| Statistical Programs |
| College of Agriculture | University of Idaho |
| Seminar Announcement |
| "Applied Statistics in Agriculture" |
|
Individual-Tree Basal Area Increment Models:
Development and Future Directions Presented By |
| William R. Wykoff |
|
USDA Forest Service Rocky Mountain Research Station Moscow, Id |
| Tuesday, January 9 3:30 P. M. Ag. Science 62 |
|
We will review development of a
model that predicts periodic basal area increment for individual
conifers of various species common to the Northern Rocky Mountains.
This model is the core component of the Forest Vegetation Simulator,
and has been modified to represent most of the forest types in the
western United States. Currently, we are revising the model to make
it more robust to variation in stand structure and differences in
sampling design. In the revised model, subplot density is used as a
local measure of competition when multiple subplots are included in a
stand inventory. Our approach recognizes that variables derived from
stocking are associated with sampling variation, which can be predicted
from tree size and inventory design. Ignoring the sampling variation
in such variables can result in biased predictions. We developed models to predict sampling variation and then used predicted variation and errors-in-variables regression to estimate coefficients of a structural increment model. Within this model, basal area is used on two levels. Background competition is represented by stand average basal area; the model of variation is simply the standard error of the mean basal area across the individual sub-plots used to inventory the stand. Local competition is represented by a transformation of sub-plot basal area; a Poisson approximation is used to estimate sampling variation. In application, the sampling variations are estimated for each tree, for each projection cycle, and the structural coefficients are adjusted to reflect changing magnitude. Compared to ordinary least squares, the structural model is more sensitive to point and stand density effects and less sensitive to crown size. Therefore, treatment of irregular stands is improved and simulated response to thinning is more realistic. Future work will focus on verification of error models and on development of regional applications. |
| Return to Seminar Listing. |