A new method to retrieve salinity profiles from sea surface salinity observed by SMOS satellite.pdfVIP

ActaOceano1．Sin．，2015，Vo1．34，No．9，P．85—93 DOI：10．1007／s13131—015—0735．3 http：／／www．hyxb．org．cn E—mail：hFxbe@263．net A newm ethodtoretrievesalinityprofilesfrom seasurface salinityobservedbySMOSsatellite YANGTingting，CHENZhong【biao，HEYijun SchoolofMarineSciences，NanjingUniversityofInformationScienceandTechnology，Nanjing210044，China Recewed23September2014；accepted28February2015 ~TheChineseSocietyofOceanographyandSpringer—VerlagBerlinHeidelberg2015 Abstract Thispaperproposesanewmethodtoretrievesalinityprofilesfromtheseasurfacesalinity(sss)observedbyhte SoilMoistureandOceanSaliniyt(SMOS)satellite．Themainverticalpatternsofthesaliniytprofilesarefirstly extractedfrom htesaliniytprofilesmeasuredbyArgousinghteempiricalorthogonalfunction．Todeterminethe timecoefficientsforeachverticalpattern，twostatisticalm odelsaredeveloped．Inthelinearmodel，atransfer functionisproposedtorelatetheSSSobservedbySMOS(SMOS—sss)withthatmeasuredbyArgo，andthena linearrelationshipbewt eentheSM OS— — SSSandthetim ecoefficientisestablished．Inthenonlinearmodel，the neuralnewt orkisutilizedtoestim atethetim ecoefficientsfrom SM OSSSS，monthsandpositionsofthesalinity — profiles．Thetwomodelsarevalidatedbycomparingthesaliniyt profilesretrievedfrom SMOSwiththose measuredbyArgoandtheclimatologicalsalinities．Theroot—mean—squareerror(RMSE)ofthelinearand nonlinearmodelareO．08—0．16and0．08-0．14forhteupper400m ，whichareO．01—0．07and0．01-0．09smallerthan hteRMSEofclimatology．Theerrorsourcesofhtemethodarealsodiscussed． Keywords：saliniytprofile，SoilMoistureandOceanSaliniyt(SMOS)data，Argodataseasurfacesaliniyt Citation：YangTingting，ChenZhongbiao，HeYijun．2015．Anewmethodtoretrievesaliniytprofilesfromseasurfacesalinityobservedby SMOSsatellite．ActaOceanologicaSinica，34(9)：85—93，doi：10．1007／s13131—015—0735—3 1Introduction ficantlyimprovethesalinity