DustDection_liuyang_Final.ppt
- 1. A Thermal Index from MODIS Data for Dust Detection Yang Liu, Ronggao Liu Institute of Geographic Sciences and Natural Resources Research, CAS 2011.7.28 • Vancouver
- 2. Outline Dust storm detection from satellite data Method Results Summary
- 3. Dust Storms in Earth Research With increasing land degradation and deforestation, dust storms spread in the world since the 20th century. air pollution public health: respiratory diseases global climate biogeochemical cycle Characterization of the dust properties and distribution at global scale helps to understand their roles on the Earth radiative budget and global biogeochemical cycle
- 4. Remote sensing in dust storm monitoring Satellite remote sensing is advantageous in monitoring the global spatial and temporal variations of dust storms. ultraviolet channels (Hsu et al., 1996; Chiapello et al., 1999) visible, infrared channels (Miller, 2003; Hsu et al., 2006) thermal infrared channels (Ackerman, 1997; Legrand et al., 1992; Schepanski et al., 2007; Walker et al., 2009)
- 5. Challenges in dust storm monitoring Visible/ultraviolet band-based algorithm: blue or ultraviolet bands are required for bright surface Problems: 1) not reliable for nighttime; 2) no observations in these bands of many sensors, such as AVHRR (long-term dataset) Thermal Infrared band-based algorithm: Dust: brightness temperature difference (eg. BTD12,11, BTD8.6,11) Problems: the BTD is also related to other factors, such as land surface temperature and emissivity
- 6. Our work Aim: develop an algorithm for dust storm monitor based on thermal infrared bands measurements of MODIS with consideration of the effects of LST on BTD. Data MODIS brightness temperature in 11 (Band 31), 12 (Band 32), 8.6 (Band 29) Method Normalized BTD12,11, BTD8.6,11 with consideration of the effects of LST on BTD Results algorithm application over major land cover types over China
- 7. Outline Dust storm detection from satellite data Method Results Summary
- 8. Spectral response: dust, cloud and surface BTD12,11=BT12-BT11 BTD12,11 could separate dust from cloud, which affected by the LST Dust (BTD12,11>0) Clear-sky surface (BTD12,11~0) Cloud (BTD12,11<0)
- 9. Spectral response: dust, cloud and surface BTD8.6,11=BT8.6-BT11 BTD8.6,11 helps to separate dust from bright surface Dust/Cloud (BTD8.6,11>0) Clear-sky surface (BTD8.6,11~0) Bright surface (BTD8.6,11<0)
- 10. (1) Construction of pixel-by-pixel relationships of BT11-BT12 and BT8.6-BT11 The ratio of BT in 8.6 and 11, as well as 12 and 11 bands are mapped pixel-by-pixel based on clear-sky observations during 2000-2008; (2) BTD normalization The BTD12,11 and BTD8.6,11 are normalized to clear-sky condition using the pixel-based ratio relationship, and then used to calculate the difference ( Δ BTD ) with the observed BTD. (3) Dust determination Methods
- 11. Methods BTD12,11 and BTD8.6,11 over clear-sky surface After normalization, BTD12,11 and BTD8.6,11 concentrated around 0
- 12. Methods DustIndex (DI) of dust, cloud and clear-sky surface Dust: DI>0; Cloud: DI<0; Clear-sky surface: DI around 0 Cloud Dust Clear-sky surface
- 13. Outline Dust storm detection from satellite data Method Results Summary
- 14. Results 1 -Normalized BTD12,11 BT11=260K BT11=280K The BTD12,11 is related to surface temperature
- 15. Results 2 -Normalized BTD8.6,11 BT11=260K BT11=280K The BTD8.6,11 is related to surface temperature
- 16. Results 3 -performance over vegetated surface a) Terra MODIS true-color image of dust storm over Northeast China on April 7, 2001; b) Dust detection results DI could separate the airborne dust from clouds, vegetated surface and ocean
- 17. Results 4 -performance over bright surface (1) a) Terra MODIS true-color image shows dust storm in Taklimakan desert. b) Dust detection results DI could separate airborne dust from sand and cloud over bright desert surface
- 18. Results 4 -performance over bright surface (2) c) Terra MODIS true-color image shows dust storm in Gobi desert; d) Dust detection results DI could separate airborne dust from sand and cloud over bright desert surface
- 19. Results 5 -with and without normalization (Day) DOY97, 2001 over China Without normalization With normalization Desert Daytime Algorithm could reduce the effects of bright surface on dust detection
- 20. Results 6 -with and without normalization (Night) Nighttime Without normalization With normalization DOY97, 2001 over China Algorithm could detect the airborn dust in nighttime
- 21. Dust storm detection from satellite data Method Results Summary Outline
- 22. Summary A new algorithm has been developed to detect dust based on satellite thermal infrared imagery; here uses the brightness temperature of three thermal infrared channels of MODIS, including 8.6, 11 and 12 The algorithm considers the effects of LST on BTD; The algorithm could distinguish airborne dust from cloud and land surface over bright and dark surface during daytime and nighttime.
- 23. Thank you ！