Barbier - CBFP-PFBC COMIFAC Brazzaville... · Barbier et al. RSE (2010) Large scale application...
Transcript of Barbier - CBFP-PFBC COMIFAC Brazzaville... · Barbier et al. RSE (2010) Large scale application...
Barbier Nicolas
IRD - UMR AMAP
ULB - FNRS
Programme Pilote de recherches: Biodiversité, Changements globaux et Santé dans les
forêts tropicales humides
• Part of a broader effort from IRD in central Africa:� Tropical forest as a focal object for interdisciplinary research ;
� Consequences of global change on biodiversity, resources and health;
� Possibly set up permanent research platforms.
• Yaoundé workshop 10/2009
�Call for partnerships…
Need to quantify canopy structure
•Valuable for :
� Carbon dynamics and forest degradation
assessment;
� Forest ecology (allometry rules, gap dynamics, etc.);
Forest-climate interaction (gaz and vapor exchange);� Forest-climate interaction (gaz and vapor exchange);
� Forest types, etc.
•Difficult to measure in the field,
� Approached via indirect estimates (DBH);
� Limited representativity.
Remote sensing methods
•Limitation of pixelwise optical and radar :
� Saturation for high biomass levels (>250 t/ha);
� Unable to detect degradation.
•High cost of airborne sensors (e.g. LiDAR).
•Potential of VHR imagery (Quickbird, Ikonos, etc.) but operational methodologies to be developed.
RCA: GeoEye pansharpened G-NIR-B
Peaks at
dominant frequencies
Fourier periodograms: quantification of image texture
10
0 m
10
0 m
Proisy et al. RSE (2007)French Guiana
2D Fourier transform textural 2D Fourier transform textural ordination (FOTO)
• 2D Fourier transform
• Radial power spectrum (r-spectrum)
= Proportions of the image variance accounted for by successive frequencybins, across azimuthal orientations.
Imagery
(metric optical)
Forest cover
Unit-windows
100 - 150 m sides
Frequencies
Fourier analysis
r-spectracomputation
Separation into
Workflow:
• Principal components of image r-spectra dataset
= Identification of main axes of textural variation in the image dataset.
Table of
r-spectra
PCA ordination
Separation into
homogeneous acquisition
groups
Global
standardisation
Partitioned
standardisation
Texture indices (PCA 1, PCA 2)
Invertedforest
parameter
(e.g. meancrown size)
Relation of Texture indices with
forest parameters for a set of
(DART) stands
10
0 m
Textural ordination: principal axes of variation
10
0 m
Couteron et al. J Appl Ecol (2005)
FOTO analysis on aerial photographs in French Guiana
Relationships with forest parameters
Couteron et al. J Appl Ecol (2005)Terra Firme forest; French Guiana
Relationships with forest parameters
No saturation!
Proisy et al. RSE (2007)
Biomass prediction using Ikonos imagery in mangrove stands; French Guiana
D1 D2 D3 D4
Q4 Q5 Q6 Q7
Q8 Q9 Q10 Q11
a
b
a b
a. Ikonos VHR.b. DART simulations
Radiative transfer modelsRadiative transfer models
• Testing the effect of acquisition conditions
• Mitigation : partitioning method
D5 D6 D7 D8
a. Ikonos VHR.b. LIDARc-d. Hillshade models
a b
c d
Extreme configurations: typical examples
a b
c d
Фs-v
= 0°; θV = 9°; θS = 59° Фs-v
= 180°; θV = 9°; θS = 59°
Hillshade effects on
LiDAR surface models
e f
Фs-v = 80°; θV = 3°; θS = 59° Фs-v
= 80°; θV = 15°; θS = 59°
Фs-v = 80°; θV = 9°; θS = 47° Фs-v = 80°; θV = 9°; θS = 71°Barbier et al. RSE (2010)
Bi-directional
texture distribution
� ‘Hotspot’ effect on texture,
� Alleviated by partitioningapproach.
-2-2-2-2
-2-2-2-2
-1-1-1-1
-1-1-1-1
-1-1-1-1 -1-1-1-1
0000
0000
00000000
11111111
1111
-1-1-1-1Sun Height = 47°
0
3
6
9
12
15
0
30
60
90
120
150
180
Sun Height = 47°
0000 00000000
0000
1111
1111 11111111
2222
2222 22222222
33333333
3333
Sun Height = 71°
0
3
6
9
12
15
0
30
60
90
120
150
180
-1-1-1-1
-1-1-1-1-1-1-1-1
0000
00000000
0000
1111
11111111
1111
22222222 2222 2222
Sun Height = 59°
0
3
6
9
12
15
0
30
60
90
120
150
180
Sun Height = 59° Sun Height = 71°180
a b c
Фs-v
θV
No
n-p
art
itio
ne
d
θS = 47° θS = 59° θS = 71°
θ = 47° θ = 59° θ = 71°
0000
0000
0000
00000000
0000
0000
0000
0000
0000
0000
0000
00000000
0000
1111
1111
Sun Height = 47°
0
3
6
9
12
15
0
30
60
90
120
150
180
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
00000000
0000
1111
1111
0000
0000
Sun Height = 59°
0
3
6
9
12
15
0
30
60
90
120
150
180
0000
00000000
0000
0000
0000
0000
Sun Height = 71°
0
3
6
9
12
15
0
30
60
90
120
150
180
d e f
Фs-v
Pa
rtitio
ne
d
θS = 47° θS = 59° θS = 71°
Barbier et al. RSE (2010)
Large scale application
maps of apparent crown size and canopy heterogeneity
Barbier et al. GEB (2010)
Amazonia
Validation: the Canopy project
• Texture analysis: The FOTO method (IRD).
• Field validation: Large forest inventory database across
Central Africa (FRM).
• Software development: Convivial interface in ArcGis
(Nev@ntropic).
Laurent Dufy (FRM)
Nicolas Bayol (FRM)
Antoine Mugnier (FRM)
Couteron Pierre (IRD-AMAP)
Proisy Christophe (IRD-AMAP)
Barbier Nicolas (IRD-AMAP, ULB-FNRS)François Marques (Nev@ntropic)
Large scale validation: test sites
FRM test sites
Possible PPR sites
Imagery
• Planet Action: Spot 5, Formosat, Kompsat
• Others: Quickbird, Ikonos, GeoEye, Orbview
FOTO
analysis
Field Data
• Geolocalized plots of 0.5 ha
• Forest concessions above 50,000 ha
• Inventoried surface areas around 1%
• All trees above 10-20 cm diameter
• Destructive biomass sampling (upcoming)
Perspectives
• Quantification of forest canopy structure.
• Valuable insights into :
�Degradation level ;
�Biomass and carbon, with no saturation ;
�Other forest structural attributes.�Other forest structural attributes.
• Large scale application and repeatable.
• Effect of acquisition conditions mitigated.
• Relative low cost.