This　paper　deals　with　the　dynamic　identification　and　finite　element　model　updating　of　Magaz　bridge,　located　in　Spain.　It　is　a　three　span　continuous　bridge　with　composite　steel-concrete　cross　section.　Operational　modal　analysis　was　performed　from　ambient　vibration　data　by　the　covariance　driven　stochastic　subspace　identification　algorithm.　A　very　accurate　three-dimensional　finite　element　model　was　built　using　SAP2000　assuming　geometrical　and　mechanical　properties　from　the　original　drawings.　At　the　first　step　of　the　calibration　process,　a　parametric　study　was　performed　to　identify　the　most　sensitive　parameters　affecting　the　finite　element　model-computed　modal　frequencies　and　mode　shapes.　The　automated　finite　element　model　updating　problem　has　then　been　formulated　as　an　optimization　problem.　The　uncertainty　on　some　physical　parameters　has　been　adjusted　through　an　iterative　process　that　aims　at　minimizing　the　objective　function,　hence　to　solve　a　nonlinear　least　squares　problem　which　has　as　its　subject　the　residuals　of　the　relative　difference　from　the　numerical　and　experimental　eigenfrequencies　and　eigenvectors.　The　outcome　of　the　optimization　is　a　significantly　improved　finite　element　model　with　very　good　agreement　in　terms　of　frequencies　and　modal　assurance　criteria　values　on　the　first　modes.　©Civil-Comp　Press,　2015.