Mari-Jose Barriola · José-Ramón Aira ·Edgar Lafuente

Abstract Larch wood is structurally classif ied in many countries as one of conifers with the highest load-bearing capacity (strength class of C30). The Spanish visual classif ication regulation only assigns a strength class to 4 pine woods: Laricio pine ( Pinus nigra Arn. var. Salzmannii), Silvestre pine ( Pinus sylvestris L.), Radiata pine( Pinus radiata D. Don), and Pinaster pine ( Pinus pinaster Ait.). This work adds to the number of structurally characterised species by creating a visual classif ication table for Japanese larch wood ( Larix kaempferi (Lamb.) Carr.)which differentiates between 2 visual classes, MEG-1 and MEG-2. Characteristic strength values were calculated for each class ( f k,MEG-1 = 31.80 MPa, f k,MEG-2 = 24.55 MPa),mean module of elasticity ( E 0,mean,MEG-1 = 13,082 MPA, E 0,mean,MEG-2 = 12,320 MPA) and density( ρ k,MEG-1 = 456.6 kg m -3 , ρ k,MEG-2 = 469.1 kg m -3 ), before f inally assigning a strength class of C30 to visual class MEG-1, and a strength class of C24 to visual class MEG-2.

Keywords Japanese l arch w ood · Visual grading ·Strength c lass · Mechanical pr operties · Density

Introduction

Each country has developed national standards for structurally classifying wood from its forest stands, establishing its own structural classif ication and nomenclature. Due to the great variety of wood species, origins and different classif ication rules, it is recommended to use the European structural timber grading system. This system is set out by the harmonised standards EN 14,081-1:2016, EN 14,081-2:2018, EN 14,081-3:2012 +A1:2018, and supporting standards. Rectangular cross-section timber is classif ied into categories (strength classes) according to three key properties:strength, stiffness and density. The strength classes and their properties are shown in standard EN 338:2016.

There are two systems for structural timber grading, visual grading and machine grading. Visual grading consists of measuring the singularities of wood, which affect its strength and stiffness, such as knots, slope of grain, width of growth rings, pith, cracks and so on. The number of classes into which one can grade is predef ined by the national standards.The valid transformation from visual grades to the corresponding strength classes are listed separately for softwoods and hardwoods, per species and source of timber, in the standard EN 1912:2012. In machine grading each specimen is evaluated by non-destructive methods in a mechanical way. A machine measures one or more physical—mechanical properties of the piece to establish its “indicating properties” (IP).

Both methods are suitable, but machine grading has the advantages that the IP obtained are more accurate predictors of wood quality than those obtained by visual grading,grading process is much faster, possible human errors are minimized, and can be used for grading into higher grades(Ridley-Ellis et al. 2016; Fortuna et al. 2018). In Spain, it is necessary to carry out numerous research works both to increase the number of timber species classif ied by visual grading and to adapt and calibrate mechanical grading machines to Spanish forest species.

Japanese larch wood (Larix kaempferi(Lamb.) Carr.)is an excellent construction material, as is shown by its use in a large number of signif icant buildings in Japan(Yamabiko Dome in Matsumoto, The Nagano ‘M Wave’arena in Nagano, Mokuzai Nakagai Kaikan office building in Osaka, or Mizu-no-Machiya Nanokamachi-Gotenzeki in Yamagata) (Dauksta 2015). However, it should be considered that Japanese larch quality depends on the location as is the case with most timber species.

According to the standard EN 1912:2012, Larch wood is structurally classified in several countries (Germany,Austria, the Czech Republic, the Scandinavian countries, the Netherlands, France, Canada, the United Kingdom and Italy. European larch (Larix decidua) often attains a strength class of C30 (Germany, Austria, the Czech Republic, the Scandinavian countries, the Netherlands) which places it to the fore in terms of conifer wood strength. However, Japanese larch is only structurally classified in the United Kingdom where it obtains a maximum strength class of C24, as is the European larch in this country.

In Spain, only 4 species of conifers are assigned to a strength class: Laricio pine (Pinus nigraArn. var.Salzmannii) that reaches a maximum strength class of C30,Silvestre pine (Pinus sylvestrisL.) with a maximum strength class of C27, Radiata pine (Pinus radiataD.Don) and Pinaster pine (Pinus pinasterAit.); both species with a maximum strength class of C24. The inclusion of this European standard requires previous visual classif ication by following the guidelines of Spanish standard UNE 56,544:2011. Nevertheless, this Spanish standard only covers the above-mentioned conifer species.

It is important to keep in mind that for some years now UK larch stands have been progressively attacked by the pathogenPhytophora ramorum. As a sanitary measure to prevent its spread, the affected masses have been cut down,which is leading to a considerable increase in the availability of this wood in sawmills. It is estimated that 3.5 million m 3 of Japanese larch will enter the market in the decade 2010-2020 from forests in Wales and Marche (Dauksta 2011). Much of this wood is used for energy and chipboard production, but it is important to consider the possible use in the structural timber market, which is a short-term competition for wood of the same species in Spain. It is also necessary to prevent and control the appearance of this pathogen in Spanish forest stands.

This work, therefore, meets the need to know the mechanical properties and to obtain the visual classif ication of Japanese larch from Spain, to make it suitable for the market.This will allow it to compete in equal terms with other timber species that have already been characterised and other widely used ecological building materials.

Materials a nd methods

The o rigins o f the materials

Standard EN 384:2016 states that test samples must be representative of the population to be analysed. Wood must be representative of the actual origin, dimensions and qualities that will be classif ied in manufacturing.

The forest stands of Japanese larch in Spain are found in the regions of Basque Country and Navarra. The usual cross-sections were considered when selecting the sample,together with the sawing technique (horizontal bandsaw with pneumatic carriage) and the maximum and minimum sizes of structural Japanese larch wood usually employed in the industry (widths from 7 to 20 cm and heights from 15 to 25 cm). Therefore, the sample was analyzed beforehand using these criteria, also considering that a single batch could not contain pieces of different cross-sections, and that pieces may exist with the same cross-section but from different origins, although they must share characteristics.

A total of 370 wooden specimens in 4 batches of 92, 92,90 and 96 were analysed, so that it was possible to obtain two visual classes, and therefore two strength classes. With this sample each visual class would be composed of a subsample of 4 batches that would have to contain at least 40 pieces each, once rejected pieces had been excluded. As there are 4 batches the standard EN 384:2016 sets an adjustment factork nwith a value of 0.97 to obtain the MOE and density, and 0.95 to obtain the MOR. Origin of the batches had to be representative of structural timber industry of this species at national level. Five sampling areas were chosen, 3 in Gipuzkoa (Basque Country region), and 2 in Navarra. The cross-sections subjected to the classif ication criteria during fabrication and the selected sawmills represent approximately 95% of the production of structural timber using this species. Plantations were chosen with the minimum requisite of having groups of adult trees with a normal diameter of about 30 cm, this having been set by the timber industry itself for using trunks. The characteristics and origins of the pieces are shown in Table 1.

In all the samples indicated in Table 1, singularities of the wood were measured and then a four-point bending test was carried out to obtain the mechanical properties (MOR,MOEG and MOE). Finally, a slice close to the breaking zone was extracted to obtain the density and moisture content.

Table 1 Characteristics and origins of the test pieces

Japanese larch has traditionally been used in Spain in upland reforestations (＞ 500-600 m in altitude), mainly among private landowners. It constitutes trees with undergrowth of great pasture interest and provides considerable growth in height, even in cold areas and soils of marked poverty.

Altitude and predominant exposure of the sampling areas are as follows: Ataun (750 m, south and west), Aldaola(850 m, south), Bera (500 m, west), Lesaka (700 m, north and north-west), Otzaurte (700 m, east). There areas are in a mountain climate (or subalpine climate), with average low temperatures (around 10 °C to 12 °C), particularly very cold in winter, with generous rainfall (more than 1500 mm per year, and in some areas more than 2,000 mm) that take the form of snow above 1500 m between autumn and spring.

Test pie ce c onditioning

In order to reduce moisture content (from 80-90 to 30%), the specimens were stored outdoors in stacks for 2-3 months.The stacks consisted of several layers of pieces between which 3 cm × 3 cm wooden strips were placed perpendicular to the pieces to allow horizontal air circulation. In the same layer, the pieces were separated from each other by about 3 cm to facilitate vertical air circulation. The stacks were oriented according to the prevailing winds so that they blew perpendicularly to them. The f irst layer of specimens was placed 50 cm above the ground to prevent wet and cold air from accumulating at the bottom of the stacks. The f loor on which the stacks were placed was clear of wood residues,bushes, grass, etc., to prevent attack by fungi and insects.Specimen moisture content was measured regularly by means of a xylohygrometer with a measuring depth of 4 cm. When moisture content was close to 30%, the samples were conditioned to a moisture content of approximately 12%. This process was carried out in accordance with EN 408:2010+A1:2012 with a drying chamber under controlled temperature and humidity conditions. As larger specimens took longer to reduce their moisture content naturally, so the order of conditioning was f irst small cross-sections, then medium cross-sections, and f inally large cross-sections.

The me asurement of s ingularities

Measurement of wood peculiarities was carried out following the guidelines of standard EN 1309-3:2018. The average width of growth rings was obtained according to the Spanish standard (UNE 56,544:2011) which uses only the 5 inner rings instead of all rings of the cross section.

Location of singularities along the piece was recorded,as pieces can be foreseen to break in the central third where moment distribution is constant, as well as in the end thirds.The singularities found in the central third were used to obtain the visual classif ication tool.

Knots were differentiated between those detected on the faces of the pieces and those on their edges. Knots detected in the central third were measured for their absolute and relative sizes, i.e., the ratio between knot size and the surface they were on, as this is the most widely used methodology in current classif ication standards. This not only takes knot size into account, as it also shows how much of a crosssection it affects, which is more important than the size of the knot itself.

The following singularities were measured: face knots,edge knots, annual rings, cracks, tangential shakes, ingrown bark, fibre deviation, wanes, pith, biological disorders,deformations and machining damage.

The obt aining of me chanical pr operties, humidity and density

Mechanical properties were obtained by bending tests according to EN 408:2010+A1:2012. Modulus of rupture(MOR), overall elastic modulus (MOEG) and modulus of elasticity (MOE) values were obtained by the Eqs. 1- 3.

whereais the distance between the load point and the closest support in mm;F max, the maximum load in N;W, the section strength modulus in mm 3 ;l, the span of piece in mm;F2-F1, the load increase from 10 to 40% of the initial lineal behaviour in N;b, the width of piece cross-section or the smallest dimension of piece cross-section in mm;h,the height of piece cross-section or the largest dimension of piece cross-section in mm;w2-w1, the deformation increase corresponding toF2-F1load increase in mm; andI, the moment of inertia in mm 4 .

After mechanical testing a defect-free slice of 5 cm thickness was cut close to the breaking point, in order to obtain the density and moisture content of specimen. The volume and wet weight of the slice were measured. Subsequently,the slice was introduced into an oven at 100 °C, measuring the weight every 6 h until the difference of 2 consecutive weighings was less than 0.1%. The last weighing was considered as the dry weight of the slice.

The characteristic values are obtained for a certain moisture content of wood, called reference moisture content(12% for most species), which is set out in the standard EN 14,081-3:2012 +A1:2018. The reference moisture content corresponds to the moisture content of wood under ambient humidity conditions of 65% and a temperature of 20 °C.The standard also indicates a number of corrections to be made when the test pieces are outside the reference moisture content.

Moisture content corrections were applied to MOE and density. When the moisture content of specimen was higher than the reference moisture content (12%), MOE was increased by 1% for each 1% increase in the moisture content of specimen relative to the reference moisture content;and was reduced by the same proportion when the moisture content was lower. Specimens that contained more than 18%moisture content were considered to contain 18%. When the moisture content of specimen was higher than the reference moisture content, density was reduced by 0.5% for each 1%increase in the moisture content of specimen relative to the reference moisture content; and increased by the same proportion when the moisture content was lower.

Regarding MOR, it is not necessary to consider any moisture content correction, albeit a correction for size is applied.In this way, the MOR values of specimens with a width less than 150 mm (and density less than or equal to 700 kg m -3 )were corrected to a 150 mm reference width. The MOR was divided by thek hfactor (which varies between 1.3 and(150/h) 0,2 , wherehis the width in mm).

The 5% percentile value of MOR (f 05,i) and density (ρ 05,i),as well as the mean MOE value (ē i), of each batch (subsamplei), were obtained according to standard EN 14,358:2016.To obtain the mean value and standard deviation of MOR,this standard stipulates different equations depending on whether the data f it better with a normal or a log-normal distribution. However, for the mean value and standard deviation of MOE and density the standard directly stipulates the equations corresponding to a normal distribution. The values obtained are then used to calculate the characteristic value of MOR (f k), the mean characteristic MOE (E 0,mean) and the characteristic density (ρ k) by considering the guidelines and equations of standard EN 384:2016.

Finally, according to standard EN 338:2016, a population is assigned a strength class when the characteristic strength values (f k), mean modulus of elasticity (E 0,mean) and density(ρ k) are equal to or greater than those indicated in the standard for the said strength class.

Statistical anal ysis

Frequency histograms of mechanical and density variables as well as the most important singularities were plotted(MOR. MOEG, MOE, density, Max knot/thickness, Max knot/height, Slope of grain, Width 5 internal growth rings).

The results were analysed to detect whether any statistically signif icant differences existed within the characterised variables (MOR, MOE and density) depending on crosssection or origin. Box and whisker plots were drawn for this purpose, and an ANOVA table was created for each variable which breaks down data variance into two component parts: one “between groups” component and one “within groups” component. The f-ratio is the coefficient between the “between groups” estimate and the “within groups”estimate, while thePvalue of theftest indicates that when its value is less than 0.05 there is a statistically signif icant difference between the averages of all of the variables, at a 5% level of signif icance. Multiple Range tests were f inally used to determine which variables were signif icantly different from others.

Results and di scussion

Frequency h istogram

Figure 1 shows the frequency histograms for the MOR,MOEG, MOE and density variables of all specimens. A Chisquared normalcy test was applied to all of them to verify that they follow normal distribution.

Fig. 1 Frequency histograms of mechanical and density variables

Fig. 2 Frequency histogram of the most important singularities

The bending behaviour of Japanese larch is excellent and generally surpasses that of many other conifers used as structural woods. 85% of the samples analysed had a strength(MOR) higher than 30 MPa, while 35% were higher than 50 MPa and values of up to 72 MPa were detected. In terms of rigidity (MOE), 90% of the pieces surpassed 9 MPa, 75%surpassed 10.5 MPa and values of 21.8 MPa were detected.Respecting density, this conifer wood is especially dense as 90% of the pieces were denser than 500 kg m -3 , reaching values of 752 kg m-3.

Figure 2 shows the frequency histogram of the most important singularities of visually graded specimens. Knots were the main defect in the wood, and a high number of grouped knots were detected on the edges of the pieces(51% of pieces) as well as on their faces (85%). This shows that this species is strongly whorled, with a high number of branches per whorl. About 90% of the pieces had less than 10% f ibre deviation, this being an important indicator of wood structural quality. Regarding the thickness of the f irst f ive annual rings, 85% of the pieces showed growth greater than 2 cm/ring, which indicates strong growth during the f irst years.

Fig. 3 Rupture caused by an isolate face knot

In the mechanical bending test, isolated or grouped edge knots were the main cause of rupture, in 41.6% of the pieces.80% of cases were caused by edge knots, 10% were due to knots on both the face and edge, and 10% were caused by face knots (Fig. 3). The second cause of rupture was f ibre deviation, at 10.3% of total ruptures (Fig. 4). It should be pointed out that 25% of the pieces broke cleanly, without any singularity favouring the start of the rupture, and this indicates the excellent structural quality of this type of wood.

Fig. 4 Rupture caused by f ibre deviation

The infl uence of cross-section and origin on the MOR,MOE and density

Figures 5 and 6 show box and whisker plots according to cross-section and to origin, respectively.

Regarding the MOR, there are signif icant differences between the 4 cross-sections tested and two homogeneous groups appear, one composed of the 14 × 18 and 20 × 25 cross-sections and the other of the 7 × 15, 10 × 15 and 14 × 18 cross-sections. The box and whisker plot shows how in spite of the fact that all of the cross-sections have similar dimensions, the smaller cross-section pieces have a far wider range as well as greater symmetry in the Q2 and Q3 quartiles respecting the mean than is the case with the larger cross-section pieces.

MOE values are more homogeneous and there are no signif icant differences between the 4 cross-sections. As the f-ratio is less than one (0.71), this indicates that the variance “between groups” is even less than the variance“within groups”. Only a few outlier values appear in the 7 × 15, 10 × 15 and 14 × 18 cross-sections, and these are not very relevant.

Signif icant differences do arise for density between the 4 cross-sections, where 2 homogeneous groups emerge: the smaller cross-sections on the one hand of 7 × 15 and 10 × 15,and on the other the larger cross-sections of 14 × 18 and 20 × 25, which have higher densities than the previous group.

Respecting the MOR, signif icant differences emerged between the 5 origins tested. Nevertheless, two quite different homogeneous groups appear: one is composed of the pieces originating in Aldaola, Bera, Lesaka and Otzaurde,between which there were no signif icant differences. The other group solely consists of pieces from Ataun, and they have higher strength values than the rest.

Fig. 5 Box and whisker plots according to cross-section

Fig. 6 Box and whisker plots according to origin

MOE values also show signif icant differences between the 5 sources. As the f-ratio is slightly higher than one (2.78),this shows that the difference is not very great. Two homogeneous groups emerge, one composed of pieces sourced in Aldaola, Lesaka, Otzaurde and Ataun; while the other is composed of those sourced in Bera and Ataun, which have slightly greater stiffness values.

In terms of density there are clear signif icant differences between the 5 sources, and 3 different homogeneous groups emerge. The f irst group would be composed solely of pieces from Aldaola, the second would contain only those from Ataun, and the third group would include those from Lesaka,Otzaurde and Bera. The pieces in this third group have the highest density values, while those in the f irst group have the lowest values.

Analysis shows that statistically signif icant differences exist between the different cross-sections and sources.Nevertheless, it also shows reveals that there is no statistical evidence that any single cross-section or origin clearly differs from the others in the 3 variables that were characterised(MOR, MOE and density).

Visual c lassif ication

Two visual classes were established for Japanese larch structural wood. They are known as MEG-1 and MEG-2, and their criteria are shown in Table 2. To determine the criteria for cracks and maximum deformations the limitations set in standard EN 14,081-1:2016 were considered.

The resulting classif ication of pieces in each batch as MEG-1, MEG-2 or rejected is shown in Table 3.

The established visual grading criteria forced the rejection of 14 of the 370 samples analysed, equivalent to a percentage of 3.8%. The 14 specimens were rejected for thefollowing reasons: Diameter of edge knots ＞ 3/5 of "b" (3 specimens of 15 × 7 cross section, and 2 of 15 × 10 cross section), Diameter of face knots ＞ 1/2 of "h" (3 specimens of 15 × 10 cross section, 1 of 18 × 14 cross section, and 2 of 25 × 20 cross section), Fibre deviation ＞ 1:6 (1 specimen of 15 × 7 cross section, and 1 specimen of 25 × 20 cross section), Width of 5 inner growth rings ＞ 11 mm/ring (1 specimen of 25 × 20 cross section). Such a small percentage of rejected pieces makes it possible to consider that the yield from this classif ication is very high and also highly suitable for usual manufacturing conditions. The 48.4% of the pieces were classif ied as MEG-1 and 47.8% as MEG-2.

Table 2 Specif ications for strength classes MEG-1 and MEG-2

Table 3 Pieces of each cross-section classif ied as MEG-1, MEG-2 or rejected

Table 4 Characteristic MOR, MOE and density values of visual class MEG-1

The characteristic values of MOR, MOE and density were determined for each batch in the samples classif ied as MEG-1 and MEG-2. The results of this determination are shown in Tables 4 and 5. To analyse whether the strength data correspond better with a normal or a log-normal distribution, each batch was subjected to a Kolmogorov—Smirnov goodness-of-f it test. Batches 2 and 4 of the test pieces assigned to MEG-2 were found to f it better with a log-normal distribution, as was the case in other studies undertaken usingLarix gmeliniifrom North Eastern China(Luo et al. 2012).

Once the strength, stiffness and density values had been obtained (f k,E 0,meanandρ k) for each visual class, the lowest one of them determines the strength class to which a population can be assigned (EN 338:2016). Thus, a strength class of C30 would be assigned to MEG-1 and a strength class of C24 would be assigned to MEG-2.

When the established classif ication criteria are compared with French standard NFB 52-001-1:2018 for European larch (Larix decidua), a priori it may be said that the limitations on the size and arrangement of knots, which are the singularity that most affects the structural quality of wood(Bunetti et al. 2016; Moriguchi et al. 2016; Jung-Kwon et al.2010; Takeda and Hashizume 1999), are similar although not the same in visual classes ST-I and MEG-1, and in classes ST-II and MEG-2. These similarities also arise in the assignation of strength classes, as ST-I is equivalent to C27 and ST-II is equivalent to C24 (EN 1912:2012).

The Italian classif ication standard UNI 11,035-2:2010 sets a visual class of S2 and higher (grouping classes S1 and S2) for European larch. This is comparable to MEG-2, which corresponds to a strength class of C22 (EN 1912:2012).This standard in general has a low classif ication yield and it undervalues the strength of this species according to a study using wood from the Italian and French Alps. Using mechanical classif ication this study obtained strength assignations of C30/C24/C18/R (Bunetti et al. 2016).

Table 5 Characteristic MOR,MOE and density values of visual class MEG-2

In the best cases the English classif ication standard BS 4978:2007+A2:2017 establishes a visual class of SS for European and Japanese larch, corresponding to a strength class of C24 (EN 1912:2012). It has to be pointed out that the English standard uses a different knot-measuring system from the one used in the European standard (EN 1309-3:2018), which enormously hinders objective comparison of it with other systems.

Respecting the Spanish classif ication standard (UNE 56,544:2011), the visual classif ication criteria for MEG class established for different species of thePinusgenus largely f it well with the limitations set in the MEG-2 visual class for Japanese larch. Face knots are the exception to this, where the limitation is set at 1/2 instead of 2/3, while for edge knots the limitation is set at 3/5 instead of 2/3.

In some investigations carried out on Japanese larch specimens from different provenances in Canada, density values between 409 and 463 kg m-3were obtained (Cáceres et al. 2017), whose higher values correspond to the values obtained in this research work. Other research carried out on Japanese larch specimens from South Korea to evaluate their orthotropic properties by digital image correlation, showed MOE values of 11,700 MPa (Jeong and Park 2016) which are close to those obtained in this work.

With regard to the width of the growth rings, investigations carried out on Japanese larch specimens from Japan average values were obtained for the entire cross-section ranging from 2.03 mm/ring to 3.28 mm/ring (Zhu et al.2000). This work obtained average value of 5-6 mm/ring,which is logically higher because it corresponds only to the average width of the 5 inner rings, and not to the entire cross-section.

Conclusion

This research work concludes with the obtaining of a visual grading table for Japanese structural larch wood (L. kaempferi(Lamb.) Carr.) from Spain, giving 2 different visual classes denominated MEG-1 and MEG-2.

The characteristic strength values were determined for each one of the visual classes (f k,MEG-1= 31.80 MPa ,f k,MEG-2= 24.55 MPa), together with the mean modulus of elasticity (E 0,mean,MEG-1= 13,082 MPA,E 0,mean,MEG-2= 12,320 MPA) and density (ρ k,MEG-1= 456.6 kg m -3 ,ρ k,MEG-2= 469.1 kg m -3 ). This f inally made it possible to assign a strength class of C30 to visual class MEG-1, and a strength class of C24 to visual class MEG-2.

AcknowledgementsWe thanks to Basque centre of research and applied innovation in vet (TKNIKA), Centre for services and promotion of Castilla y León forestry and its industry (CESEFOR), D.Bixente Dorronsoro, Gipuzkoa provincial council, and Commercial services of the wood of Guipuzkoa (SECOMA). Larra?aga sawmill(Azpeitia).