Fire Extent, Severity and Patchiness Mapping using Daily
Satellite Data
Stefan W MaiermaitecDarwin
to reduce Greenhouse Gas emissions by reducing:
1. fire severity
2. fire extent
3. fuel load
Fundamental Principle of Savanna Fire Projects
GHG emission equation (simplified):
E = P · A · FL · BE · EF
E: GHG emittedP: patchiness factor → fire severityA: fire extent / area burnt → fire extentFL: fuel load → fuel loadBE: combustion completeness (“burning efficiency”) → fire severityEF: emission factor
to reduce Greenhouse Gas emissions by reducing:
1. fire severity
2. fire extent
3. fuel load
Fundamental Principle of Savanna Fire Projects
→ these variables have to be measured (estimated) repeatedly
Daily Near Real-Time Fire Extent Mapping
Fire Severity Mapping
Daily Near Real-Time Fire Extent Mapping
Fire Severity Mapping
2-3 images per day per satellite
5 images per day
~1,800 images (1.8TB) per year
Data Source: MODIS
calibration atmospherecorrectionviewing geometry
correctionburnt areamapping
Processing Steps
near real-time mapping < 2h
Example: Near Real-Time Burnt Area Map
Example: Day of Burn
fire stopped by creek
fire jumped creek
fire jumped creek
daily progression of fire
Accuracy (Top-End)
overall accuracy
commission error
omission error
MODIS Daily Burnt Area (250m) 95.2% 3.7% 6.2%
NASA (MCD45) (500m) 90.7% 3.2% 15.8%
NAFI (manual) (250m) 94.6% 4.6% 6.7%
Conclusion – Daily Fire Extent Mapping
● automatic (twice) daily mapping of fire extent possible● fire extent maps available within 2h of satellite overpass● Top End: automatic mapping as accurate as manual mapping● accuracy in other areas has to be systematically assessed
(→ algorithm might have to be adjusted)● automatic system is treating every image exactly the same
→ accuracy does not change→ no on-going validation necessary(manual mapping treats every image slightly differently)
Daily Near Real-Time Fire Extent Mapping
Fire Severity Mapping
Fire is not Fire
fire severity
fine scale patchiness
Patchiness: Spatial Scale
northern Australia, on-ground transects within fire perimeters:
83% (EDS) / 93% (LDS) burnt
87% (EDS) / 89% (LDS) of unburnt patches are ≤ 5m
Oliveira, et al. (2015). Ecological Implications of Fine-Scale Fire Patchiness and Severity in Tropical Savannas of Northern Australia. Fire Ecology, 11, 10-31.
Patchiness: Temporal Constraints
1 day after fire
10 days after fire
Spectral Unmixing of MODIS Pixels (Fraction Pixel Burnt)
on-ground satellite sensor
30%
70%
un-mixing
pixel
unburnt burnt
Example 1: Tropical Savanna Northern Australia
Example 2: Tropical Savanna Northern Australia – Day of Burn
Example 2: Tropical Savanna Northern Australia – Fraction Pixel Burnt
day-time fire
night-time fire
Example 3: Tropical Savanna Northern Australia – Fraction Pixel Burnt
Photo: Rohan Fischer
Comparison Season vs Fraction Pixel Burnt (Patchiness)
http://www.maitec.com.au/gallery/fire_animations/index.html
Fraction Pixel Burnt WALFA
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 10
0.01
0.01
0.02
0.02
0.03
0.03
0.04
456789101112
fraction pixel burnt
frequ
ency
Trend EDS/LDS Fraction Pixel Burnt WALFA
late dry season fires
have become patchier
(smaller fraction pixel
burnt values)
not accounted for!!!
start of fire management
Conclusion – Fraction Pixel Burnt
● physical measure of fire severity (not an index)● no field calibration necessary● scale (sensor) independent● field comparison looks good, proper validation necessary● analysis of archive provides interesting insights (e.g. LDS fires patchier)
● we can measure patchiness factor (P) and combustion completeness (BE) !!!● no need to use a poor surrogate
(e.g. season, project budget, number of helicopter hours)● fire management does more than “only” shift fires from LDS to EDS● time to update the decade old method for emissions accounting !!!
Slide 1Slide 2Slide 3Slide 4Slide 5Slide 6Slide 7Slide 8Slide 9Slide 10Slide 11Slide 12Slide 13Slide 14Slide 15Slide 16Slide 17Slide 18Slide 19Slide 20Slide 21Slide 22Slide 23Slide 24Slide 25