THE GLOBCARBON INITIATIVE: A view from MPIMeteorology on Burned area

1
THE GLOBCARBON INITIATIVE A view from
MPI-Meteorology on Burned area
A. Heil and M. Schultz (MPI)
GLOBCARBON is a ESA Data User Programme Project
being executed by a Belgo-Italian consortium led
by VITO
2
Intercomparison of ATSR, AVHRR-FFPMP, GlobCarbon
fire data for Indonesia 1996-2003

• Data source and processing
• FFPMP-JICA-MoF AVHRR-FFPMP (7/97-2/05)
• Monthly accumulated hotspots SE Asia
  (http//ffpmp2.hp.infoseek.co.jp/English/)
• The AVHRR fire counts are captured during
  afternoon/evening with threshold of 315K
  (daytime) and 310 (nighttime) (Stolle et al.
  2004).
• b) AT2Old ATSR nighttime BT gt 308 Kelvins
• Source http//dup.esrin.esa.it/ionia/wfa/index.a
  sp
• c) AT1Old BT gt 312 Kelvins (Saturation)
• d) AT1Clean ATSR-cleaned Algo 1 (pers. Comm.
  Perreira 2004, ESA), 1997 to 2001
• e) GC Monthly Globcarbon Burned area ASCII data
  1/98 to 11/2000 (unvalidated test version,
  downloaded June 2005) with different Confidence
  Rating Indexes (CRI). Abbreviations used here no
  CRI applied (Index 000), CRI above 70, 80 and
  90(Index 070, 080, 090).

3
Yearly Fire Counts
The number of total fire counts (FC) shows large
differences between the different products. On
average, AT2Old and AT1Clean record 10, AT1Old
6 and GC_000 5 of the number of FFPMP-AVHRR FC.
Minimum CRI for Indonesia is 65, the average CRI
is 87).
4
Yearly Fire Counts - 2
Globcarbon data omitting CRI values below 70, 80
and 90 decreases the number of observed fire
counts within the study area Indonesia by another
14, 34 and 48, respectively. Yearly FFPMP Fire
Counts and AT2/1Old and AT1Clean Fire Counts are
strongly correlated (R² gt 0.953 at 0.01
significance level). In contrast, there is no
statistically significant correlation of GC data
with the other satellite products.
5
Monthly Fire Counts
6
Monthly Fire Counts - 2
The monthly total number of FC of the
ATSR-derived products and the GC product is much
smaller than the AVHRR-FFPMP product. Lower
monthly values are due to data gaps. The ratio of
the monthly number observed by the ATSR-derived
products and the GC product is still highly
variable, indicating some systematic bias between
these products on a monthly scale.
7
Monthly Fire Counts - 3
Correlation coefficients between the different
products for each year (based on the monthly
total counts). The FFPMP-AT2old datasets are
highly correlated during 1997 to 2002, while
there is only medium correlation in 2003. The GC
product almost not correlated with FFPMP and
AT2old in 1998, while there is medium correlation
in 1999/2000.