440 SMITHSONIAN MISCELLANEOUS COLLECTIONS VOL. 90 



The positions selected are shown in the tables which follow. Isotherms 

 were not drawn on the charts, and it was deemed inadvisable to attempt 

 readings of temperature. The isobars on the Japanese maps were in milli- 

 meters, those on the British maps were in millibars, and those on the United 

 States maps were in inches. This gave a heterogeneous set of units, and 

 in order to reduce them to something like unity the mean values were re- 

 duced to millimeters when they were in other measures. The reduction was 

 made because most of the readings were already in millimeters and less 

 changes in the observed values were needed. 



In regard to the accuracy of the readings, three classes of errors are 

 possible : 



(1) Errors of observation, including instrumental errors. 



(2) Errors in interpolation between the isobars for readings at the 

 intersection between lines of latitude and longitude. 



(3) Errors arising from the drawing of the isobars. 



It was not possible to determine the probable errors in observation, but 

 because comparison of ship barometers with standards are frequently made 

 and the observers are practiced in reading the instruments on shipboard, 

 the errors are assumed to average out in the monthly means. 



In regard to the accuracy of the means derived from interpolation between 

 the isobars, a number of tests were made by different persons and it was 

 decided that errors from this cause need not exceed o.i mm in the means 

 for a month. The errors which might arise from the drawing of isobars 

 by different men in different units and with differing amounts of observa- 

 tional material seemed more serious. In order to test the magnitude of the 

 probable error arising from this cause Mr. Namias took two sets of charts 

 for the same months and for the same areas: (i) the North Atlantic 

 charts, on which the isobars were drawn by Mr. Young of the United 

 States Weather Bureau in tenths of inches and (2) the British Northern 

 Hemisphere charts, on which the isobars were drawn for intervals of 

 4 millibars. 



The point selected for reading was Lat. 50° N., Long. 30 W., situated 

 in the middle of the Atlantic. One hundred and eight dates were used and 

 from the differences between the two sets of readings the probable error 

 of the means was determined to be 0.4 mm. From these tests it was decided 

 that the monthly means could be trusted to be correct within about 0.5 mm., 

 and it was decided to publish them in whole millimeters although the means 

 were computed to tenths of a millimeter. 



On the Pacific the interpolations for the years 1923 to 1930, inclusive, 

 were made from the Japanese daily maps of the North Pacific (Kobe Im- 

 perial Observatory). These maps were constructed from the ship reports 



