Outline Overview of recent work improving performance in most - - PowerPoint PPT Presentation

▶
outline
SMART_READER_LITE
LIVE PREVIEW

Outline Overview of recent work improving performance in most - - PowerPoint PPT Presentation

Nov 06, 2022 551 likes •873 views

Evaluation of Potential Enhancements to Version 6 Cloud Clearing and Profile Retrieval William J. Blackwell and Michael Pieper AIRS Science Team Meeting April 17, 2008 This work was sponsored by the National Oceanic and Atmospheric

slide-1
SLIDE 1

AIRS ST Apr08: 1 WJB 6/2/08

MIT Lincoln Laboratory

Evaluation of Potential Enhancements to Version 6

Cloud Clearing and Profile Retrieval

This work was sponsored by the National Oceanic and Atmospheric Administration under contract FA8721-05-C-0002. Opinions, interpretations, conclusions, and recommendations are those of the authors and are not necessarily endorsed by the United States Government.

William J. Blackwell and Michael Pieper AIRS Science Team Meeting April 17, 2008

slide-2
SLIDE 2

MIT Lincoln Laboratory

AIRS ST Apr08: 2 WJB 6/2/08

Outline

  • Overview of recent work – improving performance in most

difficult cases:

– Land – Elevated surface terrain – Near polar regions

  • SCC+NN performance comparisons with AIRS L2 Version 5

algorithm (versus ECMWF and Radiosondes)

  • IASI versus AIRS: SCC/NN temperature retrieval performance

– Importance of high SNR in SW spectral region

  • Possible Version 6 enhancements (regression first guess, etc.)
  • Future Work
slide-3
SLIDE 3

MIT Lincoln Laboratory

AIRS ST Apr08: 3 WJB 6/2/08

Retrieval Performance Validation with AIRS/AMSU

  • >1,000,000 co-located AIRS/AMSU/ECMWF observations

from ~100 days:

– Every fourth day from December 1, 2004 through January 31, 2006 – Used for training

  • ~250,000 profiles set aside for validation and testing sets
  • ~50,000 quality-controlled radiosondes from NOAA FSL

global database co-located with AIRS/AMSU observations

– Used for validation

Global: Cloudy, Land & Ocean, Day & Night Case 1: ECMWF atmospheric fields Case 2: Radiosonde data

slide-4
SLIDE 4

MIT Lincoln Laboratory

AIRS ST Apr08: 4 WJB 6/2/08

SCC/NN versus AIRS L2 (Version 5) Descending, Ocean, Edge-of-Scan, Spring05

Latitudes within ±60°

~1km vertical layers AIRS+AMSU ECMWF is “truth”

slide-5
SLIDE 5

MIT Lincoln Laboratory

AIRS ST Apr08: 5 WJB 6/2/08

SCC/NN versus AIRS L2 (Version 5) Descending, Land, Edge-of-Scan, Spring05

Latitudes within ±60°

~1km vertical layers AIRS+AMSU ECMWF is “truth”

slide-6
SLIDE 6

MIT Lincoln Laboratory

AIRS ST Apr08: 6 WJB 6/2/08

Descending, Land, Edge-of-Scan, Spring05 Versus Radiosondes

Latitudes within ±60°

~1km vertical layers AIRS+AMSU 910 radiosondes are “truth”

slide-7
SLIDE 7

MIT Lincoln Laboratory

AIRS ST Apr08: 7 WJB 6/2/08

IASI/ECMWF/SARTA Matchup Database

  • Global database spanning May07-Dec07
  • Approximately 100,000 fields-of-regard

– IASI observations (2x2) – ECMWF atmospheric fields – Radiosondes (available for some FOR’s) – IASI clear-air spectra calculated with SARTA v1.05

  • Database stratified by surface type, latitude, solar zenith

angle, sensor scan angle, surface pressure

slide-8
SLIDE 8

MIT Lincoln Laboratory

AIRS ST Apr08: 8 WJB 6/2/08

RMS IASI Cloudy Obs - Clear Calcs (i.e., Before Cloud Clearing)

Window 15-micron 4-micron Opaque 4-micron Water vapor

SCC RMS with AIRS

Ocean

RMS IASI Obs-Calcs (K)

slide-9
SLIDE 9

MIT Lincoln Laboratory

AIRS ST Apr08: 9 WJB 6/2/08

Correlation of “IASI OBS” and “IASI OBS-CALCS” Eigenvectors

Eigenvectors almost identical Indicates channels responsive to clouds Correlation decreases as atmospheric signal is removed

Ocean

slide-10
SLIDE 10

MIT Lincoln Laboratory

AIRS ST Apr08: 10 WJB 6/2/08

IASI Temperature Retrievals Over Ocean

Near-nadir scan angles, ±60° Latitude

~1km vertical layers IASI+AMSU ECMWF is “truth”

slide-11
SLIDE 11

MIT Lincoln Laboratory

AIRS ST Apr08: 11 WJB 6/2/08

AIRS versus IASI: Ocean

Near-nadir scan angles, ±60° Latitude

~1km vertical layers IASI+AMSU ECMWF is “truth”

slide-12
SLIDE 12

MIT Lincoln Laboratory

AIRS ST Apr08: 12 WJB 6/2/08

IASI Temperature Retrievals Over Land

Near-nadir scan angles, ±60° Latitude

~1km vertical layers IASI+AMSU ECMWF is “truth”

slide-13
SLIDE 13

MIT Lincoln Laboratory

AIRS ST Apr08: 13 WJB 6/2/08

AIRS versus IASI: Land

Near-nadir scan angles, ±60° Latitude AIRS is significantly better near the surface

~1km vertical layers IASI+AMSU ECMWF is “truth”

slide-14
SLIDE 14

MIT Lincoln Laboratory

AIRS ST Apr08: 14 WJB 6/2/08

AIRS versus IASI NEdT

slide-15
SLIDE 15

MIT Lincoln Laboratory

AIRS ST Apr08: 15 WJB 6/2/08

AIRS Retrieval Degradation After Adding Noise to Shortwave Channels

Near-nadir scan angles, ±60° Latitude

~1km vertical layers IASI+AMSU ECMWF is “truth”

slide-16
SLIDE 16

MIT Lincoln Laboratory

AIRS ST Apr08: 16 WJB 6/2/08

Potential Advantages SCC/NN “First Guess” Could Offer v6

  • Lower trend biases due to first guess

– Coefficients sets are derived from several stratifications:

Season, latitude, surface pressure, surface type, solar zenith angle

  • Higher yield/accuracy in critical areas

– Land – Polar – Heavy clouds

  • Lower sensitivity to changing/degrading instrument

properties

– For example, preliminary analysis of AMSU ch4 degradation has minimal impact on SCC/NN products

slide-17
SLIDE 17

MIT Lincoln Laboratory

AIRS ST Apr08: 17 WJB 6/2/08

Steps in Version 5 and Version 5.12 (Courtesy Joel Susskind)

MIT AMSU Retrieval Cloudy regression gives AMSU Retrieval using gives (now solve for T(p), only - not Ts) Determine using Determine from AMSU retrieval using gives (now solve for T(p), only - not Ts) Determine using Physical retrieval using and gives AMSU retrieval using gives determined from Physical retrieval using gives Select or Clouds, OLR determined from or Generate error estimates Do QC Steps Modified in Version 5.12 XCR XCR Xmicrowave = X0 ˆ Ri

0,0,Pc0

X0 Xreg ˆ Ri X1 ˆ Ri1 X1 ˆ Ri1 XPHYS XPHYS Xtest ˆ Ri2 XPHYS ˆ Ri2 Xfinal X0 Xfinal X0 Xfinal

X

Xreg X1

  • ˆ

R

i SCC

X NN X NN

ˆ R

i SCC

X NN

? ? ? ? ?

slide-18
SLIDE 18

MIT Lincoln Laboratory

AIRS ST Apr08: 18 WJB 6/2/08

Future Work / Conclusions

  • Additional and more extensive performance assessments

– Experiments to illuminate possible paths of integration with AIRS Level 2 algorithm (v6) – Match-ups with RAOB data

  • Algorithm optimizations, especially for IASI/CrIMSS

– Improved handling of land, including elevated surface terrain and surface emissivity

  • Comprehensive performance assessments with ECMWF and

Radiosondes continue to show encouraging results for SCC/NN

  • Potential enhancements to v6 include: Lower trend biases,

higher yield/accuracy, less sensitivity to sensor degradation

slide-19
SLIDE 19

AIRS ST Apr08: 19 WJB 6/2/08

MIT Lincoln Laboratory

Backup Slides

slide-20
SLIDE 20

MIT Lincoln Laboratory

AIRS ST Apr08: 20 WJB 6/2/08

Algorithm Overview (Part I)

  • Temperature and moisture profile retrievals are produced in all cloud conditions
  • Cloud-cleared radiance estimates are produced for all 2378 AIRS channels
  • Retrieval is global:

– All latitudes – Ocean and land – Day and night

  • Quality control has been implemented
  • IR-only option implemented
  • Very fast: Cloud-cleared radiances and retrieved profiles generated for one field
  • f regard in ~1 msec using PC!!

– Two-three orders of magnitude faster than current operational methods – One-two orders of magnitude faster than iterative, pseudochannel methods

slide-21
SLIDE 21

MIT Lincoln Laboratory

AIRS ST Apr08: 21 WJB 6/2/08

Algorithm Overview (Part II)

  • Algorithm is composed of linear and non-linear statistical
  • perators

– Projected principal components transform – Neural network estimation

  • Coefficients are derived empirically, off-line:

– Co-location of sensor measurements with “truth” (Radiosondes, NWP, etc.) – Model-generated data – Data stratification is used for:

Sensor scan angle Latitude Solar zenith angle Surface type Surface elevation

slide-22
SLIDE 22

MIT Lincoln Laboratory

AIRS ST Apr08: 22 WJB 6/2/08

Algorithm Block Diagram

R ~ P ~ T ˆ

Projected Principal Components Transform Stochastic Cloud Clearing

R ~

slide-23
SLIDE 23

MIT Lincoln Laboratory

AIRS ST Apr08: 23 WJB 6/2/08

Block Diagram of SCC Algorithm

Linear Operator A 3x3 AIRS TB’s Select/average FOV’s 5 microwave λ’s Land fraction Secant θ Linear Operator B Linear Operator C Linear Operator D Cloudy Test More cloudy Less cloudy N 1 PC Δ-cloud 2 TB’s 1

×

7

×

N 1 PC Cleared AIRS TB’s N = 2378 channels

Cho and Staelin, Aug. 2006

Quality control

slide-24
SLIDE 24

MIT Lincoln Laboratory

AIRS ST Apr08: 24 WJB 6/2/08

SCC+NN Quality Control

  • Simple, linear function of estimated radiance correction for

a set of channels

  • Framework allows for altitude-dependent quality flags
  • Yield versus accuracy trades can be easily performed

– “Qual_Good” yields approximately 80% – “Qual_Best” yields approximately 30%

slide-25
SLIDE 25

MIT Lincoln Laboratory

AIRS ST Apr08: 25 WJB 6/2/08

Stochastic Cloud Clearing Quality Control

Ocean, All latitudes

slide-26
SLIDE 26

MIT Lincoln Laboratory

AIRS ST Apr08: 26 WJB 6/2/08

SCC/NN versus AIRS L2 (Version 5) Descending, South Pole*, Edge-of-Scan, Spring

*South Pole = Latitudes < -60°

~1km vertical layers AIRS+AMSU

ECMWF is “truth” Quality is suspect

slide-27
SLIDE 27

MIT Lincoln Laboratory

AIRS ST Apr08: 27 WJB 6/2/08

IASI Eigenanalysis

Predominantly cloud effects Atmospheric and sensor “noise”

Ocean

slide-28
SLIDE 28

MIT Lincoln Laboratory

AIRS ST Apr08: 28 WJB 6/2/08

IASI “OBS” and “OBS-CALCS” Eigenvectors

slide-29
SLIDE 29

MIT Lincoln Laboratory

AIRS ST Apr08: 29 WJB 6/2/08

Stochastic Cloud Clearing of IASI

473 IASI channels were cleared Descending orbits within ±60° latitude, ocean

ECMWF is “truth”

slide-30
SLIDE 30

MIT Lincoln Laboratory

AIRS ST Apr08: 30 WJB 6/2/08

Stochastic Cloud Clearing of IASI

473 IASI channels were cleared Descending orbits within ±60° latitude, land

ECMWF is “truth”