THE NOAA SSU STRATOSPHERIC TEMPERATURE CLIMATE DATA RECORD Cheng-Zhi Zou NOAA/NESDIS/Center For Satellite Applications and Research Haifeng Qian, Lilong Zhao, Likun Wang SPARC Temperature Trends Activity Workshop, Victoria, Canada April 9-10, 2015

Outline  The SSU Instrument  NOAA SSU Processing History  NOAA SSU CDR Processing Procedure  Uncertainty Analysis  Resolving Stratospheric Temperature Trend Mystery  Conclusion

The SSU Instrument  One of the TOVS instruments (MSU, HIRS, SSU) onboard NOAA TIROS-N polar-orbiting satellite series from  Supplied by UK Met Office  Infrared radiometer using pressure modulation technique to measure atmospheric radiation from CO  m  v 2 band  Three cells of CO 2 gas were placed in the optical path with different pressure which allowed three channels for a single CO 2 band P(peak)~P(cell)/[CO 2 ] 1/2, SSU channel weighting functions determined by cell pressure values Channel Peak of WF 15mb 5mb 1.5mb

NOAA SSU Processing History/Summary—  Raw SSU data and satellite orbital information were operationally processed at NOAA/NESDIS for each satellite along with other TOVS instruments and saved in operational level-1b files  Operational level-1b files were distributed to world-wide users and archived in NOAA/NCDC Comprehensive Large Array-data Stewardship System (CLASS) ; other centers may also have archives after receiving the data  Level-1c SSU radiances can be calculated from calibration coefficients stored in those level-1b files  UK Met Office conducted instrument testing and developed its version of SSU time series  Technical notes available for NOAA operational processing and UKMO instrument testing, but information for instrument calibration was insufficient and sometimes even missing

NOAA SSU Processing History/Summary– 2008-present  NOAA/NESDIS began to develop SSU temperature time series using level-1c radiances derived from operational level-1b data in 2008  STAR released its first version of SSU time series in 2011 (Wang et al. 2012) and introduced the data to SPARC group  Thompson et al. (2012) documented SSU trend differences between NOAA and Met Office versions and their difference from climate model simulations and called for re-investigation of the SSU problems  The SPARC temperature trend group organized a meeting in 2013 with participation from both NOAA and Met Office SSU groups to discuss technical issues in SSU processing; instrument space view anomalies were identified to be missing in operational level-1b processing  STAR recalibrated the SSU level-1c radiances using improved information and developed SSU FCDR and its Version 2 SSU temperature time series in 2014 (Zou et al. 2014)

Issues in Unadjusted SSU Dataset 6  Insufficient documentation– lost or no record of many calibration parameters  Bad data from bad calibration (NOAA-7 channel 2)  Space view anomaly—directly affects level-1c radiances  Blackbody target  Gas leaking in the CO 2 cell  weighting function variation  atmospheric CO 2 variations  limb-effect  diurnal drift effect  semi-diurnal tides  Residual biases Global mean anomaly time series for each satellite for radiances before adjustment and merging

NOAA SSU FCDR and TCDR Processing Flow Chart SSU recalibrated L1c swath radiances Adjusted SSU swath radiances (L1d) Gridded SSU Tb for individual satellites Merged SSU time series SSU raw counts data in L1b files SSU Recalibration: Converting raw counts to radiances Radiance adjustment including cell pressure, diurnal drift, atmos CO 2, limb-effect Binning adjusted swath radiances into grid cells Inter-satellite bias correction; global mean bias residual used to determine space view anomaly with NOAA-6 as a reference FCDR TCDR Need two processing cycles to complete

Recalibration to Derive Radiance FCDR Scan and Calibration Cycle S Slope I Intercept Calibration offset Space view anomaly Linear calibration conducted at each earth view position  space view anomalies were estimated using residual inter-satellite biases with known values of NOAA-6 as a reference

Physics of Cell Pressure Changes 9 SSU cell pressure Weighting function Measured BT CO 2 cell pressure decreasing -> weighting function peaks higher -> because of increasing lapse rate, measured Tb increasing -> warm bias

Cell Pressure from Gas Leaking and Its Correction Time Series 10 Cell pressure time series (Data from S.Kobayashi et al. 2009) CRTM simulated brightness temperature correction time series due to cell pressure changes relative a reference close to pre-launch specification Causes dramatic effect on time series

First-Guess Adjustment—Simple Constant Bias Correction 11 Global mean time series for individual satellites after constant bias correction Constant bias correction cannot remove satellite transition jumps

Effect of cell pressure correction 12 Global mean time series for individual satellites after cell pressure correction  Cell pressure correction dramatically removed the satellite transition jumps  Suggests cell pressure correction has good accuracy  Suggests that the NOAA-14 observations just after its start were good data  Channel 2 trend significantly lower than those from constant bias correction (grey lines) Mismatch between NOAA-11 and NOAA-14 will be taken care of by diurnal drift correction

NOAA Version 2 SSU Time Series Before Adjustment and Merging After Adjustment and Merging SSU Version 2 Temperature Anomaly Time Series

Uncertainty Analysis on FCDR 14  Instrument noise was less than specification (for TIROS-N) -- good for weather application  Noise is averaged out so it does not affect CDR accuracy  Calibration uncertainty due to warm load temperature uncertainty is small Radiometric noise (Specification, NESS107) Radiometric noise (Actual, Pick and Brownscombe 1981) Calibration error due to warm load temperature Ch Ch Ch Error Budget for the SSU FCDR and TCDR SSU Tb differences from using different warm load temperatures

Uncertainty Analysis on TCDR 15  Uncertainty due to space view anomaly included those from adjustment error and those from merging  Adjustment error not known, but expected to be less than 0.1 K due to excellent matching between satellite pairs  Need to be determined in future sensitivity experiments  Matching uncertainty mainly from short overlap between NOAA-9 and NOAA-11, which was estimated to be 1.4×  Error Budget for the SSU FCDR and TCDR SSU Tb differences from using different warm load temperatures Adjustment uncertainty (Expected) Matching Uncertainty from NOAA-9 and NOAA-11 Total Calibration Uncertainty Uncertainty in global mean trend (K/Dec) Ch Ch Ch

El Chichón Mt. Pinatubo SSU2:-0.76K/dec Model Mean:-0.77K/dec SSU1:-0.69K/dec Model Mean:-0.46K/dec SSU3(a) SSU2(b) SSU1(c) Fig. 2a Global temperature anomalies time series (K) in the period of at (a) SSU3, (b) SSU2, (c) SSU1. Note SSU1~SSU3 represent the SSU observational layers. Blue thick line represents the STAR observation. Grey lines indicate the results from CMIP5,Black thick line represents model ensemble mean

El Chichón Mt. Pinatubo SSU1:-0.69K/dec Model Mean:-0.56K/dec SSU1(c) Only high top model Model Mean:-0.46K/dec SSU1:-0.69K/dec all models SSU1(c)

Some Trend Characteristics NOAA Version 2 SSU dataset Plot from Seidel et al. 2011

Conclusion Consistent radiance FCDR was developed through recalibration where space view anomaly was included; dataset ready for application in climate reanalyses and other consistent satellite retrievals  Version 2.0 SSU time series were developed through careful adjustment and merging; adjustment included effects from cell pressure changes, diurnal drift, atmospheric CO 2 variation, and viewing angle differences  Structure trend uncertainty were estimated to be 0.05, 0.06, and 0.08 K/Dec for global mean SSU1, SSU2, and SSU3, respectively  Model simulations in CMIP5 ensemble means agree with SSU Version 2.0 within the uncertainty estimates for SSU2 and SSU3; their differences were a little larger for SSU1 but not significant  Recalibration, adjustment, and merging details were well documented in Zou et al. (2014)