**4. Discussion**

Our results support our hypothesis that confocal profilometry can quantify shelling at the surface of weathered deckboards, but the methodology is not completely straight-forward. The surfaces of weathered deckboards contain features that span various scales. For example, in the *Z*-direction (height), wood cell walls and raised grain can project 10 to 50 μm from wood surfaces, respectively [16,17,27]. Weathering can lead to millimetres of erosion of wood [6], and these variable perturbations in the vertical heights of deckboard surfaces can be superimposed upon more structured surface features, such as machined profiles [21]. These multiscale features of wood surfaces have all been quantified using confocal profilometry, but probes with higher sensitivity and smaller sampling areas are required for features in the micron range as compared to those in millimeter range [16–19,27,28]. Hence, selective sampling of areas of deckboard surfaces showing shelling was required, particularly for profiled boards, to extract information on the maximum height of shelled growth rings, which was the parameter of greatest practical interest. When we took this approach, confocal profilometry was able to reveal a negative correlation between growth ring angle and severity of shelling in flat-sawn western larch boards oriented pith-side-up, and confirm previous suggestions in the literature that flat deckboards oriented pith-side-up are more susceptible to shelling than boards oriented bark-side-up [7,9–11]. Our results also accord with results of Davis, cited in Koehler that wood density and width of annual ring appear to have no effect on shelling, and also his suggestion that wood species differ in their susceptibility to shelling [9,10]. Further research is needed to fully explore species differences in susceptibility to shelling, and the influence of other factors on shelling, for example, contrast in density between earlywood and latewood [10], and the application of water-repellent finishes [29].

We showed, for the first time, that shelling occurs at the surface of profiled deckboards exposed to the weather, and we described the morphology of shelling that developed in profiled boards. The shelling of profiled boards was due to the separation of latewood tips on the peaks of profiles, and our results suggested that boards with a wavy profile were less susceptible to shelling than boards with rib profiles. One possible explanation for this observation is that boards with wavy profiles have narrower peaks (1.2 to 1.3 mm, Table 1), which could reduce the possibility of latewood tips occurring on the peaks of profiles, and subsequently separating from earlywood, when compared to rib profiles that have wider peaks (2.2 to 2.4 mm, Table 1). Shelling of profiled boards was more pronounced in deckboards oriented pith-side-up compared to those oriented bark-side-up, in accord with our observations of the shelling of flat (unprofiled) boards.

Our observations confirm that shelling at the surface of boards oriented pith-side-up occurs due to fracture at the boundary between the latewood of one growth ring and the earlywood of the subsequent growth ring [9,10]. It is at this point that there is the greatest contrast in density between latewood and earlywood tracheids [30] and, hence, the potential for maximum stress concentration due to differential swelling and shrinkage of latewood and earlywood. These stresses are responsible for delamination at growth ring boundaries according to Koehler [9]. Intra-wall rather than interfacial fracture was observed in earlywood cell walls at the growth ring boundary of a shelled growth ring in western hemlock. We sugges<sup>t</sup> that fracture at the latewood-earlywood boundary allows for unrestrained swelling of latewood, causing it to curl away from the wood surface in the same direction as the curvature of growth rings. We did not observe similar shelling in boards oriented bark-side-up although latewood projected slightly above wood surfaces, possibly due to springback or differential swelling of latewood and/or increased erosion of adjacent earlywood. Checking occurred at the interface between latewood and earlywood in boards oriented bark-side-up, but extended radially into wood, rather than propagating via the boundary between latewood and earlywood. The mechanisms that are responsible for the pronounced difference in shelling of boards oriented pith-side-up versus those oriented bark-side-up are well explained by Koehler [9] and in an anonymous publication by the U.S. Forest Products Laboratory that draws upon Koehler's work [10]. Our observations accord with Koehler's explanations [9] for the effects of growth ring orientation on the shelling of flat-sawn boards.

Shelling is a serious defect in wood species with a pronounced contrast between earlywood and latewood density [9,10]. Shelling of western larch wood was mentioned by Johnson and Bradner [12], who stated that it is objectionable because of danger from splinters, and its potential to create a tripping hazard. Our results indicate that the simplest way of avoiding the occurrence of shelling at the surface of flat western larch deckboards, is to orient boards bark-side-up rather than pith-side-up. Shelling can also be minimized, as others have pointed out, by machining boards with sharp cutter knives [9,10], and using a water-repellent treatment to reduce surface changes in wood moisture content [29]. All of these recommendations are also relevant to profiled larch deckboards, which developed shelling when exposed to the weather. An additional measure that would help reduce shelling of profiled larch deckboards is to use a wavy rather than a rib profile. Profiled deckboards should be profiled on both sides to produce a balanced board that is less susceptible to cupping [8]. A profile on the pith-side of boards that is different to that on the bark-side would make it easier to orient deckboards correctly, although the sub-surface profile would need to be able to achieve its desired aim of reducing cupping. Further research is needed to develop such sub-surface profiles, which could include saw kerfs or grooving that are employed on the undersides of flooring and some commercial deckboards [21,31].
