Monitoring Forest Management Activities using Airborne LiDAR and ALOS PALSAR
Akira Kato1, Manabu Watanabe2, Tatsuaki, Kobayashi1, Yoshio Yamaguchi3,and Joji Iisaka4
1Graduate School of Horticulture, Chiba University, Japan2Center for Northeast Asian Studies, Tohoku University, Japan
3Graduate School of Science & Technology, Niigata University,, Japan
4Department of Geography, University of Victoria, Canada
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 - Local acquisition - 10cm ~ resolution → single tree level estimation
Problem study frame⇒ ALOS PALSAR ⇔ limited field samplesBottom-up approach
State Level: Biomass change is monitored using PALSARas same quality as global scale.
District Level: Biomass change is monitored using Airborne LiDARStand Level: Biomass change is monitored using Airborne or terrestorial LiDAR
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.
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.
Study frame forest management activities⇒
2009 Summer
ALOS PALSAR data before thinning
The first airborne LiDAR acquisiton
Discrete samplesfield work
- measure trees.
Continuous samplesmodeling
Wider scalebiomass change
Ground Truth
2010 Summer
ALOS PALSAR data after thinning
The second airborne LiDAR acquisiton
2009 & 2010 Winter We thinned trees.
Terrestrial LiDAR (after thinning)
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 km2
- Dominant species is Japanese cedar (Cryptomeria japonica)-Homogeneous stands - 30 plots (20m x 20m) were set
Data – Airborne LiDARAcquisition date 1st Aug. 14th, 2009
2nd July 18th, 2010 Laser sensor Riegl LMS-Q560 Laser wavelength 1,550 nm
(Near infrared red ) Average laser point 20 points/m2
HH HV
Before thinning After thinning
Data – ALOS PALSARMode Pass Weather Acquisition dateFBD 405 Cloud 2009/7/1 13:08FBD 404 Sunny 2009/7/30 13:06FBD 404 Sunny 2009/9/14 13:07FBD 405 Sunny 2009/10/1 13:09FBD 404 Sunny 2010/6/17 13:05FBD 405 Sunny 2010/7/4 13:07FBD 404 Sunny 2010/9/17 13:04FBD 405 Sunny 2010/10/4 13:06FBD 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.
Backscattering coefficient- σ0 (dB, amplitude value)
Preprocessing – ALOS PALSAR1.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
Preprocessing – Airborne LiDARDigital Terrain Model Digital Canopy Model
⇒Tree Top locationDigital Surface Model
Preprocessing
DTM (50cm) DSM (50cm)
2010 DCM (50cm) Thinned area ⇒ white
Methodology – Identify Tree TopsStem height and location have been identified
by 0,;02 yyxxyyxxxy fffff
xx
yyxxxyxy
f
ffff
2
tan
Second order Taylor’s approximation
yyxyxx fyyfyyxxfxxyxfyxf 2000
2000 ))(2/1())(())(2/1(),(),(
~
sin)( 0 rxx cos)( 0 ryy
(Bloomenthal et al., 1997)
Tree top location and height
Before Thinning (Aug 2009) After Thinning (July 2010)
0 75 150 225 30037.5メートルm
MethodologyBiomass estimationBiomass = (stem volume = f (tree height, dbh)) × (density factor) ×(expansion factor of branch) ×(expansion factor of stem) Stem volume = α (stem density) + β (tree height) +
C
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
Results and Discussion V = 20.94 log(N) + 82.94 log(H) - 113.10
m m
Stem Volume Change (m3)
m
High: 137.03
Low: -116.04
HH
HV
Results and DiscussionALOS PALSAR
HV/HH is shifted in 9.8 degrees
X-axis: HH backscattering coefficients (σ0, dB)
Y-a
xis:
HV
bac
ksca
tter
ing
coef
fici
ents
(σ 0,
dB)
Before Thinning After Thinning
The axis is rotated towards right (when trees are thinned)
Future consideration1. Full polarization data should be utilized for the biomass change
analysis.
⇒ averaging speckle filtering requires data accumulation.
interferometric 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.
FAO global woody biomass map
Future StudyVolume Scattering ⇒ Canopy Condition
Wrapping method - Kato et al., (2009) Remote Sensing of Environment 113 : 1148-1162
Field measured crown volume (m3)
Cro
wn
volu
me f
rom
wra
pp
ing
meth
od
(m3)
Quantifying the thickness of canopy from crown volume derived by the wrapping method
Green: Low density stands Blue: High density stands
Thank you very much.Any questions?
Contact:Dr. Akira Kato
Acknowledgement This research was supported by the Environment Research and
Technology Development Fund (RF-1006) of the Ministry of the Environment, Japan.