DETERMINING THE OPTIMUM NUMBER OF STRATA FOR SAMPLING A NORMALLY DISTRIBUTED FOREST POPULATION

The purpose of the study was to determine the optimum number of strata (L) in a forest population when the underlying distribution is normal , and the variable for stratification is either the variable of intere st or a linearly related auxiliary variable. The technique used in bot h cases involved the combination of the Neyman allocation of strata sa mple sizes, Dalenius and Hodges' cumulative square root frequency meth od, and a simple cost function of the form suggested by Dalenius. Our results showed that for a normally distributed diameter at breast heig ht (dbh) and total height, eight strata, were about optimum. Gains in efficiency over the unstratified situation were 1500% for dbh and 2500 % for total height. These optimal values represented those values for which further increases in L did not produce appreciable decrease in e stimated variance of the mean of the variable of interest.