New York Weathek Bureau. 515 



almiost uniformly too high, in a large number of case® exceedimg 

 the values for February. Many of the defects are, mo doubt, due 

 to the difficulty of making the first observation precisely at isun- 

 rise^ and to-; the variable hour of the last observation, one hour 

 after sunset.* 



In the seioond period of the Regents' system of (observations, 

 between 1850, and 1868, fixed hours were employed to advantage, 

 and several valuable records 'have been obtained from this series. 



It wasi finally decided to base the system of normals upon recent 

 observations, in which istandard instruments were employed under 

 knoiwnii oonditiicins! of exposure. Cointinuous records lat the sta- 

 tions, Central Park, 'Ooiop erst own and Rochester, each covering 

 the period 1871-1891, were used as the standards of the system. 

 All shorter records were compared with these, month by month, 

 and the relation between the temperature conditionis of the 

 standard and seoondary stations thus determined. The normals 

 in the table are theui to be considered as applying to, the twenty- 

 one year period, even for observations made prior to 1870, as it 

 \wais (still possible to compare these with the record of Coopers- 

 town. 



(Several records of great merit are not oredited' in the /table of 

 normals with the full periods over which their observations ex- 

 tend, isince the continuity of the 'series was broken by important 

 changes in the location of thermometers, as became apparent at 

 once upon comparison of successive monthly and annual values 

 with the 'records of adjacent stations. In nearly all cases of this 

 nature the means given in the table were derived from the later 

 portions of the records. 



The results of table 31 show the mean of a twenty-one years' 

 period to be liable to a maximum deviation of 0.8 degrees from the 

 true normal. The record of Cooperstown might be considered to 

 afford a basis for the reduction of the whole system to a thirty- 

 eight year period, with a maximum error of 0.6 degrees; but it 

 was deemed best not to attempt this upon the authority of a single 

 record. Tlie twenty-one and thirty-eight years' averages for 



• In many cases, irregularities in the tenaperature records appear to be due to a change of 

 observers. 



