This is a repository copy of Partitioning the impacts of land use/land cover change and

climate variability on water supply over the source region of the Blue Nile Basin.

White Rose Research Online URL for this paper:

https://eprints.whiterose.ac.uk/158107/

Version: Accepted Version

Article:
Belete, M, Deng, J, Abubakar, GA et al. (6 more authors) (2020) Partitioning the impacts of
land use/land cover change and climate variability on water supply over the source region
of the Blue Nile Basin. Land Degradation & Development, 31 (15). pp. 2168-2184. ISSN
1085-3278

https://doi.org/10.1002/ldr.3589

© 2020 John Wiley & Sons, Ltd. This is the peer reviewed version of the following article:
Belete, M, Deng, J, Abubakar, GA et al. (6 more authors) (2020) Partitioning the impacts of
land use/land cover change and climate variability on water supply over the source region
of the Blue Nile Basin. Land Degradation & Development, 31 (15). pp. 2168-2184. ISSN
1085-3278, which has been published in final form at https://doi.org/10.1002/ldr.3589. This
article may be used for non-commercial purposes in accordance with Wiley Terms and
Conditions for Use of Self-Archived Versions. This article may not be enhanced, enriched
or otherwise transformed into a derivative work, without express permission from Wiley or
by statutory rights under applicable legislation. Copyright notices must not be removed,
obscured or modified. The article must be linked to Wiley’s version of record on Wiley
Online Library and any embedding, framing or otherwise making available the article or
pages thereof by third parties from platforms, services and websites other than Wiley
Online Library must be prohibited. Uploaded in accordance with the publisher's self-
Reuse
archiving
Items policy.
deposited in White Rose Research Online are protected by copyright, with all rights reserved unless
indicated otherwise. They may be downloaded and/or printed for private study, or other acts as permitted by
national copyright laws. The publisher or other rights holders may allow further reproduction and re-use of
the full text version. This is indicated by the licence information on the White Rose Research Online record
for the item.

Takedown
If you consider content in White Rose Research Online to be in breach of UK law, please notify us by
emailing [email protected] including the URL of the record and the reason for the withdrawal request.

[email protected]
https://eprints.whiterose.ac.uk/
 Partitioning the Impacts of Land Use/Land Cover change and Climate
Variability on Water Supply over the Source Region of Blue Nile Basin

Marye Belete 1,2, Jinsong Deng 1, *, Ghali Abdullahi Abubakar 1, Menberu Teshome 3, Ke Wang 1,
Muluneh Woldetsadik 4, Enyan Zhu 1, Alexis Comber 5, and Adam Gudo1
1 College of Environmental and Resource Sciences, Zhejiang University, Hangzhou 310058, China;
[email protected] (M.B.); [email protected] (G.A.A.); [email protected] (K.W.);
[email protected] (E. Z.); [email protected]
2 Department of Natural Resource Management, Debre Tabor University, Debre Tabor 272, Ethiopia.
Email: [email protected]
3 Department of Geography and Environmental Studies, Debre Tabor University, Debre Tabor 272,
Ethiopia. Email: [email protected] (M.T.)
4
Department of Geography and Environmental Studies, Addis Ababa University, Addis Ababa 1176,
Ethiopia. Email: [email protected] (M.W.)
5
School of Geography, University of Leeds, Leeds, LS2 9JT, UK. Email: [email protected] (A.C.)

* Correspondence: [email protected]; Tel.: +86-571-8898-2623

Abstract
Water plays a vital role in sustaining the natural functioning of the entire ecosystem that supports
life on earth. It plays keys role in the wellbeing of society in numerous ways. However, climate
variability and land use land cover (LULC) change have caused spatiotemporal water supply
variation. Disentangling the effects of climate variability from LULC change on water supply is
crucial for sustainable water resource management. The main purpose of this study is, therefore,
to disentangle the relative contribution of LULC change and climate variability to the overall
average annual water supply variation. Residual Trends analysis combined with Integrated
Valuation of Environmental Services and Tradeoffs (InVEST) annual water yield model were
adopted to perform simulations and disentangle the relative impacts of climate variability and
LULC change. Ground and satellite data were used in this study. The study area has experienced
a significant increasing wetness trend and significant LULC dynamics between 2003 and 2017.
As a result, an increasing water supply was observed due to the joint effects of climate variability
and LULC change in the watershed (203 mm). The contribution of climate variability was 94%
whereas LULC contributes only 6% from 2003 to 2017. Climate variability negatively led to water
supply variation while LULC change contributed positively from 2010 to 2017. Although the
ongoing soil and water conservation (SWC) practices improved vegetation cover and water
retention of the watershed, climate variability is the main driver of water supply variation.
Therefore, SWC practices should incorporate ecosystem-based climate change adaptation
strategies and scale up to community-based integrated watershed management to sustain water
supply.

Keywords: water yield – LULC change -climate variability –impact –InVEST

1. INTRODUCTION
Water provisioning is a major area of interest within the field ecosystems because it is one of the
essential ecosystem services, playing keys role in the wellbeing of society in numerous ways,
including human consumption, agriculture, fisheries, industry, recreation, and energy production
(Pessacg et al., 2015; Sahle et al., 2019). However, the growing level of human pressure on
ecosystems resulted in human-induced threats to water ecosystem services. Land Use Land Cover
(LULC) and climate changes are the main direct drivers of water supply variation (MEA, 2005;
Soytong et al., 2016) as well as many other human-induced threats. As a result, freshwater supply
has become a progressively scarce natural resource.

Several attempts have been made to investigate the impact of climate variability and LULC change
on water supply variation during the last two decades (Krysanova & White, 2015). However,
conflicting results were reported about the cause of water yield variation (Pan et al., 2015). Some
studies have reported that climate change in general and rainfall variability, in particular, is the
primary cause of water yield variation (Ayele et al., 2016; Dile et al., 2013; Zhang et al., 2016).
Whereas, other studies have reported that LULC change is the primary driving factor of water
yield dynamics (Arunyawat & Shrestha, 2016; Chakilu & Moges, 2017; Feng et al., 2012; Teklaya
et al., 2019). Soytong et al. (2016) strongly argued that both climate variability and LULC change
are responsible for water yield dynamics.

Despite substantial progress has been made to investigate the impacts of LULC and climate
changes on water supply variation, research to date has tended to focus on either climate change
or LULC change alone. For instance, Ayele et al. (2016) showed that water supply increased
substantially due to climate change. Whereas, Chakilu & Moges (2017) found that LULC change
to be the major factor of water supply change by using the Soil and Water Assessment Tool
(SWAT) model. However, an in-depth understanding of the isolated and integrated impacts of
climate variability and LULC change on water supply is crucial for optimum water supply
management (Chawla & Mujumdar, 2015). In the absence of such information, the positive
impacts of climate change or variability can mask the negative impacts of LULC change and vice
versa. Therefore, partitioning the relative contribution of LULC and climate changes to the overall
spatiotemporal water supply variation is vital for sustaining water provision service of the
ecosystem through effective land management and climate change adaptation (Pan et al., 2015).
However, little has been done to isolate the impact of climate variability and LULC change due to
lack of effective methods to investigate the combined and isolated impacts of climate and land-use
changes (Pan et al., 2015).

Although four methods (statistical techniques such as Mann–Kendall test, experimental paired
catchment approach, distributed process-based hydrological models and empirical or conceptual
models) used to study the impact of climate variability and LULC change, process-based
hydrological models are the most effective method to disentangling the effects of climate
variability from LULC change on water supply (Praskievicz & Chang, 2009; Sharp et al., 2018).
To date, there are several process-based hydrological models in the public domain developed in
America and Europe. However, most process-based hydrological models are data intensive.
Therefore, the application of such models in data-scarce areas has resulted in poor model outputs.
For example, Van Griensven et al. (2012) found several drawbacks of the SWAT model when
applied to the source region of the Nile River including different results from several studies of
the same study catchments. These drawbacks are attributed to a lack of ground data for model
calibration and result validation in Upper Nile sub-basins. However, the introduction of simple,
easy to use, and low input data requirements for models like the Integrated Valuation of
Environmental Services and Tradeoffs (InVEST) annual water yield model, makes accurate
hydrological modeling possible in data-scarce regions (Belete et al., 2018; Hamel & Guswa, 2015).
Hence, water supply change can be quantified using the InVEST annual water yield model in data-
scare regions (Sharp et al., 2018; Vogl et al., 2015).

The streamflow data collected from Ribb and Gummara hydrological stations indicate that the
average annual water supply of the Ribb-Gummara watershed exhibits an increasing trend in recent
years. Government reports show that this phenomenon is mainly because of the ongoing soil and
water conservation (SWC) practices that started in 2010. Although there is no literature, expert in
the area believe that rainfall is the primary factor for the increasing trend of water supply. What is
not yet clear is the relative and combined impact of climate variability and land-use change on
water yield dynamics. The relative impacts of climate variability and land-use change on water
dynamics are vital for water and land resource management (Kremen, 2005; Lü et al., 2012). This
study, therefore, sets out to investigate the combined and isolated impacts of LULC change and
climate variability on water supply by combining InVEST and Residual Trends analysis. The
objectives of this study are to: (i) investigate LULC change and climate variability trends, (ii)
quantify and map spatial and temporal changes of water supply and (iii) segregate the impact of
LULC change and climate variability on the average annual water supply.

2. MATERIAL AND METHODS

2.1. Study Site Description
The Ribb-Gummara watershed, drained by Ribb (130 km long) and Gummara (72 km long) Rivers,
is a watershed in the Blue Nile Sub Basin of the Nile River source area (Ayenew, 2008; Mulatu et
al., 2018). It stretches from Mount Guna to Lake Tana in the Amhara Regional State of Ethiopia
within 37.5395o-38.2474o longitude and 11.5652o-12.2389o latitude. The watershed covers a total
area of 3273 km2 with an average elevation of 2,271 m a.s.l (meter above sea level). The elevation
of the watershed ranges from 4109 m a.s.l at Mount Guna to 1783 m a.s.l at the mouths of Ribb
and Gummara Rivers (Figure 1). Higher elevation variation in the watershed results in diverse
climate, soil and vegetation cover.

The Ribb-Gummara watershed contains a wide range of climatic and landform attributes. Among
the five agro-ecological zones of the world, three of them are found in this study site. These are
sub-tropical, temperate and alpine agro-ecological zones. The upstream part of the watershed has
a temperate climate except for the top of Guna Mountain whereas the downstream portion is sub-
tropical with a distinct summer season. The mean annual temperature and rainfall were 18.78oc
and 1334.55 mm respectively between 2001 and 2017. This study site is dominated by the Fogera
plain and Guna Mountain.

According to Worqlul et al. (2015), the major soils of the Ribb-Gummara watershed are Haplic
and Chromic Luvisols as well as Eutric Fluvisols and Eutric Leptosols. Ribb sub-watershed is
covered by Eutric Fluvisols (23.9%), Eutric Leptosols (36.2%) and, Chromic Luvisols (39.7%)
having stony and gravelly characteristics. The Gummara sub-watershed, on the other hand, is
dominated mainly by Haplic Luvisols (63.4%) characterized by high clay content and Chromic
Luvisols (24.4%).

Agricultural land, forest, shrubs, grazing land, built-up area, water bodies, and wasteland are the
well-known land use and land cover types in Ribb-Gummara watershed. In 2017, nearly three
fourth of the study area was covered with food crops (73.61%). The remaining one fourth of the
area were covered with shrubland (13.53%), grassland (8.84), forestland (2.72%), and built-up
area (0.63%). Bare land (0.08%) and water bodies (0.59%) are insignificant components of the
study watershed land cover.

The Ribb-Gummara watershed has Local, National and Regional implications. This watershed
together with Gilgel Abay and Megech watersheds in Tana Basin, contribute 93% of the total
inflow of the Lake Tana which is the primary source of Blue Nile (Setegn et al., 2009). The Blue
Nile is the dominant water source of the downstream Nile riparian countries before it ends up at
the Mediterranean Sea (80 – 85 %) (Easton et al., 2010). The survival of the Grand Renaissance
Dam being constructed in Ethiopia at the Ethio-Sudan border is mostly dependent on the healthy
functioning of Ribb-Gummara ecosystem. The livelihoods of hundreds of thousands of people in
the watershed are entirely reliant on rain-fed agriculture. However, land degradation has severely
affected the land resources in the area due to intensive agriculture, free grazing, and population
pressure (Moges & Bhat, 2017).

2.2. Partitioning LULC Change and Climate Variability Impacts on Water Supply Variation
We used the biophysical component of InVEST annual time scale water yield simulation to
calculate the annual water supply of the study watershed. The model was developed by Standford
University under the Natural Capital Project with the intention of it running on any standard
computer using relatively simple, readily available, and limited input data (Sharp et al., 2018).
According to Hamel and Guswa (2015), the InVEST model has several advantages : limited input
data requirement, ease of use, and spatially explicit output data. A more detailed description of the
model is found in the InVEST user’s guide (Sharp et al., 2018).

The annual water supply capacity of each pixel in a watershed (x) is calculated as follows:

𝐴𝐸𝑇(𝑥)
𝑦(𝑥) = (1 − ) × 𝑝(𝑥) (1)
𝑝(𝑥)

Where (𝑥) and 𝐴𝐸𝑇(𝑥) are the annual precipitation and the annual actual evapotranspiration on
pixel 𝑥 respectively.

However, AET is not easily measurable. To overcome this problem, 𝐴𝐸𝑇(𝑥) can be calculated
following Fu (1981) and Zhang et al. (2004) based on the Budyko curve:

1⁄𝜔
𝐴𝐸𝑇(𝑥) 𝑃𝐸𝑇(𝑥) 𝑃𝐸𝑇(𝑥) 𝜔
𝑃(𝑥)
= 1+ − [1 + ( ) ] (2)
𝑃(𝑥) 𝑃(𝑥)

Where, 𝜔(𝑥) describes climatic-soil properties as a non-physical parameter where as 𝑃𝐸𝑇(𝑥) is the
potential evapotranspiration, both are described in detail below.

𝑃𝐸(𝑥) can be calculated as follows:

𝑃𝐸𝑇(𝑥) = 𝐾𝑐(𝑙𝑥) × 𝐸𝑇𝑜(𝑥) (3)

Where ETo is reference evapotranspiration, and kc is a crop coefficient of each pixel.

The empirical parameter 𝜔(𝑥) is also calculated following Donohue et al. (2012) as follows:
 𝐴𝑊𝐶(𝑥)
𝜔(𝑥) = 𝑍 + 1.25 (4)
𝑃(𝑥)

Where Z is Zhang constant and AWC is plant available water content.

In order to differentiate the relative impacts of LULC change and climate variability on annual
water production in the Ribb-Gummara watershed, we used Residual Trends analysis (Evans &
Geerken, 2004; Pan et al., 2015; Wessels et al., 2007). Prior to commencing the water supply
estimation, two scenarios were developed: (a) water supply under land-use change with constant
climate and (b) water supply under climate variability with constant land use. Using the calibrated
and validated annual water yield model of InVEST, we were able to calculate the annual water
supply of the study area at three different periods during the past fifteen years (2003, 2010 and
2017). After obtaining the actual water supply, we calculated the water supply variation for three
periods of the study (2003-2010, 2010-2017 and 2003-2017). Following this, the water yield
simulation was undertaken for the two scenarios by fixing the land use and climate input
parameters alternatively using InVEST water yield model. The Residual Trends were then
calculated by subtracting the scenarios trends from the actual trends. In the end, the relative
impacts of LULC change and climate variability to water yield variations were calculated. The
procedures applied are detailed below:

∆𝑌 = 𝑌𝐹 − 𝑌𝐵 (5)

Where, ∆𝑌 is the total water yield variation; 𝑌𝐹 is the water production at the end of each period;
and 𝑌𝐵 is the water production at the beginning of each period.

∆𝑌𝐿 = 𝑌𝐹 − 𝑌𝐹𝐶 (6)

Where, ∆𝑌𝐿 is the land use change impact on water production; and 𝑌𝐹𝐶 is the water yield under
only climate variability.

∆𝑌𝐶 = 𝑌𝐹 − 𝑌𝐹𝐿 (7)

Where, ∆𝑌𝐶 is the climate variability impact on water production; and 𝑌𝐹𝐿 is the water yield under
only land use change.

Δ𝑌𝐿
𝜂𝐿 = Δ𝑌
× 100 (8)
Where, 𝜂𝐿 is the relative impact of land use change on water supply.

Δ𝑌𝐶
𝜂𝐶 = × 100 (9)
Δ𝑌

Where, 𝜂𝐶 is the relative impact of climate variability on water supply.

Nine tabular and spatially explicit input data are required to run InVEST annual water yield
simulation. These are detailed in section 2.3, and summarized in Table 1.

2.3. Data Sources and Processing

2.3.1 Field data
Field observation and interviews were used to collect field data in the study area. Field observation
and key informant interviews were carried out from June to September 2017 to obtain general
insight about the land use and to collect ground truth data for training and validation. Interviews
were conducted with twelve local elderly people, three agricultural experts, and one government
official. Ancillary field observation was also undertaken between May and September 2018. Based
on the knowledge acquired during the field survey in 2017 and 2018 in our study, we have
identified seven possible land cover classes for land cover classification in Ribb-Gummara
watershed. These include cropland, shrubland, grassland, forestland, built-up area, water bodies,
and bare land (Table 2). Garmin portable Global Positioning System (GPS) apparatus was used to
collect 829 points using proportionate stratified random sampling technique (Jensen, 2016) at the
center of each class. The mean score of horizontal accuracy was 4.7 meters. Of the total 829 points,
491 were used for training and 338 points were used for accuracy assessment of the 2017 Landsat
image. For the 2003 and 2010 Landsat image, 280 training polygons from different land use types
were digitized directly from geo-referenced and high-resolution Google Earth images for each
image. Additionally, 264 polygons from different land-use classes were also digitized for use in
the accuracy assessment of each classified maps. The polygons were initially encoded as KML file
that was later converted to vector file and region of interest (ROI) through the integrated use of
ArcGIS and ENVI software. When the selected samples seem inaccurate, modification of their
boundaries or complete removal was performed.
2.3.2 LULC data and processing
Landsat images
The United States Geological Survey (USGS) has been offering Landsat images for monitoring
changes on the surface of the Earth since 1972 for free of charge (Turner et al., 2015). Three cloud-
free Landsat images for the years 2003, 2010 and 2017 were obtained from the USGS Earth
Explorer web portal. For the year 2003, we acquired Landsat 7 Enhanced Thematic Mapper
(ETM+). Landsat 5 Thematic Mapper (TM5) and Landsat 8 Operational Land Imager (OLI) were
also obtained for the years 2010 and 2017 respectively (Table 3). The year 2010 was purposefully
selected to divide the study time into two equal periods. This was the year when the government
has started SWC practices along with reclaiming illegally occupied land.

In Ethiopia, the dry season is from October to January. Therefore, we collected all cloud-free
Landsat images in December and January at anniversary dates to reduce the impact of seasonalities
on land cover change detection (Munyati, 2000).

Image Preprocessing
All the time series Landsat images used for this research were collection 1 level 1 that means the
images were corrected for geometric and terrain distortions. Although the images were
geometrically corrected, all optical satellite data are prone to atmospheric effects. Therefore, we
applied the fast line-of-sight atmospheric analysis of spectral hypercubes (FLAASH) module using
the Environment for Visualizing Images (ENVI) 5.3 software to remove atmospheric effects from
the images (I T T Visual Information Solutions, 2009). After the atmospheric correction of the
images, it was necessary to carry out a topographic correction to remove the effect of topography
using Topo-correction extension available in the ENVI 5.3 software. Following atmospheric and
topographic corrections, the images were subset to one km buffer from the spatial extent of the
study area. This helps the image not to lose data at the border during water yield simulation. After
obtaining spatially subset images, a spectral subset was carried out and then a 5-band image
consisting of all the visible (red, green, blue), near-infrared (NIR) and short-wave infrared 1
(SWIR1) bands were produced.

Optical indices can improve feature separability for land cover classification. The modified
normalized difference water index (MNDWI) and the normalized difference vegetation index
(NDVI) were generated. NDVI can easily discriminate vegetation from the other land cover types
(Dorigo et al., 2012). Likewise, MNDWI can effectively distinguish water bodies from built-up
areas in addition to restricting information from soil and vegetation and enriching the information
of water (Xu, 2006). NDVI was calculated by subtracting red band from the near infrared band
and dividing with the sum of the near-infrared band and red band. Similarly, MNDWI was
calculated by subtracting shortwave infrared band from green band and divided with the sum of
green band and short-wave infrared band. Finally, a 7-band layer stacked image was generated for
LULC classification.

Image Classification
Among other things, classification accuracy is dependent on the application of an appropriate
classification method. The introduction of machine learning approach and artificial intelligence
paved the way to the birth of many classifiers including decision tree, artificial neural network
(ANN), and support vector machine (SVM) which are highly advanced and non-parametric
(HUANG et al., 2002). Non-parametric classifiers do not include normality assumptions and have
been found to be more relevant for diverse topographic data (Rodriguez-Galiano et al., 2012).
Although object-based image analysis is best performed for Landsat OLI image, SVM is suitable
for all Landsat images (Phiri & Morgenroth, 2017). SVM is capable of producing high accuracy
classified maps than many classifiers like maximum likelihood (Mountrakis et al., 2011; Shao &
Lunetta, 2012). As a result, SVM classifier was applied in this study using the algorithm embedded
in ENVI 5.3 with the Radial Basis Function (RBF) kernel. Prior to the final image classification
with the SVM, an unsupervised classification was undertaken with the Iterative Self Organizing
Data Analysis (ISODATA) clustering algorithm intending to increase our understanding of
spectral similarities and ambiguity arising from the complex nature of the topography (Alphan et
al., 2009).

Accuracy Assessment
Image classification is usually not free from errors. The agreement between the reference data and
the classified map is expressed by an accuracy assessment (Yousefi et al., 2015). The accuracies
of classified land cover maps were checked visually and statistically. Whether the thematic map
looked right or wrong was visually judged and continuous edits were made in line with the local
experience. Then, a highly objective assessment that relied on comparing the proportion of areas
of classified land cover categories with their corresponding areas in other reference datasets was
measured using confusion matrix generated between classified pixels and their corresponding
validation data. Detailed explanations and equations for accuracy measures were found in
(Congalton, 1991). Accuracy assessment was carried out for 2003, 2010, and 2017 classified maps
using ground truth data acquired from Google Earth and field survey. A post-classification
accuracy assessment was carried out to generate a classification confusion matrix or error matrix.
Measures of producer accuracy, user accuracy, and overall accuracy were calculated from the
confusion matrix. Although Kappa Coefficients were calculated to measure the degree of
classification accuracy (Butt et al., 2015), we did not report in this study as the growing number
of literature has criticized it (Foody, 2002; Pontius & Millones, 2011).

Change detection
Change detection is used to quantify land cover transformation over time at a specific geographic
location (Biro et al., 2013; D. Lu, P. Mausel, 2004). Although numerous change detection
approaches have introduced recently; post-classification comparison, image differencing, and
principal component analysis are the most commonly used (D. Lu, P. Mausel, 2004). The post-
classification comparison approach was chosen for this study because it generates the size and
spatial distribution of changed areas as well as percentage share of individual classes within the
change area estimations (El-Hattab, 2016; Yuan et al., 2005). Loss, gain, persistence and net
change of each land cover class were calculated in three periods: 2003-2010, 2010-2017 and 2003-
2017 using cross-tabulation technique (Pontius et al., 2004).

2.3.3 Climate data

The Bahrdar branch of the National Meteorological Services Agency (NMSA) provided rainfall,
maximum and minimum temperatures for eleven meteorological stations over the period between
2001 and 2017 (Figure 1). On average, 4.7% of the data was missing during the study period which
was filled with the nearest station’s average value (Ferrari & Ozaki, 2014). Prior to commencing
the data preparation, the raw data were divided into three periods to get anniversary data with
LULC; 2001-2005, 2006-2011 and 2012-2017. Following this, both mean monthly rainfall and
temperature as well as mean monthly maximum and minimum temperatures were produced.

The average monthly reference evapotranspiration (ETo) was computed using a modified
Hargreaves’s equation (Droogers & Allen, 2002). The modified Hargreaves’s equation was chosen
because it produces better results compared with the Pennman-Montieth method in data-scarce
areas (Sharp et al., 2018). After obtaining the mean monthly ETo and rainfall, average annual
precipitation and average annual ETo were calculated. Finally, the average annual precipitation
and average annual ETo data were interpolated by Inverse Distance Weighted (IDW) with a 30 m
spatial resolution over the entire watershed (Hu et al., 2014; Yong et al., 2010).

2.3.4 Site factor data

Soil-related input data like root-restricting layer depth and plant available water content (PAWC)
are required for InVEST water yield simulation. The Harmonized World Soil Database, version
1.2, from Food and Agriculture Organization (FAO) under United Nations was acquired to
generate root-restricting layer depth and PAWC in this study (FAO/IIASA/ISRIC/ISS-CAS/JRC,
2012; Fischer et al., 2012). PAWC and root-restricting layer depth were calculated using weighted
average of AWC classes and reference soil depth respectively. Zhang constant is also required to
capture the hydrological properties and seasonal patterns of rainfall in the watershed. Despite the
existence of three techniques for Zhang constant calculation, we applied calibration to find the best
Zhang constant in the area as it was recommended by Hamel & Guswa, (2015).

2.3.5 Tabular data

Root depth, plant evapotranspiration coefficient (Kc), and land cover type (1and 0) are required
input parameters in the form of tabular data to calculate potential evapotranspiration from
reference evapotranspiration on the basis of the watershed physiological characteristics. We
calculated Kc based on Allen et al. (1998) and Sharp et al. (2018) and root depth following
Canadell et al. (1996).

2.4. Model Calibration and Validation

The simple production function is the underlying assumption of the InVEST water yield model
where all the water reached the outlet of any watershed is the remaining water from evaporative
loss. The water yield is calculated at each pixel in the watershed but applied only at the watershed
level for practical use. According to Sharp et al. (2018), model calibration should be performed
using long term, not less than ten years, streamflow data to get an accurate result from the model.
As far as data for validation is available, result validation also increases the trustworthiness of the
research findings for practical use. Unlike numerous hydrological models where both calibration
of the model and validation of the result are difficult and uncertain, InVEST model calibration and
validation is simple, straight forward, and impartial (Belete et al., 2018). We delineated Ribb sub-
watershed above Ribb stream gauge and Gummara sub-watershed above Gummara stream gauge,
where actual flow measurement has been taken place, for calibration and validation purposes.

We prepared streamflow records for Gummara at Bahir Dar from 1996 to 2006 for InVEST model
calibration and streamflow data for Ribb at Addis Zemen from 1996-2008 for model simulation
result validation. The streamflow data were aquired from the Ethiopian Ministry of Water,
Irrigation, and Electricity (MoWIE) on request. Following streamflow data preparation, water
yield calculation and simulation result comparison with the measured streamflow data have been
performed. Then consecutive adjustments of root depth, Z parameter, and plant evapotranspiration
coefficients were carried out until we got the optimum result from the model simulation based on
literature and locally available data in the study area. The value of the abovementioned model
input parameters capped through such calibration processes was used for water supply estimation
in Ribb-Gummara watershed for our study

The comparison of simulated water yield against measured streamflow data was performed using
pairwise comparative statistical methods including Bias, Root Mean Square Error (RMSE) and
Mean Error (ME) (Bayissa et al., 2017a; Bitew et al., 2012; Dinku et al., 2007). In the end, the
water supply of the whole Ribb-Gummara watershed was estimated. The final estimated water
supply was compared with observed streamflow of Ribb sub-watershed with Zonal Statistics in
Arc GIS 10.5 environment.

3. RESULTS
3.1.Model Calibration and Result Validation
Water yield model calibration resulted in a Bias of 0.97, a Root Mean Square Error (RMSE) of
0.04 km3/year and Mean Error (ME) of -0.04 km3/year, showing near a perfect agreement between
estimated and observed annual water supply. The values of ME and RMSE are near to zero and
the Bias value is also near to one. The final estimated water supply result has a 0.96 Bias value,
which is very close to the perfect score of Bias (1.00). In addition, -0.05 km3/year and 0.05
km3/year were recorded for ME and RMSE respectively for the final water supply validation,
which are near to the ideal value (0.00).

3.2.Observed Climate Variability

Patterns of rainfall and temperature are reliable indicators of climate change. Therefore, the local
patterns of mean annual temperature and annual total rainfall from 2001 to 2017 were analyzed to
detect climate change in the Ribb-Gummara watershed. The mean annual temperature of the Ribb-
Gummara watershed was 18.78oc during the study period. The year 2002 was the hottest, with an
average annual temperature of 19.85oc. Although not significant, the temperature in the study area
shows a decreasing trend. The mean annual amount of temperature decreased by about 0.023oc
each year over the past two decades (Figure 2). It is contrary to several local and global trends.
Hence, further investigation of why this happened is needed in the Ribb-Gummara watershed.

During the past two decades, the Ribb-Gummara watershed has exhibited a significant increasing
wetness trend. The annual total rainfall increased by about 12.53mm each year from 2001 to 2017
(Figure 3). However, the amount of annual total rainfall has exhibited inter-annual variability with
several dry years, including 2002 (1069mm), 2004 (1091mm) and some wet years such as 2008
(1518mm) and 2010 (1548mm).

3.3.Land Use and Land Cover Change

Table 4 summaries the accuracy assessment report for the 2003, 2010, and 2017 final classified
maps. The overall accuracies of 87.01%, 85.38%, and 88.75% were found for 2003, 2010, and
2017 classified images, showing a strong agreement between the observed and the predicted
classes.

The LULC dynamics of the Ribb-Gummara watershed are shown in Figure 4 for the periods 2003,
2010 and 2017. The LULC trend analysis of the period 2003-2010, 2010-2017, and 2003-2017
demonstrate that the study watershed has experienced high levels of land use and land cover
change.

As shown in Table 1 and Table 2, shrubland, built-up, and bare land increased continuously during
the study period. Cropland was the dominant land use type throughout the period, exhibiting the
fluctuating trends from period-to-period. Its total area was 2397 km2 (73.13%) in 2003, rose to
2749 km2 (84%) by 2010 and dropped back to 2409.05 km2 (73.61%) by the end of the study
period (2017). The growth of cropland (14.71%) in the first period changed to a decline (-12.35%)
in the second period. Illegally expanded cropland on marginal areas was also converted to
vegetated land after the introduction of SWC in 2010. This result is in agreement with the previous
study in the same watershed (Moges & Bhat, 2018), identifying SWC practices as the main reason
for cropland reduction.
Grassland and forestland were the third and the fourth dominant land cover types, after cropland
and shrubland (Figure 4 and Table 1). Contrary to the other land use types, grassland and forestland
areas showed a significant decreasing trend between 2003 and 2017 (Table 2). From 2003 to 2017,
there were a decrease in grassland from 556 km2 (17%) in 2003 to 289 km2 (9%) in 2017 and
forestland from 136 km2 (4.15%) to 89 km2 (2.72%). This finding is supported by other empirical
results (Biru et al., 2015; Hassen & Assen, 2017; Sewnet, 2016; Teferi et al., 2013; Teklaya et al.,
2019) that described the decline of vegetated areas due to conversion to cropland. However, both
grassland and forestland areas began to expand after 2010 mainly on farmland because of SWC
activities and land reclamation measures. Although its share is very small, water bodies showed
an increasing trend which is largely attributed to the Ribb irrigation dam construction on the Ribb
River.

From 2003-2017, nearly two-thirds (2,257 km2) of the watershed remained stable, while one-third
(1,016 km2) of it changed. During this period, the most notable land cover transition was the
persistent increment of shrubland in contrast to many previous findings in the Northern Highlands
of Ethiopia (Moges & Bhat, 2018; Teklaya et al., 2019; Wubie et al., 2016). In the same period,
about 311 km2 of cropland, 27 km2 of grassland and 45 km2 of forestland were converted to
shrubland (Table 3). It seems possible that these results are due to area enclosures, cut‐and‐carry
livestock feeding system, eucalyptus plantations, and agroforestry practices on soil bunds of
croplands as SWC strategies.

3.4.Water Supply Dynamics

The InVEST annual water yield model result described average annual water supply variation over
space and time in the Ribb-Gummara watershed. The average annual water supply was 2461 MCM
(Million Cubic Meter) or 752.02 mm in 2003, 3316 MCM or 1013.24 mm in 2010, and 3125 MCM
or 954.89 in 2017. The spatial variation of the water supply ranged from 191 mm in Ebnat district
to 1470 mm in Estie district. As shown in Figure 5, the southern regions generated higher water
yields, whereas the northern areas generated lower water yields in all periods of the study. This
implies that the Gummara River generated higher runoff than the Ribb River. This result is in line
with previous research findings in the area (Atanaw et al., 2015).

The average annual water supply showed a significant increasing trend between 2003 and 2017.
The water supply increased from 752 mm in 2003 to 955 mm in 2017 in the Ribb-Gumara
watershed. The water supply, which rose from 752 mm in 2003 to 1013 mm in 2010, was higher
during the first period (2003-2010). However, during the second period (2010-2017) it slightly
decreased by 58 mm. This implies that water supply in the form of runoff in the watershed
decreased since the introduction of SWC practice in 2010.

We produced the maps of water yield for three different periods of 2003-2017, 2003-2010 and
2010-2017 to clearly show the spatial changes in water supply in the watershed (Figure 6).
Although there was an overall increase in the water supply (203 mm) between 2003 and 2017, the
Ribb dam area and southeastern regions specifically showed a significant decrease. The average
annual water supply varies from -240 mm in the Ribb dam area to 762 mm in the southwestern
part (Figure 6a). From 2003 to 2010, the average annual water supply increased significantly (261
mm) whereas it slightly decreased (-58 mm) between 2010 and 2017, despite the existence of
regional variation. The regional variation pattern between 2010 and 2017 was contrary to the
spatial pattern observed between 2003 and 2010 (Figure 6b and c). For example, the average
annual water supplies increased (1054 mm) in the northwestern region of the watershed between
2003 and 2010. However, it decreased significantly (-891 mm) between 2010 and 2017 over the
same location in the watershed.

In order to differentiate the impacts of LULC change and climate variability on average annual
water supply, the water supply scenarios were subtracted from the actual water supply of different
periods. Figure 7 shows the residuals of water yield under only land-use change subtracted from
the water production at the end of each period. Although the impact of climate variability on the
water supply was positive between 2003 and 2017 (192 mm), its positive impact (253 mm) was
limited during the first period (2003-2010). The negative impact of climate variability on water
supply was observed on the Ribb-Gummara watershed in the second period (2010 - 2017) (-61
mm).

Figure 8 shows the residuals of water yield under only climate variability subtracted from water
production at the end of each period. Despite the existence of spatial variation, LULC had a
positive impact on the average annual water supply of the study watershed in all periods
considered. Its impact was 11mm from the year 2003 to 2017, 8 mm from 2003 to 2010 and 3 mm
from 2010 to 2017. This implies that the magnitude of LULC impact on water supply kept
decreasing over the last two decades. Although the impact of LULC on the water yield from 2010
to 2017 was positive, its magnitude is insignificant (3 mm).

4. Discussion
4.1.The Relative Impacts of Climate and LULC Changes on Water Production Variation
The study was primarily designed to distinguish the relative impacts of LULC change and climate
variability on the water supply variation in the study area. The results of the study suggest that
both LULC change and climate variability had a positive contribution to the average annual water
supply variation in the Ribb-Gummara watershed from 2003 to 2017 despite spatial differences
(Figure 7a and Figure 8a). These results are consistent with the findings of previous studies
(Andualem & Gebremariam, 2015; Ayele et al., 2016; Jemberie et al., 2016; Teklaya et al., 2019).
However, the impact of both LULC change and climate variability on average annual water supply
before and after 2010 was different. The overall water supply change between 2003 and 2017 was
203 mm. It was -58 mm after 2010 and 261 mm before 2010 (Table 4). There are several possible
explanations for the negative change of water supply after 2010. One possible explanation could
be that the implementation of SWC practices along with land reclamation since 2010 has improved
the watershed vegetation cover which reduces surface runoff. Another possible explanation is the
occurrence of two severe droughts in 2014-2015 and 2009-2010 (Bayissa et al., 2017b).

One unanticipated finding was that about 94% of the total water supply change was contributed
by climate variability from 2003 to 2017, contradicting government reports. The government
believes that illegally occupied land reclamation and SWC practices resulted in large water supply
increases. However, the relative contribution of LULC to the total water supply variation was only
6% during this period. This finding confirms the association between LULC change and water
supply. This finding is in agreement with Pan et al., (2015) findings which showed that the
execution of the Three Rivers Source Area Ecological Protection Project has increased the
vegetated land area and the water supply. There are two likely causes for the low relative
contribution of LULC and the high relative contribution of climate variability to the total water
supply variation in the Ribb-Gummara watershed. One is that there was a small change (13 km2)
in the dominant land use type, cropland, between 2003 and 2017 (Table 2). This implies no
significant improvement in the capacity of the land to convert runoff into groundwater (Arunyawat
& Shrestha, 2016). Another cause is a substantial increasing trend of rainfall (12.53 mm/year) and
a decreasing trend of temperature (0.023oc/year) between 2003 and 2017 (Figure 3 and Figure 2).
Therefore, increasing rainfall coupled with decreasing temperature resulted in increased water
supply in the watershed. This indicates the need for incorporating ecosystem-based climate change
adaptation strategies along with SWC practices.

Another important finding was that although the impact of LULC on water supply was positive
from 2010 to 2017, its relative contribution to the total water supply decreased from 8 mm before
2010 to 3 mm after 2010 (Table 4). This might be due to the improvement of the water holding
capacity of the watershed from ongoing SWC practices commenced in 2010 and indicates the need
to strengthen SWC practices in the watershed and to scale these up to community-based watershed
management.

4.2.The practical implications of this study

The findings of this study have several important implications for future practices. This study is
one of the few, if not the first, to distinguish the relative contribution of LULC change and climate
variability to the overall variation of water supply. Therefore, this paper will be a source of
scientific information on spatial and temporal water supplies supporting development planners and
decision-makers to make informed decisions and to ensure sustainable water supply. The results
also clearly show the effect of the ongoing SWC practices on water supply. Thus, the information
generated in this study provides a springboard to evaluate the outcome of SWC practices at the
watershed level. Contrary to expectations, this study found a significant difference between LULC
change and climate variability impact on water supply. Hence, it would help to implement
appropriate climate change adaptation strategies along with soil and water conservation practices.

Another important contribution of this study is that the simple and straightforward methodology
applied in Ribb-Gummara watershed could be efficiently transferred to other basins across the
world in general and in sub-tropical, temperate and alpine climate regions in particular. The
method is a combination of InVEST integrated catchment water yield model and Residual Trends
analysis. This method could be applied with free and globally available remote sensing data (Sharp
et al., 2018; Vogl et al., 2015). Hence, the method could apply in ungauged basins across the world
to present complex management issues.
4.3.Uncertainty in this study
There are two sources of uncertainties in this study. Some uncertainties are associated with the
InVEST water yield model. This model captures only annual average water supply and ignores
extremes and within year variations. The InVEST water yield model does not capture the seasonal
variation of the water supply, which is very important in agriculture. The water supply change in
the dry season or low flow conditions is associated with LULC, whereas the wet season or high
flow conditions is associated with climate change in general and rainfall variability in particular
(Dile et al., 2013). The other uncertainties derive from model input parameters. Despite the
application of robust classification procedures, it is difficult to generate accurate land use
classifications from Landsat images. The spatial resolution of Landsat image is 30m, which does
not allow all land cover types to be clearly differentiated. For example, grassland and shrubland
demarcation is difficult. In addition, grassland degradation that has an impact on the water supply
is hard to detect from Landsat images (Pan et al., 2015). Considered together, these uncertainties
do not influence the direction of the results acquired and conclusions drawn. However, this work
could be improved further by considering these limitations in the future.

5. Conclusions
This study aimed to separate the relative contribution of LULC change and climate variability to
the average annual water supply in the Ribb-Gummara watershed from 2003 to 2017. In order to
distinguish the impact of LULC on average annual water supplies from the impact of climate
variability, LULC change and climate variability trends were first analyzed. The results showed
an increasing trend of annual total rainfall and a slight decrease in mean annual temperatures. The
watershed also experienced significant LULC dynamics between 2003 and 2017. For example,
cropland, the dominant land use type, increased its area by 15% from 2003 to 2010 and lost 12%
of its area between 2010 and 2017.
This study found that average annual water supplies increased over the last two decades. However,
changes in the water supply show variation before and after 2010. The average annual water supply
change was positive from 2003 to 2010 while negative from 2010 to 2017. The relative
contribution of both LULC change and climate variability to the average annual water supply was
positive from 2003 to 2017. However, climate variability was found to contribute 94% of the water
supply variation, which is much higher than LULC. The impact of climate variability on water
supply was negative between 2010 and 2017. Although it is masked by climate variability, LULC
change still proved a positive impact on water supply, with a diminishing trend indicating the
meaningful impact of SWC and land reclamation practices that commenced in 2010. This implies
that SWC practice should incorporate ecosystem-based climate change adaptation strategies and
scale up to community-based integrated watershed management to support sustainable land use.

It would be interesting to assess the impacts of LULC change and climate variability on seasonal
water supplies. Hence, further study is needed to investigate the present and future impacts of
LULC and climate change on seasonal water production in the Ribb-Gummara watershed to
sustain freshwater provision services throughout the year.

Acknowledgments: This research was funded by the Zhejiang Provincial Natural Science
Foundation of China（No. LY18G030006) and National Natural Science Foundation of China
(No. 41701171). It was also supported by the International Cooperation Regional Development
Project between Zhejiang University and the University of Leeds. The authors thank Editor Shirley
Zhou and three anonymous reviewers for their constructive comments, suggestions, and help in
enhancing the manuscript.

Conflicts of Interest: The authors declare no conflict of interest.

References
Allen RG, Pereira LS, Raes D, Smith M, Ab W. 1998. Crop evapotranspiration - Guidelines for
computing crop water requirements - FAO Irrigation and drainage paper 56. 1–15

Alphan H, Doygun H, Unlukaplan YI. 2009. Post-classification comparison of land cover using
multitemporal Landsat and ASTER imagery: The case of Kahramanmaras, Turkey.
Environmental Monitoring and Assessment 151: 327–336. DOI: 10.1007/s10661-008-0274-
x

Andualem TG, Gebremariam B. 2015. Impact Of Land Use Land Cover Change On Stream
Flow And Sediment Yield : A Case Study Of Gilgel Abay Watershed , Lake Tana.
International Journal of Technology Enhancements and Emerging Engineering Research 3:
28–42

Arunyawat S, Shrestha R. 2016. Assessing Land Use Change and Its Impact on Ecosystem
Services in Northern Thailand. Sustainability 8: 768. DOI: 10.3390/su8080768
Atanaw F, Collick AS, Mcalister C, Taylor R, Steenhuis TS. 2015. Effects of Physical
Catchment Characteristics on River Flow. 1–11

Ayele HS, Li M-H, Tung C-P, Liu T-M. 2016. Assessing Climate Change Impact on Gilgel
Abbay and Gumara Watershed Hydrology, the Upper Blue Nile Basin, Ethiopia.
Terrestrial, Atmospheric and Oceanic Sciences 27: 1005. DOI:
10.3319/TAO.2016.07.30.01

Ayenew Y. 2008. Rainfall-Runoff Modelling for Sustainable Water Resources Management :

The Case of Gumara Watershed , Ethiopia.

Bayissa Y, Tadesse T, Demisse G, Shiferaw A. 2017a. Evaluation of satellite-based rainfall

estimates and application to monitor meteorological drought for the Upper Blue Nile Basin,
Ethiopia. Remote Sensing 9. DOI: 10.3390/rs9070669

Bayissa Y, Tadesse T, Demisse G, Shiferaw A. 2017b. Evaluation of satellite-based rainfall

estimates and application to monitor meteorological drought for the Upper Blue Nile Basin,
Ethiopia. Remote Sensing 9: 1–17. DOI: 10.3390/rs9070669

Belete M, Deng J, Zhou M, Wang K, You S, Hong Y, Weston M. 2018. A new approach to
modelingwater balance in Nile River Basin, Africa. Sustainability (Switzerland) 10: 1–14.
DOI: 10.3390/su10030810

Biro K, Pradhan B, Buchroithner M, Makeschin F. 2013. Land Use/Land Cover Change Analysis
and Its Impact on Soil Properties in the Northern Part of Gadarif Region, Sudan. Land
Degradation & Development 24: 90–102. DOI: 10.1002/ldr.1116

Biru MK, Minale AS, Debay AB. 2015. Multitemporal Land Use Land Cover Change and
Dynamics of Blue Nile Basin by Multitemporal Land Use Land Cover Change and
Dynamics of Blue Nile Basin by Using GIS and Remote Sensing Techniques , North-
Western Ethiopia. International Journal of Environmental Sciences Vol. 4: 81–88

Bitew MM, Gebremichael M, Ghebremichael LT, Bayissa YA. 2012. Evaluation of High-
Resolution Satellite Rainfall Products through Streamflow Simulation in a Hydrological
Modeling of a Small Mountainous Watershed in Ethiopia. Journal of Hydrometeorology 13:
338–350. DOI: 10.1175/2011JHM1292.1
Butt A, Shabbir R, Ahmad SS, Aziz N. 2015. Land use change mapping and analysis using
Remote Sensing and GIS: A case study of Simly watershed, Islamabad, Pakistan. Egyptian
Journal of Remote Sensing and Space Science 18: 251–259. DOI:
10.1016/j.ejrs.2015.07.003

Canadell J, Jackson R, Ehleringer J, Mooney HA, Sala OE, Schulze E-D. 1996. Maximum
rooting depth of vegetation types at the global scale. Oecologia 108: 583–595. DOI:
10.1007/BF00329030

Chakilu GG, Moges MA. 2017. Assessing the Land Use/Cover Dynamics and its Impact on the
Low Flow of Gumara Watershed, Upper Blue Nile Basin, Ethiopia. Hydrology: Current
Research 08: 1–6. DOI: 10.4172/2157-7587.1000268

Chawla I, Mujumdar PP. 2015. Isolating the impacts of land use and climate change on
streamflow. Hydrology and Earth System Sciences 19: 3633–3651. DOI: 10.5194/hess-19-
3633-2015

Congalton RG. 1991. A review of assessing the accuracy of classifications of remotely sensed
data. Remote Sensing of Environment 37: 35–46. DOI: 10.1016/0034-4257(91)90048-B

D. Lu , P. Mausel EB& EM. 2004. Change detection techniques. International Journal of

Remote Sensing 25: 2365–2401. DOI: 10.1080/0143116031000139863

Dile YT, Berndtsson R, Setegn SG. 2013. Hydrological Response to Climate Change for Gilgel
Abay River, in the Lake Tana Basin - Upper Blue Nile Basin of Ethiopia. PLoS ONE 8: 12–
17. DOI: 10.1371/journal.pone.0079296

Dinku T, Ceccato P, Grover‐Kopec E, Lemma M, Connor SJ, Ropelewski CF. 2007. Validation
of satellite rainfall products over East Africa’s complex topography. International Journal
of Remote Sensing 28: 1503–1526. DOI: 10.1080/01431160600954688

Donohue RJ, Roderick ML, McVicar TR. 2012. Roots, storms and soil pores: Incorporating key
ecohydrological processes into Budyko’s hydrological model. Journal of Hydrology 436–
437: 35–50. DOI: 10.1016/j.jhydrol.2012.02.033

Dorigo W, De Jeu R, Chung D, Parinussa R, Liu Y, Wagner W, Fernández-Prieto D. 2012.

Evaluating global trends (1988-2010) in harmonized multi-satellite surface soil moisture.
 Geophysical Research Letters 39: 3–9. DOI: 10.1029/2012GL052988

Droogers P, Allen R. 2002. Estimating reference evapotranspiration under inaccurate data

conditions. Irrigation and Drainage Systems 16: 33–45. DOI: 10.1023/A:1015508322413

Easton ZM, Fuka DR, White ED, Collick AS, Biruk Ashagre B, McCartney M, Awulachew SB,
Ahmed AA, Steenhuis TS. 2010. A multi basin SWAT model analysis of runoff and
sedimentation in the Blue Nile, Ethiopia. Hydrology and Earth System Sciences 14: 1827–
1841. DOI: 10.5194/hess-14-1827-2010

El-Hattab MM. 2016. Applying post classification change detection technique to monitor an
Egyptian coastal zone (Abu Qir Bay). Egyptian Journal of Remote Sensing and Space
Science 19: 23–36. DOI: 10.1016/j.ejrs.2016.02.002

Evans J, Geerken R. 2004. Discrimination between climate and human-induced dryland

degradation. Journal of Arid Environments 57: 535–554. DOI: 10.1016/S0140-
1963(03)00121-6

FAO/IIASA/ISRIC/ISS-CAS/JRC. 2012. Harmonized World Soil Database (version 1.2). FAO,

Rome, Italy and IIASA, Laxenburg, Austria, 1–50

Feng XM, Sun G, Fu BJ, Su CH, Liu Y, Lamparski H. 2012. Regional effects of vegetation
restoration on water yield across the Loess Plateau, China. Hydrology and Earth System
Sciences 16: 2617–2628. DOI: 10.5194/hess-16-2617-2012

Ferrari GT, Ozaki V. 2014. Missing Data Imputation of Climate Datasets : Implications To
Modeling Extreme Drought Events. Revista Brasileira de Meteorologia 29: 21–28. DOI:
10.1590/S0102-77862014000100003

Fischer G, Nachtergaele FO, Prieler S, Teixeira E, Toth G, van Velthuizen H, Verelst L, Wiberg
D. 2012. Global Agro-ecological Zones (GAEZ): Model Documentation. Rome, Italy

Foody GM. 2002. Status of land cover classification accuracy assessment. Remote Sensing of
Environment 80: 185–201. DOI: 10.1016/S0034-4257(01)00295-4

Fu BP. 1981. On the calculation of the evaporation from land surface. S C IE N TA T M O S P H

E R I C A S IN I C A 5: 23–31
Hamel P, Guswa AJ. 2015. Uncertainty analysis of a spatially explicit annual water-balance
model: Case study of the Cape Fear basin, North Carolina. Hydrology and Earth System
Sciences 19: 839–853. DOI: 10.5194/hess-19-839-2015

Hassen EE, Assen M. 2017. Land use / cover dynamics and its drivers in Gelda catchment , Lake
Tana watershed ,. Environmental Systems Research 1–13. DOI: 10.1186/s40068-017-0081-
x

Hu Q, Yang D, Li Z, Mishra AK, Wang Y, Yang H. 2014. Multi-scale evaluation of six high-
resolution satellite monthly rainfall estimates over a humid region in China with dense rain
gauges. International Journal of Remote Sensing 35: 1272–1294. DOI:
10.1080/01431161.2013.876118

HUANG C, DAVIS LS, TOWNSHEND JRG. 2002. An assessment of support vector machines
for land cover classification. International Journal of Remote Sensing 23: 725–749. DOI:
10.1080/01431160110040323

I T T Visual Information Solutions. 2009. Atmospheric Correction Module : QUAC and

FLAASH User ’ s Guide. 44

Jemberie M, Gebrie T, Gebremariam B. 2016. Evaluation of Land Use Land Cover Change on
Stream Flow: a Case Study Ofdedissa Sub Basin, Abay Basin, South Western Ethiopia.
International Journal of Innovations in Engineering Research and Technology 3: 2394–
3696

Jensen J. 2016. Introductory Digital Image Processing: A Remote Sensing Perspective. Pearson
Education, Inc: University of South Carolina, USA

Kremen C. 2005. Managing ecosystem services: What do we need to know about their ecology?
Ecology Letters 8: 468–479. DOI: 10.1111/j.1461-0248.2005.00751.x

Krysanova V, White M. 2015. Advances in water resources assessment with SWAT—an

overview. Hydrological Sciences Journal 60: 771–783. DOI:
10.1080/02626667.2015.1029482

Lü Y, Fu B, Feng X, Zeng Y, Liu Y, Chang R, Sun G, Wu B. 2012. A policy-driven large scale

ecological restoration: Quantifying ecosystem services changes in the loess plateau of
 China. PLoS ONE 7: 1–10. DOI: 10.1371/journal.pone.0031782

MEA. 2005. Millennium Ecosystem Assessment (MEA), 2005. Ecosystems and Human Well-
being: Synthesis. Island Press, Washington, DC. World Health. DOI:
10.1196/annals.1439.003

Moges DM, Bhat HG. 2017. Integration of geospatial technologies with RUSLE for analysis of
land use/cover change impact on soil erosion: case study in Rib watershed, north-western
highland Ethiopia. Environmental Earth Sciences 76: 1–14. DOI: 10.1007/s12665-017-
7109-4

Moges DM, Bhat HG. 2018. An insight into land use and land cover changes and their impacts in
Rib watershed, north-western highland Ethiopia. Land Degradation and Development 29:
3317–3330. DOI: 10.1002/ldr.3091

Mountrakis G, Im J, Ogole C. 2011. Support vector machines in remote sensing: A review.

ISPRS Journal of Photogrammetry and Remote Sensing 66: 247–259. DOI:
10.1016/j.isprsjprs.2010.11.001

Mulatu C, Crosato A, Moges M, Langendoen E, McClain M. 2018. Morphodynamic Trends of

the Ribb River, Ethiopia, Prior to Dam Construction. Geosciences 8: 255. DOI:
10.3390/geosciences8070255

Munyati C. 2000. Wetland change detection on the Kafue Flats, Zambia, by classification of a
multitemporal remote sensing image dataset. International Journal of Remote Sensing 21:
1787–1806. DOI: 10.1080/014311600209742

Pan T, Wu S, Liu Y. 2015. Relative contributions of land use and climate change to water supply
variations over yellow river source area in Tibetan Plateau during the past three decades.
PLoS ONE 10: 1–19. DOI: 10.1371/journal.pone.0123793

Pessacg N, Flaherty S, Brandizi L, Solman S, Pascual M. 2015. Getting water right: A case study
in water yield modelling based on precipitation data. Science of the Total Environment 537:
225–234. DOI: 10.1016/j.scitotenv.2015.07.148

Phiri D, Morgenroth J. 2017. Developments in Landsat land cover classification methods: A

review. Remote Sensing 9. DOI: 10.3390/rs9090967
Pontius RG, Millones M. 2011. Death to Kappa: Birth of quantity disagreement and allocation
disagreement for accuracy assessment. International Journal of Remote Sensing 32: 4407–
4429. DOI: 10.1080/01431161.2011.552923

Pontius RG, Shusas E, McEachern M. 2004. Detecting important categorical land changes while
accounting for persistence. Agriculture, Ecosystems and Environment 101: 251–268. DOI:
10.1016/j.agee.2003.09.008

Praskievicz S, Chang H. 2009. A review of hydrological modelling of basin-scale climate change

and urban development impacts. Progress in Physical Geography 33: 650–671. DOI:
10.1177/0309133309348098

Rodriguez-Galiano VF, Ghimire B, Rogan J, Chica-Olmo M, Rigol-Sanchez JP. 2012. An

assessment of the effectiveness of a random forest classifier for land-cover classification.
ISPRS Journal of Photogrammetry and Remote Sensing 67: 93–104. DOI:
10.1016/j.isprsjprs.2011.11.002

Sahle M, Saito O, Fürst C, Yeshitela K. 2019. Quantifying and mapping of water-related

ecosystem services for enhancing the security of the food-water-energy nexus in tropical
data–sparse catchment. Science of the Total Environment 646: 573–586. DOI:
10.1016/j.scitotenv.2018.07.347

Setegn SG, Srinivasan R, Dargahi B, Melesse AM. 2009. Spatial delineation of soil erosion
vulnerability in the Lake Tana Basin, Ethiopia. HYDROLOGICAL PROCESSES 23: 3738–
3750. DOI: 10.1002/hyp.7476

Sewnet A. 2016. Land use/cover change at Infraz watershed by using GIS and remote sensing
techniques, northwestern Ethiopia. International Journal of River Basin Management 14:
133–142. DOI: 10.1080/15715124.2015.1095199

Shao Y, Lunetta RS. 2012. Comparison of support vector machine, neural network, and CART
algorithms for the land-cover classification using limited training data points. ISPRS
Journal of Photogrammetry and Remote Sensing 70: 78–87. DOI:
10.1016/j.isprsjprs.2012.04.001

Sharp, R., Tallis, H.T., Ricketts, T., Guerry, A.D., Wood, S.A., Chaplin-Kramer, R., Nelson, E.
 E, D., Wolny, S., Olwero, N., Vigerstol, K., Pennington, D., Mendoza, G., Aukema, J.,
Foster, J., Forrest, J. C, D., Arkema, K., Lonsdorf, E., Kennedy, C., Verutes, G., Kim, C.K.,
Guannel, G., Papenfus, M., Toft, J., Marsik M, Bernhardt, J., Griffin, R., Glowinski, K.,
Chaumont, N., Perelman, A., Lacayo, M. Mandle, L., Hamel, P., Vogl AL, Rogers, L.,
Bierbower, W., Denu, D., and Douglass J. 2018. InVEST 3.5.0 User’s Guide. The Natural
Capital Project, Stanford University, University of Minnesota, The Nature Conservancy,
and World Wildlife Fund: Stanford

Soytong P, Janchidfa K, Phengphit N, Perera, Chayhard S, Ranjith. 2016. The Effects of Land
Use Change and Climate Change on Water Resources in the Eastern Region of Thailand.
12: 1695–1722

Teferi E, Bewket W, Uhlenbrook S, Wenninger J. 2013. Understanding recent land use and land
cover dynamics in the source region of the Upper Blue Nile , Ethiopia : Spatially explicit
statistical modeling of systematic transitions. “Agriculture, Ecosystems and Environment”
165: 98–117. DOI: 10.1016/j.agee.2012.11.007

Teklaya A, Dileb YT, Setegnc SG, Demissiec SS, Asfaw DH. 2019. Evaluation of static and
dynamic land use data for watershed hydrologic process simulation : A case study in
Gummara watershed , Ethiopia. Catena 172: 65–75. DOI: 10.1016/j.catena.2018.08.013

Turner W, Rondinini C, Pettorelli N, Mora B, Leidner AK, Szantoi Z, Buchanan G, Dech S,

Dwyer J, Herold M, Koh LP, Leimgruber P, Taubenboeck H, Wegmann M, Wikelski M,
Woodcock C. 2015. Free and open-access satellite data are key to biodiversity conservation.
Biological Conservation 182: 173–176. DOI: 10.1016/j.biocon.2014.11.048

Van Griensven A, Ndomba P, Yalew S, Kilonzo F. 2012. Critical review of SWAT applications
in the upper Nile basin countries. Hydrology and Earth System Sciences 16: 3371–3381.
DOI: 10.5194/hess-16-3371-2012

Vogl AL, Wolny S, Dennedy-Frank PJ, Hamel P, Johnson JA, Rogers M. 2015. Managing
catchments for hydropower services. Wealth Accounting and the Valuation of Ecosystem
Services (WAVES): Washington D.C.

Wessels KJ, Prince SD, Malherbe J, Small J, Frost PE, VanZyl D. 2007. Can human-induced
 land degradation be distinguished from the effects of rainfall variability? A case study in
South Africa. Journal of Arid Environments 68: 271–297. DOI:
10.1016/j.jaridenv.2006.05.015

Worqlul AW, Ayana EK, Atanaw Bahir F. 2015. Effect of Physical Catchment Characteristics
on River Flow: The Case of Ribb and Gumara Rivers in Upper Blue Nile, Ethiopia.
International Soil & Water Assessment Tool Conference June 24-26, 2015, 87. DOI:
10.13140/RG.2.1.2288.3680

Wubie MA, Assen M, Nicolau MD. 2016. Patterns, causes and consequences of land use/cover
dynamics in the Gumara watershed of lake Tana basin, Northwestern Ethiopia.
Environmental Systems Research 5: 8. DOI: 10.1186/s40068-016-0058-1

Xu H. 2006. Modification of normalised difference water index (NDWI) to enhance open water
features in remotely sensed imagery. International Journal of Remote Sensing 27: 3025–
3033. DOI: 10.1080/01431160600589179

Yin J, He F, Jiu Xiong Y, Yu Qiu G. 2017. Effects of land use/land cover and climate changes on
surface runoff in a semi-humid and semi-arid transition zone in northwest China. Hydrology
and Earth System Sciences 21: 183–196. DOI: 10.5194/hess-21-183-2017

Yong B, Ren LL, Hong Y, Wang JH, Gourley JJ, Jiang SH, Chen X, Wang W. 2010. Hydrologic
evaluation of Multisatellite Precipitation Analysis standard precipitation products in basins
beyond its inclined latitude band: A case study in Laohahe basin, China. Water Resources
Research 46: 1–20. DOI: 10.1029/2009WR008965

Yousefi S, Khatami R, Mountrakis G, Mirzaee S, Pourghasemi HR, Tazeh M. 2015. Accuracy

assessment of land cover/land use classifiers in dry and humid areas of Iran. Environmental
Monitoring and Assessment 187. DOI: 10.1007/s10661-015-4847-1

Yuan F, Sawaya KE, Loeffelholz BC, Bauer ME. 2005. Land cover classification and change
analysis of the Twin Cities (Minnesota) metropolitan area by multitemporal Landsat remote
sensing. Remote Sensing of Environment 98: 317–328. DOI: 10.1016/j.rse.2005.08.006

Zhang L, Hickel K, Dawes WR, Chiew FHS, Western AW, Briggs PR. 2004. A rational function
approach for estimating mean annual evapotranspiration. Water Resources Research 40:
 W025021–W02502114. DOI: 10.1029/2003WR002710

Zhang L, Nan Z, Xu Y, Li S. 2016. Hydrological impacts of land use change and climate
variability in the headwater region of the Heihe River Basin, northwest China. PLoS ONE
11: 1–25. DOI: 10.1371/journal.pone.0158394
Table 1. All the input parameters and data sources for InVEST 3.5.0 annual water yield model
Input Data Type or Format Spatial Sources
Resolution
Average annual precipitation GIS raster 30m × 30m NMSA
(mm)
Average annual ETo (mm) GIS raster 30m × 30m NMSA

Soil depth (mm) GIS raster 30m × 30m (FAO/IIASA/ISRIC/ISS-CAS/JRC,

2012; Fischer et al., 2012).

PAWC (mm) GIS raster 30m × 30m (FAO/IIASA/ISRIC/ISS-CAS/JRC,

2012; Fischer et al., 2012).

LULC GIS raster 30m × 30m USGS, (Turner et al., 2015)

Watershed and Sub- GIS - Delineated by the researchers using Arc

watersheds polygon/Shapefile Hydro tool in Arc GIS 10.5 from SRTM
30m DEM data
Root Depth Per LULC class - (Canadell et al., 1996)

Kc Per LULC class - (Allen et al., 1998; Sharp et al., 2018)

Z parameter Integer number - (Hamel & Guswa, 2015)

 Table 2. Definitions of land use/land cover classification system in the study area

Land use types Description

Crop Land A land covered with annual and perennial crops frequently found in plains, foot slopes,
plateaus, and valley floors
Grass Land A land covered with grass that found in flat areas and river banks in which water-table is
near the surface
Shrub Land A land dominantly covered by vegetation with lower than one meter height and 50% canopy
cover
Forest Land These include a remnant of high natural forests found in church fence, steep slope areas, and
eucalyptus plantations having more than 50% canopy cover
Built up Area All man-made infrastructures including buildings, roads concrete sports fields etc.
Water Body Any part of the study area covered with surface water like streams, rivers, ponds, dams, and
lakes
Bare Land Land of limited ability to support life that covered with sand and rocks
Table 3. Properties of the Landsat images used in this study

Sensor type Spatial resolution (m) Bands used Path/row Date of acquisition
Landsat 5 TM 30*30 1, 2, 3, 4, 5, 169/52 2010-01-14
Landsat 7 ETM+ 30*30 1, 2, 3, 4, 5, 169/52 2003-01-03
Landsat 8 OLI 30*30 2, 3, 4,5, 6, 169/52 2017-12-19
 Table 4. Classified map accuracy assessment report (2003, 2010, and 2017)
Land use Classes 2003 2010 2017
PA (%) UA (%) OA (%) PA (%) UA (%) OA (%) PA (%) UA (%) OA (%)
Crop Land 93.08 79.70 97.39 97.39 88.61 84.36
Grass Land 84.92 80.64 79.05 72.81 87.89 82.98
Shrub Land 63.57 79.88 35.90 93.33 55.71 79.14
Forest Land 95.64 93.57 87.01 98.87 93.33 85.38 99.89 94.71 88.75
Built up Area 48.86 78.18 59.52 100.00 57.65 58.33
Water Body 96.75 93.70 91.43 94.12 96.72 93.65
Bare Land 23.44 90.91 67.57 86.21 58.27 87.06
OA = Overall Accuracy, PA = Producer Accuracies, and UA = User Accuracies
 Table 1.Area coverage of each LULC class (2003-2017)
Land Use Type Land use (2003) Land use (2010) Land use (2017)
Area (km2) Percent Area (km2) Percent Area (km2) Percent
Crop Land 2396.57 73.13 2748.99 83.99 2409.50 73.61
Bare Area 0.00 0.00 1.06 0.03 2.51 0.08
Built up Area 4.54 0.14 4.67 0.14 20.74 0.63
Forest Land 135.76 4.15 78.21 2.38 89.02 2.72
Grass Land 556.35 17.00 113.1 3.46 289.24 8.84
Shrub Land 166.81 5.11 320 9.78 442.65 13.53
Water Body 13.15 0.41 7.15 0.22 19.52 0.59
Total 3,273.18 100 3,273.18 100 3,273.18 100
 Table 2. The rate of change for each LULC class (2003-2017)
Land Use Type Land Use Change (2003-2010) Land Use Change (2010-2017) Land Use Change (2003-2017)
Area (km2) Percent Area (km2) Percent Area (km2) Percent
Crop Land 352.42 14.71 -339.49 -12.35 12.93 0.54
Bare Area 1.06 - 1.45 136.47 2.51 -
Built up Area 0.13 2.98 16.07 344.04 16.20 357.25
Forest Land -57.55 -42.39 10.81 13.82 -46.74 -34.43
Grass Land -443.24 -79.67 176.14 155.73 -267.10 -48.01
Shrub Land 153.17 91.82 122.65 38.33 275.82 165.34
Water Body -6.00 -45.62 12.37 173.07 6.38 48.51
 Table 3. LULC conversion matrix between 2003 and 2017
Initial State (2003)
Classes Crop Grass Shrub Forest Built-up Water Bare Row Gain
Land Land Land Land Area Body Land Total
Crop Land 1952.41 328.66 88.04 31.40 3.40 5.16 0 2409.50 457.09
Final State (2017)

Grass Land 88.53 189.75 10.19 0.62 0.13 0.03 0 289.24 99.49
Shrub Land 311.02 27.39 56.37 44.78 0.05 3.02 0 442.65 386.28
Forest Land 17.97 1.35 10.64 56.12 0.00 2.93 0 89.02 32.9
Built up Area 13.88 4.30 0.79 0.83 0.90 0.04 0 20.74 19.84
Water Body 11.73 3.50 0.77 1.99 0.05 1.48 0 19.52 18.04
Bare Land 1.03 1.40 0.01 0.03 0.01 0.04 0 2.51 2.51
Column Total 2396.57 556.35 166.82 135.76 4.54 13.15 0
Loss 444.16 366.60 110.45 79.64 3.64 11.67 0
Net Change 12.93 -267.10 275.82 -46.76 16.20 6.38 2.51
Table 4. The relative contribution of Climate variability and LULC change to average annual
water supply
Periods ∆𝑌 (mm) ∆𝑌𝐶 (mm) ∆𝑌𝐿 (mm) 𝜂𝐶 (%) 𝜂𝐿 (%)
2003-2017 203 192 11 94 6
2003-2010 261 253 8 97 3
2010-2017 -58 -61 3 105 -5
Figure 1. Location map of Ribb-Gummara watershed
Figure 2. Inter-annual temperature patterns and trends of change (2001-2017)
Figure 3. Patterns of annual total rainfall and trends (2001-2017)
Figure 4. LULC map of Ribb-Gummara watershed
Figure 5. Spatial distribution of calculated annual water supply (mm)
Figure 6. Average annual water supply variation during the last two decades
Figure 7. The impact of climate change on the average annual water supply

Figure 8. The impact of LULC on the average annual water supply