Transcript Populating the Earth with Burned Area

THE GLOBCARBON INITIATIVE:
Populating the Earth with Burned
Area
Stephen Plummer (IGBP@ESA),
Olivier Arino (ESA),
Franck Ranera and Muriel Simon (SERCO)
Kevin Tansey (Univ. Leicester),
Luigi Boschetti (UMD),
H. Eva (JRC)
and Freddy Fierens (VITO Consortium)
QUEST, Exeter
27-28 October 2005
General User Needs
For Atmospheric Chemistry and Dynamic Global Vegetation Models:
 There is a particular need for information on vegetation amount
(ideally biomass or leaf area index), area burned, and vegetation
temporal variability.
 These should be global, in a consistent format, and all data products
should be available from one place.
 Consistency is more important than outright accuracy (within limits).
 The products should be multi-annual with 5 years being the minimum
but incorporating both average and extreme conditions e.g. El Ninõ.
 Products should come with spatial heterogeneity information ideally
at the highest available resolution.
 The spatial resolution requirements are 0.5°, 0.25° and 10km.
 The temporal resolution initially on a time step of 1 month but better
higher, possibly bi-weekly.
QUEST, Exeter
27-28 October 2005
Objectives
•
develop a service quasi-independent of the original Earth
Observation source.
•
focus on a system to estimate:
 Burned area
 fAPAR and LAI
 Vegetation growth cycle
•
cover six complete years: 1998 to 2003 (now up to 2007)
•
cover VEGETATION, ATSR-2, ENVISAT (AATSR, MERIS)
•
be applicable to existing archives and future satellite systems
•
be available at resolutions of ¼, ½ degree and 10km with
statistics
•
build on the existing research experience
QUEST, Exeter
27-28 October 2005
Burned Area - History
•
Year 2000 – two independent demonstrators of global burned
area: GLOBSCAR and GBA-2000
GBA-2000
•
Globscar
GLOBCARBON uses the experience of these and some of
the algorithms to produce a single burned area product –
multi-annually.
•
uses revised versions of algorithms
•
it has associated with it confidence information (detection
confidence from individual algorithms plus collocation with
available active fire products).
QUEST, Exeter
27-28 October 2005
Burned Area - Approach
GBA-2000
Globscar
•
Original =
algorithms
regional
•
•
GLOBCARBON = 1 global
and 2 regional algorithms
Original
=
2
global
algorithms and burn when
both agree.
•
it has associated with it
confidence
information
(detection confidence from
individual algorithms)
GLOBCARBON = each
algorithm and sub-parts
given a probability.
•
The resulting probability
determines occurrence of a
pixel
as
burned
(confidence information)
•
6
•
Results are merged to one product
•
Collocation with available active fire products
improves confidence
QUEST, Exeter
27-28 October 2005
Results – 1 km (Madagascar)
Confidence Rating Index (CRI)
July 1998
63-74%
75-87%
88-100%
QUEST, Exeter
27-28 October 2005
Results – 1 km (Madagascar)
Algorithm Detection (GLOBSCAR, GBA, Both)
July 1998
GLOBSCAR only
GBA only
Both algorithms
QUEST, Exeter
27-28 October 2005
Results – 1 km (Angola)
July 1998
Confidence Rating Index (CRI)
63-74%
75-87%
88-100%
QUEST, Exeter
27-28 October 2005
Results – 1 km (Angola)
July 1998
Algorithm Detection (GLOBSCAR, GBA, Both)
GLOBSCAR only
GBA only
Both algorithms
QUEST, Exeter
27-28 October 2005
MODIS Comparison – 1 km
GLOBCARBON
MODIS
July 2000
QUEST, Exeter
27-28 October 2005
MODIS Comparison – 1 km
GLOBCARBON
MODIS
Sept 2000
QUEST, Exeter
27-28 October 2005
Results –10 km (Mongolia)
2000
1-10
11-20
21-30
31-40
41-50
51-60
61-70
71-80
81-90
91-100
May 2000
Percentage of pixels with CRI > 80 in a 10*10km box
QUEST, Exeter
27-28 October 2005
Results –10 km (Australia)
1-10
11-20
21-30
31-40
41-50
51-60
61-70
71-80
81-90
91-100
100
90
QUEST, Exeter
27-28 October 2005
NW Australia – 10km May 99
May overlain with May Vectors
May overlain with May Vectors
QUEST, Exeter
27-28 October 2005
NW Australia 10km – May 99
May overlain with June Vectors
May overlain with June Vectors
QUEST, Exeter
27-28 October 2005
Validation - Approach

Theory:

Algorithm: Classification of the known (in
2000) performance of GLOBSCAR/GBA-2000

Statistical: sampling of all classes (strata)
to determine which locations to focus on
Data Needed: Each site characterised by
two Landsat ETM images 1 month apart
Problems:



Cost: 2 scenes of Landsat ETM for all
samples prohibitive

Data Availability: Often two scenes not
even available (1 year or less apart)

Solution: Use what we can afford +
seek national/continental databases
QUEST, Exeter
27-28 October 2005
Validation - Results
Montana, Stratum 1, R2 = 0.80
South Sudan, Stratum 1, R2 = 0.81
QUEST, Exeter
27-28 October 2005
Validation - Results
July 13 2000
Apr 28 2000
Aug 14 2000
Sept 15 2000
Montana, Stratum 1, R2 = 0.80
Dec 24 2000
Mar 30 2001
Sudan South, Stratum 1, R2 = 0.81
QUEST, Exeter
27-28 October 2005
LAI Results – 10 km
June 1999
water
0-0.5
0.6-1
1-2
2-3
3-4
4-5
5-6
6-7
7-8
8-9
9-10
> 10
no data
QUEST, Exeter
27-28 October 2005
Phenology Results – 10km
Leaf On 1998
water
1-50
51-100
101-125
126-150
151-200
201-300
no value
QUEST, Exeter
27-28 October 2005
Conclusions
• To feed in to the Global Carbon Project Earth observation must deliver
long time series, consistent estimates of global vegetation behaviour
complete with accuracy/quality figures.
• GLOBCARBON will deliver 10 complete years (1998 to 2007) of global
vegetation products to the DGVM and atmospheric chemistry modelling
community at resolutions of ¼, ½ degree and 10 km.
• Timeline:
GLOBCARBON has now entered the operational production phase.
Validation and inter-comparison are key elements (and being conducted
by the industrial contractor).
• Validation and inter-comparison
An ongoing process so validation data still useful – MODIS, Landsat
ETM
• Product Release (1998-2003): January 2006
QUEST, Exeter
27-28 October 2005
Acknowledgements
• Many thanks to all the people who have contributed
data for the validation of GLOBCARBON Burned
Area especially:
– Louis Giglio for TRMM hotspot information
– Robert Fraser and colleagues (FireM3 and
validation data)
– Tom Bobbe for US burn area vectors
– Chris Schmullius/Charles George for Siberia
validation
– Joao Silva for Africa validation data
– DOLA for burned area validation in Australia
QUEST, Exeter
27-28 October 2005