1. An Improved Algorithm of Wildfire Detection and a Method of Wildfire Observation for Validation Koji Nakau Ph.D (JAXA/EORC) Masami Fukuda Ph.D (Fukuyama City Univ.) Hiroki Eto (Japan Airlines Co., Ltd.) nakau.koji@jaxa.jp 1

2. Importance of Satellite Infomation on Wildfire Importance of Wildfire –Environment: Severe impact on ecology, Soil or Peat –Climate: Non-negligible GHGs emission –Disaster: Needs for saving life, welth and society Roles of Satellite imagery for Wildfire –Information for Fire Service for Efficient activity –High resolution is demanded for judge accessbility. –Map or image are not distributable in developping countries. –More verbal information is demanded. –Fire is small and rare event Getting ground based validation data is not so easy Especially comparison with observation by human eyes

3. Aim of This Research Improve fire detection algorithm –To improve, not detected fire is important Observe wildfire from cockpit –Air carrier can ovserve wide area everyday Utilization of future sensors –SGLI (Second generation Global Imager) –CIRC(Compact InfraRed Camera) 3

4. JAXA’s Activity Related to this Research 4 Fire Suppression Data Providing Fire Detection Validation Data Fire Detection Algorithm Sentinel Asia JICA-JST AFS JAL Fire Obs. AFS Fire Info Wild Fire Monitor System JICA-JST Throughout from Algorithm to Social Implementation Observ.  Algorithm  System  Social Impr. GCOM-C MODIS CIRC@ISS/ALOS2 Astronaut ? JASMES Fire Location in AK Fire Location in INA Observation by human eye High res. Thermal IR sensor Fire Suppression in AK Fire Suppression in Indonesia Distribute Disaster Infomation Global Wildfire Detection

5. Wildfire Detection Algorithm 5

6. Approach to Detect Fire 6 We need to estimate radiation from fire to detect fire. Radiation from Fire Emission from Background Reflectance of Background Total Radiance of Satellite Observation What we need is Radiation from Fire Estimated by Thermal Infrared Estimated by SWIR Rad BG.Ref Rad BG.Ems Rad Fire 4  m Rad 4  m

7. Distribution of Hotspots 7

8. Result of Fire Detection  Algorithm developed  Reflection from background land cover considered ▪Ancillary algorithms are improved ; Cloud cover, Snow mask  Doubled S/N ratio (comparing to MOD14) ▪80% more HS & 10% less False Alarm (2004 AK day MODIS) ▪Smoldering, small fire or slush and burn  Different Geographical Distribution  Next Step  Validation  Detect Smoldering (smoldering is not detected) 8

9. Wildfire observation by JAL pilots 9

10. Procedure of wildfire observation by JAL pilots When finding a wildfire, a pilot send a report to ground staff. This report is stored in database. Including location of aircraft, distance to wildfire and others. Authors estimated the location of wildfire using this report. However, utilization of these report does not go straight. –location is not so relyable Extraction of relyable location from reports needed. A pilot send a report to ground staff Ground staff input on the web form Data baseValidation

11. Sample of observed wildfire 11

12. Estimating fire location from a fire report by pilot 12 Airway

13. Wildfire reports in 2010 Wildfire observation in 2011 in Russia

14. Result of wildfire observation Total 990 reports are submitted from pilots in these 8 years. We needed trial and error to utilize this data. Therefore, analysis is still undergoing. Year Duration JPN–EUR/America daysreport 20031 Jun - 31 Jul62101 200417 May - 31 Jul7677 200523 May - 5 Aug7530 200612 Jun - 12 Aug62138 200712 Jun - 12 Sep62172 200812 Jun - 12 Sep93263 20091 Jun - 1 Sep93158 2010 27 Jul ～ 27 Sep 6251

15. Wildfire on 07 Aug 15

16. Wildfire on 07 Aug (with MODIS visible RGB on 07 Aug) 16

17. Wildfire on 07 Aug (with RGB=MODIS:4  m/2.2  m/1.2  m on 07 Aug) 17

18. Wildfire on 07 Aug (with RGB=MODIS:4  m/2.2  m/1.2  m on 07 Aug) 18

19. Fraction of detection by satellite for observed fire 19 Location of observation is not so accurate. –# of fail to detect may include such an error Overwrapped fire location might be relatively reliable. – previous slide is an example

20. Summary Proposed algorithm performed better –For the observed fire by JAL Reliable observed fire location extracted –Observed fire pixel failed to detect by satellite will be extracte based on this method.

21. Utilization of future sensors SGLI (Second generation Global Imager) CIRC(Compact InfraRed Camera) 21

22. CIRC/CALET on JEM/ISS FY2013 Swath : 80km Resolution : 115m SGLI on GCOM-C1 FY2015 Swath : 1150km Resolution : 250m CIRC on ALOS-2 FY2013 Swath : 130km Resolution : 200m Integrative Analysis of middle-high resolution IR sensor datasets. Submitting the latest fire fighting aid data JAXA’s sensors to be available for wildfire Collaboration with UNIFORM

23. MODIS 2009-09-26 CIRC (simulated from TM) 2009-09-26 Locations of burning fire can be identified as. SGLI (250m) ＋ CIRC (115m) for accurate fire location 現在主流の MODIS では、 場所が 1km 単位でしか 分からない Hotspot by MODIS

24. Summary Importance of Wildfire Management –Fire emit ¼ to ½ as much as GHG by human JAXA Wildfire Monitoring –JAXA will launch several sensors available for wildfire monitoring from FY2013 –Algorithm with better performance Wildfire observation by JAL –990 ovservation has reported. –Reliability of location is improving –Identified fire pixels failed to detect fire