Fig. 2 ” Synoptic retrieval from asynoptic nadir observations of a westward traveling wavenumber 3 mode. Recovered F 
values are exact to machine precision. (Maps at half day increments not shown.) Contour interval 0.2. 
For the purpose of demonstration, the same field has been retrieved 
without elimination of the additional aliasing due to the irregular spacing 
of combined data (Salby, 1982c). When this self-contamination is allowed to ve 
reach the synoptic products (Fig. 3) the recovered structure and evolution or! 
are rather different. Although the behavior at low latitudes is retrieved 
with reasonable fidelity, the fields at middle and high latitudes are 
radically discrepant with the true behavior. This disparity is due to the 
increasing contamination at middle and high latitudes (c.f., Salby, 1982c, 
Fig. 7). For fields whose principal variance lies outside the tropics, this 
ES can result in a complete aberration of the true behavior (see Salby, 
1982b). 
The use of spectra in evaluating field properties is predicated 
upon an assumption of "weak stationarity," i.e. spectral components do not 
change with time. Clearly this cannot be guaranteed in the context of un- 
controlled observation. Indeed, it is often situations of transience which 
are of principal interest. In the absence of stationarity, alternative 
interpretations of spectra exist (Bâth, 1974). Although founded upon spectral 
considerations, FFSM is completely detached from the question of stationarity. for 
It relies solely upon the uniqueness between synoptic and asynoptic obser- tur! 
vations, as established in the Synoptic Retrieval Theorem. Thus although lat 
caution should be exercised in interpreting the intermediate spectra, a bey 
nonstationary evolution can be retrieved equally well. The only requirement unri 
necessary is that the limitations in sampling not be violated. Sall 
con 
con 
inti 
14$ 
P Mme um 
se IE MR 
J 
N sr