Pulpwood green density prediction models and sampling-based calibration

Pulpwood arriving at the mills is mainly measured by weighing. In the loading phase of forwarding and trucking, timber is weighed using scales mounted in the grapple loader. The measured weight of timber is converted into volume using a conversion factor defined as green density (kg m(-3)). At the mill, the green density factor is determined by sampling measurements, while in connection with weighing with grapple-mounted scales during transportation, fixed green density factors are used. In this study, we developed predictive regression models for the green density of pulpwood. The models were constructed separately by pulpwood assortments: pine (contains mainly Pinus sylvestris L); spruce (mainly Picea abies (L.) Karst.); decayed spruce; birch (mainly Betula pubescens Ehrh. and Betula pendula Roth); and aspen (mainly Populus tremula L.). Study material was composed of the sampling-based measurements at the mills between 2013-2019. The models were specified as linear mixed models with both fixed and random parameters. The fixed effect produced the expected value of green density as a function of delivery week, storage time, and meteorological conditions during storage. The random effects allowed the model calibration by utilizing the previous sampling weight measurements. The model validation showed that the model predictions faithfully reproduced the observed seasonal variation in green density. They were more reliable than those obtained with the current practices. Even the uncalibrated (fixed) predictions had lower relative root mean squared prediction errors than those obtained with the current practices.

Automated summary

In this study, predictive regression models were developed for green density of different pulpwood assortments, with the models accurately reproducing seasonal variations. The models were found to be more reliable than current practices in predicting green density.