Classification of Typhoon-Destroyed Forests Based on Tree Height Change Detection Using

InSAR Technology

Haipeng Wang 1, Kazuo Ouchi 2, and Ya-Qiu Jin1

1. Key Laboratory of Wave Scattering and Remote Sensing

Fudan University, P.R. China.

2. Department of Computer Science

National Defense Academy of Japan, Japan.

• The final goal is to establish methodology to estimate the parameters of forests from high-resolution SAR data.

• This first study we carried out is to quantify the relation between high-resolution polarimetric SAR data and tree biomass of forests.

• This second one is to extract the information of typhoon-damaged forests. • The methodology is to utilize the texture information, polarimetric analysis and interferometric technique.

• The test site is the Tomakomai forests in Hokkaido, Japan.

• The SAR data were collected using the airborne Pi-SAR.

Purpose of This Study

Tomakomai National Forests

Shikotsu Lake

Larch Todo Fir Red Pine Spruce

Pi-SAR

Pi-SAR (Polarimetric interferometric - SAR) is an airborne SAR developed jointlyby NICT (National Institute of Information and Communications Technology) and JAXA (Japan Aerospace Exploration Agency).

It is equipped with two X-band antennas (frequency 9.55 GHz, wavelength 3.14 cm)and a L-band antenna operating at frequency 1.27 GHz (wavelength 23.6 cm).

The nominal resolution is 1.5 x 1.5 m at X-band, and 3 x 3 m at L-band for 4-lookazimuth multilooking.

L-band antenna

X-band main antenna

X-band sub-antenna

Pi-SAR Data

JAXA/NICT

L-band colour composite polarimetric Pi-SAR image

azimuth

range

study area

Shikotsu Lake

Tomakomai City

Tarumae Mt.

study area

Data acquired:7th, November, 2002

Data acquired:3rd, November, 2004

azimuth

range

Ground-Truth Biomass Data Collection

Field measurements were made in 2002, 2003, 2005, and in 2006.

Tree species, height, DBH (Diameter at Breast Height), basal area, soil moisture were measured within a 20 x 20 m sample area in each stands, and converted to above-ground biomass using the conversion formula provided by Project department (Stand volume table -East Japan-, Japan Forestry Investigation Committee. Tokyo: Forestry Agency, Oct. 1998)

Typhoon Songda

Typhoon Songda (Japanese No.18)Duration:August 28 – September 8

Typhoon Songda killed 20 people and injured 700 others in Japan. In addition, 15 crew members of a vessel were reported missing.

Damages from the storm amounted to $7.17 billion (2004 USD)

Time arrived at Tomakomai: September 7th, 2004

Fallen Trees After Typhoon

Ground-Truth Data

Amplitude Analysis

Image No. HH HV VV

L6407 67.1119 64.3982 66.1515

L8104 66.6437 64.4475 66.7929

L6407-L8104 0.4682 -0.0493 -0.6414

Site No. HH subtraction HV subtraction VV subtraction

196 0.8403 0.6001 -0.2851

197 0.9535 0.5802 -0.2825

198 0.8714 0.3279 -0.5437

217&218 0.8619 0.4444 -0.2188

243 0.8975 0.4549 -0.2435

245 0.9827 0.3009 -0.2797

267 0.9856 0.4899 -0.0584

271 0.8962 0.4392 -0.5062

300 0.8269 0.4759 -0.0099

302 0.7889 0.3169 -0.2331

268 0.6641 0.0218 -0.4429

303 0.5009 -0.0223 -0.6568

Test sites

Whole image

Amplitude Analysis Results

Accuracy:64.1%

Scattering Mechanisms From Forests

scattering from crown parts surface scattering

from ground

multiple reflection between ground and tree trunks/branches

multiple/volume scattering from branches

X/C-bandsL/P-bands

Three-Component Decomposition Analysis

Site No. Double subtraction Volume subtraction Surface subtraction

196 14.1248 1.1277 -22.1597

197 38.2798 1.0917 -39.4679

198 39.1910 0.4771 -34.9405

217&218 35.4332 0.7719 -32.1722

243 37.5858 0.7808 -34.0847

245 37.6906 0.4519 -25.9017

267 37.1895 0.9402 -32.4539

271 28.2527 0.8799 -27.4304

300 31.7851 0.8673 -31.1892

302 35.5875 0.4366 -27.3383

268 32.7751 -0.1122 -17.6746

303 33.6438 -0.2045 -27.5287

A. Freeman and S. L. Durden, “A three-component scattering model for polarimetric SAR data,” IEEE Trans. Geosci. Remote Sens., vol. 36, no. 3, pp. 936–973, May 1998.

Double Volume Surface

L6407 102.0232 138.2508 63.4812

L8104 74.4862 138.4537 83.7599

L6407-L8104 27.5449 -0.2026 -20.2847

Scattering Mechanism Analysis Result

Accuracy:77.7%

Pi-SAR

L-band antenna

X-band main antenna

X-band sub-antenna

Pi-SAR on Gulf Stream II X-band and L-band radomes

Two antennas within X-band main radomePartial Pol-InSAR by X-band main radome(Sub-antenna data not acquired this time)

Pi-SAR Data for Interferometric Analysis

X-band VM Pi-SAR image

Data acquired:3rd, August, 2004

Data acquired:3rd, November, 2004

M, N —— Moving window size

S —— Data

Complex Interferogram

∑ ∑∑ ∑

∑ ∑+=

−=

+=

−=

+=

−=

+=

−=

+=

−=

+=

−==Nii

Nii

Mjj

Mjjijij

Nii

Nii

Mjj

Mjjijij

Nii

Nii

Mjj

Mjjijij

SSSS

SS

*221*11

*21

Complex Degree of Coherence

Phase Image

Removal of Orbital Fringes

Courtesy of Dr. Moriyama

X7904X8104

Phase Unwrapping(X7904)

Phase Unwrapping(X8104)

Height Change

Conclusions and Future Work

Research was carried out to extract information of typhoo-damaged forests.

The accuracies of 64.1% and 77.7% were obtained for amplitude and decomposition data respectively.

InSAR processing resutls show it’s potential, but there is no quantitative results yet.

Phase unwrapping: other approaches rather than the branch cut.

Difficult to see the difference of typhoon-damaged information from the coherence data before and after the typhoon. Next step is to exam the coherence before and after the typhoon

Fusion of optical data and SAR data: QuickBird, IKONOS, Geoeye multispectrum optical data, Pi-SAR I&II data.