1
Université de Toulouse, GEODE UMR 5602 CNRS, Toulouse, France
2
Université de Franche-Comté, ThéMA UMR 6049 CNRS, Besançon, France
3
Université de Franche-Comté, FEMTO-ST UMR 6174 CNRS, Besançon, France
4
Université de Paris-Sud 11, IDES UMR 8148 CNRS , Orsay, France
Dominique LAFFLY 1 , Eric BERNARD 2 , Jean-Michel FRIEDT 3 Gilles MARTIN 3 , Christelle MARLIN 4 , Madeleine GRISELIN 2
Snow cover monitoring using combined FORMOSAT satellite imaging, in situ sensing oblique view ground-
based pictures and snow drills
(East Loven glacier, Spitsbergen, Svalbard)
Summary
Introduction
1. Field localization
2. Flood event of september 2008 3. In situ image data collection 4. In situ image geometric correction
5. Snow and ice melt quantification
Conclusion
Nevertheless, fast events appear as significant in the ice and snow budget while being ignored by satellite based studies since the slower sampling rate is unable to observe
such fast events. In this project, satellite imagery is complemented with a series of ground based autonomous automated high resolution digital cameras.
An example of the complementarity of database was presented about a flood period in september 2007.
In the frame of the Hydro-Sensor-FlOWS program funded by the French National Research Agency and the French Polar Institute (IPY 16), the East Loven glacier, located in Spitsbergen (78°N, 12°E, Norway) has been closely monitored during the last 4 years
(2007 to 2010) in order to analyze at the basin scale (10 km 2 ) and at various time scales (hour, day, month, year) the ice and snow mass-balance and their direct and
indirect hydrological consequences.
For obvious cost reasons as well as due to poor weather/cloudy conditions, daily satellite imaging coverage is not always accessible: eleven images are acquired each year but only two monthly FORMOSAT data sets have been selected as representative
of general snow cover.
La GOULE
Le DIABLE
Corbel station
Limit of the LIA moraine
Calcareous ridge
80°
78°
76°
74°
10° 15°
Spitsberg
Brogger
Field localization
East Loven
hydrological summer 2008
Runoff in w.eq
Total precipitation 129 mm
680 mm 174 mm Goule
Diable 129 mm
T°C in Ny Alesund
P mm in Ny Alesund rain
snow
Runoff in w.eq / day
Flood period at the end of 2008 Summer
1 st week of September: cold, no water in rivers it seems that it is the beginning of winter
Flood period at the end of 2008 Summer
2 d week, sudden increase of T and 10% of the yearly precipitation
Flood period at the end of 2008 Summer
Due to the rain, snow and ice melt on the glacier (flood water in the Goule river) while the moraine absorbs the rain like a sponge (Diable river stays constant)
The rain stops one day: immediate response of the Goule river. Then a new rain event gives again flood water in the Goule. The moraine is than saturated and the Diable begins to grow up.
Goule
Diable
The last week, T decreases and the rain stops. It snows. The fall in
level is slow, sustained by the sub- glacial runoff.
Flood period at the end of 2008 Summer
160 mm of melting of snow and ice 76% of summer precipitation (100 mm) 40 % of the summer runoff (260 mm)
Goule Diable
BUDGET
ELA 09/05/08
250 m ELA
09/25/08 350 m
During that flood period, the equilibrium line (ELA) of the glacier growed up from 250m to 350m.
Ablation vs Accumulation
Snow high and
quality field
measures
Only two Formosat images are available around this flood event (August 15 th and September 30 th ) ... showing the glacier totally cover of snow
15 th of august 30 th of september
Products
B&W : 2 m
Color: 2 m (pansharpened) MS (R, V, B, PIR) : 8 m
Spectral bands
P : 0,45 – 0,90 µm B1 : 0,45 – 0,52 µm (blue) B2 : 0,52 – 0,60 µm (green) B3 : 0,63 – 0,69 µm (red) B4 : 0,76 – 0,90 µm (NIR) Coverage 24 km x 24 km
Repetiivity daily
Angle lateral &front-back: +/- 45°
Programmation Yes Image dynamics 8 bits/pixel
Image size (1A level)
MS : 35 MB Pan : 137 MB
FORMOSAT
specifications
Six digital cameras are
positioned around the
glacier basin, providing
complete glacier coverage
in situ acquisition – 3 images per day…
… weather conditions + electronics: only a fraction of the available data is usable !
08/ 15/ 2008 09/ 03/ 2008 09/ 04/ 2008
09 / 08/ 2008 09 / 10/ 2008 09 / 13/ 2008
09 / 15/ 2008 09/ 17/ 2008 09/ 19/ 2008
09/20/ 2008 09/30/ 2008 10/03/ 2008
The Delaunay triangulation (rubber sheeting) model is adapted if enough reference points are available for a dense initial triangulation.
Geometrical orthorectification models are not appropriate due to the tangential views, especially since several images taken from different cameras must be combined
Hinkler, J. , Pedersen, S. B. , Rasch, M. and Hansen, B. U.(2002) 'Automatic snow cover monitoring at high temporal and spatial resolution, using images taken by a standard digital camera', International Journal of Remote Sensing, 23: 21, 4669 — 4682
Oblique views provide a qualitative information on daily glacier evolution.
In order to be used on a map, these images must be projected.
Classical calibration models fail due to
several constraints.
X1,Y1
X2,Y2
X3,Y3
Delaunay Triangular Irregular Networks (TIN), network of contiguous triangles defined so that no vertex lies within the interior of any of the circumcircles of the triangles in the network
X est =a 1 X +b 1 Y+ε 1 Y est =a 2 X +b 2 Y+ε 2
X1,Y1
X2,Y2
X3,Y3
Latitudes and longitudes are estimated from regression plane equations or a
spline surface.
In situ 22 july 2009
Formosat 22 july 2009
No reference control point on the glacier
Reference points are defined on the glacier using flags
Detail
Localisation GPS des points
GCP #9
GCP #8
GCP #7 GCP #6
GCP #5 GCP #4
GCP #3 GCP #2 GCP #1
GCP #99 GCP #98 GCP #97
GCP #95 GCP #94
GCP #93 GCP #91
GCP #90 GCP #87
GCP #86 GCP #85
GCP #84 GCP #82
GCP #81 GCP #77 GCP #76 GCP #74 GCP #70
GCP #69 GCP #68 GCP #67 GCP #66
GCP #52 GCP #51
GCP #50 GCP #49
GCP #48 GCP #46
GCP #45 GCP #44
GCP #43 GCP #42
GCP #41 GCP #40
GCP #39 GCP #38
GCP #36 GCP #35
GCP #34 GCP #33
GCP #32 GCP #27
GCP #31
GCP #30 GCP #29 GCP #28 GCP #26
GCP #25
GCP #24
GCP #23 GCP #22
GCP #20
GCP #21 GCP #19
GCP #18 GCP #17 GCP #16
GCP #15 GCP #14 GCP #13 GCP #12
GCP #11
GCP #10
GCP #286
GCP #284 GCP #283
GCP #281 GCP #280
GCP #279
GCP #276
GCP #273 GCP #274
GCP #272 GCP #270 GCP #271
GCP #269 GCP #268 GCP #267 GCP #266
GCP #265 GCP #264
GCP #263
GCP #262 GCP #261
GCP #260 GCP #259 GCP #258 GCP #257
GCP #256 GCP #255
GCP #253 GCP #251 GCP #250 GCP #249
GCP #247
GCP #245
GCP #238 GCP #237
GCP #236 GCP #234 GCP #233
GCP #232 GCP #231
GCP #228 GCP #224 GCP #223 GCP #222
GCP #221 GCP #220 GCP #218
GCP #216 GCP #215 GCP #214 GCP #213
GCP #212
GCP #211 GCP #210 GCP #209
GCP #208 GCP #207
GCP #203 GCP #202
GCP #200 GCP #199
GCP #196
GCP #193
GCP #172
GCP #174 GCP #176
GCP #188 GCP #181
GCP #187 GCP #185 GCP #184
GCP #183
GCP #180 GCP #178 GCP #177
GCP #175
GCP #173
GCP #170
GCP #164
GCP #129 GCP #127
Ground Control Point
Delaunay interpolation triangles (TIN)
Latitude and longitude simulation
Final rectification
Actual limit of the glacier
Not visible area
Combination of images provided by different cameras
(over 98 % of the glacier surface is mapped)
Images for which the geometrical model was not defined must be corrected through a double geometrical transformation to be consistent
with the mapping projection.
Geometrical models for each camera are only valid as long as images are acquired in the exact same conditions.
Technical constraints associated with the instruments and harsh environmental conditions require replacing the camera several times.
Camera replacement necessarily induces some frame change/motion.
Images are corrected to match even though the orientation of the camera was changed:
a first geometrical transform allows fitting with the parameters of the projection model.
Source image to be modified Modified image using a second order polynom, consisitent with the geometry of data acquired in 2009 used as reference for the `` rubber sheeting ’’ method
2009
2008 2009
2008
In situ images
t1 -> t2
t2 -> t3
t3 -> t4
t4 -> t5
t5 -> t6
Referencial geometry
Rubber sheeting model Delaunay triangulation
UTM 33 Database of
In situ data
Generalization to all the in situ images
(about 1000/camera/year, or a total of 26 000 images since 2007)
Very high density of ground control points
(more than 200 for each in situ image)
S
N
O
W
/
I
C
E
C
O
V
E
R
A
G
E
Snow / Ice coverage and quality estimation
Hinkler, J. , Pedersen, S. B. , Rasch, M. and Hansen, B. U.(2002)
‘Automatic snow cove r monitoring at high temporal and spatial resolution, using images taken by a standard digital camera', International Journal of Remote Sensing, 23: 21, 4669 — 4682
Normalized Difference Snow Index - NDSI
With digital camera approximation – NDSI
RGBWhere :
a and b empirical parameters depending of camera
E M P I R I C A L
N O T
E F F I C I E N T
W I T H
S E V E R A L
C A M E R A S
Binary snow and ice visual interpretation…
ICE
SNOW
Snow high/density
database
… completed by regular snow drills
field measurments (high and snow
density)
september13 th
ELA
september15 th
ELA
September 20 th
ELA
UTM 33 Database of
In situ data
-6 -5 -4 -3 -2 -1 0 1 2
Daily ELA multitemporal evolution
Daily map temperature of the glacier Positive mean temperature of the glacier
Negative mean temperature of the glacier
Daily snow high map (interpolate) Daily map precipitation
Estimating the fraction of glacier melt in the hydrologic budget
33 temperature
data logger (each hour)
FORMOSAT regular data acquisition
NDSI
RGBindex/ELA map
Water snow melt estimation Snow melt
modelling
… march – april – may – june – july – august – september – october – november – december – january – february – march – april – may – june – july – august - september …
Simple model of snow melt – day degree fusion model
k : coefficent defining the influence of natural and climatic conditions of the basin on melting, 5 mm/degC for snow and 7 mm/degC for ice.
T
i: mean air temperature or maximum daily temperature.
T
0: threshold temperature above which snow melts.
Day-degree fusion model:
Can be simplified as:
Considering precicipitations, one obtains:
α: fitted parameter, usually 0.0035 mm
-1. P
i: total daily precipitations
One should additionnaly consider the influence of progressive stock reduction and the accumulation of rain water in snow.
Both satellite images and ground pictures give a binary information concerning the presence of snow or presence of ice. This differenciation is very important to determine, for each point on the glacier surface, the melting coefficient k of the moment which determines the amount of water coming from the melting of snow and ice.
Network of Data loggers since 2008
Drills field measures since
2008
Simple model of snow melt – day degree fusion model September 15 th 2008
Mean daily air temperature IDW interpolate map
Snow/ice coverage from mosaïc of projected in situ images
Water height equivalent of snow-ice melt using model
- map
Snow Ice
3.9 °C 7.1 °C
OUTPUT
ELA
19 mm 49 mm
INPUT INPUT
Simple model of snow melt – day degree fusion model September 15 th 2008
Water height equivalent of snow-ice melt using model
_
Statistic - Diagram
1 10 100 1000 10000
19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 Total
volume (m 3 )
13.8 mm mean of water equivalent snow
melt on the drainage basin.
-
144.7 10 3 m 3 with 56% from snow (81 10 3 m 3 vs 63.7 10 3 m 3
from ice)
melt (mm) nb_pix area (m²)
volume (m3) 19
3827 15308 290.852
20
8152 32608 652.16
21
17192 68768 1444.128
2273070 292280 6430.16
239954 39816 915.768
2417897 71588 1718.112
2581230 324920 8123
26124249 496996 12921.896
2762874 251496 6790.392
2850187 200748 5620.944
2986301 345204 10010.916
30122140 488560 14656.8
3110175 40700 1261.7
3213808 55232 1767.424
3312734 50936 1680.888
3416221 64884 2206.056
3517345 69380 2428.3
3632994 131976 4751.136
3729389 117556 4349.572
3837509 150036 5701.368
3928924 115696 4512.144
4048314 193256 7730.24
4139769 159076 6522.116
4245200 180800 7593.6
4334438 137752 5923.336
4435221 140884 6198.896
4518501 74004 3330.18
4615516 62064 2854.944
4719447 77788 3656.036
486727 26908 1291.584
497363 29452 1443.148
Total1126668 4506672 144777.796
Budget
Simple model of snow melt – day degree fusion model
Each day during the flood period
We can now estimate snow melt, and hence the water equivalent thickness for each pixel, in order to define the fraction of ice and snow melt in
hydrological budgets
Conclusion
As a conclusion, we demonstrate in this presentation that the conversion of ground pictures into aerial recomposed images may be successfully made for
an Arctic glacierized system, especially during summer period when the hydrological activity is the most intense.
This original approach is very relevant for Arctic where the dynamics of processes is rarely observed and therefore is not easily quantified by
classical methods.
Thank you
ISIS
Incitation à l’utilisation Scientifique des Images Spot
