Satellite Active Fire Product Development and Validation: Generating Science

Quality Data from MODIS, VIIRS and GOES-R Instruments

Wilfrid Schroeder1, Ivan Csiszar2, Louis Giglio3, Evan Ellicott3, Christopher Justice3, Christopher Schmidt4

1 ESSIC/CICS, UMD2 STAR NOAA/NESDIS

3 Dept of Geography, UMD4 CIMMS, UW-Madison

Team BackgroundOngoing CICS Projects:

• GOES-R: • “Validation and Refinement of GOES-R ABI Fire Detection Capabilities” (GOES-R AWG)

• MODIS & VIIRS: • “Active Fire Product Evaluation and Development from MODIS and VIIRS” (NASA)• “Development of an Enhanced Active Fire Product from VIIRS” (IPO – includes NPP active fire

product validation program activities also)

Linkages and collaborations:

• Christopher Schmidt (UW-Madison) – GOES Imager/ GOES-R ABI Fire Product PI (GOES-R AWG)

• Christopher Justice and Louis Giglio (UMD/Geography) – MODIS Active Fire Product PIs (NASA)

• Ivan Csiszar (NESDIS/STAR) and Christopher Justice (UMD/Geography) – NPP/VIIRS Active Fire Product PIs (NASA, IPO)

• Wilfrid Schroeder, Christopher Schmidt, Ivan Csiszar, Elaine Prins, Christopher Justice – fire product evaluation in the Amazon and long-term fire data record (NASA LBA-ECO – recently concluded)

Progress in the Last Three Decades

Major Data Sets**

Adv Very High Res Radiometer (AVHRR) 1kmx12h within antenna range

1980

1990

2000

2010

GOES East Imager 4kmx30min Western Hemisphere

Tropical Rainfall Monitoring Mission (TRMM)2.4kmx12h ±38º

Mod Res Imaging Spectroradiometer (MODIS/Terra)(MODIS/Aqua)

1kmx12-24h global

GOES VAS13.8kmx30min Western Hemisphere

** Excluding nighttime sensors such as ATSR, DMSP

Simple Threshold (single or multi-band)

Contextual methods (x,y) (dynamically adjusted)

Contextual methods (x,y,t) (dynamically adjusted)

A few dozen images

400K+ images from GOES only

Essentials in Active Fire Monitoring

Fires are highly dynamic events

Fires may/not leave detectable scars behind

ETM+ 10am

ASTER 10:30am

0

5

10

15

20

25

30

35

40

45

0.00 0.25 0.50 0.75 1.00 1.25 1.50 1.75 2.00 2.25 2.50 2.75 3.00

Rad

ianc

e x

10 -6

(W

m-3

str-1

)

Fire x 10-2

700ºC

Sun x 10-6

6000ºC

ASTER Channel 3N

ASTER Channel 8

(m)

ETM+Channel 4

ETM+Channel 7

Active Fire Reference Data Derived from ASTER and ETM+ Imagery

ASTER

ETM+

ASTER bands 3 and 8 and ETM+ bands 4 and 7

0

10

20

30

40

50

60

70

1 2 3 4 5 6 7 8 9 10Fire Clusters per MODIS pixel

Fre

qu

en

cy

(%

)

ASTER

ETM+

Sample Size: 18 ASTER scenes

Region: South Africa

Proof of concept using fixed threshold method applied to ASTER band 9 to derive 30m resolution active fire masks

Morisette et al. 2005

Sample Size: 100 ASTER scenes

Region: Global

Development of robust active fire detection algorithm for ASTER

Giglio et al. 2008

Sample Size: 131 ASTER scenes

Region: Northern Eurasia

Development of active fire validation protocol

Csiszar et al. 2006

MODIS/Terra Active Fire ValidationC3-C4 Algorithm Version

Sample Size: 115 ASTER scenes

Region: CONUS

Validation of NOAA/NESDIS operational fire monitoring system including analyst data

Schroeder et al. 2008

Sample Size: 167 ASTER + 123 Landsat ETM+ scenes

Region : Brazilian Amazonia

Generalization of moderate-coarse resolution fire data validation (MODIS + GOES) using higher resolution imagery

Schroeder et al. 2008

MODIS/Terra Active Fire ValidationC3-C4 Algorithm Version

Sample Size: 24 ASTER + 8 Landsat ETM+ scenes

Region : Brazilian Amazonia

Assessment of short-term variation in fire behavior – implications to active fire validation

Csiszar and Schroeder 2008

MODIS/Terra C5 AlgorithmStage 3 Fire Validation

Sample Size: ~2500 ASTER scenes

Region : Global

Stage III validation of MOD14

Schroeder et al. (in preparation)

• Daytime & nighttime data

• Data equally distributed across the globe

• Multi-year analysis (2001-2006)

• ASTER SWIR anomaly May ‘07

• Omission/commission errors derived as a function of percent tree cover

Temporal Consistency of MOD14 Detection Performance

0

20

40

60

80

100

01<>20 41<>60 81<>100 121<>140 161<>180 >200

Fire Cluster Size (number of 30m ASTER fire pixels)

Pro

bab

ilit

y o

f D

etec

tio

n (

%)

2001-2002

2003-2004

2005-2006

Using a subset of points covering the range of 20-40% tree cover

No statistically significant difference over time (i.e., Dt = 0; p < 0.01)

Overall Probability of Detection

Summary curve using all data points

(125K MODIS pixels with >0 ASTER fire pixels including16K MOD14 fire pixels)

0

10

20

30

40

50

60

70

80

90

100

0 100 200 300 400 500 600 700 800 900

Fire Cluster Size (number of 30m ASTER fire pixels)

Pro

babi

lity

of D

etec

tion

(%)

Day

Night

Daytime Probability of Detection as a Function of Percentage Tree Cover**

** average value calculated using a 20x20km window centered on the target pixel

0

20

40

60

80

100

0 50 100 150 200 250 300 350 400 450 500

Fire Cluster Size (number of 30m ASTER fire pixels)

Pro

babi

lity

of D

etec

tion

(%)

vcf <20 vcf 20<>40 vcf 40<>60 vcf >60 All

ASTER (RGB 8-3-1) 21 June 2003 17:38:35UTC

Manitoba, Canada

ASTER (30m Fire Mask) 21 June 2003 17:38:35UTC

Manitoba, Canada

Commission errorsRecently burned pixels with discernable scars constitute a large fraction of the false detections. Overall fire-unrelated commission error ~2%

Nighttime commission error rate is zero.

Results – Commission ErrorsResults – Commission Errors

Schroeder et al. (in preparation)

Results – Commission ErrorsResults – Commission Errors

Typical false alarm in MOD14 data

20 Jul 2003 1407UTC21 Aug 2003 1407UTC

Commission errors can occur multiple times at the same location

MODIS/Terra was found to detect twice as many false positives as MODIS/Aqua

MIR – Initial Tests: Deriving MODIS L1B TOA Radiances usingASTER Surface Kinetic Temperature data + Radiation Transfer Model

MODIS L1B Ch21 07 Aug 2004 1405 UTC

11.7o S 56.6o W

UMD MODIS Ch21 Proxy Data07 Aug 2004 1405 UTC

11.7o S 56.6o W

Early Assessment of NPP/VIIRS Active Fire Data

Early Assessment of NPP/VIIRS Active Fire Data

MIR – Initial Tests: Deriving MODIS L1B TOA Radiances usingASTER Surface Kinetic Temperature data + Radiation Transfer Model

Initial Results

MODIS/Terra (1kmx1km) VIIRS (750m x 750m) VIIRS (250m x 750m)

Defining TIR Saturation Levels

Results being used to support VIIRS hardware and software configuration to allow optimum fire detection capabilities

Early Assessment of GOES-R/ABI Active Fire Data

Fire MaskProxy ABI (derived from MODIS L1B)

Selection of Coincident MODIS and ASTER L1B Data

0

10

20

30

40

50

60

70

80

90

100

0 50 100 150 200

Number of 30m Active Fire Pixels

Pro

ba

bili

ty o

f O

mis

sio

n (

%)

ABI_AMZ_CONUS

MOD14_Global

MOD14_AMZ

WFABBA_AMZ

Initial Results

ABI Active Fire Product Validated using Reference

ASTER Data

Probability of Detection (omission) Defined as a Function of ASTER

Fire Statistics

Results being used to assess and refine pre-flight fire detection algorithm performance and to define routine fire validation strategy for implementation during the post-launch phase

ABI

GOES

MODIS

Supporting Science Quality Data Development Regionally

1998 1999 2000 2001 2002

2003 2004 2005 2006 2007

0 100% Fraction of observations obscured by clouds (JAS)

1998 1999 2000 2001 2002

2003 2004 2005 2006 2007

Supporting Biomass Burning Emissions Products

UW

-Mad

iso

n

CIM

SS

Supporting Science Quality Data Development Globally

Global Geostationary Fire Monitoring Network

Final Remarks

• Development of MODIS active fire product continues after 10years – new versions incorporating refinements to account for problems identified during the validation analyses

• NPP/VIIRS pre-flight fire data analyses providing valuable information

• Thermal infrared band (M15) saturation issues being assessed

• Impact of pixel aggregation (M15) scheme on fire detection capabilities being quantified – results being used to support modification of platform configuration

• Results indicate that active fire product could perform better than originally thought

• GOES-R/ABI pre-flight active fire data assessment setting the stage for routine post-launch product validation

• Use of fine resolution data building on MODIS experience

• Science quality data being generated in support of regional and global fire monitoring systems

• Validation of fire characterization data (size, temp, fire radiative power) – moving beyond the binary (yes-no) fire detection information

Pending Support and Future Research

• ROSES 2010 Remote Sensing Theory: “Derivation of biomass burning properties based on the synergistic use of MODIS and ASTER global data” (PI: W. Schroeder)

• ROSES 2010 The Science of Terra and Aqua: “MODIS Collection 6 Active Fire Maintenance and Validation” (PI: L. Giglio)

• ROSES 2010 NPP Science Team for Climate Data Records: “The active fire data record from NPP VIIRS” (PI: I. Csiszar)

• GOESR3 : “Development of a blended active fire detection and characterization product from geostationary and polar orbiter satellite data” (Csiszar, Schroeder, Justice)

• Developement and support of fine resolution active fire products derived from Landsat TM, LDCM (2012), ESA Sentinel 2 (2012-2013), and HyspIRI (2017) instruments (Giglio, Csiszar, Schroeder)