buckley-smith igarss2011.pptx

Comparing RADARSAT 2 and TerraSAR-X

Quad-Pol SAR Imagery of Grasslands

Joseph R. Buckley1 and Anne M. Smith2

1Royal Military College of Canada, Kingston, ON, Canada 2Agriculture and Agri-Food Canada, Lethbridge, AB, Canada

National DéfenseDefence national

Outline

Setting The Scene Experimental Design and Procedures Some Results Conclusions and Continuing Work

“Developing Earth Observation Tools to Measure the Current and Future Spatial Extent and Productivity in Grasslands of Western Canada”

A Multi-Year, Multi-Agency ExperimentCanadian Space Agency Funded

The Big Picture

• 350 M ha North America, • 30 M ha Western Canada• 25% left• Cattle production*• Wildlife• Tourism• Biodiversity*• Greenhouse gas emissions*• Water resources*• Agriculture, oil & gas, urbanization, over grazing, invasive species

Enhancing environmental performance of the Canadian agricultural

system, enhancing economic benefits for all stakeholders

Native grasslands

Native grassland

Photographs courtesy of: G. Bourgeois, M. Didkowsky R. Bourchier, J. Nicholsen, G. Larson, C. Kloppenburg.

Image supply Mapping productsLand management

ATIC, CSA, TerraVista Earth Imaging

Airborne hyperspectralCalibration

Ecosystem modelingWeed invasion

ULeth, RMC, UND, UVic

Management decisions (stocking rates, grazing rotation)

Grazing Associations, RanchersLand management

(e.g. weed enforcement, pest control, soil

conservation)

Municipal

Provincial

Federal University

NGO

Industry

Land manager

Grassland

Health assessment Wildlife habitat

Landscape cumulative effects, Land Use Framework, BMP

ASRD, AFSC Health assessment Wildlife habitat

ACA

Risk assessment/PolicySpecies at risk

Pasture managementWeed biocontrol

Program evaluation (BMP, NCGAVS, NAHARP)

AAFC (Research Branch/AESB, EC (DND,

INAC, CFIA) Grasslands Project Team

To develop multispectral, hyperspectral and radar earth observation tools to address mapping and quantification of:

1. the spatial extent and fragmentation of grasslands,2. the net primary productivity of grasslands,3. the invasive plant species, leafy spurge, on grasslands.

Objectives

Progress to Date

We are in the third year of this two year project Test sites

– Throughout Southern Alberta Ground data

– Surveys, crop insurance and irrigation databases (2009, 2010, 2011) Remote sensing data

– Optical • Landsat-5 TM (1999, 2005, 2009)• SPOT4 (2009)

– Radarsat-2 (2009, 2010, 2011)– Airborne Hyperspectral (2009, 2010)– Lidar (2009, 2010)

8

“RADARSAT-2 Data and Products © MacDONALD, DETTWILER AND ASSOCIATES LTD. (2009) – All Rights

Reserved” “RADARSAT is an official mark of the Canadian Space Agency”

Newell County Regional Overview

potatoes

wheat

fallow

corn

alfalfa

native grassland

improved pasture

potatoes

Freeman decomposition

Radar vs Optical Imagery

10“RADARSAT-2 Data and Products © MacDONALD, DETTWILER AND ASSOCIATES LTD. (2009) – All Rights Reserved” “RADARSAT is an official mark of the Canadian Space Agency”

The POLSAR Comparison Experiment

DLR graciously accepted our request to participate in the spring 2010 DRA campaign, and acquired two three image quad-pol stripmap sets over two of our 2009-2010 test areas.

Three RADARSAT-2 fine mode quad pol images were acquired in as close to matching incidence angles and dates as could be arranged.

In mid-December 2010 we received the 6 SSC images from DLR Both RADARSAT-2 and TerraSAR-X images have been ingested,

speckle filtered (Lee Sigma) and decomposed (Yamaguchi4) with PolSARpro, georeferenced with ENVI and the tiepoints provided with the imagery, then fine co-registered by the usual cross-correlation method.

Preliminary comparative analysis has been done on the imagery.

General Conditions

Data Acquisition: 17 April to 11 May 2010 On the southern Alberta prairies

– Snow had melted– Lakes were no longer ice covered– No rain– No significant natural growth– Preparing and planting crops had not yet started

Natural rangelands were covered in senescent vegetation

Cultivated lands were either stubble, fallow or plowed

Observation Programme

TerraSAR-X April 12, 23, May 4 Stripmap Near 19.9 – 21.7

RADARSAT-2 May 1FQ17 36.4 – 38.0

RADARSAT-2 April 27FQ2 19.8 – 21.8

TerraSAR-X April 18, 29, May 10 Stripmap Far 30.8 – 32.3

RADARSAT-2 April 17FQ8 26.9 – 28.7

TerraSAR-X April 18 Stripmap Far 30.8 – 32.3

RADARSAT-2 April 17FQ8 26.9 – 28.7

Yamaguchi 4 Decomposition

SurfaceVolumeDouble

All scaled -30 0dB

RADARSAT-2 : April 27 19.8 – 21.8 TerraSAR-X: April 23 19.9 – 21.7

TerraSAR-X

RADARSAT-2

V

V

S

S

D

D

Yamaguchi 4 Decomposition

Cyan: TSXRed:R2

Surface scattering

RADARSAT-2 : April 17 26.9 – 28.7 TerraSAR-X: April 18 30.8 – 32.3

TerraSAR-X

RADARSAT-2

V

V

S

S

D

D

Yamaguchi 4 Decomposition

Cyan: TSXRed:R2

Surface scattering

RADARSAT-2 : May 1 36.4 – 38.0 TerraSAR-X: April 29 30.8 – 32.3

TerraSAR-X

RADARSAT-2

V

V

S

S

D

D

Yamaguchi 4 Decomposition

Cyan: TSXRed:R2

Surface scattering

Scatter Plots: April 17-18, Rangeland ROI

X axis: TerraSAR-XY axis: RADARSAT-2

Surface Volume

Dihedral

Observations

Both RADARSAT-2 and TerraSAR-X show the same general structure

TerraSAR-X is always brighter than RADARSAT-2 There is no correlation at the pixel level between images

from the two sensors

Sub-area for continued testing• Contains

• Water• Native rangeland• Irrigated cropland

RADARSAT-2 FQ8 April 17

TerraSAR-X April 18

TerraSAR-X April 29

RADARSAT-2 FQ17 May 1

TerraSAR-X May 10

TerraSAR-XApril 29

TerraSAR-XApril 29

April 18

April 29

May 10

RADARSAT-2 ROI Histograms

FQ8 April 17

FQ17 May 1

T33

T22

T11

TerraSAR-X May 10

April 18

April 29

May 10

Entropy:TerraSAR-X

FQ8 April 17

FQ17 May 1

Entropy: RADARSAT-2

April 18

April 29

May 10

Alpha: TerraSAR-X

FQ8 April 17

FQ17 May 1

Alpha: RADARSAT-2

FQ8 April 17

FQ17 May 1

April 18 April 29 May 10

More Observations

TerraSAR-X shows much higher entropy than RADARSAT-2

T22, T33 close to or at noise floor for TerraSAR-X cross-pol, double bounce contain little information

TerraSAR-X shows more volume scattering – Shorter wavelength or higher entropy?

Much greater dynamic range for RADARSAT-2

Conclusions

Qualitatively, RADARSAT-2 and TerraSAR-X produce similar imagery.

TerraSAR-X imagery is visibly noisier. Very high entropies for TerraSAR-X imagery indicate

little statistically significant polarimetric information

These results are specific to the environmental conditions– Is the environment of the prairies in early spring too subtle for

TerraSAR-X quad-pol?

A Cautionary note!

Thank You