Mixed models to estimate tree oven-dried cork weight in Central and Southern Portugal

Abstract

Models of oven-dried cork weight at tree level were developed using dendrometric variables and rotation cycle of cork production (9 or 10 years) as predictors. The models were based on data obtained from permanent plots laid out in five cork regions, covering most of the cork production area in Portugal. Dendrometric variables included those related to tree size (as perimeter at breast height, stem height or crown dimensions), related to management decisions in the stripping process (as stripped length in stem and branches, stripping surface, stripping intensity or stripping coefficient) and those gathered from a cork sample (as cork thickness, cork density or cork moisture). Variables were grouped according to the complexity of measurement with the aim of developing models suitable for management, that include variables with cheaper measurement, and for research, including variables with higher measurement cost. Models were structured as mixed models with random effects decomposed in regional, plot and tree-level effects, allowing higher sensitivity and more realistic variance–covariance structures comparing with fixed effects models. The variance component at tree level reflects 80–90% of total variability and at plot level is only 1–3%. The nine selected models, five for research and four for management purposes, are linear models that reflect the allometry in the relationship between cork weight and tree size.