lati 
inte 
atio 
or SI 
tudi 
and 
The 
trun 
to 1 
      
  
    
  
+ 
fw. 
2-7] 
^ 
- f 
i et 
an 
ran 
sets 
deri 
The : 
(Sal 
shal 
of i 
fiel 
equa 
the 
carr 
mini 
must 
With 
Houg 
incl 
x time 
Houg 
harm 
Fig. 7? As for Fig. 4, but for observations over the time i 
4, bu s nterval [-2.0, 3.0], extended by a cosine er to 
Zéro on the initial and final days. The correct synoptic evolution ie retrieved over the Mise effe 
observation interval. 
nece. 
  
  
  
  
  
  
  
  
  
  
  
  
  
Ackn 
5. Conclusions 
Nati 
Alias-free synoptic and asynoptic sequences have been shown to be Foun 
uniquely related via their common space-time spectra. The process of 
synoptic retrieval begins with an "irregular" set of asynoptic data. From Refe 
this, equispaced spectra are constructed, with additional contamination due 
to irregular sampling eliminated. Once the latter are derived, the correct, Bath 
"regular" synoptic sequence follows by application of the inverse space-time 
transform. Twice-daily, synoptic sequences retrieved in this fashion, 
contain "exactly" the correct spectral components resolvable in both types 
of data. In addition the information content of the combined asynoptic data 
is fully utilized. 
Chapi 
Harti 
From a practical perspective, measurement error and data gaps may 
both be handled conventionally. The constraints, i.e. the observations, may 
be relaxed in accordance with their relative certainty, viz. a least squares 
procedure which incorporates estimates of instrument error. Likewise, 
information can be projected into data voids via say bivariate interpolation 
in the asynoptic coordinates s and r (see Salby, 1982c). Although the 
evolution retrieved in gaps will depend upon a number of factors, it follows 
directly, as before, that the true synoptic evolution is recovered in regions 
where observations exist. 
Longı 
152 
| dn o0 
EC A