Dept.ofEarth,Atmospheric,andPlanetarySciences MassachusettsInstituteofTechnology Cambridge,MA02142

Tomographicimagingproblemsaretypicallyill-posedandoftenrequiretheuseofregularization techniquestoguaranteeastablesolution.Minimizationofaweightednormofmodellengthisone commonlyusedsecondaryconstraint.Tikhonovmethodsexploitlow-orderdi!erentialoperatorsto selectforsolutionsthataresmall,flat,orsmoothinoneormoredimensions.Thisclassofregularizing functionalsmaynotalwaysbeappropriate,particularlyincaseswheretheanomalybeingimagedis generatedbyanon-smoothspatialprocess.Timelapseimagingofflow-inducedseismicvelocityanomalies isonesuchcase;flowfeaturesareoftencharacterizedbyspatialcompactnessorconnectivity.Wedevelopa traveltimetomographyalgorithmwhichselectsforcompactsolutionsthroughapplicationofmodel-space iterativelyreweightedleastsquares.Ourtechniqueisanadaptationofminimumsupportregularization methodspreviouslydevelopedwithinthepotentialtheorycommunity.Weemphasizetheapplicationof compactnessconstraintstotimelapsedatasetsdi!erencedinthedatadomain,aprocesswhichallows recoveryofcompactperturbationsinmodelproperties.Wetestourinversionalgorithmonasimple syntheticdatasetgeneratedusingavelocitymodelwithseverallocalizedvelocityanomalies.Wethen demonstratethee"cacyofthealgorithmonaCO 2 sequestrationmonitoringdatasetacquiredattheFrio pilotsite.Inbothcases,theadditionofcompactnessconstraintsimprovesimagequalitybyreducing spatialsmearingduetolimitedangularapertureintheacquisitiongeometry.