METOP/AVHRR processing chain at MF/CMS G. Legendre, A. Marsouin, P. Le Borgne,
context synoptic of the chain ingest application sst application mask indicator SST calculation determination of confidence level remapping match up data base
Overview Context –OSI SAF of EUMETSAT (CDOP) Objectives Constraints –CMS operational environment –ArchiPEL framework Calendar –requirements 07/ /2006 –design 09/ /2006 –development/test 10/ /2007 –validation 06/ /2008 ( using a concurrent test chain ) productcoverageresolutionfrequencyformattimeliness FRESswath/globalfull (1km)every granule L2P4h NAREuropean seas2kmtwice a day L3P, GRIB2 4h GLBglobal0.05°12 hourlyL3P, GRIB2 6h
Functional diagram L1C data + MAIA cloud mask cloud mask control SST calculation confidence value determination full resolution SST FRES product L2P GLB product L3P, GRIB2 NAR product L3P, GRIB2 archive landmask / distance to coast maxi gradient climatology ( URI /Belkin & DMI/Ansersen) SST climatology mini (NOAA/Casey) ice edge and conc. (OSI SAF) aerosols (NAAPS) format aerosols (SDI, NAAPS) SSI (SEVIRI,GOES-E,NWP) WSP (NWP) ICE (OSI SAF, NWP) AOD (NAAPS) SST analysis (FNMOC) SSES SST climatology mean (NOAA/Casey) MDB
ArchiPEL synoptic view
ingest application check the chronology –metagranules are expected in chronological order –difference between 2 scans is expected to be 166ms ± ε –no "missing line" format in netCDF –IR and VIS channels –location, satellite and solar angles (interpolated) –cloud mask (strict and relaxed) manage the head and the tail margins metagranule nHT metagranule n+1HT
sst application 1.remap the landmask 2.remap the climatologies 3.init the mask indicator 4.compute the gradient indicator 5.compute strato aerosol indicator 6.compute the saharan dust indicator 7.compute the local value indicator 8.compute the ice indicator 9.synthesize the mask indicator 10.compute the final SST 11.check quality 12.determine the confidence level
mask control principles mask indicator initialized from the MAIA primary cloud mask set of tests –gradient –aerosol –local T –ice test method : check a test value against a limit and a critical value test indicator = 100 * ( value – limit ) / ( critical – limit ) test result →no problempotential problem critical problem unavailable test indicator0] 0,100 [100missing mask indicator100 SSTmissing test indicator value critical limit
method –calculate local T11 gradients –threshold vs front intensities atlas (URI + DMI) –applied only by night (for the moment) atlas intensities → local gradient values –atlas : gradient of SST per 5km –conversion in T11 gradient K/km : max T11 gradient = max SST gradient * transmittance / scale factor gradient indicator –limit value : max T11 gradient + max. radiometric noise –critical value : max T11 gradient + f ( distance to cloud ) gradient indicator distance to cloud ( pixels) f ( K/km) 15 noise max
method –1st estimate of local SST vs limit and critical values deduced from climatology steps –remapping upwelling atlas –compute the distance to cloud (2 nde ) –first estimate of SST –determine the indicator local SST value indicator limit = climsst_mini + klim * climsst_stddev with klim = 1.5 critical = climsst_mini + kcrit * climsst_stddev + upwelling_delta with kcrit = f( distance to cloud) upwelling_delta = f ( distance to coast ) local SST value indicator distance to cloud ( pixels) kcrit upwelling delta (K) distance to coast (km)
mask indicator synthesis weighted mean : if a test is missing it is set to 100 all weight are set to 1 cloudmask quality gradientdust aerosol strato aerosol local SSTvalue ice night day ×
SST calculation steps –determination of the algorithm to apply –smoothing of the atmospheric correction terms –determination of the aerosol correction NL : daytime or nighttime if T 37 is missing SST = a.T 11 + ( b.S θ + c.T cli )(T 11 – T 12 ) + d + e.S θ where S θ = sec(θ) – 1 and T cli the mean climatological SST T37_1 : nighttime SST = ( a + b.S θ )T 37 + ( c + d.S θ )(T 11 -T 12 ) + e.S θ + f twilight conditions weighted mean of daytime and nighttime algorithms capacity of using aerosol robust algorithms
SDI correction depends on the SST algorithm applied when 0 < SDI indicator < 100 correction = A 0 + A 1 sdi + A 2 sdi² in twilight conditions the correction is a weighted mean of daytime and nighttime ones.
determination of the confidence level use –to give the user a simple mean of filtering –to partition the MDB for deriving SSES GHRSST-PP project compliant representation of the risk factors in 2D mask indicator satellite zenith angle 50° 20 60° 70° = excellent 4 = acceptable 3 = suspect 2 = bad 1 = cloudy 0 = unprocessed
remapping gridded products –GLB cylndrical 0.05° resolution –NAR polar stereographic 2-km resolution remapping method –progressive building –average of the best confidence level candidates for a target pixel –if target pixel has been previously populated, selection according to : best confidence level illumination condition ( nighttime > daytime ) lowest satellite zenithal angle GLB : 12-hour synthesis centred at 00h00 and 12h00 UTC NAR : twice a day (METOP02 10h00 and 20h00 UTC)
quicklooks
Match up data base at day + 5, collect on the GTS measurements from ship and buoys an in-situ measurement is associated with a single pixel/metagranule validation box ( 21 pixels x 21 pixels ) criteria : 1.delta time ( ≤ 3h ) 2.distance ( ≤ 10 km ) 3.coverage of clear pixels ( ≥ 10 % ) two MDB levels : –netCDF : all satellite information –XML : statistics + histogram a 3rd ASCII level can be derived via XSLT for validation processing METOP SST validation (Anne Marsouin) on Tuesday in the SSES and validation WG
Implementation software –languages : C, Perl –XML –netCDF –ferret –gnuplot –grib api (ECMWF) operational (concurrent with another chain) –4 dual core AMD Opteron processor 885 –16 GB RAM development / test –2 dual core Intel Xeon 3.60GHz –8 GB RAM disk space : ~35 GB ( for a 1-hour latency)
Any question is welcome