Impact of Climate Change On Hydropower Generation in Rio Jubones Basin, Ecuador

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 10

Water Science and Engineering 2018, 11(2): 157e166

H O S T E D BY Available online at www.sciencedirect.com

Water Science and Engineering


journal homepage: http://www.waterjournal.cn

Impact of climate change on hydropower generation in


Rio Jubones Basin, Ecuador
Mohammad Mehedi Hasan a,*, Guido Wyseure b
a
Hydraulic Research Directorate, River Research Institute, Faridpur 7800, Bangladesh
b
Department of Earth and Environmental Sciences, KU Leuven, Leuven 3001, Belgium
Received 30 May 2017; accepted 19 February 2018
Available online 5 July 2018

Abstract

This study attempted to use the soil and water assessment tool (SWAT), integrated with geographic information systems (GIS), for assessment
of climate change impacts on hydropower generation. This methodology of climate change impact modeling was developed and demonstrated
through application to a hydropower plant in the Rio Jubones Basin in Ecuador. ArcSWAT 2012 was used to develop a model for simulating the
river flow. The model parameters were calibrated and validated on a monthly scale with respect to the hydro-meteorological inputs observed
from 1985 to 1991 and from 1992 to 1998, respectively. Statistical analyses produced Nash-Sutcliffe efficiencies (NSEs) of 0.66 and 0.61 for
model calibration and validation, respectively, which were considered acceptable. Numerical simulation with the model indicated that climate
change could alter the seasonal flow regime of the basin, and the hydropower potential could change due to the changing climate in the future.
Scenario analysis indicates that, though the hydropower generation will increase in the wet season, the plant will face a significant power
shortage during the dry season, up to 13.14% from the reference scenario, as a consequence of a 17% reduction of streamflow under an
assumption of a 2.9 C increase in temperature and a 15% decrease in rainfall. Overall, this study showed that hydrological processes are
realistically modeled with SWAT and the model can be a useful tool for predicting the impact of climate change.
© 2018 Hohai University. Production and hosting by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://
creativecommons.org/licenses/by-nc-nd/4.0/).

Keywords: SWAT; Hydropower generation; Climate change; Sensitivity analysis; Nash-Sutcliffe efficiency (NSE)

1. Introduction hydropower is the most lucrative due to renewability, lower


emissions, and longevity of infrastructure. Though the hydro-
The United Nations Framework Convention on Climate power sector makes a significant effort to cope with today's
Change (UNFCCC) defines climate change as an alteration of rising world energy demands, this is difficult due to climate
climate, accountable definitely or indefinitely to anthropogenic change effects. Madani (2011) stated that climate change will
activities, that changes the structure of the global atmosphere have a variety of effects on streamflow, involving quantity and
and is discerned over a time period, distinguishing it from the timing, temperature, sediment load, and ecosystem changes.
natural climate instability (Houghton et al., 2001). To mitigate Temperature fluctuations, rainfall patterns, floods, and droughts
climate change impacts, several solutions have been proposed are all major signs of climate change that have strong effects on
to reduce greenhouse gas emissions, including modern efficient river systems, which will consequently affect hydropower
energy alternatives and enhancing the use of sustainable energy generation. The possible recession of river flow may lead to
sources. Among alternative options for power generation, decreased hydropower generation, which in turn will have
certain influences on the economic viability of the hydropower
plant schemes. Hydropower plants with low storage capacities
* Corresponding author. are more susceptible to climate change, as a high storage
E-mail address: mmhasan@rri.gov.bd (Mohammad Mehedi Hasan).
Peer review under responsibility of Hohai University.
capacity requires more workability to operate.

https://doi.org/10.1016/j.wse.2018.07.002
1674-2370/© 2018 Hohai University. Production and hosting by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://
creativecommons.org/licenses/by-nc-nd/4.0/).
158 Mohammad Mehedi Hasan, Guido Wyseure / Water Science and Engineering 2018, 11(2): 157e166

Watershed models and general circulation models (GCMs) climate change on hydropower generation, analysis of the river
are often used to simulate the impact of climate change on flow related to climate variability is required, as it will help
watershed hydrology (Park et al., 2011). They are usually used planners and managers design and operate the hydropower
to obtain the climatic data with the help of some downscaling plant efficiently.
techniques in a particular study area. Because of the quite In this study, the soil and water assessment tool (SWAT)
coarse resolution, they are incapable of predicting regional model, which is basically a river basin model, was used to
climate scenarios. Furthermore, they are not designed to assess simulate river flow. SWAT is a process-based and spatially semi-
climate change impacts and are not used to provide a direct distributed hydrological model developed by the United States
assessment of hydrological response to climate change (Dibike Department of Agriculture (USDA) (Arnold et al., 1998). The
and Coulibaly, 2005), because of a high amount of uncertainty. tool anticipates the effects of management practices on water,
As for a regional climate model (RCM), the potential of sediment, and agrochemicals in large, complex basins with
simulating wide-scale data with significant, physically reliable changeable soils, land use types, and management conditions,
climate variations and its advanced representation of extreme over extended periods of time (Ficklin et al., 2009; Githui et al.,
climate events have been the major advantages (Huntingford 2009; Fontaine et al., 2001; Li et al., 2011; Neitsch et al., 2013).
et al., 2003). However, computations of RCMs are detailed, To simulate the physical process, a river basin may be split into a
and a small number of scenario ensembles are available, which number of sub-basins. For each sub-basin, input data are orga-
limits the model quantification to 30 years under the existing nized into the following categories: climate, hydrological
climatic conditions from 1961 to 1990 as well as from 2071 to response units (HRUs), ponds/wetlands, groundwater, and the
2100 (Fowler et al., 2007). This makes it difficult to evaluate the main channel or reach discharging to the sub-basin.
climate change impacts for other phases. Moreover, Buytaert The water balance is the key element in SWAT no matter
et al. (2010) have shown that, for some tropical regions, what case is studied in a catchment. Hydrological simulation
RCMs might yield unreliable or poor results in comparison with for a catchment may be divided into two main phases: a land
GCMs on certain scales, particularly for precipitation. phase and a water or routing phase of the hydrologic cycle. In
Andean landscapes provide a good opportunity for hydro- the land phase, the amounts of water, sediment, nutrients, and
power generation (Buytaert et al., 2006). However, energy pesticides discharging into the main channel in each sub-basin
generation is vulnerable to climate change effects (Koch et al., are simulated. The routing phase can be outlined as the
2011; Tamm et al., 2016). The Andes region, including the movement of water, sediment, nutrients, etc. to the outlet of
South Ecuadorian Andes, shows a high degree of weather and the catchment through the channel network.
climate variability (Espinoza Villar et al., 2009). Considering The hydrologic cycle in SWAT is based on the water bal-
the consequences of climate anomalies, a number of studies ance equation:
have been conducted to investigate various aspects of climate
X
t
 
change impacts on water resources of the Andes region. SWt ¼ SW0 þ Ri  Qsi  Ei  Wi  Qgi ð1Þ
Urrutia and Vuille (2009) quantified future climate change i¼1
in the tropical Andes of Ecuador by means of an RCM.
The uncertainties in projections of climate change impacts on where SWt is the final soil water content on day t (mm), SW0 is
regional water resources in the tropical Andes were evaluated the initial soil water content (mm), Ri is the amount of pre-
by Buytaert et al. (2010). Vuille et al. (2003) conducted cipitation on day i (mm), Qsi is the amount of surface runoff
another study to compare observed and modeled results on day i (mm), Ei is the amount of evapotranspiration on day i
associated with the climate change effects in the 20th century (mm), Wi is the amount of water entering the vadose zone
in the tropical Andes. More recently, Mora et al. (2014) from the soil profile on day i (mm), and Qgi is the amount of
studied the climate change trends for hydro-meteorological return flow on day i (mm).
extremes across the Paute River Basin, in the Ecuadorian The sub-division of the catchment allows the model to
Andes. However, much less attention has been paid to climate reflect the disparity of evapotranspiration (ET) from different
change impacts with regard to water availability or hydro- crops and soils. Runoff of each HRU is anticipated exclusively
power potential, which is important because hydropower has and routed to obtain the total runoff for the catchment.
become the principal source of energy in most of the Andean This enhances the precision and provides a better physical
countries (Bradley et al., 2006). description of the water balance. The model takes the values of
In Ecuador, greater hydropower generation is needed to daily atmospheric conditions as inputs from observation and
compensate for ongoing population growth. To meet the simulation. From average monthly weather data, daily values
country’s rising demand for electric energy, a new hydropower are generated. The model produces a number of weather data
plant is being constructed in the Rio Jubones Basin, in for each sub-basin.
Ecuador. This study aimed to investigate the climate change SWAT also runs with minimal data inputs, and is compu-
impact on hydropower generation in the Rio Jubones Basin. To tationally proficient and hence capable of running simulations
this end, the Minas-San Francisco Hydroelectric Project, being of large basins or management practices without wasting
constructed on the right bank of the Jubones River between computational resources. The SWAT model has been linked
the provinces of Azuay and El Oro, with an installed capacity with many systems, one of the flexible examples being its
of 275 MW, was studied. To assess the possible impact of linkage with ArcView GIS software (ArcView GIS 10.1 with
Mohammad Mehedi Hasan, Guido Wyseure / Water Science and Engineering 2018, 11(2): 157e166 159

the ArcSWAT, 2012 extension), which was used in this study. about 925 mm to about 290 mm. Dense cloud banks are almost
An inclusive interpretation of SWAT can be obtained from a daily occurrence at different elevations across the catchment.
Neitsch et al. (2013). Depending on the locality and elevation, the mean annual
In this study, possible effects of climate change on hydro- temperature varies slightly from around 15 Ce28 C. Annual
power generation in the Rio Jubones Basin, in Ecuador, average relative humidity fluctuates with elevation, from 70%
were investigated by analyzing outputs of the SWAT model. to 88%, depending on the temperature. Monthly average po-
Information regarding climate change at different hydro- tential evapotranspiration (PET) was found to be roughly
meteorological stations was analyzed. The SWAT database 95 mm, and the annual PET fluctuates from about 900 mm to
was configured with a customized local dataset as it was less than 500 mm. In this study, the upper part of the Rio
initially designed for regions of the U.S. Afterwards, a SWAT Jubones Basin, shown in Fig. 1, was taken as the study area. It
model was developed and numerical simulation was performed should be mentioned that an outlet of the basin is in a close
based on the available hydro-meteorological inputs. The SWAT proximity to the dam under construction, through which
model was calibrated and validated using the sequential un- streamflow from the entire basin passes. As the outlet was
certainty fitting (SUFI-2) algorithm, with the uncertainty and selected as the point of interest for numerical simulation
sensitivity analysis conducted. Finally, a summary was made of during watershed delineation, the lower part of the basin
the weather conditions and hydropower potentials of the study would have no meaning in modeling, and hence, it was
area. excluded in this study.
Agriculture is the prevailing activity in the Rio Jubones
2. Study area Basin. The diversified patterns of rainfall and temperature
bring about a variety of tropical crops and fruits, such as corn,
The Rio Jubones Basin, in Ecuador, ranges from the lati- barley, beans, sugarcane, wheat, potato, banana, and cacao.
tudes of 3 40 S to 3 440 S and the longitudes of 78 570 W to Discrete land use patterns (Fig. 2) exist in this basin, covered
80 10 W, with an area of 4362 km2 (Fig. 1). It lies on the with shrubs, meadow, pasture, crops, grass, paramo vegetation,
western slope in the Andes region, with an approximate forests and woodlands, barren land, and built-up areas. Par-
channel length of 180 km. Rivers in the basin discharge into amo, an alpine ecosystem found in Ecuador and other places,
the Pacific Ocean. Most of the sub-basins in this drainage is the major water source in this region and covers around 24%
basin ultimately stream towards the Jubones River, which has of the total area. Heterogeneous soil classes are observed here;
a mean annual discharge of 48.3 m3/s at the planned location among them, inceptisol and entisol are the most dominant soil
of the dam (Enerjubones, 2014). Due to the presence of the types (Fig. 3).
Andean mountain range, trade winds and ocean currents move
inward from the Pacific Ocean, and the climatic conditions 3. Materials and methods
are extremely variable. Weather situations also change rapidly.
They are generally rough with chilly wind, snow, hail, rain, 3.1. Materials and modeling software
and fog. The climate of the Rio Jubones Basin ranges from
humid to arid. From January to May, rain falls more often, and In this study, spatial, meteorological, and daily river flow
intensely, throughout the region. However, there is a marked data were prerequisites for developing a SWAT model for
decrease in rainfall during the period from June to December, the Rio Jubones Basin. The daily time series of hydro-
resulting in a desert-type climate. Annual rainfall varies from meteorological data for the period of 17 years from 1982 to

Fig. 1. Landscape of Rio Jubones Basin.


160 Mohammad Mehedi Hasan, Guido Wyseure / Water Science and Engineering 2018, 11(2): 157e166

Fig. 2. Land use classes of study area.

Both meteorological and streamflow data were taken from the


National Institute for Meteorology and Hydrology, in Ecuador,
while spatial data came from Promas at the University of
Cuenca, in Ecuador. For the spatial data, the
WGS_1984_UTM_Zone_17S coordinate projection was used.
Information regarding the hydropower plant was obtained
from the Enerjubones, the state owned power generation
holding company of Ecuador. The ArcView GIS 10.1 interface
for ArcSWAT 2012 was chosen as a tool in this study.
Fig. 4 presents the rainfall pattern of the selected stations in
the study area for the observed baseline period (1982e1998),
showing the same trend throughout the period. The rainfall
distributions at stations M419, M142, and M032 show higher
Fig. 3. Major soil classes of study area. values of rainfall than other stations. These stations are posi-
tioned at higher elevations, and high mountains radically
change the rainfall patterns naturally, due to the forcing of
1998 were taken into consideration for numerical simulation upright atmospheric motions.
based on their availability. River flow data at station H529 It is observed that the maximum and minimum daily tem-
were used for model calibration and validation. SWAT ne- peratures at the selected stations in the study area are higher at
cessitates meteorological data, consisting of daily rainfall, air M032 and M196, as compared to M142 (Table 2), since the
temperature (minimum and maximum), relative humidity, and two stations are located at relatively lower elevations and fall
wind speed, as weather inputs. In this study, wind speed, in the dry climatic zone. As a rule of thumb, temperature drops
relative humidity, and solar radiation data were derived from about 1 C for every 180 m of ascent.
the monthly statistics of weather information, since these Fig. 5 presents the mean annual streamflow for the period
variables were sporadic or not available. Seven rain gauge from 1982 to 1998 at station H529. Streamflow shows an
stations and three temperature stations (Table 1) were utilized upward trend from 1982 to 1983, followed by a downward
to obtain the weather information. Spatial data involved a trend, and then remains almost constant during the rest of
digital elevation model (DEM), soil map, and land cover map. the measured period. The observed annual streamflow at

Table 1
Observed hydro-meteorological stations in study area.
Station Site name Climatic zone Elevation (m) Latitude ( S) Longitude ( W)
H529 Jubones DJ. S. Francisco 712 3.307 79.500
M032* Santa Isabel Inamhi Dry 1450 3.247 79.277
M142* Saraguro Semi-humid 2525 3.612 79.234
M196* Minas de Huascachaca Dry 1040 3.332 79.329
M419 Giron Semi-humid 2130 3.154 79.149
M420 Nabon Inamhi Semi-humid 2750 3.334 79.066
M421 O~na Dry 2320 3.464 79.154
M423 Susudel-Caserio (L. P. CA) Dry 2600 3.399 79.185
M425 Tendales-Jubones DJ. S. Francisco Semi-humid 750 3.314 79.480
Note: H means the hydrological station, M means the meteorological station, * means the temperature station, and elevations are values in meters above sea level.
Mohammad Mehedi Hasan, Guido Wyseure / Water Science and Engineering 2018, 11(2): 157e166 161

Fig. 4. Mean monthly distribution of measured rainfall at different Fig. 5. Mean annual streamflow trend at station H529 for period from
rain gauge stations for period from 1982 to 1998. 1982 to 1998.

this station showed an average of 32.36 m3/s, with the mini- on temperature and rainfall from the investigation by
mum and maximum streamflow being 19 m3/s and 74 m3/s, Mora et al. (2014) on the impact of climate change on the
respectively. Paute River Basin in the southern Ecuadorian Andes, since the
study area is close to the Paute River Basin. The temperature
3.2. Design specifications of hydropower plant change was predicted to vary between 1.1 C and 2.9 C for
the future period from 2045 to 2065, and the rainfall was
The hydropower generation is a function of the flow predicted to vary from 5% to 15% for the same period as well.
discharge, head, and density of water. The gross hydropower The assumed climate change scenarios are listed in Table 4.
generation in watts can be calculated as follows: The basic assumption applied to rainfall was that the wet
periods become wetter and the dry periods become drier.
E ¼ rgQDhh ð2Þ
The wet season of this study area ranges from January to
where r is the density of water, with r ¼ 103 kg/m3; g is the May, and the dry season ranges from June to December. The
gravitational acceleration, with g ¼ 9.81 m/s2; Q is the flow three scenarios were assumed for a future period without
discharge (m3/s); Dh is the difference in elevations between specifying an exact period. Subsequently, the scenarios were
the water level at the intake of the dam and the outlet of the used as SWAT inputs for assessing the climate change impact
turbine (m); and h is the efficiency of turbines. The design on the hydrological extremes of the catchment.
specifications for the hydropower plant are listed in Table 3
(Enerjubones, 2014). 3.4. Model configuration

3.3. Generation of climate change scenarios SWAT database files were adjusted for the case of the Rio
Jubones Basin, as application of SWAT to other areas demands
According to Houghton et al. (2001), it is recommended to customization with regard to the local conditions. To this end,
use the GCM output data for at least a 30-year averaging a new database of soil type, crop, land use, and weather was
period. However, this study had a record of only 17 years of created. The development of the model initially required
climatic data, which limited the use of GCMs in the study automatic watershed delineation, HRUs, and weather inputs.
area. Therefore, to determine the actual impact of climate The study area was divided into sub-basins linked with
change on the water resources of the study area, it is crucial the stream network, and smaller units were named as HRUs,
to use a downscaling technique. To study the climate change signifying an arrangement of land use, soil, and slope. This
impact on the hydropower generation in the Rio Jubones enabled the SWAT model to reflect differences in hydrological
Basin, three climate change scenarios were determined based
Table 3
Basic design specifications of hydropower plant.
Table 2
Statistics of daily temperature at selected stations in study area for period from Design feature Specification
1982 to 1998. Installed capacity 275 MW
Statistics Maximum daily Minimum daily Type of turbine Pelton
temperature ( C) temperature ( C) Energy production 1290.8 GW∙h/year
Design flow discharge 65 m3/s
M142 M032 M196 M142 M032 M196 Maximum reservoir volume 14351337 m3
Mean 17.96 26.11 26.29 7.67 15.14 22.23 Minimum reservoir volume 8351337 m3
Standard deviation 2.02 1.68 1.15 2.00 1.01 3.34 Maximum sediment volume 2356267 m3
Minimum 14.00 18.50 20.20 2.40 8.50 10.34 Dead storage 1852501 m3
Median 18.00 26.40 26.30 8.00 15.20 23.80 Normal operating level 792.86 m
Maximum 24.80 36.50 34.80 12.00 20.20 26.00 Minimum operating level 783.33 m
162 Mohammad Mehedi Hasan, Guido Wyseure / Water Science and Engineering 2018, 11(2): 157e166

Table 4 linked to the SWAT model by means of SWAT-CUP software


Assumed climate change scenarios for future period (2045e2065). (Abbaspour et al., 2007; Abbaspour, 2007).
Scenario Temperature variation ( C) Rainfall variation (%) The efficacy of a hydrological model mainly depends on
1 þ1.1 ±5 how well the model is calibrated (Gupta et al., 1999). The
2 þ2.0 ±10 SWAT model can be calibrated manually and automatically
3 þ2.9 ±15 or through a combination of the two methods (Boyle et al.,
Note: þ indicates an increment of temperature, and ± indicates an increment 2000; Bardossy, 2007). In this study, automatic calibration
of rainfall in the wet season and a reduction of rainfall in the dry season. was performed with SWAT-CUP software for the period
from 1985 to 1991 at the outlet of the study area to avoid
conditions along with land use and soil type, and in turn, more time consumption in manual calibration (Eckhardt
enhanced the precision of load predictions and gave a better et al., 2005).
physical description of the water balance (Neitsch et al., Determination of the most sensitive parameters for a
2014). Considering the watershed area and the number of watershed or sub-basin is the primary step in the calibration
sub-basins generated in this study, a threshold value of 20% and validation processes of the SWAT model (Arnold et al.,
was determined for land use, which means that a certain land 2012). Over-parameterization of a complex model often
use occupying less than 20% of a particular sub-basin was leads to complications of parametric nonuniqueness and
excluded from modeling, and the areas of other land uses were equifinality in hydrological models, particularly for distributed
proportionately increased to make up 100% of the sub-basin. models, which may negatively impact prediction uncertainties
Meanwhile, a threshold value of 10% was determined for (Schoups et al., 2008). Therefore, sensitivity analysis was
both soil and slope, which means that soil types or slope conducted to lessen the number of parameters for efficient use
categories occupying less than 10% of the area of a particular of the model (van Griensven et al., 2006).
sub-basin were excluded from modeling, and the areas asso- Model validation was accomplished with SWAT-CUP soft-
ciated with the other soil types or other slope categories were ware for the period from 1992 to 1998 at the outlet of the study
proportionately increased so that they made up 100% of the area. Throughout the validation, the model was run with the
land uses or soil types within the sub-basin area. For the given same model parameters that were obtained from the calibration
threshold values, a total of 132 HRUs were generated in 19 period, to observe how well the calibrated parameters func-
sub-basins. It is essential to keep in mind that the HRUs were tioned in the validation period (Shrestha et al., 2010).
not spatially adjacent and they had clustered response units.
The pixels generating an HRU may be stretched throughout 4. Results and discussion
the sub-basin. Basic prerequisites for building the model were
achieved by defining weather input data. Files including daily 4.1. Model calibration and validation
weather information at three stations (M196, M142, and
M032) were specified in this model study. Parameter sensitivity analysis indicated that a channel
The numerical simulation was performed for a period of 17 is highly responsive to groundwater movement as well as
years from 1982 to 1998 with the SWAT model. The cali- surface and subsurface releases. The baseflow recession factor
bration and validation periods were determined to be 1985 to (ALPHA_BF) and the delay time for aquifer recharge
1991 and 1992 to 1998, respectively, considering the avail- (GW_DELAY), which control the groundwater behavior, were
ability of river flow data at the control point of the study area. found to be the most influential parameters for the sub-basin.
The first three years (1982e1984) were considered a warm-up This can be attributed to the fact that the hydrological regime
period, allowing the model to make the hydrologic cycle fully of this region is highly influenced by a slow flow response
functional and stabilize some initial model parameters. Since (Buytaert et al., 2007). Since the area is dominated by low-
the model did not perform well in daily simulations, monthly permeability layers, the influence of ALPHA_BF was antici-
data were used in this study. Model performance was assessed pated. The slope of the sub-basin (SLSUBBSN) was found to
with statistical indices and hydrograph shape during the be one of the geomorphological factors relating to the basin
calibration and validation periods. response behavior. The initial soil conservation service (SCS)
curve number for moisture condition II (CN2), the soil evap-
3.5. Sensitivity analysis, calibration, and validation of oration compensation factor (ESCO), and the effective hy-
SWAT model using SUFI-2 algorithm draulic conductivity of the channel (CH_K2) were found to
affect the surface runoff and other characteristics of the basin.
Prior to the calibration, sensitivity analysis was executed Not all of the parameters recognized by sensitivity analysis
for the control point to identify the most influential parame- were customized during the calibration process. In this study,
ters. This also assists in saving time throughout calibration only 14 parameters from the sensitivity list were used during
since it minimizes the number of parameters to be optimized the calibration process to fit the model as closely as possible to
in an over-parameterized SWAT model. It helps in recognizing natural processes (Table 5). The ranges of parameter values in
the relative significance of different parameters. The SUFI-2 the calibration process were physically realistic (Eckhardt
algorithm was used for parameter optimization, which was et al., 2005). Therefore, the model can be applied
Mohammad Mehedi Hasan, Guido Wyseure / Water Science and Engineering 2018, 11(2): 157e166 163

Table 5
Initial and calibrated parameter values of SWAT model for study area.
Parameter Modeling process Variation range Initial value Calibrated value
v__ALPHA_BF Groundwater 0e1 0.0480 0.0252
v__GW_DELAY Groundwater 0e500 31.000 98.665
v__SLSUBBSN Geomorphology 10e150 18.293 12.546
v__RCHRG_DP Groundwater 0e1 0.0500 0.1492
v__CH_K2 Channel flow 0.01e500 0 60.918
v__HRU_SLP Geomorphology 0e1 0.1871 0.8203
r__SOL_K Soil water 0e3500 17.100 0.6625
r__CN2 Runoff 35e98 37e87 39.25e92.21
v__ESCO Evaporation 0e1 0.9500 0.3750
v__REVAPMN Groundwater 0e500 1.0000 45.719
v__GWQMN Groundwater 0e5000 0 16.221
v__PLAPS Precipitation 1000e1000 0 0.8115
r__SOL_AWC Soil water 0e1 0.2500 0.2310
v__GW_REVAP Groundwater 0.02e0.20 0.0200 0.0689
Note: v means that the existing parameter value is to be replaced by a given value, and r means that the existing parameter value is multiplied by the sum of one and
a given value.

subsequently for appraising the impacts of climate change A positive or negative PBIAS value indicates a model bias of
scenarios and/or management options. underestimation or overestimation, respectively (Gupta et al.,
ALPHA_BF was changed to fine-tune the baseflow 1999). An absolute value of PBIAS of lower than 20% is
recession curve. A higher ALPHA_BF value corresponds to a considered to be good (van Liew et al., 2005). RSR values
less steep baseflow recession curve, demonstrating quick range from 0 to þ∞. The lower the RSR is, the better the
drainage and low storage. GW_DELAY was modified to simulation performance is. NSE and R2 values in Table 6
regulate the flow timing in the subsurface. SLSUBBSN was indicate that there was strong agreement between the
considered a parameter, representing the consequences of a measured and simulated results for the calibration period,
parallel terrace in an HRU. Lessening CN2 values resulted in whereas relatively smaller values were obtained throughout the
declining runoff and increasing infiltration, baseflow, and validation period. RSR and PBIAS values also demonstrated
recharge. The depth distribution of ESCO was adjusted to the strong performance of the model. However, the model
facilitate the evaporation from the soil profile. An enhanced showed poor performance in daily simulation.
SOL_AWC value led to the reduction in streamflow, because The time sequences of the measured and simulated
the ability of soil to hold water was improved. The saturated streamflow for the calibration and validation periods (Fig. 6)
hydraulic conductivity (SOL_K), a measure of ease of water were compared graphically. It can be concluded from Fig. 6
movement through the vadose zone, was adjusted to improve that, generally, there is rational agreement between the
the subsurface flow response and to facilitate the groundwater measured and simulated hydrographs. The probable reason of
storage capacity (GWQMN), which was also tuned. CH_K2, model overestimation and underestimation of the streamflow
showing the flow movement from groundwater to rivers, was might be ascribed to the fact that there are considerable un-
found to be a quite influential parameter. The precipitation certainties regarding the estimation of some SWAT parameters
laps rate (PLAPS) was used, because the study area was that control the flow through the deep and shallow aquifers
assumed to be affected by the extreme orographic effect simulated by the model. Another reason for this model
(Emck, 2007; Mu~ noz et al., 2016). The effects of the other behavior might be an inexact estimation of hydrological con-
parameters on model outputs can be found in Neitsch et al. ditions in the soil profiles.
(2014). Model performance was evaluated based on some The underestimation of the streamflow during the wet
statistical indices (Table 6). period could be ascribed to the fact that macropore flow
NSE is a reliable statistical index used widely for appraising (through worm holes, root holes, etc.) was not simulated,
the goodness of fit of hydrological models (McCuen et al., though it might be a major part of the measured peaks. The
2006). It signifies deviations between observed and simulated overestimation of flow peaks during the wet period might be
values. NSE varies between ∞ and 1.0, and the value of 1.0 is
considered the optimal value. The simulated results are Table 6
considered to be satisfactory if NSE is greater than 0.5 and Statistical indices for model performance evaluation of calibration and vali-
good if NSE ranges from 0.65 to 0.75 (Moriasi et al., 2007). R2 dation processes for monthly discharge simulation.
expresses the strength of the linear relationship between Process NSE R2 PBIAS (%) RSR (%)
simulated and measured values, with the value ranging from Calibration 0.66 0.68 3.7 52
0 to 1. A greater R2 value signifies a better agreement between Validation 0.61 0.62 5.9 6
simulated and measured values, and usually values greater than Note: NSE is the Nash-Sutcliffe efficiency, R2 is the coefficient of determi-
0.5 are considered to be satisfactory (Santhi et al., 2001; van nation, PBIAS is the percentage of bias, and RSR is the RMSE-observations
Liew et al., 2003). The optimal value of PBIAS is 0.0%. standard deviation ratio.
164 Mohammad Mehedi Hasan, Guido Wyseure / Water Science and Engineering 2018, 11(2): 157e166

Fig. 6. Measured and simulated monthly average streamflow of study area.

attributed to the facts that the curve number (CN) technique experience a power shortage due to lack of streamflow. It is
was not sufficient to produce a precise runoff prediction for observed from Fig. 7 that there is an evident increase in hy-
a day experiencing several rainstorms, and the level of soil dropower generation in the wet season for different scenarios,
moisture and the resultant runoff curve differed from storm to as compared with the reference period, and the hydropower
storm. Lastly, the availability of weather inputs played a major generation steadily decreases in the dry period, due to the
role in model performance and accuracy. scarcity of rainfall.
It is evident from Table 7 that the average annual hydro-
4.2. Analysis of climate change scenarios power generation will increase in the future. The plant will
also face a significant shortage of power in the dry season due
The model was run with each climate change scenario and the to lack of rainfall. In the case of scenario 3, the overall annual
outputs of these three scenarios were analyzed to quantify the hydropower generation is likely to increase by 7.88% from
possible climate change impact on the hydropower generation. the reference scenario due to a 13% projected augmented
In the future period, runoff will increase with precipitation streamflow in the wet season, and an estimated drop of about
in the wet season and the warming temperature. The rate of 17% in the streamflow during the dry season will lessen the
evaporation will also increase, which will consequently affect hydropower generation up to 13.14% from the reference sce-
river flow. The increased flow will contribute to hydropower nario. More repeated droughts might make the hydropower
generation. During the dry season, the hydropower plant will schemes unprofitable and more extreme rainfall events will

Fig. 7. Comparison of predicted average monthly hydropower generation in assumed climate change scenarios with that in reference scenario.

Table 7
Predicted average annual hydropower generation for different scenarios and their deviations from reference scenario.
Scenario Average annual hydropower Increment of annual Average hydropower Drop in hydropower
generation (GW∙h) hydropower generation in dry generation in dry
generation (%) season (GW∙h) season (%)
Reference 894.56 404.78
1 914.02 2.12 375.55 7.22
2 929.78 3.78 358.62 11.40
3 970.54 7.88 351.58 13.14
Mohammad Mehedi Hasan, Guido Wyseure / Water Science and Engineering 2018, 11(2): 157e166 165

cause siltation to develop, which will consequently increase generation in this region, needs to be addressed in separate
the risk of dam failures and catastrophic flood releases. Plants studies.
can generally be operated below their design limit. It is
observed that the projected average annual hydropower gen- Acknowledgements
eration is lower than the value presented in Table 3. This is
ascribed to the turbine efficiency, generator efficiency, and The first author is grateful to the authority of Vlaamse
other physical factors, which have not been incorporated in Interuniversitaire Raad, Belgium, for providing him with the
this study. Flemish Interuniversity Council Scholarship to carry out this
research. The authors would like to acknowledge the National
5. Conclusions Institute for Meteorology and Hydrology and Promas at the
University of Cuenca, in Ecuador, for providing necessary data
Application of the SWAT model in exploring the impact of and information.
climate change on future hydropower generation has been
addressed and analyzed in this study. This paper gives an References
overview of a plausible scenario of future development of the
hydroelectric output in the Rio Jubones Basin with consider- Abbaspour, K.C., 2007. SWAT-cup: SWAT Calibration and Uncertainty
ation of hydro-climatic variabilities and different parameters. Analysis Programs, a User Manual. Swiss Federal Institute of Aquatic
Science and Technology, Duebendorf.
The results show that climate change can significantly alter the Abbaspour, K.C., Yang, J., Maximov, I., Siber, R., Bogner, K., Mieleitner, J.,
flow regime of the Jubones River, which can consequently Zobrist, J., Srinivasan, R., 2007. Modelling hydrology and water quality in
affect the hydropower potential in the long run. The scenario the pre-alpine/alpine Thur watershed using SWAT. J. Hydrol. 333(2e4),
study shows that the annual hydropower generation will likely 413e430. https://doi.org/10.1016/j.jhydrol.2006.09.014.
be improved by around 8% as a result of streamflow augmented Arnold, J.G., Srinivasan, R., Muttiah, R.S., Williams, J.R., 1998. Large
area hydrologic modeling and assessment, Part I: Model development. J.
by 13% in the wet season due to a 15% increased rainfall. Am. Water Resour. Assoc. 34(1), 73e89. https://doi.org/10.1111/j.1752-
However, the plant will experience a considerable power drop, 1688.1998.tb05961.x.
by up to 13.14% from the reference scenario during the dry Arnold, J.G., Moriasi, D.N., Gassman, P.W., Abbaspour, K.C., White, M.J.,
season, owing to a 17% reduced streamflow with an increased Srinivasan, R., Santhi, C., Harmel, R.D., van Griensven, A., van
temperature of 2.9 C under the impact of climate change. Liew, M.W., et al., 2012. SWAT: Model use, calibration, and validation.
Trans. ASABE 55(4), 1345e1352. https://doi.org/10.13031/2013.42256.
A substantial reduction of rainfall and increased temperature in Bardossy, A., 2007. Calibration of hydrological model parameters for unga-
the dry season can be the largest contributor to the decrease in uged catchments. Hydrol. Earth Syst. Sci. 11(2), 703e710.
annual precipitation as well as reduced streamflow in this area, Boyle, D.P., Gupta, H.V., Sorooshian, S., 2000. Toward improved calibration
which can be attributable to power shortage. of hydrologic models: Combining the strengths of manual and automatic
In this study, the SWAT model provided satisfactory results methods. Water Resour. Res. 36(12), 3663e3674. https://doi.org/10.1029/
2000WR900207.
in simulating the hydrological processes according to the Bradley, R.S., Vuille, M., Diaz, H.F., Vergara, W., 2006. Threats to water
reference scenario, signifying that it can be applied to fore- supplies in the tropical Andes. Science 312(5781), 1755e1756. https://
casting discharge in climate change scenarios. While SWAT doi.org/10.1126/science.1128087.
has been applied effectively with the intention of attaining the Buytaert, W., Celleri, R., de Bievre, B., Cisneros, F., Wyseure, G., Deckers, J.,
goals of this study, some boundary conditions and theories of Hofstede, R., 2006. Human impact on the hydrology of the Andean
paramos. Earth Sci. Rev. 79(1e2), 53e72. https://doi.org/10.1016/
the model should be addressed to enhance the assessed results. j.earscirev.2006.06.002.
On the whole, the study reveals that SWAT is competent in Buytaert, W., I~niguez, V., de Bievre, B., 2007. The effects of afforestation and
analyzing hydrological processes reasonably, and that the cultivation on water yield in the Andean paramo. For. Ecol. Manag.
model can be utilized as a useful tool for predicting the impact 251(1e2), 22e30. https://doi.org/10.1016/j.foreco.2007.06.035.
of climate change scenarios. Buytaert, W., Vuille, M., Dewulf, A., Urrutia, R., Karmalkar, A., Celleri, R.,
2010. Uncertainties in climate change projections and regional down-
This study provides insight into the hydrological response scaling in the tropical Andes: Implications for water resources manage-
of the Jubones River Basin to various climate change sce- ment. Hydrol. Earth Syst. Sci. 14(7), 1247e1258. https://doi.org/10.5194/
narios, which is crucial for effective and sustainable water hess-14-1247-2010.
resources management. The predicted increase in river flow Dibike, Y.B., Coulibaly, P., 2005. Hydrologic impact of climate change in
will offer an opportunity to conserve water in the wet season, the Saguenay watershed: Comparison of downscaling methods and hy-
drologic models. J. Hydrol. 307(1e4), 145e163. https://doi.org/10.1016/
which will be one of the most essential adaptations to the j.jhydrol.2004.10.012.
climate change. Watershed decision makers must have other Eckhardt, K., Fohrer, N., Frede, H.G., 2005. Automatic model calibration.
operational policies that are adapted to new climatic and Hydrol. Process. 19(3), 651e658. https://doi.org/10.1002/hyp.5613.
hydrological realities. The projected water resources could be Emck, P., 2007. A Climatology of South Ecuador: With Special Focus on the
considered in planning local water management in different Major Andean Ridgeas Atlantic-Pacific Climate Divide. Ph. D. Disserta-
tion. Friedrich-Alexander-Universit€at Erlangen-Nürnberg, Erlangen.
sectors in the study region, especially irrigation, agriculture, Enerjubones, 2014. Enerjubones S.A. https://www.celec.gob.ec/enerjubones.
and hydropower generation. This study shows that there is a (Retrieved July 15, 2014).
need to improve the collection of hydro-climatic data, and also Espinoza Villar, J.C., Ronchail, J., Guyot, J.L., Cochonneau, G., Naziano, F.,
that the use of other watershed hydrological models, together Lavado, W., de Oliveira, E., Pombosa, R., Vauchel, P., 2009. Spatio-tem-
with dynamically downscaled climatic inputs for hydropower poral rainfall variability in the Amazon Basin countries (Brazil, Peru,
166 Mohammad Mehedi Hasan, Guido Wyseure / Water Science and Engineering 2018, 11(2): 157e166

Bolivia, Colombia, and Ecuador). Int. J. Climatol. 29(11), 1574e1594. Mu~noz, P., Celleri, R., Feyen, J., 2016. Effect of the resolution of tipping-
https://doi.org/10.1002/joc.1791. bucket rain gauge and calculation method on rainfall intensities in an
Ficklin, D.L., Luo, Y.Z., Luedeling, E., Zhang, M.H., 2009. Climate change Andean Mountain gradient. Water 8(11), 534. https://doi.org/10.3390/
sensitivity assessment of a highly agricultural watershed using SWAT. J. w8110534.
Hydrol. 374(1e2), 16e29. https://doi.org/10.1016/j.jhydrol.2009.05.016. Neitsch, S.L., Arnold, J.G., Kiniry, J.R., Williams, J.R., King, K.W., 2002. Soil
Fontaine, T.A., Klassen, J.F., Cruickshank, T.S., Hotchkiss, R.H., 2001. Hydro- and Water Assessment Tool, Theoretical Documentation, Version 2000.
logical response to climate change in the black hills of South Dakota, USA. Agricultural Research Service, Temple. http://swat.tamu.edu/media/1290/
Hydro. Science J. 46(1), 27e40. https://doi.org/10.1080/02626660109492798. swat2000theory.pdf (Retrieved December 21, 2013).
Fowler, H.J., Ekstr€ om, M., Blenkinsop, S., Smith, A.P., 2007. Estimating Neitsch, S.L., Arnold, J.G., Kiniry, J.R., Williams, J.R., King, K.W., 2005. Soil
change in extreme European precipitation using a multimodel ensemble. J. and Water Assessment Tool, Theoretical Documentation, Version 2005.
Geophys. Res.: Atmosphere 112(D18), D18104. https://doi.org/10.1029/ Agricultural Research Service, Temple. http://swat.tamu.edu/media/1292/
2007JD008619. swat2005theory.pdf (Retrieved January 5, 2014).
Githui, F., Gitau, W., Mutua, F., Bauwens, W., 2009. Climate change impact on Park, J.Y., Park, M.J., Joh, H.K., Shin, H.J., Kwon, H.J., Srinivasan, R.,
SWAT simulated streamflow in western Kenya. Int. J. Climatol. 29(12), Kim, S.J., 2011. Assessment of MIROC3.2 HiRes climate and CLUE-s
1823e1834. https://doi.org/10.1002/joc.1828. land use change impacts on watershed hydrology using SWAT. Trans.
Gupta, H.V., Sorooshian, S., Yapo, P.O., 1999. Status of automatic cali- ASABE 54(5), 1713e1724. https://doi.org/10.13031/2013.39842.
bration for hydrologic models: Comparison with multilevel expert cali- Santhi, C., Arnold, J.G., Williams, J.R., Hauck, L.M., Dugas, W.A., 2001.
bration. J. Hydrol. Eng. 4(2), 135e143. https://doi.org/10.1061/(ASCE) Application of a watershed model to evaluate management effects on point
1084-0699. and nonpoint source pollution. Trans. ASABE 44(6), 1559e2129. https://
Houghton, J.T., Ding, Y., Griggs, D.J., Noguer, M., van der Linden, P.J., doi.org/10.13031/2013.7041.
Dai, X., Maskell, K., Johnson, C.A., 2001. Climate Change 2001: The Schoups, G., van de Giesen, N.C., Savenije, H.H.G., 2008. Model complexity
Scientific Basis. Cambridge University Press, Cambridge. control for hydrologic prediction. Water Resour. Res. 44(12), W00B03.
Huntingford, C., Jones, R.G., Prudhomme, C., Lamb, R., Gash, J.H.C., https://doi.org/10.1029/2008WR006836.
Jones, D.A., 2003. Regional climate-model predictions of extreme rainfall Shrestha, N.K., Shakti, P.C., Gurung, P., 2010. Calibration and validation of
for a changing climate. Q. J. R. Meteorol. Soc. 129(590), 1607e1621. SWAT model for low-lying watersheds: A case study on the Kleine Nete
https://doi.org/10.1256/qj.02.97. watershed, Belgium. Hydro Nepal J. Water, Energy Environ. 6, 47e51.
Koch, F., Prasch, M., Bach, H., Mauser, W., Appel, F., Weber, M., 2011. How will https://doi.org/10.3126/hn.v6i0.4194.
hydroelectric power generation develop under climate change scenarios? A Tamm, O., Luhamaa, A., Tamm, T., 2016. Modeling future changes in the
case study in the Upper Danube Basin. Energies 4(10), 1508e1541. https:// North-Estonian hydropower production by using SWAT. Nord. Hydrol.
doi.org/10.3390/en4101508. 47(4), 835e846. https://doi.org/10.2166/nh.2015.018.
Li, T.J., Wang, G.Q., Chen, J., Wang, H., 2011. Dynamic parallelization of Urrutia, R., Vuille, M., 2009. Climate change projections for the tropical
hydrological model simulations. Environ. Model. Software 26(12), Andes using a regional climate model: Temperature and precipitation
1736e1746. https://doi.org/10.1016/j.envsoft.2011.07.015. simulations for the end of the 21st Century. J. Geophys. Res.: Atmosphere
Madani, K., 2011. Hydropower licensing and climate change: Insights from 114(D2), D02108eD02122. https://doi.org/10.1029/2008JD011021.
cooperative game theory. Adv. Water Resour. 34(2), 174e183. https:// van Griensven, A., Meixner, T., Grunwald, S., Bishop, T., Diluzio, M.,
doi.org/10.1016/j.advwatres.2010.10.003. Srinivasan, R., 2006. A global sensitivity analysis tool for the parameters
McCuen, R.H., Knight, Z., Cutter, A.G., 2006. Evaluation of the Nash- of multi-variable catchment models. J. Hydrol. 324(1e4), 10e23. https://
Sutcliffe efficiency index. J. Hydrol. Eng. 11(6), 597e602. https:// doi.org/10.1016/j.jhydrol.2005.09.008.
doi.org/10.1061/(ASCE)1084-0699. van Liew, M.W., Arnold, J.G., Garbrecht, J.D., 2003. Hydrologic simulation
Mora, D.E., Campozano, L., Cisneros, F., Wyseure, G., Willems, P., 2014. on agricultural watersheds: Choosing between two models. Trans. ASAE
Climate changes of hydrometeorological and hydrological extremes in the 46(6), 1539e1551. https://doi.org/10.13031/2013.15643.
Paute basin, ecuadorean Andes. Hydrol. Earth Syst. Sci. 18(2), 631e648. van Liew, M.W., Arnold, J.G., Bosch, D.D., 2005. Problems and potential of
https://doi.org/10.5194/hess-18-631-2014. auto calibrating a hydrologic model. Trans. ASAE 48(3), 1025e1040.
Moriasi, D.N., Arnold, J.G., van Liew, M.W., Bingner, R.L., Harmel, R.D., https://doi.org/10.13031/2013.18514.
Veith, T.L., 2007. Model evaluation guidelines for systematic quantifica- Vuille, M., Bradley, R.S., Werner, M., Keimig, F., 2003. 20th Century climate
tion of accuracy in watershed simulations. Trans. ASABE 50(3), 885e900. change in the tropical Andes: Observations and model results. Climatic
https://doi.org/10.13031/2013.23153. Change 59(1e2), 75e99. https://doi.org/10.1023/A:1024406427519.

You might also like