Rahman and Lu 2015) (2) Test Differential Evo-
cess can be incorporated, analyzed, and simulated in deter- lution Optimization (DEoptim) algorithm for parameter
ministic physically distributed models such as MIKE SHE calibration. The discretization provides critical informa-
and SWAT (Feyen et al. 2000),. Stanford Watershed Model tion about runoff, SM, and ET at each soil layer, so this is
(SWM-IV), Système Hydrologique Européen (SHE), TOP- important to understand the physical meaning of the hydro-
MODEL, Precipitation Runoff Modelling System (PRMS), logical responses in the three soil layers and the correlation
and the National Weather Service River Forecasting System between remotely detected SM. The use of a multimodel

13
Modeling Earth Systems and Environment

ensemble, in this case from the CMIP6 dataset, can help to between 12°25̍ ’ 06″–12°39′36″ N latitude and from
capture the overall range of uncertainty associated with the 37°13’06″–37°30′36″ E longitudes at the outlet point
climate projections. The SSP scenarios (Shared Socioeco- (bridge) where the discharge observation station is located
nomic Pathways) are designed to represent different possible and is shown in Fig. 1. It has a tropical savanna climate, with
future pathways for global development, including changes an average annual minimum and maximum temperature of
in population, economic development, and technology. The 16.7 °C and 27.2 °C, respectively, and a long-year average
SSP2-4.5 represents a scenario where greenhouse gas emis- annual precipitation of 1,254.50 mm shown in (Fig. A). The
sions peak around 2040 and then decline with relatively low annual flow of the river is 53.7 m 3/s and evapotranspira-
greenhouse gas emissions, leading to a radiative forcing level tion is 1280 mm. The rainfall distribution is quite uneven
of 4.5 W/m² by 2100. SSP3-7.0 represents a scenario where throughout the year in that precipitation in summer accounts
the medium level of greenhouse gas emissions continues to for more than 74%. The Dirima River is the main tributary
increase throughout the 21st century, leading to a radiative of Lake Tana, with the longest main flow path of 48.53 km.
forcing level of 7.0 W/m² by 2100. And SSP5-8.5 represents The slope class coverages are moderate and flat to gentle,
a scenario where greenhouse gas emissions increase even as shown in Fig. 1a. The soil types found in the Dirima
more rapidly than in SSP3, leading to a radiative forcing watershed include Eutric fluvisols (44.2%), Eutric lepto-
level of 8.5 W/m² by 2100 with a high level of emissions. sols (28.6%), chromic luvisols (15.6%), and Eutric vertisols
Simulating the hydrological response of a watershed under (11.6%) as shown in (Fig. 1c). Along the flowing direction
these different scenarios can help to provide insight into the of the river from north to south, the Dirima River water-
potential impacts of climate change on water resources in shed contains many agricultural areas on both sides of the
the region. This information can be used to inform water river, including fragmented irrigated lands. Regarding land
resource management decisions, such as the development use and land cover (LULC), the Dirima watershed includes
and implementation of adaptation strategies to prepare for several categories that are predominantly covered by crop-
various climate change scenarios. land (94.7%), shrubland (1.8%), and herbaceous vegetation
The Xinanjiang model is a popular choice for hydrologi- (1.5%), sources: https://​lcvie​wer.​vito.​be/​2019. In the water-
cal modeling, and it has been used to estimate hydrological shed, there is an estimated irrigated area of nearly 178 ha on
responses in many different regions around the world. How- both the right and left sides of the Dirima river.
ever, it is important to note that hydrological models need
to be calibrated and validated to ensure that the parameter Xinanjiang (XAJ) conceptual hydrological model
values are optimized for the specific watershed being stud-
ied. This is because the optimal parameter values can vary The XAJ model is a conceptual hydrological model that
depending on the characteristics of the watershed, such as predicts discharge at a basin outlet by simulating runoff,
topography, vegetation cover, soil type, and climate. Opti- generation, and concentration within a catchment (Hao et al.
mizing the parameters for the Dirima watershed in Ethiopia 2015; Ren-Jun 1992; Zhijia et al. 2013). Its major character-
using reliable algorithms can help identify optimal param- istic is the formation of runoff during storage replenishment
eter values that are specific to the watershed, thus improving (saturation excess overland flow), which implies that runoff
our understanding of the hydrological processes that occur does not occur until the SM content of the unsaturated zone
in the watershed. This, in turn, can help to improve water exceeds the field capacity, after which runoff equals the rain-
resource planning and management in the region. Overall, fall surplus without a further loss (Hao et al. 2015; Weimin
the use of hydrological models and climate change scenarios and Qian 2012; Zhijia et al. 2013; Zhu et al. 2017). The key
can provide valuable insights into the potential impacts of concept in the model is the distribution of storage capacities
climate change on water resources in a region, helping to in the soil layer. Areal average precipitation and potential
facilitate the development of more sustainable and resilient evapotranspiration serve as input to the XAJ model. Param-
water resource management strategies. eters within a group are mutually dependent, whereas param-
eters between groups appear to be mainly independent. Vari-
ation in runoff concentration parameter values, for example,
Materials and methods within a reasonable range, does not affect the optimum value
of the runoff separation parameter, and variation in runoff
Study area separation parameter values does not affect the optimized
value of the runoff production and evapotranspiration param-
The studied Dirima river watershed is a sub-water- eters (Ren-Jun 1992). Table (1) provides a list of parameters
shed located in the upper Lake Tana basin, cover- of the XAJ model and their physical implications. The model
ing an area of 162.2 ­k m 2 and an altitudinal range of assumes that potential evapotranspiration does not alter
1786–2746 masl. Geographically, the watershed is located between a “dry year” and a “wet year.“ The output is the

13
Modeling Earth Systems and Environment

Fig. 1  Dirima watershed
hydrological location with
respect to Tana Basin, a slope,
b LULC BUA: built-up area,
HW: Herbaceous wetland, Fr:
Forest, Hrbve: Herbaceous
vegetation, Shr: Shrub land, and
Cl: Cultivated land) and c soil
type of Eutric fluvisols, Eutric
leptosols, chromic luvisols, and
Eutric vertisols

runoff/discharge (Q) from the watershed outlet. and the unit four main components for each sub-catchment: (i) Evapo-
hydrograph (UH) directs the surface runoff. The linear res- transpiration, which generates the deficit of soil storage that
ervoir approach is used to route interflow and groundwater is divided into three layers, that is, upper, lower, and deep;
flow. The entire XAJ model structure is displayed in (Fig. 2). (ii) Runoff generation, which produces the runoff according
More details of the original structure of XAJ model can be to the rainfall and soil storage deficit; (iii) Runoff separa-
found in (Ren-Jun 1992). The XAJ model has 15 parameters tion, which divides the above-mentioned runoff into three
(Rahman and Lu 2015) and the outflow is estimated with components, that is, surface, subsurface, and groundwater;

13
Modeling Earth Systems and Environment

Fig. 2  Flowchart of the Xinanjiang (XAJ) model for a single watershed, Deng (2020)

Table 1  Parameters of Xinanjiang (XAJ) model and their physical meaning


Processes Physical meaning and units Lower bounds Upper bounds

Evapotranspiration KC The ratio of potential evap to pan evap (−) 0.2  1.5
WUM SM capacity of the upper layer (mm)  5  20
WLM SM capacity of the lower layer (mm) 10  90
WDM SM capacity of deep layer (mm) 10  60
C Coefficient of deep evap (−)  0  0.05
Runoff generation B The exponent of the SM storage capacity curve (−)  0.05  0.2
IM Fraction of impermeable area (%)  0.1  0.6
SM Areal mean free water capacity of the surface soil layer (mm)  10  60
Runoff separation EX The exponent of the free water capacity curve (−)  0.5  2
KI Outflow coefficient of the free water storage to interflow (−)  0.01  0.7
KG Outflow coefficient of the free water storage to GW (−)  0.01  0.7
Runoff routing /concentration/ CI Recession constant of the lower interflow storage (−)  0.5  0.9
CG Recession constant of groundwater storage (−)  0.99  0.998
N Number of reservoirs in the instantaneous unit hydrograph (IUH)  0.1  5
NK Common storage coefficient in the IUH  1  6

and (iv) Flow routing, which transfers the local runoff to the collectively transferred to QS, QI, and QG and constitute
outlet of each subbasin forming the outflow of the sub-basin the total flow to the channel network of the sub-basin. The
which shown in Fig. 2. discharge (Q) from the sub-basin is calculated using the
The state variables in the hydrological model include Muskingum method. As depicted in Fig 2, runoff (R) from
the tension water storage (W) and free water storage (S) the pervious area has three components: RS, RI, and RG,
areas. The mean tension water (W) is composed of three representing surface runoff, interflow, and groundwater
elements (WU, WL and WD) in the upper, lower, and deep runoff, respectively, while RB represents runoff from the
layers, respectively. The runoff-contributing region factor impervious area (IM). The discharge (Q) from the sub-basin
(FR) is associated with tension water storage (W), while is composed of three components: QS, QL, and QG, repre-
RB represents the immediate runoff from the impervious senting surface runoff, interflow, and groundwater runoff,
area. The pervious surface is divided into three components respectively,Table1 while FR denotes the variable runoff
(RS, RI and RG) representing surface runoff, subsurface producing area and T represents the total subbasin inflow
runoff, and groundwater runoff, respectively, which are to the stream.

13
Modeling Earth Systems and Environment

Table 2  CMIP6 climate model No. CMIP6 model name Country Nominal Res (km) Variant label References
used to evaluate hydrological
response in the watershed 1 BCC-CSM2-MR China 100 km r1i1p1f1 (Xin et al. 2019)
2 MRI-ESM2-0 Japan 101 km r1i1p1f1 (Yukimoto et al. 2019)

Model input data preparation upward or a downward trend (Tigabu et al. 2020). The
daily time series data was used to detect if the data has a
This study utilized daily precipitation and potential evap- trend or not with a 95% significance level.
oration data sets from 1994 to 2009 as input data. The Bias correction is common for simulations of regional
hydrometeorological data was collected from the National climate models (RCMs) that exhibit systematic biases in pre-
Meteorology Agency (NAM) of Ethiopia and the Ministry cipitation and temperature. To produce reliable estimators
of Water, Irrigation, and Electricity of Ethiopia (MoWIE). of local-scale climate, RCMs need to be post-processed. For
Some missing values (i.e. 6.3%) in the data series were bias correction of precipitation linear scaling (multiplica-
filled using the simple averaging method. Although the tive) methods were applied. This approach involves apply-
missing data was less than 10%, it was filled in for data ing scaling factors to the simulated precipitation data. The
clarity. The 16-year meteorological data was used because scaling factor is derived from the ratio of observed precipita-
streamflow data was only available for this period. The tion to simulation over a specific reference period. By mul-
warm-up of the model used two-year data (1994–1995), tiplying the simulation precipitation values by this scaling
calibration used 10-year data (1996–2005), and validation factor, systematic biases can be reduced and model output
used four-year data (2006–2009). The Penman-Monteith accuracy improved. Regarding temperature, linear scaling
approach for potential evapotranspiration (Allen et  al. (adaptive) methods were used. Instead of applying a single
1998) used meteorological time series data such as maxi- scaling factor throughout the temperature range, different
mum temperature, minimum temperature, solar radia- scaling factors apply to different temperature ranges. These
tion, and actual vapour pressure as input. The landscape methods were applied using climate data from 1990–2014
characteristics of a catchment were derived from a DEM as the baseline period. Before evaluating precipitation and
with a spatial resolution of 30 m by 30 m SRTM (https://​ temperature for the selected CMIP6 models, biases were cor-
search.​earth​data.​nasa.​gov/​search), and the morphometric rected over the watershed using the CMhyd tools (Rathjens
watershed attributes such as slope and watershed area were et al. 2016), Teutschbein, (2012) provides detailed math-
extracted using QGIS digital terrain analysis tools. ematical equations that are presented about all methods.
For the future analysis of the climate projection of the
watershed, data were extracted from the World Climate Model setup, calibration and validation
Research Programme (WCRP)–CMIP6 under a low forc-
ing scenario (SSP2-4.5), medium to high forcing scenario In the XAJ model, sensitive parameters were identified
(SSP3-7.0), and a strong forcing scenario (SSP5-8.5) (Xin through four distinctive processes. Daily observations data
et  al. 2019), source: https://​e sgf-​n ode.​l lnl.​g ov/​s earch/​ from 1996–2009, as shown in Fig. A were prepared using the
cmip6/. A detailed description of the scenarios is avail- hydroTSM and hydroGOF packages. The XAJ model is a con-
able in [22,24] and the CMIP6 products are summarised ceptual hydrological model that is used to simulate the main
in Table  2 which gives acronyms, resolution, sources, hydrological processes in the Dirima watershed, such as evap-
references, and the mean annual climate data distribution oration, runoff, and soil moisture from three soil layers. The
is shown in Figure B. For validation, available ground ideal initial parameter ranges were collected from literature
observation climatological data within the basin and for sources, specifically previous applications of the XAJ model
the baseline period (1990–2014) were collected from the (Bai et al. 2017; Ren-Jun 1992). Calibration and validation of
archive of the National Meteorology Agency (NMA) of model parameters were conducted using a differential evolu-
Ethiopia. The Mann Kendall and seasonal MK tau tests are tionary algorithm (DEoptim) which is a population-based, sto-
non-parametric statistical tests used to identify trends in chastic search algorithm that aims to find the global optimum
time series data. Under the following hypothesis, the tests of a given objective function (Mullen et al. 2011). DEoptim
were performed in the baseline period of the precipitation, is particularly suitable for identifying the global optimum of
evapotranspiration, and streamflow time series. Ho: the a real-valued function with real-valued parameters, as it does
time series data show neither an upward nor a downward not require the function to be continuous or differentiable. The
trend, whereas Ha: the time series data exhibit either an objective function of optimization is to minimize the Root

13
Modeling Earth Systems and Environment

Table 3  The initial and calibrated value of the XAJ parameters


Parameters KC IM WUM WLM WDM C B SM EX KI KG CI CG N NK

Initial Value 0.38 0.22 18.93 88.46 59.01 0.01 0.19 58.34 0.5 0.7 0.7 0.64 0.96 0.1 1.65
Calibrated Value 1.5 0.1 5.3 24.6 13.4 0.1 0.2 60 2 0.4 0.5 0.9 1 0.1 1
hydroGOF R2 NSE PBIS% RMSE MSE KGE VE
d m d m d m d m d d d

Calibration 0.78 0.96 0.76 0.95 10.5 1.6 12 3.03 145 0.8 0.4
Validation 0.85 1 0.85 1 9.9 0 3.65 0 13 0.9 0.6

d daily, m monthly

Mean Square Error (RMSE) between observed and simulated �∑


n
�2
streamflow. Equation (1) shows the method used to calculate (Qo − Qo)(Qs − Qs)
i
RMSE, with a lower value indicating higher accuracy of the
R2 = �2 � �2 (2)
∑n �
model results. The calibration algorithm that optimizes the i
Qo − Qo Qs − Qs
error criterion was selected as an objective function using the
provided functions. All parameters were changed within their ∑n
(Qo − Qs)2
‘reasonable’ range (lower and upper), based on literature and/ NSE = 1 − i
�2 (3)
or experience of model implementation. The simulation was
∑n �
i
Qo − Qo
conducted with a maximum of 2000 iterations.
√ ∑n
√ n (Qo − Qs)
√1 ∑ i
RMSE = √ (Qo − Qs)2 (1) pbias = 100 ∗ ∑n (4)
n t=1 i
Qo

Where ­Qo is the observed discharge shown in Fig. A ­Qs


Performance measures is the model simulated discharge. Qo is the mean measured
discharge for the entire period, Qs and is the mean simulated
Evaluation indexes are available to statistically measure the discharge for the entire period. NSE determines the rela-
performance of the hydrological model and assess the hydro- tive magnitude of the residual variance in comparison with
graph pattern. Moriasi et al. (2007) conducted comprehensive the observed data variance and is calculated using Eq. (3).
model performance evaluation guidelines, such as statistical NSE values with higher magnitudes are preferred (Car-
goodness of fit of the model, which was assessed by the coeffi- los Mendoza et al. 2021). In general, when NSE > 0.5 and
cient of determination ­(R2), RMSE, and relative squared error PBIAS < ± 25%, the performance of the model is considered
(RSE) for whose errors are measured in the different units. to be satisfactory (Amin et al. 2017; Jiang et al. 2020).
A preliminary evaluation was performed using the following
criteria: Percent Bias (PBIAS) and Nash-Sutcliffe Efficiency
(NSE) (Nash and Sutcliffe 1970; Pushpalatha et al. 2011). Results
The PBIAS formula is used to measure the variance between
observed and simulated total runoff (Pushpalatha et al. 2012). XAJ model simulation
NSE is a technique to assess mostly high-flow simulations.
PBIAS = 0 and NSE = 1 indicates that the observed and simu- The Mann–Kendall trend test was conducted on daily time
lated values are perfectly fit. The measure of the observed flow series data using the trend r-package. The resulting P value
variance explained by the simulated flow is determined using (> 0.05) indicates that there is no significant monotonic
­R2 and the values vary from 0 to 1, with values greater than trend, thereby accepting the null hypothesis. The values of
0.5 representing good performance. The three criteria may be the model parameters have been calibrated (1996–2005) and
formulated as described below: validated (2006–2009) to capture the hydrological process
of the Dirima basin. The variation between simulated and
measured streamflow is evaluated using model efficiency

13
Modeling Earth Systems and Environment

Fig. 3  The simulated and measured streamflow hydrograph (a)  and scatter plot during the calibration (1996–2005) (b)  and validation period
(2006–2009) (c) 

criteria (Li et al. 2015). The optimized parameters of the the basin does not directly contribute to the flow of water,
XAJ model are shown in Table 3 and the comparison of but is lost through various processes such as evaporation,
observed and simulated hydrograph is shown in Fig. 3. As infiltration, and percolation. Consequently, the storage
this figure shows, the observed and simulated streamflow capacity of the basin may decrease during low flow peri-
resulting from the XAJ model revealed good fitness in terms ods, indicating that there is less water in the system. Fur-
of capturing the hydrograph pattern, as well as high and thermore, this statement implies that the contribution of
low flows. Figure 5 below also shows that while winter is the base current to the measurement site (where the stream
normally dry in Dirima, March has been the month when flow is recorded) is relatively lower than the output of the
the lowest mean streamflow has been recorded. model of subsurface evaporation. In the validation period,
The areal mean tension water is 5.3, 24.6, and 13.4 in the results are also in very good agreement with the observed
upper, middle, and lower layers of the soil, respectively. The streamflow shown in Fig. 3, in the dry period the model
average runoff produced from the contribution of the imper- gives delayed underground runoff because of the SM mem-
vious area is 0.2, and the free water storage to the interflow ory of nearly 12 days. The flow furation curve (FDC) is
and groundwater is 0.9 (0.4 + 0.5) with a proportion of 0.8 shown in Fig. A (a). These figures provide a good fit at
(0.4/0.5). The SM parameter indicates that there is thick soil maximum flow, and a large gap between the simulated
in the watershed and surface runoff is sensitive to the value and the observed flow at high flow compared to medium
of this parameter. to low flow. This is due to the model overpredicting the
The streamflow hydrograph and scatter plot shows good minimum flow at which the model does not consider land
agreement in terms of correlation coefficient and pattern use/land cover and anthropogenic effects such as diversion
in both calibration and validation period. This suggests for irrigation in the upstream watershed. The FDC has
that the calibrated Xinanjiang (XAJ) model is capable good agreement at high and medium flow as compared to
of accurately simulating the streamflow dynamics in the the lower flow as shown in Fig. A (b).
watershed. Also, the watershed experiences losses from The mean daily evapotranspiration was higher than the
the incoming precipitation through surface and subsurface SM content in the upper layer, while the evaporation in the
runoff. This means that part of the precipitation falling on lower soil layer is lower as compared to the corresponding

13
Modeling Earth Systems and Environment

Fig. 4  Upper (a) and lower (b) soil layer seasonal hydrological components in the calibration period, and (c) upper and (d) (lower) in the valida-
tion period

Fig. 5  Evaporation and SM distribution in the upper soil layer and lower soil layer in (a and c) calibration and (b and d) validation period

13
Modeling Earth Systems and Environment

SM content shown in Fig. 4. This is due to radiation, and a higher NSE value in validation indicates the model’s abil-
soil heat flux might be decreased from the upper to the ity to capture the system’s response to changing conditions,
deep soil layer. As shown in Fig. 4 and Table 1 revealed making it useful for future projections and scenario analysis
that the lower SM is higher compared to the deep and (Wagener and Gupta 2005). This shows that the parameters
upper layers, and the evapotranspiration and runoff in the effectively mimic the Dirima hydrological components
upper layer are higher compared to the lower and the deep of the XAJ model shown in Table 3. Overall, the calibra-
layers, respectively. Therefore, the Dirima watershed stor- tion and validation results of the XAJ model applied to the
age capacity was reduced due to higher evapotranspiration Dirima watershed were good and thus, the corresponding
and runoff in the upper soil layer. The variability in evapo- parameters and simulation results are considered suitable for
transpiration, soil moisture, and runoff follows the patterns this study. As shown in Table 3, RMSE and MSE are meas-
of rainfall. In Ethiopia, in the seasonal1 rain cycle, most urements of the common magnitude of differences between
precipitation occurs in June (10%), July (27%), and August observed and simulated values, and lower values indicate
(37%) followed by September (8%), October (9%), and greater performance. In this case, the RMSE values of 12
November (0.9%) and March (0.9%) April (1.8%) and May during the calibration period and 3.65 during the valida-
(4.3%). During the autumn and spring seasons, additional tion period suggest that the common errors of the model are
irrigation may be optional to use the available SM in the somewhat higher during the calibration period than during
watershed as shown in Fig. 5. the validation period. Similarly, MSE values of 145 in the
The calibration results were assessed using the statistical calibration period and 13 in the validation period show a
and model efficiency criteria such as R­ 2, NSE, PBIAS, and higher total error in the calibration period compared to the
RMSE. The findings show that the XAJ model can simulate validation period. KGE is a metric that evaluates the simi-
both the magnitude and dynamics of observed flow event larity of observed and simulated data by means, variances,
occurrences, as well as distinct hydrological components to and correlation. A value of 1 indicates the ideal matching
describe the watershed hydrological responses. As a result, between the observed and simulated data. In this case, KGE
the XAJ model optimized parameters yield the NSE values values of 0.8 in calibration and 0.9 in validation suggest
of 0.76 in the calibration period and 0.85 in the validation a good agreement between the model and observed data
period indicating a reasonably good agreement between the in each period. VE represents the proportion of observed
simulated and observed flow data. If so, the PBIAS values variations captured by the model. Higher VE values indicate
of 10.5% during the calibration period and 9.9% during the that the model has a better capacity to explain the observed
validation period imply that the simulated flow values of variance. The VE values of 0.4 during the calibration period
the XAJ model are often within a range that is acceptable and 0.6 during the validation period indicate that the model
compared to the observed values. Obtaining a higher Nash- captures a significantly lower percentage of the observed
Sutcliffe Efficiency (NSE) value in validation compared to variance during the calibration period than during the vali-
calibration can have several implications. Firstly, using inde- dation period.
pendent data for validation helps assess the model’s ability
to make accurate predictions on unseen data, thus increas- Watershed hydrological responses
ing confidence in its generalization capabilities. Secondly, a
higher NSE value in validation suggests that the model has Evapotranspiration, runoff generation, and SM distribution
avoided overfitting and can provide reliable predictions on
new observations. Thirdly, it demonstrates that the model The dynamics of evapotranspiration and SM is the most
structure developed during calibration captures the dominant important component that implicitly takes care of the non-
hydrological processes, enabling it to represent the system’s linear dynamics of the variation of SM in space and hence,
behavior under different conditions. Additionally, a higher determines the spatial variation of the runoff generation
NSE value in validation indicates the robustness of the cali- mechanism. The evapotranspiration that generates the deficit
brated parameter values, which can be applied to different in soil storage is divided into three layers: upper (EU), lower
periods or regions with similar characteristics. Moreover, the (EL), and deep (ED).
representativeness of model inputs during calibration, such In the upper soil layer, the evaporation rate is higher in
as precipitation and streamflow, leads to reliable predictions all months because of direct radiation interaction to satisfy
during validation. Improved model performance in valida- the atmosphere’s vapour demand. Figure 5 shows that SM in
tion also reduces uncertainty and enhances confidence in the rainy season did not decline compared to the dry /source-
decision-making and water resources management. Finally, limited/ seasons. Figure 5a, b suggests that the Dirima water-
shed’s soil has a greater potential for water storage. The
groundwater potential is greater in watersheds with greater
1
  JJA: Summer, SON: Autumn, DJF: Winter, MAM: Spring. water storage capacity.

13
Modeling Earth Systems and Environment

Table 4  Autocorrelation (a) and Soil layers Lag. interval (a)


cross-correlation (cc) (b) with
lag values 0 1 2 3 4 5…

Upper (WU) 1.00 0.28 − 0.02 0.32 0.38 − 0.03


Lower (WL) 1.00 0.93 0.85 0.80 0.75 0.64
Deep (WD) 1.00 0.79 0.55 0.30 0.06 − 0.18
WU vs. WL 0.59 0.37 0.31 0.39 0.34 0.19
WU Vs WD 0.02 − 0.15 0.04 0.26 0.29 0.12
(b)
CC between PCP and WU 0.81 0.29 − 0.07 0.20 0.37 0.00
CC between PCP and WL 0.38 0.41 0.29 0.31 0.38 0.35
CC between PCP and WD − 0.09 − 0.05 − 0.09 0.02 0.26 0.29
CC between PCP and RS 0.94 0.22 0.27 0.21 0.22 0.30

The generation of runoff and the moisture content of the (WU vs. WL and WD) at lag k = 0 is 0.59 and k = 7 is 0.39,
soil are both important factors to consider in watershed char- respectively shown in Table 4. The correlation coefficient
acterization. The streamflow in this model includes surface at k = 0 lag between upper layer SM and precipitation is
runoff2 (QS), interflow runoff (QI), and groundwater runoff 0.89 and gradually weakened as the lag interval increases,
(QG). Surface runoff is the dominant hydrological compo- as shown in Fig. A This indicates that the cascading effects
nent of the watershed in the rainy season compared to SM of precipitation are more dominant until 34 days in the upper
in the lower and upper layers of the soil shown in Fig. (4c soil layer, which reaches a maximum value as precipitation
and d). Figure (5c and d) shows the SM3 and runoff in the occurs compared to the deep soil layer. Due to the lower soil
lower soil layer, the runoff starts earlier than the SM, this is temperature and plant cover, the impact of SM memory is
because of the effect of lower layer SM contribution. Also, greater in the deep layer than in the lower layer. Moreover,
the runoff and SM relate positively to precipitation and dur- this controls the surface runoff generation at the watershed
ing the rainy season, the SM is higher than the runoff depth. with maximum and positive cross-correlation values shown
Similarly, in a deep soil layer, SM storage is as good as in in (Fig. C (b). Additionally, the maximum autocorrelation
the lower soil layer, as groundwater flow to the surface of the value of soil moisture in each soil layer at k = 0 lag and
lateral flow to the stream shown in Fig. C (c and d). significant at k = 14 is due to the influence of precipitation
shown in Fig. D (a). In the case of the Dirima watershed,
Lag correlation analysis if there is a higher available SM in the lower and deep soil
layers due to higher evaporation in the upper soil layer, this
Pearson’s correlation coefficient of lag k = 0 (on that rainy could have implications for plant growth. Shallow-rooted
day) for daily precipitation and SM data sets during the crop plants, as the name suggests, have roots that predominantly
growing period (June to December) is evaluated. Autocor- grow in the upper soil layers, where soil moisture availability
relation is defined as the correlation between a variable and is lower due to higher evaporation. Therefore, if the available
its lag time over a set of periods (Box et al. 2008). SM is SM is limited in the upper soil layers, shallow-rooted plants
the main driver of runoff and evapotranspiration under spe- may be more susceptible to water stress and have limited
cific conditions in the watershed system. SM has a stronger growth compared to deep-rooted plants.
autocorrelation in different soil layers, and when the time
lag interval increases, the autocorrelation of SM in each soil Future hydroclimate responses
layer decreases significantly (Zhao and Yao 2020). The soil
does not hold water for long periods due to factors such as Climate change can have a significant impact on hydro-
evapotranspiration, which can cause water to be lost from logical responses, and future climate models suggest that
the soil profile relatively quickly, and there may be spa- these responses may vary depending on location and global
tial variability in soil moisture across different soil layers, temperature increase. One of the potential effects of climate
which can affect the autocorrelation of SM over time. The change on hydrological responses is an increase in evapo-
maximum cross-correlation value between each soil layer transpiration rates, which can lead to drier conditions in
some regions. This could be particularly problematic in areas
that are already experiencing water scarcity. Additionally,
2
  QS: Surface runoff, QI: interflow, QG: Groundwater flow. changes in precipitation patterns and intensity may lead to
3
  WU: upper layer, WL: lower layer, and WD: deep layer. more frequent and severe flooding events in some regions. In

13
Modeling Earth Systems and Environment

Table 5  The mean annual hydroclimate responses in the immediate, Precipitation


intermediate, and distant future under the selected scenarios
Scenarios Variables Year Range In this region, precipitation decreases by 11, 2, and 7% by
2045, 2076, and 2100 respectively, under the SSP2-4.5 sce-
2015–2045 2046–2076 2077–2100
nario, 13%, 6, and 4% under SSP3-7.0 it is likely that we
SSP2-4.5 Pcp (mm) ▼ −167.43 ▼ 25.35 ▼ −111.48 will see some significant impacts on the environment and
−11% −2% −7% human society. It may lead to reduced water availability,
Tmax (°C) ▲ 0.75 ▲ 1.78 ▲ 2.24 especially in this watershed that is already water-stressed.
+ 3% + 7% + 9% This could impact agriculture, food security, and ecosystems
Tmin(°C) ▲ 1.10 2.02 2.49 that depend on water. Furthermore, reduced precipitation
+ 7% + 14% + 17% could lead to an increase in droughts, which could further
ET (mm) ▲ 4.24 4.26 4.31 exacerbate these impacts. Also, Under the SSP5-8.5 sce-
+ 18% + 19% + 21% nario, the changes in precipitation are plausible. A decrease
Q ▼ −3.09 −2.41 ▲ 1.24 of 9% by 2046, followed by a slow increase of 9% by 2076,
−55% −43% 22% and a larger increase of 16% by 2100 would indicate a shift
SSP3-7.0 Pcp (mm) ▼ −197.72 −94.47 −61.26 in the timing and distribution of precipitation throughout the
−13% −6% −4% century shown in Table 5, Fig. 6. Changes in precipitation
Tmax (°C) ▲ 0.76 1.97 3.07 under each SSP scenario can have a significant impact on
+ 3% + 8% + 12% future water availability.
Tmin (°C) ▲ 1.13 2.41 3.45
+ 8% + 16% + 24% Evapotranspiration
ET (mm) ▲ 4.26 4.25 4.40
+ 19% + 19% + 23% The maximum temperature (Tmax) increases of 0.75 °C by
Q ▼ −3.15 −4.54 −0.43 2046, 1.78 °C by 2076, and 2.24 °C by 2100 would result
−56 −81% −8% in more frequent and intense heatwaves and droughts as
SSP5-8.5 Pcp (mm) ▼ −134.60 ▲ 54.04 38.60 shown in Fig. 7. This could lead to decreased crop yields,
−9% + 3% + 9% increased risk of vulnerability to water resources, and poten-
Tmax (°C) ▲ 0.92 2.38 4.16 tially more frequent and severe natural disasters. Addition-
+ 4% + 9% + 16% ally, an increase in minimum temperature (Tmin) by 1.1 °C
Tmin (°C) ▼ 1.27 2.81 4.46 by 2046, 2.02 °C by 2076, and 2.49 °C by 2100 would lead
+ 9% + 19% + 30% to warmer nighttime temperatures. This could exacerbate
ET (mm) ▲ 4.31 4.30 4.50 the effects of heat waves and cause discomfort and health
+ 20% + 20% + 26% issues for vulnerable populations, such as the elderly and
Q ▼ −2.97 ▲ 0.04 −1.78 those with pre-existing health conditions. Furthermore, this
−53% 1 −32% would result in increases in evapotranspiration of 18% by
2046, 19% by 2076, and 21% by 2100 as shown in Table 5.
It could lead to changes in precipitation patterns, with some
contrast, other regions may experience prolonged droughts regions experiencing more frequent and intense rainfall
due to decreased precipitation. In general, future climate while others experience more frequent droughts. This could
models suggest that hydrological responses are likely to be have significant impacts on agriculture, water availability,
complex and highly variable, and will depend on a range of and ecosystem services.
different factors, including location, temperature increases, Overall, the changes in temperature and evapotranspira-
and changes in precipitation patterns. tion could have significant impacts on the Earth’s climate
To compare future hydroclimate responses under the and ecosystems. Policymakers and individuals need to take
three SSP scenarios, we simulate the hydrological response action to reduce greenhouse gas emissions and mitigate the
of the watershed under each scenario using the calibrated impacts of climate change. Additionally, changes in evapo-
parameter values. This would involve running the hydro- transpiration rates can affect the availability of water for
logical model with inputs corresponding to each of the SSP other purposes, such as irrigation or domestic use.
scenarios, and then comparing the model output for each
scenario particularly, considering the following factors that Streamflow
can impact future water availability:
The future streamflow will be dramatically decreased by
55% by 2046, 43% by 2076, and 22% by 2100 under the

13
Modeling Earth Systems and Environment

Fig. 6  Precipitation availability under three climate change scenarios ods: immediate future (2015–2045)  (a), intermediate future (2041–
(SSP2-4.5 in light red, SSP3-7.0 in green, and SSP5-8.5 in light blue) 2070) (b), and distant future (2071–2100) (c) 
over the baseline (1990–2014 in deep blue) and three evaluation peri-

Fig. 7  Evapotranspiration rate under three climate change scenar- ods: immediate future (2015–2045) (a), intermediate future (2041–
ios (SSP2-4.5 in light red, SSP3-7.0 in green, and SSP5-8.5 in light 2070) (b), and distant future (2071–2100) (c) 
blue) over the baseline (1996–2009 in red) and three evaluation peri-

SSP2-4.5 scenario, 56% by 2026, 81% by 2076, and 8% by the immediate and distant evaluation period. These signifi-
2100, under the SSP3-7.0 scenario and 53% by 2046, 32% by cant changes in streamflow can affect water availability in
2100 under SSP5-8.5scenario. But a 1% increment is shown both surface and groundwater sources during these evalua-
by 2076 under the SSP5-8.5 scenario because maximum and tion periods. Changes in precipitation and streamflow can
minimum temperature and evapotranspiration are lower than impact groundwater recharge rates, which can, in turn, affect

13
Modeling Earth Systems and Environment

future water availability. In these watersheds, groundwater is represent different possible future scenarios with varying
an important source of water for water supply and other uses. socioeconomic and environmental conditions, the rate of
Therefore, future water uses culture and strategy should be evaporation in the upper soil layer is higher than the amount
adjusted accordingly, such as reducing streamflow by using of moisture present in that same layer. This suggests that
small-scale water harvesting, plantation, and physical struc- the relationship between the rate of evaporation and the
tures to enhance aquifer recharge. It is concerning to hear moisture content of the upper soil layer remains consistent
of the significant decrease in future flow predicted under in all scenarios, regardless of specific socioeconomic and
different scenarios. environmental conditions. In the lower and deep soil layers,
These changes in streamflow can have far-reaching the soil moisture is higher than evaporation. This finding is
impacts on water availability, affecting both surface and important because it indicates that there is a fundamental
groundwater sources. The decrease in streamflow can lead physical relationship between evaporation and the moisture
to reduced recharge rates, making it difficult to maintain content in the soil that is likely to remain unchanged in the
adequate water supply for various uses, including domestic, future, regardless of other factors that can influence the over-
agricultural, and industrial. all conditions. This information could be useful in develop-
As shown in Fig. 9 surface runoff is higher than interflow ing strategies to manage water resources in the affected areas
and groundwater flow in all ssp. scenarios, which is criti- to ensure sustainable use of available water.
cal water availability in the aquifer. This implies in all sce-
narios (SSP), the amount of water that runs off the surface
of the ground is higher than the amount of water that flows Discussions
through the soil (interflow) and the amount of water that
moves through underground aquifers (groundwater flow). The most important components of the hydrological
This is an important issue because the amount of water responses of the watershed are evapotranspiration, soil
available in the aquifer is critical and if surface runoff is moisture, and runoff, which are interconnected. In the
too high, it can reduce the amount of water that reaches the Dirima watershed, the predicted daily flow closely matches
aquifer, which can have negative impacts on water availabil- the observed daily flow. A model that systematically over
ity for human consumption and other purposes. Therefore, or under-predicts all the time will still result in “good” ­R2
this situation should be carefully monitored and managed to (close to 1), even if all predictions were not the best fit
ensure sustainable water use in affected areas. It is crucial to (Krause et al. 2005). These components are interconnected
consider these changes in water availability when develop- and play crucial roles in the water balance and overall
ing water management strategies for the future. Implement- hydrological processes of the system. The XAJ model has
ing measures to reduce runoff, such as small-scale water a variable SM storage capacity in each soil layer (WU,
harvesting, plantation, and physical structures to improve WL and WD), and it results in different evapotranspiration
aquifer recharge, can help mitigate the effects of reduced rates. This three-layer SM model was developed to more
streamflow. In addition, promoting water conservation and realistically account for SM distribution and redistribu-
efficient use can also help ensure that water resources are tion phenomena in dry and wet seasons, while also tak-
used effectively. ing runoff separation processes into account. The overall
pattern of a moderate to strong correlation magnitude and
Soil moisture a positive correlation direction strongly suggests that pre-
cipitation typically causes an increase in soil moisture.
In the Dirima watershed, the estimated mean annual hydro- The maximum correlation coefficient of precipitation and
logical responses in the upper soil layer, such as evapora- soil moisture at the upper, lower, and deep soil layers and
tion, soil moisture, and runoff, are projected to be 1.72, the corresponding time lag intervals are all k = 0 as shown
1.71, 1.79 mm, 0.63, 0.62, and 0.64 mm, and 1.07, 0.97, in Fig. 10. This indicates that soil moisture in different soil
and 1.18 mm in the SSP2-4.5, SSP3-7.0, and SSP5-8.5 sce- layers of the Dirima watershed reaches its maximum value
narios, respectively, shown in Table B. soon after a precipitation event. At a maximum delay, rain-
Based on Fig. 10, it can be observed that regardless of the fall infiltrates in the deeper soil layer, and the upper soil
specific SSP scenario, the rate of evaporation in the upper layer has a peak with a slightly lower cross-correlation
soil layer is higher than the amount of moisture present in value between SM and precipitation.
that same layer. This consistent relationship between evapo- In the Dirima watershed, runoff and SM in the upper
ration and moisture content suggests a fundamental physical layer are more dependent on the availability of precipi-
relationship that is likely to remain unchanged in the future, tation than in other soil layers, because soil layers near
irrespective of other socioeconomic and environmental fac- the surface are more vulnerable to quick changes in SM
tors. Figure 10 shows that in all three SSP scenarios, which due to precipitation. Regarding precipitation, Sehler et al.

13
Modeling Earth Systems and Environment

Fig. 8  Streamflowavailability under three climate change scenar- periods: immediate future (2015–2045) a,intermediate future (2041–
ios (SSP2-4.5 in light red,SSP3-7.0 in green, and SSP5-8.5 in light 2070) b, and distant future (2071–2100) c 
blue) over the baseline (1996–2009 iniris blue) and three evaluation

Fig. 9  Time series runoff from projected climate data a ssp2-4.5, b ssp3-7.0, and c sssp5-8.5: S = surface, I = interflow, and G = groundwater

(2019) support that the strongest positive correlations were In the Dirima watershed, there is a strong positive cor-
found in Africa, Central America, the Middle East, Asia, relation between precipitation and upper SM and runoff. It is
Australia, the eastern part of South America, and much of an essential parameter to identify the irrigation period in the
western North America. Similarly, precipitation is posi- study watershed. The SM in the lower layer is higher than
tively correlated with surface runoff at lag k = 0, as shown in the deep soil layer. When the rain stopped, evapotranspi-
in Fig. 8b and the lower and deeper soil has a delay of one ration from the lower and deeper layers of the soil slightly
to seven days lag. began. This may affect the growth of shallow roots of plants
more than the deep roots of plants in the Dirima watershed.

13
Modeling Earth Systems and Environment

Fig. 10  Soil moisture and evapotranspiration from the projected climate data a  ssp2-4.5, b  ssp3-7.0, c  ssp5-8.5 from 2015–2100 evaluation
period, u = upper, l-lower, and D = deep soil layers from the surface

Also, it implies that the watershed antecedent SM could Conclusions


enhance runoff generation, creating a cumulative effect. The
soil water holding capacity of the Dirima watershed is high The study concluded that estimating hydrological responses
and the most dominant runoff part is the interflow (RI) and based on current and future scenarios is a critical feature of
groundwater (RG). Furthermore, the study investigated the the watershed for real-time hydrological modeling. Under-
impact of climate change using shared socio-economic path- standing the historical hydrological data is critical to the
ways (SSPs). SSPs are realistic future socio-economic sto- forecasting of the hydrological response in each watershed,
ries used to forecast greenhouse gas emissions, population especially during the dry season. The analysis of time series
growth, and other socio-economic factors. By integrating the data revealed that precipitation is observed to be a relatively
SSP scenario into the model structure, the study sought to independent time series with no evident autocorrelation,
assess how changes in climate conditions caused by different meaning that it is not strongly correlated with its past val-
socio-economic channels would affect the future hydrologi- ues. On the other hand, SM in each soil layer has stronger
cal response of the watershed. It was discovered in this study autocorrelation; As the depth and time lag interval increase,
that hydrological models are substantially used to describe the autocorrelation values become significantly weakened.
the watershed hydrological responses. Also, the study is At a lag interval k = 0, the cross-correlation value between
crucial for irrigation managers, and scholars to examine the precipitation and SM in each soil layer increases. The high-
hydrogeological behavior of the watershed in the highland est correlation coefficients of precipitation and SM series at
of the upper Lake Tana basin.

13
Modeling Earth Systems and Environment

various soil layers are all found at a time lag interval of k = 0, storage capacities through the construction of reservoirs or
implying that SM in the watershed reaches its maximum the implementation of groundwater recharge projects can
value immediately after a precipitation event. Similarly, the help mitigate the effects of reduced flow during dry peri-
relationship between runoff and precipitation was highest ods. This, in turn, can increase the risk of droughts in this
at lag interval k = 0 and gradually weakened as the time lag watershed. In general, it is important to consider these poten-
interval and soil depth increased. These findings indicate the tial impacts when planning for the future and developing
immediate influence of precipitation on soil moisture and strategies to mitigate and adapt to climate change and other
runoff in the watershed. It is essential to prioritize sustain- environmental changes.
able water management practices that balance the needs of
Supplementary Information  The online version contains supplemen-
different users while ensuring the long-term availability of tary material available at https://​doi.​org/​10.​1007/​s40808-​023-​01817-z.
water resources.
Based on the average value of three future evaluation Acknowledgements  The authors thank the Africa Centre of Excellence
periods, the precipitation (PCP) decreased by 6%, tmax for Water Management (ACEWM), for partially funding the research
project. All institutes which provided relevant data are highly appreci-
increased by 6%, and Tmin increased by 13%, while ET ated. The Ministry of Water, Irrigation, and Electricity of Ethiopia
increased by 19% and Q decreased by 23% under SSP2- (MoWIE) provides streamflow data. The National Meteorology Agency
4.5. Under SSP3-7.0, PCP is projected to decrease by 8%, of Ethiopia (NMA) Amhara Design & Supervision Works Enterprise
Tmax to increase by 7%, Tmin to increase by 16%, ET to (ADSWE) provides soil data. The Copernicus Global Land Service
(CGLS) provides free land cover data. The Alaska Satellite Facility
increase by 20%, and Q to decrease by 48%. In the SSP5- (ASF) provides a free 30-meter by 30-meter digital elevation model.
8.5 scenario, PCP showed an average increase of 1%, tmax
increased by 10%, Tmin increased by 19%, ET increased by Author contributions  Conceptualization: SBA, FAZ, and TA. Project
22%, and Q decreased by 28%. To adapt to these changes in supervision: FAZ, TA. Original and revised data acquisition, Data
processing, writing original and revised, Methodology: SBA, FAZ,
streamflow, we can help build more resilient communities TA, and GTA. Writing, review, editing, and approvement of the final
and ensure a secure water supply for future generations. The version of the manuscript, all authors.
estimated values for evaporation in the upper soil layer for
these scenarios are 1.72 mm (SSP2-4.5), 1.71 mm (SSP3- Data Availability  The datasets generated and/or analyzed during the
study, as well as the codes that support the findings, can be made avail-
7.0), and 1.79 mm (SSP5-8.5). The corresponding values for able from the corresponding author upon reasonable request. Addition-
soil moisture in the upper soil layer are 0.63 mm (SSP2-4.5), ally, summarized data may also be included as supplementary files.
0.62 mm (SSP3-7.0), and 0.64 mm (SSP5-8.5). Lastly, the
estimated runoff values are 1.07 mm (SSP2-4.5), 0.97 mm Declarations 
(SSP3-7.0), and 1.18 mm (SSP5-8.5). Decision makers can
Conflict of interest  The authors declare that they have no known com-
identify how changes in precipitation, evapotranspiration, petingfinancial interests or personal relationships that could have ap-
runoff, and groundwater recharge rates may impact future peared toinfluence the work reported in this paper
water availability under each SSP scenario by comparing
the model output for each scenario. This, in turn, can help
inform water resource planning and management decisions,
particularly concerning developing adaptation strategies to References
deal with potential changes in water availability.
Streamflow reductions can occur due to changes in pre- Abdulkareem JH, Pradhan B, Sulaiman WNA, Jamil NR (2018)
Review of studies on hydrological modelling in Malaysia. Model
cipitation patterns, such as less frequent or intense rainfall Earth Syst Environ 4(4):1577–1605. https://​doi.​org/​10.​1007/​
events, and increased evapotranspiration from plants due to s40808-​018-​0509-y
warmer temperatures. These changes can lead to decreased Allen RG, Pereira LS, Raes D, Smith M, Ab W (1998) Crop evapo-
water availability in rivers, lakes, and groundwater systems, transpiration - guidelines for computing crop water requirements.
FAO Irrigation and drainage paper. https://​doi.​org/​10.​1016/j.​eja.​
which can affect water supply for human consumption, agri- 2010.​12.​001
culture, and other uses. Increased air temperatures can also Amin MGM, Veith TL, Collick AS, Karsten HD, Buda AR (2017)
exacerbate these effects, as warmer air can hold more water Simulating hydrological and nonpoint source pollution processes
vapor, leading to more evapotranspiration and drier con- in a karst watershed: a variable source area hydrology model
evaluation. Agric Water Manage 180:212–223. https://​doi.​org/​
ditions. The relationship between higher air temperatures 10.​1016/j.​agwat.​2016.​07.​011
and increased evapotranspiration is particularly important. Bai P, Liu X, Liang K, Liu X, Liu C (2017) A comparison of simple
Warmer air can hold more water vapour, which accelerates and complex versions of the Xinanjiang hydrological model in
evapotranspiration rates. As a result, the overall moisture predicting runoff in ungauged basins. Hydrol Res 48(5):1282–
1295. https://​doi.​org/​10.​2166/​nh.​2016.​094
content in the soil and vegetation decreases, contributing Box GEP, Jenkins GM, Reinsel GC (2008) Time Series Analysis fore-
to drier conditions and potentially increasing the risk of casting and control. In: EDITION F (ed) IEEE transactions on
droughts in the watershed. Furthermore, improving water Automatic Control, vol 19. Issue 6). John Wiley & Sons, Inc.,

13
Modeling Earth Systems and Environment

Hoboken, New Jersey. Published simultaneously in Canada Mullen KM, Ardia D, Gil DL, Windover D, Cline J (2011) DEoptim:
https://​doi.​org/​10.​1109/​TAC.​1974.​11007​32 an R package for global optimization by differential evolution. J
Carlos Mendoza JA, Alcazar C, T. A., Zuñiga Medina SA, (2021) Stat Softw 40(6):1–26. https://​doi.​org/​10.​18637/​JSS.​V040.​I06
Calibration and uncertainty analysis for modelling runoff in the Nash JE, Sutcliffe JV (1970) River flow forecasting through con-
Tambo River Basin, Peru, Using sequential uncertainty fitting ceptual models part I - A discussion of principles. J Hydrol
Ver-2 (SUFI-2) algorithm. Air, Soil, Water Res. https://​doi.​org/​ 10(3):282–290. https://​doi.​org/​10.​1016/​0022-​1694(70)​90255-6
10.​1177/​11786​22120​988707 Othman A, Abdelrady A, Mohamed A (2022) Monitoring mass vari-
Dau QV, Kuntiyawichai K, Adeloye AJ (2021) Future changes in water ations in Iraq using time-variable gravity data. Remote Sens
availability due to climate change projections for Huong Basin, 14(14):3346. https://​doi.​org/​10.​3390/​rs141​43346
Vietnam. Environ Processes 8(1):77–98. https://​doi.​org/​10.​1007/​ Patil NG, Pal DK, Mandal C, Mandal DK (2012) Soil water retention
s40710-​020-​00475-y characteristics of vertisols and pedotransfer functions based on
Feyen L, Vázquez R, Christiaens K, Sels O, Feyen J (2000) Application nearest neighbor and neural networks approaches to estimate
of a distributed physically-based hydrological model to a medium AWC. J Irrig Drain Eng 138(2):177–184. https://​d oi.​o rg/​1 0.​
size catchment. Hydrol Earth Syst Sci 4(1):47–63. https://d​ oi.o​ rg/​ 1061/​(asce)​ir.​1943-​4774.​00003​75
10.​5194/​HESS-4-​47-​2000 Pushpalatha R, Perrin C, Le Moine N, Mathevet T, Andréassian V
Gonçalvès J, Petersen J, Deschamps P, Hamelin B, Baba-Sy O (2013) (2011) A downward structural sensitivity analysis of hydrologi-
Quantifying the modern recharge of the “fossil” Sahara aquifers. cal models to improve low-flow simulation. J Hydrol 411(1–
Geophys Res Lett 40(11):2673–2678. https://​doi.​org/​10.​1002/g​ rl.​ 2):66–76. https://​doi.​org/​10.​1016/j.​jhydr​ol.​2011.​09.​034
50478 Pushpalatha R, Perrin C, Moine N, Le, Andréassian V (2012) A
Guo X, Fu Q, Hang Y, Lu H, Gao F, Si J (2020) Spatial variability of review of efficiency criteria suitable for evaluating low-flow
soil moisture in relation to land use types and topographic fea- simulations. J Hydrol 420–421:171–182. https://​d oi.​o rg/​1 0.​
tures on hillslopes in the black soil (mollisols) area of northeast 1016/j.​jhydr​ol.​2011.​11.​055
China. Sustain (Switzerland) 12(9):8–10. https://​doi.​org/​10.​3390/​ Qi WY, Chen J, Li L, Xu CY, Li J, Xiang Y, Zhang S (2022) Region-
SU120​93552 alization of catchment hydrological model parameters for global
Hao F, Sun M, Geng X, Huang W, Ouyang W (2015) Coupling the water resources simulations. Hydrol Res 53(3):441–466. https://​
Xinanjiang model with geomorphologic instantaneous unit doi.​org/​10.​2166/​nh.​2022.​118
hydrograph for flood forecasting in northeast China. Int Soil Rahman MM, Lu M (2015) Model spin-up behavior for wet and dry
Water Conserv Res 3(1):66–76. https://​doi.​org/​10.​1016/j.​iswcr.​ basins: a case study using the xinanjiang model. Water (Swit-
2015.​03.​004 zerland) 7(8):4256–4273. https://​doi.​org/​10.​3390/​w7084​256
Jajarmizadeh M, Harun S, Salarpour M (2012) A review on theoreti- Rathjens H, Bieger K, Srinivasan R, Arnold JG (2016) CMhyd User
cal consideration and types of Models in Hydrology. J Environ Manual. Documentation for preparing simulated climate change
Sci Technol 5(5):249–261. https://​doi.​org/​10.​3923/​jest.​2012.​ data for hydrologic impact studies. 16
249.​261 Ren-Jun Z (1992) The Xinanjiang model applied in China. J Hydrol
Jakada H, Chen Z, Luo M, Zhou H, Wang Z, Habib M (2019) Water- 135(1–4):371–381. https://​d oi.​o rg/​1 0.​1 016/​0 022-​1 694(92)​
shed characterization and hydrograph recession analysis: a com- 90096-E
parative look at a karst vs non-karst watershed and implica- Sehler R, Li J, Reager J, Ye H (2019) Investigating relationship
tions for groundwater resources in Gaolan River basin, Southern between soil moisture and precipitation globally using remote
China. Water(Switzerland). https://​doi.​org/​10.​3390/​W1104​0743 sensing observations. J Contemp Water Res Educ 168(1):106–
Jiang S, Zhang Q, Werner AD, Wellen C, Hu P, Sun J, Deng Y, 118. https://​doi.​org/​10.​1111/j.​1936-​704x.​2019.​03324.x
Rode M (2020) Modelling the impact of runoff generation on Siswanto SY, Sule MIS (2019) The Impact of slope steepness and
agricultural and urban phosphorus loading of the subtropical land use type on soil properties in Cirandu Sub-Sub Catchment,
Poyang Lake (China). J Hydrol. https://​doi.​org/​10.​1016/j.​jhydr​ Citarum Watershed. IOP Conference Series: Earth and Envi-
ol.​2020.​125490 ronmental Science. https://​doi.​org/​10.​1088/​1755-​1315/​393/1/​
Krause P, Boyle DP, Bäse F (2005) Comparison of different effi- 012059
ciency criteria for hydrological model assessment. Adv Geosci Soohong, Kim J, Kang H, Jang WS, Lim KJ (2022) Analysis of
5:89–97. https://​doi.​org/​10.​5194/​adgeo-5-​89-​2005 Water Balance Changes and Parameterization reflecting soil
Li H, Beldring S, Xu CY (2015) Stability of model performance and characteristics in a Hydrological Simulation Program—FOR-
parameter values on two catchments facing changes in climatic TRAN Model. Water 14(6):990. https://​doi.​org/​10.​3390/​w1406​
conditions. Hydrol Sci J 60(7–8):1317–1330. https://​doi.​org/​10.​ 0990
1080/​02626​667.​2014.​978333 Teutschbein C, Seibert J (2012) Bias correction of regional climate
Mohamed A, Gonçalvès J (2021) Hydro-geophysical monitoring model simulations for hydrological climate-change impact
of the North Western Sahara Aquifer System’s groundwater studies: review and evaluation of different methods. J Hydrol
resources using gravity data. J Afr Earth Sc 178:104188. https://​ 456–457:12–29. https://​doi.​org/​10.​1016/j.​jhydr​ol.​2012.​05.​052
doi.​org/​10.​1016/j.​jafre​arsci.​2021.​104188 Tigabu TB, Hörmann G, Wagner PD, Fohrer N (2020) Statistical
Moriasi, D. N., Arnold, J. G., Van Liew, M. W., Bingner, R. L., Harmel, analysis of rainfall and streamflow time series in the Lake Tana
R. D., & Veith, T. L. (2007). Model evaluation guidelines for Basin, Ethiopia. J Water Clim Change 11(1):258–273. https://​
systematic quantification of accuracy in watershed simulations. doi.​org/​10.​2166/​wcc.​2018.​008
Trans ASABE, 50(3), 885–900. Tundu C, Tumbare MJ, Onema JMK (2018) Sedimentation and its
Mounir K, Sellami H, La Jeunesse I, Elkhanchoufi A (2023) Assess- impacts/effects on river system and reservoir water quality: case
ment of future climate and hydrological changes in semi-arid study of Mazowe catchment, Zimbabwe. Proc Int Assoc Hydrol
catchment using the SWAT model and bias-corrected EURO- Sci 377:57–66. https://​doi.​org/​10.​5194/​piahs-​377-​57-​2018
CORDEX ensemble: a case of the Ouergha catchment, North Verma R, Sharif M, Husain A (2022) Application of HEC-HMS for
of Morocco. Model Earth Syst Environ. https://​doi.​org/​10.​1007/​ Hydrological modeling of Upper Sabarmati River Basin, Guja-
s40808-​023-​01775-6 rat, India. Model Earth Syst Environ 8(4):5585–5593. https://​
doi.​org/​10.​1007/​s40808-​022-​01411-9

13
Modeling Earth Systems and Environment

Wagener T, Gupta HV (2005) Model identification for hydrologi- CH4 fluxes through direct insertion in a Land Surface Model.
cal forecasting under uncertainty. Stoch Env Res Risk Assess Remote Sens 2022 14(10):2405. https://​doi.​org/​10.​3390/​RS141​
19(6):378–387. https://​doi.​org/​10.​1007/​s00477-​005-​0006-5 02405
Wang J, Bao W, Gao Q, Si W, Sun Y (2021) Coupling the Xinanjiang Zhao C, Yao S (2020) Application of time series analysis in soil
model and wavelet-based random forests method for improved moisture of fixed dune on Korqin sandy land, Northern China.
daily streamflow simulation. J Hydroinformatics 23(3):589–604. Glob Nest J 22(4):471–476. https://​d oi.​o rg/​1 0.​3 0955/​g nj.​
https://​doi.​org/​10.​2166/​hydro.​2021.​111 003373
Weimin B, Qian L (2012) Estimating selected parameters for the Zhao Ren-Jun (1992) The Xinanjiang model applied in China.
XAJ Model under multicollinearity among watershed charac- J Hydrol 135(1–4):371–381. https://​d oi.​o rg/​1 0.​1 016/​0 022-​
teristics. J Hydrol Eng 17(1):118–128. https://​doi.​org/​10.​1061/​ 1694(92)​90096-E
(asce)​he.​1943-​5584.​00004​15 Zhijia L, Penglei X, Jiahui T (2013) Study of the Xinanjiang Model
Weldegebriel ZB, Amphune BE (2017) Livelihood resilience in the parameter calibration. J Hydrol Eng 18(11):1513–1521. https://​
face of recurring floods: an empirical evidence from North- doi.​org/​10.​1061/​(asce)​he.​1943-​5584.​00005​27
west Ethiopia. Geoenviron Disasters. https://​doi.​org/​10.​1186/​ Zhu D, Das S, Ren Q (2017) Hydrological appraisal of climate
s40677-​017-​0074-0 change impacts on the water resources of the Xijiang basin.
Xin X, Wu T, Shi X, Zhang F, Li J, Chu M, Liu Q, Yan J, Ma Q, South China Water (Switzerland). https://​d oi.​o rg/​1 0.​3 390/​
Wei M (2019) BCC BCC-CSM2MR model output prepared w9100​793
for CMIP6 ScenarioMIP ssp585. Earth Syst Grid Federation.
https://​doi.​org/​10.​22033/​ESGF/​CMIP6.​3050 Publisher’s Note Springer Nature remains neutral with regard to
Yukimoto, S., Kawai, H., Koshiro, T., Oshima, N., Yoshida, K., jurisdictional claims in published maps and institutional affiliations.
Urakawa, S., Tsujino, H., Deushi, M., Tanaka, T., Hosaka, M.,
Yabu, S., Yoshimura, H., Shindo, E., Mizuta, R., Obata, A., Springer Nature or its licensor (e.g. a society or other partner) holds
Adachi, Y., & Ishii, M. (2019). The Meteorological research exclusive rights to this article under a publishing agreement with the
institute earth system model version 2.0, MRI-ESM2.0: descrip- author(s) or other rightsholder(s); author self-archiving of the accepted
tion and basic evaluation of the physical component. J Meteorol manuscript version of this article is solely governed by the terms of
Soc Japan Ser II, 97(5), 931–965. https://​doi.​org/​10.​2151/​jmsj.​ such publishing agreement and applicable law.
2019-​051
Zhang Z, Chatterjee A, Ott L, Reichle R, Feldman AF, Poulter B
(2022) Effect of assimilating SMAP soil moisture on CO2 and

Authors and Affiliations

Simir B. Atanaw1   · Fasikaw A. Zimale2 · Tenalem Ayenew3 · Gebiaw T. Ayele4

2
* Simir B. Atanaw Faculty of Civil and Water Resources Engineering, Bahir
simir.birihan@aau.edu.et; simirbirhan80@gmail.com Dar Institute of Technology, Bahir Dar University, Bahir Dar,
Ethiopia
Fasikaw A. Zimale
3
fasikaw@gmail.com; fasikaw.atanaw@bdu.edu.et School of Earth Science, Addis Ababa University,
Addis Ababa, Ethiopia
Tenalem Ayenew
4
tenalema@geol.aau.edu.et; tenualem@gmail.com Australia River Institute and School of Engineering, Griffith
University, Nathan, QLD 4111, Australia
Gebiaw T. Ayele
gebiaw.ayele@griffithuni.edu.au; g.ayele@griffith.edu.au
1
Africa Centre of Excellence for Water Management
(ACEWM), Addis Ababa University, Addis Ababa, Ethiopia

13

You might also like