1University of Bern, Department of Geography, Bern, Switzerland

1
Land Products and Applications Evaluation of
Fractional Snow Cover Maps derived from AVHRR
with ASTER Data Sets
1Nando Foppa, 1Adrian Hauser, 1David Oesch,
1Stefan Wunderle, 2Roland Meister and 2Manfred
Stähli

• 1University of Bern, Department of Geography,
  Bern, Switzerland
• 2Swiss Federal Institute for Snow and Avalanche
  Research, Davos, Switzerland

International EOS/NPP Direct Readout Meeting,
Benevento 2005
2
Outline

• Introduction
• AVHRR sub-pixel approach Method and processing
• Comparison study Determining snow cover extent
  and duration in a mountain region
• Concept of validation Evaluation based on ASTER
  data
• Application Synergy of in-situ and AVHRR data
  for improved snow depth mapping
• Conclusion and Outlook

Introduction - Sub-pixel approach - Comparison
study - Concept of Validation - Application -
Conclusion and Outlook
3
Why snow cover analysis?

• Numerical weather prediction
• Input data for meso-scale NWP models
• Climate studies
• Snow line analysis
• Hydrology
• Operational hydrological runoff modelling
• Snow depth
• Improving accuracy of snow depth maps

Introduction - Sub-pixel approach - Comparison
study - Concept of Validation - Application -
Conclusion and Outlook
4
Conventional snow cover analysis

• Surface observation
• point measurements
• limited number of stations
• spatial distribution
• temporal resolution
• in-situ method

Introduction - Sub-pixel approach - Comparison
study - Concept of Validation - Application -
Conclusion and Outlook
5
Filling the gapsSatellite-based snow detection
over continuous space-time scale
Introduction - Sub-pixel approach - Comparison
study - Concept of Validation - Application -
Conclusion and Outlook
6
Why sub-pixel analysis?

• sensors spatial resolution
• small patches of snow
• heterogeneous surface
• pixels are a mixture of different surface cover
  types!
• -gt sub-pixel snow cover estimation

Introduction - Sub-pixel approach - Comparison
study - Concept of Validation - Application -
Conclusion and Outlook
7
Linear spectral mixture modelling
-gt assuming linear mixing the spectrum of a
pixel is an area-weighted average of the
pure-element reflections called endmembers
Introduction - Sub-pixel approach - Comparison
study - Concept of Validation - Application -
Conclusion and Outlook
8
NOAA AVHRR sub-pixel processingA scene-specific
approach for operational and near-real time
Introduction - Sub-pixel approach - Comparison
study - Concept of Validation - Application -
Conclusion and Outlook
9
Pre-processing of NOAA AVHRR DataSince 2002 in
operational status data archive since the 1980s
local HRPT receiving station at Bern, Switzerland

• calibration
• geometric correction
• orthorectification
• atmospheric correction (SMAC)
• cloud detection (CASPR)

Introduction - Sub-pixel approach - Comparison
study - Concept of Validation - Application -
Conclusion and Outlook
10
AVHRR fractional snow cover productNOAA-17,
February 01, 2005
differentiated snow distribution
information! all pixels contain a minimal
amount of snow cover!
Introduction - Sub-pixel approach - Comparison
study - Concept of Validation - Application -
Conclusion and Outlook
11
Post-processingWhen is an AVHRR pixel snow-free?

• max 7 snow fraction

Introduction - Sub-pixel approach - Comparison
study - Concept of Validation - Application -
Conclusion and Outlook
12
Fractional snow cover productNOAA-17, February
01, 2005
Introduction - Sub-pixel approach - Comparison
study - Concept of Validation - Application -
Conclusion and Outlook
13
Comparison studyDetermining snow cover extent
and duration in a mountain region

• in situ measurements with a large sample of
  temperature loggers and snow station data
• remote-sensing using AVHRR-derived sub-pixel
  snow cover maps
• a distributed numerical snow hydrology model
  (Alpine3D)
• These methods were applied at two different
  scales
• the landscape scale represented by the region of
  Davos (Switzerland)
• the hillslope scale represented by forested
  slopes of approximately 1 km2 area

Introduction - Sub-pixel approach - Comparison
study - Concept of Validation - Application -
Conclusion and Outlook
14
Landscape scaleNOAA-AVHRR sub-pixel snow cover
map vs. Alpine3D
Introduction - Sub-pixel approach - Comparison
study - Concept of Validation - Application -
Conclusion and Outlook
15
Hillslope scaleSnow depletion curves
Introduction - Sub-pixel approach - Comparison
study - Concept of Validation - Application -
Conclusion and Outlook
16
Concept of ValidationEvaluation of Fractional
Snow Cover Maps with ASTER Data

• validation of the sub-pixel snow product using
  high spatial resolution data from the Advanced
  Spaceborne Thermal Emission and Reflection
  Radiometer (ASTER)
• on-board Terra (MODIS), launch Dec. 1999 swath
  width 60km, VNIR (3 channels) 15m SWIR (6
  channels) 30m TIR (5 channels) 90m
• ASTER provides the user community various
  On-Demand Data Products (ASTER On-Demand L2
  Surface Reflectance Product AST_07) -gt DAAC
• easily available and cost-effective alternative
• useful for snow classification

Introduction - Sub-pixel approach - Comparison
study - Concept of Validation - Application -
Conclusion and Outlook
17
Validation of Fractional Snow Cover MapsASTER
data sets and Areas of Interest
Introduction - Sub-pixel approach - Comparison
study - Concept of Validation - Application -
Conclusion and Outlook
18
ASTERAdvanced Spaceborne Thermal Emission and
Refl. Radiometer
1)
2)
3)
RGBch3,2,1
Introduction - Sub-pixel approach - Comparison
study - Concept of Validation - Application -
Conclusion and Outlook
19
ASTER data using as ground truth Snow fraction
determination at AVHRR scale from ASTER data
binary classif. (VIS-SWIR)
(VISSWIR)
NDSI
snow pixels vs. total pixels per 1000 m grid cell
Introduction - Sub-pixel approach - Comparison
study - Concept of Validation - Application -
Conclusion and Outlook
20
ASTER vs. AVHRR snow cover fraction
Introduction - Sub-pixel approach - Comparison
study - Concept of Validation - Application -
Conclusion and Outlook
21
ASTER vs. AVHRR snow cover fraction
Introduction - Sub-pixel approach - Comparison
study - Concept of Validation - Application -
Conclusion and Outlook
22
Near-real time applicationSynergy of in-situ
and AVHRR data for snow depth mapping

• conventional snow depth maps are based on snow
  station measurements using spatial interpolation
  technique (1 km grid)
• number of stations varies and is not uniformly
  distributed -gt overestimation of snow depth
• synergetic retrieval of snow depth using a
  combination of spatial interpolated snow depth
  values and daily snow cover extent from satellite
  imagery (AVHRR)
• using the snow-free information from the
  sub-pixel snow product for virtual snow stations
  to feed the snow depth interpolation model

Introduction - Sub-pixel approach - Comparison
study - Concept of Validation - Application -
Conclusion and Outlook
23
NOAA-17January 04, 2005, 1000 UTC
RGBch1,2,3
Introduction - Sub-pixel approach - Comparison
study - Concept of Validation - Application -
Conclusion and Outlook
24
Merging the snow information

• interested in snow-free pixels
• binary classification
• virtual station network
• intersected points with no snow values snow
  stations with 0 cm

snow depth 0 cm not included
Introduction - Sub-pixel approach - Comparison
study - Concept of Validation - Application -
Conclusion and Outlook
25
Snow depth measurements used only

• snow cover extends into the lowland with up to 25
  cm snow depth!
• high snow depth values in the central Alps and in
  the southern part.

Introduction - Sub-pixel approach - Comparison
study - Concept of Validation - Application -
Conclusion and Outlook
26
Including virtual snow stations
snow no snow border fits closer to the snow
cover extent from the satellite data snow
depth decreases in the main valleys
Introduction - Sub-pixel approach - Comparison
study - Concept of Validation - Application -
Conclusion and Outlook
27
Conclusion

• simple and automatic scene-specific technique
  suitable for operational and near real-time
  applications
• model simplifications constraints influence
  estimated snow fractions,model noise
• snow cover extent derived from AVHRR is
  plausible compared to model simulation
• AVHRR snow detection is defined by technical
  limitations and by shadow and snow beneath
  forest canopies observable snow in the AVHRR
  pixel is detected
• preliminary validation using ASTER data
  demonstrates AVHRR
  approach underestimates snow fraction for high
  snow-cover conditions.. ..and overestimates the
  snow fraction when the snow cover becomes patchy
• the application of AVHRR fractional snow cover
  maps for snow depth mapping shows promising
  improvement

Introduction - Sub-pixel approach - Comparison
study - Concept of Validation - Application -
Conclusion and Outlook
28
Outlook

• validation of the AVHRR sub-pixel snow product
  must continue
• comparison with MODIS 500 m snow fraction
  product (for forward and re- processing of all
  data from February 2000)
• incorporating a correction-model to adjust for
  over- and underestimation of the AVHRR snow
  fraction values
• additional investigations concentrate on
  understand the influence of bi- directional
  properties of snow, topography and shadow effects
• long-term monitoring of the snow cover in the
  Alps using AVHRR data
• including and evaluating the reflective part of
  AVHRR channel 3 (NOAA-12, -14, -16, -18) in the
  sub-pixel procedure

Introduction - Sub-pixel approach - Comparison
study - Concept of Validation - Application -
Conclusion and Outlook
29
Web-productshttp//saturn.unibe.ch/rsbern/noaa/d
w/realtime/index.html
30
Web-productshttp//saturn.unibe.ch/rsbern/noaa/d
w/realtime/index.html
31
Thank you for your attention and
comments... ...and very welcome for skiing
in the Swiss Alps!
Nando Foppa foppa_at_giub.unibe.ch http//saturn.unib
e.ch/rsbern/noaa/
32
(No Transcript)
33
Regionalisation of snow depth
HS snow depth cm G base value
cm A compensation value cm h altitude of
the grid cell m asl j grid cell to be
calculated
hj altitude of the grid cell m asl djj
distance between grid cell j and observation
station i km HSi measured snow depth at
station i cm G(hi) base value of modelled snow
depth at the observation station
cm hi altitude of the snow station m
asl j grid cell j, to be varied at a 1 km
grid over the whole area i observation station
i, i3 and minimum distance to the grid cell j
34
Snow cover vs. snow depth
35
Relative accuracy snow cover changes after snow
fall
36
Relative accuracy snow cover changes during
ablation
37
Real-time snow cover map for NWP March 16, 2004,
1200 UTC
38
ASTER data setAdvanced Spaceborne Thermal
Emission and Refl. Radiometer
Instrument On-board Terra (MODIS), launch Dec.
1999 Swath Width 60 Kms VNIR (3 channels)
15m SWIR (6 channels) 30m TIR (5 channels)
90m Product ASTER L1B Registered Radiance at
the Sensor 60 U ASTER On-Demand L2 Surface
Reflectance VNIR for free
39
Transect through the Alps
MODIS binary snow cover map (500 m)
40
Temporal coverage of snow maps 2003/04Real-time
processing
41
Snow Cover Difference Map, 17.01.03
42
(No Transcript)
43
Binary vs. sub-pixel classification
MODIS snow product
AVHRR ISODATA
AVHRR subpixel product
44
BRDF
45
Influence of shade and forestFebruary 10, 2004,
0937
46
Snow misclassificationNOAA-17, February 13,
2004, 1008
47
AVHRR sub-pixel snow productNOAA-17, January 16,
2005, 1019 UTC

• Sub-pixel snow product
• cloud- and water mask
• NDSI threshold
• sub-pixel estimation
• threshold
• shade fraction estimation
• shade covered snow
• combined end-product

48
AVHRR / ASTER spatial profile
49
AVHRR and ASTER spatial profile I
50
AVHRR and ASTER spatial profile II
51
Temporal variations of snow cover
boxcar-av 5 / AVHRR sub-pixel / in-situ data
52
Comparison with in-situ data
53
Snow depth difference map
areas where the snow depth increased including
the virtual snow stations -gthigh altitude areas
with decreased snow depth -gtlower and high
altitude
Introduction - Sub-pixel approach - Comparison
study - Concept of Validation - Application -
Conclusion and Outlook
54
Snow cover vs. snow depth, February 9 - 19, 2004
55
Snow cover vs. snow depth, February 9 - 19, 2004