Despite　their　crucial　importance　for　marine　engineering,　the　nonlinear　seismic　response　characteristics　of　offshore　sites　remain　poorly　understood.　Consequently,　simulating　ground-motion　at　offshore　sites　poses　a　significant　challenge.　To　address　this,　this　study　begins　with　a　dataset　comprising　stress　drops　of　70　earthquakes,　region-specific　quality　factors,　and　linear　site　amplification　factors　(AFs)　of　six　offshore　stations　in　Sagami　Bay,　Japan,　obtained　using　the　generalized　inversion　technique　(GIT).　Then,　by　incorporating　additional　offshore　accelerograms　with　focal　depths　up　to　333　km　and　peak　ground　accelerations　(PGAs)　ranging　from　0.2　to　4.2　m/　s2,　we　delve　deeper　into　the　effects　of　nonlinear　site　behaviors　on　the　high-frequency　attenuation　parameter　(kappa　0)　and　AFs,　respectively.　A　counterintuitive　decrease　in　kappa　0　was　observed　as　the　peak　ground　acceleration　(PGA)　reached　0.5-0.8　m/s2,　echoing　similar　observations　from　previous　studies　on　KiK-net　stations.　Our　results　indicate　that　the　high-frequency　attenuation　characteristics　of　offshore　sites　vary　under　strong　motions,　potentially　attributable　to　the　nonlinear　evolution　of　the　frequency-independent　quality　factor　and　S-wave　velocity　within　near-surface　sediments.　Additionally,　the　degree　of　nonlinearity　(DNL)　at　these　offshore　stations　exceeded　4　when　PGA　reached　0.2-0.3　m/s2,　a　threshold　significantly　lower　than　the　previously　reported　range　of　0.5-1.0　m/s2.　Furthermore,　we　observed　systematic　variations　in　nonlinear　behaviors　between　flat　and　steep　offshore　stations,　particularly　with　peak　frequencies　shifting　towards　lower　and　higher　frequencies,　respectively.　These　new　findings　may　be　mainly　attributed　to　the　intricate　interaction　of　topography　and　marine　sediments.　Finally,　simulations　of　two　subduction　earthquakes　(MW6.2　and　5.9)　using　the　stochastic　finite-fault　simulation　method　(SFFSM)　showed　good　agreement　with　observations　at　frequencies　above　0.1　Hz.　Notably,　nonlinear　AFs　outperformed　linear　ones　across　a　wide　PGA　range　of　0.2-1.2　m/s2,　highlighting　the　significance　of　nonlinear　site　behaviors　in　characterizing　offshore　ground-motions.　This　finding　reinforces　the　potential　of　the　simulation　framework　(integrating　GIT　and　SFFSM)　for　effectively　and　accurately　simulating　offshore　ground-motion.