Mapping burned scars in Amazon region using MODIS data Big Bear Lake, California, USA, 2011. André Lima Yosio Edemir Shimabukuro Luiz Eduardo Aragão SCGIS 2011
Dec 27, 2015
Mapping burned scars in Amazon region using MODIS data
Big Bear Lake, California, USA, 2011.
André LimaYosio Edemir ShimabukuroLuiz Eduardo Aragão
SCGIS 2011
Context
It is estimated that 75% of Brazil's emissions of CO2 from forest fires (MCT, 2004)
Context
It is estimated that 75% of Brazil's emissions of CO2 comes from forest fires (MCT, 2004);
According to Bowman et al. (2009) until 50% of GHG emissions in the world comes from forest fires.
Context
It is estimated that 75% of Brazil's emissions of CO2 from forest fires (MCT, 2004);
According to Bowman et al. (2009) until 50% of GHG emissions comes from burning the globe;
There are not systematic regional mapping of burnt scars in the tropics (Giglio et. Al., 2010, Setzer et al., 2011).
Justification
Need for data on fires in tropical forests to generate estimates of GHG emissions (IPCC, 2008).
It is possible to map fire scars using MODIS data (250m) at a level of detail appropriate to the assessments of GHG emissions caused by burning.
Hipotese
Objective
To develop methodology for mapping of burned scars using MODIS data – Surface Reflectance Daily (MOD09);
Study Area
Amazonia Legal area 5.217.423 km², equivalent 61% Brazilian territory
Study Area
Location General chateristics
Amazon Forest, the largest tropical biome of the world equivalent to 30% of remaining tropical forests;
High biodiversity;
Agriculture frontier, Deforestation, region called “Deforestation Arc”.
Material and Methods Images Selection per Brazilian Federation Unit Based on hot spot active fire frequency distribution,
PROARCO Data (http://sigma.cptec.inpe.br/queimadas/) Acre
Amapa Maranhao
Mato Grosso Para
RoraimaRondonia
Amazonas
Material and Methods
Images Selection in Rondonia State Based on hot spot active fire frequency distribution,
PROARCO Data (http://sigma.cptec.inpe.br/queimadas/)
January
February
March
April
May
June
July
August
September
October
November
December
Images correspond to months with higher occurrence of fire hotspots
Material and Methods
Used images Surface Reflectance daily 250m (Mod09 product); Spectral Bands: 1 (Red), 2 (Near-infrared), 6* (Middle-infrared);
Total images used =105.
* Spatial resolution 500 m.
Images selected
Table 01. Images used to map burnt scars occurred in 2005
0202
02
02
03020503
Spectral Linear Mixing Model (SLMM)– Decomposition (n) spectral bands in three fraction images.
Material and Methods
Vegetation Fraction
Soil Fraction
Shade Fraction
Shade Fraction – Targets with low reflectance are realced.
Water body
Burnt Scar
Materiais e Métodos
Segmentation– Region algorithm– Threshold
Area = 4 (pixels)Similarity = 8 (digital number value variation)
Shade Fraction image Segmentation
• Classification– Classification non-supervised
• Algorithm ISOSEG– Threshold 75% (probability)
Material Methods
Burnt scars boundary Burnt scar mapped
Results
Burnt scars in 2005
Results
Table 02. Burnt scars total area mapped in 2005 per State.
*Table 02. Estimates affected by the large cloud cover.
Final Considerations
Useful methodology; Support for a future detection
burnt scars program in the Amazon (DETEQ);
Important source of data for emission models of Greenhouse Gases;
Validation in process.
Obrigado.André Lima