A　gas　kick　may　occur　when　drilling　with　narrow　pressure　margin,　which　can　lead　to　significant　non-productive　time.　This　paper　proposes　an　early　gas　kick　detection　method　applicable　to　water-based　mud　(WBM),　which　integrates　a　transient　pressure　and　temperature　coupling　model　into　an　unscented　Kalman　filter　(UKF)　algorithm.　Three　pressure　factors　and　a　flow　rate　factor　are　estimated　with　the　UKF　algorithm　using　updated　measurement　data　in　a　detection　estimator,　and　a　generalized　likelihood　ratio　test　(GLRT)　is　employed　to　automatically　detect　changes　in　the　pressure　factors　and　flow　rate　factor　for　prediction　of　the　gas　kick.　Simulated　results　show　that　the　estimated　pressures　and　outlet　flow　rate　trace　synthetic　measurements　very　well　with　continuous　inversion　of　the　pressure　factors　and　flow　rate　factor.　All　of　the　estimated　pressure　factors　and　flow　rate　factor　fluctuate　to　a　little　extent　around　a　base　value　under　normal　drilling　condition,　while　the　pressure　factor　in　annulus　and　flow　rate　factor　both　deviate　from　the　base　value　when　the　gas　kick　occurs.　Performances　of　the　proposed　method,　pit　gain　monitoring　method　and　delta　flow　monitoring　method　are　contrasted.　The　kick　detection　times　are　gradually　reduced　for　all　the　methods　with　the　increase　of　gas　kick　rate.　Moreover,　the　proposed　method　herein　shows　a　much　better　performance　than　the　pit　gain　and　delta　flow　monitoring　methods,　especially　at　small　gas　kick　rate.