Cal/Val of MODIS Fire Hot Spots Using High Resolution Imagery Cal/Val of MODIS Fire Hot Spots
Using High Resolution Imagery
S. C. Liew, K. H. Lim, C. M. Shen, A. Lim and L. K. Kwoh
CRISP, National University of Singapore email: crslsc@nus.edu.sg
S. C. Liew, K. H. Lim, C. M. Shen, A. Lim and L. K. Kwoh
CRISP, National University of Singapore email: crslsc@nus.edu.sg
SEA-GOFC Fire Workshop
22 - 25 January 2002, Phuket, Thailand
Introduction Introduction
• MODIS : Key instrument on-board Terra (EOS AM-1) satellite, launched in December 18, 1999
• It has 36 spectral bands, ranging from 0.4 to 14 µµµµm
• Resolution from 250m (2 bands), 500m (5 bands), 1km (29 bands)
• One of the unique feature is its direct broadcast capability
– It transmits science data via X-band to ground
– Any ground station equipped with a 3m or larger reception system may receive the direct broadcast data
• It is particularly useful for regional ground station to perform near-real-time environmental monitoring
• MODIS : Key instrument on-board Terra (EOS AM-1) satellite, launched in December 18, 1999
• It has 36 spectral bands, ranging from 0.4 to 14 µµµµm
• Resolution from 250m (2 bands), 500m (5 bands), 1km (29 bands)
• One of the unique feature is its direct broadcast capability
– It transmits science data via X-band to ground
– Any ground station equipped with a 3m or larger reception system may receive the direct broadcast data
• It is particularly useful for regional ground station to perform near-real-time environmental monitoring
Reception and Applications of MODIS Data in Forest Fire Monitoring
Reception and Applications of MODIS Data in Forest Fire Monitoring
• CRISP has been preparing to receive MODIS data since it was launched in late 1999.
• October 2000 : 6-m antenna installed
• Processing up to level 1-B successfully implemented
• On 3 March 2001, the first MODIS data was received and processed
successfully
• Potential applications:
– Fire detection
– fire temperature and fire area retrieval – burn scar mapping
– aerosol retrieval
• CRISP has been preparing to receive MODIS data since it was launched in late 1999.
• October 2000 : 6-m antenna installed
• Processing up to level 1-B successfully implemented
• On 3 March 2001, the first MODIS data was received and processed
successfully
• Potential applications:
– Fire detection
– fire temperature and fire area retrieval – burn scar mapping
– aerosol retrieval
Daily Fire Monitoring Operation at CRISP
Daily Fire Monitoring Operation at CRISP
• Commissioned by the Ministry of Environment, Singapore
• ASEAN Regional Haze Action Plan
– Sets out co-operative measures amongst ASEAN member countries to address the problem of smoke haze in the region
arising from land and forest fires
– Singapore takes up the responsibility of coordinating the regional land and forest fire monitoring mechanisms.
• Commissioned by the Ministry of Environment, Singapore
• ASEAN Regional Haze Action Plan
– Sets out co-operative measures amongst ASEAN member countries to address the problem of smoke haze in the region
arising from land and forest fires
– Singapore takes up the responsibility of coordinating the regional land and forest fire monitoring mechanisms.
Daily Fire Monitoring Operation at CRISP
Daily Fire Monitoring Operation at CRISP
• Near real-time operation
• Determination of fire locations
• Determination of types of land cover burnt (eg. Plantations, peat swamp, natural forests, etc.)
• Production of burn scar maps
• Near real-time operation
• Determination of fire locations
• Determination of types of land cover burnt (eg. Plantations, peat swamp, natural forests, etc.)
• Production of burn scar maps
Daily Fire Monitoring Operation Daily Fire Monitoring Operation
• NOAA 14,16 AVHRR
– Global Coverage at 1km resolution – Derive ‘Hotspots’
• SPOT 1, 2 & 4
– 20m ground resolution
– Provide ‘zoom-in’ views of fire activities – Satellite image acquisition guided by
‘Hotspots’ derived from NOAA AVHRR data
• NOAA 14,16 AVHRR
– Global Coverage at 1km resolution – Derive ‘Hotspots’
• SPOT 1, 2 & 4
– 20m ground resolution
– Provide ‘zoom-in’ views of fire activities – Satellite image acquisition guided by
‘Hotspots’ derived from NOAA AVHRR data
Hotspots derived from NOAA AVHRR Data
Hot Spots
Planning SPOT Passes Using NOAA AVHRR Hotspots
Planning SPOT Passes Using NOAA AVHRR Hotspots
NOAA14 2000 Jul 14 08:59 UTC Hot Spots
Smoke Haze
Programming SPOT Satellites Programming SPOT Satellites
Standard Product
Annotated Fire Scene processed to Level 2A
Fire Report Fire Report
Modis VIS/NIR/SWIR Bands (Land/Aerosol)
Modis VIS/NIR/SWIR Bands (Land/Aerosol)
• 1 (Red) : 620 - 670 nm (250 m)
• 2 (NIR) : 841 - 876 nm (250 m)
• 3 (Blue) : 459 - 479 nm (500 m)
• 4 (Green) : 545 - 565 nm (500 m)
• 5 (SWIR) : 1230 - 1250 nm (500 m)
• 6 (SWIR) : 1628 - 1652 nm (500 m)
• 7 (SWIR) : 2105 - 2155 nm (500 m)
• 1 (Red) : 620 - 670 nm (250 m)
• 2 (NIR) : 841 - 876 nm (250 m)
• 3 (Blue) : 459 - 479 nm (500 m)
• 4 (Green) : 545 - 565 nm (500 m)
• 5 (SWIR) : 1230 - 1250 nm (500 m)
• 6 (SWIR) : 1628 - 1652 nm (500 m)
• 7 (SWIR) : 2105 - 2155 nm (500 m)
Modis MWIR/Thermal IR Bands (Surface/Cloud Temperature) Modis MWIR/Thermal IR Bands
(Surface/Cloud Temperature)
• Band 20 : 3.660 - 3.840 µµµµm
• Band 21 : 3.929 - 3.989 µµµµm (low gain)
• Band 22 : 3.929 - 3.989 µµµµm (high gain)
• Band 23 : 4.020 - 4.080 µµµµm
• Band 31 : 10.780 - 11.280 µµµµm
• Band 32 : 11.770 - 12.270 µµµµm
• Band 20 : 3.660 - 3.840 µµµµm
• Band 21 : 3.929 - 3.989 µµµµm (low gain)
• Band 22 : 3.929 - 3.989 µµµµm (high gain)
• Band 23 : 4.020 - 4.080 µµµµm
• Band 31 : 10.780 - 11.280 µµµµm
• Band 32 : 11.770 - 12.270 µµµµm
Advantages of MODIS Advantages of MODIS
• High positional accuracy
• High radiometric precision (12 bits encoding)
• Moderate spatial resolution modes available (250 m, 500 m)
• High saturation level
– 350 K (band 22) – 500 K (band 21) – 400 K (band 31)
• High positional accuracy
• High radiometric precision (12 bits encoding)
• Moderate spatial resolution modes available (250 m, 500 m)
• High saturation level
– 350 K (band 22) – 500 K (band 21) – 400 K (band 31)
MODIS Hot Spots MODIS Hot Spots
• T21 > 360 K
• Or (T31 > 320 K and T21 - T31 > 20 K)
• Flag hot spots in sun glint areas
• Exclude cloud areas
• T21 > 360 K
• Or (T31 > 320 K and T21 - T31 > 20 K)
• Flag hot spots in sun glint areas
• Exclude cloud areas
Case Study : 23 May 2001 Fire in Sarawak, Malaysia Case Study : 23 May 2001 Fire in Sarawak, Malaysia
MODIS - 23 May 2001 MODIS - 23 May 2001
SPOT Fire Locations A 2.8328N 111.9922E B 2.8344N 111.9994E C 2.8308N 111.9997E MODIS Hot Spots
2.8387N 111.9912E 2.8373N 112.0008E 2.8295N 111.9899E 2.8281N 111.9995E
MODIS Hot Spots validated with SPOT fires 2001-05-23
MODIS Hot Spots validated with SPOT fires 2001-05-23
MODIS Hot Spots validated with SPOT fires 2001-05-23
MODIS Hot Spots validated with SPOT fires 2001-05-23
SPOT Fires
D 2.5633N 111.9192E E 2.5528N 111.9158E F 2.5533N 111.9211E G 2.5550N 111.9258E H 2.5531N 111.9303E
MODIS Hot Spots
2.5642N 111.9143E 2.5628N 111.9239E 2.5563N 111.9034E 2.5549N 111.9130E 2.5535N 111.9226E 2.5521N 111.9322E 2.5508N 111.9417E 2.5456N 111.9117E 2.5443N 111.9213E
SWIR Band of SPOT4 Helps to Locate Fires Without Visible
Smoke Plumes
SWIR Band of SPOT4 Helps to Locate Fires Without Visible
Smoke Plumes
R,G,B = Bands 4,3,2 R,G,B = Bands 3,2,1
Hot Spots
Another Case: 2002 Jan 19
Fire in Riau Province, Sumatra, Indonesia Another Case: 2002 Jan 19
Fire in Riau Province, Sumatra, Indonesia
MODIS Image
SPOT - 2002 Jan 19 SPOT - 2002 Jan 19
Retrieval of Subpixel Fire Temperature and Fire Area
Retrieval of Subpixel Fire Temperature and Fire Area
• Fire pixel is assumed to radiate like a blackbody with a constant emissivity.
• Assume only two temperature fields
• Fire pixel is assumed to radiate like a blackbody with a constant emissivity.
• Assume only two temperature fields
Fire Temperature Tf Fractional area = f
Background Temperature Tb
Fractional area = 1-f
Apparent Temperature of a Fire Pixel Apparent Temperature of a Fire Pixel
Fire Area (Fraction of pixel) = f
Fire Temperature = Tf Background Temperature = Tb
Radiance, L K f
c T
f
c T
f b
= − + −
−
exp( 2 / ) 1 exp( 2 / )
1
λ λ 1
Apparent Temperature
( )
T c
= K L +
2 1
λ ln( / 1)
Two-Bands Algorithm Two-Bands Algorithm
• T1 = apparent temperature detected in band-1
• T2 = apparent temperature detected in band-2
• Given T1 and T2 of a hot-spot pixel, invert the temperature equations to obtain the fractional fire area f and subpixel fire temperature Tf. (Two unknowns, two equations).
• Analytic Solution is not possible.
• Use numerical method of solution: Newton- Raphson
• T1 = apparent temperature detected in band-1
• T2 = apparent temperature detected in band-2
• Given T1 and T2 of a hot-spot pixel, invert the temperature equations to obtain the fractional fire area f and subpixel fire temperature Tf. (Two unknowns, two equations).
• Analytic Solution is not possible.
• Use numerical method of solution: Newton- Raphson
Determination of background temperature Tb Determination of background temperature Tb
• Background temperature is taken to be the mean of a Region Of Interest(ROI) that was visually selected from the image.
• The ROI was taken from non-burning areas near the hotspot region.
• Background temperature is taken to be the mean of a Region Of Interest(ROI) that was visually selected from the image.
• The ROI was taken from non-burning areas near the hotspot region.
868.25 848.83 942.88 780.11 819.12 814.24 876.91 864.37 897.58 868.25 793.49 810.95 807.62 805.52 785.04 814.24 856.32
Tf
0.0041 308.41
338.86 123.960
-16.548
0.0034 307.91
336.82 123.966
-16.551
0.0026 307.80
346.66 123.843
-16.510
0.0037 307.74
329.66 123.822
-16.508
0.0058 309.01
333.70 126.330
-16.554
0.0093 310.11
333.19 126.336
-16.544
0.0093 310.37
339.77 126.807
-15.954
0.0052 308.96
338.45 126.812
-15.963
0.0115 310.88
341.94 125.847
-14.787
0.0111 310.70
338.86 125.854
-14.788
0.0070 309.36
331.03 126.052
-14.598
0.0131 310.80
332.85 126.023
-14.595
0.0065 309.24
332.50 126.059
-14.600
0.0102 310.27
332.28 126.030
-14.596
0.0061 308.95
330.16 127.035
-14.658
0.0119 310.64
333.19 127.042
-14.651
0.0037 311.04
351.73 114.565
-2.577
f T2
T1 longitude
latitude
Conclusions Conclusions
• MODIS hot spots can be validated using SPOT images.
• Most fires detected by SPOT are also detected by MODIS.
• Some fires in SPOT are not detected by
MODIS - small fires? Hot spots criteria too stringent? Need Adaptive thresholds.
• Some hot spots in MODIS are not detected in SPOT - No visible smoke plumes, but may be detected if SWIR band of SPOT 4 is used.
• Retrieval of subpixel fire temperature and area has been attempted, seems promising.
• MODIS hot spots can be validated using SPOT images.
• Most fires detected by SPOT are also detected by MODIS.
• Some fires in SPOT are not detected by
MODIS - small fires? Hot spots criteria too stringent? Need Adaptive thresholds.
• Some hot spots in MODIS are not detected in SPOT - No visible smoke plumes, but may be detected if SWIR band of SPOT 4 is used.
• Retrieval of subpixel fire temperature and area has been attempted, seems promising.