PARENT SESSION
Oral Session #30: Modeling.
Presiding: T. Day
Tuesday, August 6. 8:00 AM to 11:30 AM. Apache Meeting Room, TCC.
Response of an optimal tree diameter increment growth model to data uncertainty.
Bragg, Don*,1, 1 Southern Research Station, Monticello, AR
ABSTRACT- Optimal tree diameter growth equations have long formed the core of many forest gap models. One such design, the Potential Relative Increment (PRI) approach (based on Hoerl's special function), uses extensive public inventory data to estimate maximum growth. Uncertainties involved with inventory data may influence the PRI system's ability to predict potential increment, so different manipulations of the available data were tested to gauge their impact. The PRI methodology uses ordinary nonlinear least squares regression and an iterative fitting process in which hundreds to thousands of individuals are pared down to a subset (usually <20 trees) growing at the fastest rate given their initial diameter. A pool of 11,950 loblolly pines (Pinus taeda) from Texas was reduced to subsamples of 20, 10, and 7 individuals to compare how final subset size affects model form. The largest subsample predicted optimal increment most conservatively, with an earlier peak (at approximately 12 cm DBH) of 3.4 cm of growth annually, while the most restricted subset (7 trees) yield the most optimistic results, with a maximum annual increment of almost 5 cm at 20 cm DBH (the moderate subsample predicted an intermediate level). Because final subset size significantly affected growth predictions, a top-averaging approach was then tested as an alternative to a single-point maximum for each size class. Not surprisingly, the top-average approach proved more conservative, predicting optimal increments of 0.25 to 0.45 cm less than the single-point design across a range of loblolly pine diameters. While the top-averaging approach would be less sensitive to data uncertainty, it also systematically biases the fitted models towards lower growth projections, which is counter to the ultimate goal of developing "optimal" increment equations.
KEY WORDS: Pinus taeda, Potential Relative Increment, inventory, Hoerl's special function