On the Use of Double-Mass Analysis for Testing the Consistency of Meteorological Records and for Making Required Adjustments

Abstract

SHORTER CONTRIBUTIONS / "T A HE problem of correlating meteorological JL elements is basic to many research studies. In such studies it is important that the records used be homogeneous, that is, comparable throughout the period of observations. Double-mass analysis, which deals with accumulated values rather than with incremental data, provides a means of determining the consistency of observations collected over a long period of time. It also provides a means of adjusting the early records to conform to those being collected at the existing site of a station which has been moved. This method of analysis, first described by Merriam, 1 is particularly applicable to precipitation data. The general equation of a linear regression expressing the relation between two variables X and Y takes the form Y = mX + b, where m and b are constants. For some meteorological elements, such as temperature, the constant m is nearly unity. In other words, the mean monthly temperature at one site tends to be a constant number of degrees (b) higher or lower than that at a nearby site. In the case of precipitation, however, the amounts observed at two adjacent sites tend to be proportional (i.e., b = 0). If Y = mX, then accumulated values of X and Y plot as a straight line. When comparing the monthly, seasonal, or annual precipitation as observed at two nearby sites, the function is not exact and, consequently, the broken line connecting adjacent plotted points can normally be expected to weave back and forth across the mean straight line in a random manner. If, however, one of the stations has been moved during the period of record, the constant of proportionality m may have changed. In this event, the points of the double-mass plotting tend to lie along two line segments of different slope (FIG.