Fire Detection and Monitoring Akli Benali Forest Research Center, University of Lisbon aklibenali@gmail.com Conference & Summer School on Forest Fire Management Istituto Superiore Antincendi, Rome, September 2016
Outline 1. What is Fire Detection and Monitoring? 2. Why is it important? 3. What methods are available? 4. Integration of different methods
Concepts Fire Detection determining that a fire event exists detection of a previously unknown fire confirmatory detection of new fires detection of spot fires detection of rekindlings
Concepts Detection Monitoring
Concepts Fire Monitoring Mapping and characterizing the parameters of a fire (location, fire perimeter, active fronts, size, proximity to water and inhabited areas, etc) Status and changes in an active fire
Importance Fire Detection Early detection is a key element for the success of any fire management program Success of initial attack depends on : time delay, fuels, weather, fire size; Increase the probability that it can be controlled before it develops into a catastrophic event. Reduce suppression costs and negative environmental impacts
Importance Fire Monitoring Once the decision has been made to initiate suppression, emphasis is then placed on fire monitoring Critical for ongoing suppression actions, suppression planning, public warnings, etc. Reduce suppression costs and negative environmental impacts Fire detection systems are used for both the detection and monitoring phases
How do we detect fires? 1. Smoke 2. Heat 3. Flames and Light 4. Temperature & Gases
How do we detect fires? 1. Smoke Most reliable signal for daytime detection; Spectral signature (colour), motion. Characteristics vary with burning and combustion conditions Affected by clouds, fog, haze, shadows.
How do we detect fires? 1. Smoke
How do we detect fires? 2. Heat Strong signal for the detection of fire Large capabilities for nightime detection Emission follows Planck s Law Total energy follows Stefan-Boltzmann s law Short-wave to Thermal Infrared region (SWIR, MIR, TIR)
How do we detect fires? 2. Heat Source: Allison et al. 2016
How do we detect fires? 2. Heat Source: Oertel et al. In press Source: Camrascan HOTSPOT
Available Methods 1. Humans 2. Cameras & Sensors 3. Airborne 4. Spaceborne
Human-based How? Patrols, monitoring towers, general public, ground staff Direct observation of a fire or its smoke Based on the probability of human detection. Humans are involved in almost every detection method
Human-based Influencing factors Sensitivity, attention and experience Fire size and plume shape Smoke density Distance and intensity of the target Atmospheric conditions and occlusions
Human-based Examples Over 95% of fires reported to the NSW Rural Fire Service between 2004 and 2009 were reported by the general public or authorities. In Portugal between 2001 and 2014, 65% of fires were reported by general public and 10% by authorities Crowdsourcing methods (Zhong et al. 2016)
Human-based Advantages and Disadvantages Remote areas are less covered by humans Telephone network may be limited in remote areas High probability of faulty alarms Watch towers: Inflexible, costly, only have daytime coverage, limited spatial coverage.
Cameras & Sensors How? 1) Equipment 2) Detection of fire - Algorithm detects fire and decides whether to send the alarm or not (automatic) Example: Forest Fire Finder (NGNS) - Or there is a trained operator that sends the alarm (supervised) 3) Identifying the Location of the fire 4) Communicate the information
Cameras & Sensors How? Cameras Different types of cameras detect mostly smoke (visible and IR) and heat (thermal IR) Trained staff need to be dedicated to operate the system for fire detection In alternative automatic algorithms can be used Careful decision of what information is passed Detect smoke\fire in a range of 15-80km
Cameras & Sensors How? Sensors Sensors sense parameters such as temperature, pressure and humidity, and gases High monitoring frequency Easily deployed even in inaccessible places Several systems proposed have sensors linked with cameras (Lloret et al. Hartung et al.)
Cameras & Sensors Influencing factors Human error and skill, for supervised systems Technical performance of cameras and algorithms Fire size Topography Atmospheric conditions (visibility) Distance from target
Cameras & Sensors Examples FireWatch Automated fire detection system Detects smoke in a 10-40 km range (up to ~50km) Sensitive to radiation in the visible and near-infrared. Works during the night. Operational FireWatch systems are in use in Germany, Estonia, Cyprus and Mexico. Pilot scale systems are in use in the Czech Republic, Portugal, Spain, Italy, Greece, and the USA.
Cameras & Sensors Examples ForestWatch Semi-automatic detection system Smoke during the day and fire glow at night to an optimal distance of 16 20 km (up to ~60km) Workstation operator is alerted and processes or dismisses the alarm. Operational FireWatch systems : South Africa, Swaziland, USA, Canada, Chile and Slovakia.
Cameras & Sensors Examples Libellium Sensors that measure temperature, humidity and several gases. Optional Camera integrated The system analyzes the information and reacts sending an alarm GPS\GPRS integrated that delivers accurate position and time information The range varies between 500m and 40km
Cameras & Sensors Advantages/Disadvantages Detection times (10-220 min) Location errors (4-18 km) Omissions & false alarms Human observers are faster and more accurate Fail to detect small fires Importance of skilled operators Costs and installation Life span, node dependence (for sensors)
Cameras & Sensors Advantages/Disadvantages Source: Matthews et al. 2009
Cameras & Sensors Advantages/Disadvantages High monitoring frequency Absence of tower staff (i.e. at night) Data recording Remote Locations and/or locations with high value assets Fill in blind spots in existing tower networks Sense additional information (e.g. weather) Future developments are likely to improve performance
Airborne How? Human observers: smoke detection Infrared imaging, can be performed day and night; Often combined with visible light or NIR cameras UAV: visual flame detection or thermal cameras.
Airborne Influencing Factors Camera/sensors performance Altitude Maneauvrability Atmospheric conditions Monitoring frequency Period of the day Fire intensity
Airborne Examples Phoenix line scan system, scans around 4500ha per minute, detects hotspots using MIR and TIR bands as small as 15cm diameter (USFS) TIR cameras in the USFS Firewatch Cobra helicopters for detection and mapping. Red-eye system for helicopter-based detection and monitoring AMS-Wildfire on a UAV with 12 spectral bands
Airborne Advantages/Disadvantages Maneuvrable, deployable, flexible Expensive; Oportunistic detection Require planning aerial detection patrols More suitable for monitoring than detection Depending on the altitude, can be unsuitable to monitor large areas. Free ground crews for other tasks when a fire is under surveillance (mop-up)
Airborne Advantages/Disadvantges UAVs still have limited used in operational context: - Short endurance and range - Stability and Safety - On board processing - Regulation of UAV flight remain unsolved
Airborne Advantages/Disadvantges UAVs have promising capabilities: - Combination of low with high altitude for different objectives - Autonomy (precise or systematic flights) - Complement aircrafts and helicopters - Mop-up, nighttime, heavy smoke, low air traffic
Satellite-based How? Sensors that measure reflectance in the visible and infrared Atmospheric correction Algorithms to extract specific information (active fire location; burned area)
Satellite-based Sensors Fire Detection and Monitoring Landsat NPP-VIIRS MODIS SEVIRI 16 days 2 times per day 4 times per day 10 min Temporal frequency Spatial Resolution 30 m 375 m 1 km 4 km
Satellite-based Influencing factors Revisiting period Spatial resolution Sensor sensitivity / algorithm Viewing geometry Weather conditions (e.g. clouds) Vegetation type Fire size & intensity & duration
Satellite-based Examples: MODIS Terra\Aqua Automatic detection using TIR 4 times per day at a 1km nominal resolution Day and nigthtime detection Can detect small areas burning 50% detection rate for fires larger than 100ha (Hawbacker et al. 2008)
Satellite-based Examples: MODIS Terra\Aqua Detection periods
Satellite-based Deformation Examples: MODIS Terra\Aqua Spatial Resolution Nominal
Satellite-based Examples: MODIS Terra\Aqua Start Fire Date End Fire Date
Satellite-based Examples: MODIS Terra\Aqua
Satellite-based Examples: MODIS Terra\Aqua Time Elapsed
Satellite-based Examples: MODIS Terra\Aqua vs FARSITE
Satellite-based Examples: MODIS Terra\Aqua vs NPP VIIRS MODIS (4x, 1km) VIIRS (2x, 375m)
Satellite-based Examples: Geostationary (GOES, MSG) Recent advances have improved early detection - Kulthov et al. 2016 (GOES-EFD): most fires within 30min (>2ha) and 31% before they were reported - Filizolla et al. 2016 (RST-FIRES): provided the sole fire alert in 348 cases; and in 227 warnings >1h of any other source (N = 950).
Satellite-based Examples: Integrating Satellites EFFIS
Satellite-based Examples: Integrating Satellites AFIS
Satellite-based Examples: Integrating Satellites Worldview
Satellite-based Advantages/disadvantages Large spatial coverage Revisiting periods Spatial Resolution Very Low Cost Clouds Time lag between overpass and availability Particularly suited for monitoring of mediumlarge fires Automatic systems & long term continuity
Integration Complementary system that maximizes the potential of each source and minimizes its limitations Convergence of Evidence Detection Monitoring
Improvement Integration Criteria Accuracy Sampling Frequency Cost Spatial & Temporal Coverage Feasibility Values at risk Existing detection structure Detection and Monitoring system design Aplication, Implementation and Dissemination Evaluation
Thank you
Importance Situational awareness: understand how information, events, and your own actions will impact your goals and objectives, both now and in the near future. In the context of planning, preparation and suppression of wildfires: 1. knowledge of fire locations 2. understanding of current growth and behavior 3. where these fires will go in the near future. Necessary at every level of a fire suppression action
Importance
Evaluation - Ideally a system detects all fire events (high hit rate) and never detects non-fire events (low false alarm rate) - Tradeoff between measures - Sensitivity: Proportion of actual fire events that are detected - Specificity: Proportion of non-fire events that are not detected. - Difficulty of determining ground-truth