Science of The Total Environment: Contents Lists Available at

Science of the Total Environment 689 (2019) 347–365
Contents lists available at ScienceDirect
Science of the Total Environment
journal homepage: www.elsevier.com/locate/scitotenv
Hydrological responses to land use/land cover change and climate

variability in contrasting agro-ecological environments of the Upper Blue
Nile basin, Ethiopia
Mulatu Liyew Berihun a,b,⁎, Atsushi Tsunekawa c, Nigussie Haregeweyn d, Derege Tsegaye Meshesha e,
Enyew Adgo e, Mitsuru Tsubo c, Tsugiyuki Masunaga f, Ayele Almaw Fenta c, Dagnenet Sultan b,
Mesenbet Yibeltal a,b, Kindiye Ebabu a,e
a
The United Graduate School of Agricultural Sciences, Tottori University, 4-101 Koyama-Minami, Tottori 680-8553, Japan
b
Faculty of Civil and Water Resource Engineering, Bahir Dar Institute of Technology, Bahir Dar University, P.O. Box 26, Bahir Dar, Ethiopia
c
Arid Land Research Center, Tottori University, 1390 Hamasaka, Tottori 680-0001, Japan
d
International Platform for Dryland Research and Education, Tottori University, 1390 Hamasaka, Tottori 680-0001, Japan
e
College of Agriculture and Environmental Sciences, Bahir Dar University, P.O. Box 1289, Bahir Dar, Ethiopia
f
Faculty of Life and Environmental Science, Shimane University, Shimane, Matsue 690-0823, Japan
H I G H L I G H T S G R A P H I C A L A B S T R A C T
• Land use/land cover change and climate

variability can cause hydrological re-
sponses.
• We examined 35-year trends in land
use/land cover and climate in three wa-
tersheds.
• Land use/land cover change caused
higher surface runoff and lower evapo-
transpiration.
• Climate variability increased evapo-
transpiration in all three watersheds.
• Land use/land cover change had a dom-
inant role in the hydrological responses.
a r t i c l e i n f o a b s t r a c t
Article history: Land use/land cover (LULC) change and climate variability are two major factors controlling hydrological re-
Received 29 January 2019 sponses. The present study analyzed the separate and combined effects of these two factors on annual surface
Received in revised form 15 May 2019 runoff and evapotranspiration (ET) after validating the selected models in three drought–prone watersheds of
Accepted 21 June 2019
the Upper Blue Nile basin: Kasiry (highland), Kecha (midland), and Sahi (lowland). LULC maps were produced
Available online 27 June 2019
from aerial photographs and very-high-resolution satellite images from 1982, 2005/06 and 2016/17. During
Editor: Ralf Ludwig 1982–2016/17 the area covered by natural vegetation showed dramatic decreases, ranging from 60.2% in Kasiry
to 51.8% in Sahi. In contrast, increases in cultivated land ranged from 36.7% in Kasiry to 279.6% in Sahi; the smaller
Keywords: increase in Kasiry resulted from the conversion of a portion of the cultivated land to an Acacia decurrens planta-
Surface runoff tion after 2006. The observed LULC changes over the study period resulted in runoff increases ranging from 4% in
Evapotranspiration Kecha to 28.7% in Kasiry. Climate variability in terms of annual rainfall had no significant effect on estimated
⁎ Corresponding author.
E-mail address: [email protected] (M.L. Berihun).
https://doi.org/10.1016/j.scitotenv.2019.06.338
0048-9697/© 2019 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
348 M.L. Berihun et al. / Science of the Total Environment 689 (2019) 347–365
Drought-prone runoff; whereas both LULC change and climate variability had significant effect on estimated ET. Though climate
High-resolution satellite images variability increased ET from 33.6% in Kecha to 42.1% in Kasiry, the LULC change related to the reduction in nat-
Sustainable land management ural vegetation had an offsetting effect, which led to overall decreases in ET ranging from 15.8% in Kasiry to 32.8%
in Kecha watershed. As changes in LULC and climate are expected to intensify in the future, it is important to
study further hydrological responses considering these changes to devise future sustainable land and water man-
agement strategies.
© 2019 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://
creativecommons.org/licenses/by-nc-nd/4.0/).
1. Introduction agricultural land and unplanned LULC changes. Previous studies in dif-
ferent parts of the basin (e.g. Mekonnen et al., 2018a, 2018b; Gashaw
Land use/land cover (LULC) change is a major challenge facing the et al., 2018; Woldesenbet et al., 2018, 2017; Worku et al., 2017) assessed
global environment (Kates and Torrie, 1998). In particular, the rapid in- the effect of LULC change and climate variability on hydrological re-
crease in population pressure in developing countries has pronounced sponses. These studies used process-based hydrological models that
effects on the LULC dynamics mainly through deforestation aimed at in- constitute a single agro-ecological environment and uniform human ac-
creasing agricultural production (Maitima et al., 2009). Ethiopia is one tivities. As such, further research is needed in the basin to better under-
of the developing countries where agriculture is the backbone of the stand the responses of hydrological processes under LULC and climate
economy, and where agriculture is facing a major environmental chal- change at small watershed scales under different agro-ecologies (Dile
lenge from LULC change (Dessie and Kleman, 2007; Minta et al., 2018; et al., 2018).
Tekle and Hedlund, 2000). Previous studies in different parts of the Process-based hydrological models are not necessarily more useful
country, particularly in the Upper Blue Nile (UBN) basin, have shown re- than those whose parameters can be easily determined from available
markable LULC dynamics induced by human activities such as defores- data (Haregeweyn et al., 2016; Savenije, 2009). It is difficult to under-
tation or reforestation (e.g. Bewket, 2002; Gashaw et al., 2017; take detailed process-based hydrological models to analyze hydrologi-
Gebrehiwot et al., 2014; Tekle and Hedlund, 2000; Zeleke and Hurni, cal processes in the UBN basin because observed climate and
2001). On the one hand, the basin has largely been experiencing an ex- hydrological data in the basin (Awulachew et al., 2008; Conway, 2000,
pansion of agricultural land at the expense of natural vegetation cover 1997; Tekleab et al., 2014), and in the study watersheds, are limited.
(mainly forest) as people search for new land for cultivation and grazing Moreover, because as mentioned above surface runoff and ET are key
(e.g. Bewket, 2002; Gashaw et al., 2017; Zeleke and Hurni, 2001). On the components in the water balance equation and are closely linked with
other hand, some studies show that the rate of deforestation has re- changes in LULC and climate (Bosch and Hewlett, 1982; Costa et al.,
cently been reduced and the vegetation cover has improved in some 2003; Woldesenbet et al., 2018; Yang et al., 2012; Yin et al., 2017;
parts of the country because of plantation activities on degraded hill- Zhang et al., 2001), it is vital to estimate and evaluate each key compo-
sides (e.g., Wondie and Mekuria, 2018). nent individually under LULC and climate variability to assess the water
LULC change is among the most important factors contributing to al- availability in a given watershed. There have been a number of attempts
terations of the land surface across all spatial and temporal scales (Bosch to estimate these two key components only using existing climate data
and Hewlett, 1982; Conway, 2000; Legesse et al., 2003). LULC changes or the water balance equation. However, estimating ET is a complex
also have a great impact on hydrological processes such as surface run- process at watershed scale because it is affected by many factors, includ-
off, groundwater recharge, infiltration, interception, and evapotranspi- ing rainfall interception, net radiation, turbulent transport, plant avail-
ration (ET) (Costa et al., 2003; Fang et al., 2013; Gashaw et al., 2018; able water, and vegetation characteristics (Zhang et al., 2001). These
Guo et al., 2008; Legesse et al., 2003; Woldesenbet et al., 2017; Worku factors can be combined, however, to evaluate the response of ET
et al., 2017; Zhang et al., 2001). Several previous studies have shown under vegetation changes by considering their net effects, assuming
the effects of LULC change on runoff, ET, or both (key components in that ET from land surfaces is controlled by water availability and atmo-
the water balance equation), at various spatial and temporal scales spheric demand (Zhang et al., 2001; Sun et al., 2005). Thus, individually
(Bosch and Hewlett, 1982; Dong et al., 2015; Fang et al., 2013; assessing the responses of surface runoff and ET at watershed scale to
Gashaw et al., 2018; Li et al., 2017; Worku et al., 2017; Yang et al., change in LULC and climate variability using experimentally validated
2012; Yin et al., 2017; Zhang et al., 2001; Zhang et al., 2014). Most of empirical models, is vital for integrated watershed management. There-
these studies agree that the expansion of agricultural land at the ex- fore, the main objective of this study was to improve our current under-
pense of vegetation cover markedly increases the runoff potential in a standing of key hydrological responses to historical changes in LULC and
given watershed (Bosch and Hewlett, 1982; Dong et al., 2015; Fang climate variability observed over the last 35 years in three different
et al., 2013; Gashaw et al., 2018; Teklay et al., 2018; Worku et al., agro-ecological environments in the UBN basin. The specific objectives
2017). In contrast to the effect on runoff, the conversion of forest were (1) to assess LULC change and climate variability of watersheds
cover to other LULC types notably reduces ET (Fang et al., 2013; Li in different agro-ecological environments, and (2) to analyze the single
et al., 2017; Yang et al., 2012; Zhang et al., 2001). In addition to LULC, cli- and combined effects of LULC change and climate variability on surface
mate change or variability is one of the most significant factors influenc- runoff and ET in those watersheds.
ing the changes in runoff and ET (Chen et al., 2006; Dong et al., 2015;
Fenta et al., 2017a; Ficklin et al., 2010; Guo et al., 2008; Li et al., 2017; 2. Materials and methods
Ma et al., 2009; Mekonnen et al., 2018; Woldesenbet et al., 2018; Yang
et al., 2017; Yin et al., 2017; Zhang et al., 2014). However, the degree 2.1. Description of the study areas
to which LULC or climate changes influence variations in runoff and ET
varies depending on the characteristics of a watershed or basin and This study was conducted in three representative watersheds lo-
agro-ecological settings of the study sites (e.g, Dong et al., 2015; Guo cated within the UBN basin in Ethiopia (Fig. 1). The watersheds belong
et al., 2008; Li et al., 2017; Ma et al., 2009; Mekonnen et al., 2018; to different administrative districts: Kasiry is in Fagita Lekoma District
Yang et al., 2017). (10°59′30″–11°1′0″N, 36°54′0″–36°56′0″E) and Kecha is in Bahir Dar
The UBN basin is increasingly under human pressure because of a Zuria District (11°38′0″–11°40′30″N, 37°29′30″–37°31′0″E) of the Am-
rapidly growing population. This aggravates various human-induced re- hara Regional State; Sahi is in the Debatie District (10°45′30″–10°47′
source degradations, mainly because of increased demand for 0″N, 36°16′0″–36°18′0″E) of the Benishangul Gumuz Regional State.
M.L. Berihun et al. / Science of the Total Environment 689 (2019) 347–365 349
Fig. 1. Location and elevation maps of Kasiry, Kecha, and Sahi watersheds in the Upper Blue Nile basin of Ethiopia. AC: Acacia decurrens at slope 5% (AC1) and 25% (AC2) BL: bushland; CL:
cultivated land at slope 5% (CL1) and 15% (CL2); GL: grazing land at slope 15% and EP: eucalyptus plantation at slope 25%.
These three watersheds were selected to represent three different and sloping (Kecha (38.5% of area)) slope categories. The area falls on
agro-ecological environments, classified on the basis of elevation, pre- very steep slope categories of Kasiry watershed (31% of area) is higher
cipitation, and cropping systems (Hurni et al., 2016) (Table 1, Fig. 2). than Shai (10.8% of area) followed by Kecha (4.2% of area) watershed.
Each agro–ecological environment is characterized by specific topo- On the other hand, the area considered in flat and gentle slope catego-
graphical features, climate conditions, major soil types, soil and water ries (10%, 36% and 21.5% area of Kasiry, Kecha and Sahi watersheds, re-
conservation (SWC) practices and dominant agricultural practices spectively) are dominantly used for crop production and residences. In
(Table 1). The elevations range from 1501 m above sea level (m.a.s.l) 35 years (1982–2016) of recorded data obtained from the respective
at Sahi to 2857 m.a.s.l. at Kasiry. The slope was generated from SRTM– nearby meteorological stations, the mean annual rainfall generally in-
DEM and reclassified into five categories: flat (0–3% slope), gentle creased in the order Sahi (lowland) b Kecha (midland) b Kasiry (high-
(3–8% slope), sloping (8–15% slope), steep (15–30% slope) and very land) (Table 1). Even though the watersheds have located in different
steep (N30% slope). The area of the study watersheds predominantly agro-ecological environments, they are characterized by similar rainy
characterized by steep [Kasiry (40% of area) and Sahi (36.4% of area)] (Jun to October) and dry (November to May) seasons, with N86% of
Table 1
Main biophysical characteristics of the three study watersheds: Kasiry, Kecha, and Sahi in the Upper Blue Nile basin, Ethiopia.
Characteristic Kasiry (highland) Kecha (midland) Sahi (lowland)
Area (ha) 397.5 424.3 398.2

Elevation range (m.a.s.l.)a 2498–2857 1918–2126 1501–1718
Agro-ecology zone Moist subtropical (Wet Dega) Humid subtropical (Moist Weyna Dega) Tropical hot humid (Moist Kolla)
Mean monthly temp. (°C) 15–18 17–23 20–25
Mean annual rainfall (mm) 2495 1343 1022
Rainfall pattern Unimodal Unimodal Unimodal
Dominant cropsb Barley, teff, wheat, and potatoes Finger millet, teff, maize, and wheat Finger millet, teff, maize, and groundnut
Dominant livestock Cattle, sheep, donkeys, and horses Cattle, sheep, goats, and donkeys Cattle, sheep, goats, and donkeys
Sources: Ebabu et al. (2018), Sultan et al. (2018), and surveys by the authors.
a
Meters above sea level.
b
Teff (Eragrostis tef), finger millet (Eleusine coracana), wheat (Triticum aestivum), maize (Zea mays), and groundnut (Arachis hyogaea).
the rainfall concentrated during the rainy seasons. In contrast, average At present, each watershed is being managed under governmental
monthly temperature decreased from lowland to highland in the or non-governmental SWC programs, sometimes both. The Kasiry wa-
order Sahi N Kecha N Kasiry watershed (Fig. 2). The dominant soil tershed has been implementing SWC measures since 2008, supported
types (in the FAO classification system) were identified in the three by a Swiss Development Cooperation project. Similarly, the Kecha wa-
study watersheds (Mekonnen, 2018): Acrisols, Luvisols, Leptosols, and tershed has been part of the National Sustainable Land Management
Vertisols. All four soil types occur in the Kasiry watershed, whereas Programme since 2011, supported by funds obtained from the World
only three (Luvisols, Vertisols, and Leptosols) are present in Sahi and Bank. Unlike Kasiry and Kecha watersheds, Sahi has not received exter-
two (Luvisols and Leptosols) in Kecha watershed. nal support for SWC programs. In fact, however, a few physical SWC
structures have been constructed through the regular governmental ex-
tension programs or campaign-based community mobilization since
2008. The dominant agricultural practice in these three watersheds is
mixed crop–livestock farming (Table 1).
2.2. Data types and sources
2.2.1. LULC and topography data

Spatial and temporal LULC data were obtained from remotely sensed
images (aerial photographs, scale 1: 50,000, and satellite images with a
resolution of 0.5–3.2 m) covering 35 years from 1982 to 2016/17
(Table S1). Sixteen hardcopy aerial photographs were from the
Ethiopian Mapping Agency (http://www.ema.gov.et) to analyze LULC
change in addition to orthorectified satellite images provided by
AIRBUS defense and space (https://www.airbus.com/space.html). The
images were selected by considering data availability, anticipated
major LULC changes, and year consistency across the study watersheds.
All satellite images were taken during the dry season (January–March)
under clear cloud-cover conditions in three different years during the
study period (Table S1). Moreover, we used a Shuttle Radar Topo-
graphic Mission (SRTM) digital elevation model (DEM) with a spatial
resolution of 30 m (http://earthexplorer.usgs.gov/) to delineate the
boundaries of the watersheds and to describe the watersheds' topo-
graphic characteristics.
2.2.2. Temperature, rainfall, streamflow, and runoff data

Daily temperature (maximum and minimum) and rainfall data were
obtained from meteorological stations of the National Meteorology
Agency (NMA) of Ethiopia, located in Enjibara and Dangila for Kasiry,
Bahir Dar for Kecha, and Bullen for Sahi watersheds. To validate the
long-term rainfall data from these stations, the daily rainfall data were
collected in the three study watersheds during 2015 and 2016 using
manual rain gauges. For this validation purpose, the monthly rainfall
data of 2015 and 2016 from the stations and study watersheds were ag-
gregated from daily rainfall data. Based on the validation results, the re-
gression equations were developed to estimate long-term rainfall data
of the study watersheds. Three “Mini Diver” data loggers (Schlumberger
Water Services, The Netherlands) were also installed at the outlets of
the three watersheds to monitor streamflow during 2015 and 2016.
Fig. 2. Long-term (1982–2016) monthly minimum temperature (Tmin), maximum
The base flow was separated by using the WHAT Web-based
temperature (Tmax), average temperature (Tmean) and average rainfall in Kasiry, Hydrograph Analysis Tool (https://engineering.purdue.edu/mapserve/
Kecha, and Sahi watersheds. WHAT/) to obtain surface runoff in the corresponding periods.
Daily runoff during 2015 and 2016 was also measured from runoff Percentage changes (Long et al., 2009) of individual LULC types were
plots with areas of 180 m2 (30 m long × 6 m wide). We used a total of computed to describe the extent of change between two time points as:
15 plots established during 2014/15 (Sultan et al., 2018); seven in
Kasiry, four in Kecha, and four in Sahi watersheds selected by consider- A2 −A1
Percent change ð%Þ ¼ 100 ð1Þ
ing different slope gradients and land-use types: cultivated land in two A1
slope ranges (5% and 15%), grazing land (15% slope), and bushland (35%
slope). In the Kasiry watershed, we used three additional plots: two in where A1 is the area in year 1 and A2 is the area in year 2 of a LULC type
Acacia decurrens plantations (on 5% and 25% slopes), and one in a Euca- (ha). In addition, we used a transition matrix to describe the extent and
lyptus plantation (25% slope). At the lower end of each plot, a trench ca- nature of LULC changes observed and to explore the internal conversion
pable of holding a volume of 9.7 m3 was excavated; each trench was of different LULC types (Guo et al., 2009). The percentage of “conversion
trapezoidal in cross section and lined with an impermeable loss to” or “conversion gain from” were calculated in relation to the total
geomembrane plastic to allow the collection of surface runoff. A detailed loss or gain in each LULC type between 1982 and 2016/17 according to
description of the design, dimensions, instrumentation, and measure- Eqs. (2) and (3) (Guo et al., 2009; Long et al., 2009):
ments of each experimental runoff plot is given by Sultan et al.
(2018). Runoff was measured daily during the main rainy season P lossðiÞ; j ¼ pi; j −p j;i =ðpci −pri Þ 100; i≠j ð2Þ
(June to October) in each year; N86% of the rainfall in the study water-
sheds is concentrated in these months (Sultan et al., 2017). Rainfall P gainðiÞ; j ¼ p j;i −pi; j =ðpci −pri Þ 100; i≠j ð3Þ
events outside of the main rainy season are generally characterized by
smaller depth and intensity than those occurring during the rainy sea- where Ploss(i),j is the percentage taken by LULC type j in the total “con-
son. In addition, soils are generally dry and have high infiltration capac- version loss” of type row i i; Pgain(i),j is the percentage taken by type j
ity outside of the main rainy season, leading to lower runoff responses in the total “conversion gain” of type row i; pi, jpi,j andpj, i pj,i are the in-
(Descheemaeker et al., 2006; Nyssen et al., 2009; Zenebe et al., 2013). dividual entries in a given change matrix; pci is the column total of type
i; and pri is the row total of for type i.
2.3. Image classification and LULC change detection
2.4. Surface runoff estimation
Satellite images and aerial photographs were preprocessed before
image classification to enhance the quality of the image and visibility Methods and approaches to the estimation of watershed surface
of the features (Guo et al., 2009). In this study, first we employ unsuper- runoff range from simple empirical to more complex (conceptual and
vised and supervised classification techniques. However, when the re- process-based) rainfall–runoff models. However, more complex models
spective classification results were evaluated, a significant level of are not necessarily more useful than simpler models, whose parameters
error occurred whereby some areas were found to have been can easily be determined from available data (Haregeweyn et al., 2016,
misclassified because of substantial spectral similarity among some 2017; Savenije, 2009). As a whole in the UBN basin, it is difficult to apply
LULC types (e.g., forest, plantation and bushland, bushland and grazing complex hydrological models because of limited observed data for cli-
land, cultivated land and grazing land). Therefore, we followed on- matic and hydrological parameters (Awulachew et al., 2008; Conway,
screen digitization technique through intensive visual interpretation 2000, 1997; Dile et al., 2018; Haregeweyn et al., 2015a, 2015b;
of very high-resolution satellite images (Quick Bird, Worldview-2, Tekleab et al., 2014). Relatively less complex models such as unit
SPOT7, Pleiades and IKONOS, Table S1) in a GIS environment based on hydrograph derived from a spatially distributed velocity field
the LULC types presented in Table S2. This classification process was (Maidment et al., 1996) is one of the runoff estimation approaches,
done after first acquainting ourselves to the study areas trough organiz- however, this model has not been parametrized with the LULC informa-
ing field visits along with supporting information collected in the field tion which is a key input parameter in this study. Thus, we selected a
and interviews with elders in the local population. Although the digitiz- simple proportional loss model known as the runoff coefficient (RC)
ing process was found time consuming and tedious, the resulting LULC method (Geiger et al., 1987) to estimate surface runoff in this study.
classification maps being verified in the field for the 2016/17 cover are This method is widely used at plot and watershed scales in Ethiopia
more reliable than automatic classification techniques. Accuracy assess- (e.g. Conway, 2000; Descheemaeker et al., 2006; Haregeweyn et al.,
ment through the use of ground control points is important to deter- 2016, 2012; Nyssen et al., 2010; Taye et al., 2013; Zenebe et al., 2013).
mine the quality of classified LULC maps from remotely sensed data Plot-scale daily and seasonal RC (%) values were calculated by dividing
(Congalton and Green, 2009). The accuracy assessment is mostly ap- the runoff yield (R, mm) by the corresponding rainfall depth (P, mm)
plied when the classification of LULC maps is attempted through unsu- using Eq. (4).
pervised and supervised classification techniques (e.g., Congalton and
Green, 2009; Dessie and Kleman, 2007; Gashaw et al., 2017; Minta RCi ¼ Rp =P 100 ð4Þ
et al., 2018). As far as the classification is done through visual interpre-
tation by on-screen digitization technique, attempting the accuracy as- where RCi is the runoff coefficient calculated for up to six of the land-use
sessment is unnecessary (e.g., Bewket, 2002; Gebrelibanos and Assen, types i described in Section 2.2.2. P is rainfall depth (mm), and Rp is sur-
2015; Zeleke and Hurni, 2001) because it is possible to easily identify face runoff yield (mm), calculated by dividing the runoff volume mea-
the LULC types from very high-resolution satellite images (resolutions sured at the collecting trench (after subtracting the direct rainfall
finer than 3.2 m, Table S1), which is equivalent to field-based mapping. falling on the open trench) by the runoff-plot area (6 m × 30 m).
The images were classified by identifying a minimum of four and a max- Daily runoff coefficients were estimated by observing a total of 257,
imum of six LULC types for each year in each watershed (Table S2). Eu- 199 and 121 daily rainfall–runoff events in 2015 and 2016 in Kasiry,
calyptus plantation and natural forest were combined into one forest Kecha, and Sahi watersheds, respectively. Similarly, the seasonal RCs
land type in the Kecha and Sahi watersheds because the area of each for 2015 and 2016 were calculated for each land-use type by dividing
was negligible compared with the areas of the other types. In the Kasiry the seasonal runoff measured for each land-use type by the respective
watershed, natural forest, riverine trees, and plantation (uncommon in seasonal rainfall (Eq. (4)). Then the average seasonal RC for each land-
1982) were combined into the forest class in the 1982 LULC classifica- use type was calculated from the respective RCs for 2015 and 2016.
tion process because they had the same image tone, which made The average seasonal RCs for each land-use type were associated
them difficult to differentiate. Finally, LULC maps for the year 1982, with the corresponding watershed LULC types, identified from the
2005/06 and 2016/17 were produced for further analysis. LULC maps for 1982, 2005/06 and 2016/17. Because no runoff plots
were established in forest or settlements, the RCs for these LULC types data using the Mann–Kendall (MK) (Burn, 1994; Westmacott and
were adopted from the literature. Using the average seasonal RC Burn, 1997) and Pettitt's (Pettitt, 1979) tests. The MK trend and Pettitt's
(RCav) and the area for each LULC type, a weighted RC for each water- homogeneity tests have been widely applied to detect monotonic and
shed was estimated for a specific period using Eq. 5. homogeneous trends, respectively, to determine change points in
! long-term hydro-climatic time-series data (e.g. Chen et al., 2006;
X
n X Fenta et al., 2017a; Ma et al., 2009; Zhang et al., 2014). These tests
RCwt ¼ RCav Ait = Ait 100 ð5Þ
i
were selected because of their robustness with respect to missing and
tied values and to non-normality, which are common in hydro-
where RCwt is the weighted watershed RC (%) for LULC study period t climatic time series; moreover, they have the same power as their para-
(1982, 2005/06 or 2016/17) and Ait is the area (ha) of LULC type i in pe- metric counterparts (Kahya and Kalayci, 2004). The MK standard nor-
riod t (1982, 2005/06 or 2016/17). Because of a lack of long-term sea- mal Zc are given as follows:
sonal and annual surface runoff data, the daily surface runoff for 2015
s−1 Sþ1
and 2016 was estimated by multiplying the weighted RC for 2016/17 Z c ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ; SN0 0; S ¼ 0pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ; Sb0 ð10Þ
by the daily rainfall observed in the same year, and then summed varðsÞ varðsÞ
these estimates to generate monthly time-steps for the purpose of val-
idation as outlined in Section 2.5. X
n−1 X
n
S¼ sgnðxk −xi Þ ð11Þ
Lastly, we estimated long-term (1982–2016) annual surface runoff i¼1 k¼iþ1
for each watershed using the weighted watershed RC (RCwt) and long-
term annual rainfall (1982–2016) in Eq. 6: where var(S) is normal distribution variance, S is the test statistic,
x k and x i are sequential data values, n is the length of the data
Rt ¼ RCwt P t ð6Þ
series, and sgn(x) is equal to 1, 0, or −1 when
x is greater than, equal to, or less than zero, respectively. The null
where Rt is the estimated annual surface runoff (mm) and Pt is the long-
hypothesis H a0 (there is no trend) is accepted if −Z1 − α/2 ≤ Zc
term annual rainfall (mm) in year t from 1982 to 2016.
≤ Z1 − α/2 at the significance level α = 0.05.
According to Pettitt (1979), x1, x2, …. . , xn is a series of observed data
2.5. Validation of surface runoff model
that has a change point t if x1, x2, …. . , xt has a distribution function F1(x)
that is different from the distribution function F2(x) of the second part of
Long-term observed surface runoff data were not available for vali-
the series, xt+1, xt+2,….., xT. The non-parametric test statistic Ut, T for this
dation purposes; therefore, we validated the estimated watershed-
test is calculated as follows:
level daily and monthly surface runoff by using the observed surface
runoff measured at the watershed outlets under present-day conditions t X
X T
(2015 and 2016). The performance of the validated empirical model U t;T ¼ sgn xi −x j ð12Þ
was statistically evaluated by using the coefficient of determination, i¼1 j¼iþ1
R2, of linear regression curves (Moriasi et al., 2007), Nash-Sutcliffe Effi-

ciency (NSE; Saleh et al., 2000), and percent bias (PBIAS; Van Liew et al., where sgn (x) = 1 if x N 0, 0 if x = 0, and −1 if x b 0. The statistic Ut, T is
2007). NSE indicates how well the plot of observed versus estimated considered for 1 ≤ t b T. The test statistic K for the sample series with
data fits the 1:1 line; the performance is perfect if NSE = 1. PBIAS mea- length n is defined as
sures the average tendency of the estimated data to be larger or smaller
K T ¼ max U t;T ð13Þ
than their observed counterparts; the optimal value of PBIAS is 0.0, with 0≤t ≤T
low-magnitude values indicating more accurate model estimation. R2
ranges from 0.0 to 1.0, with higher values indicating better agreement. The associated probability (P) used in the test is estimated using the
We estimated these statistical indices by using the following equations: equation of Gao et al. (2010) as follows:
" #2 n o
n
2 ∑i¼1 Oi −Oavg P i −P avg P≅ exp −6ðK T Þ2 = T 3 þ T 2 ð14Þ
R ¼ n 2 n 2
ð7Þ
∑i¼1 Oi −Oavg ∑i¼1 P i −P avg
when P is smaller than the specific confidence level with significance
" #
n level α (in this study α = 0.05), then the null hypothesis Hb0 assuming
∑i¼1 ðOi −P i Þ2
NSE ¼ 1− n 2 ð8Þ the presence of homogeneity trend will be rejected.
∑i¼1 Oi −Oavg
" n # 2.6.2. Analyzing the effect of LULC change and climate variability on surface
∑i¼1 ðOi −P i Þ 100 runoff
PBIAS ¼ n ð9Þ
∑i¼1 Oi Human activities associated with LULC are the major drivers of
changes in the surface runoff response (Bosch and Hewlett, 1982;
where Oi is the ith observed value, Oavg is the average observed value for Costa et al., 2003; Fenta et al., 2017a). We used Eq. (6) to estimate the
the entire study period, Pi is the ith predicted (modeled) value, and Pavg annual surface runoff for the three watersheds during 1982–2016 ac-
is the average of the predicted value over the entire study period. cording to the LULC classification maps for 1982, 2005/06, and 2016/
17, the weighted watershed RCs, and long-term rainfall data
2.6. Analyzing surface hydrological responses to change in LULC and cli- (1982–2016). We compared the estimated surface runoff under each
mate variability LULC scenario by considering the LULC changes within and across the
watersheds. To quantify the influence of LULC changes, we evaluated
2.6.1. Trend detection in rainfall and temperature time-series variations in annual surface runoff under the LULC scenarios for 1982,
The analysis of historical trends of climate variables such as temper- 2005/06 and 2016/17, using the same climate data as the annual rainfall
ature and rainfall can help to reveal the effect of climate change or var- record from 1982 to 2016 in all watersheds. The coefficient of variation
iability on water resources (Chen et al., 2006; Fenta et al., 2017a; Guo (CV) of estimated surface runoff in each watershed was calculated to
et al., 2008; Ma et al., 2009; Woldesenbet et al., 2018; Yin et al., 2017). compare the variation of surface runoff pattern as a result of LULC sce-
We analyzed trends in annual rainfall and temperature time-series narios. We also estimated the contribution of each LULC type to mean
annual surface runoff change from the area contribution and runoff co- (SC2), 2005/06 LULC map and climate data for period 1; Scenario 3
efficient of each LULC type in the watersheds. In addition, we assessed (SC3), 1982 LULC map and climate data for period 2; and Scenario 4
the effect of climate variability as rainfall change on surface runoff (SC4), 2005/06 LULC map and climate data for period 2. Finally, after
using the methods described in Section 2.6.4. comparing the outputs from these scenarios, we determined the sepa-
rate effects of LULC change and climate variability on hydrologic re-
2.6.3. Analyzing the effect of LULC change and climate variability on actual sponses using Eqs. (17)–(19).
evapotranspiration
The amount of actual evapotranspiration (ET) in a given watershed 1
ΔH L ¼ ðΔH L1 þ ΔH L2 Þ ð17Þ
strongly depends on climate variables vegetation covers (Sun et al., 2
2006, 2005; Yang et al., 2012; Zhang et al., 2001; Zhang et al., 2014)
where ΔHL is the change in hydrological responses due to the separate
and climate forcing factors such as CO2, aerosols, greenhouse gases,
effect of LULC change and ΔHL1 and ΔHL2 are the changes in hydrological
ozone (Ainsworth and Long, 2005; Ficklin et al., 2010; Liu et al., 2016).
responses calculated as the difference between the outputs of SC2 and
However, the cumulative or net effects of climate forcing factors can
SC1 in period 1, and SC4 and SC3 in period 2, respectively. We applied
be indirectly reflected in climatic variables such as temperature and pre-
a similar approach to quantify the separate effect of climate variability
cipitation (IPCC, 2013) that are already considered in this study. In par-
on hydrological responses using the following equation:
ticular, long-term change in LULC and related management strategies,
along with climate variability are expected to have an effect on water- 1
shed ET and hence water yield and groundwater recharge. Therefore, ΔH C ¼ ðΔH C1 þ ΔH C2 Þ ð18Þ
2
quantifying long-term effects of LULC change on annual ET is extremely
important. We used an empirical model which is developed by Zhang where ΔHC is the change in hydrological responses due to the separate
et al. (2001) based on data from 250 watersheds worldwide (Eq. 15) effect of climate variability and ΔHC1 and ΔHC2 are the changes in hydro-
to quantify watershed annual actual ET under long-term LULC change logical responses calculated as the difference between outputs of SC3
situations. and SC1 under LULC of 1982 and SC4 and SC2 under LULC of 2005/06,
respectively. Finally, the total changes in hydrological responses
0 1
PET (ΔHLC) were calculated as the sum of the two effects (Eq. (19)), or, alter-
1þw
B P C natively, estimated from the difference between SC4 and SC1:
ETi ¼ @ AP ð15Þ
PET P
1þw þ
P PET ΔH LC ¼ ΔHL þ ΔH C ð19Þ
where ETi is annual actual evapotranspiration during period i (1982, Previous studies have applied similar approaches using hydrological
2005/06 or 2016/17) for one LULC type with a specific value of w, w is model simulations to evaluate the separate effects of LULC change and
the plant available water coefficient representing the relative difference climate variability on hydrological responses (e.g. Fang et al., 2013;
in the way plants use water for transpiration, P is rainfall depth (mm), Mekonnen et al., 2018a, 2018b; Guo et al., 2008; Ma et al., 2009;
and PET is potential evapotranspiration, obtained by using an equation Woldesenbet et al., 2018; Yang et al., 2017; Yin et al., 2017). Note that
developed by Hargreaves and Samani (1985). For a watershed with in our study, we replaced the general representation of hydrological re-
mixed LULC types (there are four types based on w values), ET is calcu- sponses (ΔH) with the specific hydrological component under consider-
lated as follows: ation; for example, we use ΔET for the change in actual ET and ΔR for the
X change in surface runoff.
ET ¼ ðETi Ri Þ ð16Þ
3. Results and discussion
where Ri is the proportion of each LULC type. The w parameter was set
to 0.5 for grazing and cultivated land, 2.0 for forests and plantations 3.1. Extent, trend, and conversion of LULC between 1982 and 2017
(Zhang et al., 2001), 1.0 for bushland, and 0.0 for settlements (Sun
et al., 2005). Khat (Catha edulis) (Table S2) cultivation in Kecha water- 3.1.1. Kasiry (highland) watershed
shed was included under the bushland LULC type. The contribution of LULC maps for 1982, 2006, and 2017 were prepared for Kasiry wa-
each LULC type to annual ET was also evaluated based on the relative tershed (Fig. 3a). In 1982, forest was the dominant LULC type with
area extent of each on the 1982, 2005/06 and 2016/17 LULC maps and 40.6% coverage, making this watershed relatively forest-rich among
the plant available water coefficients. the study watersheds at that time. Cultivated land and grazing land
were second (23.1%) and third (19.5%), respectively during this period
2.6.4. Framework for differentiating effects of LULC change and climate (Figs. 3a, 4a). In 2006, plantation (mainly A. decurrens) covered 10.9%
variability of this watershed, and its coverage almost quadrupled by 2017
The changes in hydrological responses in a given watershed result (39.3%). Forest, bushland, and grazing land continuously decreased to
from LULC and climate changes, which are assumed to be independent 7.1%, 5.2%, and 13.2%, respectively, in 2017 (Fig. 4a). Conversely, culti-
factors (Chen et al., 2006; Guo et al., 2008; Legesse et al., 2003; Ma et al., vated land coverage doubled from 1982 to 2006, but decreased by
2009; Woldesenbet et al., 2018; Yang et al., 2017; Yin et al., 2017; Zhang 31.5% between 2006 and 2017. Unlike other LULC types, plantation in-
et al., 2014). Differentiating the effects of LULC change and climate var- creased dramatically between 2006 and 2017, by 261.7% (Fig. 4a). Com-
iability involved three steps. First, we divided the study period into two pared to other LULC types, settlements accounted for the least coverage
periods that were determined by fixing trend change points in the cli- (b5%) but showed persistent increases during the study period (Figs. 3a,
mate time-series data for 1982–2016. The significance of the change 4a). The drastic decrease observed in cultivated land between 2006 and
points was tested by using Pettitt's test as described in Section 2.6.1. 2017 was mainly the result of the farmers' growing interest in allocating
Next, we developed four scenarios for each watershed based on the more land to plantations (primarily A. decurrens) to remedy a decline in
two identified climate periods, designated period 1 and period 2 and re- soil fertility and to provide fuelwood and charcoal (Nigussie et al.,
ferring to before and after the change point, respectively, as described in 2017).
Section 3.4.1. We used the LULC maps for 1982 and 2005/06 to repre- A LULC conversion matrix was also prepared for the Kasiry water-
sent LULC conditions during the two periods. The four scenarios are Sce- shed between 1982 and 2017 (Table S3). From 1982 to 2006 in the wa-
nario 1 (SC1), 1982 LULC map and climate data for period 1; Scenario 2 tershed, the most prominent LULC change was the conversion of forest
Fig. 3. Land use/land cover maps of the Kasiry (a), Kecha (b), and Sahi (c) watersheds for three years (1982, 2005/06, and 2016/17). BL: bushland; CL: cultivated land; FL: forest land; GL:
grazing land; PL: plantation land; KC: khat cultivation; ST: settlements.
(66 ha converted) and grazing land (27.2 ha) to cultivated land. From 3.1.2. Kecha (midland) watershed
2006 to 2017, the transition to plantation (a total of 156.5 ha) from cul- LULC change maps were prepared for the Kecha watershed for 1982,
tivated land (77.5 ha), bushland (26.5 ha), and grazing land (24.8 ha) 2005, and 2016 (Fig. 3b). In this watershed, the dominant LULC type
was the dominant conversion in the watershed (Table S3). During this was cultivated land during the entire study period, covering nearly
same period, 11.7 ha of plantation area was converted to cultivated half of the watershed area (41.7% in 1982, 69.8% in 2005, and 68.9% in
land as a result of the tree–crop rotation farming practice that is becom- 2016). Bushland (22.0%) was second in dominance in 1982, followed
ing popular in the study area (Schroth and Ruf, 2014). The gross losses by forest (18.7%) and grazing land (17.6%). Khat cultivation, which
of cultivated land (91.4 ha) were higher than those of the other LULC was first detected in 2005, accounted for b1% of the total watershed
types (collectively accounting for 121.1 ha) between 2006 and 2017. area (Fig. 4b). Settlements had the least coverage compared to other
Over the entire period (1982–2017), the major LULC conversions were LULC types over the entire study period, covering b3.5%. From 1982 to
a pronounced gain in plantation land (100%) and settlements (85.3%) 2005, cultivated land increased by 67.6% (Fig. 4b); in contrast, forest,
from bushland, cultivated land, forest, and grazing land (Table S3). bushland, and grazing land decreased by 61.3%, 54.9%, and 29.9%,
Fig. 4. Area coverage (%) by different land use/land cover types (left) and the corresponding percentage change (right) from 1982 to 2016/17 in the Kasiry (a), Kecha (b), and Sahi
(c) watersheds. Abbreviations are defined in Fig. 3.
respectively. Likewise, between 2005 and 2016, bushland and grazing Between 1982 and 2005, the gain in cultivated land (140.9 ha) was
land decreased by 39.5%, and 14.8%, respectively. During this period, due to conversions of bushland (51.7 ha), forest (44.2 ha), and grazing
however, khat cultivation showed a pronounced increase of N350% as land (45.0 ha) (Table S4). This result confirms that the transition from
a result of farmers' growing interest in increasing farm income by bushland to cultivated land is more pronounced as compared to other
shifting from dominantly cereal-based farming to cash-crop farming LULC types. Similarly, between 2005 and 2016 a substantial area of graz-
(Nigussie et al., 2017). Khat cultivation is currently common practice ing land (15.0 ha) and bushland (11.8 ha) were converted to cultivated
at Kecha, and conversion to khat is desirable for both socio-economic land (32.2 ha) (Table S4). During this period a considerable area of cul-
and agro-ecological reasons (Birhane and Birhane, 2014). Also, 3% of tivated land was converted to khat cultivation (10.5 ha), and the overall
Ethiopian production originates in the Bahir Dar Zuria District percentage gain (86.8%) of khat cultivation, from all LULC types, was rel-
(Birhane and Birhane, 2014) in which Kecha watershed is located. As atively higher than the gains of other LULC types. Through the conver-
in the Kasiry watershed, bushland, forest, and grazing land in this wa- sion of LULC types from 1982 to 2016, bushland, forest, and grazing
tershed decreased by N40% over the entire study period (1982–2016). land had losses exceeding 75% (197.2 ha). In contrast, cultivated land
On the other hand, cultivated land, khat cultivation, and settlements gained 149.7 ha from other LULC types over the same period
markedly increased by N66%. (Table S4). In this watershed, the conversion of bushland and forest to
cultivated land was the predominant change throughout the entire 3.2. Surface runoff and runoff coefficient variability at plot scale
study period.
3.2.1. Surface runoff
In the 16 runoff plots in the three agro-ecological watersheds, the
3.1.3. Sahi (lowland) watershed seasonal cumulative runoff depth in the rainy seasons of 2015 and
As in the Kasiry watershed, forest (32.4%) was the dominant LULC 2016 ranged from 252 to 635 mm in Kasiry, from 219 to 454 mm in
type in the Sahi watershed in 1982, followed by grazing land (28.6%) Kecha, and from 133 to 274 mm in Sahi (Table 2). The highest runoff
and bushland (22.9%) (Figs. 3c, 4c). Compared to the other LULC types was 635 mm in the grazing land (GL) plot on a steep slope (15%) in
in this and the other two study watersheds, cultivated land (16.2%) Kasiry, and the lowest was 133 mm in the bushland (BL) plot on a
had the least percentage cover in 1982. However, in 2006 and 2017, cul- steeper slope (35%) in Sahi, both in the 2015 rainy season. The cumula-
tivated land was the most prevalent LULC type; on average it covered tive runoff across the watersheds generally decreased from highland to
more than half of the watershed area. In contrast, in 2006 and 2017 for- lowland watersheds: Kasiry (highland) N Kecha (midland) N Sahi (low-
est had the least coverage, followed by settlements (Fig. 4c). During the land). This trend is similar to the trend in cumulative rainfall amounts
entire study period (1982–2017), all LULC types except cultivated land among the watersheds (Table 2). The ranges of the runoff values during
and settlements showed a decreasing trend. Particularly notable during both rainy seasons are comparable to the plot-scale range in other stud-
this period, forest decreased by 75.2%, more than the decreases in bush- ies in the UBN basin (seasonal runoff 180–302 mm; Alemayehu et al.,
land (32%) and grazing land (48.3%). The decreased forest coverage was 2013; Amare et al., 2014; Descheemaeker et al., 2006; Ebabu et al.,
mainly due to the expansion of cultivated land in the watershed. Thus, 2018; Sultan et al., 2018). In the same basin, Haregeweyn et al. (2016)
the area under cultivation increased by 185.0% from 1982 to 2006 and found that runoff variability at basin-scale ranged from 105 mm for
by 33.2% from 2006 to 2017, or by 279.6% over the entire study period silvipastoral land to 1601 mm for water bodies and was strongly con-
(Fig. 4c). This increase is most likely a result of the resettlement policy trolled by the LULC type.
implemented in this area in the mid-1980s during the “Derg” regime The differences in runoff among watersheds were due partly to var-
(Provisional Military Government of Socialist Ethiopia). This policy re- iations in rainfall. A previous study conducted in our study watersheds
sulted in a considerable expansion of cultivated land and a decline of (Sultan et al., 2018) showed that rainfall is linearly related to, and pro-
natural vegetation cover through deforestation to meet the increasing foundly affects, the amount of runoff. Elsewhere, studies also have
demand for agricultural production. shown that rainfall had a greater impact on surface runoff compared
A LULC conversion matrix was prepared for Sahi watershed between to change in other climate variables such as CO2 and temperature
1982 and 2017 (Table S5). Between 1982 and 2006, about 100 ha of for- (Ficklin et al., 2010).The highest seasonal runoff for grazing land in
est was converted to other LULC types: to bushland (33.0 ha), cultivated Kasiry watershed is explained by the frequent use of grazing land in
land (42.7 ha), and grazing land (23.8 ha) (Table S5). During the same this watershed by livestock that trampled the soil causing soil penetra-
period, the percentage gain in cultivated land (74.0%) was higher than tion resistance to be highest (ranging from 1990 to 2210 kPa) among
the gains other LULC types; the cultivated land gain was due primarily the land-use types (Sultan et al., 2018). In addition, surface runoff was
to the conversion of grazing land (60.9 ha, 50%), followed by forest higher on steep slopes than on flat slopes; for example, the runoff for
(42.7 ha) and bushland (32.1 ha) (Table S5). However, from 2006 to cultivated land (15% slope) is higher than for cultivated land (5%
2017, the overall percentage gain in grazing land (65.7%) was more slope) in all watersheds. This likely related to a reduction in initial ab-
than that of cultivated land (38.7%), which had the least gain in this pe- straction, a decrease in infiltration, or a reduction of the recession time
riod. Over the entire period (1982–2017), cultivated land increased by of overland flow on steeper slopes.
78.2% (191 ha), of which about 31.4%, 26.1%, and 20.5% were from the
conversion of grazing land, forest, and bushland, respectively (Table S5).
In general, between 1982 and 2016/17, the area coverage of natural
vegetation (forest, grazing land, and bushland) decreased in all of the
Table 2
study watersheds (Fig. 4). In contrast, settlements and cultivated land
Measured seasonal cumulative surface runoff depth (Rcum, mm), seasonal cumulative
show an increasing trend in all watersheds during the entire study pe- rainfall depth (PC, mm), seasonal runoff coefficient (RC, %), and seasonal average runoff co-
riod. However, the expansion of cultivated land had noticeably slowed efficient (RCav, %) for different land use types in three watersheds during the 2015 and
by 2016/17 in Kecha and Kasiry (Fig. 4a, b). This slowing was due to a 2016 rainy seasons.
dynamic increase in plantation land (mainly as A. decurrens) in Kasiry Watersheds Rainy season 2015 Rainy season 2016
after 2006, and eucalyptus plantations and khat cultivation in Kecha (Jun–Oct) (Jun–Oct)
(Fig. 4b). In Sahi, the expansion of cultivated land at the expense of nat- Land use PC Rcum RC PC Rcum RC RCav (%)
ural vegetation cover (mainly bushland and grazing land) was extreme
(slope%) mm mm % mm mm % ±SD
compared to that in Kasiry and Kecha watersheds (Fig. 4a, b). This ob-
served expansion of cultivated land at the expense of forest, bush, and CL (5%) 339.3 21.6 344.5 22.9 22.3 ± 0.9
CL (15%) 444.0 28.3 501.6 33.4 30.9 ± 3.6
grazing lands is in good agreement with results of previous studies con-
GL (15%) 635.3 40.5 537.2 35.8 38.2 ± 3.3
ducted in the UBN basin and elsewhere (e.g. Gashaw et al., 2017; Kasiry BL (35%) 1567.6 370.6 23.6 1502.1 369.5 24.6 24.1 ± 0.7
Gebrehiwot et al., 2014; Minta et al., 2018; Tekle and Hedlund, 2000). AC (5%) 469.0 29.9 353.5 23.5 26.7 ± 4.5
Also, the expansion of plantations at the expense of cultivated and graz- AC (25%) 512.7 32.7 391.1 26.0 29.4 ± 4.7
ing lands agrees with reports by Yeshaneh et al. (2013) for the Koga wa- EP (25%) 339.4 21.7 251.8 16.8 19.2 ± 3.5
CL (5%) 218.8 16.1 307.7 22.0 19.1 ± 4.2
tershed, and Bewket (2002) for the Chemoga watershed in the UBN CL (15%) 1360.7 393.8 28.9 1397.7 340.1 24.3 26.6 ± 3.3
basin. Similarly, Wondie and Mekuria (2018) reported that forest Kecha
GL (15%) 453.7 33.4 414.2 29.6 31.5 ± 2.7
cover in Fageta Lekoma District (Kasiry watershed located in this Dis- BL (35%) 281.1 20.7 339.9 24.3 22.5 ± 2.5
trict) showed a substantial increase (mainly through planting of CL (5%) 186.5 21.2 251.2 19.7 20.5 ± 1.1
CL (15%) 881.2 229.2 26.0 1272.4 221.3 17.4 21.7 ± 6.1
A. decurrens) with an annual increasing trend of 5.2% from 2010 to Sahi
GL (15%) 136.2 15.4 273.5 21.5 18.5 ± 4.3
2015. On the other hand, the conversion of grazing land to cultivated BL (35%) 133.2 15.1 206.8 16.3 15.7 ± 0.8
land in the Kasiry and Sahi watersheds is consistent with the results of
CL: Cultivated land; GL: grazing land; BL: bushland; AC: Acacia decurrens; EP: eucalyptus
Dessie and Kleman (2007) in the south-central Rift Valley region of plantation; PC: seasonal cumulative rainfall; Rcum: seasonal cumulative runoff; SD: stan-
Ethiopia, and Minta et al. (2018) in Dendi-Jeldu in the central Ethiopia dard deviation from the seasonal mean. Approximately 257, 199, and 121 daily runoff
highlands. events were evaluated for Kasiry, Kecha, and Sahi watersheds, respectively.
3.2.2. Runoff coefficient Moriasi et al. (2007) and Saleh et al. (2000). Similarly, PBIAS values
The seasonal annual RC varied across and within the watersheds in ranged from 11% to 14% at daily time-steps and from 3% to 9% at
2015 and 2016 (Table 2). The average RC ranged from 19% to 38% in monthly time-steps. These values fall under “good” and “very good”
Kasiry, from 19% to 32% in Kecha, and from 15.7% to 21.7% in Sahi. performance ratings for daily and monthly time-steps, respectively
Also, grazing land exhibited higher RCs in Kasiry (38.2%) and Kecha (Van Liew et al., 2007). Generally, across the watersheds, the model per-
(31.5%) watersheds, whereas cultivated land had a higher value formed better in Kasiry watershed, followed by Sahi and then Kecha wa-
(21.7%) in the Sahi watershed. Similarly, there were higher RCs in culti- tershed, especially at monthly time-steps. These results suggest that the
vated land on steeper slopes (15%) compared to cultivated land located model is more efficient at coarser time-steps, with the implication that
on gentle slopes in all watersheds. The literature RC values for the UBN the model can simulate surface runoff at least at a “good” level for rain-
basin and elsewhere in Ethiopia typically lie in the same range, varying fall events in the three watersheds at annual time-steps.
from 10% to 40% (e.g Descheemaeker et al., 2006; Sultan et al., 2018,
2017; Zenebe et al., 2013). On the other hand, our findings on RC and 3.4. Surface hydrological response to LULC change and climate variability
slope relationships contradict with a study conducted in northern
Ethiopia by Taye et al. (2013), who explained the relationship between 3.4.1. Trends in annual rainfall and temperature time-series
RC and slope by conducting soil particle analysis: RC increased with de- The MK time-series test for monotonic trends and Pettitt's test for
creasing slope because of an increase in the coarse-particle content of homogeneous trends were applied to annual rainfall and mean annual
the soil, because a coarse texture promotes infiltration. temperature data between 1982 and 2016 (Table 5, Fig. 6). The results
The seasonal average RC from the plot experiments and weighted RC of the MK tests show no significant long-term monotonic trend in an-
of each watershed were determined for three 1-year periods (1982, nual rainfall, with Zc of 1.3, 0.9, and 0.4, for Kasiry, Kecha, and Sahi wa-
2005/06, 2016/17; Table 3). The weighted RC increased from 1982 to tersheds, respectively (Table 5, Fig. 6). Similarly, Pettitt's test showed
2016/17 in Kasiry and Sahi watersheds, whereas in Kecha the values strong homogeneity in annual rainfall in the three watersheds, indicat-
in 2005 and 2016 were comparable. This difference might be due to ing that annual rainfall did not change significantly over the study pe-
less conversion of LULC types between the respective study periods. riod (Table 5, Fig. 6). Therefore, the null hypotheses Ha0 and Hb0 for the
On the other hand, the weighted RC in the watersheds decreased in two tests for annual rainfall in the three watersheds were accepted.
the order Kasiry (from 21.9% to 27.7%) N Kecha (from 20.9 to 22.5%) Previous studies have also confirmed that there have been no signif-
N Sahi (from 15.6 to 19.2%), following the runoff volume and rainfall icant changes in annual rainfall in either the UBN basin in particular (e.g.
pattern shown in Table 2. Conway, 2000; Mekonnen et al., 2018a, 2018b) or in Ethiopia in general
(e.g. Fenta et al., 2017b).
3.3. Validation of rainfall and surface runoff In contrast, there were significant long-term monotonic and non-
homogeneous trends in mean annual temperature (P b 0.05) in all wa-
The monthly rainfall data validation results showed that data from tersheds, and the Zc values of 3.92 in Kasiry, 2.07 in Kecha and 4.55 in
nearby stations have a good agreement with the study watersheds Sahi confirmed that there were significant changes in annual tempera-
(Fig. S1). The value of coefficient of determination(R2) varies from ture over the study period (Table 5). The mean annual temperature in-
0.78 in Sahi to 0.91 in Kasiry watershed. Across the watersheds, the re- creased by 0.04 °C per year in Kasiry watershed, 0.02 °C per year in
lationship is better in Kasiry (Kasiry with Enjibara), followed by Kecha Kecha, and 0.03 °C per year in Sahi from 1982 to 2016. Such increases
(Kecha and Bahir Dar) and Sahi (Sahi with Bullen) watershed in temperature could result in a change in ET, and a change in soil mois-
(Fig. S1). Thus, the regression equations mentioned in the figure ture and runoff. The increases in temperature observed in our study wa-
(Fig. S1) were adopted to the study watersheds to estimate long-term tersheds agree with the results of other studies in the UBN basin and
rainfall data from nearby stations. elsewhere in Ethiopia that show an increase in mean annual tempera-
The computed and observed runoff was compared at daily and ture from 0.028 °C to 1.08 °C between 1980 and 2015 (e.g. Alemayehu
monthly time-steps (Fig. 5). Pairs of observed and computed values and Bewket, 2017; Birara et al., 2018; Mekonnen et al., 2018a, 2018b).
are close to a line of perfect agreement. The performance indicators Unlike for annual rainfall, the Pettitt's test applied to mean annual
NSE and R2 varied over the same range of values: from 0.7 to 0.8 for temperature showed a change point in 2001 for Kasiry and Sahi, and
the daily comparison and from 0.7 to 0.9 for the monthly (Table 4). in 1993 for Kecha (Fig. 6, right panels). On the basis of these change
These results indicate that the model performance ranged from points, the long-term time-series climatic data were divided into two
“good” to “very good” in terms of the general rating systems of periods: period 1 (1982–2001 for Kasiry and Sahi, 1982–1993 for
Kecha) and period 2 (2002–2016 for Kasiry and Sahi, 1994–2016 for
Table 3 Kecha). We did not try to separate the effects of LULC change and cli-
Seasonal and area-weighted average annual runoff coefficients (RCs, %) for the current and mate (as rainfall variability) on the estimated surface runoff change be-
past LULC types found in the three study watersheds.
cause there was no significant trend in annual rainfall in any of the study
LULC type Seasonal average RC (%) Data watersheds. Hence, any change in annual surface runoff during the
adopted from plot results source/method study period (1982–2016) attributed to the effects of LULC change. In
Period Kasiry Kecha Sahi contrast, the significant increases in temperature, and the changes in
Bushland 2015 24.1 22.5 15.7 Plot
LULC before and after the change points, might have altered the actual
Cultivated land and 26.6 22.8 21.1 Plot ET. Therefore, we analyzed the change in the actual ET due to the iso-
Forest 2016 10.0 10.0 10.0 Geiger et al. (1987) lated effects of LULC change (ΔHL as ΔETL) and climate variability (ΔHC
Grazing land 38.1 31.5 18.5 Plot as ΔETC) was analyzed by comparing periods 1 and 2. The climate data
Plantation 28.0a 22.7b Plot
(annual rainfall and temperature) and related statistics for periods 1
Settlements 60.0 60.0 60.0 CDTc (2006)
and 2 in all watersheds are summarized in Table S6.
Area-weighted RC (%) at watershed level
1982 21.9 20.9 15.6
Weighted runoff coefficient 3.4.2. Response of annual surface runoff to LULC change
2005/06 26.8 22.3 17.6 Eq. 5
(RCwt)
2016/17 27.7 22.5 19.2 Long-term (1982–2016) estimated annual surface runoff patterns
a
varied across the three watersheds under different LULC scenarios
Runoff coefficient for Acacia decurrens and eucalyptus plantations.
b
Runoff coefficient for khat cultivation estimated by taking the average of bushland and
(Fig. 7). In Sahi, the annual estimated surface runoff pattern showed
cultivated land runoff coefficients. the clearest variation among LULC scenarios (CV = 0.18) followed by
c
California Department of Transportation. Kasiry (CV = 0.12). In these watersheds, the variation in annual surface
Fig. 5. Estimated vs. observed daily (left) and monthly (right) surface runoff in Kasiry (a), Kecha (b), and Sahi (c) watersheds. Solid line is the line of perfect fit. Dotted lines indicates a
linear relationship between observed and estimated surface runoff.
runoff between LULC scenarios from 1982 and 2006 (CV of 0.15 in Sahi surface runoff derived mainly from the expansion of cultivated land,
and 0.09 in Kasiry) is relatively higher than that between 2006 and 2017 by 99.7% and 185% in Kasiry and Sahi, respectively (Fig. 4a, c). In
(CV of 0.05 in Sahi and 0.03 in Kasiry) (Fig. 7a, c). The CV of rainfall Kecha, however, there is less variation between all LULC scenarios (CV
across the watersheds remains almost the same, ranging from 0.11 in from 0.003 to 0.024) (Fig. 7b), because of the lower conversion of
Kasiry to 0.13 in Sahi. This result indicates that the variation in annual LULC types between study periods compared to the other two
watersheds.
The estimated mean annual surface runoff increased between the
Table 4
Summary of statistical criteria for examining the model accuracy in three watersheds. 1982 and the 2005/06 LULC scenarios from 475.8 to 588.2 mm (23.6%
increase) in Kasiry, from 325.0 to 339.3 mm (4.4%) in Kecha, and from
Criteria Kasiry Kecha Sahi
207.8 to 237.6 mm (14.3%) in Sahi (Table 6). Similarly, between the
Daily Monthly Daily Monthly Daily Monthly 2005/06 and 2016/17 LULC scenarios, the mean annual surface runoff
NSE 0.7 0.9 0.7 0.7 0.7 0.7 increased to 612.4 mm (4.1% higher) in Kasiry and to 255.7 mm
R2 0.8 0.9 0.7 0.8 0.8 0.9 (7.6%) in Sahi. In Kecha, however, mean annual surface runoff during
PBIAS 10.3 3.0 13.7 8.8 11.1 6.9 this period decreased slightly to 337.9 mm (0.4%; Table 6). The increases
NSE: Nash-Sutcliff Efficiency; R2: coefficient of determination; PBIAS: Percentage bias. in mean annual surface runoff likely result from a lower rate of water
Table 5 speaking, the change in mean annual surface runoff between 1982
Monotonic trend (Mann-Kendall) test and significant change (Pettitt's homogeneity) test and 2005/06 (from 4.4% to 23.6%) is higher than that between 2005/
for two climate variables (annual rainfall and mean annual temperature time series) for
1982–2016 in three watersheds.
06 and 2016/17 (from −0.4% to 7.6%) across the watersheds
(Table 6). This resulted from the dramatic expansion of cultivated land
Climate variable Watershed Mann-Kendall test Pettitt's test between 1982 and 2005/06 compared to the period between 2005/06
Zc P Ha0 K P Hb0 and 2016/17 in all watersheds (Fig. 4).
Rainfall Kasiry 1.30 0.20 A 134.00 0.20 A The change in each LULC type over the study period makes a sepa-
Kecha 0.90 0.30 A 108.00 0.60 A rate contribution to the mean annual surface runoff in the watersheds.
Sahi 0.40 0.90 A 68.00 0.40 A From 1982 to 2006, the contribution of cultivated land to the mean an-
Temperature Kasiry 3.92 b0.0001 R 272.00 b0.0001 R nual surface runoff increased from 28% to 45% in Kasiry (Fig. 8a),
Kecha 2.07 0.04 R 172.00 0.03 R
Sahi 4.55 b0.0001 R 258.00 b0.0001 R
resulting from the expansion of cultivated land by 99.7% at the expense
of other LULC types. However, the decrease in forest by 73.7%, bushland
Ha0 is the null hypothesis that there is no monotonic trend in the time series for annual
by 24.8%, and grazing land by 24.8% reduced their collective contribu-
rainfall or mean temperature; Hb0 is the null hypothesis that there is no significant change
in the time series data for annual rainfall or mean temperature (the data are homoge- tion to annual surface runoff from 68% to 41% and then to 25% in
neous). The null hypotheses are accepted (A) or rejected (R) at significance level α = 0.05. 1982, 2006, and 2017, respectively (Figs. 4a, 8a). In contrast, the expan-
sion of plantation by 261.6%, mainly at the expense of cultivated land
(Fig. 4a), resulted in an increase in surface runoff from 11% to 39% be-
loss through ET, and to continuous deterioration of soil structural qual- tween 2006 and 2016 in Kasiry. In Kecha, cultivated land was the
ities by tillage with the expansion of cultivated land at the expense of highest contributor to changes in mean annual surface runoff from
natural vegetation between 1982 and 2016/17. Comparatively 1982 (43%) to 2016 (69%) (Fig. 8b). Similarly, during this period the
Fig. 6. Trends and changes in annual rainfall (left) and mean annual temperature (right) in the Kasiry (a), Kecha (b), and Sahi (c) watersheds from 1982 to 2016. The dotted lines indicate
the Pettitt test homogeneity trend result at a significance level α = 0.05, and also showed mean value of annual rainfall (right), and temperature before and after change point (left).
was the highest contributor to mean annual surface runoff in 1982,

followed by bushland, with a combined contribution of more than half
(57%) of the mean annual surface runoff (Fig. 8c). The contribution of
grazing land and bushland decreased from 19% to 14% and from 16%
to 13% between 2006 and 2017, respectively, because of conversion to
cultivated land. The share of cultivated land in 2006 (46% of area) and
2017 (61.3%) (Fig. 4c) also contributed 55% and 68%, respectively, to
mean annual surface runoff (Fig. 8c). The overall contribution of forest
to mean annual surface runoff decreased by a factor of five from 1982
to 2017 (from 21% to 4%).
In general, across the three watersheds the contribution of cultivated
land to annual surface runoff increased from 1982 to 2016/17 because of
the increase in agricultural demand. In Kasiry in 2017 a large portion of
the contribution of cultivated land was taken over by plantation (39%)
(Fig. 8a). Even though the expansion rate of settlement area (18% per
year on average) was higher than those of other LULC types, the contri-
bution of settlements to surface runoff was relatively minor in the three
watersheds, ranging from only 1% to 7% (Figs. 4, 8).
The increase in surface runoff resulting from the expansion of culti-
vated land, at the expense of natural vegetation, because of the in-
creased agricultural demand, agrees well with other studies in the
UBN basin in Ethiopia and elsewhere (e.g. Dong et al., 2015; Gashaw
et al., 2018; Legesse et al., 2003; Ma et al., 2009; Teklay et al., 2018;
Worku et al., 2017). For instance, the mean annual surface runoff in-
creased by between 158 and 313 mm during 1984–2015 in the UBN
basin because of the increase of cultivated land at the expense of forest,
barren land, bushland, and grassland (Gashaw et al., 2018; Teklay et al.,
2018; Worku et al., 2017). Elsewhere, studies have reported that
human-induced reduction of natural vegetation cover due to the expan-
sion of cultivated land has resulted in increases in surface runoff be-
cause of the decrease in vegetation cover, which intercepts and
reduces water loss through ET (Ma et al., 2009; Bosch and Hewlett,
1982; Costa et al., 2003; Fang et al., 2013; Zhang et al., 2014). Similarly,
the expansion of plantations increased the surface runoff in Kasiry wa-
tershed because of reduced water interception due to the cleared
ground surface and the absence of understory vegetation beneath the
A. decurrens trees (Sultan et al., 2017). This agrees with the findings of
Cheng (1999), who reported that surface runoff increased from 9% to
25% as a result of the expansion of artificial forests through reforestation
activity. As a general conclusion, the spatial and temporal conversion of
LULC types had both negative and positive influences on surface runoff
responses in the three watersheds. Moreover, a substantial LULC con-
version in the form of expanded cultivated land and plantations and re-
duced natural vegetation, mainly forest, has resulted in increased
Fig. 7. Long-term patterns in estimated annual surface runoff under different LULC
scenarios (1982, 2005/06, and 2016/17) in Kasiry (a), Kecha (b), and Sahi annual surface runoff in the study watersheds.
(c) watersheds for the period from 1982 to 2016.
3.4.3. Response of actual ET to LULC change and climate variability
As for surface runoff, the annual ET under different LULC scenarios
contribution from bushland, forest, and grazing land decreased from (1982, 2005/06 and 2016/17) and the same climate data (temperature
23% to 5%, from 9% to 5%, and from 25% to 15%, respectively, as a result and rainfall) was estimated from 1982 to 2016. In relative terms, the es-
of a continuous reduction in their respective areas. In Sahi, grazing land timated ET in the three watersheds was higher using the 1982 LULC
Table 6
Mean annual surface runoff changes in three watersheds estimated under different LULC scenarios (1982, 2005/06 and 2016/17) with the same annual rainfall data from 1982 to 2016.
Watersheds LULC scenario Runoff (mm) Mean annual rainfall (mm) Change in annual runoff (ΔR)
1982–2005/06 2005/06–2016/17 1982–2016/17
mm % mm % mm %
1982 475.8
Kasiry 2006 588.2 2145.3 112.4 +23.6 24.2 +4.1 136.6 +28.7
2017 612.4
1982 325.0
Kecha 2005 339.3 1482.5 14.3 +4.4 −1.4 −0.4 12.9 +4.0
2016 337.9
1982 207.8
Sahi 2006 237.6 1339.5 29.8 +14.3 18.1 +7.6 47.9 +23.1
2017 255.7
Fig. 8. Percentage contribution of LULC types to mean estimated annual surface runoff (1982–2016) under different LULC scenarios in Kasiry (a), Kecha (b), and Sahi (c) watersheds.
Abbreviations are defined in Fig. 3.
scenario than with the 2005/06 and 2016/17 scenarios (Fig. 9). In Kecha, 1982 and 2005/06 was greater than the offsetting increases between
there was slight variation in annual ET (CV = 0.02) between 2005 and 2005/06 and 2016/17, so that the overall mean annual ET decrease
2016 LULC scenarios because there was less conversion between LULC ranged from 1% in Kasiry to 5% in Sahi (Fig. 9, Table 7). The relatively
types (Fig. 9b, Table S4). The estimated mean annual ET decreased by low reduction of mean annual ET in Kasiry resulted from less conversion
78.6 mm (6.2%), 40.1 mm (3.9%), and 33.1 mm (3.3%) in Kasiry, of vegetation cover (mainly forest and plantation) between 1982 and
Kecha, and Sahi, respectively, between the 1982 and 2005/06 LULC sce- 2017.
narios (Fig. 9, Table 7). This decrease was mainly caused by conversion The cultivated and grazing land LULC types were the major contrib-
of natural vegetation cover such as forest, bush, and grazing lands to cul- utors to the change in annual ET under all LULC change scenarios, ac-
tivated land between 1982 and 2005/06 (Tables S3–S5), resulting in less counting for 56–80% and 41–74% in Kecha and Sahi, respectively
water lost through interception by vegetation. In contrast, the mean an- (Fig. 9b, c). However, in Kasiry, except in 2006, the forest and plantation
nual ET increased by 64.7 mm (5.5%) and 8 mm (0.8%) in Kasiry and lands were the dominant contributors in all LULC scenarios; their contri-
Kecha, respectively, between 2005/06 to 2016/17 as a result of the ex- butions ranged from 46% to 52% (Fig. 9a). The contribution of bushland
pansion of plantation through reforestation activity, mainly at the ex- to annual ET decreased from the 1982 to the 2016/17 LULC scenarios:
pense of grazing and cultivated lands. However, during this period in from 15% to 5% in Kasiry, from 23% to 9% in Kecha, and from 23% to
Sahi, the continuous increase of cultivated land relative to other LULC 17% in Sahi. Similarly, except in Kasiry in 2017, in all LULC scenarios
types resulted in the decrease of mean annual ET by 20.2 mm (2.1%) the contribution of forest and plantation lands decreased considerably
(Table 7, Fig. 4). In general, the decrease in mean annual ET between from 1982 to 2016/17 (Fig. 9a). It is worth mentioning that the increases
Fig. 9. Trend in estimated annual actual evapotranspiration (ET, left), and percentage contribution of LULC types to estimated mean annual ET (1982–2016) (right) under different LULC
scenarios in Kasiry (a), Kecha (b), and Sahi (c) watersheds. FPL: forest and plantation land; BL: bushland; GCL: grazing and cultivated land; ST: settlements. No Duplicates found.
Table 7
Mean annual actual evapotranspiration changes in three watersheds under different LULC scenarios (1982, 2005/06 and 2016/17) with the same annual rainfall data from 1982 to 2016.
Watershed LULC scenario ET (mm) Change in annual actual ET (ΔET)
1982–2005/06 2005/06–2016/17 1982–2016/17
mm % mm % mm %
1982 1260.4
Kasiry 2006 1181.8 −78.6 −6.2 +64.7 +5.5 −13.9 −1.1
2017 1246.5
1982 1019.4
Kecha 2005 979.3 −40.1 −3.9 +8.0 +0.8 −32.2 −3.2
2016 987.3
1982 1012.9
Sahi 2006 979.8 −33.1 −3.3 −20.2 −2.1 −53.3 −5.3
2017 959.6
in cultivated and grazing lands due to the conversion from forest, plan- Kasiry than in the other watersheds because of the expansion of planta-
tation land, and bushland, negatively influenced the annual ET in all wa- tions there after 2006 and the increase of annual temperature between
tersheds whereas the expansion of forest and plantation land had a periods was slightly greater compared to other watersheds (Fig. 6,
positive influence on annual ET (Fig. 9). Table S6). In general, then, the differing effects of the two factors
Overall, our results indicate that the conversion of cultivated and among the three watersheds reflect the spatial heterogeneity of LULC
grazing lands to forest as plantation through reforestation had more ef- change and climate variability.
fect than the conversion of vegetation cover (as forest and plantation) to The decline and increase of annual ET as a result of LULC change and
cultivated and grazing lands on the increase of annual ET. The decline of climate variability, respectively, in the study watersheds are in agree-
annual ET in the watersheds as a result of LULC change (mainly the ex- ment with the findings of other studies in the UBN basin in Ethiopia
pansion cultivated land with the reduction of natural vegetation) is in (e.g. Gashaw et al., 2018; Woldesenbet et al., 2017) and elsewhere
agreement with other findings in the UBN basin in Ethiopia (e.g. (e.g. Li et al., 2017; Yang et al., 2017; Yang et al., 2012). For instance,
Gashaw et al., 2018; Woldesenbet et al., 2017). Elsewhere, studies Gashaw et al. (2018) stated that LULC change in the Andassa watershed
have shown that actual ET is generally greater for forest than for non- of the UBN basin, predominantly the conversion from vegetative to non-
forest (cultivated and grazing lands) LULC types, which is attributed to vegetative cover, has resulted in the reduction of annual ET. Also, the
the reduction in soil moisture in non-forest LULC types due to the loss dominant effects of LULC change found in our study watersheds are con-
of vegetation cover (Fang et al., 2013; Li et al., 2017; Ma et al., 2009; sistent with the study by Yang et al. (2012), who found that LULC
Woldesenbet et al., 2018; Yang et al., 2017; Yang et al., 2012; Zhang change has a more pronounced effect than climate change on mean an-
et al., 2001). Moreover, because of a highly developed root system and nual ET in Shalamun River (China) watersheds. The opposite was found
higher leaf area index, forests have a much higher transpiration than by Woldesenbet et al. (2017) in Ethiopia, and by Li et al. (2017) and
other vegetation types, leading to a higher annual ET value (Yang Yang et al. (2017) in China, who reported that climate variability had
et al., 2012). a greater effect than LULC change on annual ET response. It is worth
mentioning that our study captured a reduction in annual ET caused
3.4.4. Separated effects of LULC change and climate variability on actual ET by LULC change that was relatively larger in magnitude than the offset-
The combined and separate effects of LULC and climate variability on ting increase caused by climate variability, indicating that the LULC
actual ET were analyzed (Table 8). In the three watersheds, there was a change played a dominant role affecting annual ET.
remarkable reduction of mean annual ET as a result of LULC change from
1982 to 2016: a decrease of 79.0 mm (57.9%) in Kasiry, 39.1 mm (66.4%) 4. Conclusions
in Kecha, and 33.1 mm (59.4%) in Sahi. This could be attributed to a
shortage of soil moisture as a result of the expansion of cultivated land This study investigated annual surface runoff and ET responses from
at the expense of natural vegetation cover. In contrast, climate variabil- 1982 to 2016 under LULC change (comparing 1982, 2005/06, and 2016/
ity had a positive effect on the annual ET response; mean annual ET in- 17) and climate variability scenarios in Kasiry (highland), Kecha (mid-
creased by 57.4 mm (42.1%), 19.8 mm (33.6%), and 22.6 mm (40.6%) in land), and Sahi (lowland) watersheds in the Upper Blue Nile basin in
Kasiry, Kecha, and Sahi, respectively (Table 8). The variation of mean an- Ethiopia. The LULC change results revealed that from 1982 to 2016/17,
nual ET resulting from climate variability might be due mainly to the the natural vegetation cover (forest, bushland, and grazing lands) de-
significant increases in annual temperature during the study period creased by about 32–83% in Kasiry, 40–77% in Kecha, and 32%–75% in
across the watersheds (Fig. 6). The offsetting effect of LULC change on Sahi. During the same period, cultivated land cover increased by ap-
climate variability resulted in an overall decrease of annual ET by proximately 37%, 66%, and 280% in the three respective watersheds. In
21.6 mm (15.8%), 19.3 mm (32.8%), and 10.5 mm (18.8%) in Kasiry, contrast, between 2006 and 2017, plantation land increased by 262%
Kecha, and Sahi, respectively, over the study period (1982–2016). This in the Kasiry watershed, mainly at the expense of cultivated and grazing
probably resulted from the continuous expansion of cultivated land lands. Long-term annual rainfall variability was insignificant at all three
(+99.7% in Kasiry, +67.6% in Kecha, and +185.0% in Sahi from 1982 sites. On the other hand, long-term mean annual temperature showed
to 2005/06) at the expense of natural vegetation (Fig. 4). Both LULC significant (P b 0.05) variation across the three watersheds, with the
change and climate variability had a higher influence on annual ET in mean annual temperature increasing by 0.04 °C in Kasiry, 0.02 °C in
Kecha, and 0.03 °C in Sahi from 1982 to 2016.
Table 8 LULC change and climate variability caused different hydrological re-
Changes in estimated mean annual actual evapotranspiration (ET) under separate and sponses across the watersheds. LULC change positively influenced the
combined effects of LULC change and climate variability. annual surface runoff in all three watersheds. Because there was no sig-
Watersheds Scenario LULC Climate data ET (mm) Factor ET change
nificant trend in annual rainfall, this climate factor did not significantly
affect the estimated surface runoff change. LULC change, and climate
mm %
variability in terms of temperature, had negative and positive effects, re-
SC1 1982 1982–2001 1234.7 spectively, on the changes in annual ET. However, even though climate
SC2 2006 1982–2001 1158.3 ΔETL −79.0 −57.9
Kasiry variability increased ET, from 33.6% in Kecha to 42.1% in Kasiry, LULC
SC3 1982 2002–2016 1294.6 ΔETC +57.4 +42.1
SC4 2006 2002–2016 1213.1 ΔETLC −21.6 100.0 change resulting in a reduction of natural vegetation had an offsetting
SC1 1982 1982–1993 1007.0 effect that led to an overall decrease in ET, from a 15.8% reduction in
SC2 2005 1982–1993 967.0 ΔETL −39.1 −66.4 Kasiry to 32.8% in the Kecha watershed over the 35 years. In general,
Kecha
SC3 1982 1994–2016 1025.9 ΔETC +19.8 +33.6 our results indicate that the role of LULC change is more dominant
SC4 2005 1994–2016 987.7 ΔETLC −19.3 100.0
SC1 1982 1982–2001 991.7
than that of climate variability in the annual surface runoff and ET re-
SC2 2006 1982–2001 958.7 ΔETL −33.1 −59.4 sponses. These effects are mainly attribute to LULC conversion and tem-
Sahi
SC3 1982 2002–2016 1014.4 ΔETC +22.6 +40.6 perature variation across the watersheds during the study period.
SC4 2006 2002–2016 981.2 ΔETLC −10.5 100.0 Our results suggest that LULC change and climate variability can
LULC: Land use/land cover; SC1–SC4: scenario numbers 1 to 4 (please see text); ET: modify surface runoff and ET in the UBN basin, and that runoff and ET
evapotranspiration. are important hydrological components subject to change with LULC
For ET change, the change due to changes in LULC (ΔETL) = average of [(SC2 − SC1) + and climate variation. As the changes in LULC and climate variability
(SC4 − SC3)]; the change due to climate variability (ΔETC) = average of [(SC3 – SC1) +
(SC4 − SC2)]; the total change (ΔETLC) = (SC4 − SC1). Percent (%) ΔETL = ΔETL/ΔETLC
in the basin continue to increase, further significant change can be ex-
× 100 and ΔETC = ΔETC/ΔETLC × 100. The negative sign is not considered when calculating pected in the hydrological components. Thus, further investigations of
the percentage contribution of each factor to the total change. the hydrological responses under future LULC and climate variability
scenarios, including other weather parameters besides rainfall and tem- Fang, X., Ren, L., Li, Q., Zhu, Q., Shi, P., Zhu, Y., 2013. Hydrologic response to land use and
land cover changes within the context of catchment-scale spatial information.
perature, are important for devising future sustainable land and water J. Hydrol. Eng. 18 (November), 1539–1548.
management strategies. Fenta, A.A., Yasuda, H., Shimizu, K., Haregeweyn, N., 2017a. Response of streamflow to cli-
mate variability and changes in human activities in the semiarid highlands of north-
ern Ethiopia. Reg. Environ. Chang. 17 (4), 1229–1240.
Acknowledgements Fenta, A.A., Yasuda, H., Shimizu, K., Haregeweyn, N., Kawai, T., Sultan, D., Ebabu, K., Belay,
A.S., 2017b. Spatial distribution and temporal trends of rainfall and erosivity in the
Eastern Africa region. Hydrol. Process. 31 (25), 4555–4567.
This research was funded by the Science and Technology Research
Ficklin, D.L., Luo, Y., Luedeling, E., Gatzke, S.E., Zhang, M., 2010. Sensitivity of agricultural
Partnership for Sustainable Development (SATREPS), grant number runoff loads to rising levels of CO2 and climate change in the San Joaquin Valley wa-
JPMJSA1601, Japan Science and Technology Agency (JST)/Japan Interna- tershed of California. Environ. Pollut. 158 (1), 223–234.
tional Cooperation Agency (JICA). The authors thank all of the research Gao, P., Zhang, X., Mu, X., Wang, F., Li, R., Zhang, X., 2010. Trend and change-point analy-
ses of streamflow and sediment discharge in the Yellow River during 1950–2005.
assistants for providing field support. The authors also thank the Hydrol. Sci. J. 55 (2), 275–285.
Ethiopian National Meteorological Services Agency and the Arid Land Gashaw, T., Tulu, T., Argaw, M., Worqlul, A.W., 2017. Evaluation and prediction of land
Research Centre, Tottori University, for providing meteorological data use/land cover changes in the Andassa watershed, Blue Nile Basin, Ethiopia. Environ-
mental Systems Research 6 (1), 17.
and an appropriate research environment, respectively. Gashaw, T., Tulu, T., Argaw, M., Worqlul, A.W., 2018. Modeling the hydrological impacts of
land use/land cover changes in the Andassa watershed, Blue Nile Basin, Ethiopia. Sci.
Appendix A. Supplementary data Total Environ. 619–620 (November), 1394–1408.
Gebrehiwot, S.G., Bewket, W., Gärdenäs, A.I., Bishop, K., 2014. Forest cover change over
four decades in the Blue Nile Basin, Ethiopia: comparison of three watersheds. Reg.
Supplementary data to this article can be found online at https://doi. Environ. Chang. 14 (1), 253–266.
org/10.1016/j.scitotenv.2019.06.338. Gebrelibanos, T., Assen, M., 2015. Land use/land cover dynamics and their driving forces
in the Hirmi watershed and its adjacent agro-ecosystem, highlands of Northern
Ethiopia. J. Land Use Sci. 10 (1), 81–94.
References Geiger, W., Marsalek, J., Rawls, W., Zuidema, F., 1987. Manual on drainage in urbanized
areas. Planning and Design of Drainage Systems. Studies and Reports in Hydrology.
Ainsworth, E.A., Long, S.P., 2005. What have we learned from 15 years of free-air CO2 en- vol. 1. UNESCO.
richment (FACE)? A meta-analytic review of the responses of photosynthesis, canopy Guo, H., Hu, Q., Jiang, T., 2008. Annual and seasonal streamflow responses to climate and
properties and plant production to rising CO2. New Phytol. 165 (2), 351–372. land-cover changes in the Poyang Lake basin, China. J. Hydrol. 355 (1–4), 106–122.
Alemayehu, A., Bewket, W., 2017. Local spatiotemporal variability and trends in rainfall Guo, L., Wang, D., Qiu, J., Wang, L., Liu, Y., 2009. Spatio-temporal patterns of land use
and temperature in the central highlands of Ethiopia. Geografiska Annaler, Series A: change along the Bohai Rim in China during 1985–2005. J. Geogr. Sci. 19 (5),
Physical Geography 99 (2), 85–101. 568–576.
Alemayehu, M., Amede, T., Böhme, M., Peters, K.J., 2013. Collective management on com- Haregeweyn, N., Berhe, A., Tsunekawa, A., Tsubo, M., Meshesha, D.T., 2012. Integrated wa-
munal grazing lands: its impact on vegetation attributes and soil erosion in the upper tershed management as an effective approach to curb land degradation: a case study
Blue Nile basin, northwestern Ethiopia. Livest. Sci. 157 (1), 271–279. of the Enabered watershed in Northern Ethiopia. Environ. Manag. 50 (6), 1219–1233.
Amare, T., Zegeye, A.D., Yitaferu, B., Steenhuis, T.S., Hurni, H., Zeleke, G., 2014. Combined Haregeweyn, N., Tsunekawa, A., Nyssen, J., Poesen, J., Tsubo, M., Tsegaye Meshesha, D.,
effect of soil bund with biological soil and water conservation measures in the north- Schütt, B., Adgo, E., Tegegne, F., 2015a. Soil erosion and conservation in Ethiopia.
western Ethiopian highlands. Ecohydrol. Hydrobiol. 14 (3), 192–199. Prog. Phys. Geogr. 39 (6), 750–774.
Awulachew, S.B., McCartney, M., Steenhuis, T.S., Ahmed, A.A., 2008. A Review of Hydrol- Haregeweyn, N., Tesfaye, S., Tsunekawa, A., Tsubo, M., Meshesha, D., Adgo, E., Elias, A.,
ogy, Sediment and Water Resource Use in the Blue Nile Basin (IWMI Working 2015b. Dynamics of land use and land cover and its effects on hydrologic responses:
Paper 131). vol. 131. case study of the Gilgel Tekeze catchment in the highlands of Northern Ethiopia. En-
Bewket, W., 2002. Land cover dynamics since the 1950s in Chemoga watershed, Blue Nile viron. Monit. Assess. 187, 4090.
basin, Ethiopia. Mt. Res. Dev. 22 (3), 263–269. Haregeweyn, N., Tsunekawa, A., Tsubo, M., Meshesha, D., Adgo, E., Poesen, J., Schütt, B.,
Birara, H., Pandey, R., Mishra, K., 2018. Trend and variability analysis of rainfall and tem- 2016. Analyzing the hydrologic effects of region-wide land and water development
perature in the Tana basin region, Ethiopia. Journal of Water and Climate Change 9 interventions: a case study of the Upper Blue Nile basin. Reg. Environ. Chang. 16
(3), 555–568. (4), 951–966.
Birhane, B.W., Birhane, M.W., 2014. The effect of Khat (Catha edulis) chewing on blood Haregeweyn, N., Tsunekawa, A., Poesen, J., Tsubo, M., Meshesha, D.T., Fenta, A.A., Nyssen,
pressure among male adult Chewers, Bahir Dar, Northwest Ethiopia. Science 2 (5), J., Adgo, E., 2017. Comprehensive assessment of soil erosion risk for better land use
461–468. planning in river basins: case study of the Upper Blue Nile River. Sci. Total Environ.
Bosch, J.M., Hewlett, J.D., 1982. A review of catchment experiments to determine the ef- 574, 95–108.
fect of vegetation changes on water yield and evapotranspiration. J. Hydrol. 55 Hargreaves, G.H., Samani, Z.A., 1985. Reference crop evapotranspiration from tempera-
(1–4), 3–23. ture. Appl. Eng. Agric. 1 (2), 96–99.
Burn, D.H., 1994. Hydrologic effects of climatic change in west-central Canada. J. Hydrol. Hurni, H., Wolde-Aregay, B., Chadhokar, P., Goshu, K., Zeleke, G., Danano, D., Grunder, M.,
160 (1–4), 53–70. 2016. Soil and water conservation in Ethiopia: guidlines for development agents.
Chen, Y., Takeuchi, K., Xu, C., Chen, Y., Xu, Z., 2006. Regional climate change and its effects Centre for Development and Environment (CDE). vol. 1. Bern Open Publishing
on river runoff in the Tarim Basin, China. Hydrol. Process. 20 (10), 2207–2216. (BOP), Bern, Switzerland.
Cheng, G., 1999. Yangtze change : hydrological effects in the Upper River Valley. Ambio 28 IPCC, 2013. In Stocker, T.F., Qin, D., Plattner, G.K., Tignor, M., Allen, S.K., Boschung, J.,
(5), 457–459. Nauels, A., Xia, Y., Bex, V. and Midgley. Climate change 2013. The Physical Science
Congalton, R.G., Green, K., 2009. Assessing the Accuracy of Remotely Sensed Data: Princi- Basis, Cambridge University Press, Cambridge.
ples and Practices. 2nd edition. Taylor & Francis Group, United States of America, Kahya, E., Kalayci, S., 2004. Trend analysis of streamflow in Turkey. J. Hydrol. 289 (1–4),
p. 189. 128–144.
Conway, D., 1997. A water balance model of the Upper Blue Nile in Ethiopia. Hydrol. Sci. J. Kates, R.W., Torrie, R.D., 1998. Global change in local places: how scale matters. Environ-
42 (2), 265–286. ment 40 (2), 5–8.
Conway, D., 2000. The climate and hydrology of the Upper Blue Nile river. Geogr. J. 166 Legesse, D., Vallet-Coulomb, C., Gasse, F., 2003. Hydrological response of a catchment to
(1), 49–62. climate and land use changes in tropical Africa: case study south central Ethiopia.
Costa, M.H., Botta, A., Cardille, J.A., 2003. Effects of large-scale changes in land cover on the J. Hydrol. 275 (1–2), 67–85.
discharge of the Tocantins River, Southeastern Amazonia. J. Hydrol. 283 (1–4), Li, G., Zhang, F., Jing, Y., Liu, Y., Sun, G., 2017. Response of evapotranspiration to changes in
206–217. land use and land cover and climate in China during 2001–2013. Sci. Total Environ.
Descheemaeker, K., Nyssen, J., Poesen, J., Raes, D., Haile, M., Muys, B., Deckers, S., 2006. 596–597, 256–265.
Runoff on slopes with restoring vegetation: a case study from the Tigray highlands, Liu, S., Chen, M., Zhuang, Q., 2016. Direct radiative effects of tropospheric aerosols on
Ethiopia. J. Hydrol. 331 (1–2), 219–241. changes of global surface soil moisture. Clim. Chang. 136 (2), 175–187.
Dessie, G., Kleman, J., 2007. Pattern and magnitude of deforestation in the South Central Long, H., Liu, Y., Wu, X., Dong, G., 2009. Spatio-temporal dynamic patterns of farmland
Rift Valley Region of Ethiopia. Mt. Res. Dev. 27 (2), 162–168. and rural settlements in Su-Xi-Chang region: implications for building a new coun-
Dile, Y.T., Tekleab, S., Ayana, E.K., Gebrehiwot, S.G., Worqlul, A.W., Bayabil, H.K., tryside in coastal China. Land Use Policy 26 (2), 322–333.
Srinivasan, R., 2018. Advances in water resources research in the Upper Blue Nile Ma, X., Xu, J., Luo, Y., Aggarwal, S.P., Li, J., 2009. Response of hydrological processes to
basin and the way forward: a review. J. Hydrol. 560, 407–423. land-cover and climate changes in Kejie watershed, south-west China. Hydrol. Pro-
Dong, L., Xiong, L., Lall, U., Wang, J., 2015. The effects of land use change and precipitation cess. 23 (8), 1179–1191.
change on direct runoff in Wei River watershed, China. Water Sci. Technol. 71 (2), Maidment, D.R., Olivera, F., Calver, A., Eatherall, A., Fraczek, W., 1996. Unit hydrograph de-
289–295. rived from a spatially distributed velocity field. Hydrol. Process. 10 (6), 831–844.
Ebabu, K., Tsunekawa, A., Haregeweyn, N., Adgo, E., Meshesha, D.T., Aklog, D., Masunaga, Maitima, J.M., Mugatha, S.M., Reid, R.S., Gachimbi, L.N., Majule, A., Lyaruu, H.,
T., Tsubo, M., Sultan, D., Fenta, A.A., Yibeltal, M., 2018. Effects of land use and sustain- Pomery, D., Mathai, S., Mugisha, S., 2009. The linkages between land use
able land management practices on runoff and soil loss in the Upper Blue Nile basin, change, land degradation and biodiversity across East Africa. Afr. J. Environ.
Ethiopia. Sci. Total Environ. 648 (August 2018), 1462–1475. Sci. Technol. 3 (10), 310–325.
Mekonnen, G., 2018. Soil Characterization, Classification and Mapping of Three Water- Teklay, A., Dile, Y.T., Setegn, S.G., Demissie, S.S., Asfaw, D.H., 2018. Evaluation of static and
sheds in the Upper Blue Nile Basin, Ethiopia. Amhara Design and Supervision dynamic land use data for watershed hydrologic process simulation: a case study in
Works Enterprise, Bahir Dar, Ethiopia. Gummara watershed, Ethiopia. Catena 172 (August 2018), 65–75.
Mekonnen, D.F., Duan, Z., Rientjes, T., Disse, M., 2018a. Analysis of combined and isolated Tekle, K., Hedlund, L., 2000. Land cover changes between 1958 and 1986 in Kalu District,
effects of land-use and land-cover changes and climate change on the upper Blue Nile Southern Wello, Ethiopia. Mt. Res. Dev. 20 (1), 42–51.
River basin's streamflow. Hydrol. Earth Syst. Sci. 22 (12), 6187–6207. Tekleab, S., Wenninger, J., Uhlenbrook, S., 2014. Characterisation of stable isotopes to
Mekonnen, Z., Kassa, H., Woldeamanuel, T., Asfaw, Z., 2018b. Analysis of observed and identify residence times and runoff components in two meso-scale catchments in
perceived climate change and variability in Arsi Negele District, Ethiopia. Dev. Sus- the Abay/Upper Blue Nile basin, Ethiopia. Hydrol. Earth Syst. Sci. 18 (6), 2415–2431.
tain. 20 (3), 1191–1212. Van Liew, M.W., Veith, T.L., Bosch, D.D., Arnold, J.G., 2007. Suitability of SWAT for the con-
Minta, M., Kibret, K., Thorne, P., Nigussie, T., Nigatu, L., 2018. Land use and land cover dy- servation effects assessment project: comparison on USDA Agricultural Research Ser-
namics in Dendi-Jeldu hilly-mountainous areas in the central Ethiopian highlands. vice Watersheds. J. Hydrol. Eng. 12 (2), 173–189.
Geoderma 314 (September 2017), 27–36. Westmacott, J.R., Burn, D.H., 1997. Climate change effects on the hydrologic regime within
Moriasi, D.N., Arnold, J.G., Van Liew, M.W., Bingner, R.L., Harmel, R.D., Veith, T.L., 2007. the Churchill-Nelson River Basin. J. Hydrol. 202 (1–4), 263–279.
Model evaluation guidelines for systematic quantification of accuracy in watershed Woldesenbet, T.A., Elagib, N.A., Ribbe, L., Heinrich, J., 2017. Hydrological responses to land
simulations. Trans. ASABE 50 (3), 885–900. use/cover changes in the source region of the Upper Blue Nile Basin, Ethiopia. Sci.
Nigussie, Z., Tsunekawa, A., Haregeweyn, N., Adgo, E., Nohmi, M., Tsubo, M., Aklog, D., Total Environ. 575, 724–741.
Meshesha, D.T., Abele, S., 2017. Factors affecting small-scale farmers' land allocation Woldesenbet, T.A., Elagib, N.A., Ribbe, L., Heinrich, J., 2018. Catchment response to climate
and tree density decisions in an acacia decurrens-based taungya system in Fagita and land use changes in the Upper Blue Nile sub-basins, Ethiopia. Sci. Total Environ.
Lekoma District, North-Western Ethiopia. Small-Scale Forestry 16 (2), 219–233. 644, 193–206.
Nyssen, J., Poesen, J., Haile, M., Moeyersons, J., Deckers, J., Hurni, H., 2009. Effects of land Wondie, M., Mekuria, W., 2018. Planting of Acacia decurrens and dynamics of land cover
use and land cover on sheet and rill erosion rates in the Tigray highlands, Ethiopia. change in Fagita Lekoma District in the Northwestern Highlands of Ethiopia. Mt.
Z. Geomorphol. 53 (2), 171–197. Res. Dev. 38 (3), 230–239.
Nyssen, J., Clymans, W., Descheemaeker, K., Poesen, J., Vandecasteele, I., Vanmaercke, M., Worku, T., Khare, D., Tripathi, S.K., 2017. Modeling runoff–sediment response to land use/
Walraevens, K., 2010. Impact of soil and water conservation measures on catchment land cover changes using integrated GIS and SWAT model in the Beressa watershed.
hydrological response-a case in north Ethiopia. Hydrol. Process. 24 (13), 1880–1895. Environ. Earth Sci. 76 (16), 1–14.
Pettitt, 1979. A non-parametric to the approach problem. Appl. Stat. 28 (2), 126–135. Yang, X., Ren, L., Singh, V.P., Liu, X., Yuan, F., Jiang, S., Yong, B., 2012. Impacts of land use
Saleh, A., Arnold, J.G., Gassman, P.W., Hauck, L.M., Rosenthal, W.D., Williams, J.R., and land cover changes on evapotranspiration and runoff at Shalamulun River water-
McFarland, A.M.S., 2000. Application of swat for the Upper North Bosque River Wa- shed, China. Hydrol. Res. 43 (1–2), 23.
tershed. Transactions of the ASAE 43 (5), 1077–1087. Yang, L., Feng, Q., Yin, Z., Wen, X., Si, J., Li, C., Deo, R.C., 2017. Identifying separate impacts
Savenije, H.H.G., 2009. HESS opinions “the art of hydrology”. Hydrol. Earth Syst. Sci. 13 of climate and land use/cover change on hydrological processes in upper stream of
(2), 157–161. Heihe River, Northwest China. Hydrol. Process. 31 (5), 1100–1112.
Schroth, G., Ruf, F., 2014. Farmer strategies for tree crop diversification in the humid tro- Yeshaneh, E., Wagner, W., Exner-Kittridge, M., Legesse, D., Blöschl, G., 2013. Identifying
pics. A review. Agron. Sustain. Dev. 34 (1), 139–154. land use/cover dynamics in the Koga Catchment, Ethiopia, from multi-scale data,
Sultan, D., Tsunekawa, A., Haregeweyn, N., Adgo, E., Tsubo, M., Meshesha, D.T., Masunaga, and implications for environmental change. ISPRS International Journal of Geo-
T., Aklog, D., Ebabu, K., 2017. Analyzing the runoff response to soil and water conser- Information 2 (2), 302–323.
vation measures in a tropical humid Ethiopian highland. Phys. Geogr. 38 (5), Yin, J., He, F., Jiu Xiong, Y., Yu Qiu, G., 2017. Effects of land use/land cover and climate
423–447. changes on surface runoff in a semi-humid and semi-arid transition zone in north-
Sultan, D., Tsunekawa, A., Haregeweyn, N., Adgo, E., Tsubo, M., Meshesha, D.T., Masunaga, west China. Hydrol. Earth Syst. Sci. 21 (1), 183–196.
T., Aklog, D., Fenta, A.A., Ebabu, K., 2018. Efficiency of soil and water conservation Zeleke, G., Hurni, H., 2001. Implications of land use and land cover dynamics for mountain
practices in different agro-ecological environments in the Upper Blue Nile Basin of resource degradation in the Northwestern Ethiopian Highlands. Mt. Res. Dev. 21 (1),
Ethiopia. Journal of Arid Land 10 (2), 249–263. 184–191.
Sun, G., McNulty, S.G., Lu, J., Amatya, D.M., Liang, Y., Kolka, R.K., 2005. Regional annual Zenebe, A., Vanmaercke, M., Poesen, J., Verstraeten, G., Haregeweyn, N., Haile, M., Amare,
water yield from forest lands and its response to potential deforestation across the K., Deckers, J., Nyssen, J., 2013. Spatial and temporal variability of river flows in the
southeastern United States. J. Hydrol. 308 (1–4), 258–268. degraded semi-arid tropical mountains of northern Ethiopia. Z. Geomorphol. 57 (2),
Sun, G., Zhou, G., Zhang, Z., Wei, X., McNulty, S.G., Vose, J.M., 2006. Potential water yield 143–169.
reduction due to forestation across China. J. Hydrol. 328 (3–4), 548–558. Zhang, L., Dawes, W.R., Walker, G.R., 2001. Response of mean annual evapotranspiration
Taye, G., Poesen, J., Wesemael, B. Van, Vanmaercke, M., Teka, D., Deckers, J., Goosse, T., to vegetation changes at catchment scale. Water Resour. Res. 37 (3), 701–708.
Maetens, W., Nyssen, J., Hallet, V., Haregeweyn, N., 2013. Effects of land use, slope gra- Zhang, Y., Guan, D., Jin, C., Wang, A., Wu, J., Yuan, F., 2014. Impacts of climate change and
dient, and soil and water conservation structures on runoff and soil loss in semi-arid land use change on runoff of forest catchment in northeast China. Hydrol. Process. 28
Northern Ethiopia. Phys. Geogr. 34 (3), 236–259. (2), 186–196.

Science of The Total Environment: Contents Lists Available at

Uploaded by

Copyright:

Available Formats

Science of The Total Environment: Contents Lists Available at

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Science of The Total Environment: Contents Lists Available at

Uploaded by

Copyright:

Available Formats

Science of the Total Environment 689 (2019) 347–365

Contents lists available at ScienceDirect

Science of the Total Environment

journal homepage: www.elsevier.com/locate/scitotenv

Hydrological responses to land use/land cover change and climate

• Land use/land cover change and climate

Characteristic Kasiry (highland) Kecha (midland) Sahi (lowland)

Area (ha) 397.5 424.3 398.2

2.2. Data types and sources

2.2.1. LULC and topography data

2.2.2. Temperature, rainfall, streamﬂow, and runoff data

R2, of linear regression curves (Moriasi et al., 2007), Nash-Sutcliffe Efﬁ-

was the highest contributor to mean annual surface runoff in 1982,

1982–2005/06 2005/06–2016/17 1982–2016/17

Watershed LULC scenario ET (mm) Change in annual actual ET (ΔET)

1982–2005/06 2005/06–2016/17 1982–2016/17

You might also like