Predicting resin pockets and blemishes in radiata pine lumber from log properties

Background: Resin pockets and blemishes in pruned logs of radiata pine ( Pinus radiata D.Don) can reduce the value of clear and moulding grades of lumber. External resin features (ERF) on the bark of the logs have proved an effective method of predicting the incidence of resin pockets in the lumber. Resin canals have been associated with resin blemishes in radiata pine, and could prove useful in improving the prediction of the grade recovery of lumber. Methods: Pruned butt logs of radiata pine trees from two forests in the North Island, New Zealand, were selected for low, moderate, and severe levels of external resin features (ERF) on the bark, and for low, average, and high resin canal diameter, frequency and brightness from breast height increment cores. The relationships were evaluated between these properties, and the lumber resin features and grade recovery of the logs. Results: The number of resin pockets, the blemish rating, and the percentage of boards with resin streaks and resinous heartwood increased, and the recovery of clears and moulding grade boards and the lumber value declined, with the severity of the ERF class of the logs. Multiple regression models gave good predictions of the grade recovery and loss of lumber value, using the log ERF class, volume, heartwood content, and number of Type 1 resin pockets on the ends of the logs, as independent variables. The resin canal properties did not improve the regression models. Resin blemishes were associated with Type 2 resin pockets, and were more frequent in the forest where false growth rings were present. This suggests the constitutive resin flow from resin canals, rather than the resin canal size and frequency, is more important in determining the incidence of resin blemishes. The prediction of the grade recovery of lumber using the ERFs of radiata pine logs, was supplemented by the log volume, heartwood content and number of Type 1 resin pockets on the ends of the logs. The environmental factors that drive the constitutive resin enrichment of resin canals, such as drought conditions that give rise to false growth rings, could be useful in improving the prediction of grade recovery for forest stands.

value by 20 to 68%, due to the presence of resin pockets and other resinous defects on the surface of the boards (Hughes 2007). The presence in lumber of 0.7 to 2.0 resin pockets/m 2 of sawn surface area was found to reduce the value of the lumber from pruned logs of radiata pine by 7 to 34% (McConchie 2003). The loss in lumber value has led to the development of methods for the visual grading of pruned logs for resin pockets and blemishes.
The resin pockets and blemishes that occur on the ends of the logs have been evaluated as a method for predicting the incidence of resin pockets in the lumber (Ridoutt et al. 1999). The predictive models that were developed for radiata pine logs were limited by the low incidence of resin pockets on the ends of the logs. This prevented the establishment of consistent relationships between the log end assessments and the grade recovery of lumber.
The external resin features (ERF) on the bark of the logs have been a more effective method of predicting the incidence of resin pockets in the lumber (McConchie, 2003). The presence of external resin bleeding, lesions, and galls on the surface of the logs, can be used to segregate the radiata pine logs for the grade recovery of lumber. The volume of mouldings and better grades of lumber declined by 12, 26, and 55% for pruned logs of radiata pine with low, moderate, and severe classifications of ERF on the bark of the logs (McConchie, 2003).
Resin blemishes have been associated with the frequency of resin canals in the logs of radiata pine, with a higher incidence of resin blemishes on the ends of logs, among forest sites with higher resin canal frequency (Yang et al. 2007). Resin canal frequency appears to be weakly linked to the external resin features (ERF) on the bark of radiata pine trees (Yang et al. 2007;Ananais et al. 2010), and the resin pockets and resinous patches on the ends of radiata pine logs (Yang et al. 2007), which suggests it may be useful in predicting the incidence of blemishes in the lumber.
In this study, the use of external resin features (ERF) on the bark of the logs was evaluated in combination with the resin canals from increment cores, and log end properties, to determine if these measurements could improve the prediction of the grade recovery of lumber. The pruned butt logs of radiata pine trees selected for low, moderate, and severe ERF classes, and a range of resin canal properties, were sawn into randomwidth boards and graded for appearance lumber. The relationships were evaluated between the log ERF classes, the resin canal, and log end properties, and the lumber resin features and grade recovery.

External resin features
Trees of 26-year-old radiata pine trees were selected for low, moderate, and severe levels of external resin features (ERF) on the bark of pruned butt logs, from plantation stands in Northland and coastal Hawke's Bay (Figure 1, Tables 1 and 2). Ninety trees in Parengarenga Forest, Northland, and eighty-seven trees in Tangoio Forest, coastal Hawke's Bay, were selected based on the presence of external resin bleeding, lesions, and galls on the bark of the pruned butt logs, using the ERF classification in the WQI Field Guide (McConchie, 2003).

Resin canals
The vertical resin canal properties of all the selected trees were measured using breast height (1.4 m) increment cores, taken from the outer ten growth rings from the bark. One 12 mm diameter increment core was sampled from each tree, and air-dried at room temperature. A SilviScan strip (Evans, 1994), was cut from each core, and one transverse surface was polished with 240, 320 and 1200 grit abrasive paper, and scanned at 600 DPI using a flat-bed scanner. The scanning parameters were kept the same for all the strips (image brightness = 140, image contrast = 170) to ensure the resin canal brightness could be compared between the strips.  The diameter, frequency, and brightness of the resin canals were measured using the IDL imaging routine developed at CSIRO (Chen et al. 2008, Figure 2). Resin canal size was estimated by drawing a circle over each canal and calculating the circle area using the software. Resin canal diameter was derived from circle area. The average greyscale intensity of the pixels within the boundary of each resin canal was used to represent the brightness of the resin canal.

Log selection
The butt logs of thirty trees were selected for sawmilling from each of Parengarenga and Tangoio Forest, based on the ERF classes of the logs and the resin canal properties of the breast height increment cores. Ten trees were selected from each of the low, moderate and severe ERF classes, and for each ERF class the trees were selected for combinations of low, average, and high resin canal diameter, frequency and brightness.

Log measurements
The trees were felled in April 2009 and the pruned butt logs for sawmilling were assessed for small and large end diameter, length, acoustic velocity, heartwood content, resin pockets, and blemishes. Log volume, excluding bark, was calculated using Smalian's formula (Ellis, 1995). Acoustic velocity was measured using a Director HM200 (Fibre-gen, New Zealand) connected to a laptop computer and operated in supervisor mode, to ensure the correct resonance peak was used in calculating the log acoustic velocity. Heartwood content, resin pockets, and blemishes were measured using 50 mm thick discs, cut from both ends of the logs. One transverse surface of the discs was cut smooth using a disc-surfacing saw, then wetted with water, mounted in a camera booth on a back board with calibration pins, and photographed using a digital camera (Lee and Brownlie, 2009).
Resin pockets and blemishes in the disc photos were classified by type, and the dimensions, location, and ring position were measured using an IDL imaging routine developed at CSIRO (Ottenschlaeger et al. 2012). Disc photos were corrected to a constant scale, using the back board calibration pins, and the resin pockets classified as Type 1 or 2 ( Figure 3). XY positions of the resin pockets were recorded, the annual rings identified, and the within-ring position of the resin pockets estimated as a percentage of the ring width.

Sawing and grading of boards
The pruned butt logs were sawn into 40 mm thick random-width boards, using a sawing-for-grade cutting pattern ( Figure 4). Boards were visually graded in the green rough-sawn condition, using the WWPA Factory grading rules (WWPA 1991), and assigned to the grades: clears, mouldings, No. 1 shop, No. 2 shop, No. 3 shop, and commons. Boards were assessed for: width, length, actual grade (including resin features), adjusted grade (excluding resin features), number of Type 1 and Type 2 resin pockets, blemish rating (0 -3), presence of resin streaks, resinous latewood, and galls on the worst longgrain face and edge combined.

Descriptor
Parengarenga forest   The frequencies of Type 1 and 2 resin pockets in the boards were calculated per square metre (m²) of surface area, for the boards in each log, forest and ERF class. Volume-weighted average blemish rating of the boards was calculated for each log, forest and ERF class. The number of boards with blemish ratings (0 -3), resin streaks, resinous latewood, and galls were calculated as a percentage of the total volume of boards in each log, forest and ERF class.

Grade recovery
Actual and adjusted grade recoveries of clears, mouldings, No. 1 shop, No. 2 shop, No. 3 shop, and commons, were calculated as a percentage of the total volume of boards in each log, forest and ERF class.

Lumber value
Lumber value of the actual grade and adjusted grade was calculated for each board, based on its volume and grade using the prices (NZ$): clears and moulding $780/m 3 , No. 1 shop $490/m 3 , No. 2 shop $380/m 3 , No. 3 shop $290/m 3 , commons $155/m 3 . Lumber value was summed for all the boards in each log, forest and ERF class. Value per cubic metre ($/m³) was calculated by dividing the total value of the lumber by the volume. Value loss of the resin pockets and blemishes was the difference between the value of actual and adjusted grades, expressed as a percentage of adjusted grade. Average 2001 prices for 5/4 random width lumber at the mill door in New Zealand were used to provide a comparison with McConchie and Turner (2002).

Statistical analyses
Log variables of Parengarenga and Tangoio Forest were compared using one-way analysis of variance (ANOVA) with Tukey's multiple comparison test, using Genstat statistical software (VSN International Ltd, Hemel Hempstead, UK) and the model: where: y ij denotes the log variable measured on the butt log of tree j in forest i; µ is the overall population mean; f i represents the effect of forest (fixed), t j(i) the effect of trees (random), and e ij represents the error term for the log measurements.  Correlations between log and board variables were calculated using SAS statistical software (SAS Institute, Inc., Cary, NC, USA) and the procedure PROC CORR.
Stepwise regression equations for the actual grade recovery of clears + mouldings, actual grade value, and lumber value loss, were fitted using log variables as potential independent variables. SAS procedure PROC REG was used for this analysis. A p-level of 0.05 was used for selecting variables to enter the regression, or for removing variables.

Resin canal properties
The size and frequency of the resin canals in the outer ten growth-rings at breast height (1.4 m) differed for Parengarenga Forest, Northland, and Tangoio Forest, Hawke's Bay (Table 3). Resin canals of trees in Parengarenga Forest were on average larger in diameter, of higher frequency per unit area, and of slightly lower brightness, compared with trees in Tangoio Forest.

Log properties
Volume of the logs, and numbers of Type 1 and 2 resin pockets on ends of the logs were similar at Parengarenga and Tangoio Forest. Acoustic velocity of the logs was higher at Parengarenga Forest, and number of resin blemishes on ends of the logs was higher at Tangoio Forest (Table 4).
Acoustic velocity was higher on average for low ERF class logs, compared with moderate and severe ERF class logs (Table 4). The trend was more evident for the logs at Parengarenga Forest, but there was a weak but significant correlation between acoustic velocity and ERF class for logs from both forests ( Table 5).
Number of Type 2 resin pockets and blemishes on the ends of the logs increased with the severity of the ERF class of the logs (Table 4). There were moderate correlations between the number of Type 2 resin pockets and blemishes on ends of the logs, and ERF class of the logs (Table 5).

Board properties
Presence of resin features on the sawn surface of the random-width boards was strongly influenced by the ERF class of the logs ( Table 6). Numbers of Type 1 and 2 resin pockets, blemish rating, and percentage of boards with resin streaks, and galls, increased with severity of the ERF class of logs. There were weak to moderate correlations between these resin features on the surfaces of boards, and ERF classes of the logs ( Table 7).
The number of Type 2 resin pockets and blemishes on the ends of the logs showed weaker correlations with resin features on the surface of the boards, compared with ERF class of the logs (Table 7). The number of Type 1 resin pockets on the ends of the logs, showed a stronger correlation with number of Type 1 resin pockets on the surfaces of boards, compared with ERF class of the logs. Resin canal properties of breast height increment cores was only very weakly correlated with resin features on surfaces of the boards (Table 7).

Grade recovery
Actual grade recoveries of random-width boards showed a substantial reduction in the volume of clears and moulding grades, and a corresponding increase in   Adjusted grade recovery of the random-width boards showed that if resin features on the surfaces of boards were excluded, the low, moderate, and severe ERF class logs had similar board grade recoveries (Table 8). Recovery of clears grade boards was however, much higher for logs from Parengarenga Forest, compared with those from Tangoio Forest. This was explained by larger defect cores and pruned branches in logs from Tangoio Forest, which had the effect of downgrading many boards from Tangoio    Comparison of actual and adjusted grade recoveries for the low ERF class logs, showed that when resin features were excluded the recovery of clears grade boards doubled, recovery of mouldings grade boards decreased, and overall, recovery of clears + mouldings grade boards increased by 10 and 6 percent of total board volume for Parengarenga and Tangoio Forest, respectively. For moderate ERF class logs recoveries of clears + mouldings grade boards increased by 44 and 21 percent of the total board volume for Parengarenga and Tangoio Forest, with the increase occurring in clears grade boards. For severe ERF class logs the recoveries of clears + mouldings grade boards increased by 66 and 56 percent of the total board volume for Parengarenga and Tangoio Forest, with increases in both the clears and moulding grade boards.

Lumber value
Resin features on the surface of the random-width boards reduced total lumber values by 38 percent for Parengarenga Forest logs, and 31 percent for Tangoio Forest logs. Actual grade values of the lumber were similar for Parengarenga and Tangoio Forest, but adjusted grade values (resin features excluded) were lower for Tangoio Forest, due to the larger defect cores and pruned branches in the Tangoio Forest logs (Table 9).
Moderate and severe ERF class logs were effective in segregating for lumber value (Table 9). There was greater loss of lumber value with increased severity of the log ERF class. Losses of lumber value for the low, moderate, and severe ERF class logs were 9, 42, and 57 percent at Parengarenga Forest, and 6, 28 and 57 percent at Tangoio Forest.

Prediction of grade recovery and lumber value
Linear regression models gave good predictions of actual grade recoveries of clears + moulding grade boards, actual grade value, and loss of lumber value (Table 10).
Predictions of actual grade recovery, and loss of lumber value, were weaker using numbers of resin pockets and blemishes on the ends of the logs, compared with ERF classes of logs. Resin canal properties of the logs showed no significant correlations with actual grade recovery, or loss of lumber value. Multiple regression models gave improved grade recovery and value predictions using log ERF class, volume, heartwood content, and number of Type 1 resin pockets on ends of logs as independent variables (Table 11). ERF class of log was the main contributor to the models, with the other log variables providing only small reductions in residual sums of squares. The exclusion of ERF class of log from the models resulted in poor predictions, with the number of Type 1 and 2 resin pockets and blemishes on the ends of the logs proving to be relatively ineffective predictive variables in the absence of ERF class.

Discussion
Weaker relationships between resin features on the surface of the boards and those on ends of the logs could be attributed to the small number of the resin pockets and blemishes that were sampled on the log ends. Type 1 resin pockets rarely extend more than 50-100 mm longitudinally, and while Type 2 resin pockets can extend for more than 200-250 mm (Ottenschlaeger et al. 2012), the distances are short compared with 5 m lengths of pruned butt logs. Resin bleeding and occlusion scars of Type 2 resin pockets, and lesions and galls that accumulate on the surface of the bark during the life of a tree, provide good indications of the severity of the internal resin defects that are present in the wood (McConchie 2003;McConchie et al. 2007;Ottenschlaeger et al. 2012).
Losses of lumber value for the moderate and severe ERF class logs from Parengarenga and Tangoio Forest (Table 9) were greater than the 14 and 34 percent losses for the moderate and severe ERF class logs from Tikitere Forest in the Bay of Plenty (McConchie 2003). Logs from Tikitere Forest had a lower number of resin pockets on log ends, compared with logs from Parengarenga and Tangoio Forests. Hughes (2007) found an increase in loss of lumber value with increasing numbers of resin pockets on log ends, with logs from Nelson forests having similar value losses to those of logs from Parengarenga and Tangoio Forests, when compared with the same numbers of resin pockets on the log ends. This suggests that resin pockets on the log ends could be useful indicators of the relative loss of lumber value among radiata pine plantation forests and stands.   Higher incidences of Type 1 resin pockets on surfaces of boards of the moderate and severe ERF class logs from Tangoio Forest (Table 6), could be attributed to frequent occurrences of false growth rings at this site ( Figure 1). False growth rings caused by drought conditions have been associated with presence of Type 1 resin pockets in Canterbury (Cown, 1973) and are considered zones of weakness that split tangentially along the false growth ring latewood/earlywood boundary. They are caused by axial compressive stresses (Bariska & Kučera, 1985), that occur with bending of tree stems, and this has been shown to increase the incidence of Type 1 resin pockets (Watt et al. 2009;Jones et al. 2013).

ERF class Actual grade value ($/m³) Adjusted grade value ($/m³) Lumber value loss (%) Parengarenga forest
There was a relationship between Type 2 resin pockets on the log ends and blemish rating and resin streaks on the surface of the boards (Table 7). Ottenschlaeger et al. (2012) observed that the association between Type 2 resin pockets and blemishes, and resin streaks, occurs as a result of the centripetal transport of resin from the Type 2 resin pockets towards the pith, and is a wound response to cambial damage at the time of the resin pocket formation. The blemish rating and percentage of boards with resin streaks increased for the moderate and severe ERF logs from Parengarenga and Tangoio Forest, which suggests that the severity of the cambial damage by the Type 2 resin pockets increased in these logs.
Incidences of resin blemishes on the ends of the logs were strongly associated with Type 2 resin pockets (Table 4), with both occurring in larger numbers on the ends of the logs from Tangioio Forest (Table 5). This suggests there was greater resin flow in response to cambial damage at this site. Resin flow can increase with vertical resin canals of larger diameter and higher frequency (Blanche et al. 1992;Hood & Sala, 2015), and in response to drought conditions (Blanche et al. 1992;Lombardero et al. 2000;Zas et al. 2020;Rissanen et al. 2021). Constitutive resin enrichment of the resin canals and resin flow are greater when drought conditions limit growth (Lombardero et al. 2000;Rissanen et al. 2021), as occurred with the formation of false growth rings at Tangoio Forest. Resin enrichment of the resin canals appeared to be a stronger driver of resin blemish formation than the larger diameter and frequency of the resin canals at Parengarenga Forest. This suggests that the environmental and genetic factors which increase the size and frequency of the resin canals (Reid & Watson 1965;Rigling et al. 2003;Govina et al. 2021;Rissanen et al. 2021), may not increase the constitutive resin flow, and could explain the lack of relationships between the resin canal properties and resin pockets and blemishes.

Conclusions
External resin feature (ERF) class of the pruned butt logs provided an effective method of segregating logs for differences in the number of Type 1 and 2 resin pockets and blemishes on the surface of the boards, the actual grade recovery, and the loss of lumber value associated with the resin features.
Resin pockets and blemishes on the ends of the logs were less effective at predicting the actual grade recovery and lumber value, due to the small number of the resin features on the log ends, compared with the resin bleeding, lesions and galls assessed by ERF classes on the surface of the bark.
Resin canal properties of logs did not improve predictions of actual grade recoveries and the losses of lumber value associated with resin features, but inclusion of the log heartwood content, the number of Type 1 resin pockets on log ends, and log volume, improved predictions to a small extent when combined with the ERF classes of the logs.

Log properties
Actual grade recovery of clears + mouldings