See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/319187241

Hydrological Modeling and Climate Change Impact Assessment Using HBV

Light Model: A Case Study of Narayani River Basin, Nepal

Article  in  Nature Environment and Pollution Technology · August 2017

CITATIONS READS

0 361

4 authors, including:

Santosh Bhattarai Narendra Man Shakya

China Three Gorges University Tribhuvan University
6 PUBLICATIONS   3 CITATIONS    43 PUBLICATIONS   248 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Identification of suitable plant species for soil bio-engineering towards eco-safe rural roads in Nepal.... View project

Project for Hydro-microbiological Approach for Water Security in Kathmandu Valley, Nepal WaSH-Mia/SATREPS View project

All content following this page was uploaded by Santosh Bhattarai on 05 October 2018.

The user has requested enhancement of the downloaded file.

 Nature Environment and Pollution Technology p-ISSN: 0972-6268
Vol. 17 No. 3 pp.691-702 2018
An International Quarterly Scientific Journal e-ISSN: 2395-3454
Original Research Paper Open Access

Hydrological Modelling and Climate Change Impact Assessment Using HBV

Light Model: A Case Study of Narayani River Basin, Nepal
Santosh Bhattarai*†, Yihong Zhou*†, Narendra Man Shakya** and Chunju Zhao*
*College of Hydraulic and Environmental Engineering, China Three Gorges University, Yichang, 443002, China
**Department of Civil Engineering, Water Resource Engineering, Tribhuvan University, Pulchowk Lalitpur, Nepal
†Corresponding authors: Santosh Bhattarai and Yihong Zhou

ABSTRACT
Nat. Env. & Poll. Tech.
Website: www.neptjournal.com In this study, a semi-distributed conceptual hydrological model “HBV-Light” is applied to one of the
snow fed basins “Narayani River Basin” in Nepal to estimate runoff at several gauging stations and to
Received: 16-06-2017
analyse the changes in catchment hydrology and future flood magnitude due to climate change. The
Accepted: 21-08-2017
model was calibrated for the period 1995-2005 and validated for the period 2006-2008 with satisfactory
Key Words: results producing values of Nash-Sutcliffe coefficient between 75.2% and 82.6% during calibration
Hydrological modelling and 56.3% and 87.2% during validation for all the four sub-basins. The value of coefficient of
Climate change determination (R 2) during calibration is between 0.789 and 0.844 and during validation is between
HBV light model 0.629 and 0.893. Due to the structural complexity, the model underestimates the low flows, whereas
Narayani river basin the peaks were correctly estimated except for some sharp peaks due to isolated precipitation events.
Further, the volumetric error during the calibration period is acceptable. Contribution of snowmelt to
annual, summer (March-July, MAMJJ) and winter (November-February, NDJF) runoff at the final outlet
gauging station (Narayani River at Narayanghat) is 15.72% (avg.), 26.22% (maximum in year 2002)
and 0.42% (minimum in year 1996) respectively. Sensitivity analysis (increased temperature) indicates
that global warming leading to increase in average basin temperature will significantly lead to higher
contributions to runoff from snowmelt. The model simulates an overall increase in monthly stream flow
from January to June (34% to 51%) and November-December (10% to 15%) with the output of
HADCM3 GCM, A1B scenario.

INTRODUCTION plains. Most of these rivers are glacier-fed and provide sus-
tained flow during the dry season to fulfil the water require-
Nepal is one of the richest countries of the world in water ment downstream. Global warming has a significant effect
resources (Aryal 2011). Its climate is strongly influenced by on the runoff from such glaciarized and snow fed catch-
its topography, which varies greatly between the north- ments. The changes in the runoff characteristics triggered
western border, at very high altitudes (above 8000m) in the by global warming has an impact over large areas.
Himalayan mountain range and the south-eastern edge of Glaciarized catchments are characterized by temporarily
the country which sits in the northern rim of the Gangetic storing precipitation and releasing it with a time delay.
plain at only around 300m above sea level. The lowland Winter precipitation is stored as snow and ice and released
regions of Nepal have a warm and humid sub-tropical cli- during the spring and summer seasons.
mate, with temperature around 22-27°C in summer months,
According to Aryal (2011) and Eriksson et al. (2009),
and dropping to 10-15°C in the winter. The high altitude water resource sector (hydropower production, irrigation
mountainous regions are considerably colder, at 5-15°C in facility) was heavily affected with the global warming and
summer months, and remaining well below zero in the win- temperature rise at an annual rate of 0.04-0.06°C per year
ter. Monsoon rainfall arrive in June and continue until Au- over Nepal, especially in Himalayan region. The main cause
gust or September, bringing 250-450mm of rainfall per of the increase in the frequency and magnitude of extreme
month in most parts of the country, but only 100-150mm in events like floods and droughts are triggered from climate
the north-western mountain regions whereas, the winter change (Sillmann & Roeckner 2007) worldwide. The effect
months are very dry and all regions receive less than 50mm of climate change on snow water equivalent, snowmelt run-
of rainfall per month (UNDP 2012). A recent climate classi- off, glacial melt runoff and total stream flow is examined in
fication of Nepal was carried out by Karki et al. (2015). many Himalayan rivers. Vavrus (2007) stated that climate
Nepal has a dense network of more than 6,000 rivers change impacts on hydrological systems are severe, espe-
flowing from the Himalayan Mountains to the hills and cially in mountain regions as it causes significant altera-
692 Santosh Bhattarai et al.

Fig. 1: Location of study area.

tions in the annual cycle of runoff (found that with an in- for water resource planning and management and policy
crease of 1.3°C in temperature, the annual snowmelt runoff, formulations regarding irrigation, agriculture, hydropower
glacial melt runoff and the total stream flow also increases). production and flood protection. In this scenario, the ne-
Meteorological data of the previous century also suggest a cessity of formulating snow and glacier melt runoff model
global mean temperature rise of 0.07°C per decade (Folland is reflected. Different kinds of hydrological models like,
et al. 2001, Jones & Moberg 2003). Globally observed annual HBV (Normand et al. 2010, Shrestha & Alfredsen 2011),
precipitation has reportedly increased by up to 0.98% per SRM model (Immerzeel et al. 2010, Khadka et al. 2014),
decade in the twentieth century (New et al. 2001). The SPHY model (Lutz et al. 2014) and J2000 model (Nepal et
frequency of severe floods in large river basins has increased al. 2014) were used to evaluate the scenarios of climate
during the 20th century (Milly et al. 2002). change impact on hydrological regime and on river catch-
Prediction of snow and glacier melt runoff from high ments. Hock (2003) stated that the snow and glacier melt
Himalayas is of great importance for the planning and design process in the Himalayan region can be conceptualized by
of hydropower project, flood warning system, irrigation simple or complex approaches depending upon the data
projects, dry season water management, climate change availability. Understanding the hydrologic response of the
foresight and inventory of water resources potential in local basin to physical (land use) and climatic (rainfall and air
as well as regional scale. But the direct field observations temperature) change is an important component of water
are very difficult to carry out because of rugged and remote resource planning and management (Vorosmarty et al.
mountain terrain. The quantitative assessment of snow and 2000).
glacier melt contribution to the river flow has been limited In this study, an attempt has been made to assess the
since melting process is very complicated and not well climate change impact on the future river discharge in
understood. In the Himalayan basins, where most parts are Narayani River basin with the help of semi-distributed con-
inaccessible and snow cover data from conventional methods ceptual hydrological model (HBV-Light Model;
are nonexistent, satellite remote-sensed observations Hydrologiska Byråns Vattenbalansavdelning). To achieve
provide the only viable alternative for acquiring snow cover this end, outputs from climate change scenarios [HadCM3
data necessary for hydrologic forecasting of snowmelt run- A1B scenario achieved from PRECIS (2002) developed
off. A precise, comprehensive database of climate change using the Providing Regional Climate for Impact Studies
impact is needed in order to conceptualize better strategies Regional Climate Model (PRECIS RCM)], are used as in-

Vol. 17, No. 3, 2018  Nature Environment and Pollution Technology

 HYDROLOGICAL MODELLING AND CLIMATE CHANGE IMPACT ASSESSMENT 693

and potential evapo-transpiration. The model computes

snow accumulation, snow melt, and storage in soil moisture
and groundwater and runoff from the catchment. The model
consists of different routines representing snowmelt by a
degree-day method, soil water and evaporation, groundwater
described by three linear reservoir equations and channel
routing by a triangular weighting function (Seibert 1997).
Descriptions of the model can be found in (Bergstrom 1976,
1992, Harlin & Kung 1992).
HBV-Light (Seibert 2005) is a recent version of the HBV
model. HBV-Light version 3 employed in this project
corresponds to the SMHI version 6 developed by Bergstrom.
This model has been successfully employed in several studies
evaluating the effects of climate change on river catchments
around the world. Within HBV-Light, there are process
parameters which do not necessarily have a physical
correspondence within a catchment. Reasonable ranges for
the parameter values are first estimated and then calculated
Fig. 2: Main structure of the HBV-light model
(Killingtveit & Sælthun 1995).
through calibration. The only physical features to be
specified within the model are mean catchment elevation
put into the HBV hydrological model to estimate the river and elevation of precipitation and temperature gauges. An
discharge in the present and future climate. advantage of the HBV-Light model is that Monte-Carlo
simulations can be performed using random numbers from a
STUDY AREA uniform distribution within the set ranges for each param-
eter. A “warming up” period has been included in HBV-
The study area is Narayani River basin (Fig. 1) situated in Light. The main structure of the HBV-Light model is shown
Nepal. It extends from Lat. 27°21’ to 29°20’ and Lon. 82°53’ in Fig. 2.
to Lon. 86°13’ covering an area of 26800 sq.km (in Nepal
only). The Narayani Basin includes the Himalayan range to Model input data: The model requires basic spatial input
the plains of Terai, with the elevation varying from 8200m datasets i.e., digital elevation model (DEM), evaporation
to 185m. Narayani River is a perennial, torrential, turbulent data and meteorological data. The brief methodology for
and undisturbed river that originates from the Himalayas preparation of the data is described as below.
and carries snow fed flows with significant discharge even Digital elevation model: The main applications of GIS in
in the dry period. Its final outlet point is gauged at the hydrological models are delineating watersheds and
Narayanghat. The main contribution of the flow of the streams, and defining slope, aspect, area, flow direction and
Narayani Basin is from Kaligandaki River, originated from flow length of catchment (Shrestha 2012). The GIS used for
Mustang district and from Trisuli, Madi Khola, Marsyangdi, this study is mainly shape file of point networks for obser-
and Budhigandaki rivers. The river is being used for irriga- vation of temperature, precipitation, evaporation and run-
tion at various locations and its major tributaries are also off, catchment area and DEM of catchment. The required
being used for hydropower, water supply and irrigation pur- shape files including that of the rivers are collected from the
pose (Fig. 1). Department of Hydrology and Meteorology (DHM). The
MATERIALS AND METHODS Table 1: List of temperature stations used in the study.

Brief description of HBV-light model: The precipitation Index No. Name Elevation Mean daily
(m) temperature (°C)
runoff model HBV (HBV is an acronym formed from
Hydrologiske Byrån avdeling for Vattenbalans at SMHI, 60 4 ThakMarpha 25 66 12 .6
Sweden), like most of the hydrological models, especially 80 4 Pokhara Airport 82 7 21 .3
for estimation of snowmelt, is based on the degree day 81 4 Lumle 17 40 15 .9
81 6 Chame 26 80 10 .6
method. The HBV model is a conceptual precipitation-runoff 10 38 DhuniBesi 10 58 22 .0
model which is used to simulate the runoff process in a
catchment based on the data of precipitation, air temperature Source: DHM Nepal

Nature Environment and Pollution Technology  Vol. 17, No. 3, 2018

694 Santosh Bhattarai et al.

Fig. 3: Elevation zoning of the Narayani river basin.

Fig. 4: Location of the different selected meteorological stations.

catchment delineation is performed in Arc Map 10.0 using input of HBV-Light model for this type of analysis consist
the hydrology extension in the spatial analyst tool within of daily precipitation, temperature and evaporation. These
Arc Toolbox (ESRI). The grid projection used for all the data were collected from DHM, Government of Nepal.
raster files and shape files employed in this study is the Similarly, other hydrological parameter discharge data were
Everest_Adj_1937_Transverse_Mercator projected coordi- collected for the gauzed stations within the study basin for
nate system. The Narayani River basin is divided into 4 sub calibration and validation of the model.
basins (Fig. 1) to carry out the study at different points in Evaporation data: Four evaporation stations index no: 604,
the basin. Fig. 3 shows the elevation zoning of the Narayani 804, 814 and 902 (Fig. 4) exist in the entire Narayani River
River basin. basin. For the purpose of this study, evaporation data from
Meteorological data: Meteorological data required as an 814 (Lumle) obtained from the DHM were assumed to ap-

Vol. 17, No. 3, 2018  Nature Environment and Pollution Technology

 HYDROLOGICAL MODELLING AND CLIMATE CHANGE IMPACT ASSESSMENT 695

(a) (b)

(c) (d)
Fig. 5: Observed and simulated discharge (mm/day) during calibration period: (a) Kaligandaki River at Kumalgaon, (b) Trisuli River
at Betrawati, (c) Budhigandaki at Arughat (d) Narayani River at Narayanghat.

proximately reflect the evaporation scenario of the Narayani tinuity of precipitation data are very important in statisti-
basin as whole due to lack of good quality data in other cal analyses such as time series analyses (Silva et al. 2007).
stations. In this study, generally accepted interpolation techniques
Precipitation data: For a satisfactory calibration and vali- (arithmetic mean method) are used to fill the gaps (miss-
dation of rainfall-runoff modelling using the HBV model, ing observations).
data for at least 10 years is required to increase the possi- Temperature data: Air temperature data (maximum and
bility of including both dry and wet years. Rainfall data minimum temperature) were available from 5 meteorological
were collected for 69 meteorological stations (Fig. 4) within stations (Fig. 4) within the basin at different elevation rang-
the basin. The period of data (1994 to 2008) was selected ing from 825m asl to 2680m asl. The details of the available
taking into consideration the availability of continuous temperature stations are listed in Table 1. The mean air tem-
observations. This continuity is important to compare the perature for each day is used as input to the HBV model for
average monthly rainfall over the same period and to evalu- the differentiation of the precipitation as snowfall or rain-
ate the quality of observations. The consistency and con- fall and for the computation of snow melt and potential

Nature Environment and Pollution Technology  Vol. 17, No. 3, 2018

696 Santosh Bhattarai et al.

(a) (b)

(c) (d)
Fig. 6: Scatter plots: (a) Kaligandaki River at Kumalgaon, (b) Trisuli River at Betrawati, (c) Budhigandaki River at Arughat,
(d) Narayani River at Narayanghat.

evapo-transpiration. In this study, an average value of lapse hydrological conditions and characteristics across the wa-
rate -0.6°C/100m is used to compute the temperature at el- tershed. These parameters need to be calibrated to adequately
evations different from the temperature at the measuring simulate the stream discharge.
station. Parameterization of HBV-light model: The parameters in
Model calibration and validation: The calibration and HBV-light model are process parameters i.e., they are not
validation were carried out at monthly time period using physically measurable and thus must be calibrated. Physi-
gauged discharge data available from DHM of Nepal for the cal interpretations of the parameters of a conceptual model
years from 1994 to 2008. The data from 1994 to 1995 was are normally very vague and should be regarded with sound
used for warming up and initialization of the model variables skepticism (Bergstrom 1992). As, there is no established
and this period was not used for evaluation of the model best method to estimate the model parameters, a variety of
predictions. The data from 1995 to 2005 were used for ac- different methods have been used for this purpose in previ-
tual calibration. Similarly, data from 2006 to 2008 were ous studies. Harlin & Kung (1992) estimated reasonable
used for validation of the model. The HBV-light model in- ranges of the parameter values by selecting the minimum
cludes a large number of parameters that describe different and maximum values of each parameter from eight autono-

Vol. 17, No. 3, 2018  Nature Environment and Pollution Technology

 HYDROLOGICAL MODELLING AND CLIMATE CHANGE IMPACT ASSESSMENT 697

(a) (b)

(c) (d)
Fig. 7: Observed and simulated discharge vs time (validation): (a) Kaligandaki river at Kumalgaon, (b) Trisuli River at Betrawati,
(c) Budhigandaki at Arughat and (d) Narayani River at Narayanghat.

mous calibrations of two catchments. Seibert (1999) used mination (R2): The efficiency E proposed by (Nash &
300,000 Monte Carlo runs to estimate parameter values Sutcliffe 1970) is defined as one minus the sum of the abso-
based on three objective function scores. Booij (2005) used lute squared differences between the predicted and observed
the experience of previous researchers to identify the best values normalized by the variance of the observed values
parameters. during the period under investigation, whereas, the coeffi-
Criteria for model evaluation: NSE and R2 are the most cient of determination (R2) is defined as the squared value
frequently used efficiency criteria for hydrological of the coefficient of correlation (Rodgers & Nicewander
applications and flow comparisons (Krause et al. 2005). In 1988). NSE and R2 are calculated as:
this study both the above mentioned criteria and mean dif- T
t
ference error were used to evaluate the performance of the  (Q o  Qt m )2
t 1
model. NSE  1  T
t ...(1)
Nash-Sutcliffe efficiency (NSE) and Coefficient of deter-  (Q o  Oo ) 2
t 1

Nature Environment and Pollution Technology  Vol. 17, No. 3, 2018

698 Santosh Bhattarai et al.

(a)

(b)

(c)
Fig. 8: Snowmelt contribution in percentage to stream flow: (a) March-July, (b) November-February and (c) annual.

 N 
2 satisfactory when between 75% and 36%, and unsatisfactory
  (Qoi  Oo )(Qmi  Qm )  when it is less than 36% (Nash & Sutcliffe 1970).
R2   i 1 
RESULTS AND DISCUSSION
 N N  ...(2)
  (Qoi  Oo )2  (Qmi  Qm )2 
 i 1 i 1  Model calibration: Table 2 shows the optimized sets of
parameters used in calibration period for the study area. In
Where, Qo is the observed discharge, Qm is the modelled the HBV-light model, some adjustments on the range of the
discharge and Qot is the discharge at time t. The efficiency parameter values (obtained from previous studies) were ini-
varies from 0 to 100, where 100 denoting perfect fit. tially specified and sampling was done by 5000 Monte Carlo
Generally, NSE is very good when NSE is more than 75%, runs specifying certain threshold efficiency. The model was

Vol. 17, No. 3, 2018  Nature Environment and Pollution Technology

 HYDROLOGICAL MODELLING AND CLIMATE CHANGE IMPACT ASSESSMENT 699

(a) (b)
Fig. 9: Simulation of future flow for Trisuli River at Betrawati: (a) Percentage change in future stream flow as compared to the control
period and (b) Absolute changes in stream flow for the 2040s using the HadCM3 A1B scenario.

calibrated with the number of sets of parameters yielded less with value from -54.0mm to +414.8mm for Kaligandaki
after the Monte Carlo runs, those giving acceptable Nash- basin and Narayani basin respectively.
Sutcliffe efficiency (NSE). The optimized parameter values The scatter plots shown in Fig. 6 indicate the similar
are those which give the best values of NS efficiency, coef- behaviour of the observed and HBV-light simulated dis-
ficient of determination (R2) and mean difference between charges during the calibration. The efficiency values and
the observed and the simulated stream flow values. the visual inspection of the hydrographs demonstrate that
Fig. 5 shows the observed and simulated discharge dur- the performance of the HBV-light model is satisfactory.
ing the calibration period (1995-2005) for four outlet sta- During calibration it is noted that for the river basins under
tions of the study basin one of which is the final outlet point study, threshold temperature is the most critical parameter
(Narayani River at Narayanghat). From the observed because the simulations generally show that most of the
simulations, it is seen that HBV-light model generally precipitation occurs under freezing conditions when the
underestimates the peak values and the low flow period is precipitation is in the form of snow. On the other hand, most
better simulated. of the runoff is generated in summer when temperature is
Table 3 shows the NSE, R2, mean observed and simu- above the freezing point.
lated discharge and their difference for the different river Model validation: The calibrated parameter sets were used
basins during the calibration period under the study. It has for the representation of the catchment behaviour using an
been observed that the average simulated and observed dis- independent data set for the validation period from 1st
charges are close to each other. In addition, NSE values January 2006 to 31st December 2008. Performance of the
obtained for the gauging stations: Kaligandaki River at validation results are shown in Table 3 and Fig. 7. The
Kotagaon, Trisuli River at Betrawati, and Budhigandaki acceptable range of the NS score given in Table 3 over the
River at Arughat and Narayani River at Narayanghat are validation period indicates the robustness of the model as a
82.6%, 70.5%, 76.5% and 78.9% respectively. No generally reliable simulator of catchment behaviour. One out of thir-
agreed absolute threshold exists for the performance teen catchments show a higher NS efficiency in the valida-
indicators; however, based on the previous published stud- tion period compared to the calibration period. The effi-
ies, hydrological simulation of monthly values with NSE ciency range during the validation period (0.563 to 0.872)
above 75% can be considered good, satisfactory when be- is somewhat better than that in the calibration period (0.705
tween 75 and 36%, and unsatisfactory when they are less to 0.826). The efficiency is highest for Kaligandaki River at
than 36% (Moriasi et al. 2007, Nash & Sutcliffe 1970). Fur- Kotagaon compared to the other basins in the calibration
ther, during the calibration period the value of R2 for the period. However, comparison of this value between different
basin is between 0.789 and 0.844. The mean difference of river basins needs to be done carefully as this measure is
water balance (volumetric error) during calibration is very highly influenced by runoff variability (Akhtar 2008). Dur-

Nature Environment and Pollution Technology  Vol. 17, No. 3, 2018

700 Santosh Bhattarai et al.

Table 2: Optimized Values of the different model parameters during calibration.

Parameters
Gauging PERC UZL K0 K1 K2 MAXBAS TT CFMAX SFCF CFR CWH FC LP BETA
stations

Arughat 11.05 0.19 0.03 0.14 0.02 1.35 -4.57 2.03 0.50 0.04 0.15 399.76 0.82 0.57
Betrawati 7.72 76.44 0.06 0.07 0.03 1.00 -1.23 1.72 0.01 0.47 0.35 122.68 0.48 0.11
Kotagaon 8.91 23.18 0.09 0.20 0.03 1.44 -1.09 2.26 0.15 0.54 0.74 99.75 0.79 1.34
Narayan- 7.10 67.85 0.45 0.70 0.01 1.51 -0.36 0.59 0.09 0.53 0.30 496.43 0.95 1.50
ghat

Table 3: Performance of the HBV model during calibration and validation period in different river basins.

Catchments Reff R2 Log Reff Water Balance (mm/yr)

Calibration Validation

Calibr- Valida- Calibr- Valida- Calibr- Valida- Qsim Qobs Mean Qsim Qobs Mean
ation tion ation tion ation tion Diff Diff

Budhigandaki 0.765 0.872 0.789 0.893 0.851 0.876 2533.50 2452.90 -80.6 2157.37 2165.37 8.23
Betrawati 0.705 0.589 0.844 0.845 0.219 0.428 5825.10 5333.70 -491.40 4698.69 4025.44 -673.14
Kotagaon 0.826 0.844 0.829 0.893 0.312 0.344 1224.00 1170.00 -54.0 916.82 910.83 -5.99
Narayanghat 0.781 0.563 0.838 0.629 0.805 0.643 1493.80 1908.60 414.8 1153.48 1702.57 549.14

Reff: Nash Efficiency, R 2 : Coefficient of Determination

Table 4: Snowmelt contribution (%) for increased temperature conditions.

Station Increase in Temperature 0C

0 0.5 1 1.5

Trisuli River at Betrawati MAMJJ 47 .7 48 .8 49 .7 50 .5

NDJF 1.3 1.3 1.3 1.4
Annual 30 .6 31 .1 31 .7 32 .0
Kaligandaki River at Kotagaon MAMJJ 43 .2 44 .6 46 .9 47 .7
NDJF 4.3 4.0 3.7 4.2
Annual 25 .6 26 .1 27 .2 27 .8
Narayani River at Narayanghat MAMJJ 22 .9 23 .0 24 .8 25 .3
NDJF 0.7 0.7 0.8 0.8
Annual 15 .3 15 .8 16 .2 16 .9

MAMJJ: March to May, NDJF: November to February

ing the validation period, the mean difference values show seen from Fig. 8 that the contribution of snowmelt to an-
that in most cases, the models underestimate the discharge. nual stream flow is (27.5% to 33.7%) at Trisuli River at
The acceptable value of R2 during validation is between Betrawati and (17.75% to 30.57%) at Kaligandaki River
0.629 and 0.893, which indicates that the calibration of the at Kotagaon. As we move further downstream, contribu-
model was successful. tion has lesser values (almost 15%) annually with a maxi-
Snowmelt contribution to total runoff: Fig. 8 demonstrates mum contribution to 26.22% in year 2004 of month
the percentage contribution of snowmelt (calculated us- MAMJJ and a minimum contribution of 0.42% in the year
ing a simple water balance equation assuming total in- 1996 of month NDJF at Narayani River at Narayanghat in
coming water flow to the basin is equal to the total outgo- all the calibration years.
ing water flow from the basin) to annual runoff and runoff Further, results show that for all three basins, the maxi-
at different seasons; season with less contribution (No- mum contribution of snowmelt to the stream flow is maxi-
vember to February) and that with maximum contribution mum from March to July, and minimum from November to
(March to July) for the three main gauging station of the February. This is the reason that, with the start of summer,
basin except Budhigandaki River at Arughat. It can be the accumulated snow begins to melt and from November

Vol. 17, No. 3, 2018  Nature Environment and Pollution Technology

 HYDROLOGICAL MODELLING AND CLIMATE CHANGE IMPACT ASSESSMENT 701

to February, when the snow accumulates and the snow pack CONCLUSION
rises, the contribution of snowmelt to stream flow is the
least. During the summer season, this value ranges between This study attempts to estimate the snow melt contribution
16.2% and 26.22% with an average of 22.87% at the in rainfall-runoff modelling of a watershed with significant
Narayanghat station with increase in the range as we move coverage of snow, for which the snow melt discharge was
upward (an average value of 47.66% at Trisuli River at estimated by application of general water balance method
Betrawati of summer flow and an average value of 43.21% aided with a lumped runoff model (HBV-light model). Based
at Kaligandaki River at Kotagaon). Contribution of on the analysis, the following conclusions have been drawn
snowmelt from November to February shows simultaneous from the present study:
decrease at the downstream outlets. At the most downstream 1. HBV-light model has been proven to be very effective
outlet of the basin at Narayanghat, this value ranges be- to simulate stream flow and snowmelt effectively in the
tween 0.42% and 1.26 %. snow fed basin. The model was calibrated for the period
Sensitivity analysis: Sensitivity analysis was performed for 1995-2005 and validated for the period 2006-2008 with
increasing temperature scenarios to estimate the percentage satisfactory results producing values of Nash-Sutcliffe
contribution of snow and glacier melt discharge. The model coefficient between 75.2%-82.6% during calibration and
is simulated for snow melt discharge estimated by increas- 56.3%-87.2% during validation for all four sub-basins.
ing the temperature by 0.5°C, 1°C and 1.5°C and the cali- The value of coefficient of determination (R2) during
bration period (1995-2005) is taken as the base period for the calibration is 0.789-0.844, and during validation is
this analysis. Sensitivity test has been done for the three 0.629-0.893.
main gauging stations given in Table 4. The results indi- 2. The model performance is highly sensitive on the choice
cate that the snowmelt contribution runoff increases along of the values of the parameter sets. Different sets of pa-
with an increased temperature. rameters can give same efficiency. This leads to ambigu-
ity in determining the best parameter set. So, the genera-
Simulation of future flow using the validated model: In tion of a large number of parameter sets using the Monte
this study, future flow pattern was simulated at only one Carlo method helps in prioritizing the important param-
gauging station (Trisuli River at Betrawati) under the study. eters to be used during calibration.
The HBV-light model was forced with the statistically 3. The contribution of snow melt discharge to the total
downscaled data available from HadCM3 GCM for the A1B flow decreases downstream. At the most downstream sta-
scenario. The model output was evaluated over two time tion Narayanghat, contribution of snowmelt to annual,
slices; the base period from 1970 to 2000 and 2030 to 2060. summer (March-July, MAMJJ) and winter (November-
Temperature and precipitation from the control period (1970 February, NDJF) runoff at the final outlet gauging sta-
to 2000) with the HadCM3 A1B scenario were used as input tion (Narayani River at Narayanghat) is 15.72% (avg.),
to the model. The model was run with the best parameter set 26.22% (maximum in year 2002) and 0.42% (minimum
obtained during calibration. The output from the model is in year 1996) respectively.
runoff in mm. 4. The result of sensitivity tests demonstrate that the im-
Fig. 9a shows the percentage change in monthly stream pact of climate change (i.e., increase in temperature) to
flow over the two time periods under study. The model stream flow is significant. Increase in temperature causes
shows an overall increase in monthly stream flow. In an enhancement in the annual and seasonal stream flow
Trisuli at Betrawati the model simulates increase in stream along with the snowmelt contribution to the stream flow.
flow in the months of January to June (34% to 51%). How- 5. Running the model with climate change outputs of the
ever, in the months with minimum stream flow (Novem- HadCM3 RCM simulates a significantly wetter monthly
ber-December), the model simulates a rather conservative flow regime in the 2040s, particularly in the main sea-
increase in stream flow (10% to 15% of the control period son of flooding. This highlights the changes in flood
values). Such a difference in stream flow can have large magnitude due to climate change.
consequences for water abstraction activities in the basin. 6. To simulate the snowmelt more accurately (in our study,
Fig. 9b shows the absolute changes in stream flow dis- we use simple water balance approach) it is recommended
charge in cumec in the monthly flow regime for the 2040s. that the model is run in hourly time steps. Besides, in
By 2040s, the model simulates a significantly wetter order to attain the better results, it is recommended to
monthly flow regime, particularly in the main season of consider the glacier melt by using satellite images to
flooding. This highlights the changes in flood magnitude extract the glaciarized area and the use of spatially dis-
due to climate change. tributed hydrological models for snow and glacier melt.

Nature Environment and Pollution Technology  Vol. 17, No. 3, 2018

 702 Santosh Bhattarai et al.

Further, the use of two or more climate change scenarios efficiency criteria for hydrological model assessment. Adv.
is recommended to assess the uncertainty of climate Geosci., 5: 89-97.
Lutz, A.F., Immerzeel, W.W., Shrestha, A.B. and Bierkens, M.F.P.
change impact studies that arise due to the use of differ- 2014. Consistent increase in high Asia’s runoff due to increasing
ent future climate data. glacier melt and precipitation. Natural Climate Change, 4(7): 587-
59 2
ACKNOWLEDGEMENT Milly, P.C.D., Wetherald, R.T., Dunne, K.A. and Delworth, T.L. 2002.
Increasing risk of great floods in a changing climate. Nature, 415:
We are truly grateful to the Department of Hydrology and 514-517.
Meteorology (DHM) of Nepal for providing all the necessary Moriasi, D.N., Arnold, J.G., Van Liew, M.W., Bingner, R.L., Harmel,
data for this study. R.D. and Veith, T.L. 2007. Model evaluation guidelines for sys-
tematic quantification of accuracy in watershed simulations. Trans-
REFERENCES actions of the ASABE, 50(3): 885-900.
Nash, J.E. and Sutcliffe, J.V. 1970. River flow forecasting through
Akhtar, M. 2008. The climate change impact on water resources of conceptual models, Part I - A discussion of principles. J. Hydrol.,
upper Indus basin-Pakistan. Doctor of Philosophy. Institute of 10: 282-290.
Geology, University of the Punjab, Lahore, Pakistan, 15: 66. Nepal, S., Krause, P., Flugel, W.A., Fink, M. and Fischer, C. 2014.
Aryal, R.S. (Ed.). 2011. Water resources of Nepal in the context of Understanding the hydrological system dynamics of a glaciated
climate change. Government of Nepal, Water and Energy Com- alpine catchment in the Himalaya region using the J200
mission Secretariat. hydrological model. Hydrological Process, 28: 1329-1344.
Bergström, S. 1976. Development and Application of a Conceptual New, M., Todd, M., Hulme, M. and Jones, P. 2001.Precipitation
Runoff Model for Scandinavian Catchments. SMHI Norrköping measurements and trends in the twentieth century. International
RHO 7, pp. 134. Journal of Climatology, 21: 1889-1922.
Bergström, S. 1992. The HBV Model - Its Structure and Applications. Normand, S., Konz, M. and Merz, J. 2010. An application of the HBV
SMHI Norrköping RH No 4, pp. 35. model to the Tamor basin in Eastern Nepal. Journal of Hydrology
Booij, M.J. 2002. Appropriate modeling of climate change impacts and Meteorology, 7(1): 49-58.
on river flooding. PhD Thesis, University of Twente, The Nether- PRECIS 2002. The Hadley Centre regional climate modelling system.
lands, pp. 206. PRECIS-Update 2002. Providing Regional Climates for Impacts
Eriksson, M., Jianchu, X., Shrestha, A.B., Vaidya, R.A., Nepal, S. and Studies. Meteorological Office, Hadley Centre, Bracknell, UK.
Sandstrom, K. 2009. The Changing Himalayas: impact of climate Rodgers, J.L. and Nicewander, W.A. 1988. Thirteen ways to look at
change on water resources and livelihood in the greater Himala- the correlation coefficient. The American Statistician, 42: 59-66,
yas, ICIMOD, Kathmandu. doi: 10.2307/2685263.
Folland, C.K., Rayner, N.A., Brown, S.J., Smith, T.M., Shen, S.S.P., Seibert, J. 1997. Estimation of parameter uncertainty in the HBV
Parker, D.E., Macadam, I., Jones, P.D., Jones, R.N., Nicholls, N. model. Nord. Hydrol., 28(4/5): 247-262.
and Sexton, D.M.H. 2001. Global temperature change and its Seibert, J. 1999. Regionalization of parameters for a conceptual rain-
uncertainties since 1861. Geophysical Research Letters, 28(13): fall-runoff model. Agricultural and Forest Meteorology, 98-99:
2621-2624. 27-293
Harlin, J. and Kung, C.S. 1992. Parameter uncertainty and simulation Seibert, J. 2005. HBV Light Version 2 User’s Manual. University of
of design floods in Sweden. J. Hydrol., 137: 209-230. Stockholm.
Hock, R. 2003. Temperature index modeling in mountain areas. Journal Shrestha, J.P. 2012. Regional Modeling for estimation of runoff from
of Hydrology, 28(1): 104-115. ungauged catchment, Case study of the Saptakoshi basin, Nepal.
Immerzeel, W.W., Van Beek, L.P.H. and Bier Kens, M.F.P. 2010. Master's thesis, Institutt for Vann-og Miljøteknikk.
Climate change will affect the Asian water towers. Science, 328: Shrestha, S. and Alfredsen, K. 2012. Application of HBV model in
1381-1385. hydrological studies of Nepali river basins: A case study. Hydro
Jones, P.D. and Moberg, M. 2003. Global surface temperatures over Nepal: Journal of Water, Energy and Environment, 8: 38-43.
the past two millennia. Geophysical Research Letters, 30(15): Sillmann, J. and Roeckner, E. 2007 Indices for extreme events in
1820, doi: 10.1029/2003GL017814 projections of anthropogenic climate change. Climatic Change,
Karki, R., Talchabhadel, R., Aalto, J. and Baidya, S.K. 2015. New 86: 83-104
climate classification of Nepal. Theoretical and Applied Climatol- Silva, R.P., Dayawansa, N.D.K. and Ratnasriri, M.D. 2007. A com-
ogy, 125(3-4): 799-808. DOI 10:1007/s00704-015-1549-0 parison of methods used in estimating missing rainfall data. Jour-
Khadka, D., Babel, M.S., Shrestha, S. and Tripathi, N.K. 2004. Cli- nal of Agricultural Sciences, 3(2).
mate change impact on glacier and snowmelt and runoff in UNDP 2012. UNDP Climate Change Country Profiles, Nepal. http://
Tamakoshi basin in the Hindu Kush Himalayas (HKH) region. country-profiles.geog.ox.ac.uk.
Journal of Hydrology, 511: 49-60. Vorosmarty, C.J., Green, P., Salisbury J. and Lammers, R.B. 2000.
Killingtveit, Å. and Sælthun, N.R. 1995. Hydrology: Hydropower Global water resources: Vulnerability from climate change and
Development. Vol. 7, Trondheim, Norway: Norwegian Institute population growth. Science, 289: 284-288.
of Technology, Division of Hydraulic Engineering. Vavrus, S. 2007. The role of terrestrial snow covers in the climate
Krause, P., Boyle, D. P. and Base, F. 2005. Comparison of different system. Climate Dynamics, 29: 73-88.

Vol. 17, No. 3, 2018  Nature Environment and Pollution Technology

View publication stats