PowerPointPrsentation - PowerPoint PPT Presentation

1 / 21
About This Presentation
Title:

PowerPointPrsentation

Description:

Oliver Cartus (1), Christiane Schmullius (1) Maurizio Santoro (2), Pang ... (1) Department of Earth Observation, Friedrich-Schiller University. Jena, Germany ... – PowerPoint PPT presentation

Number of Views:57
Avg rating:3.0/5.0
Slides: 22
Provided by: c1s4
Category:

less

Transcript and Presenter's Notes

Title: PowerPointPrsentation


1
CREATION OF LARGE AREA FOREST BIOMASS MAPS FOR NE
CHINA USING ERS-1/2 TANDEM COHERENCE
Oliver Cartus (1), Christiane Schmullius (1)
Maurizio Santoro (2), Pang Yong (3), Li Zengyuan
(3)
(1) Department of Earth Observation,
Friedrich-Schiller University Jena, Germany
(3) Chinese Academy of Forestry, Institute of
Forest Resource Information Technique Beijing,
China
(2) Gamma Remote Sensing Gümligen, Switzerland
2
Background
The ERS-1/2 tandem mission has created a huge
interferometric dataset (1995-2000) It is known
that ERS-1/2 tandem coherence can be used for
biomass estimation in boreal forest with high
accuracy
for small managed test sites
Kättböle, Sweden, RMSE 21 m3/ha Conclusion
multi-temporal winter coherence data is most
suitable (Santoro et al, 2002)
Coherence depends on meteorological and
environmental conditions ? The behaviour of
coherence found in a small test site cannot be
transferred to large areas automatically
3
Background
Coherence - stem volume relationship strongly
varies with meteorological and environmental
conditions
4
Background
It could be shown that ERS-1/2 tandem coherence
can be used for biomass estimation in boreal
forest at large scale
with an ERS-1/2 tandem dataset acquired only in
fall and with a narrow range of baselines
Histogram-based training of an empirical model,
which relates coherence to stem volume, could be
done
SIBERIA Project Central Siberia (Wagner et al.,
2003) Area covered 1.000.000 km2 Accuracy gt
90
Method cannot be used for multi-seasonal
multi-baseline data
5
Overview
  • Data Overview of test sites and ERS-1/2
    coherence imagery
  • Coherence measurements at the test sites
  • Coherence modelling
  • Model training A new VCF-based model training
    procedure
  • Regression-based vs. VCF-based training
    procedure
  • Classification Accuracy
  • Application of the new approach for Northeast
    China

6
Forest inventory data
For each stand measurements of Stem volume
m3/ha Height, DBH, dominant Species, Relative
Stocking RS are
available.
7
ERS-1/2 tandem data
Processing Co-registration, 2x10 multi-looking,
common-band filtering, adaptive coherence
estimation (3x3 to 9x9), Geo-coding using the
SRTM-C DEM, Pixel size 50x50 m
ERS-1/2 Mosaic R Coherence G Sigma nought
(ERS-1) B Sigma nought ratio
223 coherence scenes Baselines 0 - 400 m
8
Coherence measurements at the test sites
r -0.746
r -0.895
RS gt 50 Area gt 3 ha
r -0.678
r -0.746
RS gt 30 Area gt 3 ha
(Santoro et al. 2007)
9
Interferometric Water Cloud Model
ground coherence ? temporal decorrelation canopy
coherence ? temporal and volume decorrelation
Forest coherence is the sum of
Ground contribution
Vegetation contribution
  • ?gr and ?0gr represent ground temporal coherence
    and backscatter
  • ?veg and ?0veg represent vegetation temporal
    coherence and backscatter
  • ? is related to the forest transmissivity
    (0.003 - 0.007 for ERS)
  • Volume decorrelation related to
  • h, Height ? allometric equation to express it
    as a function of stem volume
  • Bn, perpendicular baseline
  • a, two-way tree attenuation ? 1 2 dB/m
    depending on season (Askne
    et al. 1997)

10
Question How to calculate the unknowns of the
model for each frame without ground-truth data?
11
Model training based on VCF
12
Temporal decorrelation
13
Forest transmissivity ß
Regression-based estimation of all 5 unknowns
14
Regression- vs. VCF-based model training
Dashed line- regression Solid line - VCF
15
Variability of coherence within frames
Variability of ground coherence
Variability of coherence of dense
canopies
Sandy soils, Peat soils
16
Variability of coherence within frames
Training for the whole frame
Restricted
17
Stem volume retrieval
gt3ha
gt 6ha
18
Classification accuracy
Classes according to the SIBERIA
map 0-20,20-50,50-80,gt80 m3/ha
Green VCF-based training Red Regression-based
training
19
Forest Map of Northeast China
20
Conclusions
  • The new VCF-based classification approach is a
    fast and easy to apply method to map forest stem
    volume
  • Weak points
  • 1) Low accuracy of intermediate classes
    (20-50,50-80 m3/ha)
  • ? multi-temporal combination of results
    obtained from winter
    coherence images unfortunately not
    possible with the ERS dataset available
  • 2) Siberian boreal forest Chinese
    cold-temperate forests Are there differences in
    coherence?

21
Topography
Increasing influence of spatial decorrelation for
longer baselines
Topographic modification of temporal
decorrelation (wind field?) of dense forests
Write a Comment
User Comments (0)
About PowerShow.com