EUMETSAT 2004 March 24 th Earth Observation Dep

EUMETSAT 2004, March 24 th Earth Observation Dep. t Monitoring the summer of 2003 forest fires of Spain and Portugal, part I Automatic Fire Detection and Characterization by MSG/SEVIRI A. Bartoloni, E. Cisbani, E. Zappitelli (Telespazio/Rome) B. Greco (ESA/ESRIN)

EUMETSAT 2004, March 24 th TOC / Credit Earth Observation Dep. t v Approach § Assumptions § Fire Equation v Implementation § SEVIRI data § Processor Flow Chart v Performance on real data Credit: - ESA/CDMC: Prototype processor implementation and § Validated Data analysis of real SEVIRI data § Comparison § Fire Evolution Characterization (this presentation) § False Alarm / Efficiency - ESA/Fires. Med: Simulated data v Conclusions § User Needs § Prospect analysis - FIRES/RIT-NASA: Original ideas (for GOES)

EUMETSAT 2004, March 24 th Rationale Earth Observation Dep. t Use of existing Satellite Sensors for fire detection and characterization v Need of frequent revisit § Geo-stationary Polar sensors v Need of high spatial sensitivity § Smart processing § Use Polar sensors data Optimal Sensors (non existing) Geo-stationary sensors

EUMETSAT 2004, March 24 th Approach Earth Observation Dep. t v Procedure based on a simple Radiative Transfer Model Include atmospheric transmissivity, ground emissivity, solar term v Fast changing quantities from geostationary sensor/SEVIRI Background/Fire Temperatures/Size, Atmospheric water vapor content v Improved spatial sensitivity by multi band, sub-pixel exploitation SWIR/MIR/TIR atmospheric windows v Contrast analysis for fire temporal changes enhancement Looking for thermal ground changes v Quasi-stationary parameters from polar sensors Ground emissivity

EUMETSAT 2004, March 24 th Fire Equation Fire Background (t+Dt) Earth Observation Dep. t = (t) Effective Pixel Fraction Change Fire temperature Background temperature Atmospheric and Solar changes neglected (except SWIR) One equation for each atmospheric window in IR/TIR The technique is sensible to fire variations

EUMETSAT 2004, March 24 th Input Data: SEVIRI Radiances MODIS SW IR ch 3 IR ch 4 M s R TI ch 0 7, 1 9, v Overall Excellent Image Quality v Geolocation pretty good § Frame coregistration second order effect in detection v Saturation effects on very large fires (see images) SEVIRI Earth Observation Dep. t

EUMETSAT 2004, March 24 th Implementation: Processor SEVIRI Radiances Ground Emissivity Compute Acquisition Geometry Earth Observation Dep. t Processor: Pixel Based, fully automated and parallelizzable Minimize Fire System of Equations Sea/Land plain Cloud Masking Estimate Bck. Temperature Cuts on: - Minimization Residue - Fire Temperature - Minimum Fire Size - Minimum Power Estimate TPW Evaluate Solar Terms Detected Fire Characteristics

EUMETSAT 2004, March 24 th Data for Processor Characterization Earth Observation Dep. t v SEVIRI/MSG Data: § sparse temporal coverage due to acquisition station malfunctioning § Commissioning data! v Raw Validated Data: § MODIS: ch 21/22 + ch 31 § BIRD: MIR 3. 4 -4. 2 + TIR 8. 5 -9. 3 mm Heavy Cloudy v Test Case: § Portugal Fires beginning of August 2003 v No Ground Truth analyzed ! BIRD Calibrated/Geolocated Images from BIRD Team @ DLR

EUMETSAT 2004, March 24 th Validation Fires v Visually identify hot spots on BIRD MIR image and corresponding ones on MODIS Red: BIRD Yellow: MODIS (4 Aug) Earth Observation Dep. t v Define background and fire ROIs Background Fire v Apply the traditional sub-pixel Dozier method (MIR/TIR) to the fire ROI v Assume as firing pixels those with temperature above 350 -400 K (starting plateau) v Retrieve hot spot characteristics: § fire / background temperature § burning area § fire power § location 04/Aug/03 12: 03 BIRD/MIR

EUMETSAT 2004, March 24 th Validated Data Consistency BIRD – MODIS comparison Earth Observation Dep. t

EUMETSAT 2004, March 24 th SEVIRI Detection on 04/Aug/03 Earth Observation Dep. t 09: 00 09: 15 10: 45 11: 00 11: 15 11: 30 11: 45 12: 00 12: 15 B T 12: 30 12: 45 14: 30 A Available SEVIRI frames (15 minutes apart) on 4/Aug/2003 T: Terra MODIS B: Bird A: Aqua MODIS

EUMETSAT 2004, March 24 th SEVIRI Detection BIRD 12: 03 04/Aug/03 downsampled Earth Observation Dep. t Red: BIRD Yellow: MODIS (4 Aug) Green: SEVIRI (4 Aug) SEVIRI 04/Aug/03 12: 00 upsampled

EUMETSAT 2004, March 24 th BIRD MIR detail Performance Earth Observation Dep. t Box = SEVIRI fire detection (04/Aug/03) Box Area ~ Fire Power SEVIRI 04/08/03 12: 00 upsampled

EUMETSAT 2004, March 24 th Performance Earth Observation Dep. t

EUMETSAT 2004, March 24 th Performance 158 Frame Processed = 898700 pixels 40 -60 False Hits False Alarms: 6· 10 -5/pixel NOTE: - several false alarms associated to cloud boundaries (next) - a bunch of them due to SEVIRI saturation v Processing Time: 20 min whole Iberian Peninsula (on Athlon 1. 7 GHz) Earth Observation Dep. t

EUMETSAT 2004, March 24 th Long lasting hot spots v Detected on Aug 1 st by SEVIRI v Size and Power variation (SEVIRI) generally smaller than corresponding MODIS/BIRD quantities Earth Observation Dep. t

EUMETSAT 2004, March 24 th Large, 1 day fire Earth Observation Dep. t

EUMETSAT 2004, March 24 th A different fire Earth Observation Dep. t

EUMETSAT 2004, March 24 th Long lasting, moving fire v Fire moves from SW to NE v MODIS/SEVIRI/BIRD geolocations do not fully agree Earth Observation Dep. t

EUMETSAT 2004, March 24 th Fire Parameters Distributions 04/Aug/2003 Fires Earth Observation Dep. t

EUMETSAT 2004, March 24 th MIR Movie VIS 08 SWIR TIR 8. 7 TIR 11 Earth Observation Dep. t NOTE: v several frames are mission v The detection is sensible to fire condition changes v Very High radiances in TIR have been not processed

EUMETSAT 2004, March 24 th MIR Noise from clouds VIS 08 SWIR TIR 8. 7 TIR 11 Earth Observation Dep. t

EUMETSAT 2004, March 24 th SEVIRI-MODIS Detection Efficiency Search MODIS images for fire outbreaks between two consecutive acquisitions: TERRA 10: 45 AQUA 14: 00 v 24 fires identified (1 -5 Aug) v 17 Detected by SEVIRI (histogram) v SEVIRI/MODIS Efficiency ~ 70% § some SEVIRI frames are missing Earth Observation Dep. t

EUMETSAT 2004, March 24 th User Needs and Detection Perf. Earth Observation Dep. t Up to 2 hour before MODIS Processor False Alarms: better than 10 -4/pixel HRV channel can be used to improve the location ~70% MODIS efficiency

EUMETSAT 2004, March 24 th Conclusions Earth Observation Dep. t • First results of the prototyped, fully automated, geostationary fire detection processor on SEVIRI commissioning data show: – comparable efficiency with MODIS (visual analysis) – reasonable false alarms rate – fire characterization capabilities • Margin of improvements (at least): – effective cloud masking (reduce false alarms) – better integration of the SWIR channel (replace MIR on saturated pixel) – moving toward a contextual analysis – validation on ground truth data (fire reports) Next generation of SEVIRI sensors may fulfill the tight user requirements