Estimation of Volume in Timber Stands by Strip Sampling

Open Access
Publisher Website
Google Scholar
Abstract
Several methods of sampling timber stands and statistical treatments of the samples are considered. Data from a complete inventory of 4,402 acres of the Blacks Mountain Exptl. Forest in N. California serve for testing the methods. The usual method of estimating from strip samples taken within non-rectangular blocks of timber gives biased estimates, unless the linear regression of vol. on strip length passes through the origin of coordinates. This condition is not a safe one to assume, so methods are sought that are free from this restriction. Appropriate formulas for the best linear unbiased estimates are deduced under various combinations of the following assumptions: (1) That the regression of timber vol. on strip length is strictly linear, but may or may not pass through the origin of coordinates; (2) that the values of the (S.D.)2 of timber vol. on strips of equal lengths are (a) constant for different strip lengths, (b) proportional to strip length, or (c) proportional to the square root of strip length; (3) that the number of strips of a given length in each block is (a) finite, or (b) infinite. Assumption (b) of (2) gives slightly better results than either (a) or (c). An extensive sampling expt. is made to test whether the smallness of the samples combined with the conflicts between assumptions of the theory and the actual facts influences the validity of the normal theory. When the sample did not exhaust strips of a given length, formulas based on the assumptions that the populations of such strips are finite and that they are infinite both work satisfactorily. Confidence intervals based on the former assumption are narrower, however. Where the sample exhausts strips of a given length, treatment of the number of such strips as finite results in marked disagreement between the actual distribs. of statistics and those based on normal theory. This disagreement does not exist in statistics based on an infinity of strips of a given length. The formulas can be applied as well to line plots as to strips. The most efficient sampling will be obtained when the av. sample-strip length is close to the av. for the population, but where the variation among sample strips is the maximum.

This publication has 1 reference indexed in Scilit: