aQo 
The Weather Bureau data are arranged on a monthly basis for “squares” 
bounded by five degrees of latitude and longitude. In some months there were 
several hundred observations in some individual squares, in others only a few, 
sometimes less than ten, These uneven frequencies plus the unreliability of 
many of the measurements made careful processing necessaryo! 
At first, means for each month were computed for each square and these 
means averaged over the whole period. The resulting mean monthly values were 
then plotted on graphso These yielded smcoth seasonal variations in almost 
all squarese Occasional irregular values could be traced to isolated extremely 
low or high temperature reports, or months with very few reports. They were 
considered incorrect and replaced by values on the curve. Figo 1 shows the 
mean monthly temperatures o* 
Secular Variation 
Before proceeding to monthly temperature anomalies it was necessary to 
determine the secular trend because of the large climatic variation known to 
have occurred in the Atlantic in recent decades. This was done by computing 
annual temperatures in each square and expressing these as deviations from the 
mean annual temperature for the whole pericd. Mean annual temperatures were 
calculated from the data of fige lo The result is shown in figs 2 which also 
indicates the periods for which observations were available in each squareo 
One observes relatively strong year to year variations in many squares, 
some of these quasi-periodic. Very large changes occurred in the region of 
1 Tt could not be ascertained to what extent, if any, changes in 
methods of measurement have affected the temperatures as recordede No 
allowance has been made for this factor 
é The Weather Bureau data extend beyond 50°N, but the polar part 
of the ocean was not studied in this projecto 
