ALTER: RAINFALL AND SUN-SPOT PERIODS. 29 



tain a perfectly homogeneous record, since volunteer observers must 

 sometimes fail, often through no fault of their own, but I would ven- 

 ture to suggest a method by which the records may be reduced to a 

 near homogeneity. The sum of the actual rainfall for all the stations 

 used may be divided by the sum of the normals of the several sta- 

 tions and the quotient published as the percentage of normal which 

 fell that month. The means of the normals of stations chosen for 

 accuracy of records and geographical distribution may then well be 

 taken as the normal of the strate, and when multiplied by this quo- 

 tient will give a weighted mean of the state that will be practically 

 homogeneous from year to year. This lack of homogeneity in state 

 records is much more serious in investigation of long periodicities 

 such as the Bruckner and eleven-year cycles, and might easily show 

 entirely negative results where the period actually exists. An ex- 

 ample of the reduction of scattered material to homogeneity is given 

 in this paper in the treatment of Chile, where long records are 

 available from five towns with widely differing normals. These 

 records begin in different years and omit certain years irregularly. 

 The sums of the actual rainfall given were tabulated for the fifteen- 

 month periodicity, as were also the sums of the normals for each 

 month that a station was used. These sums were then added 

 through each half of the data for each phase, and the quotient of 

 actual by normal was taken. These tables are Nos. 19 and 20. In 

 the eastern part of the United States the normals from one part of 

 a state to another vary by small enough amounts that the records 

 are not seriously impaired. For the western part I felt it best to 

 take instead the stations on the coast having perfect records ex- 

 tending as far back as 1880. All such were used except where sta- 

 tions in California happened to be very close together, in which 

 cases one was always omitted in order not to give that small section 

 of the coast undue weight. Nineteen such stations in California and 

 western Oregon were available. No station in Washington had such 

 a long record without break. This procedure also has the advantage 

 of almost doubling the length of record over the published state 

 averages. The results from these stations are shown as tables 10 to 

 12. The names of the stations will be found at the heads of these 

 tables. The Adelaide Observatory in South Australia seems to have 

 kept the most ideal record from 1861 to 1907. They averaged the 

 same fifty towns, apparently, from the beginning to the end of that 

 period. Unfortunately, this method was discontinued and the 

 present one of averaging all available stations, as, in the United 



