General circulation models (GCMs) can be used to develop diagnostics for identifying weather regimes. The author has looked for three-dimensional (3D) weather regimes associated with a 10-yr run of the U.K. UGAMP GCM with perpetual January boundary conditions; 3D low-pass empirical orthogonal functions (EOFs), using both the 500- and 250-mb streamfunctions (psi) have been computed. These EOFs provide a low-order phase space in which weather regimes are studied. The technique here is an extension to 3D of that of Haines and Hannachi. They found, within the 500-mb psi EOF phase space, two local minima of area-averaged psi-tendency (based on barotropic vorticity dynamics), which were identified as +/-Pacific-North America (PNA). In this work, the author demands that both the flow and its tendency be within the phase space spanned by the 3D EOFs. The streamfunction tendency is computed from the two-level quasigeostrophic potential vorticity equation and projected onto the EOF phase space. This projection produces a finite dynamical system whose singular points are identified as the quasi-stationary states. Two blocking solutions and one zonal solution are found over the Pacific. The first blocking solution is closer to the west coast of North America than the other blocking, which is shifted slightly westward and has a larger scale, rather similar to the +PNA pattern, indicating that blocking over the Pacific may have two phases in the model. Further investigation of the GCM trajectory within the EOF phase space using a mixture analysis shows the existence of realistic three-dimensional weather regimes similar to the singular points. The same solutions were found when the transient eddy contributions to the climatological quasigeostrophic potential vorticity budget were included. It is also shown that this extended technique allows a direct study of the stability of these quasistationary states and helps in drawing transition pictures and determining the transition times between them.
