The influence of site quality, silviculture and region on wood density mixed model in Quercus petraea Liebl.

Abstract

Three linear mixed-effect models for wood density were constructed with average ring density data measured by X-ray microdensitometry on 10427 heartwood rings collected from 82 sessile oaks sampled in five French regions (Alsace, Lorraine, Orne et Sarthe, Allier, Loir-et-Cher). Two types of forest management (coppice-with-standards and high forest) and three site qualities were represented in each region. A model (1) between air dry wood density and radial tree growth was established. Model (2), with radial growth and tree environment (region, silviculture, site quality, and their interactions) as fixed effects, tested whether wood density varied with changing region, silviculture, or site quality when tree radial growth was kept constant. Model (3), derived from model (2), consisted of the effects or interactions found to be significant in model (2). The effects specific to a tree were accounted for by a random tree effect in the three models. We calculated that the fixed effects in models (1), (2) and (3) explained, respectively, 48, 50 and 48% of the total variation of wood density and that the random tree effects were responsible for 31, 29, and 31% of the total variation of wood density, respectively. The comparison of the three models indicated that the fixed effects of the region, silviculture, and site quality and the interactions among them, all present in model (2) or partly present in model (3), explain only a small percentage (2% and less than 1%, respectively) of the total wood density variation when tree radial growth is statistically controlled. Consequently, wood density hardly changed with changing environment and type of forest management when ring width and cambial age were kept constant. It implied that compressed trees in high forests and trees with relatively free radial growth from coppice-with-standards follow similar wood density mixed models. Thus, a single wood density model can be used to predict wood density efficiently in oak stems for several situations with contrasting silviculture, site quality, and geographic location.