Remote Sensing Systems Climate Satellite Program

Frank J. Wentz and Carl MearsRemote Sensing Systems, Santa Rosa, CA

Supported in part by : NASA’s Earth Science Division

NOAA’s Scientific Data Stewardship and Climate Program Office

Presented at: CICS Mtg., Univ. of Maryland, September 8, 2010

Engineering Climate Data Records: The Challenges

Large Volume: Over 150 satellite-years of observations from Microwave Radiometers

Microwave Imagers (13) SSM/I: F08, F10, F11 F13 ,F14, F15 SSM/IS: F16, F17, F18 TMI AMSR-E and AMSR WindSat

Microwave Sounders (16) MSU: Tiros-N, NOAA 6, 7, 8, 9, 10, 11, 12, and 14 AMSU: NOAA 15, 16 ,17 18, and 19, Aqua, MetOP A

Difficult Calibration: Each sensor has its unique set of problems

Sensor pointing and S/C attitude errors Antenna pattern knowledge error Scan-dependent errors Sun intrusion and thermal gradients in Hot Load Emissive Antenna

High Precision Required: Climate Variability typical 1% of the Mean

Engineering Climate Data Records: The Method

Maintain a High Level of ConsistencyThe steadfast adherence to the same principles and methods in data processing all the way from radiometer counts to climate time series, including:

GeolocationRadiometer calibrationGeophysical retrievals

On-Going and Comprehension ValidationRadiosondes temperature and vaporGPS vaporOcean buoy winds and SSTScatterometer wind retrievalsSST, Cloud and Rain retrievals from satellites IR sensors

Much of this validation is done by the User community via peer reviewed papers.

Sensor Calibration and Geophysical Retrievals Use Same Radiative Transfer Model = Same Retrieval Algorithm for all Sensors

All MW imagers are calibrated to the same RTMAll retrievals come from the inverse of the RTM (RTM-1)

The Fundamental, Fundamental Data Record: Radiometer Counts

Consistency in data processing requires the antenna temperatures or brightness temperature obtained from various data providers like TDRs from NOAA, NASA Level 1 dataset , JAXA L1A data be reversed engineered back to the original radiometer counts in the telemetry downlink. Also satellite position, velocity, and attitude vectors are required

RSS maintains a complete archive of radiometer counts (both earth view and calibration), and associated spacecraft ephemeris and housekeeping data (thermistors).

Dataset is fully q/c and organized into time-ordered orbital files.

Volume size is relatively small

Non-Controversial

Stable

Basis for all product generation at RSS (Ta, Tb, CDR, gridded products)

Geophysical Retrievals

Validation EP Adjustments (i.e., clear sky bias, high vapor bias)

Retrieval Algorithm Radiative Transfer Model

Simulated Antenna Temperatures

Sensor Antenna Temperatures

Sensor Adjustments RTM Adjustments

Automatic

Calibration

Methodology: Continuous Updating and Reprocessing

Cycle Time ≈ ½ Year

Radiometer Counts

Same ΔTa (simulated minus measured) plotted versus different parameters Same color scale: ΔTa goes from -3K to +3K

0.5 1 1.5 2 2.5

x 104

0102030405060708090

100110120130140150160170180190200210220230240250260270280290300310320330340350360

-3

-2

-1

0

1

2

3

0.5 1 1.5 2 2.5

x 104

0102030405060708090

100110120130140150160170180190200210220230240250260270280290300310320330340350360

-3

-2

-1

0

1

2

3

0.5 1 1.5 2 2.5

x 104

0102030405060708090

100110120130140150160170180190200210220230240250260270280290300310320330340350360

-3

-2

-1

0

1

2

3

0.5 1 1.5 2 2.5

x 104

0102030405060708090

100110120130140150160170180190200210220230240250260270280290300310320330340350360

-3

-2

-1

0

1

2

3

0 10 20 300

10

20

30

-2

0

2

0 10 20 300

10

20

30

-2

0

2

0 10 20 300

10

20

30

-2

0

2

0 10 20 300

10

20

30

-2

0

2

0 20 40 600

10

20

30

-2

0

2

0 20 40 600

10

20

30

-2

0

2

0 20 40 600

10

20

30

-2

0

2

0 20 40 600

10

20

30

-2

0

2

0 10 20 300

20

40

60

-2

0

2

0 10 20 300

20

40

60

-2

0

2

0 10 20 300

20

40

60

-2

0

2

0 10 20 300

20

40

60

-2

0

2

40 60 80 100 120 140

40

60

80

100

120

140

-3

-2

-1

0

1

2

3

40 60 80 100 120 140

40

60

80

100

120

140

-3

-2

-1

0

1

2

3

40 60 80 100 120 140

40

60

80

100

120

140

-3

-2

-1

0

1

2

3

40 60 80 100 120 140

40

60

80

100

120

140

-3

-2

-1

0

1

2

3

Y=Wind, X=SST Y=Wind, X=Vapor Y=Vapor, X=SST

Y=Orbit Position, South Pole to South Poel, X=Orbit number (5 years) Y=Sun Polar Angle, X=Sun Azimuth Angle

RTM Error Diagnostics

Sensor CalibrationError Diagnostics

Distinguishing Sensor Errors from RTM Errors

Sun intrudingInto hot load

240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 27520

21

22

23

24

25

Wa

ter

Va

por

(mm

)

Months (240=Dec 2006)

240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 27520

21

22

23

24

25

Wa

ter

Va

por

(mm

)

Months (240=Dec 2006)

240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 27520

21

22

23

24

25

Wa

ter

Va

por

(mm

)

Months (240=Dec 2006)

F17F13

F17F13

F17F13

Comparing F17 and F13 Water Vapors (Monthly Averages)

Version 7

Version 7

Version 7

Bias (SSMI minus GPS) = -0.07 mmStd. Dev. = 1.9 mmNum Obs = 53,730Credit: Carl Mears (to be published)

Geophysical Validation: Columnar Water Vapor

Version – 7 100 Satellite-Years of Earth System Data Records Consistently Processed with Common AlgorithmsSame RTM Used for Calibration and Retrievals for all Sensors

The Products

SST (not all sensors) Wind Speed Columnar Water Vapor Columnar Liquid Water Rain Rate

The Sensors

SSM/I: F08, F10, F11, F13, F14, F15 SSM/IS: F16, F17, F18 TMI AMSR-E and AMSR-A WindSat

AvailabilityBackbone: F13, WindSat, F16, and F17 now available

Remaining Sensors will be made publicly available end of 2010.

Products will be hosted at www.remss.com

MSU/AMSU Climate Data Records

Trend Map, Middle Troposphere (1979-2009)

Time Series for Each Channel

MSU/AMSU Uncertainty AnalysisTwo types of Analysis

“Bottom Up”Estimated Errors Propagated Through Merging Procedure using Monte-Carlo Techniques• Sampling• Adjustments (local time, etc)• Residual Calibration Errors

“Top Down”Compare to other measurements of the same parameters

• Other sounder datasets

• Radisondes

• GPS-RORange of simulated errors

Satellite, Radiosonde and ReanalysisComparison

BACKUP SLIDES

Individual overpasses: RMS variation of SSMI minus bouy difference = 1.0 m/sMonthly Averages: RMS variation of SSMI minus bouy difference = 0.1 m/s ( lag-1 correlation =0.46)Estimate error bar is 0.05 m/s/decade at 95% confidence SSM/I trend minus the buoy trend is 0.02 m/s/decade.

Yearly corrections (0.05-0.10 m/s) are applied to SSMI winds (equivalent to 50 mK TB adjustments)These results indicate we are doing better than 0.1 K/decade

Geophysical Validation: Wind Speed Trends

Figure from : Wentz, Ricciardulli, Hilburn, Mears: Science, July 13, 2007, How Much More Rain Will Global Warming Bring?

Climate Constraint: Constant Relative Humidity

Correlation Coef.W::TS = 0.84W::TA = 0.71TA::TS = 0.69

Correlation Coef.W::TS = 0.98W::TA = 0.93TA::TS = 0.94

Scaling Coef.W/TS = 9.2%/KW/TA = 6.7%/KTA/TS = 1.4

Scaling Coef.W/TS = 8.5%/KW/TA = 6.8%/KTA/TS = 1.2

Scaling Coef.W/TS = 8.6%/KW/TA = 4.6%/KTA/TS = 1.9

Correlation Coef.W::TS = 0.73W::TA = 0.36TA::TS = 0.56

WTS

TA/1.6

Decadal TrendW = 1.9%/decTS = 0.20K/decTA = 0.04K/dec

WTS

TA/1.6

WTS

TA/1.6

Decadal TrendW = 2.1%/decTS = 0.23K/decTA = 0.22K/dec

Decadal TrendW = 1.0%/decTS = 0.03K/decTA = 0.00K/dec

Tem

per

atu

re (

K)

Vap

or

(%)

-0.4

-0.2

0.0

0.2

0.4

-4

-2

0

2

420N to 60N

Tem

per

atu

re (

K)

Vap

or

(%)

-0.4

-0.2

0.0

0.2

0.4

-4

-2

0

2

420S to 20N

Year

Tem

per

atu

re (

K)

Vap

or

(%)

79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98

-0.4

-0.2

0.0

0.2

0.4

-4

-2

0

2

460S to 20S

Figure from : Wentz and Schabel: Nature, 403,2000, Precise climate monitoring using complementary satellite data sets.

Nearly all climate models predict a near constant Relative Humidity with global warming.This means total water vapor will increase with rising air temperatures at a rate ≈ 7%/K

Original MSU air temperatures showed little warming while SSM/I water vapors showed significant moistening.

This contradiction was due to errors in the original calibration of MSU.

Results here show with proper calibration MSU air temperature and SSM/I water vapors agree.

The “MSU controversy” was resolved.

Parameter Mean Std. Dev. Trend Evaporation 961 mm/yr 10.1 mm/yr (1.1%) 12.6 mm/yr/decade (1.3%/decade) Precipitation 950 mm/yr 12.7 mm/yr (1.3%) 13.2 mm/yr/decade (1.4%/decade) Total Water 28.5 mm 0.292 mm (1.0%) 0.354 mm/decade (1.2%/decade)

Climate Constraint: Evaporation = Precipitation

Figure from : Wentz, Ricciardulli, Hilburn, Mears: Science, July 13, 2007, How Much More Rain Will Global Warming Bring?

On global, monthly time scales, Evaporation must equal Precipitation (Variability in storage term is extremely small)

Climate Constraint: Radiative Cooling Limit on Evaporation

Climate Models: 4 mm/year/decadeWentz et al., Science, July 13, 2007: 13 mm/year/decade (C-C rate)Yu and Weller, BAMS, Vol 88, No 4, April 2007: 50 mm/year/decade J. R. Christy et al., Geophys. Res. Lett., 28, 183, 2001: 21 mm/year/decade Hamburg Ocean Atmos. Parameters & Fluxesfrom Satellite Data (IGARSS 2006 presentation) very large

Explanation of Christy Result

Found a trend in the Tropical Air-Sea Temperature Difference: -0.1 K/decade (1979-1999)

Used to support their MSU Lower Tropospheric Temperature Retrieval, which indicated: In the tropics, the troposphere is not warming as fast as the surface. (Later was found to be incorrect)

From the standpoint of evaporation:A trend of -0.1 K/decade in the air-sea temperature difference would greatly increase the trend in global evaporation

Nearly all climate models predict an enhanced radiative cooling that is balance by an increase in latent heat from precipitation

Geolocation

Analysis

Attitude

Adjustment

Along-Scan

Correction

Absolute

Calibration

Hot Load

Correction

Antenna

Emissivity

Resampling

Algorithm

Rain

Threshold

OOB

Q/C

SSM/I F08, F10, F11 F13 ,F13, F15

NRL/RSS No Yes APC Yes 0 Opt. Intrep. 0.18 mm Ocean RTM

TMI Goddard Dynamic Yes TA Offsets No 3.5% Opt. Intrep. 0.18 mm Ocean RTM

AMSR-E RSS Fixed Yes APC Yes 0 Opt. Intrep. 0.18 mm Ocean RTM

AMSR-A RSS Dynamic Yes APC Yes 0 Opt. Intrep. 0.18 mm Ocean RTM

WindSat NRL/RSS Fixed Yes APC Yes 0 Earth-Grid Weighted Average

0.18 mm Ocean RTM

SSM/IS

F16, F17

RSS No Yes APC Yes 1 - 4%

Note: 1

Opt. Intrep. 0.18 mm Ocean RTM

STEP-1: All Level-1 (NASA) or TDR (DMSP) datasets are reversed engineered back to radiometer counts

Level-1 Processing: Counts-to-Antenna Temperature

STEP-2: Apply a completely consistent set of Level-1 Routines to produce calibrated antenna temperatures

Note 1: Over 19 – 92 GHz

Preliminary Results Over Land

266 268 270 272 274 276 278 280 282 284 2860

0.2

0.4

0.6

0.8

1

Antenna Temperaure (K)

PD

F w

ith m

axim

um

se

t to

1

evening F13 SSMIevening F16 SSMISevening F17 SSMISevening Windsat

266 268 270 272 274 276 278 280 282 284 2860

0.2

0.4

0.6

0.8

1

Antenna Temperaure (K)

PD

F w

ith m

axim

um

se

t to

1

morning F13 SSMImorning F16 SSMISmorning F17 SSMISmorning Windsat

270 272 274 276 278 280 282 284 286 288 2900

0.2

0.4

0.6

0.8

1

Brightness Temperaure (K)

PD

F w

ith m

axim

um

se

t to

1

evening F13 SSMIevening F16 SSMISevening F17 SSMISevening Windsat

270 272 274 276 278 280 282 284 286 288 2900

0.2

0.4

0.6

0.8

1

Brightness Temperaure (K)

PD

F w

ith m

axim

um

se

t to

1

morning F13 SSMImorning F16 SSMISmorning F17 SSMISmorning Windsat

Calibration is done just over ocean.Land can be considered a withheld data set.How well does the ocean calibration extrapolate to land?Key factors: Radiometer linearity and accuracy of ocean RTM

Histograms of 19 GHz V-pol at the high end: TA above and TB below.F13, F16, F17 consistent to about 0.25 K. Windsat has different spatial resolution