6 
filter coeti dents for the moving average application are given only as a guide. For some 
.ngle proti.es of particularly bad quality, the processing may differ under certain circum 
stances. For instance, in the case of profiles disturbed by heavy noise, stronger median 
filtering (Sy, 1985) was carried out using q > 15 filter weights. Such doubtful profiles 
are indicated in the station lists. 
The final depth-interval-averaging with J\z = 1 m takes no account of the reduced 
statistical independence caused by the previous filtering. Compaction on 1 m steps 
should guarantee a uniform standard for archived data. Whatever the requirements of 
scientific analysis may be, the relatively high nominal vertical resolution of the final data 
allows investigators to adjust data according to their own needs. 
The final data are archived in the Deutsches Ozeanographisches Datenzentrum (DOD) 
which is operated by BSH. These data are available for personal use, in particular for 
WOCE purposes. However, any use of these data should be agreed with the originators 
prior to publication, but reservations apply to issues such as statistical analysis, heat con 
tent, wintertime convection, and North Atlantic Current which are the primary reason 
for operating TWI line AX-3. Data from TWI line AX-11 are not subject to reserva 
tions. 
The section plots were drawn by means of a new objective analysis method developed 
by Paul (1994). The model is generated by performing a least-square fit for each grid 
point with orthogonal polynomial surfaces using data surrounding each grid point in a 
defined area. The orthogonal polynomial surface analytical method has the properties 
of a wave number filter, so that the spatially coherent structures of the analysed field are 
identified below a selected wave number maximum. Because of the numerical precision 
and the high computational velocity, the method is suitable for processing large volumes 
of oceanographic data. A further advantage is that the field properties, e.g. the 
correlation or the "grand mean" over the analysed area need not already be known 
before the analysis starts as is the case for statistically objective analytical methods.