1 / 30

The NOAA Unique CrIS ATMS Processing System (NUCAPS) Algorithm Status February 26th, 2013

The NOAA Unique CrIS ATMS Processing System (NUCAPS) Algorithm Status February 26th, 2013. Antonia Gambacorta 1 , Chris Barnet 2 , Walter Wolf 2 , Thomas King 1 , Eric Maddy 3 , Nick Nalli 1 , Murty Divakarla 1 , Kexin Zang 1 , Xiaozhen Xiong 1 1 IMSG 2 NOAA/NESDIS/STAR

lawson
Download Presentation

The NOAA Unique CrIS ATMS Processing System (NUCAPS) Algorithm Status February 26th, 2013

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. The NOAA Unique CrIS ATMS Processing System (NUCAPS) Algorithm StatusFebruary 26th, 2013 Antonia Gambacorta1, Chris Barnet2, Walter Wolf2, Thomas King1, Eric Maddy3, Nick Nalli1, MurtyDivakarla1,KexinZang1, XiaozhenXiong1 1 IMSG 2 NOAA/NESDIS/STAR 3 STC

  2. Introduction • The NOAA Unique CrIS/ATMS Processing System (NUCAPS) is a heritage algorithm of the AIRS Science Team and NOAA IASI hyper spectral retrieval methodology • The NOAA/NESDIS/STAR implementation of this algorithm is a modular architecture that was specifically designed to be compatible with multiple instruments: the same retrieval algorithm and the same underlying spectroscopy are currently employed to process the AIRS/AMSU, IASI/AMS/MHS and CrIS/ATMS suite of instruments. • Inputs: CrIS and ATMS radiance • Outputs: AVTP, AVMP, AVPP, O3, + cloud products (cloud fraction and height), surface products (spectral emissivity and surface temperature) and other trace gases (carbon monoxide, carbon dioxide, methane, nitric acid, nitrous oxide) • The NUCAPS Algorithm Readiness Review was held on January 14th 2013, when the system has been successfully approved by NOAA senior management for transition into operations (scheduled to begin in the Fall 2013).

  3. Outline • Temperature, water vapor , ozone validation results: • Global, Ocean/Land regimes validation versus: • collocated ECMWF and AVN analyses • AIRS operational (“version 6”) retrievals (uses same spectroscopy as NUCAPS, neural network first guess, improved surface emissivity first guess) • AIRS “version 5.9” retrievals (uses same spectroscopy and retrieval algorithm as NUCAPS) • CrIMSS EDR algorithm • Focus day used: 05-15-2012 (high coincidence between AIRS and CrIS footprints) • Back up slides: Tropical, Mid-Latitude, Polar; Day/Night;

  4. Overview of CrIMSS EDR Product Specifications • Atmospheric Vertical Temperature Profile (AVTP). • Lower tropospheric temperature are Key Performance Parameters (KPPs). 4

  5. Overview of CrIMSS EDR Product Specifications • Atmospheric Vertical Moisture Profile (AVMP). • Lower tropospheric moisture layers are Key Performance Parameters (KPPs) . 5

  6. Overview of CrIMSS EDR Product Specifications

  7. T, q Retrieval Statistics vs ECWMF; o3 vs AVNNUCAPS: ECMWF trained ccr FG (dash), final RET (solid)AIRS v5.9: ECMWF trained ccr FG (dash), final RET (solid)AIRS v6: NN FG (dash), final RET (solid) GLOBAL RMS vertical red bars indicate specification values (listed on previous slides 7,8 and 9) Significance: NUCAPS T, q, and O3 profiles meet specifications almost everywhere. Column integrated values fully meet specifications. See next slide for further comments on statistical performance.

  8. Temperature, Water Vapor, Ozone Statistics vs Model Analyses • NUCAPS global RMS and BIAS temperature, water vapor and O3 profile statistics generally meet specifications. Column integrated values fully meet specifications. • After only ~1 year in orbit, NUCAPS T, q and O3 RMS and BIAS statistical performance is comparable to AIRS v6 and AIRS v5.9 (10 year maturity product) . • NUCAPS global acceptance yield is ~60%, AIRS v6 yield is 86% (different rejection criteria than NUCAPS) and AIRS v5.9 is 75% (same rejection criteria as NUCAPS). Possible sources of difference in the acceptance yield and retrieval performance (future work for NUCAPS): • a) AIRS v6 has an improved surface emissivity first guess; • b) AIRS v5.9 has a multi-year temperature and water vapor first guess regression training; • c) AIRS radiance tuning uses a dedicated raob training sample ; • d) AIRS retrieval parameters and QAs are fully optimized. • Work in progress: use the independent intensive cal/val RAOB data set to fully understand and improve NUCAPS lower troposphere temperature RMS and BIAS features.

  9. T, q Retrieval Statistics vs ECWMF; o3 vs AVNNUCAPS: ECMWF trained ccr FG (dash), final RET (solid)AIRS v5.9: ECMWF trained ccr FG (dash), final RET (solid)AIRS v6: NN FG (dash), final RET (solid) GLOBAL BIAS See slide 8 for comments on statistical performance

  10. T, q Retrieval Statistics vs ECWMF; o3 vs AVNNUCAPS: Global (solid), Land (dash), Ocean (dash-dot)AIRS v5.9: Global (solid), Land (dash), Ocean (dash-dot) AIRS v6: Global (solid), Land (dash), Ocean (dash-dot) Ocean/Land RMS Significance: NUCAPS performance appears stable over multiple geophysical regimes (see back up slides for tropical, mid-latitude, polar, day/night validation results)

  11. T, q Retrieval Statistics vs ECWMF; o3 vs AVNNUCAPS: Global (solid), Land (dash), Ocean (dash-dot) AIRS v5.9:Global (solid), Land (dash), Ocean (dash-dot) AIRS v6: Global (solid), Land (dash), Ocean (dash-dot) Ocean/Land BIAS Significance: NUCAPS performance appears stable over multiple geophysical regimes (see back up slides for tropical, mid-latitude, polar, day/night validation results)

  12. 05/15 vs07/13 focus day RMS statistics Significance: NUCAPS performance is stable and robust over multiple focus days, including those not used for tuning and regression training: 05/15 focus day (red curves) was used for tuning and training, 07/13 (green curves) was not.

  13. 05/15vs07/13 focus day BIAS statistics Significance: NUCAPS performance is stable and robust over multiple focus days, including those not used for tuning and regression training: 05/15 focus day was used for tuning and training, 07/13 was not.

  14. NUCAPS vsCrIMSS-EDR GLOBAL NUCAPS AIRS v5.9 NUCAPS Phys-only CrIMSS EDR RMS NUCAPS NUCAPS Phys-only CrIMSS EDR AIRS v5.9 AIRS v6 • NUCAPS (red) is generally more stable than CrIMSS-EDR (green) • NUCAPS –Phys only (MW retrieval as first guess) and CrIMSS EDR have similar yield and performance

  15. NUCAPS vsCrIMSS-EDR GLOBAL BIAS NUCAPS NUCAPS Phys-only CrIMSS EDR AIRS v5.9 AIRS v6 • NUCAPS (red) has generally lower bias than CrIMSS-EDR (green) • NUCAPS –Phys only (MW retrieval as first guess) and CrIMSS EDR have similar yield and performance

  16. Summary and Conclusions • We have shown NUCAPS temperature, water vapor and ozone validation results • Global, Ocean/Land regimes (back upslides: tropical, mid-latitude, polar, day/night) validation versus 1) collocated ECMWF and AVN analyses, 2) AIRS retrievals (operational version 6 and version “5.9” ) and 3) CrIMSS-EDR: • NUCAPS meets specification: • Surface – 300mb 1.6K/1-km layer; 300mb- 30mb 1.5K/3-km layer for global temperature • Surface- 600mb 20%/2-km; 600mb – 300mb -100mb 35% /2-km layer for global water vapor • 260mb – 4mb 20%/5-km layer for global ozone • After only ~1 year in orbit, NUCAPS T, q and O3 RMS and BIAS statistical performance is comparable to AIRS v6 and AIRS v5.9 (10 year maturity product) • NUCAPS RMS and BIAS appears generally more stable than CrIMSS-EDR

  17. Future work • Ongoing optimization study on NUCAPS includes channels, perturbation functions,  first guess and damping parameter. • Use dedicated cal/val field campaign in situ measurements (AEROSE, Aerospace, ARM sites) and operational radiosondes (NPROVS) to fully assess NUCAPS retrieval performance of temperature, water vapor, cloud cleared radiance, cloud parameters and trace gases. • Leverage on ongoing scientific collaborations (low cost activities for NOAA) to perform trace gas validation.

  18. Back up slides

  19. T, q Retrieval Statistics vs ECWMF; o3 vs AVNNUCAPS: ECMWF trained ccr FG (dash), final RET (solid)AIRS v5.9: ECMWF trained ccr FG (dash), final RET (solid)AIRS v6: NN FG (dash), final RET (solid) TROPICS RMS • Possible sources of difference in the performance and acceptance yield: • AIRS v6 improved surface emissivity first guess; AIRS v5.9 multi-year regression training; dedicated raob based tuning • Future work (phase II): NUCAPS QAs optimization ; multi-seasonal regression and tuning training

  20. T, q Retrieval Statistics vs ECWMF; o3 vs AVNNUCAPS: ECMWF trained ccr FG (dash), final RET (solid)AIRS v5.9: ECMWF trained ccr FG (dash), final RET (solid)AIRS v6: NN FG (dash), final RET (solid) TROPICS BIAS • Possible sources of difference in the performance and acceptance yield: • AIRS v6 improved surface emissivity first guess; AIRS v5.9 multi-year regression training; dedicated raob based tuning • Future work (phase II): NUCAPS QAs optimization ; multi-seasonal regression and tuning training

  21. T, q Retrieval Statistics vs ECWMF; o3 vs AVNNUCAPS: ECMWF trained ccr FG (dash), final RET (solid)AIRS v5.9: ECMWF trained ccr FG (dash), final RET (solid)AIRS v6: NN FG (dash), final RET (solid) MIDLAT RMS • Possible sources of difference in the acceptance yield: • AIRS v6 improved surface emissivity first guess; AIRS v5.9 multi-year regression training; dedicated raob based tuning • Future work (phase II): NUCAPS QAs optimization ; multi-seasonal regression and tuning training

  22. T, q Retrieval Statistics vs ECWMF; o3 vs AVNNUCAPS: ECMWF trained ccr FG (dash), final RET (solid)AIRS v5.9: ECMWF trained ccr FG (dash), final RET (solid)AIRS v6: NN FG (dash), final RET (solid) MIDLAT BIAS • Possible sources of difference in the acceptance yield: • AIRS v6 improved surface emissivity first guess; AIRS v5.9 multi-year regression training; dedicated raob based tuning • Future work (phase II): NUCAPS QAs optimization ; multi-seasonal regression and tuning training

  23. T, q Retrieval Statistics vs ECWMF; o3 vs AVNNUCAPS: ECMWF trained ccr FG (dash), final RET (solid)AIRS v5.9: ECMWF trained ccr FG (dash), final RET (solid)AIRS v6: NN FG (dash), final RET (solid) POLAR RMS • Possible sources of difference in the acceptance yield: • AIRS v6 improved surface emissivity first guess; AIRS v5.9 multi-year regression training; dedicated raob based tuning • Future work (phase II): NUCAPS QAs optimization ; multi-seasonal regression and tuning training

  24. T, q Retrieval Statistics vs ECWMF; o3 vs AVNNUCAPS: ECMWF trained ccr FG (dash), final RET (solid)AIRS v5.9: ECMWF trained ccr FG (dash), final RET (solid)AIRS v6: NN FG (dash), final RET (solid) POLAR BIAS • Possible sources of difference in the acceptance yield: • AIRS v6 improved surface emissivity first guess; AIRS v5.9 multi-year regression training; dedicated raob based tuning • Future work (phase II): NUCAPS QAs optimization ; multi-seasonal regression and tuning training

  25. T, q Retrieval Statistics vs ECWMF; o3 vs AVNNUCAPS: Global (solid), Night (dash), Day (dash-dot) AIRS v5.9:Global (solid), Night (dash), Day (dash-dot) AIRS v6: Global (solid), Night (dash), Day (dash-dot) Day/Night RMS • Possible sources of difference in the acceptance yield: • AIRS v6 improved surface emissivity first guess; AIRS v5.9 multi-year regression training; dedicated raob based tuning • Future work (phase II): NUCAPS QAs optimization ; multi-seasonal regression and tuning training

  26. T, q Retrieval Statistics vs ECWMF; o3 vs AVNNUCAPS: Global (solid), Night (dash), Day (dash-dot) AIRS v5.9:Global (solid), Night (dash), Day (dash-dot) AIRS v6: Global (solid), Night (dash), Day (dash-dot) Day/Night BIAS • Possible sources of difference in the acceptance yield: • AIRS v6 improved surface emissivity first guess; AIRS v5.9 multi-year regression training; dedicated raob based tuning • Future work (phase II): NUCAPS QAs optimization ; multi-seasonal regression and tuning training

  27. Part II: (1) Temperature and (2) Geopotential Height Statistics vs Dedicated Radiosondes • ARM Sites (n = 450) • Tropical Western Pacific (TWP, island) (90) • Southern Great Plains (SGP) (180) • North Slope Alaska (NSA) (180) • Jul–Sep 2012 • JPSS funded • Kauai, Hawaii (PMRF) Site (n = 20+) • Tropical Central Pacific (island) • May 2012 (20), • Collocated lidar • Collaborator: The Aerospace Corp. • Beltsville, MD (BCCSO) Site (n = 10+) • Urban midlatitude • Jun–Sep 2012 • Collaborator: HU/NCAS • NOAA AEROSE Cruise (n ≈ 60) • Tropical Atlantic (ship) • September 2012 • Possible HS3 Campaign AC overflight • JPSS funded • Collaborators: HU/NCAS, NOAA/ESRL Picture courtesy of Nick Nalli • Significance: • (1) Dedicated RAOBs can provide independent correlative data not assimilated into NWP models • (2) Dedicated RAOBs can provide detailed performance specification

  28. Kauai, Hawaii validation campaign results • Kauai, Hawaii (PMRF) Site • Tropical Central Pacific Ocean Regime • May 2012 13,14,15,18,19,25,29 • Collaborator: The Aerospace Corporation.

  29. (1) Temperature coarse layer (1km) statisticsNUCAPS, ECMWFvs RAOB(+/- 3 hours; <= 200km; 7 RAOBs profiles, 188 collocated retrievals) • NUCAPS temperature statistics against model independent and dedicated RAOBs over the tropical ocean regime meets specs (and shows improved lower troposphere RMS and BIAS performance wrt NUCAPS against ECMWF statistics shown on previous slides)

  30. NUCAPS vs AIRS v59 acceptance yield(blue = accepted; red = rejected) NUCAPS AIRS v59 • NUCAPS global acceptance yield is ~60% (focus day 2012/05/15) • AIRS v59 global acceptance yield is ~75% (focus day 2012/05/15)

More Related