1 / 33

VIIRS SM Performance Evaluation

VIIRS SM Performance Evaluation. NESDIS STAR CAL/VAL Team February 12 2013. Outline. Background – VIIRS SM Algorithm The performance evaluation of VIIRS SM product Discussions and Experiments Summary. VIIRS Suspended Matter (SM). Land Algorithm VCM  IP AOT  Ash Flag  SM=“ Ash ”

axl
Download Presentation

VIIRS SM Performance Evaluation

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. VIIRS SM Performance Evaluation NESDIS STAR CAL/VAL Team February 12 2013

  2. Outline • Background – VIIRS SM Algorithm • The performance evaluation of VIIRS SM product • Discussions and Experiments • Summary

  3. VIIRS Suspended Matter (SM) • Land Algorithm • VCM  IP AOT  Ash Flag  SM=“Ash” • AOT Not Produced  No SM type assigned SM=“Fill Value (255)” • AOT550 ≤ 0.15  SM=“None” • Depends on the land aerosol model used to retrieve the AOT • Dust  SM=“Dust” • Smoke High Absorption  SM=“Smoke” • Smoke Low Absorption  SM=“Smoke” • Urban High Absorption  SM=“Smoke” • Urban Low Absorption  SM=“Smoke” • Quality flag of SM Typing defined according to AOT QF and values. • Compute Smoke Concentration if SM=“Smoke” • Oceanic Algorithm : Next Slide OAD – VIIRS Aerosol Products IP/EDR Table 23

  4. VIIRS Suspended Matter (SM) continue • Ocean Algorithm • VCM  IP AOT  Ash Flag  SM=“Ash” • AOT Not Produced  No SM type assigned SM=“Fill Value (255)” • AOT550 ≤ 0.15  SM=“None” • Depends on the Small Mode Fraction (SMF) • SMF < 0.2  SM=“Dust” • 0.2 ≤ SMF < 0.5 • IP AOT < 0.3  SM=“Sea-Salt” • IP AOT ≥ 0.3  SM=“Unknown” or “Undetermined” • SMF ≥ 0.5 • Ocean small mode model 1,2,3  SM=“Smoke” • *** Ocean small mode model 4  SM=“Smoke”  (OAD Figure 5 is inconsistent with code, slide 29, which SM=“unknown”) • Quality Flag of SM Typing defined according to AOT QF and values. • Compute Smoke Concentration if SM=“Smoke”. OAD – VIIRS Aerosol Products IP/EDR Figure 5 (Slide 30)

  5. The VIIRS Components that Impact SM Type • VIIRS Cloud Mask (VCM) • Ash Flag -> SM=“Ash” • Confident Cloudy  AOT NOT Produced  SM=“Fill Value” or “Not Produced” • Heavy Aerosol Flag  reset Confident Cloudy to Confident Clear  AOT retrievals  potential SM type assigned • VIIRS Aerosol AOT Routine (IP AOT) • Retrieved AOT value • SM typing Threshold (0.15) • Sea-Salt or not (0.3) • SM Quality Flag • Land Aerosol Model Selected  direct assignment • Ocean Small/Large models combination (SMF) selected • 4 small mode models • 5 large mode models • 101 fractional amount • 2020 possible aerosol models VCM Aerosol AOT SM

  6. SM Type Reference • Using CALIPSO observations for the evaluation and improvement of VIIRS SM typing.

  7. What we found so far • Most prominent aerosol cases (dust/smoke) both over the land and ocean were treated as confident cloudy by VCM. It leads to NO SM type identified. Two examples shown on following slides • 09/03 2012 US Fire case • 10/18 2012 US Dust case • AOT NOT PRODUCED for CLOUDY and BRIGHT SURFACE (except for pixel with VCM Heavy Aerosol Flag) • Nothing else we can do till VCM can better differentiate between “Cloud” and “Heavy Aerosol”

  8. 09/03/2012 VIIRS RGB IP AOT550 Best Quality VIIRS Suspended Matters Type VCM Cloud Confident Flag

  9. Due to the code error, it is confident clear almost all over the granule. It allows the AOT to be retrieved and SM type to be identified. It shows some skill of VIIRS SM typing by identifying some dust pixels of the dust plume. 10/08/2012 VIIRS RGB IP AOT550 Best Quality VIIRS Suspended Matters Type VCM Cloud Confident Flag This case actually is in the period of “VCM array size was read incorrectly due to code error”. If the code is correct, once can expected the AOT is not retrieved over the dust plume as well as SM typing

  10. SMF Over the Ocean • In addition to cloud impact, SMF is the biggest source of error for the SM type over the ocean • VIIRS SMF is too high as compared to MODIS (< 0.52 for dust) • Most of the SM type over the ocean is “SMOKE” • (a) SM can do better with present oceanic algorithm if Aerosol AOT algorithm can provide better SMF information. • (b) Experiment with different SMF criteria for SM type

  11. 09/13/2012 VIIRS RGB IP AOT550 Best Quality VIIRS Suspended Matters Type IP Small Mode Fraction

  12. 09/13/2012 VCM Cloud Confident Flag VCM Heavy Aerosol Flag CALIPSO TRACK – green section (figure to the right) Dust was found in the target area

  13. 09/13/12 VIIRS MODIS MODIS Small Mode Fraction VIIRS VIIRS SMF is more than double the values of MODIS From Lorraine Remer

  14. SMF Criteria Experiments • By raising the SMF criteria for dust • it is arbitrary • literature values indicate different thresholds, and • 20% is likely the correct one for near source regions • By adding a pre-test for dust prior to present routine

  15. How to read this scale 0 20 50 100 SMF AOT Criteria 0.3 (1) Using the color coded SM type at the bottom of each slide • (2) Should be interpreted as • SMF < 20% is dust • 20%≤SMF<50% is sea-salt or unknown type depends on AOD 550 (≥ or < 0.3) • SMF ≥ 50% is Smoke • (3) Should be interpreted as • SMF < 75% is dust • 75% ≤ SMF<95% is sea-salt or unknown type depends on AOD 550 (≥ or < 0.3) • SMF ≥ 95% is Smoke 0 100 75 95 SMF Criteria 0.3 AOT

  16. (a) (c) 0 20 50 100 SMF AOT Criteria 0.3 0 100 65 85 SMF Criteria 0.3 AOT (b) (d) 0 100 68 88 SMF Criteria 0.3 AOT 0 100 70 90 SMF (c) (e) Criteria 0.3 AOT 0 100 72 92 SMF Criteria 0.3 AOT 0 100 75 95 SMF Criteria 0.3 AOT

  17. Use Information from Dust Aerosol Index to refine the SMF criteria • Dust Aerosol Index (DAI) – STAR Deep Blue Dust Index algorithm (P. Ciren et. al., in preparation) • Use DAI routine to identify dust pixels from VIIRS granules. • Compute statistic of SMF of dust pixels using IP small mode fraction data (IVAMI, not available from PEATE; Can only obtain it from GRAVITE and within <30 days from present day) • STAR CAL/VAL team re-runs the two granules using ADL and produce the IVAMI output

  18. npp_d20120913_t1506465 npp_d20120913_t1508119 From Pubu Ciren

  19. Over ocean AOT>0.15, good and degraded quality AOT>0.15, good and degraded quality Dust pixels identified by DBDI Non-Dust pixels identified by DBDI Fine mode weight from IVAMI. From Pubu Ciren

  20. Discussion • By simply adjusting the criteria of small mode fraction for dust (SMF <x%), dust becomes the dominated aerosol over the ocean. • Using the accumulate frequency plot of dust and non-dust detected by DAI (see previous slide), the single SMF adjusting methodology will have high detection rate for dust but also with high false alarm rate by counting non-dust particle as dust particle, e.g, with SMF < 70% for dust we can detect ~98% of dust particles but also miss-identified >95% of non-dust particle as dust. • Using the same chart, assuming that we do not need to detect 100% of dust and allow some error for detecting both dust and non-dust particles, i.e., by suffering some small inaccuracy, an additional step is inserted prior to the present algorithm • With the assumption that particles over the ocean with 50% ≤ SMF < 75% is dust, we cover ~70% of dust particle (~30% error; to the left) and only treat ~20% (to the right) of non-dust as dust. • Apply original criteria on the rest of particles (SMF <50% and SMF ≥ 75%). Hopefully, some dust will be recovered from the 30% of error (w/ SMF < 0.2) we initially ignored. • The results of using 65% ≤ SMF < 75% and 50% ≤ SMF < 75% as the step be applied prior to present SM code ( after Ash, AOD QF <3, AOD 550 > 0.15) are shown in next slide.

  21. Ash? AOD produced ? AOD > 0.15 ? yes Dust x %≤SMF<75%

  22. 0 20 50 75 100 0 20 50 65 75 100 SMF 0 20 50 100 SMF SMF AOT Using only original algorithm in white area Criteria 0.3 AOT Criteria 0.3 AOT Criteria 0.3 x %≤SMF<75% Dust With pre-test for dust incorporated X=65 X=50

  23. Consistent SMF Distribution for Dust • Although previous results are encouraging, but it relies on the consistent global SMF distribution of dust over the ocean. • The dust SMF distribution for different day (09/09/12, next slide) shows the main dust SMF is shifted to [0.25 ,0.55] as compared to [0.65,0.75] in previous slide. • Dust plume from different dust sources? (both chemical and physical composition) **Figures for 09/09/12 case can be found in slides 31-33

  24. Over ocean AOT>0.15, good quality AOT>0.15, good quality Dust pixels identified by DBDI Non-Dust pixels identified by DBDI Fine mode weight from IVAMI

  25. Next Step • Find the monthly/seasonal mean of dust SMF range global distribution. • Find the monthly/seasonal mean of dust global distribution from CALIPSO observations. • Compare these two and see if a SMF range can be found for identifying dust. (cover most of the dust).

  26. Summary • Using CALIPSO product as reference. • VIIRS Land and Ocean SM algorithms are different. • Aerosol model selected over land. • Small mode fraction selected over ocean. • VIIRS SM algorithms fundamentally is using the output of VCM and AOT for SM typing. No other M-Band SDR data are used. • VCM capability of differentiating cloud and aerosol is important to VIIRS SM type performance. • VIIRS SMF over the ocean is relatively higher than that of MODIS and published articles. • Present SMF criteria failed to identify the oceanic SM type correctly. • More investigations are in progress.

  27. END

  28. More slides for your own reference

  29. 7818 //------------------------------------------------------------------------------ 7819 // SelOceanAeroType 7820 //------------------------------------------------------------------------------ 7821 void ProEdrViirsAerosol::SelOceanAeroType 7822 ( 7823 AOT_STRUCT *aot, /* Structure with AOT Data */ 7824 QF_STRUCT *qf, /* Structure with QF Data */ 7825 Int32 pixel /* Pixel Being Processed */ 7826 ) const 7827 { 7828 Float32 percent = (float)(*(aot->combo + pixel)) / 100.f;  small mode fraction 7829 ::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::: ::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::: ::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::: 7863 else if(percent >= COM_LOW && percent < COM_HIGH)  COM_LOW=20%, COM_HIGH=50% 7864 { 7865 if(*(aot->faot550 + pixel) < AOT_THRESH)  AOT_THRESH = 0.3 7866 { 7867 *(aot->susmat + pixel) = SUSMAT_TYPE_SEASALT; /* seasalt */ 7868 } 7869 else 7870 { 7871 *(aot->susmat + pixel) = SUSMAT_TYPE_UNKNOWN; /* undetermined */ 7872 } 7873 } 7874 else 7875 { 7876 if(*(aot->smmdl + pixel) <= 3)  small mode model 1,2, and 3 7877 { 7878 *(aot->susmat + pixel) = SUSMAT_TYPE_SMOKE; /* smoke */ 7879 } 7880 else 7881 { 7882 *(aot->susmat + pixel) = SUSMAT_TYPE_UNKNOWN; /* undetermined */  small mode model 4 7883 } 7884 }

  30. OAD – VIIRS Aerosol Products IP/EDR. Figure 5 (Page 42)

  31. 09/09/2012 VIIRS RGB IP AOT550 Best Quality IP Small Mode Fraction VIIRS Suspended Matters Type

  32. 09/09/2012 VCM Cloud Confident Flag VCM Heavy Aerosol Flag CALIPSO TRACK – green section (figure to the right) Dust was found in the region (slightly further out away from the coastline)

  33. npp_d20120909_t1441290 npp_d20120909_t1442544 33

More Related