Stem moisture content prediction model for Larix olgensis based on beta regression

Area of Science:

• Forestry Science
• Wood Science
• Ecological Modeling

Background:

• Understanding wood moisture content is crucial for timber quality and forest management.
• Longitudinal variation in moisture content within tree stems influences wood properties and processing.
• Artificial larch (Larix olgensis) plantations are significant for timber production.

Purpose of the Study:

• To investigate longitudinal variation patterns of sapwood, heartwood, bark, and stem moisture content in Larix olgensis.
• To develop and validate mixed-effect beta regression models for predicting tree moisture content.
• To identify key factors influencing moisture content distribution along the tree trunk.

Main Methods:

• Construction of two-level mixed-effect beta regression models incorporating plot and tree effects.
• Calibration of models using two sampling schemes: unrestricted relative height (Scheme I) and limited height (< 2 m) (Scheme II).
• Evaluation of model accuracy using Mean Absolute Percentage Error (MAPE) with varying numbers of sampled discs.

Main Results:

• Sapwood and stem moisture content increase longitudinally; heartwood shows a slight decrease then increase; bark moisture content increases and levels off.
• Relative height, crown base height, stand density, age, and dominant height are key drivers of moisture content.
• Scheme I achieved stable prediction accuracy with 2-3 discs (MAPE up to 7.4%), while Scheme II was effective with discs at 1.3 and 2 m (MAPE up to 7.1%).

Conclusions:

• Mixed-effect beta regression models provide accurate predictions of Larix olgensis moisture content.
• The models effectively account for plot and tree variability, enhancing predictive power.
• The findings support optimized sampling strategies for moisture content assessment in forest inventories and wood quality evaluations.