Conventional　soil　models　with　single　spring　and　damping　components　can　not　properly　simulate　the　frequency-dependent　behavior　of　foundation　soils.　In　this　study,　the　foundation　soil　is　modeled　by　a　two-degrees-of-freedom　mechanical　model　(2DOF)　with　eight　integrated　constant　parameters.　The　2DOF　model　provides　a　reasonable　basis　for　the　nonlinear　dynamic　analysis　of　soil-structure　interaction　(SSI)　system　in　the　time　domain.　To　determine　the　values　of　the　integrated　parameters　in　the　2DOF　model,　a　system　identification　technique　using　the　extended　Kalman　filter　(EKF)　is　developed　in　this　study.　To　improve　the　computational　accuracy　and　simplicity,　the　first　order　polynomial　approximation　is　applied　in　the　EKF　procedure.　A　realistic　identification　example　is　given　for　two　soil　sites　in　Anchorage,　Alaska.　The　recorded　seismic　data　from　two　relatively　long-period　earthquakes,　the　Nanana　earthquake　(23rd　October　2002)　and　the　Denali　earthquake　(3rd　November　2002),　are　used　for　the　identification　example.　Results　of　the　example　demonstrate　the　feasibility　and　practicality　of　the　proposed　identification　technique.　Based　on　the　identified　integrated　parameters,　the　site　impedance　functions　and　predominant　frequencies　can　be　estimated.　The　2DOF　soil　model　and　the　identification　technique　can　be　used　in　time-domain　for　nonlinear　analysis　of　soil-structure　interaction　in　various　complex　soil　conditions.　The　identified　parameters　can　be　stored　in　a　database　for　other　similar　soil　conditions.　Developing　a　broad　database　may　provide　an　useful　insight　for　investigating　effects　of　SSI.