ScientistsinventnewalgorithmtoimproveaerosolmonitoringonChina'sFY-4Asatellite_GuangmingOnlineScientistsinventnewalgorithmtoimproveaerosolmonitoringonChina'sFY-4AsatelliteBEIJING,Feb.12(Xinhua)--ScientistshaverecentlyintroducedanewalgorithmthatcombinesdeeplearningandtransferlearningtoimproveaerosolmonitoringonChina'sFY-4Asatellite.Thestudy,publishedinthejournalEngineering,wasconductedthroughacollaborativeeffortbytheInstituteofAtmosphericPhysics(IAP)oftheChineseAcademyofSciences,NationalSatelliteMeteorologicalCenter,theHarbinInstituteofTechnology,andotherinstitutes.ScientistsbelievethataccuratemeasurementsofatmosphericaerosolsarepivotalinunderstandingEarth'sradiationbalance,climatechange,andairquality.AboardChina'sgeostationarymeteorologicalsatelliteFY-4A,theAdvancedGeostationaryRadiationImager(AGRI)scansChinaeveryfiveminutes,providingcrucialdataformonitoringaerosolspatiotemporalvariations.However,theinflexibilityoftraditionalphysicalretrievalsalgorithms,coupledwiththeinsufficientnumberofground-basedsunphotometersites,poseschallengesinmeetingtheextensivesamplerequirementsformachinelearninginaerosolopticaldepth(AOD)retrievals.0000000000Inresponsetothesechallenges,thescientistsdevelopedaninnovativeAODretrievalsalgorithmthatcombinesdeeplearningandtransferlearning.Thenewalgorithmincorporateskeyconceptsfromthedarktargetanddeepbluealgorithmstofacilitatefeatureselectionformachinelearning.Accordingtothestudy,independentvalidationconfirmsthatthealgorithmishighlyaccurateinestimatingAGRIaerosollevels.Theresultsshowastrongcorrelationwithexpectedvalues,indicatingthealgorithm'sreliabilityinpredictingaerosolopticaldepth."Ourstudyshowcasesthesignificantpotentialofmergingthephysicalapproachwithdeeplearningingeoscientificanalysis,"saidtheleadauthorFuDisongfromtheIAP."Theproposedalgorithmholdspromiseforapplicationtoothermulti-spectralsensorsaboardgeostationarysatellites,"Fuadded.Editor:DisclaimerTheviewsandopinionsexpressedinthisarticlearethoseoftheauthor's,GMW.cnmakesnorepresentationsastoaccuracy,suitability,orvalidityofanyinformationonthissiteandwillnotbeliableforanyerrors,omissions,ordelaysinthisinformation.