Epsilon Archive for Student Projects

Abstract

This MSc thesis is divided into two parts. The first part deals with the need for information
from forest campartments selected for clear cutting, in connection with planning in an
integrated forest- and sawmill enterprise. A review of literature is followed by a case stud y,
where present conditions in a Swedish forest company are compared with current research in
the area. The seeond part consists of a simulation stud y, which aims at camparing the
accuracy of inventory estimates using different methods. The estimates are delivery forecasts,
covering one month consumptian of timber in a large sawmill. The range of inventory designs
includes systematic circular plot sampling as well as the use of register data in combination
with a data base containing tree data. The results, i. e the forecasts and their standard errors,
are presented separately for tree species, log type and diameter class.

In many Swedish sawmill campanies of today, there is a trend in productian philosophies
towards a greater adaption to customer demands. As a component in this process, the
requirements on the flow of raw material will change. The right assortment, of the right
quality, must be delivered at the exact right time. Due to the time span between harvest and
delivery of timber from the forest, the mills will depend upon adequate forecasts. This implies
controlied bucking to meet the demands from the mill, as well as improvment of existing
models for predieting log yield from single forest stands. The stand data must allow flexible
computerized bucking simulation, allowing adjustment to specific requirements at different
points in time, generating prognoses for the distribution of logs into log classes after
harves ting. Data used in the models can be collected through forest in ventory or by using
stand register data. In the latter case, the diameter distribution must be constructed from the
known stand charactaristics, by the use of mathematical models or by assignment from a data
base with stem data.

It might seem obvious that the information used in the models should be the best possible, but
inventory actions are costly. The "total cost" of an inventory consists of two parts, one being
the cost of performing the inventory and the other the expected loss, due to inoptimal
decisions caused by erroneous data. The optimal solution is given by the minimization of the
two components. Customer orientated production in a sawmill might entail great losses, due
to lack of precision in prognoses regarding raw material, for products produced in close
cooperation with important customers compared to standard products. Thus, different data
acquisition policies should be used in stands containing different kinds of timber, to create
prognoses adapted to the actual requirements in the industry.

The integration between forest management and the sawmill industry, creating a customer
oriented flow of raw material, is slowed down because of difficulties in expressing the
demands on the raw material - the timber. In addition to this, the ability to forecast log yield
is unsufficient, as well as the routines for follow-up between sawmills and forest departments.
Large Swedish forest enterprices have the best prospects for overcoming these obstacles, due
to large forest holdings as well as a modern sawmill industry. The use of information from
forest stands harvested, for adapting the composition of the timber deliveries to the demands
from the mills, will constitute a competitive edge for these companies. The development of
systems such as computerized bucking control, information technology, including cordless
communications, and positioning by satellite, creates the technical apportunities for tailoring
the wood flow to customer requirements.

Normally, when studying sampling methods for estimating log yield after clear cutting, the
results are presented as total volumes of timber for pine and spruce. The results from the
stud y indicate that the marginal benefit of sampling 25 sample plats, instead of 10, is small.
The standard error of the estimate of total volumes is lowered by only l%, although the cost is
more than doubled - from 40 000 SEK to 92 000 SEK, for a total volume of 50 000 m3fub. If no inventory action is performed, and instead register data connected to a data base containing
tree data is used, the estimates are still useful. There is no variable cost using this method. The
method for using register data outlined here is very straight forward and simple. By
developing the constitution of the data base and the routines for assigning substitute data, it
should be possible to improve the technique substantially.

However, if the results are broken down on estimates of volumes distributed in to
combinations of tree species, log type and diameter class, large variation becomes visible. This
is most evident for butt logs with large diameter, where the standard error might amount to
50%, although the standard error for the total volume is only 3-4%. This implies that more
intensive inventory designs should be used in compartments with large mean diameter. The
term "adapted inventory design" is used to describe an inventory scheme in w hi ch different
designs are used depending on the properties of the compartment More intensive designs are
allocated to compartments with high values for mean diameter, mean height and stock. In
other compartments more extensive inventory designs are used, e. g register data connected to
a data base. Thus, the resources are allocated to compartments containing desired timber
trees, and the estimates for selected segments are improved. The results indicate that it is
possible to lower the standard errors for estimates of volumes in selected segments by using
adapted inventory designs. This is done without increasing the over all inventory cost,
compared to a common method in practice today.