Analysis of Aqua AMSR-E derived Snow Water
Equivalent over Himalayan Snow Covered Regions
Vijay Kumar ,Y. S. Rao , G.Venkataraman
Centre of Studies in Resources Engineering
Indian Institute of Technology, Bombay
Mumbai 400076 India
[email protected]
Abstract— We have made an endeavor to investigate the snow
water equivalent (SWE) variations in Himalayan mountain
region which is the most difficult terrain to access during winter
seasons. The area also covers the large glaciers such as Siachen
and Gangotri. A time series multi scale of SWE L3 product
derived from Aqua Advanced Microwave Scanning Radiometer
(AMSR-E) data have been analysed for three consecutive winters
during 2002-2005 for Himalayan Snow cover region. A major
emphasis is on the study of SWE trend in the two glacier areas
viz. Siachen and Gangotri. It has been observed through AMSRE 5-day product that snow cover area (SCA) over whole
Himalayan region is very dynamic. AMSR-E derived 5-day SWE
is analysed at the two test sites viz. Patsio (Lat 320 45’ 17.89″ N
and Lon 770, 15′, 43.13″E) and Dhundi (Lat 320, 22′, 05″N and Lon
770, 15′E) concurrently with the available in-situ data .The result
indicates that in the peak winter days only we found reasonable
co-relation. At the Patsio, minimum and maximum SWE
observed are 26mm and 108mm respectively using in situ data.
Keywords- Snow water equivalent, Snow cover area , Snow
depth
R. N. Sarwade , Snehmani
Snow and Avalanche study Establishment
Him Parisar, Plot No. 1
Chandigarh 160036, INDIA
[email protected]
underestimate the SCA as compared to estimates from visibleinfrared snow mapping methods [4]. The perceived need by
water resource managers and land surface and climate
modelers is for high accuracy, local scale estimates of snow
volume on a daily basis. Unfortunately, the spatial resolution of
the SMMR and SSM/I instruments tends to restrict their
effective use to regional-scale studies. Furthermore, currently
available SSM/I data is acquired twice daily only at high
latitudes with coverage more restrictive at lower latitudes. The
Advanced Microwave Scanning Radiometer—Earth Observing
System (AMSR-E) aboard Aqua, which was launched in 2002,
has been giving data at high spatial resolution and at more
frequently to overcome some of these drawbacks. Our aim is to
detect snow cover area and analyzing the SWE in the
Himalayan region .The microwave brightness temperature
emitted from snow cover is related to the snow mass which
can be represented by the combined snow density and depth, or
the SWE (a hydrological quantity that is obtained from the
product of SD and density).
II.
STUDY AREA
I.
INTRODUCTION
Spaceborne passive microwave remote sensing can provide
useful information about snow cover characteristics for various
hydrological climatical and meteorological applications[1] .The
most important parameter related to snow is snow cover extent
and snow water equivalent (SWE) that are very important for
number of applications such as hydro power stations, resource
management, irrigation requirements and flood forecasting.
Many rivers in north India originate from the snow covered
areas and form a major source for drinking and irrigation
requirements of millions of people. It also helps in forecasting
snow avalanches and mitigating land slides. Snow cover can
affect dynamic Himalayan climate because snow cover over
land or mountain reduces the amount of sunlight absorbed by
the Earth and also restricts flow of heat from the surface
beneath the snow cover.
In this study entire Himalayan region falling between Lat
280N to 460N and Lon 66.970 E to 103.840E is covered. Area of
interest is selected keeping in mind two big glaciers like
Siachen and Gangotri. The longest river of India, the Ganges
originates from Gangotri glacier. In the western region of the
Himalaya we have observatories for observing snow
parameters. These observatories measure in Situ data of snow
grain size, temperature, snow density and depth. It is well
known that snow parameters are region dependent [5].
Unavailability of in-situ information will affect the study.
Snow cover study in the Himalaya region is vital in the context
of real time use in hydrological, climatological and agricultural
applications and for long term monitoring climate variability
and detection of change.
Earlier work has been done for retrieving snow depth (SD)
or SWE through the available spaceborne radiometers such as
the Scanning Multichannel Microwave Radiometer (SMMR)
and the Special Sensor Microwave Imager (SSM/I). Neither
instruments were designed explicitly for snow applications but
have been found to be effective for this application [2], [3]. For
snow detection, passive microwave instruments tend to
The study is carried out using AMSR-E derived SWE and
observatory snow parameters like SD, snow temperature, grain
size and wetness. Microwaves have the capability to penetrate
dry snow pack and depending on the brightness temperature
and frequency used it is possible to determine the depth of the
snow.
III.
DATA SETS AND METHODOLOGY
0-7803-9510-7/06/$20.00 © 2006 IEEE
Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY BOMBAY. Downloaded on June 25, 2009 at 00:46 from IEEE Xplore. Restrictions apply.
706
AQUA AMSR-E has operational frequencies 6.9GHz,
10.7GHz, 18.7 GHz, 23.8 GHz and 36.5 GHz. All Channels
give brightness temperature in horizontal as well as vertical
polarization. It is found that 18.7 GHz and 36.5 GHz channels
are the most suitable for snow depth and SWE determination.
Snow studies group at National Snow and Ice Data Centre
(NSIDC) has used difference of these two frequencies for
developing SWE algorithm. A positive difference is regarded
as a scatterer and might possibly have been emitted by the
snow [6]. Generally, the greater this difference, and hence, the
scattering signal, the greater the snow volume assumed to be
present. Unfortunately, a major problem with this assumption
is that in nature, changes to snow pack physical properties can
also cause changes to the microwave scattering response of the
pack; a change in the observed scattering equally might be
caused by an increased snow volume or a change in the
physical structure of the snow resulting from snow pack
metamorphism. For homogeneous snow packs, the scattered
signal can be converted to SWE or snow depth using an
empirical algorithm or physically based static algorithm.
SWE (mm) = 4.8 × (
T 18v − T 36v
)
1.0 − 0.2 × ff
and Sept. 2004. The location of the points studied for the work
was selected at the centre of the glacier area and there is no
ground truth available at those places because those areas are
inaccessible.
Monthly (Aug. 2004) average of SWE
(a)
Monthly (Feb. 2004) average of SWE
(1)
Where ff is representation of the fractional forest cover for
that pixel obtained from the data set.
Monthly and 5-day average SWE products for 4 years were
downloaded and converted into binary format using software
provided by NSIDC. We developed our own program for
converting the northern hemisphere SWE product to ASCII
format. The ASCII data contains the information of latitude,
longitude and SWE for each pixel in our region of interest.
With the help of ERDAS Imagine, ASCII file is converted into
raster format with a grid of 0.25 x 0.25 degree.
IV.
(b)
5 day (Sept. 3-8, 2004) average of SWE
(c)
RESULTS AND DISCUSSION
SWE has been studied from 2002 to 2005 using 5-day as
well as monthly products. It has been observed through 5-day
products that SCA over the whole Himalayan region is very
dynamic on both time scales as shown in Fig.1 (a) (b) (c) and
(d). It is revealed from 5-day products of all years that
maximum SCA was observed in the month of February. From
March last week onwards, it starts decreasing and becomes
minimum in August. SCA in the peak of winter that is in
months of December and January remains almost same. From
middle of August, SCA starts increasing and becomes
maximum in February. We have made a detailed study on
annual SWE trend at specific locations on the two major
glaciers (Gangotri and Siachen). Although SCA was maximum
in February, but SWE values were observed maximum (64mm,
66mm, 55mm) in the fourth week of May 2003, second week
of March 2004 and third week of April 2005 at Gangotri area
respectively. Minimum values (3mm, 3mm, 14mm) SWE at
Gangotri were observed in the second week of Sept. 2003 and
second week of Aug. 2004 respectively as shown in Fig.2 (a)
(b) and (c). For Siachen glacier area, maximum SWE values
(76mm, 79mm) were observed in the third week of July 2003,
second week of July 2004 respectively. Minimum SWE value
at the Siachen point is observed in fourth week of Sept. 2003
5 day (Sept. 23-28, 2004) average of SWE
(d)
Fig. 1(a) and (b) show minimum and maximum snow
coverage in the year 2004 respectively while 1(c) and (d)
show cover in the first week and last week of sep.2004
respectively on the five day average basis.
AMSR-E derived 5-day SWE is analysed at the two test
sites Patsio (Lat 320 45′ 17.89″ N and Lon 770 15′ 43.13″E) and
Dhundi (Lat 320 22′ 05″N and Lon 770 15′E) in three
consecutive winters 2002/03, 2003/04 and 2004/05. The two
test sites are located at different Himalayan regions and are
0-7803-9510-7/06/$20.00 © 2006 IEEE
Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY BOMBAY. Downloaded on June 25, 2009 at 00:46 from IEEE Xplore. Restrictions apply.
707
having different climatic conditions. Patseo is located at the
altitude of 3700 m and surround by barren mountains, whereas
Dhundi was located at the altitude of 3000 m and surround by
Gangotri 5day Avg SWE for the year 2003
70
Jan
60
Feb
Mar
S WE (m m )
50
Apr
May
40
Jun
Jul
30
Aug
Sep
20
Oct
Nov
10
Dec
0
5
10
15
20
25
30
DAys of a month
(a)
Gangotri 5dayAv g SWE for the ye ar 2004
Jan
SWE ( in m m )
Feb
70
March
60
April
50
May
40
Jun
30
July
20
Aug
10
Sep
Oct
0
5
10
15
20
25
30
Nov
Dece
Days of a m onth
(b)
Gangotri 5dayAvg SWE for the year 2005
60
Jan
S W E ( in m m )
50
Feb
40
Mar
30
Apr
20
May
Jun
10
0
5
10
15
20
25
30
Days of a m onth
(c)
Fig2.(a) (b) (c) and (d) SWE versus days of different months of the years
2002, 2003, 2004 and 2005 respectively at Gangotri glacier.
high peak mountains with and without vegetation cover. At
the two stations, observatory data are available in terms of
standing snow, fresh snow, temperature, wind speed, etc. both
at morning and evening.
Seasonal trend of SWE at the two places is same in all the
three winters. At Patsio maximum value (108 mm) of SWE is
observed in third week of Nov. 2004. From Nov. onwards it
follows increasing and decreasing trend and becomes minimum
(26 mm) in third week of April 2005. But at the same place
scenario is different in 2002/03 winter. The SWE in the first
week of Nov. 2002 is observed about 60 mm and starts
decreasing to 40 mm on 10th of Nov 2002 and then increased to
maximum value 64 mm on 20th Nov. 2002. From this
onwards, it decreased to 44 mm on 20th December 2002 and
finally becomes minimum (16mm) on Feb. 25, 2003. Similar
trend of SWE is observed for Dhundi test site except 2003/04
winter season. We observed that the trend of SWE in two
winters 2002/03 and 2004/05 is same at two test sites but trend
of 2003/04 is entirely different. At patsio, the maximum height
of standing snow varied from 233 cm in 2002/03 to 120 cm in
2004/05 winter. But at Dhundi, maximum values were quite
large and varied from 252 cm in 2003/04 winter to 319 cm in
2004/05 winter.
We analyzed 5-day average SWE and observatory record of
standing snow of the day after averaging five day record. We
found correlation in Jan. Feb. and March observations of
2002/03 and 2004/05 only at two test sites [Fig.3 (a) (c)]. But
for winter 2003/04 we found correlation in Feb. and March for
both places [Fig.4 (b)]. Density of snow is the governing factor
in the determination of SWE and it is varying throughout the
winter. Algorithm observes SWE as a bulk property of the
snow pack and density is considered constant. It seems due to
dynamic nature of snow density we are not finding good
correlation between SWE and standing snow.
In the north-west Indian Himalayan region, there were
reports of vast devastations to property and lives due to heavy
snowfall in Feb 2005 in a short span of time. Data were
analysed of the same period day by day and weekly average.
We could not find any abrupt change in snow cover and SWE
in vast affected region. In order to verify the estimated SWE
with ground truth measurements, it is necessary to conduct
field work at different locations of the above test sites
synchronous with AMSR-E passes.
V.
CONCLUSION
The SWE over Himalayan regions during 2003-2005 have
been studied along with ground-truth data available at some
observatories. SWE results indicate that the snow cover in the
Himalayan region is very dynamic on both weekly and
monthly time scales. The SWE in the middle of the glaciers
remained constant in Aug. third week of 2002 and 2003. But in
Dec. (peak of winter) we found difference of SWE as large as
20 mm between two years. These observations should be seen
in the context of climatical change on the regional scale. Since
passive microwave radiometers have their own limitations in
measuring the geophysical parameters we could not find good
correlation between satellite observed SWE and in-situ SD
except peak winter months. It might be due to variability of the
snow density with time.
VI.
ACKNOWLEDGMENT
The authors would like to thank the National snow and Ice
Data Centre (NSIDC) for providing AMSR-E snow data
products used in this paper.
0-7803-9510-7/06/$20.00 © 2006 IEEE
Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY BOMBAY. Downloaded on June 25, 2009 at 00:46 from IEEE Xplore. Restrictions apply.
708
SWE(mm)
80
250
200
150
100
50
0
60
40
20
5
10
15
20
25
30
35
40
45
50
55
60
65
70
75
80
85
90
95
100
105
110
115
120
125
130
135
140
145
0
Standing
Snow (cm)
SWE and standing snow verses days, Dece02-April03 at
Dhundi
SW E
Da ys of the months
Std_Snow
(a)
250
200
150
100
50
0
60
40
20
0
5
10
15
20
25
30
35
40
45
50
55
60
65
70
75
80
85
90
95
100
105
110
115
120
125
130
135
140
145
150
155
160
165
170
180
SWE (mm)
80
Standing
Snow (cm)
SWE and standing snow verses days of the months, Nov03April04 at Dhundi
SW E
Days of the months
Std_Snow
(b)
SWE and standing snow verses Days of months, Nov04-April05
at Dhundi
400
80
300
60
200
40
100
20
0
5
10
15
20
25
30
35
40
45
50
55
60
65
70
75
80
85
90
95
100
105
110
115
120
125
130
135
140
145
150
155
160
165
170
175
180
0
Standing
Snow (cm)
SWE (mm)
100
Da ys of the months
SW E
Std_Snow
(c)
Fig. 3. (a) (b) and (c) shows SWE and standing snow versus days of the winter months of 2002-03, 2003-04 and 2004-05 respectively at the Dundi test site.
[4]
REFERENCES
[1]
[2]
[3]
J. Pulliainean, M. Hallikainen, S. Anttila and S. Metsamaki,
“Estimation of snow water equivalent and snow depth in boreal forests
by
assimilating
AMSR-E
observations
with
in
situ
observations”.Geoscience and Remote Sensing Synposium, vol. 4, 25-29
July , pp. 2641 -2642, 2005.
R.L Armstrong, A.T.C. Chang, A. Rango and E.G. Josberger, “Snow
depth and grain size relationships with relevance for passive microwave
studies,” . Annals of Glaciology 17, pp. 171-17 , 1993
A. T. C. Chang, J. L. Foster, and D. K. Hall, “Nimbus-7 derived global
Snow cover parameters,” Ann. Glaciol., vol. 9, pp. 39–44, 1987.
[5]
[6]
C. Chen, B. Nijssen, J. Guo, L. Tsang, A. W. Wood, J. Hwang, and D. P.
Lettenmaier, “Passive microwave remote sensing of snow constrained
by hydrological simulations,” IEEE Trans. Geosci. Remote Sensing, vol.
39, pp. 1744–1756, Aug. 2001
R. L. Armstrong and M. J. Brodzik, “Recent Northern Hemisphere snow
Extent: A comparison of data derived from visible and microwave
satellite Sensors,” Geophys. Res. Lett., vol. 28, pp. 3673–3676, 2001.
R. E. Kelly, Alfred, T. Chang, L. Tsang, and J. L. Foster, “A
prototype AMSR-E global snow area and snow depth algorithm,” IEEE
Transas. Geosci.and Remote Sensing, vol. 41, pp. 230-231. Feb. 2003.
0-7803-9510-7/06/$20.00 © 2006 IEEE
Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY BOMBAY. Downloaded on June 25, 2009 at 00:46 from IEEE Xplore. Restrictions apply.
709