A comparison of prediction models for cover type transitions and their effects on harvest schedules in coastal Oregon

by Eng, Helge

There is considerable uncertainty associated with stand

establishment in coastal Oregon. Several species can gain

control of the site depending on stand conditions as well as

management practices. In this study, three cover type

transition models were developed using data from the Siuslaw

National Forest to predict stand establishment patterns in

aggregate forest planning.

Two discrete alternative probability models were

analyzed. The full model included the cover type and age of

the mature stand before harvest as well as vegetation

management treatments. The reduced model included only the

cover type of the mature stand. A third model, the naive

model, consisted of assigning the old-stand cover type to

the regenerated stand with probability one.

Discrete alternative probability models constitute a

challenge in evaluation and validation. A residualgenerating

procedure developed in this study allowed graphic

analysis of residual plots. Although specification bias in

these models remains a potential concern, the full model

gave the best fit of the models analyzed in this study.

The different cover type transition models developed

here resulted in substantially different forest plans in

terms of harvest levels, financial return, and forest

structure. The full model produced forest plans with lower

harvest levels and financial returns, but greater biological

diversity, than the other models investigated in this study.

These results were achieved primarily through the typeconversion

effects of vegetation management treatments.

The attractiveness of vegetation management depended on

minimum rotation age requirements as well as cover type

conversion policies in the forest planning formulation.

Vegetation management was most effective at short minimum

rotation ages. Type conversion policies requiring either

constant or non-decreasing biological diversity over time

resulted in forest plans in which vegetation management was

consistently a marginal or unprofitable investment.

Given the substantial impact of different cover type

transition models on forest plan results, and the

indications of systematic bias in the models tested here,

additional study on predicting cover type transition

patterns is recommended. Recognition of the spatial

configuration of hardwood seed sources surrounding the

reforestation areas is likely to be an important area of

research.