Title: Fire Detection and Mapping using Satellite Remote Sensing: Sensors, Systems, and Solutions
1Fire Detection and Mapping using Satellite Remote
Sensing Sensors, Systems, and Solutions
- Robert H. Fraser CCRS
- Robert Landry CCRS
- Ron Hall CFS
- Dieter Oertel DLR
2Rob Fraser Research Scientist Canada Centre for
Remote Sensing Natural Resources Canada
3Sensors and Systems for Active Fire Detection
- (A) NOAA/AVHRR
- Visible, NIR, MIR, 2 Thermal channels at 1 km
- Meteorological sensor, although workhorse for
fire detection for 20 years - Inexpensive to set up (most commonly used sensor
for national fire monitoring systems) - Channel 3 is shared in post N14 satellites (1.6
?m 3a operated during daytime, 3.7 ?m 3b during
night)
4Sensors and Systems for Active Fire Detection
- (A) NOAA/AVHRR Products and Systems
- IGBP-DIS Global Fire Product (1992-1993)
- World Fire Web (1998-2001)
- NOAA NESDIS Operational Significant Event
Imagery and SSD Experimental Fire Analysis page
22 World Fire Web HRPT stations
SSD Experimental Fire Analysis Page
NOAA Significant Event Imagery
5Sensors and Systems for Active Fire Detection
Fire M3 Regional Product over Quebec
- (A) NOAA/AVHRR Products and Systems
- Canadian Fire M3 System (NRCan)
- North American historical (1985-2001) fire
mapping project (NASA LULUC) - National AVHRR fire monitoring systems in Brazil,
Indonesia, Mexico, Australia FIREWATCH, Finland
VTT/ESA/FMI, Russia, and others
Brazil INPE
Australia Firewatch
Finland FireAlarm System
Russia Avialesookhrana
6Sensors and Systems for Active Fire Detection
- (B) ERS-2/ATSR-2
- Channels similar to AVHRR
- ESA/ESRIN created ATSR World Fire Atlas (1995-Jan
2002 latest) - Night time imagery
- Two algorithms based on 3.7 ?m thresholds
- Validated using a network approach
ATSR World Fire Atlas August 2001 ATSR
Algorithm 1
7Sensors and Systems for Active Fire Detection
- (C) GOES
- Geostationary perspective captures diurnal fire
cycle - Experimental Wildfire ABBA Fire Product (CIMSS
and NOAA) - Hotspot monitoring web site at Hawaii Inst.
Geophysics
Wildfire ABBA Regional Overview July 9, 2002
Hawaii Hotspots from Int. of Geophysics
8Sensors and Systems for Active Fire Detection
- (D) DMSP OLS
- NOAA-NESDIS NGDC night time detection using
visible channel and database of stable lights - Provide near real time delivery of OLS data and
algorithms to regional centers
9Sensors and Systems for Active Fire Detection
- (E) Terra and Aqua with MODIS
- Launched in 1999 aboard EOS/Terra, 2002 aboard
Aqua - Only sensor providing global daily fire products
in near real-time - Two channels specifically designed for fire
detection (band 21 _at_ 4 ?m saturates at 450 K,
band 31 at 11 ?m saturates at 400 K) - Sensor will be used for fire detection,
estimating rate of emissions, and
smouldering/flaming ratio
10Sensors and Systems for Active Fire Detection
- (F) MODIS Fire Products and Systems
- Land Rapid Response
- (NASA GSFC / Umd / USFS)
Rapid Response (250m false color) Quebec July
7
University of Maryland Web Fire Maps
July 7-9
11Sensors and Systems for Active Fire Detection
- (F) MODIS Fire Products and Systems
- Land Rapid Response
- at USFS RSAC
USFS Remote Sensing Application Center
Regional Cartographic Products
12Sensors and Systems for Active Fire Detection
- (F) MODIS Fire Products and Systems
- (iii) GEOMAC internet-based mapping tool designed
for fire managers
GEOMAC Internet Map Server
13Sensors and Systems for Active Fire Detection
Other sensors FUEGO constellation BIRD
(DLR) Meteosat Second Generation (MSG) MTSAT-1R
Eventual Geostationary Coverage
14Outstanding Issues for Active Fire Detection
- Validation of satellite fire products
- AVHRR channel 3a/3b sharing
- Need information systems to archive hotspot data
from various sensors/systems (e.g. a global
hotspot clearinghouse building on efforts of NOAA
Fire Detection Program and Global Fire Monitoring
Centre) - Continued research on active fire products
contribution to emissions modelling
15Small Satellite Mission BIRD DLR Hot Spot Images
16Dieter Oertel Institute of Space Sensor
Technology and Planetary Exploration
17The BIRD Payload
- Payload platform of the flight model
- Total mass 30.2 kg
18 Hot spots in Italy (BIRD, Nov. 5, 2001)
Equivalent fire temperature
19Hot spot detection by GOES and BIRD (Nov. 23,
2001)
GOES
BIRD
20BIRD MIR images of bush fires, Sydney area,
Australia obtained on a - January 4, 2002 b -
January 5, 2002 c - January 9, 2002
21Fragment of BIRD bush fire images (Australia,
January 2002)
Equivalent fire temperature
MIR, Jan. 9, 2002
22Typical characteristics of fire fronts (BIRD,
Australia, Jan. 5, 2002)
23 Fire detection by MODIS and BIRD (Australia,
January 5, 2002)
MODIS Fire map
BIRD Fire map
24Industrial hot spots Munich area (BIRD, Jan. 29,
2002)
0.1 1 10 MW
MIR
Radiative energy release
25BIRD Detects Hot Spots in and around Munich
- Infrared Image of Munich region 29 of Jan.2002,
local time 1010 h
Hot spot Wood waste that burned for several
hours (4m diameter, hot temperature) by Farmer J.
Kranz (written in his working diary)
26China coal seam fires (BIRD, January-February
2002)
Radiative energy release
MIR, Feb. 6, 2002
27Easter Fires (BIRD, Steiermark -Kaernten, Austria
March 30, 2002)
Energy release
MIR
28Industrial Hot Spots (BIRD, Ruhr area, Germany,
Mar. 28, 2002)
Thyssen T 530 K, A 2.4 Ha, E 101 MW
T 470 K, A 2.1 Ha, E 51 MW
Energy release
MIR
29USA Fires June 2002
BIRD MIR image obtained at night over New Mexico
/ Colorado on 17 June 2002. Swath width 190 km,
swath length 380 km. Fire are visible in the
upper part (red color) - probably in the San Juan
Natiotional Forest - which is about 290km
North-north-west of Albuquerque. Albuquerque is
shown in the image by blue color, using a 2 K
temperature thresholding to the background
temperature.
30USA Fires June 2002
The San Juan park fire analysis shows 620 MIR
pixels are occupied by fire. The MIR fire-pixel
brightness temperatures are between 300 and
400K. The Bi-spectral method allowed to retrieve
fire temperatures from 650 to 1000 K
31USA Fires June 2002
BIRD MIR image of the Denver / Pike Forest area
obtained 21 June 2002
32USA Fires June 2002
22 remaining small hot spots of the Pike Forest
fire could be identified in the MIR image of 21
June 2002
33VGT Scar Mapping Validation with Landsat-TM
Scar Mapping
- Robert Landry, Heather MacLeod, Don Raymond
- Canada Centre for Remote Sensing
- Ron Hall
- Canadian Forest Service
34Robert Landry Research Scientist Canada Centre
for Remote Sensing Natural Resoures Canada
35Virginia Hills burnt forest
36(No Transcript)
37Fire events distributionper Ecozone
38Burn area product validationExperimental design
Factor 1 Ecozone (6 classes) 4,5,6,9,12,15 Facto
r 2 Fire size (3 classes) 1) 200-400, 2)
400-600, 3) 600 ha) Factor 3 Fuel type (3
classes) 1) conifers, 2) deciduous, 3)
mixed Replicate minimum 3 replicates
39Burn area validationDataset characteristics
- Landsat-TM
- 1998,1999 - over 55 imagery
- Number of fires mapped (over 200 ha. single
polygon detected only) - 197 fires were mapped
40Burn area validation - Sampling dataset
Landsat-TM 1998,1999 - over 55
imagery Number of fires mapped - over 200
ha. - 197 fires were mapped
Landsat-TM dataset for SPOT-VGT mapping
validation
41Burn area product validationDeliverables
- Statistical reports
- UTM - total burn area, polygon size distribution
(fragmentation) - LCC - total burn area, fuel class distribution,
ecozone distribution - Burn vectors
- UTM - smooth vectors (e00)
- LCC - full resolution and 75 meters resampled
(during UTM to LCC projection) for Canada-wide
composite - LCC database, all VGT and TM derived vector are
compiled within the same ArcView project
42Burn area product validation where are we ?
- Data preparation
- Fire event selection based on experiment design
(completed) - Extract burn vectors from Landsat-TM scenes
(completed) - Compile statistics from TM and VGT (completed)
- Define and identify common fire event between TM
and VGT - Analysis of false alarms from VGT/TM (TBD)
- Identify problematic fires (TM and VGT)
(completed) - Reprocess (feedback loop) problematic fires
(completed) - Validation and modeling
- Definition of control fire sample set, and
validation fire sample set - Statistical analysis (regression, 3-factor ANOVA)
- Modeling analysis
- Modeling testing with the validation fire sample
set (loop process) - Results
43Burn area product validation
44Burn area product validationobservations
- Fire history increases ambiguities in the mapping
process, - Humid areas are difficult to map because of
different fuel type, - The algorithm maps complete burn, not partial
burn... In deciduous area (regeneration within
old fire), partial burn signature seems to
dominate, so fragmentation becomes high so more
problematic to map, - False alarms detected from VGT and TM (beaches,
recent depletions)
45Burn area product validationobservations (TM
related)
- within complete burnt area, TM NIR and Mid-IR
spectral signatures seems to be correlated to
stem density, and not necessarily to burn
severity, - time lapse between Landsat-TM acquisition and
fire occurrence has significant impact on
spectral signature, therefore on our ability to
correlate the signature to severity, - fire occurrence within the fire season has an
impact on the observed signature responses
(contribution of the fireweed), - partial burn (lt50 crown burnt, with any
combination of ground burnt) is challenging to
assess (research ongoing)
46Conclusions
- Validation of SPOT-Vegetation with Landsat-TM for
the fire seasons 1998 1999 - over 197 fires across Canada areas comprising
six eco-zones and different fuel types, - still under progress
- plans to present preliminary results (modelling)
during the spring - Fire scar mapping with SPOT-Vegetation will
fulfil other requirements of the relevant
federal, provincial, and forest management
agencies - FireM3 has demonstrated the feasibility of a
complex end-to-end process from satellite
acquisition to web-product, available in almost
real time
47Conclusions
- Fire scar mapping with SPOT-Vegetation and the
derived products will contribute to - National Forest Information System (NFIS)
- The State of Canadas Forests Reporting
Next steps...
- Carbon emission estimate from fire
- through a pilot project...burn severity
48Defense Support Program (DSP) Satellites The
AeroSpace Corporation Source D. Pack et al.
Civilian Uses of Surveillance Satellites
- Features of DSP Satellites
- designed to track missile launches
- scans nearly entire earth every 10 seconds
- uses infra-red detectors
- dynamic range of sensor can measure wide range
- in fire intensity without sensor saturation
- not clear if system can be used to track
- non-USA fires
49Fire Intensity Plot from DSP Satellite of
Topanga-Malibu Fire Nov. 2, 1993 (source D. Pack
The Aerpspace Corp.)
50Diurnal Intensity from DSP Satellite of Burning
Across South Africa July 3, 1992 (source D. Pack
et al. The Aerospace Corp.)
51Thank You