^{1}

^{1}

^{2}

^{1}

^{*}

This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution license (http://creativecommons.org/licenses/by/3.0/).

This study investigates how the use of a Hitman ST300 acoustic sensor can help identify the best forest stands to be used as supply sources for the production of Machine Stress-Rated (MSR) lumber. Using two piezoelectric sensors, the ST300 measures the velocity of a mechanical wave induced in a standing tree. Measurements were made on 333 black spruce (^{2} = 0.41). Results suggest that, at a regional level, 92% of the black spruce trees meet the requirements of MSR grade 1650Fb-1.5E, whilst 64% and 34% meet the 2100Fb-1.8E and 2400Fb-2.0E, respectively. Mature stands with a TSF < 150 years had 11 and 18% more boards in the latter two categories, respectively, and therefore represented the best supply source for MSR lumber.

The boreal forest zone of northeastern North America is a major source of softwood lumber for both domestic and export markets [

The black spruce (

For these reasons, forest managers are increasingly recognizing the importance of resource quality characterization earlier along the value chain. Ideally this information will be used to inform planning decisions so that raw material can be allocated to appropriate markets according to its predicted end-use properties [

Recently, new acoustic sensors have been developed for the rapid nondestructive evaluation (NDE) of standing trees and logs. These instruments are used to estimate MOE through its relationship with stress-wave velocity or acoustic resonance [

The aim of this study was to use standing tree assessments of wood stiffness to predict structural grade outturn in black spruce trees. Specific objectives were as follows: (1) to develop a linear regression equation linking mean tree MOE with stem diameter and acoustic velocity; (2) to use logistic regression to predict the proportion of boards that meet the requirements of certain MSR grades as a function of the predicted tree-level MOE, and (3) to combine steps 1 and 2 using inventory data to predict the MSR potential of the black spruce resource at the regional scale. The effect of stand structure on MSR grade potential was also tested by sampling from stands at different post-fire successional stages.

The Hitman ST300 (Fibre-gen, Christchurch, New Zealand) is a portable device designed to measure the velocity of mechanical stress-waves in standing trees [^{−1}), _{D}^{−2}) and ^{−3}). Since the tool does not provide a measure of green density, it is usually assumed to be constant for a given species and time of year, to account for seasonal fluctuations in moisture content [_{D}

The tool contains two Monitran MTN/P100 accelerometers, each attached to a probe inserted into the lower part of the stem at a 45-degree angle and aligned vertically between 50 and 120 cm apart (

The accelerometer in the lower probe detects the stress-wave induced by a hammer blow, while the second accelerometer records the arrival time of the stress wave. The transit time

The sample trees were located over a large area (20,000 km^{2}) in the North Shore region of Quebec, Canada (

Since moisture content affects density and therefore the speed of propagation of a stress-wave in wood, acoustic measurements were taken on the sample trees around 1.3 m from ground level with the probes inserted into the stem at two different depths, namely 1.5 cm and 3 cm, which conform to previously published data on sapwood and heartwood thickness [

The sampling locations along the chronosequence were selected on the basis of accessibility to the existing road network. In addition, every effort was made to sample trees from distinct recorded fire events. At each location, a random distance (50 m to 200 m from the road) and a random azimuth were chosen to determine the exact location of the plot. Variable-radius plots were established using a prism of factor 2 [

The 39 harvested logs were sawn using a portable sawmill into 1.8 m-long boards with nominal dimensions of 38 mm × 89 mm in cross section, after drying and planing. The sawn pieces were stored at 20 °C and 65% RH for five months, until they reached an equilibrium moisture content of approximately 12%. Mechanical properties were assessed in accordance with ASTM D4761 [^{−2}) was calculated using the stress values recorded between 10% and 45% of the maximum load at failure for each piece, thus ensuring the elastic limit was not reached. Values were adjusted to a moisture content of 15% following ASTM D1990 [

First, the averaged MOE data were used to develop a multivariate linear regression model describing mean tree MOE as a function of the acoustic velocity squared and tree diameter. An interaction term between these variables was also included in the model, and the root mean square error (RMSE) and mean percentage error (E%) calculated from the observed and predicted values. The explanatory variables were centered on their mean values prior to model-fitting to reduce the effects of multicollinearity and increase the interpretability of the model coefficients [_{est}, kN·mm^{−2}) for each tree in the larger NDE dataset. Next, analysis of variance (ANOVA) was carried out to determine if there were significant differences in mean acoustic velocity and mean MOE_{est} between the TSF classes. Multiple comparisons were made using Tukey's honestly significant difference (HSD) tests, which identify any significant differences between groups.

Since each board was assigned an MSR grade based on its static MOE value, logistic regression was used to model the proportion of boards meeting the requirements of certain MSR classes as a function of the predicted tree-level stiffness values (MOE_{est}). Logistic regression can model dichotomous outcomes as binary variables that are coded as either 1 or 0, with the proportion of 1s representing the probability (between 0 and 1) of the event of interest occurring [

Subsequently, the information from the linear and logistic regression equations was used to estimate the proportion of sawn pieces that could be expected to meet a given MSR grade for each TSF class. This was achieved in a 3-stage process. First, the measurements made in the variable-radius plot were expressed on a per hectare basis. Since the plot radius varied with the DBH of the stems, observations had to be weighted using a correction factor of 80,000/(π·DBH^{2}) [_{est} values overlaid with normal density curves confirmed that the data in each TSF class could be assumed to be normally distributed (_{est} to be calculated for each TSF class. The relative frequency distribution function of a variable _{1} + _{2} × _{est}_{1} and _{2} for each MSR grade, modelled as a function of MOE_{est}. Thirdly, predicted pass rates for each MSR grade in each TSF class were calculated as the integral of the product of Equations

All analyses were performed using functions contained in the R statistical programming environment (R Development Core Team, 2012).

Acoustic velocities at probe depths of 1.5 and 3 cm were highly correlated (^{2} = 0.91) and thus appeared to have a similar potential for MOE predictions. However, they were consistently higher at a depth of 3 cm than at 1.5 cm, with mean values of 4.39 and 4.32, respectively (

Mean tree static MOE was positively related to the acoustic velocity squared and negatively to tree diameter. There was also an interaction between the velocity squared and diameter terms. The equation for MOE_{est} was given by:

Because the explanatory variables were centered on their mean values, the model intercept of 10.680 is the estimated MOE when both explanatory variables are set to their mean values. The coefficient of determination (^{2}) of the relationship between the observed and predicted MOE was 0.41, which is lower than the values of 0.65 and 0.55 previously reported by Mora ^{2} values of 0.44 and 0.45, respectively [^{−2} and 0.01%, respectively). The overall fit of the model was hence judged sufficient to pursue our analyses.

The interaction term in our model indicates that the effect of acoustic velocity on MOE is mediated by stem diameter. This may occur due to the anisotropic nature and heterogeneous structure of wood [

When the wave is generated by the impact of the hammer, it first spreads through the stem at an angle of 45 degrees (dilatational wave). However, the wave front soon evolves into a quasi-plane wave propagating in the longitudinal direction of the stem [

Values of acoustic velocity, DBH and MOE_{est} in the NDE dataset (n = 333) showed some variation according to TSF classes (_{est}, values in the two youngest TSF classes were significantly higher than in stands with a TSF greater than 200 years. The only significant differences in mean DBH were between stands in the 50–99 year TSF class and all the older classes. The results suggest that even-aged, uniform stands initiated by a forest fire may produce wood with higher structural performance than stands in which the last fire occurred more than 200 years ago. Further investigations are underway that will provide more data to elucidate the influence of TSF on the structural wood properties of this important resource, particularly considering the high proportion (>60%) of stands with a TSF greater than 200 years in our study area [

The parameters of the logistic regression equations for the proportion of boards meeting the requirements of each MSR grade are presented in _{est} did not significantly influence the proportion of boards meeting the design specifications of the 1650Fb-1.5E grade, although it was a significant predictor of the two higher grades.

The stand-level predictions of the percentage of the resource meeting the minimum requirements of each selected MSR grade, calculated using

An illustrative example of the methodology used to calculate pass rates for MSR grade 2400Fb-2.0E for the combined TSF classes is shown in

The propagation speed of mechanical waves is fairly well correlated with wood stiffness and, in combination with tree diameter, can be used to make unbiased predictions of tree-level MOE, despite the moderate fit of the initial regression model of static MOE as a function of acoustic velocity and tree diameter (^{2}

Results from our case study confirm that black spruce stands in the North Shore region of Quebec have the potential to produce lumber that meets the requirements for MSR grade 1650Fb-1.5E at a high pass rate, regardless of stand structure. However, we noticed differences between stands of different TSF classes for the MSR grades with higher structural requirements. The results show that the wood from forests with a TSF < 150 years would generally have better mechanical properties than wood from forests with a TSF > 150 years, up to 11% and 18% more for MSR grades 2100Fb-1.8E and 2400Fb-2.0E, respectively. Further work is being conducted to identify the biological causes of this difference. A first hypothesis being tested is that wood properties decline as a result of gradual changes in site and stand characteristics in the absence of fire [

The information obtained from acoustic sensors can help inform stand selection decisions according to the current demand for specific MSR lumber grades. While it will be important to increase our understanding of the effect of TSF in our sampling area, the method presented in this study is applicable to other forest types where different factors may affect wood properties. Standing-tree acoustic sensors such as the ST-300 could be used as part of conventional forest surveys to obtain low-cost assessments of wood properties at the regional scale. For such uses, the strength of the relationship between the acoustic velocity measured on a given tree and the MOE of lumber pieces it produces should not be used as the main indicator of standing-tree tool performance. Instead, future efforts should focus on calibrating this relationship at the population level,

The authors would like to thank Jean-Sébastien Perron where his help was greatly appreciated in the field. Our thanks to Josée Dugal who made illustrations showing the various sensors used in the Hitman ST300 and Philippe Goulet (NSERC-University Laval industrial research chair in silviculture and wildlife) for making the maps using ArcGIS software. Funding was provided by ForValueNet, an NSERC strategic network on forest management for added-value products.

Schematic diagram of the ST300 operating principle. (

Location of sample plots in the North Shore region of Quebec, Canada.

Normal distribution curves overlaid on histograms of MOE_{est} for each TSF class. The dashed line on each graph represents the median and the black line represents the mean.

Relationship between standing tree acoustic velocity measured at horizontal depths of 1.5 and 3.0 cm in the stem (R^{2} = 0.91).

(_{est} assigned to a tree using the ST-300. (_{est} values. (

Historical fire map from the North Shore region (data obtained from [

Mean and standard deviation of acoustic velocity (tree level) and MOE_{est} (15% MC) for each TSF class (n = 333). Identical letter codes indicates no significant difference in means between TSF classes in Tukey HSD tests (α = 0.05).

^{−1}) |
_{est}(kN·mm^{−2}) |
|||||||||
---|---|---|---|---|---|---|---|---|---|---|

50–99 | 75 | 4.48 | 0.35 | B | 11.16 | 1.61 | A | 15.36 | 2.73 | A |

100–149 | 79 | 4.66 | 0.35 | A | 11.12 | 1.45 | A | 19.65 | 5.16 | B |

150–199 | 79 | 4.40 | 0.43 | B | 10.61 | 1.39 | AB | 20.53 | 5.22 | B |

>200 | 100 | 4.42 | 0.35 | B | 10.55 | 1.05 | B | 20.77 | 5.26 | B |

Parameter estimates (θ_{1} and θ_{2}, _{est} at 15% MC).

_{est} | ||
---|---|---|

1650Fb-1.5E | 2.2442 | - |

2100Fb-1.8E | −9.6062 | 0.9413 |

2400Fb-2.0E | −31.2750 | 2.6765 |

Percentage of the resource meeting the minimum requirements of certain MSR grades, grouped by TSF (time since last fire) class.

50–99 | 92.0 | 67.3 | 41.8 |

100–149 | 92.0 | 71.1 | 44.0 |

150–199 | 92.0 | 57.7 | 28.3 |

>200 | 92.0 | 57.9 | 20.4 |