THE GLOBCARBON INITIATIVE: Populating the Earth with Burned Area

1. 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)

2. 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.

3. 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

4. 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).

5. Burned Area - Approach GBA-2000 Globscar • Original = 6 regional algorithms • GLOBCARBON = 1 global and 2 regional algorithms • it has associated with it confidence information (detection confidence from individual algorithms) • Original = 2 global algorithms and burn when both agree. • GLOBCARBON = each algorithm and sub-parts given a probability. • The resulting probability determines occurrence of a pixel as burned (confidence information) • Results are merged to one product • Collocation with available active fire products improves confidence

6. Results – 1 km (Madagascar) Confidence Rating Index (CRI) July 1998 63-74% 75-87% 88-100%

7. Results – 1 km (Madagascar) Algorithm Detection (GLOBSCAR, GBA, Both) July 1998 GLOBSCAR only GBA only Both algorithms

8. Results – 1 km (Angola) July 1998 Confidence Rating Index (CRI) 63-74% 75-87% 88-100%

9. Results – 1 km (Angola) July 1998 Algorithm Detection (GLOBSCAR, GBA, Both) GLOBSCAR only GBA only Both algorithms

10. MODIS Comparison – 1 km MODIS GLOBCARBON July 2000

11. MODIS Comparison – 1 km MODIS GLOBCARBON Sept 2000

12. Results –10 km (Mongolia) 1-10 11-20 21-30 31-40 41-50 51-60 61-70 71-80 81-90 91-100 2000 May 2000 Percentage of pixels with CRI > 80 in a 10*10km box

13. 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

14. NW Australia – 10km May 99 May overlain with May Vectors May overlain with May Vectors

15. NW Australia 10km – May 99 May overlain with June Vectors May overlain with June Vectors

16. 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

17. Validation - Results Montana, Stratum 1, R2 = 0.80 South Sudan, Stratum 1, R2 = 0.81

18. Validation - Results July 13 2000 Apr 28 2000 Aug 14 2000 Dec 24 2000 Sept 15 2000 Mar 30 2001 Montana, Stratum 1, R2 = 0.80 Sudan South, Stratum 1, R2 = 0.81

19. LAI Results – 10 km 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 June 1999

20. Phenology Results – 10km 1-50 water 51-100 201-300 151-200 126-150 101-125 no value Leaf On 1998

21. 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

22. 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