An extended empirical orthogonal function analysis technique is described which expands a data set in terms of functions which are the “best” representation of that data set for a sequence of time points. The method takes advantage of the fact that geophysical fields are often significantly correlated in both space and time. Two examples of applications of this technique are given which suggest it may be a highly useful tool for diagnosing the modes of variation of dominant sequences of events. In the first, an analysis of 300 mb relative vorticity, fairly regular advection of the major features of the spatial patterns is evident. Westward speeds of between 0.3 and 0.4 m s−1 are inferred. The second example illustrates extended functions of tropical Pacific Ocean surface temperatures. The dominant function, which is associated with El Niño, shows a high degree of persistence over a six-month sequence. The second most important function suggests opposing variations in the influences of the North a... Abstract An extended empirical orthogonal function analysis technique is described which expands a data set in terms of functions which are the “best” representation of that data set for a sequence of time points. The method takes advantage of the fact that geophysical fields are often significantly correlated in both space and time. Two examples of applications of this technique are given which suggest it may be a highly useful tool for diagnosing the modes of variation of dominant sequences of events. In the first, an analysis of 300 mb relative vorticity, fairly regular advection of the major features of the spatial patterns is evident. Westward speeds of between 0.3 and 0.4 m s−1 are inferred. The second example illustrates extended functions of tropical Pacific Ocean surface temperatures. The dominant function, which is associated with El Niño, shows a high degree of persistence over a six-month sequence. The second most important function suggests opposing variations in the influences of the North a...