Environmental Challenges 5 (2021) 100273 

Contents lists available at ScienceDirect 

Environmental Challenges 

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

Impacts of climate variability and changing land use/land cover on River 

Mpanga flows in Uganda, East Africa 

Charles Onyutha 

∗ , Catherine Turyahabwe , Paul Kaweesa 

Department of Civil and Environmental Engineering, Kyambogo University, P.O. Box 1, Kyambogo, Kampala, Uganda 

a r t i c l e i n f o 

Keywords: 

SWAT 

Land use/land cover (LULC) changes 

Hydrological modeling 

River Mpanga catchment 

partitioning river flow drivers 

Uganda 

a b s t r a c t 

We analyzed River Mpanga Catchment (RMC) land use/land cover (LULC) types based on Landsat images for 

2000, 2008 and 2014. Soil and Water Assessment Tool (SWAT) was driven by daily meteorological data from 

2000 to 2011 to investigate impacts of LULC changes on river flow variation. In 2000, 2008, and 2014, cropland 

covered 33.0%, 69.1%, and 72.2% of RMC area, respectively. However, the fractions of the RMC area covered by 

grassland in 2000, 2008, and 2014 were 39.4%, 12.5%, and 10.4%, respectively. The portion of RMC area covered 

by human settlement increased from 0.2% in 2000 to 0.5% by 2014. RMC was characterized by increasing trends 

in annual rainfall and river flows. SWAT calibration and validation at daily scale over the periods 2000–2005 

and 2006–2011 yielded Nash Sutcliffe Efficiency of 0.77 and 0.75, respectively. Contribution from transitions in 

LULC types to river flow changes over the period 2000–2008 was 7.65%. Generally, 70.46% of the total river flow 

variation was contributed by climate variability in terms of changes in climatic conditions. However, 21.89% of 

the total river flow variance remained unexplained and this could be attributed to other factors not considered 

in this study including extra impacts of human activities such water abstractions for agricultural, industrial and 

domestic needs. These findings are important for planning predictive land and water resources management 

amidst impacts of climate variability and human activities on water resources. 

1

 

s  

m  

p  

a  

t  

t  

a  

m  

(

 

v  

fl  

e  

X  

t  

t  

2  

r  

r  

t  

2  

B  

c  

v

d  

2  

a  

o  

e  

t  

a  

i  

n  

C  

c  

I  

a  

i  

o  

m  

g  

h

R

2

. Introduction 

Sustainable management of the Earth’s surface including water re-

ources and land remains a critical environmental challenge that society

ust address ( Mustard et al., 2004 ). This challenge arises due to com-

eting water demands for agriculture, mining, tourism, urbanization,

nd industries. Due to continuous population growth and other factors,

here have been dramatic land use/land cover (LULC) changes across

he various watersheds in the world thereby putting water resources

nd land under increasing stress ( Global Water Partnership 2009 ). Cli-

ate variability and changes in LULC types affect quantity of river flow

 Pirnia et al., 2019 ). 

Several studies across the different regions of the world showed

arying extents to which climate variability and human activities in-

uenced changes in river flows (see e.g. ( Pirnia et al., 2019 , Kumar

t al., 2018 , Wang et al., 2016 , Zhao et al., 2016 , Tan et al., 2015 ,

u et al., 2014 , Zhang et al., 2011 , Wang et al., 2009 )). Contribu-

ions of climate variability and human activities to the decrease in

he river flows of Guantai catchment over the period 1950–2005 were

6.1% and 73.9%, respectively ( Bao et al., 2012 ). Over the two pe-

iods 1961–1966 and 1973–2001, 35% and 68% of the decrease in

ainfall-runoff across the Chao River basin in China were attributed
∗ Corresponding author. 

E-mail address: conyutha@kyu.ac.ug (C. Onyutha). 

ttps://doi.org/10.1016/j.envc.2021.100273 

eceived 1 April 2021; Received in revised form 1 September 2021; Accepted 2 Sept

667-0100/© 2021 The Author(s). Published by Elsevier B.V. This is an open access a
o climate variations and human activities, respectively ( Wang et al.,

009 ). Changes in the water resources availability of the Weihe River

asin were more attributable to climate variability than land use

hange ( Zhao et al., 2016 ). For Haihe basin, the impacts of climate

ariation and LULC changes were found to account for the runoff

ecrease by 26.9% and 73.1% on average, respectively ( Xu et al.,

014 ). In the Hun–Tai River basin, the contributions of climate vari-

bility and human activities to the reduction of annual streamflow

ver the period 1961–2006 were 43% and 57%, respectively ( Zhang

t al., 2011 ). Climate variability and human activities contributed

o the decrease in annual runoff of the Mian River basin by 23.9%

nd 76.1%, respectively ( Fan et al., 2010 ). For the Johor River basin

n Malaysia, climate and human activities were found to raise an-

ual stream flow by 4.4% and 0.06%, respectively ( Tan et al., 2015 ).

ontributions of climatic variability and human activities to the de-

rease in the stream flow changes in the Haraz River basin, northern

ran was 34.78% and 65.21%, respectively ( Pirnia et al., 2019 ). The

bove information shows that the extents to which climate variabil-

ty and LULC changes impact on the changes in river flows vary from

ne catchment to another and across regions. This is because catch-

ents are spatially different with respect to size, topography, soils,

eology, space–time variability of climatic variables such as rainfall
ember 2021 

rticle under the CC BY license ( http://creativecommons.org/licenses/by/4.0/ ) 

https://doi.org/10.1016/j.envc.2021.100273
http://www.ScienceDirect.com
http://www.elsevier.com/locate/envc
http://crossmark.crossref.org/dialog/?doi=10.1016/j.envc.2021.100273&domain=pdf
mailto:conyutha@kyu.ac.ug
https://doi.org/10.1016/j.envc.2021.100273
http://creativecommons.org/licenses/by/4.0/


C. Onyutha, C. Turyahabwe and P. Kaweesa Environmental Challenges 5 (2021) 100273 

a  

c  

2

 

t  

c  

B  

a  

s  

e  

m  

c  

(  

(  

s  

i  

f  

(  

g  

s  

p  

fl  

t  

o  

i  

h  

R

 

i  

i  

a  

o  

i  

a  

f  

t  

a  

b  

s  

w  

T  

i  

o  

(  

w  

c  

g  

o  

H  

t  

i  

l  

f  

m  

e  

s  

W  

2  

t

 

(  

U  

S  

A  

f  

e  

c  

w  

m  

r

 

i  

t  

f  

r  

c  

t  

a  

i  

d  

t  

a  

e  

d  

e  

W  

c

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

nd evapotranspiration and also the levels of human involvements in

hanging LULC types are not the same for various watersheds ( Onyutha,

016 ). 

In East Africa or the region where the current study area is located,

here were several researchers who investigated the influence of LULC

hanges on river flows of various watersheds some of which include the

lue Nile basin ( Gebrehiwot et al., 2013 , Rientjes et al., 2011 , Bewket

nd Sterk, 2005 ), Lake Bunyonyi catchment ( Kizza et al., 2017 ), Murchi-

on Bay catchment Anaba et al., (2017) , River Muzizi catchment ( Bahati

t al., 2021 ), Nyando catchment ( Olang and Fürst, 2011 ), Mara catch-

ent ( Mwangi et al., 2016 , Mango et al., 2011 , Mati et al., 2008 ), Nzoia

atchment ( Odira et al., 2010 ), Malagarasi River catchment (Tanzania)

 Kashaigili and Majaliwa, 2013 ), Njoro catchment Baker and Miller,

2013) , and Wami River Basin ( Nobert and Jeremiah, 2012 ). For in-

tance, Baker and Miller, (2013) reported that LULC changes resulted

nto an increased surface runoff and decreased groundwater recharge

or the Njoro watershed located in Kenya’s Rift Valley. Mango et al.

2011) warned that any further conversion of forests to agriculture and

rassland in the Upper Mara River Basin headwaters would reduce dry

eason flows and increase peak flows. Mati et al. ( Mati et al., 2008 ) re-

orted that changes in LULC types in Mara catchment increased the peak

ow by 7%. Further information regarding impacts of LULC changes on

he surface runoff of the various catchments across East Africa can be

btained in a review paper by Guzha et al., (2018) . Despite the above

nformation, there were no studies which could be found in literature to

ave been conducted on the impacts of LULC changes on river flow of

iver Mpanga catchment. 

Available data shows that River Mpanga flows have been increas-

ng since 2000. Attempts to explain such trend can be through answer-

ng some questions. For instance, is there evidence of LULC changes

cross the River Mpanga catchment (RMC)? What could be the causes

f such LULC changes? What are the contributions of climate variabil-

ty and LULC changes to the river flow variation in the RMC? There

re several human activities, such as deforestation, sand mining, poor

arming practices, wetland encroachment, and overgrazing which lead

o LULC changes in the RMC. Impacts of LULC changes and climate vari-

bility on runoff across the RMC have never been investigated studied

efore our study. Furthermore, implementation of Integrated Water Re-

ources Management (IWRM) in the study area remains thin coupled

ith limited exercise of IWRM in practice ( Nicol and Odinga, 2016 ).

his is because the embedding of catchment management institutions

n local political and other processes with respect to the integration

f land- and water-related aspects is still a big challenge in Uganda

 Nicol and Odinga, 2016 ). One form of support for planning of proper

atershed management is provision of information on historical LULC

hanges and its potential effects on water resources. A typical way to

et such information is through analysis of high-quality aerial photos

r satellite images of LULC types archived over a long period of time.

owever, archives of such data with high spatial and temporal resolu-

ions are difficult to obtain for the study area. Besides, the use of aerial

mages alone cannot directly indicate how the changes in a particular

and cover type are affecting river flows. Thus, the available LULC data

rom satellite images can be complimented with catchment hydrological

odeling ( Onyutha and Willems, 2018 ). For such modeling studies (see

.g. ( Kumar et al., 2018 ), ( Kumar et al., 2017 ), ( Narsimlu et al., 2015 )),

emi-distributed macro-scale hydrological models such as, the Soil and

ater Assessment Tool (SWAT) ( Arnold et al., 1993 ), ( Neitsch et al.,

002 ) and Variable Infiltration Capacity model ( Liang et al., 1994 ) tend

o be used. 

At this time, only three studies Anaba et al., (2017) , Bahati et al.,

2021) , Mutenyo et al., (2011) applied SWAT to some catchments in

ganda. Mutenyo et al., (2011) focused on assessing the effectiveness of

WAT in predicting the flows of River Manafwa in the eastern Uganda.

naba et al. ( Anaba et al., 2017 ) applied SWAT to simulate river flows

or estimating sediment yield from the Murchison Bay catchment. Bahati

t al. ( Bahati et al., 2021 ) applied SWAT to investigate the effect of LULC
2 
hanges on hydropower of Muzizi hydropower plant. The main reason

hy applications of physically-based or semi-distributed hydrological

odels to Ugandan watersheds remain limited is scarcity of quality data

equired by the models at fine temporal and spatial resolutions. 

It is worth noting that impacts of LULC on hydrology can be analyzed

n context of climate change or climate variability. The difference be-

ween climate variability and climate change lies in the time scales used

or analysis. For climate change impact investigation, two long-term pe-

iods (say, at least 30 years) one under current and the other from future

limatic conditions are considered. Furthermore, climate change tends

o focus on the stationarity of the climate system. Under climate vari-

bility, wet and dry climatic conditions tend to occur in a clustered way

n time ( Onyutha and Willems, 2017 ). The epochs over which wet or

ry conditions occur can vary from annual to decadal (or multi-decadal)

ime scales. Under climate variability we focus on the oscillation lows

nd highs of climatic conditions which are known to engender extreme

vents such as droughts and floods. The idea is that the extent of the

amage from such hydrological extremes (like flooding events) can be

xacerbated by how LULC types are modified through human activities.

hether to consider climate change or variability, hydrological changes

an be linked to human activities in a number of ways as follows. 

i) In the context of climate change, the use of predicted (of future)

LULC information is combined with outputs from scenario out-

puts from general circulation models for hydrological analysis.

An example of such analysis for a catchment within the same re-

gion where our study area is located was conducted by Bahati

et al. ( Bahati et al., 2021 ) in which scenarios from four climate

models were combined with future LULC types in River Muzizi

catchment for 2060 projected using the MOLUSCE (Module for

Land Use Change Evaluation). 

ii) In case there is a significant hydrological change (for instance,

step jump in mean in river flow), analysis of LULC types and

river flow before and after the change-point can be used for

the hydrological attribution. Typical examples of such analysis

(though not for the current study area) can be found in Pirnia

et al., (2019) and Wang et al., (2016) . A hydrological sensitivity

elasticity-based method can also be applied such that the river

flow sensitivity to precipitation and PET is computed with respect

to dryness index and plant available water coefficient (see e.g.

( Koster and Suarez, 1999 , Zhang et al., 2001 , Milly and Dunne,

2002 , Sun et al., 2005 )). 

iii) We can use one LULC map under changing climatic conditions

and this can be good for scenario analysis to investigate the sen-

sitivity of a catchment to the changes in climatic conditions. This

method can be used to answer, for instance, the question on the

amount by which the maximum annual flows will change when

precipitation intensities are maintained as they occurred over the

given period while potential evapotranspiration (PET) rates are

increased, say, by 5% compared to the historical values. 

iv) A number of LULC maps can also be used under constant climatic

condition over a particular period. Here, the differences in results

of simulations based on various LULC maps characterize impacts

of LULC changes on hydrology during the period under consider-

ation. If there is only one LULC map, other LULC information can

be generated for scenario analysis. Such an approach can be used

to answer a question such as: To what extent will the river flow

change if 5% of forest across a catchment is converted to crop-

land while areas for all the other LULC types remain constant or

minimally affected? 

v) A sub-period can be selected as the baseline such that a

(semi)distributed model like SWAT is run to simulate hydrologi-

cal conditions over various epochs especially before and after the

baseline. Using this approach, Kumar et al., (2018) investigated

impacts of LULC on water availability of Tons River Basin, Mad-

hya Pradesh, India. The distributed model can also be run over



C. Onyutha, C. Turyahabwe and P. Kaweesa Environmental Challenges 5 (2021) 100273 

 

 

 

 

 

 

i  

i  

a  

s  

c  

i  

d  

p

2

2

 

o  

U  

t  

G  

d  

3  

G  

t  

1  

t  

3

 

a  

F  

i  

d  

a  

2  

i  

r  

e

 

t  

m  

2  

m  

d  

6

2

2

 

S  

a  

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

2

 

E  

f

 

 

 

 

2

 

R  

G  

J  

s  

t  

o  

o  

b  

a  

a  

fi  

t  

d  

i

2

2

 

w  

q  

v  

C  

v  
the various sub-periods but with most of the model parameters

kept at their optimal values while a few of them with physical

meanings such as the curve number (CN) are allowed to change.

Details on CN can be found in Rawat and Singh, (2017) . The idea

would be to determine whether variation in the relevant param-

eters like CN can explain changes in simulated flow. 

Therefore, this study aimed at undertaking SWAT-based quantitative

nvestigation of the contributions from LULC changes, climate variabil-

ty and other factors to the changing river flows of RMC. In this study,

pplication of SWAT was based on a number of LULC maps under the as-

umption of constant climatic conditions over a particular period. LULC

hanges were deemed to impact on runoff through differential alteration

n evaporation rate across the catchment. Due to limitation of observed

ata, the use of reanalysis or satellite rainfall products was explored

rior to the hydrological modeling. 

. Materials and methods 

.1. Study area 

River Mpanga in Uganda within East Africa originates from the foot

f the Mountain Rwenzori. RMC is located in the southwestern part of

ganda ( Fig. 1 ). River Mpanga is endorheic and flows through the dis-

ricts of Kabarole, Kyenjojo and Kamwenge before discharging into Lake

eorge. RMC has various sub-catchments including Upper Mpanga, Mid-

le Mpanga, Lower Mpanga, and Rushango with drainage areas equal to

84, 1174, 477, and 3170 km 

2 , respectively. The RMC area up to Lake

eorge is 5205 km 

2 . Sub-catchment areas upstream of hydrological sta-

ions at Kampala-Fort Portal and Fort Portal-Ibanda roads are 401 and

484 km 

2 , respectively ( Fig. 1 ). Elevation from the source of the river to

he gauge station considered in this study ranges from 1153 m to nearly

000 m above sea level. 

The River Mpanga is the source of domestic piped water supplied to

 number of urban areas in the Southwestern part of Uganda including

ort Portal city, Kamwenge and Ibanda towns. Furthermore, the river is

mportant for the Mpanga Falls hydroelectric power station. Mpanga hy-

ropower plant was commissioned in 2011 in Kamwenge District with

 capacity of 18 Megawatts to serve 20000 households ( Daily Monitor

011 ). The hydropower plant has not been able to generate power to

ts full capacity throughout the year ( M. Ministry of Water and Envi-

onment 2014 ) due to reduced volumes of water flowing in the river

specially in dry seasons. 

At the foot of Rwenzori Mountain where River Mpanga emanates,

he climate is wet year-round with annual rainfall up to about 3000

m in some years. Annually, the study area receives rainfall of about

500 mm on average ( Ministry for Water and Environment 2020 ). On a

onthly scale, average temperature varies over the range 27–31°C and

ue to high evaporation rates, average monthly humidity varies from

0 to 80% ( National Environment Management Authority 2009 ). 

.2. Data 

.2.1. Hydro-meteorological data 

Hydro-meteorological data required for hydrological modeling using

WAT included precipitation, wind, relative humidity, solar radiation,

nd river discharge. These datasets were obtained from various sources.

i) Daily river flows observed at gauge stations numbered 84215

(Fort Portal-Ibanda Road) and 84212 (Kampala-Fort Portal Road)

with series over the periods 2000–2011 and 1999–2018, respec-

tively, were obtained from Ministry of Water and Environment

in Uganda. 

ii) Daily rainfall data at four weather stations including Kiburara

Prison Farm, Isunga Estate, Kyehara, and Kilembe Mines (see

Fig. 1 for their locations) were obtained from Uganda National
3 
Meteorological Authority. Reliable data from these stations used

in this study were between 1999 and 2012. 

iii) Due to few stations at which observed rainfall series were

obtained, meteorological datasets were required at many

other locations within the catchment. One option would have

been to interpolate observed rainfall data to fine spatial res-

olution or grid points. However, the interpolation would be

characterized by uncertainty due to the few weather stations

available for this study. Eventually, meteorological series

were obtained from other sources. Daily rainfall (mm/day)

series were obtained from the Japanese 55-year Reanalysis

(JRA-55) ( Kobayashi et al., 2015 ), Climate Hazards group

InfraRed Precipitation with Stations (CHIRPS) version 2.0 (or

CHIRPS v2.0) ( Funk et al., 2015 ) and National Centers for

Environmental Prediction’s (NCEP’s) Climate Forecast System

Reanalysis (CFSR) ( Saha et al., 2014 ). The data for JRA-55

(1°×1° grid), CHIRPS v2.0 (0.25°×0.25° grid), and CFSR

(0.25°×0.25°) were over periods 1958–2017, 1981–2019, 1979–

2013, respectively. The JRA-55, CHIRPS v2.0 and CFSR were

downloaded via ftp://ftp.climserv.ipsl.polytechnique.fr/FROGs

(accessed: 15 th December 2020), https://climexp.knmi.nl/

selectdailyfield2.cgi?id = someone@somewhere (accessed: 27 th 

May 2021), and https://globalweather.tamu.edu/ (accessed:

27 th January 2021), respectively. Other CFSR series downloaded

were daily, wind (m/s), minimum and maximum temperature

(°C), relative humidity, and solar radiation (MJ/m 

2 ). These

additional CFSR datasets also covered the period 1979–2013. 

.2.1. Topographical data and soil map 

Some of the spatial data required to run SWAT included the Digital

levation Model (DEM), and soil map. These datasets were obtained

rom different sources. 

i) The hole-filled DEM of 30 m ×30 m ( Jarvis et al., 2008 ) was

downloaded online via https://lta.cr.usgs.gov (accessed: 4 th June

2019). 

ii) The soil map was acquired from the soil database of the United

States’ Food Agricultural Organization ( FAO-UNESCO 2003 ). The

soil map was at a scale of 1:5,000,000 ( FAO-UNESCO 1977 ). 

.2.2. LULC map 

Another spatial data required to run SWAT comprised LULC map.

emotely sensed LULC data can be downloaded from the United States

eological Survey website https://earthexplorer.usgs.gov (accessed: 12

uly 2021). In this case, one needs to classify LULC types either using

upervised or unsupervised classification tool. However, in this study,

he classified 2000, 2008 and 2014 National Land Cover maps at scale

f 1:250,000 were obtained from the National Forestry Authority (NFA)

f Uganda. In the next step, each LULC map was reclassified into a num-

er of important classes in the SWAT LULC classification table including

griculture/cropland, pasture/grasslands, human settlements (built-up

reas), forested areas, water bodies, wetlands and other lands. Reclassi-

cation involved modifying values from the obtained raster into classes

hat characterized LULC types corresponding to those of SWAT LULC

atabase. The reclassified LULC maps 2000, 2008 and 2014 were here-

nafter denoted as LuM2000, LuM2008, and LuM2014, respectively. 

.3. Hydrological modeling using SWAT 

.3.1. Selection of the rainfall series 

As mentioned before, sufficient number of observed rainfall series

as lacking for the study area. Furthermore, it was deemed that the

uality of the data from the various sources (JRA-55, CFSR, and CHIRPS

2.0) was not the same. Another series coined from JRA-55, CFSR, and

HIRPS v2.0 (and hereinafter denoted as JCC) was used. Consider that u ,

 and w denote values of JRA-55, CFSR, and CHIRPS v2.0, respectively.

http://ftp://ftp.climserv.ipsl.polytechnique.fr/FROGs
https://climexp.knmi.nl/selectdailyfield2.cgi?id=someone@somewhere
https://globalweather.tamu.edu/
https://lta.cr.usgs.gov
https://earthexplorer.usgs.gov


C. Onyutha, C. Turyahabwe and P. Kaweesa Environmental Challenges 5 (2021) 100273 

Fig. 1. Location of River Mpanga as well as 

rainfall and streamflow measuring stations. 

F  

t  

t  

t  

F

𝐽  

w  

T  

r  

r  

O  

o

2

 

fi  

C  

b  

m  

i  

b  

a  

A  

t  

2  

i  

c  

t  
urthermore, let q represent observed daily river flow while n denotes

he sample size of u , v or w . Since reanalysis datasets tend to underes-

imate observed extreme rainfall, a threshold of q 75 m 

3 /s (representing

he 75 th percentile of observed river flow) was selected to compute JCC.

inally, the i th value of JCC was computed using 

𝐶 𝐶 𝑖 = 

{ 

3 −1 ×
(
𝑢 𝑖 + 𝑣 𝑖 + 𝑤 𝑖 

)
if q 𝑖 < q 75 

𝑚𝑎𝑥 
(
𝑢 𝑖 , 𝑣 𝑖 , 𝑤 𝑖 

)
if q 𝑖 ≥ q 75 

for 1 ≤ 𝑖 ≤ 𝑛 (1)

here max denotes the maximum of the three i th values of u , v or w .

o obtain the most suitable reanalysis to use along with the observed

ainfall, analysis of correlation was performed between river flow and

ainfall series of (i) JRA-55, (ii) CFSR, (iii) CHIRPS v2.0, and (iv) JCC.

ther SWAT meteorological inputs apart from rainfall series were based

n the CFSR data. 
4 
.3.2. Model build-up 

SWAT 2012 version was used for the hydrological modeling. The

rst step in the model setup was to delineate the catchment using DEM.

atchment delineation was done to determine the area of land drained

y a river and its tributaries. To identify drainage patterns, the catch-

ent was divided into sub-basins. The sub-basins were further divided

nto smaller parts called hydrologic response units (HRUs). HRUs can

e taken to mean unique combinations of LULC types, soil attributes,

nd/or slope classes distributed over a sub-basin ( Neitsch et al., 2011 ).

fter creation of HRUs, the next step was to import weather data into

he model. Lastly, the model was run on daily time step with data from

000 to 2005. SWAT requires warm up period at the beginning of the

nput series. Since the data record period was short in this study, we

reated and transferred copies of input series for three years (from 2002

o 2004) to the beginning of the datasets. This procedure lengthened



C. Onyutha, C. Turyahabwe and P. Kaweesa Environmental Challenges 5 (2021) 100273 

Table 1 

Conditions for differences among sets of model outputs. 

Combination Assumed simulation condition Cause(s) of the differences in sets of model outputs 

(1) (2) (3) 

i) True True False Changes in LUCL types 

ii) True False True Changes in climatic conditions 

iii) False True True Equifinality of model parameters 

iv) False False True a) Equifinality of model parameters 

b) Changes in climatic conditions 

v) False True False a) Equifinality of model parameters 

b) Changes in LUCL types 

vi) True False False a) Changes in climatic conditions 

b) Changes in LUCL types 

vii) False False False a)Equifinality of model parameters 

b) Changes in climatic conditions 

c) Changes in LULC types 

Assumed simulation conditions: 

(1) Optimal set of parameters is kept constant during each simulation. 

(2) The same hydrometeorological inputs are used in each simulation. 

(3) The various LULC maps are totally identical. 

t  

w  

c  

i  

f  

a  

t  

(

2

 

t  

c  

t  

c  

c  

(

 

w  

s  

c  

a

𝑁  

𝑃  

𝑅  

𝑅  

w  

𝑠  

n  

A  

s

 

r  

d  

c  

u

2

 

p  

t  

2  

m  

v  

b

2

 

u  

L  

t  

r  

i  

w  

a  

o  

 

r  

q  

s  

e  

e  

w  

s  

u  

s  

a  

a  

t  

i  

o  

u  

p  

t

2

v

 

c  

t  

i  

n  

t  
he input series by 3 years. Thus, data of 3-years was used for the model

arm up. The methods available in SWAT to simulate excess rainfall in-

lude CN and Green-Ampt approach. In this study, CN was used due to

ts relevance for the study objectives. Assessment of inputs and outputs

rom SWAT was made based on the information provided by Arnold et

l., (2012) . Before calibration, sensitivity analysis was performed using

he global approach in semi-automated Sequential Uncertainty Fitting

SUFI2) algorithm ( Abbaspour, 2015 ). 

.3.3. Calibration and validation 

SWAT was driven by daily hydro-meteorological data from 2000

o 2005 with the spatial data comprising soil map and LuM2000. For

alibration, SUFI-2 method within the SWAT Calibration and Uncer-

ainty Procedures (SWAT-CUP) ( Abbaspour et al., 2007 ) was used to

alibrate SWAT against observed daily flow. SUFI-2 method tends to be

ommonly applied due to its suitability for calibrating SWAT (see e.g.

 Kumar et al., 2017 ), ( Narsimlu et al., 2015 )). 

Since this study considered variability of river flow (a phenomenon

hich comprises both high and low flows), model performance was as-

essed in terms of the Nash-Sutcliffe Efficiency (NSE) ( Nash and Sut-

liffe, 1970 ), percentage bias (PBias), root mean squared error (RMSE)

nd coefficient of determination ( R 

2 ) such that 

𝑆𝐸 = 1 − 

∑𝑛 

𝑖 =1 
(
𝑞 𝑖 − 𝑠 𝑖 

)2 
∑𝑛 

𝑖 =1 
(
𝑞 𝑖 − 𝑞 

)2 (2)

 𝐵𝑖𝑎𝑠 = 

∑𝑛 

𝑖 =1 
(
𝑞 𝑖 − 𝑠 𝑖 

)
∑𝑛 

𝑖 =1 𝑞 𝑖 
× 100 (3)

𝑀𝑆𝐸 = 

√ 

𝑛 −1 ×
∑𝑛 

𝑖 =1 

(
𝑞 𝑖 − 𝑠 𝑖 

)2 
(4)

 

2 = 

(∑𝑛 

𝑖 =1 
(
𝑞 𝑖 − 𝑞 

)(
𝑠 𝑖 − ̄𝑠 

))2 
∑𝑛 

𝑖 =1 
(
𝑞 𝑖 − 𝑞 

)2 ∑𝑛 

𝑖 =1 
(
𝑠 𝑖 − ̄𝑠 

)2 (5)

here q denotes observed data, s represents modeled series, while ̄𝑞 and

̄ are the mean values of the q i ’s and s i ’s, respectively. NSE varies from

egative infinity to one while the values of R 

2 exist over the range 0 –1.

pplication of R 

2 is based on the assumption that observed modelled

eries are linearly related ( Onyutha, 2020 ). 

Validation requires driving a model using independent data (or se-

ies not used for calibration). In this study, SWAT was validated using

aily hydro-meteorological data over the period 2006–2011. Like for

alibration, performance of the model during validation was assessed

sing NSE and R 

2 . 
5 
.3.4. Simulations 

Optimal parameter values obtained during calibration were used to

arameterize SWAT. In the next step, there were three SWAT simula-

ions. In other words, SWAT was driven by daily series over the period

006–2011 but separately run using LuM2000, LuM2008 and LuM2014

aps. During each simulation, parameters were fixed at their optimal

alues and the soil data remained the same as those used during cali-

ration. 

.3.5. Contribution from changes in LULC types to river flow variation 

In attribution of hydrological changes, it is expected that the sim-

lated river flow series obtained based on LuM2000, LuM2008, and

uM2014 would be the same so long as (1) the optimal set of parame-

ers is kept constant during each simulation, (2) the same hydrometeo-

ological inputs (such as precipitation and PET series) are used as model

nputs, and (3) there are no differences among the LULC maps (in other

ords, the spatial information from LuM2000, LuM2008, and LuM2014

re totally the same). However, sets of simulated results can vary based

n whether any of the conditions (1)–(3) are true as detailed in Table 1 .

If a model is calibrated several times, the various sets of model pa-

ameters (when assumed conditions 2 and 3 are true) can be used to

uantify uncertainties on simulated flows. This can be done, for in-

tance, through the generalized likelihood uncertainty estimation strat-

gy ( Beven and Binley, 1992 ). It becomes complex to unravel the syn-

rgistic contributions of many conditions, like combinations (iv)–(vi),

hen considered at the same time. Thus, combinations (i)–(ii) were con-

idered in this study. For combination (i) in Table 1 , there were two sim-

lations, one based on LuM2008 and the other using LuM2014. The ab-

olute difference between the mean of model outputs based on LuM2000

nd the mean of simulated series obtained using LuM2008 (as a percent-

ge of the mean of modeled series based on LuM2000) was considered

o be due to the changes in LULC types over the period 2000–2008. Sim-

larly, the ratio of the absolute difference between the mean of model

utputs based on LuM2000 and the mean of simulated series obtained

sing LuM2014 to the mean of modeled series based on LuM2000 (ex-

ressed in percentage) was considered to be due to the changes in LULC

ypes over the period 2000–2014. 

.3.6. Contribution from changes in climatic conditions to river flow 

ariation 

The null hypothesis H 0 (no correlation) between river flow and

atchment-wide averaged rainfall was tested. This was on the assump-

ion that river flow changes can largely be attributed to rainfall variation

f the overall rainfall-runoff generation processes of the catchments are

ot significantly impacted upon, for instance through the human ac-

ivities like changes in LULC types ( Onyutha and Willems, 2018 ). The



C. Onyutha, C. Turyahabwe and P. Kaweesa Environmental Challenges 5 (2021) 100273 

Fig. 2. Mean of long-term (2000-2011) monthly rainfall total in RMC. 

s  

fl  

r  

c  

c  

r  

t  

v  

m  

h  

i

g  

t  

c  

b  

m  

c  

t  

n  

p  

t  

c  

m  

c  

p  

t  

fl

 

i  

t  

e  

C  

t  

o

1  

B  

a

3

3

 

e  

I  

"  

S  

F  

"  

m  

H  

t

 

C  

fl  

r  

t  

r  

B  

R  

(  

r  

d  

t

2  

w  

e  

m  

a  

T  

K  

i  

m  

P  

t  

c

3

 

a  

c  

w  

N  

t  

c  

S

 

s  

s  

o  

c  

n  

f

econd assumption was that the relationship between rainfall and river

ow is linear. It turns out that there are several factors which influence

ainfall-runoff generation such as rates of infiltration, evaporation, per-

olation, river water abstraction, and volume of industrial effluents dis-

harged into the river. Thus, the relationships among these factors and

iver flow are complex and may not necessarily be linear in nature. Fur-

hermore, variation of rainfall alone cannot explain the total river flow

ariance. Therefore, to characterize the influence of the variation in cli-

atic conditions on river flow changes, variables such as precipitation,

umidity, temperature or PET should be considered. In this way, if we

gnore the limitation of the model to accurately capture rainfall-runoff

eneration processes (due to imperfections in model structure, parame-

er uncertainty, and observation errors on inputs) and assume that the

hanges in LULC types or river flow is minimal, the simulated series can

e considered to be indicative of the contribution from the variation in

eteorological conditions on river flow changes. In other words, if the

atchment remains in its natural form without impacts of human activi-

ies, and assuming climatic variables such as precipitation and PET have

o observational error, the river flow would naturally be accurate if a

erfect model is used. Here, we consider model outputs to characterize

he various rainfall-runoff generation processes and changes in climatic

onditions to be captured by the model structure. The assumption of

inimal impacts of LULC changes on river flow changes means that we

an use one LULC map for the hydrological modeling. Thus, model out-

uts over validation period 2006–2011 based on LuM2000 were used to

est the H 0 (no correlation) between daily simulated and observed river

ow. 

Consider that the contributions due to climate variability, changes

n LULC types, and other factors are A , B and C %, respectively. Let

he amount of the total variance in observed flow explained by mod-

led series to be D %. After quantifying the term B , the value of

 can be given by C = (100–B %). Finally, the term A can be ob-

ained using A = D ×C/ 100. It is vital to check that A + B + C % ≤ 100%,

therwise the values should be rescaled to be within the range 0–

00%. For instance, rescaling can be done using A r = A /( A + B + C ) ×100,

 r = B /( A + B + C ) ×100, and C r = C /( A + B + C ) ×100 where A r , B r and C r %

re rescaled values of A , B and C %, respectively. 

. Results and discussion 

.1. Rainfall and river flow properties 

Fig. 2 shows monthly rainfall pattern based on catchment-wide av-

raged series. The rainfall across the study area is of bimodal pattern.

n other words, there exist two rainy seasons ( Fig. 2 ). The "long" and

short" rains occur over the months of March-April-May (MAM) and

eptember-October-November (SON), respectively. December-January-
6 
ebruary (DJF) and June-July-August (JJA) seasons are characterized by

short" and "long" dry conditions, respectively. The bimodal pattern of

onthly rainfall was well reproduced by series from the various sources.

owever, observed monthly totals were more comparable (in magni-

ude) to CFSR series than those of JRA-55 and CHIRPS v2.0. 

Fig. 3 shows variation in annual rainfall and river flows over time.

onsidering data the station at Fort Portal-Ibanda Road, annual river

ows increased steadily from 2000 to 2011 ( Fig. 3 a). River flow data

ecording at this station stopped before the end of 2012 because of

he destruction during construction of Fort Portal-Ibanda road. The

ate of increase in the river flow at this station was 0.47 cumecs/year.

oth observed and JCC-based annual rainfall series averaged over the

MC were characterized by increasing trends like for the river flows

 Fig. 3 c-d). The rates of increase in observed and JCC annual rainfall se-

ies were 52.08 and 56.93 mm/year, respectively. However, river flow

ata observed at Fort Portal-Kampala road station ( Fig. 3 b) showed that

he mean values of the sub-series over the periods 2000–2010 and 2011–

018 were different. River flows at this hydrological station ( Fig. 3 b)

ere found to be unusually high in 2011 and 2012 (and 2013 to some

xtent). Whereas such unusual flow data could indicate the effect of hu-

an intervention on hydrology, both observed and JCC annual rainfall

cross the RMC did not show step jump in mean in 2011 ( Fig. 3 c, d).

hus, the inhomogeneity in the river flow observed at the Fort Portal-

ampala road station required application of correction factors before

ts use for analysis. Given that a bigger catchment area was required for

odeling and due to the questionable river flow data at Kampala-Fort

ortal Road hydrological station especially over the period 2011-2013,

he station at Fort Portal-Ibanda Road was adopted as the outlet of the

atchment considered for this study. 

.2. Soil and LULC maps 

Fig. 4 shows spatial information required as SWAT inputs and for

ssessment of the changes in LULC types. Geo-referenced and clipped

atchment soil map ( Fig. 4 a) produced three (5) original soil groups

hich fell under three FAO soil categories including Gleyic Aerosols,

itisols and Lixic Ferrasols and they covered 7.6%, 34.6% and 57.8% of

he catchment area, respectively. The SWAT database names of FAO soil

ategorized as Gleyic Aerosols, Nitisols and Lixic Ferrasols were Benson,

wanton, and Weider, respectively. 

Some of the common LULC types included bushland, cropland, and

paces taken up for human settlement ( Fig. 4 b–d). To simplify analysis,

ome of the LULC types were grouped. For instance, closed bushland and

pen bushland were grouped as bushland. Similarly, open grassland and

losed grassland were combined to become grassland. Finally, moderate

atural forest, dense natural forest, plantation forest, and sparse natural

orest were combined into one class called forest. 



C. Onyutha, C. Turyahabwe and P. Kaweesa Environmental Challenges 5 (2021) 100273 

Fig. 3. Time series plots of annual (a-b) river flows, and (c-d) rainfall of RMC. 

 

(  

l  

m  

s  

c  

m  

f  

c  

a  

l  

s  

w  

O  

t

 

o  

i  

U  

1  

p  

e  

p  

t  

t  

U  

r  

b  

(  

l  

t  

p  

c  

o  

e  

Table 2 

Population changes in some districts which share RMC. 

SNo District Year 

1991 2002 2014 

1 Kabarole 299,573 356,914 469,236 

2 Kamwenge 201,654 263,730 414,454 

3 Kyenjojo 182,026 266,246 422,204 

f  

e  

j  

c  

(  

t  

r  

U  

m  

m  

p  

t  

U  

t

 

t  

i  

c  

b  

r  

t  

e  

2  

u  

p  
Fig. 5 shows distribution of the various LULC types. In 2000

 Fig. 5 a), the catchment area mainly comprised grassland (39.5%), crop-

and or farmland (33.0%), and forest (22.0%). Bushland was taken to

ean land that supports remnant vegetation or covered with a few

hort trees, shrubs, or natural vegetation. In cropland areas, the main

rops grown in the study area comprise bananas, and cereals (maize,

illet, sorghum) as well as a few perennial cash crops, such as cof-

ee, tea and fruit trees. Based on LuM2008 ( Fig. 5 b), much area was

overed by cropland/farmland (69.2%) followed by grassland (12.5%),

nd forest (11.8%). In 2014 ( Fig. 5 c), dominant LULC types were crop-

and/farmland (72.2%), forest (10.7%), and grassland (10.4%). Fig. 5 d

hows summary of the LULC changes in the study area. LULC types

hich increased in area were cropland, built-up areas, and woodland.

ther LULC types including forest, grassland, and wetland reduced in

heir areas of coverage. 

We think that the temporal alterations in LULC types in RMC and

ther parts of Uganda can generally be thought of in terms of laws and

nstitutional policies or transition in land tenure. The key land laws after

ganda got independence included the 1969 Public Land Act and the

975 land Reform Decree. According to the 1969 Public Land Act, all

ublic lands were brought under the Uganda Lands Commission. How-

ver, the 1975 Land Reform Decree declared all land in Uganda to be

ublic thereby abolishing the mailo tenure. Article 237 (1) of the Consti-

ution of Uganda 1995 as well as Article 3 of the Land Act 1998 declared

hat "land in Uganda belonged to the citizens of Uganda" ( Republic of

ganda 1995 ), ( Republic of Uganda 1998 ). This declaration meant that

adical titles of land were given to the citizens something which had

een abolished by the 1975 Land Reform Decree. The 1998 Land Act

 Republic of Uganda 1998 ) was aimed at reforming the land tenure re-

ations in Uganda by recognizing land tenure systems including the cus-

omary (communal), mailo , freehold and leasehold. Mailo tenure com-

rised allotments of land to the Kabaka (the King of Buganda) and his

hiefs following the 1900 agreement. Therefore, there were increased

pportunities for citizens in the early 2000s in utilizing their land. How-

ver, after 2000 there were escalating conflicts, land grabbing and force-
7 
ul evictions of tenants by land owners. To address issues of widespread

victions, the Land (Amendment) Act 2010 was instituted with the ob-

ective of enhancing security of occupancy of the lawful bona fide oc-

upants (or tenants on registered land plus people on customary land)

 Republic of Uganda 2010 ). In February 2013, the final document of

he National Land Policy of Uganda Republic of Uganda, (2013) was

atified. One of the principles underpinning the National Land Policy of

ganda 2013 was that "Land must be productively used and sustainably

anaged for increased contribution to economic productivity and com-

ercial competitiveness" Republic of Uganda, (2013) . Another princi-

le was that "Management of land resources must mitigate environmen-

al effects, reverse decline in soil quality and land quality" Republic of

ganda, (2013) . Therefore, changes in land laws could have contributed

o the LULC changes. 

Another important reason for the changes in LULC across RMC was

he population increase. To show that population of the RMC has been

ncreasing, we can considered a few districts into which RMC extends in-

luding Kabarole, Kyenjoyo, Kamwenge districts among others ( Table 2 )

ased on information from the Uganda Bureau of Statistics ( Uganda Bu-

eau of Statistics 2016 ). Population increase could partly be ascribed to

he massive influx of refugees (particularly the Congolese) into the west-

rn Uganda especially over the last two decades. Pursuant to Uganda

006 Refugees Act ( Republic of Uganda, 2006 ) and 2010 Refugees Reg-

lations ( Republic of Uganda, 2006 ), each refugee family was given a

iece of land for their own exclusive use in terms of agriculture. Increase



C. Onyutha, C. Turyahabwe and P. Kaweesa Environmental Challenges 5 (2021) 100273 

Fig. 4. Spatial data of a) soil map, and b) LuM2000, c) LuM2008, and d) LuM2014. 

i  

r  

2

 

i  

a  

a  

2  

2  

R

2  

i  

A  

2  

f  

g  

t  

N  
n population prompted deforestation and the clearing of grasses along

iver banks within the RMC mainly for cultivation and settlement ( BRLi

015 ). 

The translation of pressure from population increase into changes

n LULC types across East Africa as a region where the present study

rea is located has been well documented; see for instance ( Rientjes et

l., 2011 , Bewket and Sterk, 2005 , Kizza et al., 2017 ), ( Mwangi et al.,

016 ), ( Kashaigili and Majaliwa, 2013 ), ( Enku et al., 2014 , Wagesho,

014 , Barasa et al., 2011 ). For the Gilgel Abbay catchment of Ethiopia,
8 
ientjes et al. ( Rientjes et al., 2011 ) showed that in the period 1986–

001 forest land decreased from 32.9% to 16.7% while agricultural land

ncreased from 40.2% to 62.7 %. Furthermore, forest cover area in Gilgel

bay was 51% in 1973 but decreased to 17% of the catchment area in

001 ( Enku et al., 2014 ). Bilate catchment in Ethiopia exhibited 68%

orest cover loss over the period 1976–2000 ( Wagesho, 2014 ). Nyan-

ores catchment in Kenya registered 50% forest cover loss from 1976

o 2006 ( Mwangi et al., 2016 ). In Chemoga watershed within the Blue

ile basin, there was agricultural conversion of 79% of the Riverine



C. Onyutha, C. Turyahabwe and P. Kaweesa Environmental Challenges 5 (2021) 100273 

Fig. 5. LULC types based on a) LuM2000, b) LuM2008, c) LuM2014, and associated d) LULC changes. The vertical axis of the chart (d) was plotted in logarithmic 

scale for clarity. 

f  

M  

0  

W  

f  

f  

b  

m  

b  

K  

i  

m  
orests in about 40 years (1957–1998) (Bewket and Sterk, 2005). In

alagarasi catchment of Tanzania, forest cover increased in area by

.4% over the period 1984–2001 ( Kashaigili and Majaliwa, 2013 ). In

eru-Weru catchment of Tanzania, there was a forest cover loss by 12%

rom 1990 to 2008 ( Chiwa, 2008 ). Small-scale farmland gained variably

rom all the LULC types across the Lake Bunyonyi catchment in Uganda
9 
etween 1999 and 2005 (Kizza et al., 2017). In the Murchison Bay catch-

ent, 26.7% increase in runoff over over the period 1997-2008 could

e ascribed to the rapidly changing LULC types ( Anaba et al., 2017 ). In

anungu District of the southwestern Uganda, areas covered by Trop-

cal high forest decreased by 16% between 1975 and 1987 while the

agnitude of small scale farming increased by 5% from 1975 to 1999



C. Onyutha, C. Turyahabwe and P. Kaweesa Environmental Challenges 5 (2021) 100273 

Table 3 

LULC change summary for RMC. 

ReclassifiedLULC Total area (ha) Fractionof catchment (%) Change (%) 

2000 2008 2014 2000 2008 2014 2000–2008 2008–2014 

Bushland 6047 7087 7843 2.8 3.3 3.6 0.5 0.3 

Grassland 85401 27068 22536 39.4 12.5 10.4 -26.9 -2.1 

Forest 47710 25474 23181 22.0 11.8 10.7 -10.3 -1.1 

Woodland 1404 3000 3946 0.6 1.4 1.8 0.7 0.4 

Cropland/Farmland 71435 149778 156312 33.0 69.1 72.2 36.2 3.0 

Settlement 395 750 1036 0.2 0.3 0.5 0.2 0.1 

Open Water 125 118 110 0.1 0.1 0.1 < 0.1 < 0.1 

Wetland 4012 3328 1639 1.9 1.5 0.8 -0.3 -0.8 

Table 4 

List of sensitive parameters. 

No. Parameter name Description t -Stat p -value 

1 CN2 Soil moisture condition II curve number -6.323 < 0.001 

2 SURLAG Surface runoff lag coefficient -2.567 0.018 

3 GW REVAP Groundwater evapotranspiration coefficient -2.232 0.037 

4 OV_N Manning’s " n " value for overland flow -1.678 0.109 

5 GW DELAY Ground delay (days) 1.445 0.164 

6 ALPHA BF Base flow alpha factor (days) -1.223 0.236 

7 SOL AWC Available water capacity of the soil layer (mm H 2 O /mm soil) -0.898 0.380 

(  

R  

c  

b

 

t  

a  

u  

a  

l  

a  

f  

t  

t  

a  

w  

1  

t  

t  

t  

i  

o  

T  

t  

m  

l

3

3

 

f  

s  

c  

b  

b  

o  

p  

o

3

 

t  

w  

b  

M  

C  

s  

l

3

 

a  

(

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 Barasa et al., 2011 ). Forested area (cropland as a percentage of the

iver Muzizi catchment within the same region where our study area de-

reased (increased) from 41.48% (40.16%) in 2000 to 31.12% (50.02%)

y 2014 ( Bahati et al., 2021 ). 

Percentage changes in LULC types can be found in ( Table 3 ). Be-

ween 2000 and 2008, areas covered by cropland, settlement, woodland

nd bushland increased. The largest increment (36.2%) was in the area

nder farmland/cropland. Areas of grassland, forest areas, open water

nd wetlands decreased. The outstanding decline (26.9%) was for grass-

and. This was followed by forests (10.3%). Between 2008 and 2014, the

rea under cropland/farmland increased by 3.0%. However, grassland,

orest, and wetland areas reduced by 2.1%, 1.1%, and 0.78%, respec-

ively. The decline in forest cover was due to increased cutting down of

rees for various uses such as firewood and timber ( Ministry of Water

nd Environment, 2013 ). Percentage of the catchment area covered by

etland (open water) in 2000, 2008 and 2014 was 1.85% (0.058%),

.54% (0.054%), and 0.76% (0.051%), respectively. The decrease in

he areas covered by open water and wetland decreased over time in

he RMC reflects the increasing level of encroachment on wetlands over

he study period. Wetland encroachments are due to failure by the cit-

zens to adhere to the relevant policies or regulations for conservation

r management of wetlands, riverbanks and lakeshores in the country.

hus, the National Environment Management Authority in liaison with

he Ministry of Water and Environment should devise and implement

easures that can impede further increasing trend in the levels of wet-

and encroachments. 

.3. Hydrological modeling 

.3.1. Selection of the rainfall series to use for modeling 

The coefficients of correlation between daily river flow and rain-

all from CFSR, CHIRPS, and JRA55 were 0.199, 0.098 and 0.204, re-

pectively. This indicated that performances of CFSR and JRA-55 were

omparable and better than that of CHIRPS v2.0. However, correlation

etween river flow and JCC was 0.439. This showed that series obtained

y combining various rainfall products could be better correlated with

bserved rainfall than in the case when reanalysis or satellite rainfall

roduct is used individually. Eventually, JCC-based rainfall was used at

ther locations where there were observed data were lacking. 
10 
.3.2. Sensitivity analysis 

Table 4 shows results of sensitivity analysis in terms of Student’s t -

est statistic and probability or p -values. The most sensitive parameter

as CN2 followed by SURLAG, and GW REVAP. CN2 was also found to

e the most sensitive in previous studies (see e.g. ( Anaba et al., 2017 ),

utenyo et al., (2011) ) that applied SWAT to two catchments in Uganda.

N2 tends to be a function of soil permeability, LULC and the antecedent

oil moisture thereby influencing runoff generation. An increase in CN2

eads to large volumes of stream flow. 

.3.3. Calibration and validation 

Fig. 6 and Table 5 show results of model performance. It is notice-

ble that the model captured well the variation in observed river flow

 Fig. 6 ). Results in Fig. 6 and Table 5 indicated that: 

i) the model performed well and could therefore be applied to quan-

titatively investigate the contributions of changes in climatic con-

ditions and LULC types to the temporal variation in river flow of

the RMC, and 

ii) the use of reanalysis or satellite products to drive SWAT for RMC

comprises a partial solution to the problem of data limitation that

hampers hydrological modeling studies on investigating changes

in water resources especially in Uganda (and perhaps other de-

veloping countries). We remark that apart from CFSR, JRA-55

and CHIRPS v2.0 there are several other reanalysis or satellite

datasets which can be applied for hydrological modeling in data

scarce regions. For instance, Mubialiwo et al. (2021) applied

daily precipitation, minimum and maximum temperature data

from the Princeton Global Forcing ( Sheffield et al., 2006 ) (af-

ter bias correction) to River Malaba sub-catchment and obtained

NSE of up to 0.8. Capacity to reproduce observed climatology (for

instance, with respect to extreme climatic events) varies among

existing reanalysis or satellite products. For instance, CHIRPS was

found to perform better than CFSR in reproducing observed rain-

fall across the River Muzizi catchment Bahati et al. (2021) . Differ-

ences among reanalysis or satellite products can be large when

(i) fine (e.g. hourly or daily) scale is used, and (ii) focus is on

extreme climatic events. A part from random errors, reanalysis

or satellite products can also be characterized by bias. The bias,

for instance in satellite products, could stem from a number of



C. Onyutha, C. Turyahabwe and P. Kaweesa Environmental Challenges 5 (2021) 100273 

Fig. 6. Time series plot of observed versus simulated daily flow. 

Fig. 7. Attributes of river flow changes in terms of average daily flow in each month and contributions from a) LULC changes and b) other factors such as climate 

variability. 

Table 5 

Statistical “goodness-of-fit ” measures. 

Metric Period 

Calibration(2000-2005) Validation(2006-2011) Full data record (2000-2011) 

NSE 0.769 0.751 0.761 

PBias -0.018 -0.002 -0.010 

R 2 0.779 0.777 0.763 

RMSE (m 

3 /s) 5.451 5.632 5.542 

 

 

 

 

 

3

 

c  

r  

f  

(  
sources such as imperfections in rainfall retrieval algorithms, and

poor distribution of sensors. This means that bias correction may

be required in reanalysis or satellite products. Finally, careful se-

lection of which reanalysis or satellite product should be used

can be made on a case by case basis with respect to the purpose

of the study. 
11 
.3.4. Contribution from changes in LULC types to river flow variation 

Fig. 7 shows attributes of mean monthly flow. Contributions by LULC

hanges to the variation in river flow over the entire period 2000–2014

anged from 3.3% (November) to 9.4% (August) ( Fig. 7 a). However,

rom 2000 to 2008 the contributions varied from 8.6% (April) to 13.5%

August). On average, the amount of contribution from LULC changes



C. Onyutha, C. Turyahabwe and P. Kaweesa Environmental Challenges 5 (2021) 100273 

w  

s  

2  

p  

c  

(

3

v

 

w  

t  

t

2  

fl  

w  

l  

e  

fl  

c  

s  

c

 

o  

c  

b  

a  

c  

r  

a  

o  

c  

r  

i  

fl  

i  

p  

r  

c  

i

4

 

c  

R  

s  

m

 

l  

t  

r  

c  

s  

0  

w  

l

 

s  

o  

fl  

o  

s  

t

 

a  

i  

t  

r  

l  

t  

f  

i  

M  

s  

t  

B  

t  

c  

o  

p  

t  

w

 

t  

w  

t  

t  

f  

o  

m  

e  

u  

s  

d  

c  

w  

w  

e

 

m  

o  

l  

r  

w  

t

F

A

 

i  

e  

c  

w  

y

D

A

 

5  

g  

C  

m  

(  
as 9.1%% and 6.2% over the periods 2000–2008 and 2000–2014, re-

pectively. Thus, 7.65% of the changes in river flow over the period

000–2014 was due to changes in LULC types on average. The remaining

ercentage in each month was due to other factors. These other factors

ollectively contributed close to 90% considering the period 2002–2014

 Fig. 7 b). 

.3.5. Contribution from changes in climatic conditions to river flow 

ariation 

The percentage of the total variance in observed daily flow which

as explained by the variation in catchment-wide average rainfall over

he period 2000–2011 was 43.9%. The R 

2 value for the relationship be-

ween observed and simulated series over the entire data period 2000–

011 was 0.763 or 76.3% ( Table 5 ). This showed that variation in river

ow did not result from changes in rainfall only. Other climatic variables

hich determine rates of evapotranspiration (including temperature, so-

ar radiation, wind speed, and relative humidity) are also important for

xplaining variation in flow. Therefore, the mismatch between observed

ow and output of a distributed hydrological model like SWAT (which

onsiders the various factors or climatic conditions) is expected to be

maller (and thus, high R 

2 value) than that in the case when individual

limatic variable is used in the regression analysis 

By taking into account results regarding the changes in LULC types

ver the period 2000–2014 (9.1% and 6.2% with average of 7.65%),

ontributions from other factors were 92.35% (i.e. 100-7.65%). Contri-

ution from changes in climatic conditions or inter-annual climate vari-

bility was 70.46% ( ≈ 76.3% of 92.35 ≈0.763 ×92.35 ≈70.46%). Thus,

hanges in runoff were dominantly driven by climate variability. The

emaining or unexplained 21.89% (i.e. 100-70.46-7.65) of the total vari-

bility in river flow can be thought of in terms of (i) additional impacts

f human activities apart from changes in LULC types, and (ii) limited

apacity of the hydrological model to capture complexities in rainfall-

unoff generation processes. Two examples of additional human factors

nclude a) possible flow returns into the river through discharge of ef-

uents from industries, and b) abstraction of water at various locations

n the RMC for irrigation, industrial and domestic supplies. Limited ca-

acity of the hydrological model arises because of a) measurement er-

ors on model inputs, b) observation errors on the data against which

alibration is to be made, c) inaccuracies in model parameters, and d)

mperfection of model structure. 

. Conclusions 

The aim of this study was to quantitatively assess contributions of

limate variability and LULC changes to the increasing runoff in the

MC. This was done while also testing the suitability of reanalysis or

atellite products in driving SWAT as a semi-distributed hydrological

odel when applied under data-scarcity. 

In 2000, most of the RMC area comprised grassland (39.5%), crop-

and/farmland (33.0%), and forest (22.0%). Due to changes in land

enure and increased pressure from rapidly growing population on natu-

al resources, the fractions of the catchment area covered by grassland,

ropland/farmland, and forest became 10.4%, 72.2%, and 10.7%, re-

pectively. In 2000, 2008 and 2014, wetland covered 1.85%, 1.54%, and

.76% of the catchment area, respectively. Furthermore, LULC changes

ere possibly driven by the policy-institutional factors such as shifts in

and laws. 

SWAT calibration (2000–2005) and validation (2006–2011) at daily

cale yielded NSE of 0.77 and 0.75, respectively. This indicated that

utputs from SWAT were suitable for investigating attributes of river

ow variation in the RMC. Furthermore, availability of the reanalysis

r satellite meteorological products can offer partial solution to the data

carcity which hampers physically-based hydrological modeling for wa-

ersheds in developing countries like Uganda. 

Analysis showed that from 2000 to 2011, annual runoff increased at

 rate of 0.47 cumecs/year. This increase was mainly due to the pos-
12 
tive trend in rainfall which was at a rate of 52.08 mm/year. Positive

rend in rainfall implies increasing surface runoff. Given the increasing

ate of deforestation across the catchment, increasing rainfall-runoff can

ead to erosion and sedimentation and these can (i) be detrimental to

he turbine in the hydropower plant, and (ii) lower the quality of water

rom River Mpanga to be supplied to the nearby towns thereby increas-

ng water treatment cost. Physical and chemical water qualities of River

panga were already confirmed to be largely influenced by (i) the river

ediment extraction and (ii) transportation of suspended solids and nu-

rients from upstream due to erosion of river bed and bank erosions ( Van

utsel et al., 2017 ). Enforcing regulations that prevent deforestation and

he implementation of tree planting programmes or projects should be

onsidered in the River Mpanga catchment management plan. Creation

f riparian buffer zones along the river can also be good management

ractice to deal with possible soil erosion and sedimentation. Measures

o control rampart LULC changes should be implemented in the frame-

ork of the IWRM. 

Over the period 2000–2008 (2000–2014), contribution from transi-

ion in LULC types to changes in river flow was 9.1% (6.2%). In other

ords, human factors influenced this runoff increase more from 2000

o 2008 than over the period 2008 to 2014. At least 43.9% of the to-

al variance in river flow could be explained by the variation in rain-

all. When various changes in climatic variables (such as temperature

r evapotranspiration) were considered to collectively characterize cli-

ate variability, up to 70.46% of the total variance in river flow was

xplained. Of the total variance in river flow changes, 21.89% remained

nexplained. This could be due to other factors not considered in this

tudy including (i) abstraction of water for irrigation, industrial use, and

omestic consumption or by population in the urban centers within the

atchment, and (ii) possible flow returns into the river from industries

ithin the catchment. It was recently found that abstraction of water

ithin the River Mpanga catchment affect river flow variation to the

xtents which lead to losses of hydropower ( Onyutha et al., 2021 ). 

These findings can support planning of predictive environmental

anagement amidst impacts of climate variability and human activities

n water resources. We recommended other related studies (on hydro-

ogical quantification of LULC changes and human activities on water

esources) to be conducted for the different catchments in Uganda. This

ould be important to support developments of management plans for

he various catchments in the country. 

unding 

This research received no external funding. 

uthor contributions 

Charles Onyutha : Conceptualization, data curation, formal analysis,

nvestigation, methodology, validation, supervision, writing-review &

diting; Catherine Turyahabwe : Conceptualization, data acquisition, data

uration, formal analysis, investigation, methodology, validation, and

riting draft manuscript; Paul Kaweesa : Conceptualization, formal anal-

sis, proofreading draft manuscript, and assistantship in supervision. 

eclaration of Competing Interest 

The authors declare that there are no conflicts of interest. 

cknowledgments 

The authors acknowledge that this study made use of the Japanese

5-year Reanalysis (JRA-55) ( Kobayashi et al., 2015 ), Climate Hazards

roup InfraRed Precipitation with Stations (CHIRPS) version 2.0 (or

HIRPS v2.0) ( Funk et al., 2015 ), and National Centers for Environ-

ental Prediction’s (NCEP’s) Climate Forecast System Reanalysis (CFSR)

 Saha et al., 2014 ). The authors recognize that this paper was partially



C. Onyutha, C. Turyahabwe and P. Kaweesa Environmental Challenges 5 (2021) 100273 

b  

u  

K

R

A  

 

K  

A  

 

A  

J  

 

B  

 

B  

B  

 

B  

 

B  

B  

 

B  

C  

 

 

D  

E  

 

F  

 

F  

F  

F  

 

G  

 

 

G  

 

G  

 

A  

 

K  

 

K  

 

K  

K  

K  

 

K  

 

 

L  

 

M  

 

M  

 

 

M  

 

M  

 

M  

 

 

M  

 

M  

 

M  

 

 

M  

 

M  

 

 

N  

 

 

N  

 

N  

S  

 

N  

 

N  

N  

 

O  

 

O  

 

O  

 

O  

O  

 

O  

O  

 

P  

 

 

ased on a Master of Science study by Turyahabwe ( Turyahabwe, 2019 )

nder the supervision of Charles Onyutha (Ph.D.) assisted by Eng. Paul

aweesa. 

eferences 

bbaspour, K.C., et al., 2007. Modelling hydrology and water quality in the pre-

alpine/alpine Thur watershed using SWAT. J. Hydrol. 333 (2–4), 413–430.

doi: 10.1016/j.jhydrol.2006.09.014 , Feb . 

.C. Abbaspour, “SWAT-calibration and uncertainty programs (CUP) - a User Manual.

Swiss Federal Institute of Aquatic Science and Technology, Eawag, Duebendorf, ”

2015. 

naba, L.A., Banadda, N., Kiggundu, N., Wanyama, J., Engel, B., Moriasi, D., 2017. Appli-

cation of SWAT to assess the effects of land use change in the Murchison bay catch-

ment in Uganda. CWEEE 06 (01), 24–40. doi: 10.4236/cweee.2017.61003 . 

rnold, J.G., Allen, P.M., Bernhardt, G., 1993. A comprehensive surface-groundwater flow

model. J. Hydrol. 142 (1–4), 47–69. doi: 10.1016/0022-1694(93)90004-S . 

.G. Arnold, J.R. Kiniry, R. Srinivasan, J.R. Williams, E.B. Haney, and S.L. Neitsch, “Soil &

Water Assessment Tool: Input/Output Documentation. Version 2012. TR–439, Texas

Water Resources Institute, College Station, ” Texas, 2012. 

ahati, H.K., Ogenrwoth, A., Sempewo, J.I., 2021. Quantifying the potential impacts of

land-use and climate change on hydropower reliability of Muzizi hydropower plant,

Uganda. J. Water Clim. Change doi: 10.2166/wcc.2021.273 , p. jwc2021273 . 

aker, T.J., Miller, S.N., 2013. Using the Soil and Water Assessment Tool (SWAT) to assess

land use impact on water resources in an East African watershed. J. Hydrol. 486, 100–

111. doi: 10.1016/j.jhydrol.2013.01.041 . 

ao, Z., et al., 2012. Attribution for decreasing streamflow of the Haihe River basin,

northern China: climate variability or human activities? J. Hydrol. 460–461, 117–129.

doi: 10.1016/j.jhydrol.2012.06.054 , Aug . 

arasa, B., Egeru, A., Okello, P., Mutuzo, F., 2011. Dynamics of Land Use/Cover Trends

in Kanungu District, South-western Uganda. J. Appl. Sci. Environ. Manag. 14 (4).

doi: 10.4314/jasem.v14i4.63260 . 

even, K.J. , Binley, A.M. , 1992. The future role of distributed models: model calibration

and predictive uncertainty. Hydrol. Processes 6, 279–298 . 

ewket, W., Sterk, G., 2005. Dynamics in land cover and its effect on stream flow in the

Chemoga watershed, Blue Nile basin, Ethiopia. Hydrol. Processes 19 (2), 445–458.

doi: 10.1002/hyp.5542 , Feb . 

RLi, “Study on current and future potential water resources, under different climate sce-

narios, for the Mpanga River basin (Uganda), ” 2015. 

hiwa, R.J., 2008. Effects of Land Use And Land Cover Changes On The Hy-

drology Of Weruweru-Kiladeda Sub-Catchment In Pangani River Basin, Tan-

zania. Kenyatta University, Kenya [Online]. Available http://ir-library.ku.ac.ke/

handle/123456789/6852 . 

aily Monitor, “Mpanga hydro power plant commissioned, ” 2011. http://www.

monitor.co.ug/Business/Technology/-/688612/1106832/-/9huacbz/-/index.html 

(accessed Mar. 06, 2020). 

nku, T., et al., 2014. Biohydrology of low flows in the humid Ethiopian

highlands: the Gilgel Abay catchment. Biologia (Bratisl) 69 (11), 1502–1509.

doi: 10.2478/s11756-014-0462-9 , Nov . 

an, J., Tian, F., Yang, Y., Han, S., Qiu, G., 2010. Quantifying the magnitude of the impact

of climate change and human activity on runoff decline in Mian River Basin, China.

Water Sci. Technol. 62 (4), 783–791. doi: 10.2166/wst.2010.294 . 

AO-UNESCO, 1977. FAO-UNESCO Soil Map of the World, 1:5000000: Africa. United

Nations Educational, Scientific and Cultural Organization . 

AO-UNESCO, 2003. The Digital Soil Map of the World, Version 3.6. Land and Water

Development Division. FAO, Rome . 

unk, C., et al., 2015. The climate hazards infrared precipitation with stations —a

new environmental record for monitoring extremes. Sci Data 2 (1), 150066.

doi: 10.1038/sdata.2015.66 , Dec . 

ebrehiwot, S.G., Seibert, J., Gärdenäs, A.I., Mellander, P.-E., Bishop, K., 2013. Hydrolog-

ical change detection using modeling: Half a century of runoff from four rivers in the

Blue Nile Basin: hydrological Change Detection Using Modeling. Water Resour. Res.

49 (6), 3842–3851. doi: 10.1002/wrcr.20319 , Jun . 

lobal Water Partnership, “A Handbook for Integrated Water Resources Management

in Basins, ” 2009. https://www.inbo-news.org/IMG/pdf/GWP-INBOHandbookFor

IWRMinBasins.pdf (accessed Feb. 20, 2020). 

uzha, A.C., Rufino, M.C., Okoth, S., Jacobs, S., Nóbrega, R.L.B., 2018. Impacts of land

use and land cover change on surface runoff, discharge and low flows: evidence from

East Africa. J. Hydrol. 15, 49–67. doi: 10.1016/j.ejrh.2017.11.005 . 

. Jarvis, H.I. Reuter, A. Nelson, and E. Guevara, “Hole-filled seamless SRTM data V4,

International Centre for Tropical Agriculture (CIAT), ” 2008. http://srtm.csi.cgiar.org

(accessed Jun. 04, 2019). 

ashaigili, J.J., Majaliwa, A.M., 2013. Implications of land use and land cover changes

on hydrological regimes of the Malagarasi River, Tanzania. JASA 02 (01), 45–50.

doi: 10.14511/jasa.2013.020107 . 

izza, C.L., et al., 2017. Land use/cover change patterns in highland ecosystems

of Lake Bunyonyi Catchment in western Uganda. Afr. Crop Sci. J. 25 (1), 43.

doi: 10.4314/acsj.v25i1.4S . 

obayashi, S., et al., 2015. The JRA-55 Reanalysis: general specifications and basic char-

acteristics. J. Meteorol. Soc. Jpn. 93 (1), 5–48. doi: 10.2151/jmsj.2015-001 . 

oster, R.D. , Suarez, M.J. , 1999. A simple framework for examining the interannual vari-

ability of land surface moisture fluxes. J. Clim. 12, 1911–1917 . 

umar, N., Singh, S.K., Singh, V.G., Dzwairo, B., 2018. Investigation of impacts of land

use/land cover change on water availability of Tons River Basin, Madhya Pradesh,

India. Model. Earth Syst. Environ. 4 (1), 295–310. doi: 10.1007/s40808-018-0425-1 . 
13 
umar, N., Singh, S.K., Srivastava, P.K., Narsimlu, B., 2017. SWAT Model calibration and

uncertainty analysis for streamflow prediction of the Tons River Basin, India, using

Sequential Uncertainty Fitting (SUFI-2) algorithm, ”. Model. Earth Syst. Environ. 3 (1),

30. doi: 10.1007/s40808-017-0306-z , Apr . 

iang, X., Lettenmaier, D.P., Wood, E.F., Burges, S.J., 1994. A simple hydrologically based

model of land surface water and energy fluxes for general circulation models. J. Geo-

phys. Res. 99 (D7), 14415. doi: 10.1029/94JD00483 . 

. Ministry of Water and Environment, 2014. Uganda Catchment Management Planning

Guidelines. Republic of Uganda, Ministry of Water and Environment, Directorate of

Water Resources Management . 

ango, L.M., Melesse, A.M., McClain, M.E., Gann, D., Setegn, S.G., 2011. Land use and

climate change impacts on the hydrology of the upper Mara River Basin, Kenya: results

of a modeling study to support better resource management. Hydrol. Earth Syst. Sci.

15 (7), 2245–2258. doi: 10.5194/hess-15-2245-2011 , Jul . 

ati, B.M., Mutie, S., Gadain, H., Home, P., Mtalo, F., 2008. Impacts of land-use/cover

changes on the hydrology of the transboundary Mara River, Kenya/Tanzania: Land

use change, hydrology, and the Mara basin. Lakes Reservoirs: Res. Manag. 13 (2), 169–

177. doi: 10.1111/j.1440-1770.2008.00367.x . 

illy, P.C.D., Dunne, K.A., 2002. Macroscale water fluxes 2. Water and energy supply

control of their interannual variability: CONTROLS OF WATER FLUX VARIABILITY.

Water Resour. Res. 38 (10). doi: 10.1029/2001WR000760 , pp. 24-1-24–9 Oct . 

inistry for Water and Environment, “Ministry for Water and Environment Catch-

ment based Water Resources Management (CbWRM) in Uganda, ” https://www.

mwe.go.ug/library/catchment-based-water-resources-management-cbwrm , 2020.

https://www.mwe.go.ug/library/catchment-based-water-resources-management- 

cbwrm (accessed Nov. 26, 2020). 

inistry of Water and Environment, 2013. Framework and Guidelines For Water

Source Protection-Volume 1: Framework for Water Source Protection. Government

of Uganda, Kampala . 

ubialiwo, A., Abebe, A., Onyutha, C., 2021. Performance of rainfall–runoff models in

reproducing hydrological extremes: a case of the River Malaba sub-catchment. SN

Appl. Sci. 3 (4), 515. doi: 10.1007/s42452-021-04514-7 , Apr . 

ustard, J. , DeFries, R. , Fisher, T. , Moran, E.F. , 2004. Land use and land cover change

pathways and impacts. In: Cochrane, M.A. (Ed.), Land Change Science: Observ-

ing, Monitoring, and Understanding Trajectories of Change on the Earth’s Surface.

Springer-Verlag, Dordrecht, the Netherlands, pp. 431–447 . 

utenyo, I., Nejadhashemi, P.A., Woznicki, S.A., Giri, S., 2011. Evaluation of SWAT per-

formance on a mountainous watershed in tropical Africa. Hydrol. Curr. Res. S14 (001),

1–7. doi: 10.4172/2157-7587.S14-001 . 

wangi, H.M., Julich, S., Patil, S.D., McDonald, M.A., Feger, K.-H., 2016. Modelling the

impact of agroforestry on hydrology of Mara River Basin in East Africa: impact of

Agroforestry on Hydrology of Mara River Basin. Hydrol. Process. 30 (18), 3139–3155.

doi: 10.1002/hyp.10852 . 

arsimlu, B., Gosain, A.K., Chahar, B.R., Singh, S.K., Srivastava, P.K., 2015. SWAT Model

Calibration and Uncertainty Analysis for Streamflow Prediction in the Kunwari River

Basin, India, Using Sequential Uncertainty Fitting. Environ. Process. 2 (1), 79–95.

doi: 10.1007/s40710-015-0064-8 , Mar . 

ash, J.E., Sutcliffe, J.V., 1970. River flow forecasting through conceptual

models part I — a discussion of principles. J. Hydrol. 10 (3), 282–290.

doi: 10.1016/0022-1694(70)90255-6 , Apr . 

ational Environment Management Authority, 2009. Environmental Sensitivity Atlas for

the Albertine Graben, 1st ed. NEMA, Kampala, Uganda . 

.L. Neitsch, J.G. Arnold, J.R. Kiniry, and J.R. Williams, “Soil & water assessment tool

theoretical documentation version 2009. Technical Report, Texas Water Resources

Institute, College Station, ” Texas, 2011. 

eitsch, S.L. , Arnold, J.G. , Kiniry, J.R. , Williams, J.R. , King, K.W. , 2002. Soil and Water

Assessment Tool Theoretical Documentation. USDA-ARS Publication GSWRL 02-01

BRC 02-05 TR-01, ”. 

icol, A. , Odinga, W. , 2016. IWRM in Uganda – progress after decades of implementation.

Water Alternatives 9 (3), 627–643 . 

obert, J., Jeremiah, J., 2012. Hydrological Response of Watershed Systems to

Land Use/Cover Change. A Case of Wami River Basin. TOHYDJ 6 (1), 78–87.

doi: 10.2174/1874378101206010078 . 

dira, P.M.A. , Nyadawa, M.O. , Ndwallah, B.O. , Juma, N.A. , Obiero, J.P. , 2010. Impact

of land-use /cover dynamics on streamflow: a case of Nzoia River catchment, Kenya.

Nile Basin Water Sci. Eng. J. 3 (2), 64–78 . 

lang, L.O., Fürst, J., 2011. Effects of land cover change on flood peak discharges and

runoff volumes: model estimates for the Nyando River Basin, Kenya. Hydrol. Processes

25 (1), 80–89. doi: 10.1002/hyp.7821 , Jan . 

nyutha, C., 2016. Influence of hydrological model selection on simulation of moderate

and extreme flow events: a case study of the Blue Nile Basin. Adv. Meteorol. 2016

(Article ID 7148326), 1–28. doi: 10.1155/2016/7148326 . 

nyutha, C. , 2020. From R-squared to coefficient of model accuracy for assessing ‘good-

ness-of-fits. Geoscientific Model Development Discussions, pp. 1–25 . 

nyutha, C., Kerudong, P.A., Guma, B.E., Mugisha, C., 2021. Impacts of upstream wa-

ter abstraction and climate variability on River Mpanga hydropower production in

Uganda. Int. J. Energ. Water Res. doi: 10.1007/s42108-021-00137-1 . 

nyutha, C., Willems, P., 2017. Space-time variability of extreme rainfall in the River Nile

basin. Int. J. Climatol 37 (14), 4915–4924. doi: 10.1002/joc.5132 , Nov . 

nyutha, C., Willems, P., 2018. Investigation of flow-rainfall co-variation for catchments

selected based on the two main sources of River Nile. Stoch. Environ. Res. Risk Assess.

32 (3), 623–641. doi: 10.1007/s00477-017-1397-9 . 

irnia, A., Darabi, H., Choubin, B., Omidvar, E., Onyutha, C., Haghighi, A.T.,

2019. Contribution of climatic variability and human activities to stream flow

changes in the Haraz River basin, northern Iran. J. Hydroenviron. Res. 25, 12–24.

doi: 10.1016/j.jher.2019.05.001 , Jun . 

https://doi.org/10.1016/j.jhydrol.2006.09.014
https://doi.org/10.4236/cweee.2017.61003
https://doi.org/10.1016/0022-1694(93)90004-S
https://doi.org/10.2166/wcc.2021.273
https://doi.org/10.1016/j.jhydrol.2013.01.041
https://doi.org/10.1016/j.jhydrol.2012.06.054
https://doi.org/10.4314/jasem.v14i4.63260
http://refhub.elsevier.com/S2667-0100(21)00252-3/sbref0010
http://refhub.elsevier.com/S2667-0100(21)00252-3/sbref0010
http://refhub.elsevier.com/S2667-0100(21)00252-3/sbref0010
https://doi.org/10.1002/hyp.5542
http://ir-library.ku.ac.ke/handle/123456789/6852
http://www.monitor.co.ug/Business/Technology/-/688612/1106832/-/9huacbz/-/index.html
https://doi.org/10.2478/s11756-014-0462-9
https://doi.org/10.2166/wst.2010.294
http://refhub.elsevier.com/S2667-0100(21)00252-3/sbref0017
http://refhub.elsevier.com/S2667-0100(21)00252-3/sbref0018
https://doi.org/10.1038/sdata.2015.66
https://doi.org/10.1002/wrcr.20319
https://www.inbo-news.org/IMG/pdf/GWP-INBOHandbookForIWRMinBasins.pdf
https://doi.org/10.1016/j.ejrh.2017.11.005
http://srtm.csi.cgiar.org
https://doi.org/10.14511/jasa.2013.020107
https://doi.org/10.4314/acsj.v25i1.4S
https://doi.org/10.2151/jmsj.2015-001
http://refhub.elsevier.com/S2667-0100(21)00252-3/sbref0027
http://refhub.elsevier.com/S2667-0100(21)00252-3/sbref0027
http://refhub.elsevier.com/S2667-0100(21)00252-3/sbref0027
https://doi.org/10.1007/s40808-018-0425-1
https://doi.org/10.1007/s40808-017-0306-z
https://doi.org/10.1029/94JD00483
http://refhub.elsevier.com/S2667-0100(21)00252-3/sbref0031
https://doi.org/10.5194/hess-15-2245-2011
https://doi.org/10.1111/j.1440-1770.2008.00367.x
https://doi.org/10.1029/2001WR000760
https://www.mwe.go.ug/library/catchment-based-water-resources-management-cbwrm
http://refhub.elsevier.com/S2667-0100(21)00252-3/sbref0036
https://doi.org/10.1007/s42452-021-04514-7
http://refhub.elsevier.com/S2667-0100(21)00252-3/sbref0038
http://refhub.elsevier.com/S2667-0100(21)00252-3/sbref0038
http://refhub.elsevier.com/S2667-0100(21)00252-3/sbref0038
http://refhub.elsevier.com/S2667-0100(21)00252-3/sbref0038
http://refhub.elsevier.com/S2667-0100(21)00252-3/sbref0038
https://doi.org/10.4172/2157-7587.S14-001
https://doi.org/10.1002/hyp.10852
https://doi.org/10.1007/s40710-015-0064-8
https://doi.org/10.1016/0022-1694(70)90255-6
http://refhub.elsevier.com/S2667-0100(21)00252-3/sbref0043
http://refhub.elsevier.com/S2667-0100(21)00252-3/sbref0045
http://refhub.elsevier.com/S2667-0100(21)00252-3/sbref0045
http://refhub.elsevier.com/S2667-0100(21)00252-3/sbref0045
http://refhub.elsevier.com/S2667-0100(21)00252-3/sbref0045
http://refhub.elsevier.com/S2667-0100(21)00252-3/sbref0045
http://refhub.elsevier.com/S2667-0100(21)00252-3/sbref0045
http://refhub.elsevier.com/S2667-0100(21)00252-3/sbref0046
http://refhub.elsevier.com/S2667-0100(21)00252-3/sbref0046
http://refhub.elsevier.com/S2667-0100(21)00252-3/sbref0046
https://doi.org/10.2174/1874378101206010078
http://refhub.elsevier.com/S2667-0100(21)00252-3/sbref0048
http://refhub.elsevier.com/S2667-0100(21)00252-3/sbref0048
http://refhub.elsevier.com/S2667-0100(21)00252-3/sbref0048
http://refhub.elsevier.com/S2667-0100(21)00252-3/sbref0048
http://refhub.elsevier.com/S2667-0100(21)00252-3/sbref0048
http://refhub.elsevier.com/S2667-0100(21)00252-3/sbref0048
https://doi.org/10.1002/hyp.7821
https://doi.org/10.1155/2016/7148326
http://refhub.elsevier.com/S2667-0100(21)00252-3/sbref0051
http://refhub.elsevier.com/S2667-0100(21)00252-3/sbref0051
https://doi.org/10.1007/s42108-021-00137-1
https://doi.org/10.1002/joc.5132
https://doi.org/10.1007/s00477-017-1397-9
https://doi.org/10.1016/j.jher.2019.05.001


C. Onyutha, C. Turyahabwe and P. Kaweesa Environmental Challenges 5 (2021) 100273 

R  

 

R  

 

R  

 

R  

R  

 

R  

 

R  

 

 

S

S  

S  

 

T  

 

T  

U

V  

W  

 

W  

 

 

W  

 

X  

 

Z  

 

Z  

 

Z  

 

awat, K.S., Singh, S.K., 2017. Estimation of surface runoff from semi-arid ungauged agri-

cultural watershed using SCS-CN method and earth observation data sets. Water Con-

serv. Sci. Eng. 1 (4), 233–247. doi: 10.1007/s41101-017-0016-4 . 

epublic of Uganda, 1995. The Constitution of the Republic of Uganda. Ministry of Jus-

tice and Constitutional Affairs, Kampala, Uganda. Uganda Printing and Publishing

Corporation. Government of Uganda, Entebbe . 

epublic of Uganda, 1998. Guidelines on the Management of Land and other Related Issues

under the Land Act, 1998. Ministry of Water, Lands and Environment. Government

of Uganda, Kampala . 

epublic of Uganda, 2006. The Refugee Act 2006, 24 May 2006. National Legislative

Bodies /National Authorities. Government of Uganda, Kampala . 

epublic of Uganda, 2010. Uganda: The Refugees Regulations, 2010, 27 October 2010,

S.I. 2010 No. 9, National Legislative Bodies /National Authorities. Government of

Uganda, Kampala . 

epublic of Uganda, 2013. National Land Policy, 2013.National Legislative Bodies /Na-

tional Authorities, Ministry of Water, Lands and Environment. Government of Uganda,

Kampala . 

ientjes, T.H.M., Haile, A.T., Kebede, E., Mannaerts, C.M.M., Habib, E., Steenhuis, T.S.,

2011. Changes in land cover, rainfall and stream flow in Upper Gilgel Abbay

catchment, Blue Nile basin – Ethiopia. Hydrol. Earth Syst. Sci. 15 (6), 1979–1989.

doi: 10.5194/hess-15-1979-2011 , Jun . 

aha, S., et al., 2014. The NCEP Climate Forecast System Version 2. J. Clim. 27 (6), 2185–

2208. doi: 10.1175/JCLI-D-12-00823.1 . 

heffield, J., Goteti, G., Wood, E.F., 2006. Development of a 50-year high-resolution global

dataset of meteorological forcings for land surface modeling. J. Clim. 19 (13), 3088–

3111. doi: 10.1175/JCLI3790.1 , Jul . 

un, G. , McNulty, S.G. , Lu, J. , Amatya, D.M. , Liang, Y. , Kolka, R.K. , 2005. Regional annual

water yield from forest lands and its response to potential deforestation across the

southeastern United States. J. Hydrol. 308, 258–268 . 

an, M.L., Ibrahim, A.L., Yusop, Z., Duan, Z., Ling, L., 2015. Impacts of land-use and

climate variability on hydrological components in the Johor River basin, Malaysia.

Hydrol. Sci. J. 1–17. doi: 10.1080/02626667.2014.967246 . 
14 
uryahabwe, C. , 2019. Assessment of impacts of land use changes on water resources of

River Mpanga catchment. Kyambogo University, Kyambogo, Kampala . 

ganda Bureau of Statistics, 2016. The national population and housing census 2014 –

main report. Government of Uganda, Kampala . 

an Butsel, J. , et al. , 2017. Report: Ecological water quality assessment of the Mpanga

catchment, Western Uganda.. Ghent University Ghent University . 

agesho, N., 2014. Catchment dynamics and its impact on runoff generation: Coupling

watershed modelling and statistical analysis to detect catchment responses. Int. J.

Water Res. Environ. Eng. 6 (2), 73–87. doi: 10.5897/IJWREE2013.0449 . 

ang, G., Xia, J., Chen, J., 2009. Quantification of effects of climate variations and human

activities on runoff by a monthly water balance model: A case study of the Chaobai

River basin in northern China: CLIMATE VARIATIONS AND HUMAN ACTIVITIES.

Water Resour. Res. 45 (7). doi: 10.1029/2007WR006768 , Jul . 

ang, H., Chen, L., Yu, X., 2016. Distinguishing human and climate influences

on streamflow changes in Luan River basin in China. Catena 136, 182–188.

doi: 10.1016/j.catena.2015.02.013 , Jan . 

u, X., Yang, D., Yang, H., Lei, H., 2014. Attribution analysis based on the Budyko hy-

pothesis for detecting the dominant cause of runoff decline in Haihe basin. J. Hydrol.

510, 530–540. doi: 10.1016/j.jhydrol.2013.12.052 , Mar . 

hang, L., Dawes, W.R., Walker, G.R., 2001. Response of mean annual evapotranspira-

tion to vegetation changes at catchment scale. Water Resour. Res. 37 (3), 701–708.

doi: 10.1029/2000WR900325 , Mar . 

hang, Y., Guan, D., Jin, C., Wang, A., Wu, J., Yuan, F., 2011. Analysis of impacts of

climate variability and human activity on streamflow for a river basin in northeast

China. J. Hydrol. 410 (3–4), 239–247. doi: 10.1016/j.jhydrol.2011.09.023 , Nov . 

hao, A., Zhu, X., Liu, X., Pan, Y., Zuo, D., 2016. Impacts of land use change and climate

variability on green and blue water resources in the Weihe River Basin of northwest

China. Catena 137, 318–327. doi: 10.1016/j.catena.2015.09.018 , Feb . 

https://doi.org/10.1007/s41101-017-0016-4
http://refhub.elsevier.com/S2667-0100(21)00252-3/sbref0057
http://refhub.elsevier.com/S2667-0100(21)00252-3/sbref0058
http://refhub.elsevier.com/S2667-0100(21)00252-3/sbref0059
http://refhub.elsevier.com/S2667-0100(21)00252-3/sbref0060
http://refhub.elsevier.com/S2667-0100(21)00252-3/sbref0061
https://doi.org/10.5194/hess-15-1979-2011
https://doi.org/10.1175/JCLI-D-12-00823.1
https://doi.org/10.1175/JCLI3790.1
http://refhub.elsevier.com/S2667-0100(21)00252-3/sbref0065
http://refhub.elsevier.com/S2667-0100(21)00252-3/sbref0065
http://refhub.elsevier.com/S2667-0100(21)00252-3/sbref0065
http://refhub.elsevier.com/S2667-0100(21)00252-3/sbref0065
http://refhub.elsevier.com/S2667-0100(21)00252-3/sbref0065
http://refhub.elsevier.com/S2667-0100(21)00252-3/sbref0065
http://refhub.elsevier.com/S2667-0100(21)00252-3/sbref0065
https://doi.org/10.1080/02626667.2014.967246
http://refhub.elsevier.com/S2667-0100(21)00252-3/sbref0067
http://refhub.elsevier.com/S2667-0100(21)00252-3/sbref0067
http://refhub.elsevier.com/S2667-0100(21)00252-3/sbref0068
http://refhub.elsevier.com/S2667-0100(21)00252-3/sbref0069
http://refhub.elsevier.com/S2667-0100(21)00252-3/sbref0069
http://refhub.elsevier.com/S2667-0100(21)00252-3/sbref0069
https://doi.org/10.5897/IJWREE2013.0449
https://doi.org/10.1029/2007WR006768
https://doi.org/10.1016/j.catena.2015.02.013
https://doi.org/10.1016/j.jhydrol.2013.12.052
https://doi.org/10.1029/2000WR900325
https://doi.org/10.1016/j.jhydrol.2011.09.023
https://doi.org/10.1016/j.catena.2015.09.018

	Impacts of climate variability and changing land use/land cover on River Mpanga flows in Uganda, East Africa
	1 Introduction
	2 Materials and methods
	2.1 Study area
	2.2 Data
	2.2.1 Hydro-meteorological data
	2.2.1 Topographical data and soil map
	2.2.2 LULC map

	2.3 Hydrological modeling using SWAT
	2.3.1 Selection of the rainfall series
	2.3.2 Model build-up
	2.3.3 Calibration and validation
	2.3.4 Simulations
	2.3.5 Contribution from changes in LULC types to river flow variation
	2.3.6 Contribution from changes in climatic conditions to river flow variation


	3 Results and discussion
	3.1 Rainfall and river flow properties
	3.2 Soil and LULC maps
	3.3 Hydrological modeling
	3.3.1 Selection of the rainfall series to use for modeling
	3.3.2 Sensitivity analysis
	3.3.3 Calibration and validation
	3.3.4 Contribution from changes in LULC types to river flow variation
	3.3.5 Contribution from changes in climatic conditions to river flow variation


	4 Conclusions
	Funding
	Author contributions
	Declaration of Competing Interest
	Acknowledgments
	References