Monitoring Forest Management Activities using Airborne - PowerPoint Presentation

Slide1

Monitoring Forest Management Activities using Airborne LiDAR and ALOS PALSAR

Akira Kato

1

, Manabu Watanabe

2

,

Tatsuaki

, Kobayashi

1

,

Yoshio Yamaguchi

3

,and Joji Iisaka

4

1

Graduate School of Horticulture, Chiba University, Japan

2

Center for Northeast Asian Studies, Tohoku University, Japan

3

Graduate School of Science & Technology, Niigata University,, Japan

4

Department of Geography, University of Victoria, CanadaSlide2

ALOS PALSAR ⇔Airborne LiDAR

ALOS PALSAR

- L-band radar

→　

polarization

(

indirect measurement

)

- Multi-temporal data

-

Low

cost

-

Global

acquisition

-

15m

～

resolution

→　

plot level estimation

Airborne

LiDAR

- Near-infrared red laser

→　

direct measurement

- (Multi-) temporal data

-

High

cost

-

Loca

l acquisition

-

1

0cm

～　

resolution

→

single

tree level estimationSlide3

Problem ⇒ study frame

ALOS PALSAR

⇔

limited field samples

Bottom-up approach

State Level:

Biomass

change is monitored using

PALSAR

a

s same quality as global scale.

District Level:

Biomass

change is monitored using

Airborne

LiDAR

Stand Level:

Biomass

change is monitored using

Airborne or

terrestorial

LiDARSlide4

Forest Biomass ⇔ Volume Scattering

Past studies

1.

Saturation level of forest biomass using L-band

100 ton/ha in homogeneous pine forest

(

Imhoff et al.

, 1995)

⇒　Approx.

5 meters spacing of 20 m height trees.

40

ton/ha in broadleaf evergreen forest (Lucas

et al., 2006)

2. HV polarization is higher correlation with forest biomass (Lucas et al.

, 2006) ALOS PALSAR is a good sensor to detect the forest management activities, but correlation between backscattering coefficient and the change is still unknown. Slide5

Volume Scattering ⇔stand condition

Stand condition

is defined by

-

stem

density - tree height

- tree forms (the shape of tree crown) - tree age

⇒ airborne

LiDAR is used to bridge between field measurement and backscattering coefficient of ALOS PALSAR as the ground truth

. Slide6

Study frame

⇒

forest management activities

２００９

Summer

ALOS PALSAR data before thinning

The first airborne

LiDAR

acquisiton

Discrete samples

f

ield work

- measure trees.

Continuous samples

modeling

Wider

scale

b

iomass change

Ground Truth

２０１０

Summer

ALOS PALSAR data after thinning

The second airborne

LiDAR

acquisiton

２００９

&

２０１０

Ｗｉｎｔｅｒ

W

e thinned trees.Slide7

Terrestrial LiDAR (after thinning)Slide8

Study AreaSanmu

City, Chiba Prefecture, JAPAN

→

Commercial timber production area

Name

Number

d.b.h(cm)

Tree Height(m)

Cryptomeria

japonica

718

10.3～69.7

10.4～34.3

Chamaecyparis

obtuse

179

8～72

2.9～31.8

Chamaecyparis

pisifera

38

17.1～90.7

14.6～34.9

Quercus

myrsinaefolia

9

4～67.7

5.9～29.8

Research area is around 9 km

2

- Dominant

species is Japanese cedar

(

Cryptomeria

japonica

)

Homogeneous

stands

- 30 plots (20m x 20m) were setSlide9

Data – Airborne LiDAR

Acquisition date

1

st

Aug

. 14

th

, 2009

2

nd

July

18

th

, 2010

Laser sensor

Riegl LMS-Q560

Laser wavelength

1,550 nm

(Near infrared red )

Average laser point

20 points/m

2

HH

HV

Before thinning

After thinningSlide10

Data – ALOS PALSAR

Mode

Pass

Weather

Acquisition date

FBD

405

Cloud

2009/7/1 13:08

FBD

404

Sunny

2009/7/30 13:06

FBD

404

Sunny

2009/9/14 13:07

FBD

405

Sunny

2009/10/1 13:09

FBD

404

Sunny

2010/6/17 13:05

FBD

405

Sunny

2010/7/4 13:07

FBD

404

Sunny

2010/9/17 13:04

FBD

405

Sunny

2010/10/4 13:06

FBD

405

Cloud

2010/11/19 13:05

L-band FBD (Fine beam Double Polarization)

Resolution:

20m

Before

thinning

After

thinning

HH

HV

ALOS satellite ended

at May 2011.

- 20 m resolution L-band SAR.

- 46 days observation cycle.

ALOS 2 will be launched at 2013.

1

～

3 m resolution L-band SAR.

16 days observation cycle.

B

ackscattering coefficient

-

σ

0

(dB,

amplitude value) Slide11

Preprocessing – ALOS PALSAR

１．

Geometric and terrain correction

⇒

MapReady

(Alaska Satellite Facility, ver

2.3, 2010). 2. layover / shadow regions for the terrain correction

　　⇒

5m resolution DEM provided by Geospatial Information Authority of Japan

3. Speckle filtering

⇒Averaging the values of multi-temporal data. The data before thinning (before August 2010) and after thinning (after August 2010) are averaged separately.

4. Pixel alignment

⇒Manual geo-referencing was applied to match the images with less than half pixel of error (10m) among the multi-temporal data Slide12

Preprocessing – Airborne LiDAR

Digital Terrain Model

Digital

Canopy Model

⇒

Tree

Top location

Digital

Surface

ModelSlide13

Preprocessing

DTM (50cm)

DSM (50cm)

2010 DCM

(50cm)

Thinned area

⇒

whiteSlide14

Methodology – Identify Tree Tops

Stem

height and location have

been

identified by

Second order Taylor’s approximation

(

Bloomenthal

et al.

,

1997)Slide15

Tree top location

and height

Before Thinning (

Aug 2009)

After Thinning (

July 2010)

mSlide16

Methodology

Biomass estimation

Biomass = (

stem

volume =

f

(tree height,

dbh

))

× (density factor) ×(expansion factor of branch)

×(expansion factor of stem) Stem volume = α

(stem density) + β (tree height

) + CSlide17

Results and DiscussionAirborne

LiDAR

Stem density Tree height

Stem density correction:

y

= 2.5034x -

12.41

where x: the number of stems derived from airborne

lidar

y: the corrected number of stems Slide18

Results and Discussion

V

= 20.94 log(N) + 82.94 log(H) - 113.10

m

mSlide19

Stem

V

olume Change (m

3

)

m

H

igh: 137.03

Low

: -116.04

HH

HVSlide20

Results and Discussion

ALOS PALSAR

HV/HH

is

shifted

in

9.8 degrees

X-axis: HH backscattering coefficients (σ

0

, dB)

Y-axis: HV backscattering

coefficients

(σ

0

, dB)

Before Thinning After Thinning

The axis is rotated towards right (when trees are thinned) Slide21

Future consideration

1. Full polarization data should be utilized for the biomass change analysis.

⇒

averaging speckle filtering requires

data accumulation

.

i

nterferometric analysis needs

the shorter observation cycle.

2. Full polarization

interferometry analysis can raise the saturation level (more than 100 ton / ha).

⇒　registration among multi-temporal images should be accurate enough

.3. World biomass map shows the limitation to use the backscattering coefficient for the biomass stock, but the biomass change can be monitored.Slide22

FAO global woody biomass mapSlide23

Future StudyVolume Scattering ⇒

Canopy Condition

Wrapping method - Kato

et al.

, (2009)

Remote Sensing of Environment

113 : 1148-1162

Field measured crown volume (m

3

)

C

rown volume from

wrapping method(m

3

)

Quantifying the thickness of canopy from crown volume derived

by the wrapping method Green: Low density stands Blue: High density standsSlide24

Thank you very much.Any questions?

Contact:

Dr. Akira Kato

akiran@faculty.chiba-u.jp

Acknowledgement

This research was supported by the Environment Research and Technology Development Fund (RF-1006) of the Ministry of the Environment, Japan.