FIMMA
Fire Identification, Mapping and Monitoring from AVHRR

 

I. Csiszar

 

CIRA@NOAA/NESDIS/ORA

 

 

 

 

 


FIMMA status

 

•      Current system: fully automated on NT and LINUX platforms including improved navigation

 

 


Automatic fire monitoring from AVHRR

 

•      Input data

 

•      Preprocessing

 

•      Fire detection

 

•      Data dissemination

 

 


Input data

 

•      AVHRR level-1b files from the WWB Polar Data Server

–    HRPT: High Resolution Picture Transmission

•    direct-readout

–    LAC: Local Area Coverage

•    recorded on board the satellite

 

 

 


Preprocessing

 

•      Radiometric calibration

–    vis/NIR: time-dependent coefficients from vicarious calibration

–    IR: on-flight coefficients

 

•      (Atmospheric and angular corrections)

–    (not done for fire detection)

 

•      Geo-referencing

–    currently: interpolation between anchor points as provided within the level-1b files (from NESDIS operational navigation) -> error can be up to 5 km -> further correction is needed for fire detection

–    future: automatic landmark-based geo-referencing

 

 


Automatic landmark-based geo-referencing

 

•      Commercial software: PCI/GeoComp-n

–    successfully used at the Canada Center for Remote Sensing for operational fire monitoring

 

•      Main processing steps:

–    data input:

•   conversion of level-1b files into PCI “.pix” database format

–    ephemeris update

•   current TBUS for good “first guess” navigation

 

•      Calibration

–    standard procedure as in current operational processing

–    geocoding

•   ground control points (GCP’s) are found automatically from a image chip database - no manual selection of GCP’s is needed

•   geocoded scenes in pre-defined projection

 

 


Automatic nighttime geocoding

 

 


 

Nighttime fire detection (?)

 

•      Nighttime geocoding:

–    criteria for valid geocoding need to be relaxed

•    number of GCP’s found

•    minimum GCP span

•    maximum RMS

–    good “fake” channel data -> more GCP’s found

–    stable hot spots (e.g. Weirton Iron Works etc.)

 

•      Algorithm:

–    low saturation level of ch. 3 on pre-KLM satellites

–    higher saturation level, but no daytime ch. 3b data on KLM satellites -> need to develop near-terminator and nighttime algorithms and nighttime geocoding

 

•      Validation:

–    no quantitative validation yet

 

 

 


 

Automatic processing

 

•      In NT/MS-DOS environment (Perl scripts and EASI/PACE routines)

–    Batch processes for

•   acquisition of latest TBUS bulletins (“TBUS”)

•   acquisition of latest level-1b files and preparing for geocoding (“Level1b”)

•   submitting files for geocoding (“Geocode”)

•   exporting geocoded scenes into binary format

 

•      In LINUX environment (shell scripts)

–    Cron jobs for

•   running fire detection

•   forwarding results

 

 


 

The fire detection algorithm: physical basis

 

•      Rapid increase of mid-IR  (3.5-4.0 um) radiance with fire temperature
B(
8, T)=C1/ [85(eC2/8T -1)]

•      AVHRR channel 3(b): 3.7 micron

 

 


 

Problems in AVHRR-based fire detection

 

•      Obscuration

–    clouds

–    thick smoke (wet)

 

•      False signals

–    hot/bright surfaces: heat + solar reflection

–    sun glint (direct reflection) from water bodies

–    cloud edges: solar reflection + emission from underlying hot surface

 

 


 

Current Algorithm: Fire Identification, Monitoring and Mapping from AVHRR (FIMMA)

 

•       Based on NASA Goddard Space Flight Center two-step day-time scheme
(Justice et al. 1996) 

–    Potential fire pixels     Original    Modified
                  T3 >316K; T4 >280K   T3 >316K; T4 >280K
                                            T3-T4 >12K
                                            if glint angle<40o then
                                            R1<30% and R2<30%
Valid background   T3<316; T4>280   T3<316; T4>280  
Calculate mean and
F for                                              {)T},{T3},{T4}                                          
    
)T=T3-T4

–    Classification
)T' ={)T}+2F({)T})   )T> )T' ; T4> {T4}   )T> )T'; T>{T4}      or 5K
T'3 ={T3}+ 2
F({T3})                 T3>T'3 or 325K                                                      
                                                                
                                     
                                                                                                                                    

 

 


 

Remaining issues

 

•      Algorithm-related:

–    imperfect cloud detection

–    commission errors

–    omission errors

 

•      Sensor-related:

–    low saturation level of ch. 3 on pre-KLM satellites

–    higher saturation level, but no daytime ch. 3b data on KLM satellites -> need to develop near-terminator and nighttime algorithms and nighttime geocoding

 

•      Validation:

–    no quantitative validation yet

–    limited ground truth data

 

 


 

Data dissemination

 

•      Experimental/semi-operational

–    ASCII files with latitude, longitude of fire locations

–    being moved into operations by OSDPD/SSD

 

•      Data are being integrated into the Hazard Mapping System (HMS)