CERES meeting, Newport News, May 2001 - PDF document

CERES meeting, Newport News, May 2001 Homogenisation of GERB andCERES fluxes.S. DewitteRoyal Meteorological Institute of Belgium CERES meeting, Newport News, May 2001 Overviewu0. Context: Climate Monitoring SAFu1. Homogenisation of GERB and CERES fluxesu2. Nature of expected errorsu3. General homogenisation methodologyu4. Radiance homogenisationu5. Flux homogenisationu6. Needed datau7. Conclusions CERES meeting, Newport News, May 2001 0. Context: Climate MonitoringSAFuSattelite Aplication Facilities (SAF Õs): projectsinitiated by EUropean METeosat SATellite(EUMETSAT) organisation for betterexploitation of (future) satellite datauClimate Monitoring SAF: aims to derivesatellite products with good quality and whichare consistent in time CERES meeting, Newport News, May 2001 Role of RMIB in CM SAFuDerive radiative fluxes at top of atmosphereuInput sources for incoming solar irradiance:solar irradiance measurementsuInput sources for reflected solar irradiance andemitted thermal irradiance: GERB and CERES CERES meeting, Newport News, May 2001 Flowchart for incoming solarirradiance Reference TSI measurements X Continuous TSI measurements TSI homogenisation Daily mean SARR solar irradiance at 1 A.U. Three hours mean incident solar flux over grid boxes Daily/monthly and monthly diurnal Averaging CERES meeting, Newport News, May 2001 Flowchart for reflected solar andemitted thermal irradiance A posteriori correction Three hours mean daily/monthly and monthly diurnal Averaging A posteriori correction 15' GERB footprint fluxes Objective characterisation of systematic errors Statistical analysis Database of collocated flux pairs Collocation High resolution 15' GERB fluxes Instantaneous CERES footprint fluxes feedback for a priori correction CERES meeting, Newport News, May 2001 1. Homogenisation of GERB andCERES fluxesuHomogenisation = merge datasets withoutintroducing discontinuitiesÐstatistical analysis : estimation of systematicdifferences in function of known parametersÐa posteriori correction : removal of systematicdifferences CERES meeting, Newport News, May 2001 2. Nature of expected errorsuSatellite measurement �- Unfiltered radianceÐprocessing: calibration, unfilteringÐexpected errors depend on scene typeuUnfiltered radiance� - FluxÐprocessing: angular modellingÐexpected errors depend on scene type and viewingangles CERES meeting, Newport News, May 2001 3.1 General homogenisationmethodologyuTo homogenise the data form two sources, acomparison and the choice of a reference isneeded.uDifference = source 1 - source 2= (source 1 - reference) - (reference - source 2)= error 1 -error 2 CERES meeting, Newport News, May 2001 3.2 Definition of comparison casesand binsuComparison/homogenisation can be doneindependently for number of cases c:Ðradiances, thermal flux: 3 surface scene types ocean,land, desertÐsolar flux: 3 surface scene types x solar zenith angleintervals CERES meeting, Newport News, May 2001 Comparison method = regressionue.g. flux comparison FCERES=A + B FGERBuperfect &#x-000;agreement ucloud classes are treated implicitelyÐsolar: low &#x-000;values clear sky high&#x-000; values cloudy skyÐthermal: low &#x-000;values cloudy sky high&#x-000; values clear sky CERES meeting, Newport News, May 2001 uFor every comparison case c data has to becompared for different angular bins b:Ðradiances: viewing zenith angle intervalsÐfluxes: viewing zenith and relative azimuth angleintervals CERES meeting, Newport News, May 2001 4.1 Radiance homogenisationuuse co-angular radiances onlyureference = (GERB + CERES)/2uerrorGERB= (GERB - CERES)/2uerrorCERES=(CERES-GERB)/2 CERES meeting, Newport News, May 2001 4.2 Practical implementationuregress CERES versus GERB radiancesLCERES=A + B LGERBuhomogenise radiances Lhomog.CERES=-A/2 + [1+(1- B)/2] LCERES Lhomog.GERB=A/2 + [1-(1- B)/2] LGERBuhomogenise fluxes - step 1 Fhomog.CERES=-pA/2 + [1+(1- B)/2] FCERES Fhomog.GERB=pA/2 + [1-(1- B)/2] FGERB CERES meeting, Newport News, May 2001 5.1 Flux homogenisationugood reference = mean flux averaged over allviewing anglesÐremoves most of the systematic errors dependent onanglesuproblem GERB: mostly backscattermeasurementsèreference = SbCERES cos(qvz) sin(qvz) / Sbcos(qvz) sin(qvz) CERES meeting, Newport News, May 2001 5.2 CERES flux homogenisationuchoose GERB data for one fixed GERB viewingangle bin bGERB as intermediate referenceuFor every possible CERES viewing angle bin b: regress CERES fluxes versus GERB fluxes forfixed GERB viewing angle bin bGERB : FCERES(b)=A(b) + B(b) FGERB(bGERB) CERES meeting, Newport News, May 2001 ucalculate reference regression parametersA= Sb A(b) cos(qvz) sin(qvz) / Sb cos(qvz) sin(qvz)B= Sb B(b) cos(qvz) sin(qvz) / Sb cos(qvz) sin(qvz)uhomogenise CERES fluxes relative to reference Fhomog.CERES = A - A(b) + (1+ B - B(b)) FCERES(b) CERES meeting, Newport News, May 2001 5.3 GERB flux homogenisationuGERB fluxes for all possible bins can behomogenised by regression againsthomogenised CERES fluxes CERES meeting, Newport News, May 2001 6. Needed datauCERES in RAPS mode : all viewing zenithangles and relative azimuth angles are covereduAll surface scene types and solar zenith angleintervals need to be covered in METEOSATfield of viewue.g. 6 RAPS days in August 1998 for TRMM CERES meeting, Newport News, May 2001 7. ConclusionsuA method has been proposed to homogeniseGERB and CERES fluxes.uThe method removes the angular dependentsystematic differences between GERB andCERES.uThe method will be tested using the 6 CERESRAPS days in August 1998 using GERB likedata derived from METEOSAT.