As presented in Table 1, the MCm was significantly lower for the samples with a film-forming coating than the samples with a non-film-forming coating. A similarly lower result of the film-forming coatings is also presented in Fig. 1a, which shows the MC fluctuations of the samples with the film-forming coatings (as an average of all samples with coating systems A, B and C) and with the non-film-forming coating (as an average of all samples with coating system D). All samples followed the variation in ambient RH and precipitation (Fig. 1b), but the curves of the two film-forming categories diverged at higher MC values and cohered at lower values. The higher MC of the samples with a non-film-forming coating confirms the higher moisture permeability of such coating, resulting in faster water sorption during periods with high precipitation and RH and likewise, faster desorption during periods of drier weather. Figure 2 shows boxplots with the median MCm values of each wood–coating combination. The results in the boxplot further support the higher permeability of the non-film-forming coating.
Figure 3 presents the average MC of the samples with a film-forming coating. The influence of density is pronounced in the figure. As can be seen, the low-density wood displayed larger MC changes compared to the high-density wood. Figure 4 presents the MC fluctuations of the corresponding samples with the non-film-forming coating (D). A similar pattern in MC variations’ relative density was observed for samples with a calcimine coating. The deviation in MC between high- and low-density samples with a non-film-forming coating was larger than those of samples with a film-forming coating.
Figure 5 presents the linear coefficient (k value) of the wood and coating parameters, relative to the MCm of the total replicates (MCmT), R2 = 0.42, Q2 = 0.86. R2 is the coefficient of determination and Q2 is the predictive ability of the model. A negative k value means that the parameter is likely to result in a lower MCm, and the opposite for a positive value. The k value confirms previous results, showing that all three film-forming coatings (A, B and C) had a negative value and led to a lower MCm, while the non-film-forming coating (D) had a positive value and a higher MCm.
Furthermore, the k value supports the results from Figs. 3 and 4, with a positive value for low-density wood and a negative value for high-density wood (95% significance). However, the k value of the parameters of heartwood and sapwood had no significant effect on the MCm.
It seems like the variation in MC of coated spruce is more related to wood density than to heartwood and sapwood characteristics, and the influence of density is independent of the coating permeability. Low-density spruce with larger lumens and thin cell walls facilitates an increased diffusion of water through the wood structure [23, 24], which is also likely to increase MC fluctuations in coated low-density wood. It is, therefore, suggested that the wood characteristics such as density and growth ring width impact the water sorption behaviour of coated spruce both in terms of seasonal MC fluctuations and average MC over time (MCm). The sorption behaviour of coated spruce was furthermore similar to that of uncoated wood, which has an increased diffusion with decreased density [12, 24]. The uncoated backside of the samples might have contributed to this similarity.
Moving to PCA, the model used all measured data along the timeline as input, R2 = 0.77 and Q2 = 0.69. This means that each sample contributed with multiple values that have been measured during the test period. The following score plots in Figs. 6, 7 and 8 contain the same data but are coloured according to different parameters. Figure 6 is the score plot presenting the pattern of different coatings. The figure shows a clear division of the two clusters, i.e. the film-forming (A, B, C) and non-film-forming (D) coatings. The cluster of film-forming coatings is further separated into the different coatings. Replicates with coating B had the most positive PC2 values, followed by coatings C and A (in decreasing order). The score plot does not reveal the magnitude of MC relative to different variables; rather, the position of each replicate illustrates how close they are relative to others in MC. Hence, coatings A and B had the most contrasting MC among the film-forming coatings since they are farthest from each other as shown in Fig. 6. Next, Fig. 7 is coloured according to the parameter of high- and low-density wood. Like previous figures, Fig. 7 shows a separation among different groups, but in this case, the focus is on density conditions.
Comparing the pattern seen in Fig. 6 with the pattern in Fig. 7, a separation between high- and low-density wood can be seen within each cluster of film-forming and non-film-forming coatings. The third score plot, Fig. 8, illustrates the colouring related to the heartwood and sapwood characteristics. Figure 8 is the only score plot with no clear clusters of colours or pattern. Hence, the results of this work did not indicate any impact of heartwood or sapwood characteristics on the MC of coated spruce exposed to outdoor conditions. In contrast to this study, other studies have shown that uncoated heartwood spruce absorbs a lower amount of water as compared to sapwood [16, 25]. However, all the aforementioned studies are related to uncoated spruce wood. It seems that a coating system reduces such differences in water sorption behaviour between the heartwood and sapwood of spruce, which is probably related to a changed sorption mechanism involving free and bond water diffusion and wetting properties. However, more studies are needed to produce a deeper understanding of the sorption mechanisms involved for coated wood.
The ANOVA analysis with a subsequent Tukey’s test showed a significantly different MCm among some of the samples. Coating B had a significantly different MCm between high-density heartwood and low-density sapwood. Coating C had a significantly different MCm between high-density sapwood and low-density heart- and sapwood. Coating D had a significantly different MCm between low-density sapwood and high-density heart- and sapwood. However, the scattered ANOVA analysis did not show any general trends. The overall impression was that depending on the coating, different wood combinations contributed to a significantly different MCm. Coatings do not change the EMC of the wood; it only delays the water transportation through the film. The boundary conditions for the wood are, however, changed with a coating on the top. Hence, the scattered ANOVA suggests that the water sorption of coated wood depends not only on the sorption capacity of the wood, but on its interaction with the properties of the coating.
Table 1 also presents the length of cracks on the samples at the end of the experiment in June 2012. At this point, none of the samples with a film-forming coating had any visual signs of cracks, which can be due to the relatively low MC fluctuations during the exposure period. The first signs of cracks on the non-film-forming coating D appeared during the second year of exposure. As can be seen in Table 1, the high-density sapwood with coating D had a substantially higher crack length than the other wood samples. The dimensional movement of wood is related to the density and MC of the wood [19], and there is a tendency of a higher number of cracks on the high-density samples. However, the two samples of high-density heartwood and sapwood with coating D had a large difference in crack length despite a similar density and MC (heartwood: ρ = 431 kg/m3, MC = 20.04%; sapwood: ρ = 471 kg/m3, MC = 20.05%). Sandberg [26] studied the crack development of coated and uncoated wood and found a higher number of crack lengths on sapwood samples, irrespective of whether they were coated or not. Her finding strengthens the results in this study; however, the combination of large differences in crack formation with a similar MC was still unexpected. The differences in crack formation might be related to the moisture gradients of the coated heartwood and sapwood samples. Future studies of moisture gradients in outdoor-exposed spruce boards are, therefore, suggested to investigate if such relationship exists since this study only measured the average MC of the board. Svensson and Mårtensson [27] described, for instance, the stress formation of an uncoated sapwood board during drying, with increased drying stress at the surface when the board reached zero moisture gradient. Growth ring orientation also contributes to differences in crack development. For example, radially sawn wood with vertical growth rings shows less crack developments, especially when compared to flat sawn wood with horizontal growth rings [28]. The wood samples in this work had the same type of growth ring orientation (horizontal).