332 R- B- Montgomery 



The weather ships, occupying nearly fixed positions, are potential sources of data 

 of great climatological and oceanographic interest. Surface water at Weather Ship 

 J in the North Atlantic Ocean is chosen for the present example, because published 

 observations are available for three years in Bulletin Hydrographique pour Vannee 

 1948, 1949, 1950. The data are credited to the Fisheries Laboratory, Lowestoft, 

 Great Britain. Station J lies west of Ireland and south of Iceland. The usual position 

 up to March 1950 was in the 1 -degree quadrangle with southeast corner at 53° N 

 18° W and from April 1950 was in the quadrangle with southeast corner at 52° N 

 20° W. 



For most days there is one observation of temperature and sahnity, but for some 

 days there is none. In order to obtain a homogeneous series from the somewhat 

 irregular original series, 15 observations have been chosen for each of the 36 months. 

 If available, the odd days 1, 3, ... 29 have been used. Missing odd days have been 

 replaced with other days or with observations from nearby positions. January 1948 

 has the fewest observations, only 10, and has been completed by including 4 from 

 late December 1947 {Bulletin Hydrographique pour Vannee 1947) and 1 from early 

 February 1948. 



The homogeneous series of 540 observations serves as the basis for the statistics 

 presented on the temperature-salinity diagram in Fig. 1 . The frequency for each two- 

 dimensional class with temperature interval 0-5 C and salinity interval 0-1 per mille 

 is tabulated in per mille. 



By arranging the frequencies in order of decreasing magnitude and summing cumu- 

 latively, the result is found that about half (5 1 per cent) of the observations fall in 

 classes with frequency 22 per mille and greater; these classes are enclosed by a heavy 

 hne. About nine tenths (89 per cent) fall in classes with frequency 6 per mille and 

 greater; these are enclosed by a fine Une.* 



Summing the two-dimensional distribution horizontally and vertically respectively 

 gives the one-dimensional frequency distributions of sahnity and temperature, 

 shown by the histograms at the left and top of Fig. 1 . 



Three-year monthly means for Weather Ship J are shown as points in Fig. 1 . This 

 method of showing monthly means was used by Neumann (1940, Abb. 25). His 

 Feld 1 (Abb. 25, Nr. 6), the 5-degree quadrangle with northeast corner at 50° N 

 20° W, is represented by a circuit of similar shape (nearly isohaline), but with tem- 

 perature range 1-5 C greater, mean temperature 2C higher, and mean sahnity 0-4 

 per mille higher. 



Specific volume of surface water at Weather Ship J is shown by including isopleths 

 of specific-volume anomaly. This quantity (which for surface water equals ther- 

 mosteric anomaly) is chosen in preference to density or sigma-/ for the reasons 

 presented by Montgomery and Wooster (1954). Any other temperature-salinity 



* Two items of terminology are needed for two-dimensional frequency distributions. First, for a 

 frequency isopleth of stated frequency a handier term than Bilham's " constant-frequency graph " 

 (Brooks and Carruthers, 1953) is desirable. Secondly, an even more useful term would represent 

 a frequency isopleth that encloses a stated proportion of the occurrences. If the first isopleth were 

 called alpha and the second beta, the 00055-alpha would separate 5-per-mille classes from 6-per-mille 

 classes, while the 0-9-beta would enclose nine-tenths of the occurrences; in Fig. 1, these two isopleths 

 coincide. The betas have the advantage of being approximately independent of class intervals, for 

 the alphas expand as the class intervals increase. It may be noted that beta is not similar to the term 

 quantile (Brooks and Carruthers, 1953) used with regard to one-dimensional frequency distribu- 

 tions, because the maximum frequency coincides with 0-beta but with an intermediate quantile. 



