Symposium presentation

The ALOS Kyoto & Carbon Initiative
An international science collaboration
led by JAXA EORC
Ake Rosenqvist
EU Joint Research Centre
Inst. for Environment and Sustainability
2nd ALOS PI Symposium,
Rhodes, Greece, Nov. 3-7, 2008
Heritage and objectives
The Kyoto & Carbon Initiative is an international collaborative project
initiated by JAXA EORC in 2002, forming the continuation of the JERS-1
SAR GRFM/GBFM project into the ALOS PALSAR era.
Aims to support information needs posed by the “3 C’s”:
• The terrestrial Carbon cycle science community (CO2 & CH4
sources and sinks);
* Multinational Environmental Conventions and Declarations:
UNFCCC Kyoto Protocol & Post-Kyoto/REDD (Forest and Land Cover Change);
Ramsar Convention (wetland characteristics and disturbances);
* Environmental Conservation
Project Management
JAXA EORC / JRC
Forest & CO2
.
Theme Coordination:
Univ. Wales / U.K.
Product Leader 1
Product Team 1
Product Leader 2
Product Team 2
...
Wetlands & CH4
Theme Coordination:
UCSB/USA
ERISS/Australia
Desert & Water
.
Theme Coordination:
U. Bordeaux/France
Mosaic Products Theme
.
Theme Coordination:
Jet Propulsion Lab./USA
K&C Science Team:
JAXA EORC (Japan); JRC (E.U.); INPE (Brazil); JPL (USA);
DLR (Germany); Int’l Wetlands Management Inst (Ethiopia);
CESBIO (France); FSU-Jena (Germany); U-Sheffield (U.K.);
SLU (Sweden); U-Wales Aberystwyth (U.K.); UCSB (USA);
Borneo Orangutang Salvation Foundation (Indonesia);
U-Victoria (Canada); UNSW (Australia); U-Bordeaux (France);
AGS (USA); sarmap (Switzerland); ERISS (Australia)
(~25 active members / 13 countries)
The Forest Theme
Support to the Kyoto Protocol
and Terrestrial Carbon Cycle Science
Data
Development of systematic observation systems and data archives
(Kyoto Protocol Art. 5, 10d) ( --> ALOS observation strategy)
Derived information
Detection and spatial quantification of annual changes in forest- and land cover
(KP Art. 3.3, 3.4, 4, 12)
Incremental changes in [regenerating] above-ground biomass
(KP Art. 3.3, 3.4, 4, 12);
The Wetlands Theme
Support to the Ramsar Convention on Wetlands*
Ramsar Convention information requirements (conservation focus)
(1) Spatial and temporal characteristics of flooding patterns in Ramsar
designated (and other) wetland areas
Ð Spatial extent;
Ð Temporal cycle (seasonal/annual/decadal…);
(2) Identification of natural- and human-induced disturbances in wetlands;
(3) Support to the Ramsar global wetlands inventory.
Carbon Cycle (wetlands as sources of CH4)
(4) Monitoring of anthropogenic and natural sources of CH4:
Time schedule
1. Methodology development - Year 1-3 (2006-2008)
• Development of algorithms and methods required to support the specific
information requirements identified (e.g. annual biomass change, wetland
flooding extents, irrigated rice spatial distribution, etc.);
2. Regional “prototype” demonstration - Year 1-3 (2006-2008)
• Operational demonstration of the methodology to a “moderately large”
geographical region, covering several PALSAR swaths;
3. Upscaling and extension - Year 4+ (2009-2011)
• Regional/continental-scale application of the methods and work approaches
developed to other, or larger, geographical regions.
Mapping of seasonal inundation
in the Congo river basin
using PALSAR ScanSAR
Ake Rosenqvist
European Commission
Joint Research Centre
ALOS PI Symposium
Rhodes, Greece, Nov. 3-7,2008
Institute for Environment and Sustainability
Global Vegetation Monitoring Unit
ScanSAR use for inundation monitoring
• ScanSAR is being used
for intensive monitoring of
seasonal inundation
dynamics in major
wetlands and river basins
• Every-cycle (46-days)
monitoring during one or
more seasonal cycles
ScanSAR Wetlands
Descending
7
6
5
West Siberia
Lena Delta
Volga Delta
Amur
East Asia paddy
India paddy
Mainland SE-Asia
Insular SE-Asia
PNG
North Australia
Pripet-Biebrza
Niber Basin
Congo Basin
Okavango-Mozambique
ASF mask
BOREAS SSA
Quebec-Everglades
Amazon Basin
Pantanal
2006
2005
8
6
9
1 0 1 1 1 2 1
7
8
9
2
10
3
4
11
5
12
6
2007
7
13
8
14
9
1 0 1 1 1 2 1
15
16
17
2
18
3
4
19
5
20
6
2008
7
21
8
22
9
1 0 1 1 1 2 1
23
24
25
2
26
3
4
27
5
28
6
7
29
Institute for Environment and Sustainability
Global Vegetation Monitoring Unit
8
30
9
1 0 1 1 1 2 1
31
32
33
Congo river basin
• Second largest watershed in the world ~3.8 million km2
• The Congo crosses the equator twice -> complex seasonal inundation patterns
alternating seasonality in N and S hemisphere
• Seasonal flooding a significant source of non-anthropogenic GHG emissions,
BUT magnitude and variability are unknown due to the lack of adequate
measurement techniques to map below-canopy flooding.
_____________________________________________________________________________________
JERS-1 SAR 1996
Institute
for
Global Rain Forest Mapping
(GRFM) project
2006-11-04
Environment and Sustainability
Global Vegetation Monitoring Unit
Objectives
Use ALOS PALSAR data mode to map and monitor the spatial extent and temporal
dynamics of flooding in the Congo Basin, using PALSAR ScanSAR (WB1) path data,
processed by JAXA EORC.
Initial analysis and prototype classification undertaken on a time series of standard
product ScanSAR data (L 1.5) covering the central part of the basin.
7 out of 8 possible observations, acquired every 46-days between July 2007 and
June 2008 (May 2008 missing).
271/3600
PALSAR ScanSAR
2006-11-04
Institute for Environment and Sustainability
Global Vegetation Monitoring Unit
Central Congo river basin - PALSAR ScanSAR
2007-07-18
Institute for 271/3600
Environment and Sustainability
Global Vegetation Monitoring Unit
Central Congo river basin - PALSAR ScanSAR
2007-07-18
2007-09-02
Institute for 271/3600
Environment and Sustainability
Global Vegetation Monitoring Unit
Central Congo river basin - PALSAR ScanSAR
2007-07-18
2007-09-02
2007-10-18
Institute for 271/3600
Environment and Sustainability
Global Vegetation Monitoring Unit
Central Congo river basin - PALSAR ScanSAR
2007-07-18
2007-09-02
2007-10-18
2007-12-03
Institute for 271/3600
Environment and Sustainability
Global Vegetation Monitoring Unit
Central Congo river basin - PALSAR ScanSAR
2007-07-18
2007-09-02
2007-10-18
2007-12-03
2008-01-18
Institute for 271/3600
Environment and Sustainability
Global Vegetation Monitoring Unit
Central Congo river basin - PALSAR ScanSAR
2007-07-18
2007-09-02
2007-10-18
2007-12-03
2008-01-18
2008-03-04
Institute for 271/3600
Environment and Sustainability
Global Vegetation Monitoring Unit
Central Congo river basin - PALSAR ScanSAR
2007-07-18
2007-09-02
2007-10-18
2007-12-03
2008-01-18
2008-03-04
2008-06-04
Institute for 271/3600
Environment and Sustainability
Global Vegetation Monitoring Unit
Simple approach
• Radiometric normalisation of the ScanSAR scenes
• Classification of flooded vs. non-flooded forest
2006-11-04
Institute for Environment and Sustainability
Global Vegetation Monitoring Unit
Radiometric calibration
Homogeneous non-flooded
(terra firme) forest used to
assess absolute (between
scenes) and relative (scanto-scan) radiometric
calibration.
20 reference areas
(~3800 pixels) selected
between near and far range
(3 points/scan + overlap)
421/3650
#5
#4
#3
#2
Difficult to identify stable
terra-firme forest reference
areas
Scan #1
2006-11-04
Institute for Environment and Sustainability
Global Vegetation Monitoring Unit
Absolute & relative radiometry
Obs. date
20070115
20070718
20070902
20071018
20071203
20080118
20080304
20080604
Tot mean
Tot st.dev
WB1 271/3600, Congo basin
Mean (forest)
St.dev
-7.13
0.39
-7.09
0.42
-6.87
0.45
-6.50
0.50
-6.27
0.51
-7.12
0.27
-7.29
0.34
-7.29
0.48
-6.95
0.42
0.37
271/3600 (Congo Basin)
Δ bias
-0.18
-0.15
0.07
0.44
0.67
-0.17
-0.34
-0.34
Proc. date
20080313
20071204
20071204
20071204
20071204
20080225
20080313
20080609
Estimated absolute accuracy
Total mean (all scenes) = -6.95 dB
St.devWB1 = 0.37 dB
<Max-min> = 1.01 dB
Relative (within-scene) accuracy
0.27 ~ 0.51 dB
-(Δ bias): Normalisation factor
applied to each scene
Institute for Environment and Sustainability
Global Vegetation Monitoring Unit
Scan-to-scan radiometry
>0.1 dB over mean
within +/- 0.1 dB of mean
<-0.1 dB over mean
• 0.6 ~ 1.9 dB difference between scans.
• Scan 1 consistently darker in all scenes
• Far range trail-off effect in scan 5 in all
scenes
• Scan overlaps zones darker
Institute for Environment and Sustainability
Global Vegetation Monitoring Unit
Preliminary classification
Preliminary classification of flooded vs. non-flooded forest was undertaken by a
simple threshold method, applied on each of the normalised scenes
Since no in-situ data over the Congo were available, a threshold value of -4.6 dB was
used - verified valid for flooded forest in the Amazon in JERS-1 SAR data. Solitary
pixels were removed by a 3x3 median filter.
Stack, add and colourcode.
2006-11-04
Institute for Environment and Sustainability
Global Vegetation Monitoring Unit
Flood duration map (prototype)
Flood duration
1 – 45 days
46 – 91 days
92 - 137 days
138 – 183 days
184 – 229 days
Temporal resolution:
1 satellite cycle (46 days)
Institute for Environment and Sustainability
Global Vegetation Monitoring Unit
Flood duration map (prototype)
Flood duration
1 – 45 days
46 – 91 days
92 - 137 days
138 – 183 days
184 – 229 days
Institute for Environment and Sustainability
Global Vegetation Monitoring Unit
Results
First trial of flood duration mapping in the Congo Basin undertaken.
Due to the lack of ground truth information on flooding in the Congo Basin, it
is acknowledged that the threshold method yields large uncertainties in the
area estimations which cannot be verified, and that a more sophisticated
classification method - possibly based on backscatter changes – therefore
may need to be applied.
The prototype flood duration map to the right nevertheless illustrates the
unique potential of using PALSAR ScanSAR data for flood duration
mapping:
.
* L-band HH-polarisation
.
* Flooded forest is detectable across the entire 360 km image range (18-43
deg)
* Systematic observation strategy implemented for ALOS PALSAR,
observing globally significant wetlands every 46-day cycle.
2006-11-04
Institute for Environment and Sustainability
Global Vegetation Monitoring Unit
Thank you
Institute for Environment and Sustainability
Global Vegetation Monitoring Unit