10
1
SNOWMELT RUNOFF ANALYSIS AND IMPACT
2
ASSESSMENT OF TEMPERATURE CHANGE IN THE
3
UPPER PUNATSANG CHU BASIN, BHUTAN
4
5
Running head: Snowmelt Runoff Simulation and impact of
6
temperature change on runoff in the Upper Punatsang Chu Basin,
7
Bhutan
8
9
Jigme Tenzin1 and Suwit Ongsomwang1
10
11
12
13
Abstract
14
In the area like Bhutan, accessing and monitoring of glacier and
15
snow melt is difficult due to its unfriendly and rugged terrain, thus
16
Snowmelt Runoff Model (SRM) with remote sensing data offers the
17
potential for furnishing information to improve water resources
18
management and decision making. The main objective of the study is
1
School of Remote Sensing, Institute of Science, Suranaree University of Technology, Nakhon
Ratchasima 30000, Thailand.

Corresponding Author, E-mail: suwit@sut.ac.th
11
19
to estimate runoff during snowmelt period and impact of hypothetical
20
temperature change on streamflow. Herewith the model input data
21
include basin characteristic, variables and parameters to execute the
22
model. The processes are routinely operated by calibration and
23
validation process and accuracy assesses with standard measurement.
24
The output include runoff volume and average runoff with
25
hydrograph for a melting season (April- August) of year 2005-2009.
26
Besides, the impact of temperature change on the streamflow are
27
investigated using three different hypothetical scenarios: (1). T + 1˚C,
28
(2) T + 2˚C and (3) T + 3˚C.
29
The simulated runoff volume were 5,713.29, 5,719.19, 5,750.92,
30
6,516.85 and 5,400.42 million cu. m, respectively for year 2005-
31
2009.The computed discharge is then correlated with measured
32
discharge and it was found that Nash-Sutcliffe efficiency ranging:
33
70.56 to 90.82%, absolute percent bias: ranging 0.3580 to 3.0431% and
34
volume different ranging: -3.0937 to 3.0431% for all hydrological
35
years.
36
Based on the hydrograph, it was observed that the SRM model
37
has simulated the daily flows reasonably well showing generally a good
38
agreement with the daily observed flows except few peaks. However,
39
it was found that SRM model has some limitation to model the period
40
where there is occurrence of extreme weather condition like cyclone,
12
41
storm and heavy rainfall. In case of impact of temperature change on
42
the streamflow, it was observed that with increase in every 1˚C of
43
average temperature, an average runoff increased by 14.36%.
44
In conclusion, the results achieved by the SRM model for the
45
basin considerable display good agreement and proved to be an
46
efficient tool to simulate snowmelt runoff and study impact of
47
temperature change on streamflow. The output can be used as a
48
guideline for water resources management, hydraulic system design
49
and mitigation plan to combat climate change effect.
50
51
Introduction
52
Snow is an important environment parameter, not only influencing the Earth’s
53
radiation balance but also playing a significant role in river discharge. Snowmelt and
54
snow covered area (SCA) has been the major source of runoff and groundwater
55
recharge in middle and higher latitudes areas (Jain, et al., 2010). The process of
56
converting snow and ice into water, known as snowmelt, needs input of energy (heat).
57
Hence, snowmelt is linked to the flow and storage of energy into and through the
58
snowpack (USACE, 1998). Therefore, estimation of snowmelt runoff is very important
59
for regulating the flow from the reservoirs, estimating flood flow for the design of
60
hydraulic structures and for other water resource development activities in the
61
Himalayan region.
62
Singh and Jain (2003) affirmed that the snowpack depletes either fully or
63
partially during the forthcoming summer season depending up the climatic conditions.
13
64
Attributing to the climatic condition, there is change in areal extent of snow covered
65
area (SCA) and snow free area (SFA) over the time, and the contribution from the rain
66
and snow to the stream flow varies with season. However, the precipitation like rain
67
dominates the lower altitude part of the basins (<2,000 m amsl.) and, rain and snow in
68
middle and higher altitude region of the basins (about > 2,000 m amsl.) with change in
69
altitude. With increase in altitude of the basin, the rain contribution to stream flow
70
reduces and the snowmelt contribution increases, therefore, runoff is dominated by the
71
snowmelt runoff above 3,000 m (amsl.) altitude (Singh and Jain, 2003). In higher
72
altitude and latitude regions where snowfall is predominated, runoff depends on the
73
heat supplied to snowmelt rather than just the timing of precipitation. Hence, to
74
understand the hydrological behavior and simulate the stream flow in such area, it is
75
very important to model the snowmelt runoff (Jain et. al., 2012).
76
In line with research on climate change by Liu and Rasul (2007) clearly
77
mentioned that according to Intercontinental Panel on Climate Change (IPCC) report
78
the climate change is a major concern in the Himalayas because of potential impacts on
79
the economy, ecology, and environment of the Himalayas and areas downstream.
80
Bhutan, being part of the eastern Himalaya region is adversely affected by climate
81
change causing the snow and glacier residing on mountains to melt faster in larger
82
extent as compared to other part of the world. This causes change in the hydrological
83
cycle which may further disturb river runoff, accelerate water-related hazards, and
84
affect agriculture, vegetation, forests, biodiversity and health. However, the
85
vulnerability of the Himalayas is unclear because of the lack of data and knowledge at
86
the regional level (Chettri et al., 2010).
87
Bhutan has witnessed flash floods and glacier outburst floods devastating acres
88
of agriculture lands and infrastructure properties, destruction to historical monuments
14
89
and causing threat to people living downstream in the Punatsang Chu basin in the years
90
1957, 1960 and 1994.
91
The basin shelters Punakha-Wangdue fertile valley along two major rivers:
92
Pho (Male) Chu and Mo (female) Chu fed by snow and glacier in the upper region of
93
the basin. After the confluence of these two rivers, the main river is called Punatsang
94
Chu which follows to the south entering Indian Territory and joining the Brahmaputra
95
River. The Punatsang Chu basin is the second largest basin amongst the five basins of
96
Bhutan. Taking advantage of its topographical features - rugged, steep terrain and fast
97
flowing rivers the basin is declared as home to the biggest ongoing hydropower
98
projects: Punatsang Chu Hydropower Project Phase I (1,200 MW) and Phase II (1,000
99
MW), thus, from the economic perspectives the hydropower plants have been the major
100
contributor to the economy of the country accounting for the increase in the overall
101
gross domestic product.
102
Therefore, the information on spatio-temporal variation of snow and the
103
snowmelt runoff can be applied practically to build hydraulics infrastructure for the
104
future hydropower project after the completion of this research. Furthermore, it can
105
provide sufficient information on water availability during different seasons advancing
106
in the field of water resource planning and management.
107
The specific objectives of the research are to estimate runoff from snowmelt
108
to the river during snow melting period and to assess the impact of temperature change
109
on stream flow by simulating the stream flow under different future temperature change
110
scenarios.
111
112
Concept of Snowmelt Runoff Model (SRM)
15
113
There are several temperature index based snowmelt models like the SSARR
114
Model, the HEC-1 and HEC-1F Models, the NWSRFS Model, the PRMS Model, the
115
SRM and the GAWSER Model (Singh and Jain, 2003 and Jain et al., 2012). Among
116
many models, snowmelt runoff model (SRM), which uses the snow cover information
117
as input, has been the most widely used for both simulation and forecasting (Martinec
118
and Rango, 1989; Rango and van Katwijk, 1990; Ferguson, 1999; Matinec et al., 2007;
119
DeWalle and Rango, 2008; Butt and Bilal, 2011). SRM or variations of it were applied
120
over 100 basins in 25 countries at latitudes 32–60-N, 33–54-S with basin sizes varying
121
from <1 to 120,000 km2 and documented in about 80 scientific journals (Seidel and
122
Martinec, 2004). This model has proved to be valuable for use in Himalayan regions
123
where meteorological and gauging field networks are sparse (Immerzeel, et al., 2010).
124
SRM is a conceptual, deterministic, degree day hydrologic model used to
125
simulate and/or forecast daily runoff resulting from snowmelt and rainfall in
126
mountainous regions. SRM requires daily temperature, precipitation and daily snow-
127
covered area values as input parameters.
128
129
130
Based on the input values, SRM computes the daily stream flow for a lag time
of 18 h (Matinec et al., 2007) as:
𝑄𝑛+1 = [𝐶𝑆𝑛 . 𝑛 (𝑇𝑛 + ∆𝑇𝑛 )𝑆𝑛 + 𝐶𝑅𝑛 . 𝑃𝑛 ]. 𝐴
10000
86400
(1 − 𝑘𝑛+1 ) + 𝑄𝑛 . 𝑘𝑛+1 ]
(1)
131
According to Eq. (1), the daily average discharge (Q) on day n+1 is computed
132
by summation of snowmelt and precipitation that contributes to runoff with the
133
discharge on the preceding day. Snowmelt from the preceding date is found by
134
multiplication of the degree day factor, , (cm C-1 d-1), zonal degree days (T+T)(C)
135
and the snow-covered area percentage (S). To determine the percentage that contributes
136
to runoff, the result of the above multiplication is further multiplied with CS, snowmelt
137
runoff coefficient and the total area of the zone, A (km2).
16
138
Measured/forecasted precipitation (P) is multiplied by CR, rainfall runoff
139
coefficient and the zonal area to calculate the precipitation contributing to runoff.
140
Discharge computed on preceding date is multiplied by recession coefficient (k) to
141
calculate the effect on today’s runoff. Eq. (1) is applied to each zone of the basin when
142
the model is applied in semi-distributed manner, the basin is subdivided into zones, and
143
then the discharges are summed up. SRM adjusts the input data if lag time other than
144
18 h is used (Martinec et al., 2007).
145
For accuracy assessment of the model performance, Nash-Sutcliffe efficiency
146
(NSE), percent bias (PBIAS) and volume difference (Dv) are used with following
147
equations.
∑𝑛
(𝑄 −𝑄𝑖′ ) 2
𝑁𝑆𝐸 = 1 − ∑𝑖𝑛=1(𝑄𝑖
148
𝑖 =1
′ 2
𝑖 −𝑄 )
(2)
149
where 𝑄𝑖 is measured daily discharge, 𝑄𝑖′ is computed daily discharge, 𝑄′ is average
150
measured discharge of the season under study, and n is the number of daily discharge
151
values.
152
153
154
𝑃𝐵𝐼𝐴𝑆 =
′ 2
∑𝑛
1=1(𝑄𝑖 −𝑄𝑖 )
2
∑𝑛
1=1(𝑄𝑖 )
× 100
(3)
where 𝑄𝑖 is the measured daily discharge, 𝑄𝑖′ is the computed daily discharge
𝐷𝑉 [%] =
(𝑉𝑅 − 𝑉𝑅′ )
𝑉𝑅
. 100
(4)
155
where 𝑉𝑅 is measured yearly or seasonal runoff volume, and 𝑉𝑅′ is computed yearly or
156
seasonal runoff volume.
157
SRM was successfully applied for runoff simulation by many researchers in
158
various basins, namely, Ganges, Toutunhe, Gongnisi, Beas-Thalot, Brahmaputra,
159
Parbati, Beas-Manali, Kabul River under Himalayan region. The accuracies obtained
160
from the mentioned studies are vary from 0.66 to 0.94 for NSE and -7.5-12 for Dv,
161
(Martinec et al., 2007).
17
162
163
Materials and Methods
164
Study Area
165
The study area is the upper region of Punatsang Chu basin covering three
166
districts: Gasa, Punakha and Wangduephodrang (partially), with a total area of 5,636.95
167
sq. km encompassing geographical area between 28° 14'N and 27° 27' N and 89° 19'E
168
and 90° 22' E is dissected by discharge gauging station located at latitude and longitude
169
of 27° 27' N and 89° 54' E from the overall basin (Figure 1). The topography of the
170
study area varies from altitude of 1,180 to 7,087 m (amsl).
171
Data and tools
172
Summary of data and tools used in this study as follows:
173
(1) Temperature and precipitation. Daily temperature and precipitation of
174
weather stations at Wangdue RNRRC (13640046) were collected from Ministry of
175
Economic Affairs (MoEA) and examined for extrapolating the average temperature to
176
each zonal hypsometric elevation. Wangdue RNRRC and Thanza stations were used
177
for calculating temperature lapse rate of the study area.
178
179
(2) Discharge. Daily discharge data at station Wangdue (13490045) was
collected from MoEA for assessing the SRM model accuracy.
180
(3) Digital elevation model (DEM). SRTM DEM with 90 m resolution, which
181
is more accurate than ASTER DEM with 30 m resolution (Forkuor and Matthuis, 2012)
182
was chosen and downloaded from the CGIAR Consortium for Spatial Information
183
website (http://www.cgiar-csi.org ) for this study. This data is used to generate basin
184
characteristics.
185
(4) MODIS data. MODIS snow products, MOD10A2 (Terra) and MYD10A2
186
(Aqua) with spatial and temporal resolution of 500 m and 8-day were downloaded from
18
187
NASA website (http://reverb.echo.nasa.gov/reverb) to calculate zonal snow cover area
188
based on the developed algorithm of Hall et al. (1995) and Hall et al. (2001). In this
189
study, 225 scenes of MOD10A2 and 225 scenes of MYD10A2 were downloaded (see
190
an example in Figure 2). This products demonstrated the capability to detect and
191
discriminate snow from clouds (Tekeli et al. 2005).
192
193
(5) WinSRM model. The WinSRM model is used for studying the snowmelt
runoff and impact of changing temperature on the snow cover area.
194
(6) ESRI ArcMap. This software is used for processing DEM for delineating
195
the watershed boundary and generating the hypsometric zones of the study area. In
196
addition, ArcMap Model Builder module is used to create semi-automate model to
197
extract zonal snow cover area from multiple MODIS data.
198
(7) MODIS Reprojection Tools (MRT). MRT enables users to read data files
199
in HDF-EOS format specify a geographic subset or specific science data sets as input
200
to processing, perform geographic transformation to a different coordinate
201
system/cartographic projection, and write the output to file formats other than HDF-
202
EOS.
203
204
(8) MODIS Snow Tool. It was developed by MENRIS, ICIMOD in order to
facilitate processing and analysis of daily and 8-day standard MODIS snow products.
205
(9) MATLAB. It is a high-level language and interactive environment for
206
numerical computation, visualization, and programming. This software is used to
207
interpolate the snow cover area of missing day between two consecutive 8-day snow
208
products by Piecewise Cubic Hermite Interpolation technique (Li and Williams, 2008).
209
Research methodology
210
The schematic workflow of research methodology which consists of three
211
components: (1) input data preparation (2) runoff simulation during snowmelt period
19
212
by SRM model and (3) impact of temperature change on stream flow is schematically
213
displayed in Figure 3. Major tasks of each component is separately summarized in
214
following sections.
215
1. Input data preparation
216
1.1 Hydro-meteorological data. Daily temperature, precipitation and
217
discharge which are recorded manually and supplied in raw format were converted to
218
time series format using MS Excel for executing the model. Daily average temperature
219
is derived using observed maximum and minimum temperature reading. Daily average
220
temperature and precipitation during snowmelt period for year 2005-2009 is displayed
221
in Figure 4.
222
1.2 Basin characteristics. If the elevation range of the basin exceeds 500 m
223
then the basin should be subdivided into elevation zones (Matinec et al., 2007).
224
Herewith, the basin boundary is subdivided into 3 different elevation zones using
225
Reclassify tool and the area of each zone is calculated (Table 1). Herein, a curve is
226
plotted between the cumulative zone area and elevation range, and the zonal mean
227
hypsometric elevation is calculated from the curve by balancing the areas above and
228
below the mean elevation (Figure 5). The individual zone area and its mean hypsometric
229
elevation are the basic basin characteristics used for setting up the model with zone
230
wise approach
231
1.3 Snow cover area (SCA) extraction. The MODIS products obtained for
232
the study area in a HDF format are firstly re-projected to UTM, Zone 45 N projection
233
with reference datum WGS1984 and converted from HDF to *.tiff format using MRT
234
tool software. Then both MOD10A2 and MYD10A2 snow product are combine and
235
apply cloud removal and spatial filtering using MODIS Snow Tool. The output obtained
236
after applying cloud removal algorithm is then used as input to ArcMap Model Builder
20
237
module (Figure 6) to extract snow cover area as shown in Figure 7. The derived SCA
238
for year 2005-2009 is displayed as zonal snow depletion curve in Figure 8 and Table 2
239
summarized the ratio area of SCA of melting season which is the significant variable to
240
execute the model.
241
1.4 Initial input parameters. Initial input parameters besides basin
242
characteristic and variables which are required to execute the SRM model are
243
synthesized from the literature reviews and calculated based on hydro-meteorological
244
data of the study area (Table 3).
245
2. Runoff simulation during snowmelt period by SRM model
246
2.1 Model Calibration. For model calibration, initial parameters sets
247
available in the SRM literature include α, TCRIT , CS, CR, and RCA are tested with
248
multiple variables and parameters configuration by trial and error to understand the
249
relationship between inputs and their simulated hydrographs. In this study zone-wise
250
approach was used with parameters adjustment at a daily or period time step. The
251
permissible range of value for parameters adjustment during the calibration mode
252
should be strictly monitored. The model is iteratively calibrated by accessing NSE value
253
until it achieves equal or more than 65% as suggested by Kult et al. (2014) for obtaining
254
the optimum range of local parameters.
255
2.2 Model validation. The derived optimum range of parameters value of
256
calibration year is further used to validate model by estimating runoff during snowmelt
257
period for year 2007, 2008 and 2009 by accessing the accuracy using NSE. During the
258
validation process, there is a constant need of changing the parameters value due to
259
change in the average temperature and snow covered areas of validation years.
260
2.3 Data output. Main derived output products from data processing under
261
SRM model are estimated runoff from snowmelt during snow melting period from
21
262
calibration and validate periods with its optimum range of parameters and accuracy
263
assessment.
264
3. Impact of temperature change on stream flow
265
Other than simulating the snowmelt contribution to river discharge from all
266
hydrological years, the impact of temperature change on the streamflow are investigated
267
using three different hypothetical scenarios: (1) average temperature + 1˚C (2) average
268
temperature + 2 ˚C and (3) average temperature +3 ˚C for calibration and validation
269
periods. This impact was investigated by maintaining all derived parameters and
270
variables constants, except the average temperature. The output achieved from the
271
investigation are effect of hypothetical scenarios on the average streamflow.
272
273
Results and Discussion
274
1. SRM runoff simulation under calibration process
275
The SRM model with all its required input data is calibrated iteratively
276
varying the value of parameters on trial and error method and the NSE value obtained
277
were 85.34 and 79.53%, and absolute PBIAS value were 0.6342 and 0.358% for the
278
hydrological year 2005 and 2006, respectively. The results of calibration year include
279
the simulation hydrograph and statistics data of runoff with accuracy assessment are
280
displayed and summarized in Figure 9 and Table 4. The optimum range of parameter
281
value derived from calibration periods is summarized in Table 5.
282
As results, it revealed that the simulated runoff for year 2005 and 2006 from
283
SRM model shows very good performance rating of NSE. The simulated average runoff
284
volume for year 2005 slightly overestimates with DV of -0.6342% but the simulated
285
runoff volume for year 2006 slightly underestimates with DV of 0.358%.
286
2. SRM runoff simulation under validation process
22
287
Using the derived basin characteristics, local recession coefficient value and
288
initial parameter range from calibration process, the model was set up for validation
289
period (2007, 2008 and 2009) and their NSE values vary between 70.56 and 90.82%.
290
The validation hydrograph are shown in Figure 10 and statistical results in Table 6. The
291
optimum range of local parameters during validation period is displayed in Table 7.
292
The results revealed that the validated runoff for year 2007, 2008, and 2009
293
show NSE higher than the defined range of performance rating. However, the validated
294
runoff for year 2009 comparatively show less NSE value of 70.56% than other
295
hydrological years. The low NSE value for hydrological year 2009 was mainly
296
triggered by Cyclone Aila which hit the Bay of Bengal on 25-26 May, 2009, had a
297
disastrous effect causing flash floods and river flooding events over Bhutan (Figure 11).
298
Thus, the river water levels in Wangdue and Punakha districts exceeded the water level
299
recorded in 1994 Glacier Lake Outburst Flood (Tenzing, 2009). Tahir et al. (2011)
300
stated that it is difficult to model the period where there is occurrence of extreme
301
weather condition like cyclone, storm and heavy rainfall. Herewith the simulated
302
average runoff volume for year 2009 is slightly underestimated with DV of 3.0431%.
303
3. Efficiency of SRM for runoff simulation and recommended local parameter
304
The SRM model has been applied for simulating the daily flows for the snow
305
melting season of Upper Punatshang Chu Basin for five years. The flow data for the
306
year 2005 and 2006 have been considered for calibrating the model whereas the year
307
2007, 2008 and 2009 have been considered for validating the model and the obtained
308
results of the efficiency of the model are summarized in Table 8.
309
The efficiency of the model has been computed based on the daily simulated
310
and observed flow values for five years. The values of the model efficiency, NSE are
311
85.34, 79.53, 90.82, 86.65 and 70.56% and absolute PBIAS are 0.6342, 0.358, 3.0937,
23
312
1.16 and 3.0431%, respectively for years 2005, 2006, 2007, 2008 and 2009. Similarly,
313
DV are -0.6342, 0.358, -3.0937, 1.16 and 3.0431%. It is observed that hydrologic year
314
2007 provides the maximum NSE value of 90.82% and minimum value of 70.56% in
315
hydrologic year 2009 caused mainly due to extreme event ‘Cyclone Aila’. The overall
316
average NSE value is 82.58% for the whole study period. It is also observed maximum
317
DV value of 3.0431% for hydrologic year 2009 underestimated the average runoff
318
compared to average measured runoff. On the contrary, the hydrologic year 2007 with
319
minimum DV value of -3.0937% overestimated the average computed runoff compared
320
to average measured runoff. The overall average difference in observed and simulated
321
runoff, DV is 0.1664% for the entire study period. The daily simulated and observed
322
flow hydrograph comparison for the study period as shown in Figures 7 and 8 show
323
that the model has simulated the daily flow reasonably well showing a good agreement
324
with the daily observed flow except few peaks.
325
At the global level it was found that the least NSE of 70.56 % for year 2009
326
obtained in this study proved to be more accurate than 42 SRM case studies of 112 that
327
have been applied over 112 river basins, located in 29 different countries which was
328
compiled by Martinec et al., 2007.
329
Likewise, at the regional level, many researchers have applied SRM in
330
Himalayan region, the efficiency rating of year 2009 proved to be more accurate than
331
7 out of 13 SRM applications and the average NSE value of 82.58% for year 2005-2009
332
is higher than 12 out of 13 applications in Himalaya region as compiled by Martinec et
333
al. (2007).
334
Since there is no hydrological study carried out using the SRM model in
335
Bhutan, therefore, the study is compared with three test sites nearby Bhutan. Firstly, the
336
results comparison was made against the research work of Silwal (2014) carried out in
24
337
Dudhkoshi River, Nepal, whose average NSE and DV are 84 and 4.5727 % respectively,
338
with difference observed in value of NSE is 1.42% and DV is 4.4063%.
339
The results of the simulation with average NSE of 82.58% proved better than
340
the research done by (1) Aggarwal et al. (2014) applied SRM model in Bhagirathi river
341
basin in upper Ganga catchment and the average NSE achieved is 80%, (2) Arya,
342
Gautam, and Murukar (2014) in Dhualigang River, India whose average NSE value for
343
calibration and validation periods resulted 75 and 73% respectively and (3) Zhang et
344
al.(2014) applied SRM model in Qinghai Lake basin and its average NSE value of 73%.
345
Thus, the simulation results achieved for this research work is reasonably
346
good when compared to the above stated results. Herewith, the recommended range of
347
optimum local parameter of SRM for Bhutan are as follows:
348
 degree day factor varies between 0.40 and 2.4 cm ˚C-1d-1,
349
 runoff coefficient value for snow varies from 0.17 to 0.90,
350
 runoff coefficient value for rain varies from 0.15 to 0.90, and
351
 temperature lapse rate is 0.57 ˚C/100 m.
352
4. Snowmelt simulation and its contribution to river discharge
353
Based on the above simulation with the defined range of parameters, the
354
amount of contribution of snowmelt depth to river discharge is summarized in Table 9.
355
It is observed that the total snowmelt depth of year 2006 is relatively low when
356
compared with remaining four hydrological years. Ramamoorthi (1987) and Prasad
357
(2005) mentioned in their research work about the regression relationship between the
358
SCA and runoff, which are extensively applied to places where detail snow studies are
359
not carried out. This agreement can be confirmed by the relationship between snowmelt
360
depth and average snow cover area between 2005 and 2009 by simple linear equation
361
with coefficient of determination (R2) of 81.19% as shown in Figure 12.
25
362
5. Sensitivity analysis
363
Sensitivity analysis of the SRM simulation for the Upper Punatshang Chu
364
basin, Bhutan is carried out to check the sensitivity of its parameters. For this purpose,
365
all the SRM parameters were varied by ± 10% of its calibrated value used in the SRM
366
runoff simulation of year 2007 (NSE = 90.82% and DV = -3.0937). It is clearly evident
367
that γ is most sensitive parameter followed by α, CS, and CR. On other hand, parameters
368
include TCRIT, RCA and L are least sensitive parameters (Table 10). The pattern of NSE
369
difference of α and CS are identical as these parameters are strongly related to snowmelt.
370
In addition, this finding was similar to the previous work of Haq (2008), who
371
applied SRM model for flood forecast and water resource management of River Swat
372
in Kalam Basin, Pakistan. Similar pattern of sensitivity test was observed in the work
373
of Bilal (2010) who applied SRM model in the Upper Indus Basin for water resource
374
management of Astore River, Northern Pakistan.
375
6. Impact of temperature change on streamflow
376
The result of the impact of temperature change on streamflow under three
377
simulated scenarios based on hydrological year 2005-2009 is displayed as hydrograph
378
in Figure 13 and summarized the percent increase of discharge volume in Table 11.
379
As results representing in Table 11, there is evident increase in snowmelt
380
runoff of approximately 13.96, 28.37 and 43.07% under a hypothetical scenarios by
381
increasing temperature by 1, 2 and 3C respectively. Hereby, it is observed an increase
382
of average temperature by 1C, the streamflow is expected to rise approximately by
383
11.84-16.67% from the simulated runoff.
384
In 1990, Rango and Katwijk applied SRM model to study climate change
385
effect in Western North America Mountain basins by increasing the mean temperature
386
by 1, 3 and 5˚C. Evidently there was an increase runoff during snowmelt season in Rio
26
387
Grande basin by 2.7, 8.3 and 14.3%, respectively. Similarly, there was increase in
388
snowmelt season runoff in Illecillewaet basin by 4.5, 11.1 and 16.3 %, respectively.
389
Likewise, Tahir et al. (2011) studied about the temperature change impact on
390
snow runoff in Hunza river basin, northern Pakistan and concluded with a finding that
391
there is increase of 33% of summer discharge resulted from increase of 1˚C and 64%
392
from 2˚C.
393
Silwal (2014) studied about the climate change in Dudhkoshi River basin,
394
Nepal and concluded that a rise in 1˚C in the mean temperature resulted in a 0.37%
395
increase in annual runoff volume. Regmi (2011) studied on the impact of climate
396
change by varying temperature from mean measured temperature and observed there is
397
rise in runoff approximately at rate of 2% in winter, 5% in summer and 4% annually
398
under the projected temperature rise of 1˚C.
399
Unlike, Singh and Kumar (1997) carried out an analytical studies using UBC
400
watershed model representing temperature increase of 1-3˚C in the western Himalayan
401
region suggest an increase in glacial melt runoff by 16-50%. Archer (2003) who applied
402
a linear regression analysis for climate variable and streamflow indicated that a 1˚C rise
403
in the mean summer temperature resulted in a 16% increase runoff into the Hunza and
404
Shyok River due to accelerated glacier melt.
405
For Upper Punatshang Chu basin, an increase in temperature by 1˚C resulted
406
in 14.36% increase of snowmelt runoff approximately. Thus, the results of impact of
407
temperature change on snowmelt associated with the basin contradicted with the above
408
studies. The discrepancy between the results obtained by different studies may be
409
possible due to the methods, hypothesis and limitations. Moreover, these results may
410
be specific to a particular region because the catchment response to the climate warming
411
may not be the same in other catchments as explained by Tahir et al. (2011).
27
412
413
CONCLUSION
414
SRM model is one kind of model which takes SCA as input instead of snow
415
depth and applicable to mountainous area with scarce hydro-meteorological data to
416
simulate and forecast runoff and study the effect of climate change on runoff. With
417
these abilities of the model, it is commonly applied in Himalayan regions for application
418
such as, flood mitigation, climate change effect and water management program.
419
The SRM model applied in Upper Puntshang Chu basin resulted a good
420
agreement between the measured and simulated runoff with NSE ranging: 70.56-
421
90.82%, absolute PBIAS ranging: 0.358-3.0937% and DV ranging: –3.0937–3.0431%
422
for hydrological year 2005-2009. Hereby, the optimum range of local parameters for
423
melt season: α = 0.4-2.4 cm˚C-1d-1, CR = 0.15-0.85, CS = 0.17-0.90, RCA = 1, γ =
424
0.57˚C/100m and TCRIT =1.2.
425
It was also observed that there is increase in streamflow by 14.36% with rise
426
of temperature by 1˚C. With this information, the future hydropower project dam can
427
be built with a storage capacity to hold all the melt and thus, increase the power
428
generation. And more over the flood mitigation program should consider the rise of
429
14.36% when preparing to meet future flood. Finally, the information on the impact of
430
temperature change on streamflow can be used in the water resources planning and
431
management purposes.
432
In conclusion, the model proved to be a good tool to simulate snowmelt
433
runoff and to study about the impact of temperature change on stream flow with
434
requirement of very less and very much available input data.
435
436
References
28
437
Aggarwal, S.P., Thakur, P.K., Nikam, B.R., and Garg, V. (2014). Integrated approach
438
for snowmelt runoff estimation using temperature index model, remote sensing
439
and GIS. Current Science, 106(3):397-407.
440
441
Archer, D. R. (2003). Contrasting hydrological regimes in the upper Indus Basin. J.
Hydrol., 274:198–210.
442
Arya, D.S., Gautam, A.K., and Murukar, A.R. (2014). Snowmelt modelling of
443
Dhauligang River using snowmelt runoff model. 11th International Conference
444
on Hydroinformatics, HIC 2014, New York City, USA.
445
446
Butt, M. J., and Bilal, M. (2011). Application of snowmelt runoff model for water
resource management. Hydrol. Processes, 25:3735-3747.
447
Chettri, N, Sharma, E., Shakya, B., Thapa, R., Bajracharya, B., Uddin, K., Oli, K.P.,
448
and Choudhury, D. (2010). Biodiversity in the Eastern Himalayas; Status,
449
Trends and Vulnerability to Climate Change: Climate Change Impact and
450
Vulnerability in the Eastern Himalayas. ICIMOD Technical Report 2.
451
Kathmandu, Nepal.
452
Dai, A. (2008). Temperature and pressure dependence of the rain-snow phase transition
453
over
454
doi:10.1029/2008GL033295.
455
456
land
and
ocean.
Geophys.
Res.
Lett.,
35:L12802,
DeWalle, D.R., and Rango, A. (2008). Principal of Snow Hydrology. Cambridge
University Press. 428p.
457
Duran-Ballen, S.A., Shrestha, M., Wang, L., Yoshimura, K., and Koike, T. (2012).
458
Snow cover modelling at the Puna Tsang river basin in Bhutan with corrected
29
459
JRA-25 temperature. Journal of Japan Society of Civil Engineers. 68(4):I_235-
460
I_240.
461
Ferguson, R.I. (1999). Snowmelt runoff models. Prog. Phys. Geog., 23(2): 205-227.
462
Forkuor, G., and Maathuis, B. (2012). Comparison of SRTM and ASTER derived
463
Digital Elevation Models over two regions in Ghana – Implication for
464
Hydrological and Environmental Modeling, Studies on Environmental and
465
Applied Geomorphology, Dr. Tommaso Placentin(Ed.), ISBN: 978-935-51-
466
0361-5,
467
http://www.intechopen.com/books/studies-on-environmental-and
468
appliedgeomorphology/comparison-of-srtm-and-aster-derived-digital-
469
elevation-models-over-two-regions-in-ghana .
470
InTech,
[Online]
Available:
Hall, D.K., Riggs G.A., and Salomonson, V.V. (1995). Development of methods of
471
mapping
472
spectroradiometer data. Remote Sens. Environ. 54:127-140.
global
snow
cover
using
moderate
resolution
imaging
473
Hall, D.K., Riggs, G.A., and Salomonson, V.V. (2001). Algorithm Theoretical Basis
474
Document (ATBD) for the MODIS snow and sea ice – mapping algorithms.
475
[Online] Available: http://modis.gsfc.nasa.gov/data/atbd/atbd_mod10.pdf
476
477
Hock, R. (2003). Temperature index melt modelling in mountain areas. J. Hydrol.,
282:104-115.
478
Immerzeel, W.W, Droogers, P., de Jong, S.M, and Bierkens, M.F.P. (2010). Satellite
479
derived snow and runoff dynamics in the Upper Indus River basin. Grazer
480
Schriften der Geographie und Raumforschung. 45: 303-312.
30
481
Jain, S.K., Goswami, A., and Saraf, A.K. (2010). Snowmelt runoff modelling in a
482
Himalayan basin with the aid of satellite data. Int. J. Remote Sens., 31(24):6603-
483
6618.
484
Jain, S.K., Lohani, A.K., and Singh, R.D. (2012). Snowmelt runoff modeling in a basin
485
located in Bhutan Himalaya. India Water Week 2012- Water, Energy and Food
486
Security: Call for Solution: 1-13pp.
487
488
489
490
491
492
493
494
495
496
497
498
499
500
Kult, J., Choi, W., and Choi, J. (2014). Sensitivity of the snowmelt runoff model to
snow covered area and temperature inputs. Appl. Geogr., 55: 30-38.
Li, X., and Williams, M.W. (2008). Snowmelt runoff modelling in an arid mountain
watershed. Tarim Basin, China. Hydrol. Processes, doi10.1002/hyp
Liu, J., and Rasul, G. (2007). Climate Change, the Himalayan Mountians, and ICIMOD.
Sustainable Mountain Development. 53:11-14.
Martinec, J., and Rango, A. (1986). Parameter values for snowmelt runoff modelling.
J. Hydrol., 84: 197-219.
Martinec, J., and Rango, A. (1989). Merits of Statistical criteria for the performance of
hydrological models. Water Resources Bulletin. 25(2): 421-432.
Martinec, J., Rango, A., and Major, E. (1983). The Snowmelt-Runoff Model (SRM)
User’s Manual. NASA Reference Publication 1100.
Martinec, J., Rango, A., and Roberts, R. (2007). Snowmelt Runoff Model (SRM) User’s
Manual Version 1.11. 172p.
501
Pandy, P.K., Williams, C.A., Frey, K.E., and Brown, M.E. (2013).Application and
502
evaluation of a snowmelt runoff model in the Tamor River basin, Eastern
31
503
Himalaya using Markove Chain Monte Carlo (MCMC) data assimilation
504
approach. Hydrol Process. doi: 10.1002/hyp.10005.
505
506
Prasad, V.H., and Roy, P.S. (2005). Estimation of Snowmelt Runoff in Beas Basin,
India. Geocarto International. 20(2):41-47.
507
Ramamoorthi, A.S. (1987). Snow cover area (SCA) is the main factor in forecasting
508
snowmelt runoff from major river basins. Large scale effects of seasonal snow
509
cover in Proceedings of the Vancouver Symposium.166: 187-198.
510
511
Rango, A., and Katwijk, V.V. (1990). Development and testing of a Snowmelt Runoff
Forecasting Technique. Water Resources Bulletin. 26(1):135-144.
512
Regmi, D. (2011). Impact of climate change on water resources in view of contribution
513
of snowmelt in stream flow: A case study from Langtang Basin Nepal. Ph.D.
514
Dissertation, Tribhuvan University, Kirtipur, Kathmandu, Nepal.
515
516
Seidel, K., and Martinec, J. (2004). Remote sensing in snow hydrology: Runoff
Modelling, Effect of Climate Change. Springer. 150p.
517
Silwal, G. (2014). Modelling snow and icemelt runoff in the context of climate change:
518
A case study of Dudhkoshi river basin, Nepal. Master Thesis. Tribhuvan
519
University, Kirtipur, Kathmandu, Nepal.
520
Singh, P., and Jain, S.K. (2003). Modelling of stream flow and its components for large
521
Himalayan basin with predominant snowmelts yields. Hydrol. Sci. J.,
522
48(2):257-276.
32
523
Singh, P., and Kumar, N. (1997). Impact assessment of climate change on hydrological
524
response of a snow and glacier melt runoff dominated Himalayan river. J.
525
Hydrol., 193: 316-350.
526
Tahir, A.A., Chevallier, P., Arnaud, Y., Neppel, L., and Ahmand, B. (2011). Modeling
527
snowmelt runoff under climate scenarios in the Hunza River basin, Karakoram
528
Ranger, Morthern Pakistan. J. Hydrol., 409:104-117.
529
Tekeli, A.E., Akyurek, Z. Sorman, A.A., Sensoy, A., and Sorman, A. (2005). Using
530
MODIS snow cover maps in modeling snowmelt runoff process in the eastern
531
part of Turkey. Remote Sens. Environ., 97:216-230.
532
Tenzing Lamsang (2009, 27 May). Flood kills 4. Kuensel. [Online] Available:
533
http://www.bhutan-switzerland.org/pdf/Kuensel_27-05-09.pdf Accessed data
534
May5, 2015.
535
Tiwari, S., Kar, S.C., and Bhatla, R. (2015). Examination of snowmelt over Western
536
Himalayan
537
10.1007/s00704-15-1506-y.
using
remote
sensing
data.
Theor
Appl Climatol. doi
538
USACE (1956). Snow Hydrology: Summary report of snow investigations,
539
Washington D.C.: U.S. Department of Commerce Office of Technical Services
540
PB 151660.
541
USACE (1998). Engineering and Design Runoff from snowmelt. [Online] available:
542
http://www.usace.army.mil/publications/eng-manuals/em1110-2-1406/toc.htm
543
Accessed
date May 5, 2014.
33
544
USDA NRCS. (2004). National Engineering Handbook. [Online] available:
545
http://www.nrcs.usda.gov/wps/portal/nrcs/detailfull/national/home/?cid=stelpr
546
db1043063. Accessed date November 21, 2014.
547
Zhang , G., Xie, H., Yao, T., Li, H., and Duan, S. (2014). Quantitative water resources
548
assessment of Qinghai Lake basin using snowmelt runoff model (SRM). J.
549
Hydrol.. 519: 976-987.
550
551
552
553
Zhang, Y., Liu, S., and Ding, Y. (2006). Observed degree-day factors and their spatial
variation on glaciers in western China. Annals of Glacialogy. 43:301-306.
34
554
Table 1 Summary of the hypsometric zones of Upper Punatsang Chu Basin.
Mean Hypsometric
555
Zone Elevation Band (m) Area (sq.km) %Zone Area
Elevation (m)
A
1,180-2,500
839.96
14.921
2,019
B
2,501-4,000
1,624.54
28.579
3,278
C
4,001-7,087
3,199.92
56.500
4,678
35
556
Table 2 Ratio of SCA of melt season (April- August) for different years and their
557
change with comparison to base year 2005.
Ratio of SCA of year
558
Zone
2005
2006
2007
2008
2009
A
7.428
1.947
7.094
9.444
7.628
B
20.448
7.610
18.882
22.208
19.324
C
59.426
52.809
53.698
56.205
59.791
36
559
Table 3 Summary of initial input parameters.
Parameter
Value of range
Reference
Critical Temperature (TCRIT)
1.2˚C
1.2˚C from Dai (2008)
+1.5˚C to 0˚C from USACE (1956)
Runoff coefficient for rain (CR)
0.1-0.9
USDA NRCS (2004)
Runoff coefficient for snow (CS)
0.1-0.9
USDA NRCS (2004)
RCA (rainfall contributing area)
1
Degree day factor ()
0.2-2.4 cm ˚C d
Martinec et al. (2007)
-1 -1
Hock (2003); Zhang et al. (2006);
Tahir et al. (2011); Butt and Bilal
(2011); Zhang et al. (2014), Tiwari et
al.(2015)
Temperature lapse rate ()
0.57C/100 m
Calculate from weather stations
Time lag
Zone A =11 h,
Modified from Martinec and Rango
Zone B =12 h and
(1986)
Zone C =13 h
Recession coefficient (k)
x=0.899; y = 0.010 for
Derived
2005
(𝐾𝑛+1 = x𝑄𝑛 )
x=1.051; y = 0.033 for
discharge data as suggested by
2006
Martinec et al. (1983) and Martinec
x=1.048; y = 0.032 for
and Rango (1986)
2007
x=0.884; y = 0.008 for
2008
x=1.007; y = 0.047 for
2009
560
from
−𝑦
generic
equation
with
historical
37
561
Table 4 Statistics data of SRM simulated runoff and its accuracy for year 2005 and
562
2006.
Statistics
Measured Runoff Volume (10^6 m3)
Average Measured Runoff (m3/s)
Computed Runoff Volume (10^6 m3)
Average Computed Runoff (m3/s)
NSE (%)
|PBIAS| (%)
Volume Difference (%)
563
2005
2006
5,677.29
5,739.73
426.68
431.38
5,713.29
5,719.19
429.39
429.83
85.34
79.53
0.6342
0.358
-0.6342
0.358
38
564
Table 5 Range of optimum local parameters of calibration period.
Calibrated parameters
565
Year
α ( cm ˚C-1 d-1)
CS
CR
RCA
2005
0.4–1.2
0.20-0.90
0.15-0.85
1
2006
0.4–2.4
0.20-0.90
0.20-0.90
1
39
566
Table 6 Statistics data of SRM simulated runoff and its accuracy for year 2007, 2008
567
and 2009.
Statistics
Measured Runoff Volume (10^6 m3)
Average Measured Runoff (m3/s)
Computed Runoff Volume (10^6 m3)
Average Computed Runoff (m3/s)
NSE (%)
Absolute PBIAS (%)
Volume Difference (%)
568
2007
2008
2009
5,578.34
6,593.33
5,569.92
419.25
495.53
418.61
5,750.92
6,516.85
5,400.42
432.22
489.78
405.88
90.82
86.65
70.56
3.0937
1.16
3.0431
-3.0937
1.16
3.0431
40
569
Table 7 Range of optimum local parameters for validation period.
Validated parameters
570
Year
a ( cm ˚C-1 d-1)
CS
CR
RCA
2007
0.40 - 2.40
0.20 - 0.90
0.20 - 0.85
1
2008
0.40 - 1.20
0.17 - 0.85
0.20 - 0.85
1
2009
0.40 - 1.08
0.20 - 0.85
0.20 - 0.85
1
41
571
Table 8 Summary of the efficiency of the model of the study.
Year
Performance access
NSE (%)
Absolute PBIAS (%)
DV (%)
572
573
2005
2006
2007
2008
2009
85.34
79.53
90.82
86.65
70.56
0.6342
0.358
3.0937
1.16
3.0431
-0.6342
0.358
-3.0937
1.16
3.0431
42
574
Table 9 Contribution of snowmelt to river discharge.
Contribution of snowmelt in hypsometric elevation zone
Zone A
Year
575
576
Zone B
snow snow (new) snow
Zone C
snow (new) snow
snow (new)
Total
2005
102
0
193.81
0
165.25
7.86
468.87
2006
24.07
0
59.61
0
119.46
8.53
211.67
2007
67.99
0
120.86
0
125
3.83
317.68
2008
145.3
0
220.06
0
171.32
1.67
538.39
2009
110.8
0
168.24
0
173.9
0.24
453.20
43
577
Table 10 Sensitivity analysis result for the SRM simulation year 2007.
No
1
2
3
4
5
6
578
579
SRM Parameters
Simulated
NSE (%)
Sensitivity test NSE
NSE (%)
Difference
+10% of Lapse rate ()
90.82
71.50
19.32
-10% of Lapse rate ()
90.82
64.25
26.57
+10% of TCRIT
90.82
90.82
0.00
-10% of TCRIT
90.82
90.82
0.00
+ 10% of Degree day factor (α) 90.82
87.63
3.19
- 10% of Degree day factor (α)
90.82
90.45
0.37
+10% of Lag time (L)
90.82
90.73
0.09
-10% of Lag time (L)
90.82
90.90
-0.08
+10% of CS
90.82
87.53
3.29
-10% of CS
90.82
90.51
0.31
+10% of CR
90.82
90.12
0.70
-10% of CR
90.82
90.98
-0.16
RCA = 1
90.82
90.82
0.00
RCA = 0
90.82
90.79
0.03
44
580
Table 11 Comparison of runoff volume and its percent increase and average percent
581
increase of streamflow under the hypothetical scenarios of year 2005-2009.
Year
2005
Hydrological
statics
Simulated
Scenario 1
(+1°C)
Scenario 2
(+2°C)
Scenario 3
(+3°C)
Runoff volume
(×106 m3)
5,713.29
6,554.82
7,441.37
8,337.06
14.73
30.25
45.92
6,557.17
7,452.98
8,406.13
14.65
30.32
46.98
6,637.17
7,528.50
8,419.91
15.41
30.91
46.41
7,373.94
8,237.60
9,114.29
13.15
26.40
39.86
6,040.00
6,694.76
7,353.64
11.84
23.97
36.17
13.96
28.37
43.07
% Increase
2006
Runoff volume
(×106 m3)
5,719.19
% Increase
2007
Runoff volume
(×106 m3)
5,750.92
% Increase
2008
Runoff volume
(×106 m3)
6,516.85
% Increase
2009
Runoff volume
(×106 m3)
% Increase
Average
582
583
5,400.42
45
584
585
586
Figure 1 The study area and its location.
46
(a) MOD10A2
587
588
589
(b) MYD10A2
Figure2 Example of MOD10A2 and MYD10A2 snow products
47
PART 1 Input data preparation







Daily average temperature
Daily precipitation
Daily discharge
Basin characteristic
SCA data using NDSI algorithm
Recession coefficient
Time lag
PART 2 Runoff simulation during snowmelt period by SRM model
INPUT
Basin Characteristic

Hypsometric elevation zone

Basin zone area

Basin station elevation
Variables

Daily temperature

Daily precipitation

Zonal snow cover area
Parameters
 Runoff coefficient
 Recession coefficient
 Temperature lapse rate
 Critical temperature
 Time lag
 Degree day factor
 Rainfall contributing area
PART 3 Impact of temperature
change on streamflow
PROCESS
Model calibration
HYPOTHETICAL SCENARIO
No



NSE  0.65
Average temperature by + 1C,
Average temperature by + 2C,
Average temperature by + 3C.
Yes
Optimum range of local parameters
Model validation
NSE  0.65
PROCESS
No
Yes
Optimum range of local parameters
OUTPUT



Simulation analysis
Hydrograph
Accuracy assessment
590
591
592
Figure 3 Workflow diagram of research methodology
OUTPUT


Streamflow change analysis
Hydrograph
48
(a)
(b)
(c)
(d)
(e)
593
Figure 4 Daily average temperature and precipitation during snowmelt period for year
594
(a) 2005, (b) 2006, (c) 2007, (d) 2008 and (e) 2009.
595
49
596
597
598
Figure 5 Hypsometric curve of Upper Punatsang Chu Basin.
50
599
600
Figure 6 Schematic diagram of input, process and output of Model builder for zonal
601
SCA extraction.
602
51
603
604
605
606
Figure 7 Example of snow cover area map of June 25, 2010.
52
(a)
(b)
(c)
(d)
(e)
607
Figure 8 Zonal snow cover depletion curve for year (a) 2005, (b) 2006, (c) 2007, (d)
608
2008 and (e) 2009.
609
53
(a)
(b)
610
Figure 9 Comparison of measured and computed hydrograph for year: (a) 2005, and
611
(b) 2006
612
54
(a)
(b)
(c)
613
Figure 10 Comparison of measured and computed hydrograph for year: (a) 2007, (b)
614
2008, and (c) 2009.
55
615
616
Figure 11 Superimposed data of measured and calculated runoff and rainfall of May
617
2009.
618
56
619
620
621
Figure 12 Relationship between snowmelt depth and average snow covered area.
57
(a)
(b)
(c)
(d)
(e)
622
Figure 13 Comparison of simulated streamflow in each scenario based on simulation
623
data for year: (a) 2005, (b) 2006, (c) 2007, (d) 2008, and (e) 2009.