Examining bias in estimating the response variable and assessing the effect of using alternative plot designs to measure predictor variables in diameter growth modeling

by Zumrawi, Abdel Azim

Abstract (Summary)
Diameter growth of Douglas-fir (Pseudotsuqa menziesii (Mirb.) Franco) estimated from increment cores was compared with that obtained from repeated measurements of tree diameter on permanent plots located in two Douglas-fir study areas in the central Coast Range of Oregon. Growth was measured for a 6-year period (1979-

1985). Diameter growth measured from two increment cores taken opposite to each other, provided an unbiased estimate of the stand average diameter-growth as determined from repeated measurements of diameter. However, a statistically significant trend was found in the differences in individual tree diameter-growth between the two methods of measurements. A nonlinear

model was used to characterize these differences. The practical significance of the observed trend and the use of the developed model as a calibration tool, depend upon the reliability desired by the particular user. The second part was a simulation study to examine the effect upon growth model predictions of using

alternative sample plot designs to measure predictor variables. Five forest stands were generated through computer simulation by use of field data and random spatial distributions. Two variable-radius plot designs and four fixed-radius circular plots were used to sample simultaneously the generated stands . Sample data then were used to simulate diameter and gross-basal-area growth in both a single-tree/distance-independent growth model and a whole-stand/diameter-free growth model. In comparing the growth predictions of each model, the plot design used to develop the model was the standard against which alternative plot designs were evaluated. Both fixed and variable area plots provide, with varying degree of precision, unbiased estimates of stand-level predictor variables. For both models, average gross basal area

growth-rate predictions from 50 samples of each alternative design were not significantly different from the standard design. However, large differences in individual predictions may occur as a result of using a different plot design. The magnitude of these differences depend on the stand size and density.

Bibliographical Information:

Advisor:Hann, David W.; Thomas, David; Peters, Dawn; Paine, David; Verts, B. J.; Brodie, Douglas

School:Oregon State University

School Location:USA - Oregon

Source Type:Master's Thesis

Keywords:douglas fir growth measurement mathematical models oregon coast ranges


Date of Publication:03/01/1990

© 2009 All Rights Reserved.