Climate studies using 12+ years of AIRS radiancesS. De Souza-Machado1, L. L. Strow1, A. Tangborn1, C. Hepplewhite1, H. Motteler1, S. Buczkowski1, P. Schou1

1Joint Center for Earth System Technology, Univ. of Maryland Baltimore County

1 AIRS instrument

ñAIRS: Grating spectrometer sounder on EOS-AQUA, 2378 channels with spectralresolution ∼0.5-2.0 cm−1 from 650-2665 cm−1. 1:30 am/pm orbit. AIRSoperational for 12+ years, intensively validated.

ñ Look at zonally averaged radiance changes from two different subsetsñ Calibration Stability DataSet which are clear-sky scenes over oceanñ All-sky scenes along the nadir track, ocean/land

AIRS, IASI, CrIS producing low noise, highly accurate spectra for 12+ years.Can be used for trending at climate level of WV, temperature and trace gases12 years not enough for climate, so at this stage we are mostly testing AIRSobservations against ERA re-analysis fieldsAlso examine gridded (4x4 ◦grids) AIRS radiance data for evidence ofstochastic forcing

2 Approach for trends

ñCreate zonally averaged daily subsets (all i=1 .. 2378 channels)ñMatch to re-analysis T(z), WV(z), stemp (and cloud fields) from ERA-Interimñ Find linear trend (dBT/dt) from time-seriesñ The linear trends bi(ν) of the AIRS spectral radiances are used to retrieve a

variety of geophysical trends using an optimal estimation retrieval approachñClear sky bins span about ± 70 deg, narrowest bins at tropicsñAll-sky zonal bins are 5 deg wide, spanning -90S to +90N, over land/ocean

3. Calibration Stability Data Set

Graphs showing (L) sample clear sky rates and (R) fitting errors

1000 1500 2000 2500

−0.3

−0.2

−0.1

0

0.1

0.2

Wavenumber (cm−1

)

∆ B

(T)

in K

/ye

ar

55 S

1.2 N

29 N

500 1000 1500 2000 25000

0.01

0.02

0.03

0.04

0.05

0.06

Wavenumber (cm−1

)

Rate

2−

σ E

rror

(K/y

ear)

Serial Correlation Corrected

Raw Fit Error

4. CO2 and Surface Temp results

Graphs showing (L) CO2 rates and (R) surface temp ratesCO2 rates have highly accurate global reference data. BUT CO2, T(z) Jacobianvary Co-Linearly, so need to do an Obs-correction to accurately fit CO2 rates

−60 −40 −20 0 20 40 60

1.4

1.6

1.8

2

2.2

2.4

Latitude

CO

2 G

row

th R

ate

(ppm

/year)

Obs

Obs−ERA

In−Situ

In−Situ Interp

ñ red : CO2 trends after Obs-correction

−60 −40 −20 0 20 40 60−0.1

−0.08

−0.06

−0.04

−0.02

0

0.02

0.04

0.06

Latitude

SS

T R

ate

(K

/ye

ar)

Obs (no SW)

ERA

Obs (all ν)

ñ Retrieved SST Rates versus ERA rates

5. All-Sky rate retrievals

ñCloud/geophysical Jacobians fromERA

ñMinor gas, cloud parameters(OD/size for liquid, ice clouds)

ñ Zero A-priori, Tikhonov L1

regularization.

Approach works as on average, scenesare almost-clear!!!! (10 K cloud effect)

−100 −50 0 50 100220

240

260

280

300

320

Latitude

Me

an

BT

12

31

(K

)

AIRS Obs

SARTA CLD

SARTA Clr

BT (K) at 1231 cm−1 vs Latitude

6.Greenhouse Effect

Greenhouse Spectral Changes in Radiative Forcing

Wavenumber cm-1

800 1000 1200 1400 1600 1800 2000 2200 2400 2600

d(B

T)/

dt (K

/yr)

-0.08

-0.06

-0.04

-0.02

0

0.02

0.04

Cld feedback

Greenhouse Gas Forcing

Synthesized ClrSky rates

Observed AIRS AllSky rates

Red curve is AIRS allsky rates, averged over whole globeBlue curve is global synthesized clear sky ratesThe negative trends from 650 cm−1 to 800 cm−1 are due to CO2 increases, whichdecrease radiation to spaceThe 800 cm−1- 1200 cm−1 positive trends could be cloud feedbacks

7. T(z) and WV(z) All-Sky rate retrievals

Geophysical rates derived from AIRS all-sky radiance trends (UMBC-AIRS) agreebetter with re-analysis (ERA and MERRA) rates, than with AIRS L3 rates. The L3product is derived from AIRS L2 retrievals.

T(z) rate from UMBC-AIRS WV(z) rate from UMBC-AIRS

−50 0 50

10

100

1000

d(T)/dt K/yr

−0.1

−0.05

0

0.05

0.1

0.15

−50 0 50

10

100

1000

d(W)/dt frac/yr

−0.02

−0.01

0

0.01

0.02

8. Skewness and Kurtosis of window channel centered at 1231 cm−1

(L) Skewness and (R) log10(Kurtosis) shown on a 4 × 4 grid. Regions of highSkewness and Kurtosis are where non-Gaussian extreme events are more likely.

Longitude [deg]

-100 0 100

Latitu

de [deg]

-50

0

50

-8

-6

-4

-2

0

Longitude [deg]

-100 0 100

Latitu

de [deg]

-50

0

50

-4

-3

-2

-1

0

1

2

9. Theoretical Background

ñ Physical processes can be separated into into fast and slow modes. The slowmodes are deterministically represented while the fast modes are approximatedby a state dependent noise process.The system can then be modeled by a stochastic different equations (SDE):

dxdt= A(x) + B(x)η(t) (1)

Here the slow processes are described by A(x) and B(x)η(t) is anapproximation to the fast processes where η(t) is a noise vector. The stochasticforcing is then state dependent (through B(x)) which has been shown to be onepossible scenario that leads to non-Gaussian PDFs.

ñWhen Correlated Additive and Multiplicative (CAM) noise is present, thestochastic model of Equation 1 yields a relationship between the excess kurtosis(K =< x >4 /σ4 − 3) and skewness (S =< x >3 /σ3):

K ≥ 32

S2 − r (2)

10. Skewness vs. Kurtosis for 1231 cm−1 channel

AIRS observations BT Calculations from ERA

−2 −1.5 −1 −0.5 0 0.5 1 1.5 2−2

0

2

4

6

8

10

Channel = 1291, Observations

Skewness

Excess K

urt

osis

−2 −1.5 −1 −0.5 0 0.5 1 1.5 2−2

0

2

4

6

8

10

Channel = 1291, Calculations

Skewness

Excess K

urt

osis

Clear sky 1231 cm−1 Observations and ERA simulations look very similar. Thereare a number of data points below the curve; this possibly arises from using asimple model to describe complex multidimensional dynamics.

http://asl.umbc.edu sergio@umbc.edu