Establishing a model for the dry density of heartwood of
Norway spruce by parameters industrially measurable on green
logs
Abstract: In this study different models for the prediction of dry density from
parameters measured on green wood was tried on different datasets.
Data from old literature have been utilized to derive a multivariate PLS
model and to compare the significance of different variables. The data was
presented as average data for different stands from four different locations
in Sweden.
Validation was performed by applying the models on two different datasets:
One small sample from southern Finland and the large data gathered by STFI
and Skogforsk in the project “Skog Massa Papper”. The Finnish data was
acquired by measuring properties of log stumps from CT-scanned images.
Derived models were compared with an algebraic derived model and the density
correction suggested by EN384. The multivariate model, using green density
and position in the stem, can predict dry density of heartwood of Swedish
Spruce with a R² of about 60% and a standard error of prediction of 27.4
kg/m³ on a sample disc as its best. This is slightly better compared to
single variable models only utilizing the green density as variable. The
correlation on the testset was 78% which is promising when considering that
mill specific models should be made in case of industrial implementation
which should also improve the models fit on a validation data.
The model should also be tested with respect to the X-ray log scanners
ability to measure the variables and the measurement error connected with
the measurements.
CLICK HERE TO DOWNLOAD THE WHOLE ESSAY. (in PDF format)