Capabilities of airborne infrared remote sensing systems to detect hotspots

Transcription

1 NOT Restricted to FERIC Members nd Prtners Mrch 2004 Contents 1 Introduction 2 Objectives 2 Study methods 3 Test flight detils 4 Results nd discussion 10 Conclusions 10 References 11 Acknowledgements Cpbilities of irborne infrred remote sensing systems to detect hotspots Abstrct The Forest Engineering Reserch Institute of Cnd (FERIC) conducted study to ssess the relibility, effectiveness, nd cost of irborne infrred remote sensing, mpping, nd nlysis systems used in Albert. Field trils were conducted to develop ssessment methods nd to fcilitte preliminry tests of the hotspot detection cpbilities of two irborne infrred remote sensing systems: n AWIS (Airborne Wildfire Intelligence System) mounted on twin-engine ircrft, nd FLIR 2000 mounted on birddog ircrft. Logistic regression ws used to develop probbility of detection models for ech of the sensor nd ircrft pltforms evluted. This report lso describes the development of simple model tht llows fire mngers to ssess the cost effectiveness of lterntive technologies. Keywords Infrred, Fire control, Fire detection, AWIS, FLIR, Remote sensing, Sensors, Thermogrphy. Author Judi Beck, Protection Brnch, B.C. Ministry of Forests. Formerly with FERIC s Wildlnd Fire Opertions Reserch Group. Introduction Knowledge of the fire behviour chrcteristics t vrious loctions long the fire perimeter is vitl to the sfe nd efficient deployment of fire suppression resources, but wildfire detection nd mpping cn be difficult when smoke, hze, nd drkness reduce visibility. The infrred (IR) energy rdited by fire penetrtes smoke, drkness, nd most hze, nd this llows n infrred detector to locte fires tht would otherwise be obscured. Albert Sustinble Reserch Development s (SRD) Forest Protection Division sked FERIC to ssess the detection cpbilities nd limittions of two irborne infrred remote sensing systems: n AWIS (Airborne Wildfire Intelligence System) 1 mounted on twin-engine ircrft, nd FLIR mounted on SRD s birddog ircrft. In Albert, helicopters nd fixed-wing ircrft hve been used to crry infrred sensors (Niederleitner 1976; Ogilvie nd Young 1989). Airborne infrred remote sensing hs been used to support mny fire mngement ctivities such s fire detection, lrge fire reconnissnce, fire perimeter mpping, hotspot detection, nd identifiction of wek points nd breks in fire control lines (Aldrich 1979; Billings 1986; Mtthews 1997; Cmpbell et l. 2002). Hndheld infrred systems, such s severl AGA 3 models, hve been used from helicopter t low ltitude (less thn 500 feet bove ground level) to confirm tht logging residue piles burned in the spring re fully extinguished. Helicopter infrred scnning missions hve lso been 1 AWIS nd the Airborne Wildfire Intelligence System re trdemrks of Rnge nd Bering Environmentl Resource Mpping Corportion, Pender Islnd, B.C. 2 FLIR 2000 is trdemrk of FLIR Systems, Inc., Portlnd, Ore. 3 AGA is trdemrk of FLIR Systems, Inc., Portlnd, Ore.

2 pplied to confirm tht mop-up opertions on lrge fires hve completely extinguished ll hotspots. In 1991, the first gimbled forwrdlooking infrred unit (FLIR 2000) ws mounted on birddog ircrft, which is used in Albert to led n ir ttck group. The birddog crries n ir ttck officer who directs ll irtnker (wter bomber) opertions nd employs the infrred sensor technology to detect gps or wek points in the fire retrdnt line tht is lid by irtnkers long the fire perimeter. Becuse of the potentil to increse the cost effectiveness of irtnker opertions (Woodrd et l. 1993), Albert s entire fleet of birddog ircrft ws equipped with FLIR sensors in These units hve lso been used for missions to scn springburned piles nd to support hotspot detection on lrge fires. The birddog-mounted FLIR (BD FLIR) is combined with printer nd VHS recorder. Tpes for ech mission re kept for evlution nd printed imges cn be dropped to ground crews, lthough the primry function of the birddog is to support irtnker opertions rther thn incident mpping. Moreover, the ppliction of the imged informtion is limited in the bsence of trined thermogrpher or infrred interpreter. In 1999, the AWIS unit becme opertionl. AWIS cquires digitl therml infrred imges nd records of inerti using twinengine ircrft tht is IFR (Instrument Flight Rules) cpble, lthough most scnning is conducted under VFR (Visul Flight Rules) conditions. Proprietry processing softwre is pplied to perform geometric corrections nd produce geo-referenced therml imge mosics, which re then distributed vi the Internet. FERIC s study of the AWIS nd BD FLIR is prt of lrger reserch project to develop methods to evlute nd quntify the relibility, effectiveness, nd cost of vrious irborne infrred remote sensing systems used in wildfire suppression opertions in Albert. Objectives The objectives of this study were to: Develop ssessment methods nd mke preliminry quntittive tests of the detection cpbilities of AWIS nd BD FLIR infrred remote sensing systems. Develop model tht would llow fire mngers to evlute the detection cpbilities of vrious irborne infrred remote sensing systems. Develop model tht would llow fire mngers to ssess the cost effectiveness of lterntive technologies. Study methods Experimentl fieldwork ws crried out ner Colspur, 52 km south of Hinton, Albert. A col sem fire is permnently ctive within the study re. Artificil het sources were lso set up in n djcent re, in the open nd beneth cnopy of mture pine, to provide trgets of known temperture nd size. Litter nd surfce fuels in the open re were very sprse, nd the forest understorey ws comprised lrgely of moss. Trget sizes nd tempertures were vried to determine how these prmeters ffect the detection cpbilities of given sensor. Trget tempertures were designed to rech those typicl of the vrious combustion phses of forest fuels, with tempertures of bout 370, 590, nd 860 C for smouldering, glowing, nd flming combustion, respectively. Cmp Forest Engineering Reserch Institute of Cnd (FERIC) Estern Division nd Hed Office Western Division 580 boul. St-Jen 2601 Est Mll Pointe-Clire, QC, H9R 3J9 Vncouver, BC, V6T 1Z4 (514) (514) dmin@mtl.feric.c (604) (604) dmin@vcr.feric.c Disclimer Advntge is published solely to disseminte informtion to FERIC s members nd prtners. It is not intended s n endorsement or pprovl of ny product or service to the exclusion of others tht my be suitble. Copyright Printed in Cnd on recycled pper. 2 Mrch 2004 Advntge ISSN

3 fuel, propne, nd smll sterno stoves produced trgets of , , nd C, respectively. The sterno stove yields very smll (5.2 cm in dimeter) but hot, open-flme trget. Subtler but lrger hotspots typicl of smouldering nd glowing combustion were emulted using cmp fuel nd propne stoves. To produce uniform het source of known size, round plte of cst iron ws plced on top of the cmp fuel (26.5 or 30 cm in dimeter) nd propne (20 cm in dimeter) stoves. Thermocouples nd dtloggers were used to monitor trget tempertures t one-second intervls for the durtion of the test flight. The trgets were rndomly distributed throughout the study re. The study re ws pproximtely m, nd ws designed so tht it could be detected within single imge t the lowest flight ltitude (pproximtely 2000 feet bove ground level). This would result in minimum of two infrred imge pixels (pproximtely 20 m) between trgets. The ltitude nd longitude of ech trget were recorded using hndheld GPS unit. Smoke bombs were lit to obscure the generl study re nd to serve s generl nvigtionl id for the ircrft. Air temperture nd reltive humidity were mesured periodiclly on the ground throughout the flight trils. Test flight detils Test flights were crried out in August 2001 in the erly morning t pproximtely 05:00 DST, nd in the lte fternoon t pproximtely 17:15. Wether conditions were clm nd cler, but light cloud ws encountered t bout 8000 feet during the morning AWIS test flight. Imge cquisition for ll ltitudes smpled took plce over pproximtely one hour. The detection performnce of both AWIS nd BD FLIR ws worse in the fternoon thn in the erly morning, which confirmed tht infrred detection flights should tke plce in the erly morning. Therefore, the field tril results reflect the morning test flights only, when optimum conditions for infrred scnning existed. Sensor performnce ws evluted t ltitudes of 2000 to 6000 feet. Although AWIS is used for scnning t higher ltitudes, cloud cover precluded tests bove 6000 feet on the morning of the field tril. A summry of the key mission detils is given in Tble 1. Figures 1 nd 2 re typicl imges produced by the AWIS nd BD FLIR systems for these detection missions. Note tht results for the BD FLIR system re vilble immeditely upon ircrft lnding. However, AWIS dt must be processed before nlyticl results re mde vilble, nd imge-processing time increses with the re scnned. b Tble 1. Slient mission detils for AWIS nd FLIR infrred remote sensing field trils Infrred Flight Mission Air Reltive Assumed Ground system time ltitude temperture humidity emissivity pixel re b (ft./m) ( C) (%) (cm 2 ) AWIS 05:03 5: / / / BD FLIR 07:23 08: / n / n / n Emissivity is the rtio of the rdition emitted by surfce to tht emitted by blck body t the sme temperture. Ground pixel re is mesure of the re viewed by single cell within detector rry of sensor on scnning system t given ltitude in given instnt in time. Advntge Mrch

4 Figure 1. Morning therml imge nd hotspots identified by AWIS t 3000 feet AGL. The finl product includes colour mpping detils, not reproduced here, tht highlight fetures such s rods, rivers, nd elevtion contours. Figure 2. Morning therml imge nd hotspots identified by BD FLIR t 4000 feet AGL. N S Results nd discussion Field tril results Eleven trgets were estblished on the ground for the morning AWIS flight trils, nd twelve were estblished for the BD FLIR. The size nd temperture of ech re given t the time of imging (Tbles 2 nd 3). The detection sttus of ech trget indictes whether or not it ws successfully detected. The thermocouples nd dtloggers filed to record trget tempertures t trget sites 1, 11, nd 13 for the AWIS flight trils nd t site 3 for the BD FLIR flight trils. In the cse of dtlogger filures, those trgets not detected hve been clssified with n Unknown detection sttus becuse it could not be confirmed tht trget tempertures were bove mbient. Where stove ws not turned on or rn out of fuel, nd the thermocouple dt indicted tht the temperture of trget ws not significntly higher thn mbient, trgets were clssified s Undetectble. The methods of trget detection for these trils were S typicl of those for opertionl hotspot detection using AWIS nd BD FLIR. To identify trgets, AWIS processes rdiometric vlues utomticlly, N while locl contextul visul ssessment is used to control the qulity of the utomted clssifiction. Dt processing is pplied to detect hotspots utomticlly when trget s rdiometric vlue registers bove specified threshold for the mosic scene (clssified by Threshold detection sttus). Other trgets re identified using visul interprettion to determine when rdiometric or imge vlues exceed bckground vlues in the locl vicinity (clssified by Bckground detection sttus). Therefore, Threshold or Bckground detection sttus indictes tht the sensor ws successful in detecting trget. AWIS pplies other digitl geogrphic informtion to confirm hotspots contextully, to minimize the number of flse trgets identified. The BD FLIR only llows visul interprettion of trgets equivlent to the Bckground detection sttus of AWIS. At ll three test ltitudes, both AWIS nd BD FLIR identified lrge hotspot t the loction of the ctive col sem fire. The very 4 Mrch 2004 Advntge

5 Tble 2. Size nd temperture of ech trget s reported t the time of imging by AWIS Trget Actul Imged Mission Trget Trget Time of Trget Trget to pixel trget trget Detection ltitude identifier exposure smple dimeter re re rtio temperture temperture sttus (feet) (cm) (cm 2 ) (%) ( C) ( C) Open 5:46: Unknown 40.1 Threshold 2 Open 5:46: Not detected 3 Open 5:46: Not detected 6 Open 5:46: Undetectble 7 Open 5:46: Threshold 8 Beneth cnopy 5:51: Threshold 9 Beneth cnopy 5:51: Threshold 10 Beneth cnopy 5:46: Undetectble 11 Beneth cnopy 5:46: Unknown 34.8 Threshold 12 Beneth cnopy 5:46: Bckground 13 Beneth cnopy 5:46: Unknown Unknown Open 5:25: Unknown 28.2 Threshold 2 Open 5:25: Bckground 3 Open 5:25: Not detected 6 Open 5:25: Undetectble 7 Open 5:25: Threshold 8 Beneth cnopy 5:25: Threshold 9 Beneth cnopy 5:25: Bckground 10 Beneth cnopy 5:25: Not detected 11 Beneth cnopy 5:25: Unknown 27.1 Threshold 12 Beneth cnopy 5:25: Bckground 13 Beneth cnopy 5:25: Unknown Unknown Open 5:03: Unknown 13.0 Threshold 2 Open 5:03: Bckground 3 Open 5:03: Not detected 6 Open 5:03: Not detected 7 Open 5:03: Bckground 8 Beneth cnopy 5:03: Bckground 9 Beneth cnopy 5:03: Not detected 10 Beneth cnopy 5:03: Not detected 11 Beneth cnopy 5:03: Unknown 12.5 Bckground 12 Beneth cnopy 5:03: Not detected 13 Beneth cnopy 5:03: Unknown Unknown Trget number 4 filed completely, nd trget 5 ws non-trget. rough terrin (70% slope) ssocited with the col sem precluded ccurte mesurements, but it ws pproximtely m on the ground with highly vrible ctul surfce tempertures of 55 to 250 C. Of the confirmed detectble rtificil trgets, AWIS detected 75, 78, nd 50% of the confirmed detectble rtificil trgets t ltitudes of 3000, 4000, nd 6000 feet, respectively. BD FLIR detected 60, 30, nd 11% of the trgets t ltitudes of 2000, 4000, nd 6000 feet, respectively. The interprettion of BD FLIR informtion lso resulted in the identifiction of one flse trget t 4000 feet nd nother t 6000 feet, wheres no flse trgets were identified by AWIS. Tril results suggest tht hotspot detection cpbilities decline rpidly with ltitude (Tbles 2 nd 3). Mny of the detection filures cn be ttributed to AWIS s inbility to detect the smllest (5.2 cm in dimeter) yet hottest (greter thn 600 C) trgets t ll test ltitudes. The presence or bsence of forest cnopy did not pper to gretly influence detectbility during these trils (Tble 2), lthough trgets were intention- Advntge Mrch

6 Tble 3. Size nd temperture of ech trget s reported t the time of imging by BD FLIR Trget Actul Mission Trget Trget Time of Trget Trget to pixel trget Detection ltitude identifier exposure smple dimeter re re rtio temperture sttus (feet) (cm) (cm 2 ) (%) (C) Open 7:32: Bckground 2 Open 7:32: Not detected 3 Open 7:32: Unknown Unknown 4 Open 7:32: Bckground 6 Open 7:32: Undetected 7 Open 7:32: Bckground 8 Beneth cnopy 7:32: Bckground 9 Beneth cnopy 7:32: Not detected 10 Beneth cnopy 7:32: Not detected 11 Beneth cnopy 7:32: Bckground 12 Beneth cnopy 7:32: Bckground 13 Beneth cnopy 7:32: Not detected Open 7:56: Bckground 2 Open 7:56: Not detected 3 Open 7:56: Unknown Unknown 4 Open 7:56: Not detected 6 Open 7:56: Undetected 7 Open 7:56: Not detected 8 Beneth cnopy 7:56: Bckground 9 Beneth cnopy 7:56: Not detected 10 Beneth cnopy 7:56: Not detected 11 Beneth cnopy 7:56: Not detected 12 Beneth cnopy 7:56: Bckground 13 Beneth cnopy 7:56: Not detected Open 8:08: Bckground 2 Open 8:08: Not detected 3 Open 8:08: Unknown Unknown 4 Open 8:08: Not detected 6 Open 8:08: Undetectble 7 Open 8:08: Not detected 8 Beneth cnopy 8:08: Not detected 9 Beneth cnopy 8:08: Not detected 10 Beneth cnopy 8:08: Not detected 11 Beneth cnopy 8:08: Not detected 12 Beneth cnopy 8:08: Not detected 13 Beneth cnopy 8:08: Undetectble Trget 5 ws non-trget. lly plced within the forest so they would not be directly obscured plnimetriclly. Imged tempertures reported by AWIS, while perhps pplicble for reltive ctegoriztion, do not ccurtely reflect ctul tempertures. This my be of concern if imgery is to support documenttion during litigtion. The low imged tempertures result becuse only smll portion of the imge element (pixel) receives energy from the trget surfce. The energy emitted from the trget surfce is diffused over the whole imge or over multiple pixels due to distnce tht lso ttenutes the signl. This diffusion is pprent in the imgery nd mny pixels (up to bout 20) re ffected by the trget s energy, lthough in theory only mximum of four pixels should be prtilly ffected. More work is required to estblish methods to minimize this diffusion, nd develop ccurte reltionships between imged nd ctul tempertures. 6 Mrch 2004 Advntge

7 Models for probbility of hotspot detection Probbility of detection ws modelled for AWIS nd BD FLIR using logistic regression. Explntory vribles included hotspot temperture (Htemp, C) nd the size of the trget hotspot s proportion of single imge pixel on the ground. The trget-to-pixel re (TPA) defines the proportion of single imge pixel tht is occupied by given hotspot, clculted s: TPA = Hsize/PA 100 where: Hsize = size of the hotspot (cm 2 ) PA = dimension of single imge pixel on the ground (cm 2 ) The re (PA, in cm 2 ) of single imge pixel on the ground is defined by ltitude of the ircrft nd the specifictions of the imge sensor (Mtthews 1997): PA= (TAN(FOVx/2 π/180) Altitude / Px) (TAN(FOVy/2 π/180) Altitude / Py) where: FOVx = FOVy = Px = Py = sensor field of view in degrees in x direction sensor field of view in degrees in y direction number of imge pixels in x direction number of imge pixels in y direction Altitude = ltitude of ircrft in feet bove ground level TAN = tngent of ngle where ngle is given in rdins Seprte models were fitted to the dt given in Tbles 2 nd 3 to describe the probbility of detecting hotspot using AWIS nd BD FLIR: DetectionProb AWIS = 1/(1+EXP(-( LN(TPA) Htemp))) 100 McFdden s Rho-Squred = DetectionProb BD FLIR =1/(1+EXP(-( Htemp LN(TPA)))) 100 McFdden s Rho-Squred = where: EXP= e to the power of given number LN = nturl logrithm of number Slightly different sttisticl models were likely required becuse AWIS involved more imge processing nd interprettion thn those normlly pplied during BD FLIR opertions. Both of these models hve been implemented to produce simple tbles tht cn be used to derive detection probbilities for vriety of scenrios (Appendix I). Both dtsets were lso merged to develop generic probbility of hotspot detection model. The model could be used to ssess the potentil detection performnce of ny sensor t given ltitude, when trined thermogrpher is vilble to interpret the imgery. DetectionProb = 1/(1+EXP(-( LN(TPA) Htemp))) 100 McFdden s Rho-Squred = By vrying sensor specifictions nd flight ltitudes, this theoreticl model cn be used to evlute vriety of sensors in vriety of ircrft, including sensors used in helicopters. Models for cost effectiveness A simple model hs been developed to llow fire mngers to compre the costs of vrious infrred remote sensing systems nd worklods, in the form of flight hours. The model ws not designed to produce definitive mission costs, but rther cn be used to compre the reltive costs nd mission flight times for vrious sensors nd ircrft pltforms. Users re required to define the detection probbility objectives for the mission, e.g., to detect 90 out of 100 smouldering hotspots (300 C) tht re 25 cm 2 in size or greter. Advntge Mrch

8 The probbility of detection models developed in this study re then pplied, using the pproprite sensor specifictions (FOVx, FOVy, Px, nd Py), to determine the highest flight ltitude tht will meet the detection probbility objectives for the mission. In Tble 4, smple user inputs re given for severl infrred sensors the AWIS nd FLIR, s well s the AGA 750 hndheld sensor mounted in helicopter. Preliminry model outputs re clculted to determine the optimum mission ltitude nd re-scnning rte (re/h) for ech sensor/pltform configurtion. Although forml field trils hve not been crried out to evlute the performnce of rotry wing nd hndheld scnning opertions, theoreticl costing scenrio hs been included here to illustrte how the models cn be expnded to consider how vrious ircrft nd infrred scnner specifictions might impct scnning costs. The scene width of rotry wing nd hndheld scnner opertions my be expnded gretly, possibly up to double, by the opertor pnning the scnner. However, this scenrio uses conservtive single scene width becuse the exct impct of pnning on effective scene width is unknown. The model clcultes the time required by specific ircrft to scn given re, bsed on the optimum ltitude nd the opertionl scnning speed of the ircrft. Opertionl ircrft scnning speeds (km/h), ircrft costs ($/h), nd ferry times re entered by the user nd pplied to derive totl mission flight time (h) nd cost for fire of given re or perimeter length. Smple model outputs re given in Tble 5. Model users must check the scene width to ensure it meets minimum scn width of 100 m. This width would be required to support mop-up opertions. In this cse, the flight times nd costs for the perimeter scnning missions must be tripled since the scene width is less thn the minimum 100 m required. This cn only be chieved by three psses t optimum ltitude. Users should lso be wre tht minimum flight ltitudes for BD ircrft re 75 feet under VFR conditions during the dy, nd 500 or 1000 feet under VFR or IFR conditions over flt ground t night. AWIS missions re lso limited to 1000 feet under IFR conditions over flt ground t night, but opertionl missions re rrely undertken below 3000 feet bove ground level. In both cses, nighttime Tble 4. Smple user inputs used to clculte optimum mission ltitude, width, nd re-scnning rte Aircrft nd infrred sensor specifictions AWIS BD FLIR 2000 Rotry wing & AGA 750 Inputs FOVx n FOVy n Pixels cross (x) n Pixels down (y) n Hotspot size (cm 2 ) Probbility of detection Hotspot temperture ( C) Aircrft scnning ($/h) Scnning speed (km/h) Ferry time (h) Clculted vlues Altitude (feet bove ground level) Scene width (m) Are/h (m 2 /h) AWIS sensor specifictions re proprietry informtion nd hence re not included here. 8 Mrch 2004 Advntge

9 Tble 5. Totl mission flight times nd costs for fire of given re or perimeter length when ircrft scnning speeds, ferry times, nd ircrft costs re input Fire Equivlent fire Fire mission flight time (h) Flight costs ($) re perimeter Rotry wing Rotry wing (h) (km) AWIS BD FLIR & AGA 750 AWIS BD FLIR & AGA Flight times in bold indicte tht it is unlikely tht the mission will be completed in single morning. missions would usully be crried out under VFR conditions. Severl ssumptions hve been mde in the development of this model: The mximum fesible morning mission durtion is 1 h for VFR ircrft. The mximum fesible morning mission durtion is 4 h for IFR ircrft. Flight times nd costs re bsed on fire re or perimeter scns ssuming perimeter scn width of 100 m. The rotry wing nd hndheld scnner scenrio does not reflect the fct tht the scene width cn be expnded if the opertor pns the scnner. Costs re bsed on ircrft costs only (fuel in). There is zero turnround time t the edge of flight pths. The results suggest tht for fires less thn 150 h or when perimeter scnning in support of mop-up opertions is less thn bout 15 km, rotry wing missions with n AGA 750 re more cost effective thn AWIS or BD FLIR. The cost of BD FLIR opertions is less thn hlf tht of AWIS until fires exceed bout 2000 h or perimeter scnning exceeds 200 km. Advntge Mrch

10 Conclusions Field trils were used to develop methods nd models to ssess nd quntify the detection cpbilities of infrred systems such s AWIS nd BD FLIR. The trils showed tht AWIS detected 75, 78, nd 50% of the detectble rtificil trgets t ltitudes of 3000, 4000, nd 6000 feet, respectively. BD FLIR detected 60, 30, nd 11% of the trgets t ltitudes of 2000, 4000, nd 6000 feet, respectively. The interprettion of BD FLIR informtion lso resulted in the identifiction of one flse trget t 4000 feet nd nother t 6000 feet, wheres no flse trgets were identified by AWIS. Field tril results suggest tht there re limittions for the ppliction of AWIS nd BD FLIR to support mop-up opertions by wy of the detection of subtle hotspots. Such detection limittions my or my not be criticl depending on user expecttions. Subsequent field trils could include trgetto-pixel re rtios of bout 5 10%, which would expnd the rnge of conditions evluted. Nonetheless, the dt collected hve been pplied to estblish detection threshold models, which cn be used to indicte the temperture nd size of minimum detectble hotspots for given ltitude. These models re useful for mission nd flight plnning during wildfire opertion. Tril results suggest tht hotspot detection cpbilities decline rpidly with ltitude, which is expected given the decrese in trget-to-pixel re rtio. Mny of the detection filures cn be ttributed to the inbility of AWIS to detect the smllest (5.2 cm in dimeter) yet hottest trgets (greter thn 600 C) t ll test ltitudes. The presence or bsence of forest cnopy did not pper to gretly influence the success of detection during these trils, lthough trgets were intentionlly plced within the forest so they would not be obscured plnimetriclly. The probbility of detecting hotspot using infrred remote sensing techniques depends on the size nd temperture of the hotspot, the specifictions of the infrred sensor, nd the ltitude t which scnning is conducted. The probbility of detection models developed cn be pplied to set infrred remote sensing mission objectives, nd to help fire mngers ssess the cost effectiveness of lterntive detection technologies. Results of the costing models developed suggest tht rotry wing missions with n AGA 750 re more cost effective thn AWIS or BD FLIR for fires less thn 150 h or when perimeter scnning in support of mop-up opertions is less thn bout 15 km. BD FLIR opertions cost less thn hlf tht of AWIS until fires exceed bout 2000 h or perimeter scnning exceeds 200 km. Therefter, the costs for AWIS nd BD FLIR missions become more similr, lthough the products nd informtion delivered to fireline stff using these services my differ gretly. References Aldrich, R Remote sensing of wildlnd resources: stte-of-the-rt review. U.S. Deprtment of Agriculture, Forest Service, Rocky Mountin Forest nd Rnge Experiment Sttion, Fort Collins, Colo. Generl Technicl Report RM-71. Billings, P Opertionl spects of the infr-red line scnner. Deprtment of Conservtion nd Environment, Victori, Austrli. Fire Protection Brnch Reserch Report No. 26. Cmpbell, D.; Born, W.; Beck, J.; Beresk, B.; Frederick, K.; Hu, S The irborne wildfire intelligence system: decision support tool for wildlnd fire mngers in Albert. Society of Photo- Opticl Instrumenttion Engineers (SPIE). Pper number Mtthews, A FIRESCAN: technique for irborne infrred mpping of wildfires. Ph.D. thesis prepred for the Deprtment of Erth Sciences, Monsh University, Austrli. 10 Mrch 2004 Advntge

11 Niederleitner, J Detecting holdover fires with the AGA Thermovision 750 infrred scnner. Cndin Forest Service, Northern Forest Reserch Centre, Edmonton, Alt. Informtion Report Nor-X-151. Ogilvie, C.J.; Young, R.W Dedlus line scnner trils in Albert, 1985 nd Forestry Cnd, Northern Forestry Centre, Edmonton, Alt. Informtion Report NOR-X-298. Woodrd, P.M.; Admowicz, W.L.; Bolster, O.J An economic evlution of forwrd looking infrred (FLIR) technology to enhnce eril suppression of forest fires in Albert. Forestry Cnd, Northern Forestry Centre, Edmonton, Alt. Cnd-Albert Prtnership Agreement in Forestry Report 105. Acknowledgements Thnks to those who ssisted with field tril work including Ry Ault, Gry Dkin, Greg Bxter, nd Rex Hsieh of FERIC; Chuck Ogilvie, formerly of the Cndin Forest Service; Weldwood of Cnd Limited, Hinton Division; Mrk Ackermn nd Shirley Niven of the University of Albert; Dennis Driscoll nd Wlly Born of Albert Sustinble Resource Development; nd stff t the Edson Wildfire Mngement Are. Advntge Mrch

12 Appendix I Tbles to predict probbility of hotspot detection Probbility of hotspot detection using AWIS Altitude b Hotspot size Hotspot temperture ( C) (ft.) (cm 2 ) b Numbers indicte tht the probbility of detection is greter thn 95%, except those tht re itlicized indicte 0 70% probbility, nd those tht re bolded indicte 70 95% probbility. Above ground level. 12 Mrch 2004 Advntge

13 Appendix I Tbles to predict probbility of hotspot detection (continued) Probbility of hotspot detection using AWIS Altitude b Hotspot size Hotspot temperture ( C) (ft.) (cm 2 ) b Numbers indicte tht the probbility of detection is greter thn 95%, except those tht re itlicized indicte 0 70% probbility, nd those tht re bolded indicte 70 95% probbility. Above ground level. Advntge Mrch

14 Appendix I Tbles to predict probbility of hotspot detection (continued) Probbility of hotspot detection using FLIR 2000 Altitude b Hotspot size Hotspot temperture ( C) (ft.) (cm 2 ) b Numbers indicte tht the probbility of detection is greter thn 95%, except those tht re itlicized indicte 0 70% probbility, nd those tht re bolded indicte 70 95% probbility. Above ground level. 14 Mrch 2004 Advntge

15 Appendix I Tbles to predict probbility of hotspot detection (continued) Probbility of hotspot detection using FLIR 2000 Altitude b Hotspot size Hotspot temperture ( C) (ft.) (cm 2 ) b Numbers indicte tht the probbility of detection is greter thn 95%, except those tht re itlicized indicte 0 70% probbility, nd those tht re bolded indicte 70 95% probbility. Above ground level. Advntge Mrch