-
Vol.:(0123456789)1 3
Applied Water Science (2019) 9:110
https://doi.org/10.1007/s13201-019-0993-x
ORIGINAL ARTICLE
A case study for the assessment
of the suitability of gridded reanalysis weather
data for hydrological simulation in Beas river basin
of North Western Himalaya
Tanmoyee Bhattacharya1 · Deepak Khare1 ·
Manohar Arora2
Received: 16 May 2018 / Accepted: 16 May 2019 / Published
online: 24 May 2019 © The Author(s) 2019
AbstractThe major problem of estimating snowmelt runoff for Beas
river basin is inadequacy of observed meteorological data
dis-tributed across the basin. In this study, ERA-Interim global
reanalysis data have been used for assessing the stream flow and
sediment yield in Beas river basin of North Western Himalaya. The
snow module of ARCSWAT hydrology model has been simulated by
integration of subbasin-wise elevation band files for modeling
snowmelt runoff process including sedi-ment yield due to rainfall
and temperature change for different elevation bands varying from
361 to 6188 m. The gridded reanalysis (0.125° × 0.125°)
dataset produces a decreased maximum and minimum temperature and
increased precipitation at higher elevation in comparison with IMD
gridded weather data. The outcome of this study conveys that the
reanalysis data represent better snowmelt runoff (NSE = 0.76, 0.70
and R2 = 0.80, 0.70) and sediment yield (NSE = 0.50, 0.53 and R2 =
0.72, 0.57) mechanism at Pong and Pandoh dams than IMD gridded
weather data (NSE = 0.50, 0.47 and R2 = 0.65, 0.60) for stream flow
and (NSE = 0.50, 0.53 and R2 = 0.65, 0.60) sediment yield during
the period 1996–1999 and 1999–2002 for these two locations.
Keywords Water balance · Beas river basin · Snowmelt
runoff · Sediment yield · Himalaya
Introduction
Weather, climate and global water cycle have a fundamental
relation. The precise prediction of stream flow is greatly
influenced by the quality of input weather data for hydrology
models (Duncan et al. 1993; Fekete et al. 2004). However,
for numerous regions, like vast snow clad areas of Himalaya, which
is at high altitude, uneven and inaccessible, there is a scanty
distribution of weather stations and excellence of his-torical
measurement is questionable. So there is a challenge of finding
proper data for hydrological modeling. Accord-ing to Konig
et al. (2001) in recent decade, remote sensing
through satellite has been used over remote areas for study-ing
various processes related to hydrology because of its large
contiguous coverage. Gridded data give global cover-age compared to
traditional ground observation data as well as freely accessible
for scientific research and used for those regions where standard
in place measurement is not possible just as the Himalayas Region.
According to Berezowski et al. (2015), the finer-resolution
regional datasets are better fitted for hydrological modeling than
the global dataset having coarser resolution. The reanalysis data
with higher spatial and temporal scale provide large assemble of
climate vari-able including global extent over several decades
(Rienecker et al. 2011; Mesinger et al. 2006). There is
an inconsistency of observational network compared to constant
assimilation scheme. Moreover, reanalysis data have a significant
pros-pect for hydrology modeling (Choi et al. 2007; Fuka
et al. 2013; Vu et al. 2012). Siam et al. 2013 used
reanalysis data-sets for simulating hydrology models to regenerate
observed stream flow and prove the most exact process to examine
how the hydrological simulations done by reanalysis forecast
models. In high-altitude areas of Indus basins, gridded data-sets
use only a few station data having commonly available
* Tanmoyee Bhattacharya [email protected]
Deepak Khare [email protected]
Manohar Arora [email protected]
1 Indian Institute of Technology, Roorkee, Uttarakhand,
India2 National Institute of Hydrology, Roorkee, Uttarakhand,
India
http://crossmark.crossref.org/dialog/?doi=10.1007/s13201-019-0993-x&domain=pdf
-
Applied Water Science (2019) 9:110
1 3
110 Page 2 of 15
old observatories which don’t exhibit the topographic and
contiguous variance of precipitation in those areas (Reggiani and
Rientjes 2015). Tanaka and Kotsuki (2013) discussed estimation of
Southeast Asia runoff using four precipita-tion datasets. Getirana
et al. (2011) performed a research in Negro River Basin,
Amazon, using different precipita-tion datasets including
reanalysis data. For distributed earth system modeling, i.e.,
hydrology modeling, the high-res-olution rainfall and temperature
data are becoming highly demanded day by day. In case of
rainfall–runoff models, e.g., ARCSWAT, WetSpa and TOPMODEL,
precipitation can be used as (Arnold et al. 1998; Beven 1995;
Liu and Smedt 2004) an important factor for hydrological processes
(Berezowski et al. 2016). Vu et al. (2012) used ARCSWAT
model for comparison of output using rainfall data from five
different sources at Vietnam river basin and found that grid-ded
dataset can capture hydrological responses in this basin perfectly.
At a recent time, a lumped hydrology model was calibrated using
weather data from CFSR, NARR, MERRA, ERA-Interim reanalysis and
gridded observation database to estimate the stream flow in the
continental USA. The study reveals that the performance of NARR
forcing is better than gridded observations (Essou et al.
2016). The main focus of the study is to assess the impacts of
ERA-Interim global climatic reanalysis data and gridded IMD dataset
on ARC-SWAT hydrological model simulation. The study also
inves-tigates the vigorous changes in extreme weather condition
based on ERA-Interim reanalysis and gridded observation dataset
(IMD gridded data). The research was conducted at Beas river basin
to estimate snowmelt runoff and sediment yield at Pong dam. In
catchments with less available data, the adequacy of the study is
to validate the use of reanalysis dataset depending on high spatial
variation of precipitation and temperature. The study evaluates the
possible influence of climate change on stream flow and sediment
yield of Beas river basin up to Pong dam using Geographical
Information System and hydrological modeling tool. Spatial input
data layers needed to run the model are weather data, digital
ele-vation model (DEM), land use data and soil data. The DEM
available from Advanced Spaceborne Thermal Emission and Reflection
Radiometer Japanese sensor developed by joint operation
by NASA and Japan’s Ministry of Economy is used for
delineating the boundary and analyzing the drain-age patterns of
the watershed. Channel slope, slope length, stream length and
stream width are derived from DEM. And National Bureau of Soil
Survey and Land Use Planning (NBSS&LUP) at 1:250000 is used for
deriving soil prop-erty. Model calibration was performed by SUFI-2
against observed data obtained by many gauging stations having
potential to examine a considerable number of parameters. The IMD
meteorological data of 1° × 1° are obtained by interpolating as
ground inspections-based gridded data. Dee et al. (2011) first
described configuration of reanalysis
data ERA-Interim. The climate variables extracted from
ERA-Interim are surface air temperature, precipitation, wind speed
at 10 m, surface downwelling solar irradiance and relative
humidity data. Depending on the variable, the temporal resolution
of ERA-Interim is either 3 h (forecast) or 6 h (analysis)
(Dee et al. 2011).
Study area
The area selected for the study is the Beas basin (up to Pong
dam) that lies in the northwestern part of the Indian Himalayan
Region (Fig. 1). The study area cov-ers four districts of
Himachal Pradesh, namely Kullu, Mandi, Kangra and Hamirpur (Lakhera
et al. 2016). It has an elevation of 4361 m (14308 ft)
and is situated at geo-graphical coordinates 32°21′59″–31°16′09″N
latitude and 77°05′08″E–74°58′31″E. Beas is a main tributary of
Sutlej river, a perennial river originates from southern face of
Roh-tang Pass in Kullu district and flows about 470 km to join
Satluj river in the state of Punjab (256 km up to Pong
dam).
The Beas River is mainly fed by snowmelt during sum-mer. Some
major tributaries of the river are Tirthan, Parvati and Sainj
rivers near Sabarimala (near Kullu), Larji and Bakhil Khad (near
Pandoh dam). During winter, a consid-erable portion of river
becomes snow-covered. The biggest glacier of this basin is Parvati
glacier. The altitude of the basin varies from 802 m (near
Pandoh dam) up to 6600 m (on Northwest Border) (Lakhera
et al. 2016). The catchment area of Beas River is about
20,303 km2 out of which only 777 km2 is under permanent
snow and glaciers (http://india -wris.nrsc.gov.in).
Data and methodology
Data used
The fundamental information required for modeling is a Soil and
Water Assessment Tool (SWAT) model that includes digital elevation
model (DEM) data, land cover data, soil data, hydrological data and
meteorological data. For the study area, the datasets used are
basin, elevation and slope derived from Aster DEM at 30-m
resolution (Fig. 2a, b). Land use land cover data were
obtained from NOAA-AVHRR, global 1 km land cover product
(Fig. 2c). Soil data were obtained from National Bureau of
Soil Survey and Land Use Planning (NBSS&LUP) at 1:250,000 scale
(Fig. 2d) (Aggarwal et al. 2016). The vegetation classes
for this study area obtained from AVHRR global 1 km land cover
product are water, snow, grassland, orchards/vine-yards, deciduous
forest, evergreen forest, cropland and shrubland.
http://india-wris.nrsc.gov.inhttp://india-wris.nrsc.gov.in
-
Applied Water Science (2019) 9:110
1 3
Page 3 of 15 110
Land cover classes are converted to their equivalent SWAT LULC
classes (Table 1) to be compatible with the model’s LULC data
input requirement. However, soil tex-ture information from
NBSS&LUP is used to generate the soil parameter. According to
USDA taxonomy, viz. Typic Dystrudepts, Typic Cryopsamments, Typic
Udorthents, rock outcrops, water bodies, and glaciers and rock
outcrops, soil is classified into different categories.
Elevation band parameters for SWAT model obtained from DEM
(Fig. 3). ARCSWAT allow elevation band up to ten in each
subbasin.
The meteorological data used for the study are IMD dataset
produced by the Indian Meteorological Depart-ment for the whole
India from 1951 to 2007. Shepard method was used to interpolate the
data from weather sta-tions (Shepard 1968). The climate variables
acquired from IMD weather data are maximum temperature, minimum
temperature, rainfall, relative humidity and solar radiation. The
spatial resolution for maximum, minimum tempera-ture and solar
radiation obtained from IMD is 1° × 1°. The
maximum and minimum temperatures of IMD weather data vary from
11.5 to 35.5 °C and 2 to 25.5 °C from 1990 to 1996, and
for ERA-Interim the values are 5–31 °C and − 12 to 22 °C,
respectively.
The ERA-Interim (Dee et al. 2011) datasets are of high
spatial and temporal resolution with multiple climate vari-ables.
The ERA-Interim produced by the ECMWF (Euro-pean Centre for
Medium-Range Weather Forecasts) is atmospheric reanalysis data
available globally (Dee et al. 2011). The dataset is available
from 1979 to 2016 on a daily basis. Maximum temperature, minimum
temperature, rainfall, wind speed, relative humidity and snow water
equivalent (SWE) are obtained from ERA-Interim with 0.125°
resolution. The variation of maximum and mini-mum temperatures from
1990 to 1996 for both datasets is shown in Fig. 4a, b.
Rainfall data obtained from IMD for this study are of 0.25°
resolution. Figure 5a, b shows the spatial variation of
rainfall at monsoon, post-monsoon and pre-monsoon period for both
IMD (0.25°) and ERA-Interim datasets (0.125°).
Fig. 1 Study area of Beas river basin
-
Applied Water Science (2019) 9:110
1 3
110 Page 4 of 15
Daily precipitation, observed river discharge and sedi-ment
yield measurements for study area were obtained from BBMB (Bhakra
Beas Management Board) cover-ing the period 1989–2005 at two
discharge stations, Pong dam and Pandoh dam. The observed rainfall
data for two rain gauge stations, Pong dam (31.9676°N,
75.9471°E)
and Pandoh dam (31.6721°N, 77.0662°E), are shown in
Fig. 6a, b. Average annual rainfall of Pong dam and Pandoh dam
for the study period is 1347.49 mm and 1348.66 mm,
respectively.
Fig. 2 a DEM map, b slope map, c LULC map and d soil map of Beas
basin
Table 1 SWAT LULC classes corresponding to LULC classes from
NOAA-AVHRR
NOAA-AVHRR LULC SWAT LULC code Description
Water WATR WaterCropland AGRR Agricultural row cropsBare soil
SWRN Bare rock/sand/clayDeciduous forest FRSD Deciduous
forestEvergreen forest FRSE Evergreen forestOrchards/vineyards ORCD
Orchards/vineyardsGrassland RNGE Range grassesShrubland PAST
Pasture/hay
Fig. 3 Elevation zone map of the basin
-
Applied Water Science (2019) 9:110
1 3
Page 5 of 15 110
Fig. 4 Variation of maximum and minimum temperatures from a IMD,
b ERA-Interim of Beas basin
Fig. 5 Spatial variation of a ERA-Interim and b IMD rainfall
during monsoon (June–September), post-monsoon (October–January) and
pre-monsoon (February–May)
Fig. 6 Observed rainfall data from rain gauge station of Beas
river basin of a Pong dam and b Pandoh dam
-
Applied Water Science (2019) 9:110
1 3
110 Page 6 of 15
Methodology used
Comparison of the IMD and ERA‑Interim
For running a hydrological model like ARCSWAT, meteoro-logical
data have a great influence. In recent time, a major challenge is
finding high-quality hydrometeorological data. This study
investigates whether global reanalysis data prod-ucts can replace
observed hydrometeorological data for use in hydrological modeling
for areas where there is a lack of observed data. Moreover, spatial
resolution of regional cli-mate models is definitely important in
the mountainous areas (Ménégoz et al. 2013). It is a challenge
to obtain improved quality precipitation data due to high
heterogeneity in topog-raphy, sparse network and cold climate.
There is an impor-tance to evaluate spatial pattern and temporal
variation of existing precipitation datasets. Glacierized
basins are more sensitive to temperature because of
positive lapse rate as the temperature decreases with elevation.
When the temperature increases with elevation, it is negative, and
the temperature is zero when it has no change with elevation.
Moreover, at
high-altitude topography, orographic lifting, a process which
forces a mass of air upward can generate snowfall, cloudi-ness and
rainfall. Here the performance of IMD and ERA-Interim datasets is
compared with conventional ARCSWAT model. The ERA-Interim data are
available at 3-hourly time step, and we calculated the daily mean
for all climatic vari-ables obtained from it. In glaciated Beas
river basin where the altitude is high, we investigate the
relationship between elevation, temperature and rainfall from these
two data-sets. ArcGIS software is used to carry out a zonal
analysis between interpolated mean temperature and rainfall of
cli-mate dataset with elevation. The validation of ERA rainfall
with observed rainfall data from rain gauge station has also done
for three rain gauge stations.
Model setup, calibration and validation
ARCSWAT hydrology model has been utilized for this study to
analyze snowmelt runoff and sediment yield of Beas river basin
using IMD and ERA-Interim datasets. To get a suc-cessful ARCSWAT
run, projection and coordinate system should be properly defined.
Digital elevation model is used to extract river network. The
minimum catchment area of 2000 ha (20 km2) was set to
obtain suitable amount of sub-basin and number of hydrological
response units (HRUs). Thirty-five subbasins and 315 HRUs were
finally generated by ARCSWAT. A 10% threshold was set to land use,
soil type and slope gradient. According to Sindhu et al.
(2013) for ARCSWAT, surface runoff was estimated by Soil
Conservation Service Curve Number (SCS-CN) method.
Table 2 SWAT input variables to estimate snowmelt runoff
Variable name Description
TLAPS Temperature lapse rateELEVB Elevation at centre of the
elevation bandSNOEB Initial snow water equivalence in the elevation
bandELEVB_FR Fraction of subbasin area within the elevation
band
Table 3 Parameters for sensitivity analysis in ARCSWAT
Components Parameters Initial value Description
Minimum value Maximum value
Runoff CN2 35 98 Initial SCS runoff curve number for moisture
condition IIESCO 0 1 Soil evaporation compensation factorSOL_AWC 0
1 Available water capacity of the soil layerALPHA_BF 0 1 Base flow
alpha factorSURLAG 0.05 24 Surface runoff lag coefficientSOL_BD 0.9
2.5 Moist bulk densityESCO 0 1 Soil evaporation compensation
factorEPCO 0 1 Plant uptake compensation factorTIMP 0 1 Snowmelt
temperature lag factorSFTMP − 20 20 Snowfall temperature
Sediment USLE_K 0 0.65 Soil erodability factor in USLEUSLE_P 0 1
USLE support practice factorSPCON 0.0001 0.01 Linear parameter for
calculating the maximum amount of sediment that can be
reentrained during channel sediment routingSPEXP 1 1.5 Exponent
parameter for calculating sediment reentrained in channel
sediment
routingADJ_PKR 0.5 2 Peak rate adjustment factor for sediment
routing in subbasin
-
Applied Water Science (2019) 9:110
1 3
Page 7 of 15 110
Penman–Monteith method was used to estimate evapotran-spiration
by ARCSWAT hydrology model. For ARCSWAT hydrology model, HRUs are
segregated into several subwa-tersheds (Neitsch et al. 2005).
HRUs are the combination of similar spatial units of similar
hydrological properties and similar landforms (Neitsch et al.
2005). HRU is the cluster of similar LULC, soils and slopes for
individual subbasins and based upon user-defined thresholds. The
flow generated from SWAT model is routing of HRUs to watershed
outlet. For estimating snowmelt runoff in ARCSWAT, we have to
prepare elevation band up to ten from DEM. These eleva-tion
bands allow the model to describe
the snowmelt pro-cess based on basin topography (Hartman
et al. 1999). For
estimating snowmelt runoff by ARCSWAT, we have to esti-mate
certain variables from elevation band file (Table 2).
Lapse rate including the difference of average and gage
elevation is used as a function for calculating maximum
tem-perature, minimum temperature and precipitation for each band.
By zonal statistics of DEM and classified DEM, elevation at centre
of elevation band is calculated. Fraction of subbasin area within
each elevation band is calculated by zonal statistics with
shapefile of subbasin. Snow water equivalence obtained from
ERA-Interim for each elevation band is calculated. The lapse rate
is taken as − 6.5(°C/Km) is the decrease in atmospheric temperature
with increase of altitude. According to Liu et al. (2007),
Soil Loss Equation (MUSLE) was used for estimating
Fig. 7 Variation of mean temperature with elevation for monsoon,
pre-monsoon and post-monsoon period for IMD (a–c) and ERA-Interim
(d–f)
R² = 0.4848
18
19
20
21
22
23
0 2000 4000 6000 8000
Mea
n Te
mpe
ratu
re (°
C)
Eleva�on
R² = 0.801
-10
-5
0
5
10
15
20
25
0 2000 4000 6000 8000Mea
n Te
mpe
ratu
re (°
C)
Elevation
R² = 0.5573
24
25
26
27
28
29
30
31
32
0 2000 4000 6000 8000
Mea
n Te
mpe
ratu
re (°
C)
Elevaon
R² = 0.7995
0
5
10
15
20
25
30
0 2000 4000 6000 8000
Mea
n Te
mpe
ratu
re (°
C)
Elevation
R² = 0.4712
13
14
15
16
0 2000 4000 6000
Mea
n Te
mpe
ratu
re (°
C)
Elevaon
R² = 0.8096
-10
-5
0
5
10
15
0 2000 4000 6000 8000
Mea
n Te
mpe
ratu
re (°
C)
Elevation
(b)
(a)
(c)
(d)
(e)
(f)
-
Applied Water Science (2019) 9:110
1 3
110 Page 8 of 15
sediment yield and replacing rainfall energy factor used in
Uni-versal Soil Loss Equation (USLE) with the runoff. Channel
sediment routing computed by Bagnold’s stream power defini-tion
involves two concurrent processes, deposition and degra-dation
(Williams 1980). SWAT model has been calibrated and validated for
monthly simulated stream flows by comparing with the observed
stream flows on the Pong dam and Pandoh dam gauge stations. The
model was simulated for 1993–1996 and validated for 1999–2002.
SWAT-CUP interface has been used for model sensitivity, uncertainty
analysis, calibration and validation. For sensitivity analysis, a
comparative change in model output is observed by changing input
parameter one by
one at one time. Sensitivity analysis leads to efficient
calibra-tion due to a significant effect on the simulation output.
The simulation results at Pong dam and Pandoh dam include 15
sensitive parameters used for sensitivity analysis. Table 3
shows the parameters selected for the calibration.
Model evaluation
Model parameters were calibrated and validated for runoff and
sediment yield using the simulation and observation. In addition to
relative error, RMSE as an objective function, R2 and
Nash–Sutcliffe efficiency (Eqs. 1, 2, 3) are computed.
0
500
1000
1500
2000
2500
3000
3500
4000
0 1000 2000 3000 4000 5000 6000
Rain
fall
(mm
)
Elevation (m)
0
200
400
600
800
1000
1200
0 1000 2000 3000 4000 5000 6000
Rain
fall
(mm
)
Eleva�on (m)
0
200
400
600
0 1000 2000 3000 4000 5000 6000
Rain
fall
(mm
)
Eleva�on (m)
(a) (c)
(b) (d)
Fig. 8 Variation of rainfall with elevation for monsoon and
winter period for ERA-Interim (a, b) and IMD (c, d)
Fig. 9 Comparison of ERA-Interim rainfall data with observed
rainfall data at a Pong dam and b Pandoh dam
R² = 0.5918
0100200300400500600700800900
1000
0 500 1000 1500
ERA-
INTE
RIM
rain
fall
(mm
)
Observed rainfall(mm)
R² = 0.6821
0100200300400500600700
0 500 1000
ERA-
INTE
RIM
rain
fall
(mm
)
Observed rainfall(mm)
(a) (b)
-
Applied Water Science (2019) 9:110
1 3
Page 9 of 15 110
(1)Relative error = 1∕NN∑
i=1
||Qobs − Qsim||∕Qobs
(2)RMSE =√
1∕N∑
(Qobs
− Qsim
)2where RMSE and NSE are the root-mean-square error and
Nash–Sutcliffe efficiency, Qobs is the observed stream flow, Qsim
is the simulated stream flow, N is the total number of
(3)NSE = 1 −
[N∑
I=1
(Qobs
− Qsim
)2∕
N∑
I=1
(Qobs
− Qmeanobs
)2
]
Fig. 10 Monthly rainfall data from 1990 to 2004 from ERA-Interim
data for a Pong dam and b Pandoh dam and IMD for c Pong dam and d
Pandoh dam
0100200300400500600700800900
1000
Jan-
90
Jul-9
1
Jan-
93
Jul-9
4
Jan-
96
Jul-9
7
Jan-
99
Jul-0
0
Jan-
02
Jul-0
3
Rain
fall
(mm
)
Date
0100200300400500600700
Jan-
90
Sep-
91
May
-93
Jan-
95
Sep-
96
May
-98
Jan-
00
Sep-
01
May
-03
Rain
fall
(mm
)
Date
0
200
400
600
800
1000
Jan-
90
Jul-9
1
Jan-
93
Jul-9
4
Jan-
96
Jul-9
7
Jan-
99
Jul-0
0
Jan-
02
Jul-0
3
Rain
fall
(mm
)
Date
0
100
200
300
400
500
600
Jan-
90
Jul-9
1
Jan-
93
Jul-9
4
Jan-
96
Jul-9
7
Jan-
99
Jul-0
0
Jan-
02
Jul-0
3
Rain
fall
(mm
)
Date
(a) (c)
(b) (d)
Fig. 11 Hydrograph of the observed and calibrated monthly flow
for the calibration period of Pong dam for a ERA-Interim and b IMD
weather data
0
200
400
600
800
1000
1200
Jan-
93
Nov
-93
Sep-
94
Jul-9
5
May
-96
Stre
anflo
w (m
3 /s)
Date
0
200
400
600
800
1000
1200
Jan-
93
Nov
-93
Sep-
94
Jul-9
5
May
-96
Stre
amflo
w (m
3/s)
Date
(a) (b)Observed_StreamflowSimulated_Streamflow
Observed_Streamflow
Simulated_Streamflow
-
Applied Water Science (2019) 9:110
1 3
110 Page 10 of 15
observations and Qmeanobs
is the average of the observed stream flow.
Results and discussion
A scatter plot between elevation and temperature from IMD and
ERA-Interim shows a linear pattern for mon-soon (June–September),
pre-monsoon (February–May) and post-monsoon period
(October–January). There is a decreasing trend of temperature with
elevation for IMD with R2 value 0.50 at monsoon, 0.55 at
pre-monsoon and 0.50 at post-monsoon period. For ERA-Interim data,
we observe a sharp falling trend of temperature with R2 value 0.80
at monsoon, 0.79 at pre-monsoon and 0.81
at post-monsoon period (Fig. 7a–c, e–f). ERA-Interim
temperature varies from 7 to 25 °C during pre-monsoon, − 6.65
to 13.25 during post-monsoon and − 2.70 to 19.48 during monsoon
season with a elevation of 361 m to 6188 m. IMD
temperature varies from 25.5 to 31.6 °C during summer, 13.5 to
15.50 °C during winter and 18.5 to 22.8 °C during monsoon
with the same elevation range as ERA-Interim. After analyzing the
result, a strong linear relationship has been observed between mean
temperature and elevation for all three seasons. The regression
coef-ficient is high, 0.81, during winter for ERA-Interim and for
IMD it is high at monsoon, 0.55. The standard deviation of IMD and
ERA-Interim mean temperature is 0.87 and 3.5 which indicates that
ERA-Interim data have more varia-tion. Over the flat topography,
the precipitation decreases with latitude (Li et al. 2018).
According to Uhlenbrook
Fig. 12 Hydrograph of the observed and validated monthly flow
for the validation period of Pong dam for a ERA-Interim and b IMD
weather data
0
100
200
300
400
500
600
Jan-
99
Sep-
99
May
-00
Jan-
01
Sep-
01
May
-02
Stre
amflo
w (m
3 /s)
Date
0
100
200
300
400
500
600
Jan-
99
Sep-
99
May
-00
Jan-
01
Sep-
01
May
-02
Stre
amflo
w (m
3 /s)
Date
(a) (b)
Observed_Streamflow
Simulated_Streamflow
Observed_Streamflow
Simulated_Streamflow
Fig. 13 Hydrograph of the observed and calibrated monthly flow
for the calibra-tion period of Pandoh dam for a ERA-Interim and b
IMD weather data
0
100
200
300
400
500
600
Jan-
96
Aug-
96
Mar
-97
Oct
-97
May
-98
Dec-
98
Jul-9
9
Stre
amflo
w (m
3 /s)
Date
0
100
200
300
400
500
600
Jan-
96
Aug-
96
Mar
-97
Oct
-97
May
-98
Dec-
98
Jul-9
9
Stre
amflo
w(m
3/s)
Date
(a) (b)
Observed_Streamflow
Simulated_Streamflow
Observed_Streamflow
Simulated_Streamflow
-
Applied Water Science (2019) 9:110
1 3
Page 11 of 15 110
et al. (2004), the results of hydrology model were
signifi-cantly influenced by topographic influence on the rainfall.
We select annual rainfall of four months in each season to evaluate
precipitation pattern of monsoon (June–Sep-tember) and winter
monsoon (October–January) against elevation. Figure 8a–d shows
the variation of rainfall with elevation. According to Fig. 8,
the precipitation has an increasing trend with elevation up to
2000 m. Beyond this elevation, the precipitation range becomes
larger. A pat-tern of varying precipitation with elevation is
clearer in monsoon than winter for both IMD and ERA-Interim.
Pre-cipitation from ERA-Interim varies from 1800 to 3600 mm
during monsoon season and 300 to 1100 mm during winter
season with the elevation range 361–6188 m. The varia-tions
of rainfall from IMD for these two seasons are less than
ERA-Interim. The values are 700–2700 mm during monsoon and
250–450 mm, respectively, with elevation. In spite of a
huge variance of ERA-Interim rainfall, a sig-nificant falling
trend was observed for these two rainfall datasets at high
elevation than in low elevation.
The correlation between reanalysis and IMD data is below 0.5
during monsoon. There is a good correlation between ERA-Interim
precipitation and observed rainfall data than IMD (R2 less than
0.5). Figure 9a, b shows the good match of ERA-Interim and
observed rainfall datasets with regression coefficient (R2) of 0.60
and 0.70 at two rain
Fig. 14 Hydrograph of the observed and calibrated monthly flow
for the calibra-tion period of Pandoh dam for a ERA-Interim and b
IMD weather data
0
100
200
300
400
500
600
700
800
900
Jan-
02
Sep-
02
May
-03
Jan-
04
Sep-
04
May
-05
Stre
amflo
w (m
3/s)
Date
Observed_Streamflow
Simulated_Streamflow
0
100
200
300
400
500
600
700
800
900
Jan-
02
Sep-
02
May
-03
Jan-
04
Sep-
04
May
-05
Stre
amflo
w (m
3/s)
Date
Observed_Streamflow
Simulated_Streamflow
(a) (b)
Table 4 Relative error (%), RMSE (m3/s), NSE and R2 for
simulated river discharge
Performance indicators Location Period IMD ERA-Interim
Stream flow Relative error Pong dam Calibration 29.16
17.34Validation 25.62 18.45
Pandoh dam Calibration 150.17 87.34Validation 51.45 25.04
RMSE Pong dam Calibration 116.26 50.12Validation 90.64 55.71
Pandoh dam Calibration 124.00 93.56Validation 100.29 77.76
NSE Pong dam Calibration 0.22 0.68Validation 0.50 0.76
Pandoh dam Calibration 0.46 0.69Validation 0.47 0.70
R2 Pong dam Calibration 0.50 0.70Validation 0.65 0.80
Pandoh dam Calibration 0.50 0.75Validation 0.60 0.70
-
Applied Water Science (2019) 9:110
1 3
110 Page 12 of 15
gauge stations such as lower-magnitude station Pong dam and
middle-altitude station Pandoh dam. There is a good trend obtained
between precipitations from ERA-Interim and observed data for
middle-altitude station Pandoh dam. It proves the authenticity of
ERA-Interim precipitation data. Figure 10a–d shows the monthly
rainfall from 1990 to 2004 for both ERA-Interim and well-known
weather datasets at Pong dam and Pandoh dam. It can be observed
from Fig. 10 that the variability in the dataset is not the
same because maximum and minimum rainfall occurs in almost
differ-ent years except in a less cases. The average annual
rainfall from 1990 to 2004 for ERA-Interim is 2706.76 mm and
2061.196 mm where for IMD theses values are 1253.12 mm
and 976.46 mm, respectively, at Pong dam and Pandoh dam.
The calibration and validation of stream flow for Pong dam and
Pandoh dam are shown in Figs. 11, 12, 13 and 14 for monthly
time steps. According to Moriasi et al. (2007) model appraisal
standard at both time steps for most hydro-metric stations, it can
be observed from Tables 4 and 5 that ERA-Interim dataset
produces NSE and R2 values assessed to be satisfactory. However,
relative errors are 17.34%, 87.34% for Pong dam and Pandoh dam
during calibration, whereas 25.04% and 55.71% during validation for
ERA-Interim. For IMD the relative error changed to 29.16% and
150.17% during calibration and 25.62% and 51.45% during validation
for two observation stations. The result of ERA-Interim shows a
better performance in terms of RMSE and relative error during both
calibration and validation periods.
Table 5 Relative error (%), RMSE (m3/s), NSE and R2 for
simulated river sediment yield
Performance indicators Location Period IMD ERA-Interim
Sediment yield Relative error Pong dam Calibration 30.23
20.34Validation 25.67 18.23
Pandoh dam Calibration 73.34 55.23Validation 93.56 72.34
RMSE Pong dam Calibration 120.56 52.34Validation 92.34 55.34
Pandoh dam Calibration 190.56 135.23Validation 145.32 135.12
NSE Pong dam Calibration 0.38 0.42Validation 0.45 0.50
Pandoh dam Calibration 0.38 0.52Validation 0.40 0.48
R2 Pong dam Calibration 0.50 0.65Validation 0.60 0.72
Pandoh dam Calibration 0.51 0.67Validation 0.47 0.57
Fig. 15 Observed and simulated monthly sediment yield for model
calibration (1993–1996) of Pong dam for a ERA-Interim and b IMD
weather data
0
1
2
3
4
5
6
7
8
9
Jan-
93
Aug-
93
Mar
-94
Oct
-94
May
-95
Dec-
95
Jul-9
6
Sedi
men
t yie
ld (t
/ha)
Date
Observed(t/ha)
Simulated(t/ha)
0
1
2
3
4
5
6
7
8
9
Jan-
93
Aug-
93
Mar
-94
Oct
-94
May
-95
Dec-
95
Jul-9
6
Sedi
men
t yie
ld (t
/ha)
Date
Observed(t/ha)
Simulated(t/ha)(a) (b)
-
Applied Water Science (2019) 9:110
1 3
Page 13 of 15 110
Fig. 16 Observed and simulated monthly sediment yield for model
validation (1999–2002) of Pong dam for a ERA-Interim and b IMD
weather data
0
2
4
6
8
10
12
14
Jan-
99
Aug-
99
Mar
-00
Oct
-00
May
-01
Dec-
01
Jul-0
2
Sedi
men
t yie
ld (t
/ha)
Date
Observed(t/ha)
Simulated(t/ha)
0
2
4
6
8
10
12
14
Jan-
99
Aug-
99
Mar
-00
Oct
-00
May
-01
Dec-
01
Jul-0
2
Sedi
men
t yie
ld (t
/ha)
Date
Observed(t/ha)
Simulated(t/ha)(a) (b)
Fig. 17 Observed and simulated monthly sediment yield for model
calibration (1996–1999) of Pandoh dam for a ERA-Interim and b IMD
weather data
Observed(t/ha)
Simulated(t/ha)
Observed(t/ha)
Simulated(t/ha)
0
2
4
6
8
10
12
14
Jan-
96
Aug-
96
Mar
-97
Oct
-97
May
-98
Dec-
98
Jul-9
9
Sedi
men
t yie
ld(t
/ha)
Date
0
2
4
6
8
10
12
Jan-
96
Aug-
96
Mar
-97
Oct
-97
May
-98
Dec-
98
Jul-9
9
Sedi
men
t yie
ld(t
/ha)
Date
(a) (b)
Fig. 18 Observed and simulated monthly sediment yield for model
validation (2002–2005) of Pandoh dam for a ERA-Interim and b IMD
weather data
0
2
4
6
8
10
12
Jan-
02
Aug-
02
Mar
-03
Oct
-03
May
-04
Dec-
04
Jul-0
5
Sedi
men
t yie
ld (t
/ha)
Date
Observed(t/ha)
Simulated(t/ha)
0
2
4
6
8
10
12
Jan-
02
Aug-
02
Mar
-03
Oct
-03
May
-04
Dec-
04
Jul-0
5
Sedi
men
t yie
ld (t
/ha)
Date
Observed(t/ha)
Simulated(t/ha)
(a) (b)
-
Applied Water Science (2019) 9:110
1 3
110 Page 14 of 15
The statistical indicators for sediment yield using ERA-Interim
are R2 = 0.65, NSE = 0.42, relative error = 20.34, RMSE = 52.34
during calibration at Pong dam, and for Pandoh dam the values are
R2 = 0.67, NSE = 0.52, rela-tive error = 55.23, RMSE = 135.23.
During validation R2, NSE, relative error and RMSE values found are
0.72, 0.50, 18.23 and 55.34, respectively, for Pong dam and 0.57,
0.48, 72.34% and 135.12 for Pandoh dam. For IMD the values are R2 =
0.50, 0.51, NSE = 0.38, 0.40, relative error = 30.23%, 73.34% and
RMSE = 120.56, 190.56 for Pong dam and Pandoh dam during
calibration, whereas during validation the values are R2 = 0.60,
0.47, NSE = 0.45, 0.40, relative error = 25.67%, 93.56% and RMSE =
92.34, 145.32, respec-tively, for two stations (Figs. 15, 16,
17, 18). The statisti-cal indicator during calibration and
validation justifies the model’s better execution using ERA-Interim
data for sedi-ment yield.
This study has been done in the time period of 1993–1996 for
calibration and 1999–2002 for validation at Pong dam and 1996–1999
for calibration and 2002–2005 for valida-tion at Pandoh dam as per
observed data availability. The Nash–Sutcliffe efficiency and
regression coefficient are always better for ERA-Interim than IMD
at two locations for both stream flow and sediment yield. Moreover,
the mean monthly stream flow and sediment yield during study period
vary from 270–950 m3/s and 0–12 t/ha for Pong dam and
20–768 m3/s and 0–12.5 t/ha for Pandoh dam from
observa-tional data, whereas by using ERA-Interim and IMD data-set,
the stream flow varies from 245 to 975 m3/s and 243 to
750 m3/s at Pong dam and from 0.78 to 375 m3/s and 0.45
to 380 m3/s at Pandoh dam. For sediment yield the values
obtained are 0–9.5 t/ha and 0–7.8 t/ha at Pong dam and
0–12 t/ha and 0–7.5 t/ha at Pandoh dam, respectively. As
a conclusion, the prediction of hydrological variables using
ERA-Interim is quite satisfactory for Beas river basin up to Pong
dam.
Conclusion
In this research, an attempt has been done to investigate the
trend of ERA-Interim and IMD weather data with elevation and their
variability. The ERA-Interim temperature and rain-fall data give a
satisfactory trend and variation with elevation than IMD for all
time period in terms of standard devia-tion and correlation
coefficient. To simulate stream flow of hydrology model accurately,
it is essential to give accurate rainfall as input in hydrology
model. A good power expo-nent relationship has been found between
ERA-Interim and observed rainfall data for this study area.
Moreover, in Beas river basin where snowmelt runoff has a great
influence on generating river flow, snowmelt and corresponding
runoff response can be modified by changes in air temperature.
ERA-Interim which has a significant variation of tempera-ture
from − 2.70 to 25 °C has a great influence on snowmelt runoff
generation. This current study is a very preliminary attempt that
has taken to investigate the performance of ERA-Interim weather
data for hydrology modeling by com-paring stream flow and sediment
yield. There are many other hydrology parameters also like ET, base
flow, soil moisture, snow cover, etc. In the future study, these
parameters will be tried to be analyzed by hydrology modeling.
Moreover, SWAT hydrology model is a river basin scale hydrology
model.
Open Access This article is distributed under the terms of the
Crea-tive Commons Attribution 4.0 International License
(http://creat iveco mmons .org/licen ses/by/4.0/), which permits
unrestricted use, distribu-tion, and reproduction in any medium,
provided you give appropriate credit to the original author(s) and
the source, provide a link to the Creative Commons license, and
indicate if changes were made.
References
Aggarwal SP, Thakur PK, Garg V, Nikam B, Chouksey A, Dhote P,
Bhattacharya T (2016) Water Resources status and availability
assessment in current and future climate change scenarios for Beas
river basin of North Western Himalaya. In: The interna-tional
archives of the photogrammetry, remote sensing and spatial
information Sciences, XLI-B8, pp 1389–1396
Arnold JG, Srinivasan R, Muttiah S, Williams JR (1998) Large
area hydrologic modeling and assessment: part I. Model development.
J Am Water Resour Assoc 34(1):73–89
Berezowski T, Szczesniak M, Kardel I, Michałowski R, Piniewski M
(2015) CPLFD-GDPT5: high-resolution gridded daily precipita-tion
and temperature dataset for two largest Polish river basins. Earth
Syst Sci Data Discuss 8:1021–1060
Berezowski T, Szcześniak M, Kardel I, Michałowski R, Okruszko T,
Mezghani A, Piniewski M (2016) CPLFD-GDPT5: high-resolu-tion
gridded daily precipitation and temperature data set for two
largest Polish river basins. Earth Syst Sci Data 8(1):127–139.
https ://doi.org/10.5194/essd-8-127-2016
Beven KJ (1995) Linking parameters across scales: sungrid,
param-eterisations and scale dependent hydro logical models. Hydrol
Process 9:507–525
Choi W, Moore A, Rasmussen PF (2007) Evaluation of temperature
and precipitation data from NCEP-NCAR global and regional
Reanalysis for hydrological modelling in Manitoba. In: Paper
presented at the 18th Canadian Hydrotechnical Conference,
Winnipeg
Dee DP et al (2011) The ERA-Interim reanalysis:
configuration and performance of the data assimilation system. Q J
R Meteorol Soc 137(2011):553–597. https
://doi.org/10.1002/Qj.828
Duncan M, Austin B, Fabry F, Austin G (1993) The effect of gauge
sampling density on the accuracy of stream flow prediction for
rural catchments. J Hydrol 142:445–476
Essou GR, Sabarly F, Lucas-Picher P, Brissette F, Poulin A
(2016) Can precipitation and temperature from meteoroloical
reanalysis be used for hydrological modeling. J. Hydrometeorol
17:1929–1950
Fekete BM, Vörösmarty CJ, Roads JO, Willmott CJ (2004)
Uncer-tainties in precipitation and their impacts on runoff
estimates. J Climatol 17:294–304
http://creativecommons.org/licenses/by/4.0/http://creativecommons.org/licenses/by/4.0/https://doi.org/10.5194/essd-8-127-2016https://doi.org/10.1002/Qj.828
-
Applied Water Science (2019) 9:110
1 3
Page 15 of 15 110
Fuka DR, Walter MT, MacAlister C, Degaetano AT, Steenhuis TS
et al (2013) Using the Climate Forecast System Reanalysis as
weather input data for watershed models. Hydrol Process
28:5613–5623
Getirana ACV, Espinoza JC, Ronchail J, Rotunno O (2011)
Assessment of different precipitation datasets and their impacts on
the water balance of the Negro River basin. J Hydrol
404(3–4):304–322. https ://doi.org/10.1016/j.jhydr
ol.2011.04.037
Hartman MD, Baron JS, Lammers RB, Cline DW, Band LE, Liston GE,
Tague C (1999) Simulations of snow distribution and hydrology in a
mountain basin. Water Resour Res 35(5):1587–1603
Konig M, Winther J, Isaksson E (2001) Measuring snow and glacier
ice properties from satellite. Rev Geophys 39(1):1–27
Lakhera N, Kumar N, Shankar V (2016) Water balance study of Beas
river, Himachal Pradesh, using Arcgis technique. In: Proceedings of
national conference: civil engineering conference—innovation for
sustainability
Li H, Haugen JE, Xu C-Y (2018) Precipitation pattern in the
West-ern Himalayas revealed by four datasets. Hydrol Earth Syst Sci
22:5097–5110. https ://doi.org/10.5194/hess-22-5097-2018
Liu Y, Smedt F (2004) A GIS-based hydrological model for flood
prediction and watershed management. Documentation and User
Manual
Liu J, Williams JR, Zehnder AJB, Yang H (2007) GEPIC—modeling
wheat yield and crop water productivity with high resolution on a
global scale. Agric Syst 94(2):478–493
Ménégoz M, Gallée H, Jacobi HW (2013) Precipitation and snow
cover in the Himalaya: from reanalysis to regional climate
simulations. Hydrol Earth Syst Sci 17:3921–3936. https
://doi.org/10.5194/hess-17-3921-2013(2013)
Mesinger F et al (2006) North American regional
reanalysis. Bull Am Meteorol Soc 87:343–360. https
://doi.org/10.1175/BAMS-87-3-343
Moriasi DN, Arnold JG, Van Liew MW, Bingner RL, Harmel RD, Veith
TL (2007) Model evaluation guidelines for systematic
quan-tification of accuracy in watershed simulations. Trans ASABE
50:885–900
Neitsch SL, Arnold JG, Kiniry JR, Williams JR (2005) Soil and
water assessment tool theoretical documentation, version 2005.
Springer, Berlin
Reggiani P, Rientjes THM (2015) A reflection on the long-term
water balance of the Upper Indus Basin. Hydrol Res
46(3):446–462
Rienecker MM et al (2011) MERRA—NASA’s modern-era
retrospec-tive analysis for research and applications. J Clim
24:3624–3648. https ://doi.org/10.1175/JCLI-D-11-00015 .1
Shepard D (1968) A two-dimensional interpolation for
irregularly-spaced data. In: Proceedings, ACM national conference,
pp 517–524
Siam MS, Demory M, Eltahir EAB (2013) Hydrological cycles over
the Congo and upper Blue Nile basins: evaluation of general
circulation model simulations and reanalysis products. J Clim
26:8881–8894
Sindhu D, Shivakumar BL, Ravikumar AS (2013) estimation of
surface runoff in nallur amanikere watershed using SCS-CN method.
In: International journal of research in engineering and
technology, IC-RICE conference issue, pp 404–409
Tanaka K, Kotsuki S (2013) Uncertainties of precipitation
products and their impacts on runoff estimates through hydrological
land surface simulation in Southeast Asia. Hydrol Res Lett
7:79–84
Uhlenbrook S, Roser S, Tilch N (2004) Hydrological process
represen-tation at the meso-scale: the potential of a distributed,
conceptual catchment model. J Hydrol 291:278–296
Vu MT, Raghavan SV, Liong SY (2012) SWAT use of gridded
obser-vations for simulating runoff—a Vietnam river basin study.
Hydrol Earth Syst Sci 16:2801–2811. https
://doi.org/10.5194/hess-16-2801-2012
Williams JR (1980) SPNM, a model for predicting sediment,
phospho-rus and nitrogen yields from agricultural basins. Water Res
Bull AWRA 16(5):843–848
Publisher’s Note Springer Nature remains neutral with regard to
jurisdictional claims in published maps and institutional
affiliations.
https://doi.org/10.1016/j.jhydrol.2011.04.037https://doi.org/10.5194/hess-22-5097-2018https://doi.org/10.5194/hess-17-3921-2013(2013)https://doi.org/10.5194/hess-17-3921-2013(2013)https://doi.org/10.1175/BAMS-87-3-343https://doi.org/10.1175/BAMS-87-3-343https://doi.org/10.1175/JCLI-D-11-00015.1https://doi.org/10.5194/hess-16-2801-2012https://doi.org/10.5194/hess-16-2801-2012
A case study for the assessment
of the suitability of gridded reanalysis weather
data for hydrological simulation in Beas river basin
of North Western HimalayaAbstractIntroductionStudy areaData
and methodologyData usedMethodology usedComparison
of the IMD and ERA-InterimModel setup, calibration
and validationModel evaluation
Results and discussionConclusionReferences
