REPORT ON ATMOSPHERIC CIRCULATION. 43 



Correction for Mean Temperature. — The period selected for the mean temperature 

 observations is the fifteen years adopted for pressure, beginning with 1870 and ending 

 with 1884. From the remarkable extension of meteorological observation in recent 

 years, data of greater fulness and of higher quality are now available for drawing 

 isothermals over the globe, which therefore represent the geographical distribution of 

 temperature with a degree of approximation to the truth not previously attainable. 



The methods of discussing the observations are, to a large extent, the same as 

 those detailed and explained in dealing with the observations of atmospheric pressure, 

 with, however, several important differences. 



Since the observations made use of preferentially in this inquiry are the daily 

 maximum and minimum temperatures, special attention was given in making the 

 extracts of the monthly means to detect, where possible, any cases that may have 

 occurred of the minimum thermometer having got out of order, as not unfrequently 

 happens, and allowed, from inadvertence, to remain out of order for some time. These 

 errors, together with typographical errors and many of the errors of computation, were 

 the more readily detected by the practice adopted of extracting the means of the 

 separate years in succession for each country or region by itself, so that the curve of 

 monthly variation of each year being easily kept in mind, any deviation from it was 

 seen with little difficulty. 



"When observations are read to the tenth of a degree, the personal errors of 

 observation may be neglected. But when the readings are only to whole degrees, two 

 kinds' of errors are certain to occur where provision is not made to secure that each 

 observer is properly taught. These two sorts of error are, (l) taking the degree which 

 the mercury or spirit has j ust passed ; or (2) taking the degree immediately above the top 

 of the mercury or spirit. In the former case, the means deduced from the observations 

 will be half a degree too low, and, in the latter case, half a degree too high. In many 

 cases these faulty methods of observing may be detected from the annual means, 

 corrected for height, entered in maps of the country whose temperature is being 

 discussed. 



By the same method the errors of faulty thermometers may be detected. In all 

 cases where for this assumed cause the means have been corrected to the extent 

 of 1° or upwards, the amount of the correction is stated in the last column of the 

 table under " Corrections applied." In such cases as Portland in Victoria, Australia, 

 where the published mean temperatures were for many years about 5° too high, but 

 where the error was rectified some time ago, the correction was applied to the 

 observation of the years in error, but no note is made of it in the last column of the 

 table. 



Again, in cases where " mean temperatures " alone are published, and no informa- 

 tion given whence these have been derived, a change of hours sometimes takes place 



