Comparing and Selecting Models

Table 1 lists and compares a number of model characteristics and features that may be helpful in selecting an appropriate model for a given situation. Even though there are several models available, it is entirely possible that there may be situations where none of them are appropriate or where even the "best" one should be used with caution. There are a few simple, intuitive steps or questions that can help us select growth and yield models.

• Keep in mind your intended application, for example, is it silviculture, inventory or planning?

• What type of stand do you have and what kind of information do you currently have about it? You can set aside models that are not calibrated for your species and stand types. You can also determine if you currently have sufficient data to run a particular model, as some require more detailed data than others (e.g., permanent sample plot data versus inventory data).

• What type of density management treatment(s) you are interested in and how much treatment flexibility do you require? Although all the models in the table have some capability for stand density management, most only support a limited number of options.

• Finally, you will no doubt be exposed to conflicting results regarding these and other growth and yield models. Some differences are due to the models themselves but others stem from an emphasis on biology versus economics, or different interpretations of the historic theory versus science. Data from reproducible experiments should always be the ultimate arbitrator for both scientific theories and models. "In God we trust; all others bring data" (K. Iles, personnal communication)

Link to or Download MS Word Version (you may have to right-click to save) of
Table 1:  Model Comparison Table (Stearns-Smith, 1999)

Managed vs. Unmanaged Stands
What Model Might be used?

The following information will help planners identify managed vs. unmanaged stands (polygons), for the application of appropriate yield models (curves). The distinction is important because managed stands tend to be more productive. A decision tree is provided to help make the distinction between managed and unmanaged stands and suggest yield models for various stand types (Nussbaum, 1998).

Managed stands

Managed stands are even-aged stands which have benefited from management activities to encourage their growth potential. They have known establishment conditions including species, density, and distribution of stems. Managed stands can be:

• planted or natural origin but have not experienced repression or overstory competition.
• harvested stands regenerated after 1986 which have achieved "free growing" status, as specified in the regional free growing guide books. For stands regenerated before 1987, silviculture records, management plans, and local knowledge are needed to determine if stands are managed.
• young spaced and fertilized stands if the establishment conditions can be approximated
• partially harvested stands, such as commercially thinned stands if they were unsuppressed or unrepressed during establishment, establishment conditions can be approximated, and removals are documented.

Unmanaged Stands

Even-aged stands have not had the benefits of management and their establishment conditions are unknown. Although some stands may achieve their potential, others may have inadequate stocking, experience overstory competition, or repression.

Uneven-aged stands are considered unmanaged for this exercise, as the concept of establishment conditions holds little meaning, and a large number of stems could be suppressed by an overstory. These stands generally contribute to non-timber objectives where maximizing growth is not the primary concern. They have historically been handled as "naturals".

Models Decision Tree

Forest growth and yield models have been developed for many different purposes. It is important to choose the proper model and understand it’s assumptions and limitations. Models can be sophisticated computer models or simple yield tables derived from appropriate data. The above distinction between managed and unmanaged stands should help in choosing an appropriate yield model for your particular application.

See Models Decision Tree Flow Chart

A word of caution
about models!

Forestry and statistically-based biological experimentation are both relatively new sciences whose joint development is governed largely by the (slow) rate of tree growth. Seeming contradictions among the limited existing experiments serve to highlight our imperfect understanding of complex biological systems and discourage risk-laden investment decisions based on limited (or select) information. Decision making given imperfect information requires risk analyses which take into account the uncertainties regarding future biological and economic consequences. Models can be important tools, but we should not rely solely on them for making decisions. Use your professional judgement to examine your own data and assumptions before making the final management decision.

Model Application and Use

Selecting a model is only half the battle. Proper use of a model also depends on proper selection and preparation of the input data and proper interpretation of the model output. This is why most regulatory agencies avoid any open or implied sanctioning of specific models in favor of yield table approvals.

The main uses of growth and yield predictions are to:

• increment and update forest inventories
• compare silviculture treatments by simulating treatments and predicting outcomes
• influence stand and forest level decision making
• provide input for forest management planning including timber supply analysis, Allowable Annual Cut (AAC) determinations and policy making
• assess the impact of timber losses due to pests and fire
• allow extrapolations for missing or inadequate data
• explore and teach tree and stand dynamics

The application of any model in silvicultural decision support also requires a clear statement of management objectives translated into appropriate quantitative values that can be identified in model output. Care must be taken to understand the implications and limitations of using various quantitative measures as surrogates for management objectives. (Click here for an example)