The stand generator, STAG, is a sophisticated processor developed to ensure
that data sets destined for use in the CACTOS projection system are complete.
Since there are different levels of data availability, the stand generator
was designed to: (1) fill in missing measurements of tree height and/or 
height to crown base found in inventory field data; (2) convert stand table
data, numbers of trees by DBH classes and species, to individual tree records,
and (3) generate stands from summary statistics. Each of these capabilities
is discussed in brief detail below.

III.A. Filling in Missing or Incomplete Data

Inventory field data collected by forest land managers for use with CACTOS
often have missing values. The values may be missing at random, or missing 
because the sample design called for sub-sampling total height and/or live 
crown ratio. In either case, STAG is able to fill in missing height or live
crown ratio variables. If desired, the user has the option of adding random
errors to the estimates to reflect the variability found in the modelling
data set.

When heights are sub-sampled, the available height data can be used to 
localize the height prediction model. This is accomplished using either an 
ad-hoc approach or a pseudo-Bayesian approach that adjust the amount of 
change to the model parameters by a constant ratio between 0 and 1. Within
the ad-hoc procedure, a weight of zero causes the update routine to abort 
(no update), while a weight of one places all the emphasis on the local 
sample to determine the coefficients for the height prediction equation are 
quite close to the database values. If, however, there is a large local
sample, then the pseudo-Bayesian estimates are a compromise between the 
database values and those determined from the local sample. See Van Deusen
(1984) and Biging and others (1991) for more information on the updating
procedure.

STAG, unlike CACTOS, allows four different types of tree height measurements:
1) total heights; 2) heights to a merchantable top (<6.5 in); 3) heights 
measured to whole (16.5 ft.) logs; or 4) heights measured to half logs 
(8.25 ft.). Within a STAG stand description file all heights must be of the 
same measurement standard. STAG uses a taper equation for the six major 
conifer species (species numbers 1-6) to solve for total height to a 
merchantable top or number of 16.5 ft. logs is supplied.

There are several cases when STAG does not estimate total height from 
merchantable height or number of logs. These cases include: 1) when taper
equations do not exist for a species (note: there are no taper equations for
any species other than the six major conifer species); 2) when the 
merchantable top is greater than 6.5 inches (total height predictions become
inaccurate when the merchantable top is too large); and 3) when tree DBH is
less than 8 inches (a tree is considered unmerchantable if its DBH is too 
small). In these cases, height is set to missing and can be filled in using
the standard missing data techniques within STAG.


III.B. Converting Stand Tables into Complete Tree Lists

Stand table information can be used to develop distributions from which 
individual trees and their accompanying dimensions (diameter, heights and 
crown ratios) can be generated. The created tree list has the property that
reclassification of the diameters would reproduce the original stand table.
Also, the sum of the tree weights of pseudo-individual trees within a 
diameter class equals the original tree weight from a diameter class of the
stand table. The tree weight represents (in this case) the probability of 
the tree having actually been on the plot. This new tree list mimics the
inherent variability found in inventory data and produces a list of trees 
having diameters different than the diameter class midpoints of the stand
table. The procedure used to create the tree list completes the tree list by
supplying height, crown ratio and tree weight (expansion factor) of each 
generated tree. In essence, this process produces a facsimile of a permanent 
plot using only the numbers of trees by diameter class and species. The 
resultant tree list is quite large, but is currently limited to be between 
100-500 records. Converting stand tables into complete tree lists is only
recommended when the width of the diameter classes is less than or equal to 
two inches so that an accurate portrayal of the diameter distribution can be
obtained.

III.C. Generating Stand Descriptions

STAG has the additional ability of generating complete tree lists when only
summary ststistics are available to descibe the stand of interest. This 
capability has obvious utility for transforming summary information into a 
form that can be used with the CACTOS individual tree growth simulation 
system. It also can be used to produce stand descriptions so that investigation 
of forest growth dynamics can be performed in the absence of field data 
observations. While this is not generally recommended, it can be important
when the manager is evaluting the performance of stands that currently do not 
exist, but which he or she may wish to establish in the future. These 
hypothetical stands can easily be created within the interactive environment 
of STAG. Appendix B.4 gives a demonstration of how this interactive process 
can be used.

STAG requires certain information when either the Weibull or the negative
exponential diameter distribution models are used. This information includes:
stand elevation, basal area per acre by species, the number of trees per acre
by species, and site index for each species of interest. In the case of the 
negative exponential diameter distribution the "Q value", or diminution
quotient, must also be specified.

To keep the number of tree records for selected diameter classes manageable
the understory data can be collapsed into classes for DBH, total height, and
live-crown raio by invoking the collapsing algorithm with the cu command
issued from the sg routine.


Weibull distribution model

It is possible to iteratively generate various components of the stand using 
a Weibull diameter distribution until the user has built a stand matching 
the desired description. Each species that is represented in the overstory 
is considered to be a separate component, and the understory in its entirety 
is considered to be another component.

Because stands of trees are often simulated for over 30 years with the 
CACTOS System it is essential to be able to generate an understory component
that matures with relatively long simulations. When generating understories 
using the Weibull distribution the user is given several options. In one 
opeion the user specifies the information on the overstory such as the 
average diameter of trees over 5.5 inches in DBH, and the number of trees 
per acre greater than 5.5 inches in DBH. This information is used to 
generate a Weibull distribution for understory trees. Alternatively, the 
overstory trees can be used directly to predict the Weibull coefficients for
the distribution of diameters of small trees. The user can then either 
specify directly the number of understory trees to generate, or they can use 
prediction models for estimating the number of small trees to be generated.
The user must also specify the species composition.


Negative exponential distribution model

The user may also generate a diameter distribution for any species using the
negative exponential distribution model by specifying the diminution quotient
or "Q" value. When this model is employed, the user has the option either to 
generate understory trees of a given species simultaneously with the over-
story trees, or to generate understory trees separately from the overstory 
as is done when using the Weibull distribution model.