3-Way Comparison between AIRS, ECMWF and GPS Temperatures in Upper - - PowerPoint PPT Presentation

▶
3 way comparison between airs ecmwf and gps temperatures
SMART_READER_LITE
LIVE PREVIEW

3-Way Comparison between AIRS, ECMWF and GPS Temperatures in Upper - - PowerPoint PPT Presentation

Aug 21, 2023 156 likes •515 views

Comparison of AIRS-ECMWF-GPS Temperature Profiles 3-Way Comparison between AIRS, ECMWF and GPS Temperatures in Upper Troposphere and Stratosphere Tom Yunck, Eric Fetzer, Anthony Mannucci, Chi Ao, Bill Irion, Brian Wilson and Gerald Manipon Jet

slide-1
SLIDE 1

Comparison of AIRS-ECMWF-GPS Temperature Profiles

29 March 2007 AIRS Science Meeting, Pasadena

1

3-Way Comparison between AIRS, ECMWF and GPS Temperatures in Upper Troposphere and Stratosphere

Tom Yunck, Eric Fetzer, Anthony Mannucci, Chi Ao, Bill Irion, Brian Wilson and Gerald Manipon

Jet Propulsion Laboratory

California Institute of Technology

AIRS Science Meeting

Pasadena, CA

29 March 2007

slide-2
SLIDE 2

Comparison of AIRS-ECMWF-GPS Temperature Profiles

29 March 2007 AIRS Science Meeting, Pasadena

2

slide-3
SLIDE 3

Comparison of AIRS-ECMWF-GPS Temperature Profiles

29 March 2007 AIRS Science Meeting, Pasadena

3

  • Adjustments to data are generally subjective and can be large

compared to the actual trend.

  • Reported trends vary considerably between groups using the same

data owing to differing adjustment methods.

  • Most models predict greater warming in the troposphere; most
  • bservations show greater warming at the surface. The likely

cause is errors in the tropospheric observations.

  • Recent adjustments have brought satellite observations into closer

agreement with models.

  • “Satellite observations tend to be bias-corrected to the model.”

(Healy & Thépault, 2006) [ECMWF] Some Key Points

slide-4
SLIDE 4

Comparison of AIRS-ECMWF-GPS Temperature Profiles

29 March 2007 AIRS Science Meeting, Pasadena

4

One Approach

slide-5
SLIDE 5

Comparison of AIRS-ECMWF-GPS Temperature Profiles

29 March 2007 AIRS Science Meeting, Pasadena

5

Ben Ho (UCAR), Jan 07

±1 K (single instr.) ~1 K at 30 km

GPS-GPS Comparison Stats from COSMIC

Bias <0.05 K

  • Std. Dev.

No corrections

  • r calibrations

required Performance traceable to an absolute SI standard No cloud or weather sensitivity

slide-6
SLIDE 6

Comparison of AIRS-ECMWF-GPS Temperature Profiles

29 March 2007 AIRS Science Meeting, Pasadena

6

TLS Comparison with GPS Sensors Launched 6 yrs Apart

N18 TLS est. from COSMIC v. N18 AMSU TLS est. from CHAMP Ben Ho (UCAR), Jan 07

slide-7
SLIDE 7

Comparison of AIRS-ECMWF-GPS Temperature Profiles

29 March 2007 AIRS Science Meeting, Pasadena

7

Modeled GPSRO Temperature Sigma

Modeled GPS Temperature Sigma

slide-8
SLIDE 8

Comparison of AIRS-ECMWF-GPS Temperature Profiles

29 March 2007 AIRS Science Meeting, Pasadena

8

Controlled Three-Way Comparison:

  • AIRS-ECMWF-GPS temperatures
  • Common set of 3-way match-ups
  • For all of 2003 (Champ, SAC-C)

ECMWF “Sweet Spot” First comparisons: 30°-60° North (“Mid North”)

Match-up criteria: <200 km, <2 hrs apart

slide-9
SLIDE 9

Comparison of AIRS-ECMWF-GPS Temperature Profiles

29 March 2007 AIRS Science Meeting, Pasadena

9

Pairwise RMS deviations

Match-ups: 766 AIRS Quality: 0, 1

A Puzzle

slide-10
SLIDE 10

Comparison of AIRS-ECMWF-GPS Temperature Profiles

29 March 2007 AIRS Science Meeting, Pasadena

10

Derived AIRS & ECMWF RMS deviations

slide-11
SLIDE 11

Comparison of AIRS-ECMWF-GPS Temperature Profiles

29 March 2007 AIRS Science Meeting, Pasadena

11

Derived AIRS & ECMWF RMS deviations

“Expected” AIRS-ECMWF RMS deviation

slide-12
SLIDE 12

Comparison of AIRS-ECMWF-GPS Temperature Profiles

29 March 2007 AIRS Science Meeting, Pasadena

12

Actual

AIRS-ECMWF RMS deviation

Var(X-Y) = Var(X) + Var(Y) – 2Cov(X,Y)

Derived AIRS & ECMWF RMS deviations

“Expected” AIRS-ECMWF RMS deviation

slide-13
SLIDE 13

Comparison of AIRS-ECMWF-GPS Temperature Profiles

29 March 2007 AIRS Science Meeting, Pasadena

13

AIRS-ECMWF “Covariation”

Derived AIRS & ECMWF RMS deviations

Match-ups: 766 AIRS Quality: 0, 1

slide-14
SLIDE 14

Comparison of AIRS-ECMWF-GPS Temperature Profiles

29 March 2007 AIRS Science Meeting, Pasadena

14

AIRS-ECMWF

  • Corr. Coeff.

Derived AIRS & ECMWF RMS deviations

Match-ups: 766 AIRS Quality: 0, 1

slide-15
SLIDE 15

Comparison of AIRS-ECMWF-GPS Temperature Profiles

29 March 2007 AIRS Science Meeting, Pasadena

15

I – The Tropics How else can we examine this question? Compare regional performance variations, where a particular technique is known to vary in a particular way

slide-16
SLIDE 16

Comparison of AIRS-ECMWF-GPS Temperature Profiles

29 March 2007 AIRS Science Meeting, Pasadena

16

ECMWF-GPS RMS Deviations, 2003

Smoothing of Sharp Tropopause

slide-17
SLIDE 17

Comparison of AIRS-ECMWF-GPS Temperature Profiles

29 March 2007 AIRS Science Meeting, Pasadena

17

ECMWF-GPS RMS Deviations and Means, 2003

Match-ups: 2203 AIRS Quality: 0,1

slide-18
SLIDE 18

Comparison of AIRS-ECMWF-GPS Temperature Profiles

29 March 2007 AIRS Science Meeting, Pasadena

18

Regional Performance Variations – II Vertical Bias Patterns

Mid North Far North Far South

slide-19
SLIDE 19

Comparison of AIRS-ECMWF-GPS Temperature Profiles

29 March 2007 AIRS Science Meeting, Pasadena

19

ECMWF-GPS Means, All GPS, 2003

slide-20
SLIDE 20

Comparison of AIRS-ECMWF-GPS Temperature Profiles

29 March 2007 AIRS Science Meeting, Pasadena

20

ECMWF & AIRS Means vs GPS, All 2003

slide-21
SLIDE 21

Comparison of AIRS-ECMWF-GPS Temperature Profiles

29 March 2007 AIRS Science Meeting, Pasadena

21

Conclusion

IF our GPSRO error model is accurate, then:

3-way comparisons show significant correlation between AIRS and ECMWF temperature errors. Next: Repeat the analysis with AIRS V5 and COSMIC data and true AIRS smoothing functions. Caveat: We should adopt actual AIRS smoothing functions to ensure we are comparing like quantities.

slide-22
SLIDE 22

Comparison of AIRS-ECMWF-GPS Temperature Profiles

29 March 2007 AIRS Science Meeting, Pasadena

22

Backups

slide-23
SLIDE 23

Comparison of AIRS-ECMWF-GPS Temperature Profiles

29 March 2007 AIRS Science Meeting, Pasadena

23

Match-ups: 766 Quality: 0, 1 Match-ups: 1388 Quality: 0, 1

RMS Deviations & Covariation, All 2003

slide-24
SLIDE 24

Comparison of AIRS-ECMWF-GPS Temperature Profiles

29 March 2007 AIRS Science Meeting, Pasadena

24

RMS Deviations & Covariation, All 2003

Match-ups: 888 Quality: 0, 1 Match-ups: 500 Quality: 0, 1

slide-25
SLIDE 25

Comparison of AIRS-ECMWF-GPS Temperature Profiles

29 March 2007 AIRS Science Meeting, Pasadena

25

RMS Deviations & Covariation, All 2003

Match-ups: 1731 Quality: 2 Match-ups: 1388 Quality: 0, 1

AIRS Quality 0,1 AIRS Quality 2

slide-26
SLIDE 26

Comparison of AIRS-ECMWF-GPS Temperature Profiles

29 March 2007 AIRS Science Meeting, Pasadena

26

Measurement error = bias (b) + zero mean random error (e): Me = b + e Measurement difference M1 - M2 is therefore: M1,2 = b1 - b2 + e1 - e2

= b1,2 + e1 - e2

The mean (expected) squared (MS) difference is therefore: MS1,2 = b1,2

2 + σ1 2 + σ2 2

(assuming e1 and e2 uncorrelated)

Simple Analysis

slide-27
SLIDE 27

Comparison of AIRS-ECMWF-GPS Temperature Profiles

29 March 2007 AIRS Science Meeting, Pasadena

27

Or: MS1,2 - b1,2

2 = σ1 2 + σ2 2

(i.e., Var = MS - square of the mean) For the three-way comparison we have: (1) MS1,2 - b1,2

2 = σ1 2 + σ2 2

(2) MS2,3 - b2,3

2 = σ2 2 + σ3 2

(3) MS1,3 - b1,3

2 = σ1 2 + σ3 2

slide-28
SLIDE 28

Comparison of AIRS-ECMWF-GPS Temperature Profiles

29 March 2007 AIRS Science Meeting, Pasadena

28

Measurement difference M1 - M2 is: M1,2 = b1,2 + e1 - e2 The mean squared (MS) difference is: MS1,2 = b1,2

2 + σ1 2 + σ2 2

(assuming e1 and e2 uncorrelated)

  • 2E[e1e2]

What this means:

slide-29
SLIDE 29

Comparison of AIRS-ECMWF-GPS Temperature Profiles

29 March 2007 AIRS Science Meeting, Pasadena

29

For the three-way comparison we now have: (1) MSE,A - bE,A

2 = σE 2 + σA 2 - 2γE,A 2

(2) MSA,G - bA,G

2 - σG 2 = σA 2

(3) MSE,G - bE,G

2 - σG 2 = σE 2

Revised Analysis

slide-30
SLIDE 30

Comparison of AIRS-ECMWF-GPS Temperature Profiles

29 March 2007 AIRS Science Meeting, Pasadena

30

10 K 20 K 10 K 20 K

Examples of AIRS-ECMWF Temperature Error Similarity

slide-31
SLIDE 31

Comparison of AIRS-ECMWF-GPS Temperature Profiles

29 March 2007 AIRS Science Meeting, Pasadena

31

Working Hypothesis

  • 1. AIRS “first guess” temperatures, trained on ECMWF model, closely

reproduce location-dependent ECMWF bias characteristics.

  • 2. Where temperature gradients are small, AIRS retrieval information

is weak and departures from the first guess are small.

  • 3. This leads to significant correlation in location-dependent biases.
  • 4. Where temperature gradients are steep (near the tropopause), AIRS

retrievals are allowed to depart more from the a priori, reducing the AIRS-ECMWF correlation.

slide-32
SLIDE 32

Comparison of AIRS-ECMWF-GPS Temperature Profiles

29 March 2007 AIRS Science Meeting, Pasadena

32

Simple Sigma Solutions, Mid North

Match-ups: 766 AIRS Quality: 0, 1

slide-33
SLIDE 33

Comparison of AIRS-ECMWF-GPS Temperature Profiles

29 March 2007 AIRS Science Meeting, Pasadena

33

ECMWF-GPS Means, 3-Way Matchups, 2003

slide-34
SLIDE 34

Comparison of AIRS-ECMWF-GPS Temperature Profiles

29 March 2007 AIRS Science Meeting, Pasadena

34

Smoothing of Sharp Tropopause

Match-ups: 2203 AIRS Quality: 0,1

  • 1

1

GPS-GPS Comparison Stats (COSMIC) Temperature (K)

25 20 15 10 5 Altitude (km)

slide-35
SLIDE 35

Comparison of AIRS-ECMWF-GPS Temperature Profiles

29 March 2007 AIRS Science Meeting, Pasadena

35

Typical 3-Way Matchup For consistent comparisons GPS & AIRS were smoothed to 2 km vertical resolution