SEVIRI Active Fire products HansJoachim Lutz hansjoachimeumatsat int

SEVIRI Active Fire products Hans-Joachim Lutz hansjoachim@eumatsat. int Slide: 1 Fire Workshop, Sofia, 7 – 10 Sep 2009

SEVIRI Active Fire products • Introduction • Description of the algorithm • Results • Some critical issues • Future plans • MTG • Summary and Conclusions Slide: 2 Fire Workshop, Sofia, 7 -10 Sep 2009

SEVIRI Active Fire products Description of the algorithm Slide: 3 Fire Workshop, Sofia, 7 -10 Sep 2009

SEVIRI Active Fire products The active fire detection algorithm is based on a simple threshold test technique, using SEVIRI channels 4 (3. 9 μm) and 9 (10. 8 μm). The algorithm is applied for all SEVIRI repeat cycles (15 Minutes) and for all land surface pixels, excluding desert/bare soil surface pixels and coastal pixels. The algorithm does not need any cloud masking as an input. However it uses SEVIRI channel 1 (0. 6 μm) to eliminate (low) clouds with a high reflectance. Slide: 4 Fire Workshop, Sofia, 7 -10 Sep 2009

SEVIRI Active Fire products VIS 0. 6 global view and AOI from 15 Jan 2009, 12 UTC Slide: 5 Fire Workshop, Sofia, 7 -10 Sep 2009

SEVIRI Active Fire products IR 10. 8 Slide: 6 Fire Workshop, Sofia, 7 -10 Sep 2009 15 Jan 2009, 12 UTC IR 3. 9

SEVIRI Active Fire products The following threshold tests are used: 1. Brightness temperature IR 3. 9 (2. Standard Deviation IR 3. 9) (3. Standard Deviation IR 10. 8) 4. Standard Deviation difference (Std. Dev 3. 9–Std. Dev 10. 8) 5. Brightness temperature difference (IR 3. 9 -IR 10. 8) Slide: 7 Fire Workshop, Sofia, 7 -10 Sep 2009

SEVIRI Active Fire products Std. Dev 3. 9 Slide: 8 Fire Workshop, Sofia, 7 -10 Sep 2009 15 Jan 2009, 12 UTC Std. Dev 3. 9 -Std. Dev 10. 8

SEVIRI Active Fire products BTD(3. 9 -10. 8) Slide: 9 Fire Workshop, Sofia, 7 -10 Sep 2009 15 Jan 2009, 12 UTC Std. Dev 3. 9 -Std. Dev 10. 8

SEVIRI Active Fire products The setting of the thresholds is crucial for the fire detection !!! In our scheme we use: Forecast data and RTTOV for tests 1 and 5 Static thresholds (corrected for viewing angle) for tests 2, 3, and 4. HOWEVER: Slide: 10 Fire Workshop, Sofia, 7 -10 Sep 2009

SEVIRI Active Fire products Emissivity plays an important rôle • Emissivity more variable near 3. 9 m • Sandy areas appear 5 -10 K cooler at IR 3. 9 than at IR 10. 8 (at night, dry atmosphere) IR 3. 9 Dry sand: 0. 8 Slide: 11 Fire Workshop, Sofia, 7 -10 Sep 2009 IR 10. 8 0. 95

SEVIRI Active Fire products Emissivity map channel 4 (3. 9 µm) for January Slide: 12 Fire Workshop, Sofia, 7 -10 Sep 2009

SEVIRI Active Fire products 1. Use RTM output (RTTOV) with the following parameter: TOA clear sky radiance (ε=1) Surface skin radiance (ε=1) Downward radiance at surface level Total Transmittance 2. Interpolate the gridded data to pixel location 3. Calculate the atmospheric contribution to the TOA radiance 4. Calculate the correct surface skin radiance using the emissivity, the reflected downward radiance at surface level and the reflected solar radiance (at daytime) Slide: 13 Fire Workshop, Sofia, 7 -10 Sep 2009

SEVIRI Active Fire products r. TOA = B(Ts). s + rdown. (1 - s). s + rsolar. s. (1 - s). s + r. ATM please note: s has been calculated for the satellite viewing angle and is not corrected for the solar zenith angle. that no bi-directional effects are included (1 - s) Slide: 14 Fire Workshop, Sofia, 7 -10 Sep 2009

SEVIRI Active Fire products Results Slide: 15 Fire Workshop, Sofia, 7 -10 Sep 2009

SEVIRI Active Fire products The following threshold tests are used: 1. Brightness temperature IR 3. 9 (2. Standard Deviation IR 3. 9) (3. Standard Deviation IR 10. 8) 4. Standard Deviation difference (Std. Dev 3. 9–Std. Dev 10. 8) 5. Brightness temperature difference (IR 3. 9 -IR 10. 8) Slide: 16 Fire Workshop, Sofia, 7 -10 Sep 2009

TB (3. 9 m) - TB (10. 8 m) SEVIRI Active Fire products Slide: 17 Fire Workshop, Sofia, 7 -10 Sep 2009 • Its strong sensitivity to sub-pixel "hot areas" makes the IR 3. 9 channel very useful in fire detection. • If only 5% of the pixel is at 500 K, the IR 3. 9 channel measures 360 K, while the IR 10. 8 measures less than 320 K.

SEVIRI Active Fire products BTD (3. 9 - 10. 8) 15 Jan 2009, 12 UTC For fire (and most clouds) BT 3. 9 >> BT 10. 8 For most other surfaces BT 3. 9 ≈ BT 10. 8 Slide: 18 Fire Workshop, Sofia, 7 -10 Sep 2009

SEVIRI Active Fire products Results of test 5 (BTD of ch. 4 and 9) for possible fire 15 January 2009 12 UTC Slide: 19 Fire Workshop, Sofia, 7 -10 Sep 2009

SEVIRI Active Fire products Std. Dev 3. 9 -Std. Dev 10. 8 15 Jan 2009, 12 UTC For fire Std. Dev 3. 9 >> Std. Dev 10. 8 For clouds/cloud edges Std. Dev 3. 9 ≤ Std. Dev 10. 8 For other surfaces Std. Dev 3. 9 ≈ Std. Dev 10. 8 Slide: 20 Fire Workshop, Sofia, 7 -10 Sep 2009

SEVIRI Active Fire products Results of test 4 (Standard Dev. Diff. of ch. 4 and 9) added to test 5 for possible fire 15 January 2009 12 UTC Slide: 21 Fire Workshop, Sofia, 7 -10 Sep 2009

SEVIRI Active Fire products Results of test 2 and 3 (Standard Dev. of ch. 4 and 9) added to tests 4 and 5 for possible fire 15 January 2009 12 UTC Slide: 22 Fire Workshop, Sofia, 7 -10 Sep 2009

SEVIRI Active Fire products BT 3. 9 15 Jan 2009, 12 UTC Slide: 23 Fire Workshop, Sofia, 7 -10 Sep 2009

SEVIRI Active Fire products Results of test 1 (BT of ch. 4) added to tests 2 - 5 for possible fire 15 January 2009 12 UTC Slide: 24 Fire Workshop, Sofia, 7 -10 Sep 2009

SEVIRI Active Fire products Results of test 5 (BTD of ch. 4 and 9) for probable fire 15 January 2009 12 UTC Slide: 25 Fire Workshop, Sofia, 7 -10 Sep 2009

SEVIRI Active Fire products Results of test 4 (Standard Dev. Diff. of ch. 4 and 9) added to test 5 for probable fire 15 January 2009 12 UTC Slide: 26 Fire Workshop, Sofia, 7 -10 Sep 2009

SEVIRI Active Fire products Results of test 2 and 3 (Standard Dev. of ch. 4 and 9) added to tests 4 and 5 for probable fire 15 January 2009 12 UTC Slide: 27 Fire Workshop, Sofia, 7 -10 Sep 2009

SEVIRI Active Fire products Results of test 1 (BT of ch. 4) added to tests 2 - 5 for probable fire 15 January 2009 12 UTC Slide: 28 Fire Workshop, Sofia, 7 -10 Sep 2009

SEVIRI Active Fire products Final results combining all possible and probable fires 15 January 2009 12 UTC Slide: 29 Fire Workshop, Sofia, 7 -10 Sep 2009

SEVIRI Active Fire products Diurnal cycle of number of fires for 15 January 2009 Slide: 30 Fire Workshop, Sofia, 7 -10 Sep 2009

SEVIRI Active Fire products Some critical issues Slide: 31 Fire Workshop, Sofia, 7 -10 Sep 2009

SEVIRI Active Fire products Artefacts coming from Digital Filter 18: 00 18: 15 18: 30 18: 45 MSG-1, 7 August 2006, IR 3. 9 Channel (inverted) Slide: 32 Fire Workshop, Sofia, 7 -10 Sep 2009

SEVIRI Active Fire products Impact of Sunglint • There is very strong reflection of solar radiation at 3. 9 m (sunglint) • Features such as rivers in sunglint are obvious, but “illuminated” lakes could take on the appearance of fires Slide: 33 Fire Workshop, Sofia, 7 -10 Sep 2009

SEVIRI Active Fire products Midday sun glint over the Congo river MSG-1, 24 March 2004, 09: 00 UTC, Channel 04 (3. 9 m) Slide: 34 Fire Workshop, Sofia, 7 -10 Sep 2009

SEVIRI Active Fire products Frequency of Saturated IR 3. 9 Pixels Slide: 35 Fire Workshop, Sofia, 7 -10 Sep 2009

SEVIRI Active Fire products Working without cloud mask – “hot clouds” 15 January 2009 12 UTC Low-level water clouds are relatively warm They have a high reflectance in ch. 4 (3. 9 µm) These clouds appear as “hot spots” in ch. 4 Slide: 36 Fire Workshop, Sofia, 7 -10 Sep 2009

SEVIRI Active Fire products Working without cloud mask – “hot clouds” 15 January 2009 12 UTC The BTD ch. 4 (3. 9 µm) – ch. 9 (10. 8 µm) These clouds appear as “hot spots” in ch. 4 Slide: 37 Fire Workshop, Sofia, 7 -10 Sep 2009

SEVIRI Active Fire products (Not really) Working without cloud mask – “hot clouds” 15 January 2009 12 UTC Using channel VIS 0. 6, helps to filter out all “hot clouds” Slide: 38 Fire Workshop, Sofia, 7 -10 Sep 2009

SEVIRI Active Fire products Future plans Slide: 39 Fire Workshop, Sofia, 7 -10 Sep 2009

SEVIRI Active Fire products 1. Use the improved ECMWF forecast model The operational algorithm works with a 1º / 6 -hourly forecast The prototype algorithm uses a 0. 25º / 3 -hourly forecast In the near future ECMWF will provide a 0. 125º forecast grid 2. Include solar transmittance 3. The transmittance calculation for channel 4 is only done with the viewing 4. 5. Slide: 40 angle, but needs also to be done with the solar zenith angle for the downwelling solar part. Fire Workshop, Sofia, 7 -10 Sep 2009

SEVIRI Active Fire products 3. Create a MSG emissivity map 4. There are plans to create a “MSG” emissivity map instead of using a map 5. derived from other sources and then interpolated in space and spectra 6. to the MSG channels 4. Make use of the temporal filtering 5. Currently the algorithm is not using temporal filtering for both the diurnal 6. cycle of the temperatures and the tracking of fires for a single pixel 7. throughout the day Slide: 41 Fire Workshop, Sofia, 7 -10 Sep 2009

SEVIRI Active Fire products MTG Slide: 42 Fire Workshop, Sofia, 7 -10 Sep 2009

SEVIRI Active Fire products MTG Imager - Full Disk Mission The imager mission (FCI) of the third generation of Meteosat (MTG) will have 16 channels which are scanning the full disk every 10 minutes. The channels will be located at the following wavelengths: Slide: 43 Fire Workshop, Sofia, 7 -10 Sep 2009

SEVIRI Active Fire products MTG Imager - Full Disk Mission MTG Imagery Band Central Wavel. (µm) Spatial resolution (km) Objective FDHSI-1 0. 440 1 Aerosol optical depth, aerosol particle size, volcanic ash concentration FDHSI-2 0. 510 1 Aerosol optical depth, aerosol particle size, volcanic ash concentration FDHSI-3 0. 645 1 Cloud detection, AMV, cloud type, cloud optical depth, snow cover, vegetation stress, smoke detection, volcanic ash concentration FDHSI-4 0. 860 1 Cloud detection, AMV, cloud type, cloud phase, cloud optical depth, cloud microphysics, snow cover, vegetation stress, smoke detection, volcanic ash concentration FDHSI-5 0. 91 1 Total column humidity FDHSI-6 1. 375 1 Cirrus detection, water vapour imagery FDHSI-7 1. 61 1 Cloud phase, cloud microphysics, cloud detection, cloud type, snow cover, vegetation stress, smoke detection FDHSI-8 2. 26 1 Cloud microphysics FDHSI-9 3. 80 2 Cloud detection, cloud type, AMV, LST/SST, cloud phase, cloud microphysics, detection/monitoring, volcanic ash detection, volcanic ash concentration FDHSI-10 6. 30 2 AMV, tracer heights, instability (grad T/Hu), column humidities FDHSI-11 7. 35 2 AMV, tracer heights, instability (grad T/Hu), column humidities FDHSI-12 8. 70 2 Cloud detection, cloud type, cloud top height, LST/SST, cloud phase, cloud drop size, instability (grad T/Hu), sea ice temperature, volcanic SO 2 detection, sand/dust storm detection FDHSI-13 9. 66 2 Total Column ozone FDHSI-14 10. 5 2 Cloud detection, cloud type, cloud top height, AMV, tracer heights, LST/SST, instability (grad T/Hu), total column humidity, sea ice temperature, fire detection/monitoring, volcanic ash detection, volcanic ash concentration, sand/dust storm detection FDHSI-15 12. 3 2 Cloud detection, cloud type, cloud top height, AMV, tracer heights, LST/SST, instability (grad T/Hu), total column humidity, sea ice temperature, volcanic ash detection, volcanic ash concentration, sand/dust storm detection FDHSI-16 13. 3 2 Cloud top height, tracer heights, instability (grad T/Hu), volcanic ash detection Slide: 44 Fire Workshop, Sofia, 7 -10 Sep 2009 snow cover, sea ice temperature, fire

SEVIRI Active Fire products MTG Imager - Full Disk Mission True Color 1 km Slide: 45 Vertical integrierter Wassrdampf über Land 1 km Extrem dünne Zirrusbewölkung 1 km Wolkenteilchengröße und Phase 0. 5 km Fire Workshop, Sofia, 7 -10 Sep 2009

SEVIRI Active Fire products MTG Imager Full Disk Mission MTG hat die gleichen Kanäle wie MSG im thermischen Bereich 2 km nominale Auflösung 1 km Auflösung für 3. 8µm and 10. 5 mm Kanäle (rapid scan) Slide: 46 Fire Workshop, Sofia, 7 -10 Sep 2009

SEVIRI Active Fire products MTG Imager - High Resolution Mission It will be possible to setup a high resolution rapid scanning for the northern 25% of the disk (Europe) at 4 channels which are scanning the area every 2. 5 minutes. These channels will be located at the following wavelengths: Slide: 47 Fire Workshop, Sofia, 7 -10 Sep 2009

SEVIRI Active Fire products MTG Imager - High Resolution Mission MTG Imagery Band Slide: 48 Central Wavel. (µm) Spatial resolution (km) Objective HRFI-1 0. 645 0. 5 Cloud detection, AMV HRFI-2 2. 26 0. 5 Cloud microphysics HRFI-3 3. 80 1 Cloud detection, cloud microphysics, (fire detection/monitoring) HRFI-4 10. 5 1 Cloud detection, cloud top height, AMV, (fire detection/monitoring) Fire Workshop, Sofia, 7 -10 Sep 2009

SEVIRI Active Fire products MODIS ch. 7 (2. 1 µm) 1 km resolution (28 Aug 2007, 11: 10 UTC) Similar to the planned MTG ch. FDHSI-8 at 2. 26 µm The hot spots appear already at 2. 1 m Slide: 49 Fire Workshop, Sofia, 7 -10 Sep 2009

SEVIRI Active Fire products MTG Imager But the main (big) points of the MTG Imager for the fire community are: • The “fire channel” centered at 3. 8 µm avoiding the CO 2 absorption • The temporal and spatial resolution will be higher (10 minutes and 2 km) • The dynamic range of the 3. 8 µm channel will be extended to 450 K Slide: 50 Fire Workshop, Sofia, 7 -10 Sep 2009

SEVIRI Active Fire products Summary and Conclusions Slide: 51 Fire Workshop, Sofia, 7 -10 Sep 2009

SEVIRI Active Fire products Active Fire detection considers: • Spectral signals and its spatial distribution (channels IR 3. 9, IR 10. 8, IR 3. 9 -IR 10. 8 ) • Uses ECMWF Forecast data to derive thresholds • Uses channel VIS 0. 6 to eliminate “hot clouds” Slide: 52 Fire Workshop, Sofia, 7 -10 Sep 2009

SEVIRI Active Fire products Problems still to be solved are: • Sunglint areas • Better forecast/emissivity information • Use of temporal signals (temporal filtering) Slide: 53 Fire Workshop, Sofia, 7 -10 Sep 2009