1. The MODIS Active Fire and Thermal Anomalies Products
Louis Giglio
Science Systems & Applications, Inc. / University of Maryland
NIMH Bulgaria - EUMETSAT Training Workshop
7-10 September 2009

2. Topics Background MODIS Sensor Products Distribution Caveats
Validation
Example Applications and Results

3. Satellite-Based Active Fire Products
Identify where fires are actively burning at time of satellite overpass (and implicitly when they are burning)
Possibly provide additional information about fires at time of satellite overpass
Intensity, average temperature, instantaneous size, rate of combustion, injection height, etc.

4. Satellite-Based Burned Area Products
Map spatial extent of post-fire burn scars over a particular time period
Approximate date of burn
Possibly provide additional information about area burned
Fire severity, burn severity

5. Southeast Australia, 10 Dec. 2006, 03:45 UTC
MODIS Rapid Response System

6. Fire Radiative Power

7. MODIS Instrument (1 of 2)
“Moderate Resolution Imaging Spectroradiometer”
On board AM-1 (“Terra”) and PM-1 (“Aqua”) polar orbiters
Terra 10:30 & 22:30 local overpass
Aqua 01:30 & 13:30 local overpass

8. MODIS Instrument (2 of 2) 36 spectral bands covering 0.4 to 14.4 µm
Two 250-m bands
Five 500-m bands
Twenty nine 1-km bands
Enable comprehensive daily evaluation of land, ocean, and atmosphere

9. MODIS Active-Fire Channels
Dedicated 1 km fire bands
Channel 21: 3.96 µm, ≈ 500 K saturation
Channel 31: 11.0 µm, ≈ 400 K saturation
Lower noise, lower quantization error, 1-km multipurpose band
Channel 22: 3.96 µm, ≈ 330 K saturation

10. Fire Detection Approach
Look for elevated signal at 4 m; use 11 m channel to help exclude warm, non-fire surfaces
Adaptive thresholds based on local statistics so algorithm can be applied globally
Use additional tests and channels to help reject false alarms
Small convective clouds, sun glint, heterogeneous surfaces, land cover/vegetation boundaries, coastline, desert and other hot/reflective surfaces, urban areas*

11. Products

12. Preliminaries MODIS “Collection” MODIS Tile
Reprocessing round or version
Currently at Collection 5
Collection 6 reprocessing early next year
MODIS Tile
Global sinusoidal grid

13. MODIS Tiles

14. MODIS Fire Products (1 of 3)
Active Fire (Level 2 -- native swath)
Fire mask
QA layer
Detailed information about each fire pixel
MOD14, MYD14
Active Fire (Level 3 -- tiled, sinusoidal grid)
Cumulative fire mask
Maximum fire radiative power (FRP)
Daily and 8-day summaries
MOD14A1, MYD14A1, MOD14A2, MYD14A2

15. Example 8-Day L3 Fire Product Angola
Water
Unknown
Non-Fire
Missing
2000 day 241
Tile h19v10

16. MODIS Fire Products (2 of 3)
Climate Modeling Grid
Global, 0.5°
Monthly, 8-day
MOD14CMH, MYD14CMH MOD14C8H, MYD14C8H
Fire Locations
ASCII
Monthly
MCD14ML

17. MODIS Fire Products (3 of 3)
Rapid Response Products
Imagery (JPEG)
Near-real time active fire locations (ASCII)

18. Siberia, 22 May 2001

19. Plume length > 200 km
Australia, 4 Dec. 2002

20. Distribution

21. https://lpdaac.usgs.gov/lpdaac/get_data

24. CMG and Fire-Location Products
fuoco.geog.umd.edu
Login: fire Password: burnt
cd modis

25. Caveats

26. Product Caveats and Issues
Pre-November 2000 data quality
Envelope of detectable fires
False alarms
Fire locations
Active fire vs. burned area

27. Pre-November 2000 Data Quality
MWIR channels suffer from crosstalk, poor calibration, “dead” detectors
SWIR and TIR channels also affected (but to a lesser extent)
Obvious artifacts in many products

28. Early Terra MODIS Artifacts

29. Terra Band 21, Post-November 2000
Vietnam
:40

30. Factors Affecting Detection
Fire size
Fire temperature
Biome
Season
Surface temperature
Cloudiness
Types of clouds
Position of sun
Viewing geometry
Characteristics of smoke
Instrument issues
Other

31. False Alarm Sources Small convective clouds Sun glint
Land cover/vegetation boundaries
Coastline
Desert and other hot/reflective surfaces
Heterogeneous surfaces
Urban areas*

32. False Alarms
Terra MODIS
Shanghai
18 July 2004,
02:45 UTC

33. Fire-Location Limitations
“Fire location”-only products omit useful information
ASCII (“text”) files
Lack spatial context
cloud cover, missing data, water
Sometimes inappropriate for research purposes (yet used anyway!)

34. Active Fire vs. Burned Area Data
In general they are not the same:
1 km2 “fire pixel”  1 km2 area burning
1 km2 “fire pixel”  1 km2 area burned

36. Validation

37. MODIS Active Fire Detection Validation

38. MODIS Active Fire Validation (1)
Primary method is to use coincident, high resolution ASTER imagery
“Advanced Spaceborne Thermal Emission and Reflection Radiometer”
Terra only (no Aqua ASTER)
14 high resolution channels
15 m, 30 m, 90 m
None ideal for observing fires

39. MODIS Active Fire Validation (2)
Fire mask using ASTER imagery
Sometimes degraded by frequent ASTER saturation
Fire radiative power retrieval using ASTER
Cannot use simple middle-infrared band approach used for MODIS with ASTER
Often degraded by frequent ASTER saturation
Simulated MODIS imagery
Independent burn scar maps
Usually not able to evaluate false alarm rate

40. ASTER Fire Detection Approach
Use two reflective bands
One sensitive to emissive radiation from fires
SWIR band 8 (2.3 m)
One insensitive to emissive radiation from fires
NIR band 3N (0.86 m)
Otherwise highly correlated and sensitive to reflective, non-fire radiation
Approach is reasonable for small ASTER pixels, but useless at coarser spatial resolution

41. Example Scene: Eastern Cambodia
ASTER Bands 8 (2.33 µm),
3N (0.82 µm), 1 (0.56 µm)
Active Fire Mask
12 km

42. Example Scene: Eastern Cambodia
Fires
Clouds
Burn Scars
Vegetation

43. ASTER Band 9 Southern Africa, 17 Aug. 2001
Grid delineates “edges” of 1-km MODIS pixels.
Southern Africa, 17 Aug. 2001

44. MODIS Active Fire Detection Validation via ASTER

45. MODIS Active Fire Detection Validation
Post-Fire Burn Scar Maps
Coincident 30-m ASTER Imagery
de Klerk (2008)

46. Example Simulation Results
MODIS
Temperate deciduous rainforest
Night
0° scan angle
Summer
No background fires

47. Collection 6 Plans

48. Algorithm Refinements
Reduce false alarms in problem areas
Recognize optically thick smoke (vs. cloud)
Smoldering ratio
Approximate state of combustion (smoldering, flaming, mixed)
Improve detection confidence estimate
Improved land/sea mask*
Input data improvements*

49. Applications and Results

50. Terra MODIS Monthly Active Fires

51. Terra MODIS Peak Fire Month
Giglio et al. (2006)

52. Terra MODIS Mean Fire Radiative Power
Giglio et al. (2006)

53. Autocorrelation
Giglio et al. (2006)

54. Aqua Fire Fraction STRONGER WEAKER DIURNAL CYCLE STRONGER
Giglio et al. (2006)

55. Global Fire Emissions Database (GFED)

56. McCarty et al. (2007)

57. Plume Heights
Mazzoni et al. (2007)

58. Additional Information
MODIS Fire and Thermal Anomalies
MODIS Active Fire User’s Guide
MODIS Land Rapid Response System
WIST (order MODIS land products)