10 likes | 139 Views
Calibrate High-Quality (HQ) Estimates to “Best”. Instant-aneous SSM/I TRMM AMSR AMSU. HQ coefficients. Merge HQ Estimates. 3-hourly merged HQ. Match IR and HQ, generate coeffs. 3-hourly IR Tb. 30-day IR coefficients. Apply IR coefficients. Hourly HQ-calib IR precip.
E N D
Calibrate High-Quality (HQ) Estimates to “Best” Instant-aneous SSM/I TRMM AMSR AMSU HQ coefficients Merge HQ Estimates 3-hourly merged HQ Match IR and HQ, generate coeffs 3-hourly IR Tb 30-day IR coefficients Apply IR coefficients Hourly HQ-calib IR precip Merge IR and HQ estimates 3-hourly multi-satellite (MS) Compute monthly satellite-gauge combination (SG) Monthly gauges Monthly SG Monthly Climo. Adj. Coeff. Rescale 3-hourly MS to monthly SG Rescaled 3-hourly MS Upgrades to the Real-Time TMPA Poster PT-2 G.J. Huffman1,2, D.T. Bolvin1,2, EJ. Nelkin1,2, R.F. Adler3, E.F. Stocker1 1: NASA/GSFC Earth Sciences Division 2: Science Systems and Applications, Inc. 3: Univ. of Maryland College Park/ESSIC http://precip.gsfc.nasa.gov A Note on Terminology General approach: TRMM Multi-satellite Precipitation Analysis [TMPA] Real-time product: 3B42RT [RT], providing two precipitation fields • original multi-satellite, without climatological calibration [RT(U)] • multi-satellite with climatological calibration, [RT(C)] Version 7 post-real-time production product: 3B42 [P] 1. TMPA SYSTEM - Data Sources and Processing 2. RT ENHANCEMENTS FOR VERSION 7 A diverse, changing, uncoordinated set of input precip estimates, with various • periods of record • regions of coverage • sensor-specific strengths, limitations Seek the longest, most detailed record of “global“ precip Combine the input estimates into a “best” data set Run twice: • (near-)real time (~8 hr after observation time) • research quality (~2 months later, after “all” data are received) • the focus here is the real-time Version 7 RT Will Include Retrospective Processing Upgraded Input Data Should Yield a Better Product in Version 7 V7 vs. V6 P – July 2008 Users requested this to provide • an extended data record • with consistent processing • for quantitative work (hydrology, crop forecasting) Most existing input data sets are upgraded Additional periods of record are included (boxes) • some parts of the data records are not suitable – shown with lighter-color lines SSMIS is used for the first time Dotted vertical line indicates start of CPC Merged IR dataset • the P before this time depends on a different IR data set not considered in the RT 2002 2006 2004 2000 2008 2010 2012 Lacking Version 7 RT, we compare an example month of Versions 6 and 7 P, which should be similar to comparing Versions 6 and 7 RT Comparing Version 6 to Version 7 P for July 2008 (general product name is 3B43) • this is the first sample month • Version 7 tends to be somewhat higher over ocean • the Southern Ocean is lower, perhaps due to seasonal effects? • changes over land should be driven by differences in the gauge analysis • Version 6 gauge for this month is CAMS; Version 7 gauge is GPCC throughout • dramatic changes in the Indian Monsoon area are likely due to data gaps in CAMS • Version 7 values are considered better TMI SSM/I F13 SSM/I F14 SSM/I F15 SSMIS F16 SSMIS F17 V7 3B43 Jul 2008 (mm/d) SSMIS F18 Processing Sequence Both RT and P computed on a 3-hr 0.25° grid Microwave precip: • intercalibrate, combine IR precip: • calibrate with microwave Combined microwave/IR: • IR fills gaps in microwave Both RT and P calibrate the initial 3-hr MS • RT does monthly climatological calibration on 1° grid • TMI-TCI calibration first, then gauge • uses 10 years of match-ups, broken into monthly fields Both RT and P will archive both uncal. and cal. 3-hr MS AMSR-E AMSU N15 AMSU N16 V6 3B43 Jul 2008 (mm/d) AMSU N17 The previous RT record was inhomogeneous in time • black hash marks denote substantial changes in input data types, input data algorithms, or RT processing • the only fully consistent record was October 2008 – present • without reprocessing, the changes listed above would have implied starting over with a new, different data series when Version 7 RT starts (in 2012) Retrospective processing provides a uniformly processed 12-year record • P is still the dataset of choice for research – the goal here is to facilitate calibration of real-time applications There are still caveats: • changes in the satellite constellation still create potential inhomogeneities • the retrospective processing for RT will be run on the P input data, introducing the possibility of small differences compared to running in real time on the RT input data MHS N18 MHS N19 MHS MetOp (V7) – (V6) Jul 2008 (mm/d) Comparing Version 6 and Version 7 P to atoll gauge data for July 2008 • atoll data from the PACRAIN archive at Univ. of Oklahoma • the western tropical Pacific yielded 8 atolls, each corresponding to a 0.25°x0.25° grid box of P data • the average biases are: • Version 6: -1.60 mm/d (-20%) • Version 7: -0.56 mm/d (-7%) • consistent with maps above CPC Merged IR 2002 2006 2004 2000 2008 2010 2012 Original RT Retrospective Version 7 RT Upgrades and Reprocessing Episodically, the TMPA is upgraded to take advantage of • new/revised input data • improved processing concepts Ideally, upgrades are accompanied by reprocessing to ensure consistency of the entire data set • built in to P system • new concept for RT – this is the first time 3. RT FEATURES THAT REMAIN THE SAME 4. FINAL REMARKS RT Format Remains the Same RT Is Again Climatologically Calibrated to P Native binary • trmmopen.gsfc.nasa.gov • 3B40RT: microwave precipitation, error, pixel counts, source • 3B41RT: IR precipitation, error, pixel counts • 3B42RT: combined microwave/IR precipitation (both original and calibrated to 3B42), error, source KMZ (for Google Earth) • trmmopen.gsfc.nasa.gov • 3B42RT: combined microwave/IR precipitation(calibrated to 3B42) Flat binary • through TOVAS • 3B42RT: combined microwave/IR precipitation (calibrated to 3B42) Text • through TOVAS • 3B42RT: combined microwave/IR precipitation (calibrated to 3B42) TOVAS on-line visualization • 3B40RT: microwave precipitation • 3B41RT: IR precipitation • 3B42RT: combined microwave/IR precipitation (calibrated to 3B42) Computed on coincident 3-hr, 0.25° data in the band 50°N-S for uncalibrated (RT(U)) and calibrated 3B42RT (RT(C)) against the production Version 6 3B42 (P) Lacking Version 7, we show the performance for Version 6, which should be similar Better values are colored in green Bias • mixed results • global values are just a small residue of larger regional differences • worse global-average bias could result from improvement mainly in regions of one sign RMS • calibrated version almost always better • implies improvement in most regions Boreal cold season (Jan 2007, Dec 2008) • calibrated land is marginally worse for both Bias and RMS • most land is in the Northern Hemisphere – believed to reflect relatively large cold-season errors The new Version 7 Real-Time TMPA is being readied for release • approx. calibrated to the Version 7 production post-real-time TMPA • includes upgrades to most input data sets • includes SSMIS input data for the first time • provides retrospectively processed data starting in Feb. 2000 The expected release date for RT is late Spring 2012 • the pacing item is completing computation of climatological calibrations to the Version 7 calibrators • thereafter, both retrospective processing and processing for new data can begin • Version 6 RT processing will continue during the switch-over Dec’08 Nov’08 Oct’08 Oct’07 Jul’07 Apr’07 Jan’07 Questions or to to be put on the notification e-mail list: george.j.huffman@nasa.gov Web pages: http://precip.gsfc.nasa.gov http://trmm.gsfc.nasa.gov 3B42/43 document: ftp://meso.gsfc.nasa.gov/pub/trmmdocs/ 3B42_3B43_doc.pdf 3B4xRT documents: ftp://meso.gsfc.nasa.gov/pub/trmmdocs/rt/