Evaluation of Double-Sampling Estimators of Subalpine Herbage Production

Abstract

Electronic capacitance metering, relative and dry weight estimations, and canopy cover estimation of herbaceous standing crops were statistically evaluated with respect to sampling costs, precisions, and the vegetal and environmental factors which affected their double-sampling correlations. Twenty-four factors were investigated using stepwise regression analysis. Minimum sampling costs were achieved when the double-sampling estimator technique was compatible with the characteristics of the sampling site. Relative and dry weight estimations were found to be consistently precise estimators in meadow, aspen, fir, and spruce-flr, and spruce-fir vegetation types. Both were successfully used by workers with no prior experience or training. The sampling techniques were capable of providing, at equal sampling cost, up to a 4-fold increase in sample size over that of clipping along, depending on the vegetation type. The need for an inexpensive and precise estimator of herbage yield has become a classic topic in the range literature. A great deal of attention has been paid to the subject, primarily because of the difficulty in adequately sampling the highly variable standing crops encountered in many vegetation types. Since harvesting herbage is expensive and destructive, a technique which uses a large sample of acceptably lower precision estimates is generally better than a small sample of precise clipped measurements. The statistical parameters obtained with an estimator, however, must not be biased (Haydock and Shaw 1975) and the sample must also be less expensive to obtain than clipping, at equal levels of precision (Green 1949). A variety of both direct and indirect estimation techniques have been proposed which use a doublesampling with regression procedure (Cochran 1977) to minimize bias. While most of the techniques have been tested in the field, an uncertainty in how they compare with respect to cost, precision, and the factors influencing their performance, h,ave made it difficult to choose the most efficient technique in a given sampling situation. The need for analysis of the factors influencing the precision of various double-sampling estimators has been recognized by several authors (Morley et al. 1964; Bryant et al. 1971; and Michalk and Herbert 1977). Jones et al. (1977) recently investigated the effects of tropical grassland composition on The authors are former graduate research assistant, former research technician and professor, respectively, Department of Range Science, Utah State University, Logan, UJtah 84322. Reese is now ecologist, Division of Energy, Missouri Department of Natural Resources, Jefferson City. Bayn is now research technician, Department of Biology, Utah State University. This research was supported by National Science Foundation Grant DEB 75-13966 to James A. MacMahon and Utah Agr. Exp. Sta. Project 766. Submitted as Utah Agr. Exp. Sta. Journal Paper 2272. The authors wish to thank Mr. K.P. Haydock, C.S.I.R.O., Div. of Mathematics and Statistics, Cunningham Laboratory, St. Lucia, Queensland and Dr. Walter F. Mueggler, U.S.D.A., Forest Serv., Forestry Sciences Laboratory, Logan, UJtah, for their critical comments on the manuscript. Debbie Ketchie, Susan Kadlec, and Marilyn Pratt assisted with field work. electronic capacitance metering. However, similar regression studies on other double-sampling estimators are