SEE THE FORESTS WITH DIFFERENT EYES

CHUE POH TAN

ARMANDO MARINO

IAIN WOODHOUSE

SHANE CLOUDE

JUAN SUAREZ

COLIN EDWARDS

A contribution from radar polarimetry

Overview

OUTLINE Combination of sensors (RADAR + LIDAR) may

be a powerful solution

This work will treat meanly the RADAR part of the problem

Backscatter-biomass regression is problematic on mountainous terrain and sparse woodlands

Polarimetry can be our SAVIOUR !

STUDY AREA – GLEN AFFRIC

STUDY AREA – GLEN AFFRIC

FIELDWORK

FIELDWORK

ALOS PALSAR

• L-band (1.27 GHz)

• Quad-pol data

• Ascending mode

• Incidence Angle 23o

• Data acquisition 17 April 2007 & 08 June 2009

ALOS PALSAR

• X-band (9.6 GHz)

• Quad-pol data

• Ascending mode

• Incidence Angle 43o

• Data acquisition 13 April 2010

TerraSAR-X

•Ability to capture high resolution images quickly and relatively independent of lighting conditions

•Ability to collect 20000 to 75000 records per second

•High vertical accuracy with 15-20cm RMS and horizontal accuracy with 20-30cm RMS

•Data acquisition 9 June 2007

LiDAR: Optech ALTM 3100 LiDAR system

• Many studies on forest volume/ biomass estimation in flat areas(Kurvonen et al., 1999; Saatchi et al., 2001; Austin et al., 2003)

•Kurvonen L. et al., “Retrieval of Biomass in Boreal Forests from Multitempotal ERS-1 and JERS-1 SAR Images,” IEEE Transactions on Geoscience and Remote Sensing, Vol. 37, pp. 198-206, 1999.•Figure Reference: Mette, T.   Papathanassiou et al., “Forest biomass estimation using polarimetric SAR interferometry,” IEEE International Geoscience and Remote Sensing Symposium, Vol.2, pp. 817- 819, 2002.

PREVIOUS WORK

•The level of saturation has been found to vary as a function of polarization and forest types at L-band. (M. L. Imhoff ,1995;Luckman ,1998;L. Kurvonen,1999)•Studies have been done on biomass estimation using backscatter-biomass relationship (Kasischke et al., 1995; Ranson et al.,1996; Mitchard et. al)

Figure Reference: http://www.eci.ox.ac.uk/research/biodiversity/linkcarbon.php Mitchard, E.T.A., S.S. Saatchi, I.H. Woodhouse, T.R. Feldpausch, S.L. Lewis, B. Sonké, C. Rowland, & P. Meir. In press. Measuring biomass changes due to woody encroachment and deforestation/degradation in a forest-savanna boundary region of central Africa using multi-temporal L-band radar backscatter. In press, Remote Sensing of Environment.)

REGRESSION ANALYSIS – HV

HV=-14.2151+1.8251*(1-exp(0.0488*AGB)), R2=0.36

REGRESSION ANALYSIS – HV

Methodology

METHODOLOGY

STUDY AREA

Aerial Photograph DEM

1500m

1500

mArea

Google Earth

LIDAR DATA PROCESSING

CSM

Canopy

Heig

ht M

odel

CHM

Canopy

Surfa

ce M

odel

LIDAR

LIDAR DATA PROCESSING

Height derived from LiDAR

Biomass LiDAR

Canopy Cover derived from LiDAR

LIDAR

Polarisation: Scatterers have different polarisation behaviour

M-POL 19/35

Radar Polarimetric Observations

Cylin

der:

An

isotr

op

ic Ground:

No depolarisationGround:No depolarisation

Sphere:isotropic

+ …

Cylin

der:

An

isotr

op

ic

Ground:No depolarisation

Ground:No depolarisation

Sphere:isotropic

+ …

RADAR

RADAR POLARIMETRYAZIMUTH SLOPE REMOVAL

SRR =(SHH-SVV+ i SHV)/2SLL =(SVV-SHH+ i SHV)/2

θ=[Arg(< SRRSLL *>) + π ]/4]

For θ> π /4, θ= θ - π /2

22

*1

4

Re4tan

4

1

HVVVHH

HVVVHH

SSS

SSS

]][][[][ Tnew UTUT

2cos2sin0

2sin2cos0

001

U

Ref: Lee, J. S., Shuler, D. & Ainsworth, T. (2000). Polarimetric SAR Data Compensation for Terrain Azimuth Slope Variation, IEEE Transactions on Geoscience and Remote Sensing, Vol. 38(5), pp. 2153-2163.

Circular Polarisation Method

RADAR

RADAR POLARIMETRY YAMAGUCHI MODEL

• Model-based Decomposition• Satisfy non reflection symmetric case <SHHSHV

*>≠0 and <SHVSVV*>≠0

helixcvolumevdoubledsurfaces TfTfTfTfT

10

10

000

2100

010

002

4000

01

0

000

0

01*

2

2

*

j

jff

ffT cvds

- Processed using POLSARpro, courtesy of ESA -

RADAR

YAMAGUCHI MODEL

)1(2 ss fP

vv fP )1(2 dd fP

Volume Scattering (Pv) Double Bounce Scattering (Pd)Surface Scattering (Ps)

HVch fP

Helix Scattering (Ph)Google aerial photograph

RADAR

Results

Biomass

Regression Analysis

Biomass fieldwork

Pv/ Ps

BIOMASS REGRESSION ALGORITHM

R2 =0.74

AGB

S

V eP

P 0104.015363.03604.0

Yam

aguc

hi V

ol/O

dd

AGB: Above Ground Biomass

BIOMASS ESTIMATION: DENSE FOREST

Aerial Photograph Canopy Cover derived from LiDAR Biomass derived from LiDAR

DEM Height derived from LiDAR Biomass derived from ALOS PALSAR

BIOMASS ESTIMATION: SPARSE AREA

Aerial Photograph Canopy Cover derived from LiDAR Biomass derived from LiDAR

Height derived from LiDAR Biomass derived from ALOS PALSARDEM

BIOMASS ESTIMATION: SHADOW EFFECT

Aerial Photograph Canopy Cover derived from LiDAR Biomass derived from LiDAR

Height derived from LiDAR Biomass derived from ALOS PALSARDEM

TerraSAR-X ALOS PALSAR

R2 =0.74

Regression Analysis : TerraSAR-X

AGB

S

V eP

P 0104.015363.03604.0 AGB

S

V eP

P 0158.012605.04498.0

R2 =0.56

TerraSAR-X ALOS PALSAR

10 20 30 40 50

5

10

15

20

25

30

35

40

45

50 0

50

100

150

•Data acquisition: 13 April 2010 08 June 2009

BIOMASS ESTIMATION

CONCLUSION• ALOS PALSAR has the potential for biomass

estimation over wide coverage area but it is limited to biomass less that 150t/ha

• TerraSAR-X can be used to detect the existence of vegetation.

• LiDAR is able to estimate biomass at stand level over small coverage footprint

• It can be extended to be incorporated into the existing measurement models to estimate the biomass and to compare it with other remote sensing data.

HUNGRY

FUN !

SLOPY

Rain!

GHOST?

SPIKY No

Signal

THANK YOU QUESTIONS?