The　quality　of　satellite　radar　altimetric　data　is　very　important　in　studies　of　geodesy,　geophysics,　and　oceanography.　Over　coastal　oceans,　altimeter　waveforms　are　contaminated　by　the　terrain　and　physical　environments　so　that　the　accuracy　of　altimeter　data　is　lower　than　that　over　open　oceans.　Here　we　develop　a　new　multi-subwaveform　parametric　retracker　(MSPR)　to　improve　the　quality　of　altimeter　data　for　the　recovery　of　gravity　anomaly　in　coastal　oceans.　The　least　squares　collocation　method　is　used　to　recover　the　residual　gravity　anomaly　over　the　coastal　water　from　altimetric　data.　The　waveform　data　records　from　Geosat/GM　around　Taiwan　Island　are　practically　retracked　with　MSPR.　When　compared　with　the　Taiwan　geoid　height,　the　results　retracked　by　MSPR　are　more　accurate　than　those　retracked　by　the　well-known　beta-5-parmeter　method　and　from　the　geophysical　data　records　(GDRs).　The　gravity　anomalies　over　Taiwan　coastal　waters　are　calculated　from　the　retracked　altimeter　data　with　the　least　squares　collocation.　When　we　compared　gravity　anomalies　computed　using　altimeter　GDRs　with　the　ship-borne　gravity　data　over　Taiwan　coastal　ocean,　we　found　that　the　results　from　retracked　data　are　more　accurate　than　those　from　GDRs.