methods of estimating wind speeds must have 

 resulted in considerable differences among 

 individual observers. Excellent results have 

 been obtained, however, in oceanographic com- 

 putations using ships' wind data by assuming 

 that the observed values for a given Beaufort 

 force form a Gaussian distribution about its 

 midpoint and have a standard deviation of half 

 a Beaufort interval {Reid 1948). 



Apart from estimating the reliability of 

 single observations on wind force is the prob- 

 lenn of setting the mininnum number of obser- 

 vations needed to give a reliable estinnate of 

 the wind forces over a given 5-degree square 

 of ocean during a given month; that is, assum- 

 ing observations are randomly distributed in 

 time, how many are needed to furnish a reliable 

 description of the winds in a 5-degree block of 

 ocean during one nnonth? Since the estimated 

 standard deviation of a single observation is 

 one-half a Beaufort force class, it was arbi- 

 trarily decided that the data from a 5-degree 

 square for a particular month would not be 

 considered unless there were enough observa - 

 tions to furnish a mean having a high probabil- 

 ity of lying within 0. 5 of a Beaufort force of the 

 true mean. Following Snedecor (1946, p. 457) 

 this minimum sample size was estimated by 

 computing the standard deviation and standard 

 error of the mean of a square from midocean 

 represented by 427 observations. From these 

 statistics n was computed for the 99-percent 

 and 95-percent confidence levels as 37 and 27 

 respectively. In constructing the monthly 

 charts all squares having less than 27 observa- 

 tions were omitted and the contours of the areas 

 having between 27 and 37 observations were 

 drawn with dotted lines. 



PROCEDURES 



Since marine weather reports and fore- 

 casts are now made in knots rather than in 

 Beaufort scale, knots were used in the charts 

 in this atlas, which show for each nnonth the 

 percentage of observed winds of 20, 25, 30, 

 and 35 knots or less. Twenty knots was selected 

 as the lowest speed for contouring, since at this 

 wind speed any seagoing fishing vessel should 

 still be able to carry out all types of fishing 

 operations. It is slightly greater than the mid- 

 point of Beaufort force 7 (17-21 knots) which is 

 defined as the speed at which "smacks shorten 

 sail" (Bowditch 194 3, p. 52) and at which the 

 U. S. Coast Guard begins to display small-craft 

 warnings. Thirty-five knots was selected as 

 the maxinnum wind speed for fishing operations 

 because it lies in the lower limits of Beaufort 

 force 8 (34-40 knots), which is defined as the 



speed at which "smacks remain in harbor, and 

 those at sea lie to." The two intermediate 

 velocities have been included as a basis for 

 estimating the percentage of time larger ves- 

 sels should be able to carry out limited opera- 

 tions such as trolling, or that smaller vessels 

 must "lie to, " 



The experience of POFI vessels in the 

 North Pacific has shown the validity of these 

 limits. At wind speeds of 20 knots, gill nets 

 begin to tangle and are difficult to retrieve, and 

 longlining becomes extremely difficult. The 

 records show that longline has been laid in 

 winds up to 25 knots but only when the 30-hour 

 weather forecast indicated that the weather 

 would moderate, and there were other signs of 

 improving weather. Both gill nets and longline s 

 have been retrieved in winds up to 35 knots 

 when it was a matter of saving the gear, but the 

 operation was exceedingly difficult and hazard- 

 ous. Thirty knots represents the upper limit 

 of trolling; above this speed the lines tangle and 

 do not fish properly. It has been POFI's expe- 

 rience that, above 35 knots, her research ves- 

 sels must suspend fishing activities and either 

 lie to or jog into or before the sea, depending 

 on their particular characteristics. 



The percentages used in preparing the con- 

 tour charts were obtained from cumulative per- 

 centage distribution curves of the wind speeds . 

 Monthly cunnulative distribution curves for each 

 square were constructed by first plotting the 

 cumulative percentages for each Beaufort force 

 at the speed in knots equivalent to the extreme 

 upper limit; e.g., 16.5 knots for force 4. These 

 points were then used as a basis for drawing 

 smoothed ogive curves (Guilford 1950, pp. 121- 

 126). Actually the smoothed curves passed 

 through the points in the majority of cases. The 

 percentiles for the 20-, 25-, 30-, and 35-knot 

 levels were then taken fronn these curves. 



The percentages were plotted at the mid- 

 point of each square and the contours were 

 drawn assuming that they represented the fre- 

 quency at this point. After the values were 

 plotted, examination showed that some of the 

 squares off North America and Japan, which 

 were largely land, e.g., the square bounded by 

 50°-55°N., 125'-130°W., and the confined 

 waters between Japan and Asia had a much 

 larger percentage of winds of lower speeds than 

 the adjacent areas offshore. It was believed 

 that this apparent increase in percentage of 

 winds of lower speeds was due to vessels taking 

 advantage of shelter when operating locally or 

 awaiting good weather before departing from 

 port. Data for these squares were therefore 



