Calibration and Validation Studies for Aquarius Salinity Retrieval

Shannon Brown and Sidharth MisraJet Propulsion Laboratory, California Institute of Technology

7th Aquarius SAC-D Science Meeting11-13, April 2012

Buenos Aires, Argentina

Project Overview

• Developing methods to track Aquarius calibration over Antarctica and rainforest regions

– Enables characterization of drift into gain and offset components– Ensures well calibrated brightness temperatures over full TB range

• Important for other applications (e.g. soil moisture)– Follows approach developed for altimeter radiometers (Topex, Jason)

• Investigation of roughness correction algorithms– Evaluating v1.2.3 roughness correction– Assessing dependence on sea state through match-ups with radar

altimeters

Aquarius Ocean Drift• Represents drift only at one TB – need another reference to determine whether it is a gain

or offset drift– Offset drift means constant drift at all TBs (e.g. front end path loss drift)– Gain drift means drift largest at cold TBs that approaches zero for warm scenes (e.g. ND drift)

3

Horn 1 V-pol

Horn 1 H-pol

Horn 2 V-pol

Horn 2 H-pol

Horn 3 V-pol

Horn 3 H-pol

Natural On-Earth Calibration Targets for Stability Tracking

• Rainforest (warm end TB ~280K)– Select depolarized heavily vegetated areas within the Aquarius swath– Use TMI and WindSat to determine canopy temperature to track

Aquarius calibration

4

6 V-H

• Antarctica (mid-range TB ~200K)– Select areas with stable temperature (V-pol TB) and

snow structure (H-pol TB)– Radiative transfer model used to determine L-band

TB over time using in-situ temperature and higher frequency microwave observations as input

Aquarius L-band Temporal Stability (Sep’11-Feb’12)

5

q= 28.7o

q= 37.8o

q= 45.6o

V-pol

H-pol

V-pol

H-pol

V-pol

H-pol

Ice Model

• Temperature profile determined from in-situ surface temperature data coupled with a heat transport model to determine T(z,t)

• AMSR-E 6.9-37 GHz V&H-pol observations constrain ice structure, ice dielectric model and thermal diffusivity

• MEMLS model (Wiesmann and Matzler, 1999) used to compute upwelling TB

6

Drift over Antarctica

7

Horn 2 V pol

Horn 2 H pol

-Aquarius minus Model over Antarctica

- Scaled ocean drift

• Aquarius – model TB shows drift that is approximately half the ocean drift for all channels which is consistent with an instrument gain drift

• Computed Aquarius – Model TB for each channel

Rainforest Regions

• Select regions in Amazon and Congo that exhibit small polarization signature at 6 and 10 GHz and contain Aquarius swath

– Opaque canopy obscures the surface– Exhibit little polarization or incidence angle

dependence– Brightness temperature closely tracks canopy

physical temperature

• 5 regions identified

Amazon Region TMI 10 GHz De-polarization

1

231

2

Congo Region TMI 10 GHz De-polarization

6 V-H

~ Canopy Temperature Variations

~ Vegetation properties variations

0.05 K2 K

Drift over Amazon

• Used simple parametric model to estimate Aquarius TBs from TMI/WindSat 10.7 GHz TB time series to estimate residual drift over warm rainforest regions

– TMI and WindSat data filtered for rain using flag based on 37-10 GHz TB difference

– Applied 30-day smoothing

• Negligible drift observed over warm regions – consistent with gain drift

Rainforest regions show potential to monitor drift at

the warm end

Roughness Correction Investigation

10

• Generating database of Aquarius match-ups with Jason-1/2 radar altimeters

– Find altimeter observations that fall within Aquarius footprint separated by < 1 hour

– Analyze roughness correction as a function of altimeter WS, SWH

• Database used to evaluate v1.2.3 roughness correction

11

Aquarius/Altimeter Match-ups

Number of match-ups per 1o bin – all horns

2-dimensional lookup table:[wind speed, σ0] → ΔTB

V1.2.3 Surface Roughness Correction:Wind Speed + Scatterometer σ0

Meissner and Wentz, Aquarius Cal/Val Workshop March 2012

V1.2.3 Excess TB Compared to Altimeter Winds

• Compared V1.2.3 specular TB (rad_TB_rc) to model using ancillary SST,SSS

• Generally unbiased with respect to altimeter wind speed– 0.1K level biases observed in some channels for winds < 2m/s and

>15m/s

13

Excess TB-V Bias vs Alt. WS Excess TB-H Bias vs Alt. WS

V1.2.3 Excess TB Compared to Altimeter Winds

• Standard deviation typically between 0.3-0.5K (slightly higher for H-pol)

• Standard deviation minimum near 9m/s - larger for low and high winds

– More pronounced for H-pol

14

Excess TB-V StdDev vs Alt. WS Excess TB-H StdDev vs Alt. WS

V1.2.3 Excess TB Compared to Altimeter WS/SWH

• Binned v1.2.3 model differences vs altimeter WS and SWH

• Roughness correction underestimated for low winds/high waves

• Over-estimated for high winds/low waves

• More pronounced for V-pol

15

Low winds/high waves

Low winds/low

waves

High winds/low

waves

High winds/high

waves

Horn 1 V-pol Horn 3 V-pol

All Channels

• Largest residual correlation with SWH in Horn1 (V&H) and Horn 3(V)

16

Horn1 V-pol

Horn1 H-pol

Horn2 V-pol

Horn2 H-pol

Horn3 V-pol

Horn3 H-pol

V1.2.3 SSS Bias and StdDev

• Lowest residuals when scatterometer is viewing surface in the along-wind direction

• Highest residuals in cross-wind direction for high wind speeds

17

Bias (psu) Std. Dev. (psu)

V1.2.3 Excess TB vs WS & Dir

• Highest residuals in excess TB when scatterometer viewing in cross-wind direction

• Suggests that hybrid approach may reduce errors– Weight observations for L3 product based on relative wind direction– Supplement scatterometer correction in cross-winds case with ancillary

data

18

Bias (psu) Std. Dev. (psu)

Summary

• Method developed to track Aquarius radiometer TB drift over Antarctica and Amazon

– References independent from ocean– Used to determine Aquarius drift is a gain drift

• Evaluation of v.1.2.3 roughness correction shows little bias with respect to altimeter WS but some residual correlation with SWH

• Highest residuals from scatterometer correction when Aquarius is viewing cross wind

– Suggest improvements can be gained from hybrid algorithm relying more on ancillary data in cross-wind cases

19

• Level of gain vs offset drift will depend on which components are changing

• Example: Drift in noise diode brightness creates gain drift

• Largest drift for cold TBs, small drift at warmer TBs that are close to internal reference load temperature

Separation of Gain vs Offset Drift

22

ND

NDREFBBB T

tTTTtTT = ,

5.0

300100300200

,100,200

=

tKTTtKTT

BB

BB

= ,,,, LTTLtTTTtTT Offset

ND

NDREFBBB

ND drift will cause TB drift over Antarctica which is ~0.5 x ocean drift

05.0

,100,290

tKTTtKTT

BB

BB

TB drift will approach zero over warm regions

23

Tracking Aquarius TB over Rainforest

• Ferrazzoli and Guerriero (1999) and other modeling studies show that for high biomass, optical depth is >>1 and emissivities (defined by scattering from the forest) are within few %) between L-band , C-band and X-band

• Track variation of L-band TB as a function of time from TMI and WindSat 10.7 GHz observations for morning passes (4-7 LST)– Annual surface temperature variations

less than 2K in the morning over these regions so a few percent uncertainty in scaling factor not critical for estimating temporal variability

24

1~1

2

12

=

a

bfaTfT BB

Drift over Rainforest• Heavily vegetated regions act like pseudo-blackbodies

– Opaque canopy obscures the surface– Exhibit little polarization or incidence angle dependence– Brightness temperature closely tracks canopy physical temperature

• Estimated canopy TB over rainforest regions as a function of time from TMI and WindSat 10.7 GHz observations for morning passes (4-7 LST)

• Negligible drift observed over warm regions – consistent with gain drift

25

5 Regions identified in Amazon and Congo which exhibit little

polarization signature from 6.9-37 GHz

See Brown and Misra talk Thursday at 16:10 for details

Tracking Aquarius at Warm TBs

• Differenced Aquarius TBs from TMI/WindSat 10.7 GHz TB time series to estimate residual drift over warm rainforest regions

– TMI and WindSat data filtered for rain using flag based on 37-10 GHz TB difference

– Applied 30-day smoothing

• Congo region 1 showed largest temporal variability (correlated in both TMI and Aquarius data)

V-pol Horn 1 All Regions

Warm Reference

• Over warm regions (TB ~ 280K) we should see almost no drift in Aquarius TBs if it is a gain drift

• Heavily vegetated regions act like pseudo-blackbodies (Brown and Ruf, 2005)

– Opaque canopy obscures the surface– Exhibit little polarization or incidence angle dependence– Brightness temperature closely tracks canopy physical temperature– TB near 280K

sfcscanopySB TeTeeT canopycanopycanopy qqq secsecsec 1111 =weakly scattering canopy

canopyB TffT )(1)( =

1for

Developing On-Earth TB Calibration References at L-band

• Natural targets for L-band radiometer calibration over on-Earth dynamic range

– Calm, flat ocean scenes – Cold reference– Ice sheets: Antarctica, Greenland – Mid-range reference– Land areas: flat, dry deserts; homogeneous heavily

vegetated regions – Hot reference

• Use to assess absolute calibration, monitor stability and assess residual instrument calibration errors

37 V-H

23 V-H

18 V-H

10 V-H

6 V-H

AMSR-E De-polarization

Stability of Regions

• TBs stable to ~2K over these regions over several years– C-band to Ka-band highly correlated

• 6.9 GHz polarization difference stable to 0.05K– Vegetation microwave properties vary little with time

AMSR-E TBs Amazon Region 1April 2008 – October 2010

AMSR-E 6.9 GHz TBV-TBHApril 2008 – October 2010

0.05 K

TB with Model – All Channels

• All channels in good agreement with model after applying gain drift correction based on ocean TA drift estimates

30

V1.1 TB not corrected for

drift

TB after applying gain

drift correction (v1.2.3)