Experimental design
Data were collected between July 2008 and March 2009 from a radiata pine clonal variety trial at Esk Forest, Hawke’s Bay, New Zealand (latitude 39.274 S, longitude 176.789 E, elevation 180 m). The trial was established in 2001 as an incomplete block, single tree plot design, and consists of 42 plots each containing 48 trees planted at a 4 x 4 m spacing (nominally 625 stems ha-1). The site is in its second rotation of radiata pine. The trial has been pruned in three lifts, occurring between 2005 and 2007, to a height of around 6 m. Hydrated sodium calcium borate hydroxide (Ulexite) was applied in 2005 at a rate of 6 kg ha-1 in response to a foliar boron deficiency but no other applications of fertiliser were made. Over the trial period, this site has had a mean annual rainfall of 1500 mm and a mean annual temperature of 13°C.
Twelve clones were selected (Mike Carson, pers. comm.) to span the range of available breeding values for diameter (at breast height), height, wood density, acoustic velocity, dothistroma resistance, branching, malformation and straightness. Selected clones also covered the range in outerwood densities and stress wave velocities (as assessed using a standing tree instrument). For each clone, two ramets were randomly selected giving a total of 24 trees for measurements.
The mean (and range in brackets) for diameter at breast height, height and crown mass of the selected trees were respectively 20.95 cm (17.6 – 24.3 cm), 13.25 m (11.7 – 16.5 m) and 66.2 kg (29 – 146.2 kg).
Measurements
Each selected tree was felled then cut into sequential, parallel 30-mm discs from the butt to a top diameter of approximately 50 mm. A single disc was retained for spiral grain measurements from each annual growth unit meaning that discs were taken at approximately 1.8 m intervals up each stem from a height of about 0.3 m to a height of 7.6 m – 12.0 m depending on the height of the tree, i.e. from 5 to 8 (average 6.6) discs per tree. These discs were selected to avoid grain deviation around branches. The discs were kiln dried under conditions of 60°C dry bulb/56°C wet bulb to a moisture content of between 18 and 25%. They were then further conditioned at 25°C and 65% relative humidity for 2–3 weeks, resulting in an equilibrium moisture content of approximately 10%. Four equally-spaced, 25-mm-wide, radial strip samples (in the N, E, S and W directions) were cut from each disc.
Spiral grain was measured at 5 mm intervals along each radial strip by exploiting the so-called “T2 effect” (McGunnigle 2009; Marschner et al. 2005; Matthews and Soest 1984). Due to the cellular nature of wood, light specularly reflected from a wooden surface is scattered in amounts that differ with fibre direction. By measuring the relative reflected intensities over a range of directions, the grain orientation vector (usually expressed as a pair of surface and dive angles) can be estimated. Careful surface preparation is critical to the success of this technique, since it relies on the microscopic texture and any preparation that alters the nature of this surface (e.g. sawing) will influence the results (Shen et al. 2000; Eastin and Johnson 1993). For this study, surfaces were first sawn then hand planed with a carefully honed blade.
A red laser (635 nm, 5 mm spot size) was projected at 90° to the longitudinal-radial surface of the radial strips. The laser was moved from pith to bark along each radial sample in 5 mm steps. The raw surface (i.e. deviation from vertical in the longitudinal-radial plane) and dive (i.e. deviation from vertical in the longitudinal-tangential plane) angles were obtained from the specular reflection peaks. A peak-fitting algorithm was used to find peak locations based on the intersection of the maximum slopes. Analyses were carried out on grain values in relation to distance from the pith – not annual rings as in most previous studies.
Previous experience indicates that tilt correction significantly reduces the observed variation in spiral grain (Cown et al. 2010). Consequently, tilt adjusted values of surface and dive angle were also calculated, where surface angles from adjacent radii were used to adjust the measured dive angles. The tilt adjustment was carried out using the radii averages. The surface angle averages for the N and S radii were adjusted (by rotating the disc mathematically) to make the N and S radii equal and opposite in sign with the corresponding adjustment of measured dive angles in the E and W radii. Similarly E and W radii were used to adjust the N and S dive angles.
The most common convention is to consider the left-hand, or “S” pattern (as viewed from the bark side) as a positive angle and right-hand “Z” pattern as negative (Harris, 1989) and this was adopted for this study.
Analyses
Data were analysed using R (R Development Core Team 2011) and SAS Version 9.2 (SAS-Institute-Inc. 2000). An initial exploratory analysis was conducted using graphical procedures. These were used to examine the raw data, the probability distribution of SGA, and to produce box plots showing within- and between-tree variation in SGA. Means, standard deviations, and skewness and kurtosis parameters were calculated for each tree, and compared between clones using a one-way analysis of variance (ANOVA).
To determine how SGA varied within a typical stem, the following random coefficients mixed effects model was fitted using the MIXED procedure:
(1)
where, i = tree (1,…,24)
j = sample within tree
SGA
ij
= grain angle of the ijth sample
h
ij
= height within stem of the ijth sample
d
ij
= distance from pith of the ijth sample
In this model, a
i
, b
i
and c
i
were assumed to vary between tree and to be independent identically distributed normal variables with means α, β and γ respectively, and with an unstructured covariance matrix, while e
ij
is a randomly distributed error term. When fitting this model, the independent variables were standardised by subtracting their approximate mean values. Thus, 50 mm was subtracted from distance from pith, and 3 m subtracted from height within the stem. The effect of this approach was to make the random intercept term correspond roughly to a random mean for each tree. The model was used to test whether grain angle varied significantly in either the radial or longitudinal directions within a typical stem (by testing whether β or γ differed significantly from zero). Nonlinear and interaction effects were also modelled by adding and testing the significance of quadratic and interaction terms to Model (1).
To test for differences in grain angle between circumferential direction (N, S, E, W), mean grain angle was calculated by direction and tree, and a two-way ANOVA fitted with terms for tree and direction.