mapping of snow cover extent over mountainous terrain in the...

Mapping of Snow Cover Extent over Mountainous Terrain in the Swiss National Park using

Multi-temporal & Multi-angular CHRIS Imagery

Vittal Boggaram & Ray Merton

© School of Biological, Earth, and Environmental Sciences (BEES) The University of New South Wales

Sydney, AUSTRALIA



Vittal Boggaram & Ray Merton


Sydney, AUSTRALIA

4th CHRIS/Proba Principal Investigators Workshop 2006. European Space Agency (ESRIN), Frascati, Italy. 19 – 21 September 2006.


HyperspectralApplications

SensorsAstroVision.

Hyperion, CHRIS/Proba, MSMISat.AVIRIS, HyMap, CASI.

ASD.

SensorsAstroVision.

Hyperion, CHRIS/Proba, MSMISat.AVIRIS, HyMap, CASI.

ASD.

Multi-temporalHysteresis

Community successionNear-shore topography

Multi-temporalHysteresis

Community successionNear-shore topography

Multi-angularESA CHRIS

Topographic normalisationCoastal wave analysis

Goniometers

Multi-angularESA CHRIS

Topographic normalisationCoastal wave analysis

Goniometers

WaterBathymetryCoral reefs

Phytoplankton fluorescenceAquatic Ecosystems

VegetationBiochemistryBiogeochem

StressLidar

Canopy geometryGrapes/viticulture

Fire fuel load

Dr Ray MertonResearch Thrusts

Dr Ray MertonResearch Thrusts

Dr Ray Merton

1. Boggaram, V., & Merton, R. N. (2006) Mapping snow cover extent over mountainous terrain in the Swiss National Park using multi-temporal and multi-angular CHRIS imagery.

2. Brace, B., & Merton, R. N. (2006) CHRIS/Proba hyperspectral image analysis for marine bathymetry determinations.

3. David, N. M., & Merton, R. N. (2006) Bathymetric mapping of coral reef communities in the presence of partial cloud shadow, using CHRIS/Proba hyperspectral data.

4. Davies, S. W., & Merton, R. N. (2006) Bathymetric & benthic investigation of coastal lake environments using multi-angle CHRIS/Proba hyperspectral data.

5. Holmes, W., & Merton, R. N. (2006) Investigation of the BRDF response of Sydney’s Northern Beaches sands by CHRIS/Proba hyperspectral imagery.

6. Merton, R. N. (2006) Multi-angular mapping techniques for the enhancement of steep terrain classification.

7. Miller, I., & Merton, R. N. (2006) Assessing the ultility of multi-angle hyperspectral CHRIS data for the detection of a Trichodesmium bloom, Great Barrier Reef, Australia.

8. Miller, I., & Merton, R. N. (2006) Multi-angular mapping of a coral reef using satellite hyperspectral BRDF data.

9. Mitchell, B., Merton, R. N. (2006) Contributions of multi-view angle remote sensing in determining vegetation structure and classification using CHRIS/Proba hyperspectral data.

10. Rowe, S., & Merton, R. N. (2006) Wave train analysis for inverse bathymetry estimationusing CHRIS/Proba multi-angular satellite imagery.

11. Vidal, K., & Merton, R. N. (2006) Deriving bathymetric contours from CHRIS/Proba hyperspectral imagery.

1. Boggaram, V., & Merton, R. N. (2006) Mapping snow cover extent over mountainous terrain in the Swiss National Park using multi-temporal and multi-angular CHRIS imagery.

2. Brace, B., & Merton, R. N. (2006) CHRIS/Proba hyperspectral image analysis for marine bathymetry determinations.

3. David, N. M., & Merton, R. N. (2006) Bathymetric mapping of coral reef communities in the presence of partial cloud shadow, using CHRIS/Proba hyperspectral data.

4. Davies, S. W., & Merton, R. N. (2006) Bathymetric & benthic investigation of coastal lake environments using multi-angle CHRIS/Proba hyperspectral data.

5. Holmes, W., & Merton, R. N. (2006) Investigation of the BRDF response of Sydney’s Northern Beaches sands by CHRIS/Proba hyperspectral imagery.

6. Merton, R. N. (2006) Multi-angular mapping techniques for the enhancement of steep terrain classification.

7. Miller, I., & Merton, R. N. (2006) Assessing the ultility of multi-angle hyperspectral CHRIS data for the detection of a Trichodesmium bloom, Great Barrier Reef, Australia.

8. Miller, I., & Merton, R. N. (2006) Multi-angular mapping of a coral reef using satellite hyperspectral BRDF data.

9. Mitchell, B., Merton, R. N. (2006) Contributions of multi-view angle remote sensing in determining vegetation structure and classification using CHRIS/Proba hyperspectral data.

10. Rowe, S., & Merton, R. N. (2006) Wave train analysis for inverse bathymetry estimationusing CHRIS/Proba multi-angular satellite imagery.

11. Vidal, K., & Merton, R. N. (2006) Deriving bathymetric contours from CHRIS/Proba hyperspectral imagery.

2006 ESA CHRIS/Proba Papers2006 ESA CHRIS/Proba Papers

…plus MSMISat.

Vittal Boggaram & Ray MertonVittal Boggaram & Ray Merton





Dr Ray Merton

Study AreaStudy Area

• The SNP is located to the southeast of Switzerland and to the north of the Italian border. It is the largest naturally protected area in Switzerland and covers an area of approximately 172.4 km2. The national park is located in the central Alps at elevationsranging from 1,400 m (Clemgia Gorge) to 3,174 m (Piz Pisoc) above sea level.

• The climate of the region is dry, harsh, with strong solar radiation and low humidity. The average annual precipitation is 950 mm and the average annual temperature is 0° C.

• SNP can be broadly divided into forest areas (28%), alpine and sub-alpine grassland (21%) and mountainous terrain (51%). The forest areas predominantly contain coniferous pine species like

– Mountain Pine (Pinus montana), Swiss Stone pine (Pinus cembra), European Larch (Larixdecidua), Scots Pine (Pinus sylvestris) and Norway Spruce (Picea abies).

– The dominant Mountain pine stands cover large areas of the mountainous terrain. The tree linein certain areas of the SNP is approximately 2,300 m a.s.l, which is comparatively higher than the average of 1,900 m a.s.l in Switzerland.

• A rational for selecting the study area was the relatively high tree line ranging from approximately 2,150 m to 2,250 m a.s.l and distinct permanent and seasonal snow covered areas. Satellite imagery of the area was acquired on a multi-temporal seasonal sequence to map the extent of the permanent and seasonal snow cover above the tree line on the mountain slopes.

• The SNP is located to the southeast of Switzerland and to the north of the Italian border. It is the largest naturally protected area in Switzerland and covers an area of approximately 172.4 km2. The national park is located in the central Alps at elevationsranging from 1,400 m (Clemgia Gorge) to 3,174 m (Piz Pisoc) above sea level.

• The climate of the region is dry, harsh, with strong solar radiation and low humidity. The average annual precipitation is 950 mm and the average annual temperature is 0° C.

• SNP can be broadly divided into forest areas (28%), alpine and sub-alpine grassland (21%) and mountainous terrain (51%). The forest areas predominantly contain coniferous pine species like

– Mountain Pine (Pinus montana), Swiss Stone pine (Pinus cembra), European Larch (Larixdecidua), Scots Pine (Pinus sylvestris) and Norway Spruce (Picea abies).

– The dominant Mountain pine stands cover large areas of the mountainous terrain. The tree linein certain areas of the SNP is approximately 2,300 m a.s.l, which is comparatively higher than the average of 1,900 m a.s.l in Switzerland.

• A rational for selecting the study area was the relatively high tree line ranging from approximately 2,150 m to 2,250 m a.s.l and distinct permanent and seasonal snow covered areas. Satellite imagery of the area was acquired on a multi-temporal seasonal sequence to map the extent of the permanent and seasonal snow cover above the tree line on the mountain slopes.

Dr Ray Merton

Multi-temporalMulti-temporal22 May 200528 May 200510 July 20052 Sept 200529 Nov 2005

Dr Ray Merton

Multi-angular DataMulti-angular Data

Snow ReflectanceSnow Reflectance

Curves Showing Snow Reflectance (%)

0

20

40

60

80

100

120

0.3 0.4 0.5 0.6 0.7 0.8 1.8 2.27 2.32 2.37 2.43Wavelength (Micrometres)

medium reflectance fine reflectance coarse reflectanceGrain Size:

Reflectance of Snow & CloudsReflectance of Snow & Clouds

Modified From Jensen (2000)

Normalized Difference Snow Index (NDSI)


• The high albedo of snow and clouds in the visible region of the EMS makes it difficult to differentiate between the two (Hall, et al., 1998). The reduction in the reflectance of snow in the mid-IR region can be used in differentiating snow from cloud cover (Jensen, 2000). Cloud cover is predominant in the SNP images acquired during the summer month of July. Accurate estimation of snow cover in the July images was difficult as cloud & snow cover overlap. The multi-angular nature of the CHRIS images generates shadows in mountain terrain posing a greater challenge in estimating the snow cover area.

• NDSI is based on the NDVI concept used for mapping vegetation in remote sensing imagery (Hinkler, et al., 2000; Hall, et al., 1998). NDSI uses spectral band ratios to determine the relative band intensity to differentiate snow from cloud cover (Hinkler, et al., 2000; Hall, et al., 1998). The NDSI band ratio is advantageous in mapping the snow line in mountainous terrain. The spectral band ratio can enhance features by eliminating atmospheric effects and viewing geometry of the image (Gupta, 2005).

• The NDSI band ratio is generated by calculating the difference between the visible band (0.52µm -0.59µm) and the near-IR band (1.55µm – 1.7µm) and dividing the result by the sum of the two bands (Gupta, 2005; Hall, et al., 1998). The algorithm is designed to detect snow in each pixel. The NDSI algorithm was modified for trial in this study as:

NDSI = CHRIS B3 – CHRIS B18CHRIS B3 + CHRIS B18

• July images were processed using the NDSI spectral band ratio. The July NDSI images were not accurate in differentiating snow from cloud cover as the wavelength range (0.4µm to 1.05µm) of the CHRIS images pose limitations. Wavelength bands in the near and mid-IR regions (1.55µm – 1.7µm) of the EMS can generate better results. The NDSI images can be combined into a colour-ratio-composite (CRC) image, which helps in determining the approximate spectral shape for each pixel.

• The high albedo of snow and clouds in the visible region of the EMS makes it difficult to differentiate between the two (Hall, et al., 1998). The reduction in the reflectance of snow in the mid-IR region can be used in differentiating snow from cloud cover (Jensen, 2000). Cloud cover is predominant in the SNP images acquired during the summer month of July. Accurate estimation of snow cover in the July images was difficult as cloud & snow cover overlap. The multi-angular nature of the CHRIS images generates shadows in mountain terrain posing a greater challenge in estimating the snow cover area.

• NDSI is based on the NDVI concept used for mapping vegetation in remote sensing imagery (Hinkler, et al., 2000; Hall, et al., 1998). NDSI uses spectral band ratios to determine the relative band intensity to differentiate snow from cloud cover (Hinkler, et al., 2000; Hall, et al., 1998). The NDSI band ratio is advantageous in mapping the snow line in mountainous terrain. The spectral band ratio can enhance features by eliminating atmospheric effects and viewing geometry of the image (Gupta, 2005).

• The NDSI band ratio is generated by calculating the difference between the visible band (0.52µm -0.59µm) and the near-IR band (1.55µm – 1.7µm) and dividing the result by the sum of the two bands (Gupta, 2005; Hall, et al., 1998). The algorithm is designed to detect snow in each pixel. The NDSI algorithm was modified for trial in this study as:

NDSI = CHRIS B3 – CHRIS B18CHRIS B3 + CHRIS B18

• July images were processed using the NDSI spectral band ratio. The July NDSI images were not accurate in differentiating snow from cloud cover as the wavelength range (0.4µm to 1.05µm) of the CHRIS images pose limitations. Wavelength bands in the near and mid-IR regions (1.55µm – 1.7µm) of the EMS can generate better results. The NDSI images can be combined into a colour-ratio-composite (CRC) image, which helps in determining the approximate spectral shape for each pixel.

Dr Ray Merton

SAMSAM

SAM image with features classed together. Cloud cover classed as snow. SAM image with features classed together. Cloud cover classed as snow.

0º

Dr Ray Merton

SAMSAM

SAM image with features classed together. Cloud cover classed as snow. SAM image with features classed together. Cloud cover classed as snow.

-55º

Dr Ray Merton

Minimum DistanceMinimum Distance 0º

Dr Ray Merton

Minimum DistanceMinimum Distance -55º

Dr Ray Merton

NDSI Minimum DistanceNDSI Minimum Distance 0º

Dr Ray Merton

July 2004 - Snow Cover

0

50000

100000

150000

200000

250000

-55 -33 0 33 55

CHRIS Image Angles

Num

ber o

f Pix

els

SAM

MinimumDistance

November 2004 - Snow Cover

0

50000

100000

150000

200000

250000

300000

350000

400000

-55 -33 0 33 55

CHRIS Image Angels

Num

ber o

f Pix

els

SAM

MinimumDistance

July 2004 - Snow Cover - SAM

0

50000

100000

150000

200000

250000

-55 -33 0 33 55

Image Angles

Num

ber o

f Pix

els

November 2004 - Snow Cover - SAM

0

50000

100000

150000

200000

250000

300000

350000

-55 -33 0 33 55

Image Angles

Num

ber o

f Pix

els

July 2004-Snow Cover- Minimum Distance

0

10000

20000

30000

40000

50000

60000

-55 -33 0 33 55

Image Angles

Num

ber o

f Pix

els

November 2004 - Snow Cover - Minimum Distance

0

50000

100000

150000

200000

250000

300000

350000

400000

-55 -33 0 33 55

Image Angles

Num

ber o

f Pix

els

Classified Vector ImagesClassified Vector Images 0º

Dr Ray Merton

0º

-55º

+36º

-36º

+55º

Classified Vector ImagesJuly Datasets

Classified Vector ImagesJuly Datasets

Dr Ray Merton

Multi-Angular Mapping Techniques for the Enhancement of Steep Terrain Classification

Ray Merton


Sydney, AUSTRALIA

Multi-Angular Mapping Techniques for the Enhancement of Steep Terrain Classification

Ray Merton


Sydney, AUSTRALIA



Table Mountain-Cape Town, South Africa

Dr Ray Merton

-36-36

+36+36

SSNN

CHRIS/ProbaSteep Terrain Normalisation

CHRIS/ProbaSteep Terrain Normalisation

ab

b

e.g. 32% more pixels

nadir

• +36º datasets for N slopes• -36º datasets for S slopes• Across track capability for E-W• Across+along track options

• +36º datasets for N slopes• -36º datasets for S slopes• Across track capability for E-W• Across+along track options

Dr Ray Merton

0º

target

-36º-36º+36º+36º

Observation Zenith AnglesObservation Zenith Angles

+55º+55º -55º-55º

-20º-20º

+50º+50º

+20º+20º

-50º-50º

Dr Ray Merton

-36º-36º

0º0º

target

+36º+36º

-20º-20º

+50º+50º

+20º+20º

-50º-50º

Observation Zenith AnglesObservation Zenith Angles

20º29º

39º

-55º-55º+55º+55º

50º50º

±36º±36º ±55º±55ºnadirnadir

Slopes 21 - 50º

Slopes 51 - 90ºCliff face

90º90º

Elev. = 20ºElev. = 20º

Slopes 0 - 20º

Obs. Angles vs. Slope Elevation AnglesObs. Angles vs. Slope Elevation Angles

…this also applies to across-track observations.

Level ground0º0º

50º50º

±36º±36º ±55º±55ºnadirnadir

Slopes 21 - 50º

Slopes 21 - 50ºSlopes 51 - 90º

Slopes 51 - 90º Cliff face

90º90º

20º20º

Slopes 0 - 20ºSlopes 0 - 20º

Observation Angles vs. Slope AnglesObservation Angles vs. Slope Angles

…this also applies to across-track observations.

Level ground0º0º

Dr Ray Merton

Mountainous TerrainTable Mountain, Cape Town

Mountainous TerrainTable Mountain, Cape Town

…analogous to multi-angular ROI’s

Dr Ray Merton

Mountainous TerrainCape Town, South Africa

Mountainous TerrainCape Town, South Africa

Lion’s HeadLion’s Head

Cable Car top station33.957394S, 18.402964E

Cable Car top station33.957394S, 18.402964EDevil’s PeakDevil’s Peak

= Priority 2 research area= Priority 2 research area= Priority 3 research area= Priority 3 research area

= Priority 1 research area= Priority 1 research area

Dr Ray MertonCHRIS/Proba PI, AustraliaDr Ray MertonCHRIS/Proba PI, Australia

Topographic normalisation technique, veget. fire regrowth, multi-angular corrections.Topographic normalisation technique, veget. fire regrowth, multi-angular corrections.

Table Mountain-Cape Town, South AfricaTable Mountain-Cape Town, South Africa

+55[63% of full scene]

+55[63% of full scene]

+36[37%

of full scene]

+36[37%

of full scene]

-36[63%]-36[63%]

-55[0%]

(not acquired??)

-55[0%]

(not acquired??)

Nadir[100%]

…perfect coverage

Nadir[100%]

…perfect coverage

Lion’s HeadLion’s Head

Priority 1Priority 1

Priority 2Priority 2

TargetingPriority 3TargetingPriority 3

HRC image centrecoordinate

33.942299S, 18.395176E

HRC image centrecoordinate

33.942299S, 18.395176E

New CHRIS image centre coord33.957394S, 18.402964E(Cable Car top station)

New CHRIS image centre coord33.957394S, 18.402964E(Cable Car top station)

Devil’s PeakDevil’s Peak

CHRIS_TM_060203

Dr Ray Merton

5 X-angularCHRIS Images

BRDF Modelling AspectPolygons

5 Classification ImagesDEM/TIN

Single layerComposite Image

Steep Terrain Normalisation ModelSteep Terrain Normalisation Model

5 ROI subset Classification Images

Lion’s Head, Cape Town(HRC PAN imagery + DEM)

Lion’s Head, Cape Town(HRC PAN imagery + DEM)

Cape Town, South AfricaCape Town, South Africa

mapping of snow cover extent over mountainous terrain in the...

Documents