- Email: [email protected]
Predicting softwood quality attributes from climate data in interior British Columbia, Canada
Forest Ecology and Management 361 (2016) 81–89
Contents lists available at ScienceDirect
Forest Ecology and Management journal homepage: www.elsevier.com/locate/foreco
Predicting softwood quality attributes from climate data in interior British Columbia, Canada Lisa J. Wood a,⇑, Dan J. Smith a, Ian D. Hartley b a b
University of Victoria Tree-Ring Laboratory, Department of Geography, University of Victoria, Victoria, British Columbia V8W 3R4, Canada University of Northern British Columbia, Ecosystem Science and Management Program, 3333 University Way, Prince George, British Columbia V2N 4Z9, Canada
a r t i c l e
i n f o
Article history: Received 18 June 2015 Received in revised form 26 October 2015 Accepted 1 November 2015
Keywords: Wood quality Wood density Cell wall thickness Microfibril angle Climate Dendroclimatology
a b s t r a c t Ongoing and future climate changes are expected to result in fundamental shifts in forest productivity and wood quality over wide regions of interior British Columbia. This study was conducted to investigate the relationships between climate and wood property attributes within Douglas-fir and spruce forests, and to use those relationships to develop a non-invasive approach wood quality attribute prediction. Historical climate station data was correlated to tree-ring samples collected at five sites and climate–tree growth relationships were established to measure and predict wood density, cell-wall thickness, and microfibril angle attributes. Time series models were developed to reconstruct the measured wood properties, and a strong correspondence between the predicted and measured wood attributes was verified. The results confirm that climate parameters provide a useful index for assigning wood quality attributes to forest stands at many sites in BC’s interior. The findings of this research provide insight into the impact future climates may have on wood quality characteristics and could be applied elsewhere to investigate the impacts of climate change on forest fibre supplies. Ó 2015 Elsevier B.V. All rights reserved.
1. Introduction and background Over the next century, mean annual air temperatures in British Columbia (BC) are expected to increase from 1 to 4 °C, and be accompanied by precipitation events of increased intensity and severity (IPCC, 2013). These climate changes are certain to have an impact on BC’s forests (Kozlowski, 1979; Larsen, 1993; Wang et al., 2006) and are expected to initiate long-term alterations in cellular wood structure that could result in fundamental shifts in productivity and wood quality over wide regions (Williamson et al., 2009; Adams, 2014). To accommodate the accompanying social and economic impacts of these shifts, BC needs to develop informed stand management plans that take into account any climatically-induced wood quality changes in future forests (Spittlehouse, 2007). The term wood quality is often used in discussions concerning the contributions of cell growth and maturation to the radial development of a tree. Higher wood densities are most often equated with wood structure and properties of superior quality for the production and manufacturing of wood products (Haygreen and Bowyer, 1996). For example, wood with long tracheids and lower
⇑ Corresponding author. E-mail address: [email protected] (L.J. Wood). http://dx.doi.org/10.1016/j.foreco.2015.11.004 0378-1127/Ó 2015 Elsevier B.V. All rights reserved.
lignin content is preferred for producing most paper types (Taylor et al., 1982). Likewise, high-density wood with a high percentage of latewood in each annual ring, or a high cell wall-to-cell lumen diameter ratio, is often desired by solid wood and pulp and paper manufacturers because of its increased strength properties and high fibre yield (Haygreen and Bowyer, 1996). Fundamentally, every anatomical characteristic of a wood fibre, including the tracheid diameter, the lumen diameter, the cell-wall thickness, and the microfibril angle (MFA), controls the mechanical property of wood and the quality of any fibre-related products (Burdon et al., 2004; Vahey et al., 2007). Functional properties of wood products, such as axial stiffness and longitudinal shrinkage, are characteristics that rely heavily upon MFA in the S2 layer of the secondary cell wall (Ansell, 2011). As MFA increases, longitudinal shrinkage increases exponentially and stiffness in the axial direction decreases. Cave (1968) notes that cell-wall stiffness increases five-fold with a decrease from 40° to 10° in mean MFA. Microfibril angles greater than 35° lead to shrinkage as high as 5% of the total board length. For solid wood of good quality, it is important then that MFAs are as low as possible with respect to the longitudinal direction. Low stiffness leads to strength issues and high shrinkage leads to warping of timber, which are both serious wood quality issues (Barnett and Jeronimidis, 2003). Climate conditions throughout the growing season have a large influence on the development of radial cells and related wood
82
L.J. Wood et al. / Forest Ecology and Management 361 (2016) 81–89
anatomical characteristics in temperature-limited coniferous trees (Tardif et al., 2001; Deslauriers et al., 2007). Annual tree ring growth begins with the production of large, thin-walled earlywood cells (Creber and Chaloner, 1984; Barnett and Jeronimidis, 2003). As environmental conditions change through the summer months, the cells transition to small thick-walled latewood with decreasing lumen size (Brown et al., 1949; Wort, 1962; Creber and Chaloner, 1984; Barnett and Jeronimidis, 2003). Consequently, where temperature is the primary limiting factor of growth (D’Arrigo et al., 1992; Davi et al., 2002; Watson and Luckman, 2002; Savva et al., 2010), developing an understanding of conditions at the time of ring formation may provide the insight needed to assess specific wood quality attributes in a specific area (Fonti et al., 2010). Climate variables such as air temperature and precipitation directly affect energy accumulation within a tree by influencing net photosynthesis, and thus influence the resources available for construction of tracheids and cell walls. During periods of low net photosynthesis, radial growth slows significantly and the cells produced are smaller and/or thinner-walled. For example, Tardif et al. (2001) reported on the radial growth of seven boreal tree species in response to increases in soil and air temperature. Periods of substantial and continuous radial diameter were positively associated with precipitation and soil temperature, and negatively associated with the photoperiod (Tardif et al., 2001). Because of the existing relationship between climate and cell formation, intraspecific tree-ring and tracheid variability can be used to determine variability in climate and ecosystem conditions (Fonti et al., 2010). By identifying trends in the relationship between climate parameters and wood anatomical characteristics over time, it should be possible to predict how climate variables will effect wood production in specific locations, and therefore certain wood quality attributes produced at that site in the future. These insights would be valuable for forest managers as they design management and rotation strategies tuned to future climates (Williamson et al., 2009; MFLNRO, 2013; IPCC, 2013). The objectives of this study were to investigate the relationships between climate and wood anatomical development, and to use the relationships identified to predict certain wood quality attributes in interior BC forest stands. Time series were developed to predict wood properties based on correlations to historical climate station data, and the predicted values were compared to measured wood attributes. For the purposes of this study, we focus solely on cellular wood quality attributes including density, cell wall thickness, and MFA. The findings of the research provide preliminary insight into whether changing climates are likely to have a positive or negative impact on wood quality attributes within future forests in interior BC.
2. Methods Tree-ring samples were collected from five sites: four sites in northern BC and one site in southern BC (Fig. 1). All of the sites contain mature, open canopy, mixed-species forests dominated by softwoods. Sites were located on well-drained, nutrient-rich sites found in close proximity to long-term climate stations. High-elevation or northern latitude sites were identified to ensure that climate was the primary factor limiting tree growth (Fritts, 1976). Samples were collected from three hybrid Engelmann spruce (Picea glauca (Moench) Voss x engelmannii (Parry)) stands in the Smithers area (sites A (NSx1), B (NSx2), and C (NSx3)) and from Douglas-fir (Psuedotsuga menziesii (Mirb.) Franco) trees located at sites near Babine Lake (site D (NDf4)) and Pemberton (site E (SDf7)) (Fig. 1). Two 5.2 mm cores from opposite and one 12 mm core were collected from each tree at breast height (Stokes and Smiley, 1968).
The 5.2 mm cores were allowed to air dry, glued to a grooved mounting board, and were sanded until the annual ring boundaries were clearly visible (Stokes and Smiley, 1968). Mounted cores were scanned with a high resolution Epson XL1000 flatbed scanner to create digital images, which were then used to measure the width of each annual ring to 0.001 mm using WindendroÒ software (Version 2006). Annual rings that were exceptionally narrow or unclear were measured to 0.001 mm using a VelmexÒ tree-ring measurement system equipped with a trinocular boom-mounted microscope and CCD video display. Following air drying, each 12 mm core was prepared for densitometric analysis by gluing it flush to the surface of a 2.5 cm-wide fibreboard block. Once dry, a 2 mm thick wood lath was cut (pith to bark) with a Waltech high precision twin-bladed saw to reveal the radial surface of the core (Haygreen and Bowyer, 1996). Resins add to wood’s structural mass and must be removed prior to wood density measurement (Lenz et al., 1976). In this instance, wood resins were chemically extracted using an acetone Soxhlet apparatus (Jensen, 2007). Acetone was placed in a round-bottomed flask resting on a hot plate, and the wood lathes were placed in the Soxhlet chamber sealed by a condenser cycling cold tap water. Acetone was cycled over the samples for five hours by consecutively evaporating and condensing the liquid to remove the resins (Schweingruber et al., 1978; Grabner et al., 2005). The samples were air-dried after resin extraction and mounted vertically on pins in an ITRAX scanning densitometer. To ensure accurate light attenuation by the laser scanner, care was taken to orient samples perpendicular to the X-ray beam. Scanned images were measured using WindendroÒ ITRAX density software (Version 2008b). Sectioned and extracted 12 mm 2 mm samples were sent to Dr. R. Evans at the Australian Commonwealth Scientific and Research Organization (CSIRO) for SilviScan analysis. The SilviScan system allowed for precise measurement of density, microfibril angle and tracheid radial and tangential diameters, and also calculates cell-wall thickness (Eq. (1)) (Vahey et al., 2007; Lundgren, 2004; Jones et al., 2005). The 12 mm radial width of the core was trimmed to a precise 7 mm prior to scanning with SilviScan. Various wood property measurements were obtained as treering series: total ring width (RW); maximum, mean, and minimum density (MXD, MD, MND); maximum, mean, and minimum microfibril angle (XMFA, MMFA, NMFA); maximum, mean, and minimum cell radial and tangential diameters (XRD, MRD, NRD and XTD, MTD, NTD); and maximum, mean, and minimum fibre coarseness (XC, MC, NC). Data for maximum, mean, and minimum cell-wall thickness (XCWT, MCWT, NCWT) were calculated by SilviScan as a function of the measured density (d), coarseness (C), and cell perimeter distance (P) (Eq. (1)); cell perimeter (P) is calculated using radial vs. tangential cell diameter dimensions (Eq. (2)) (Vahey et al., 2007; Lundgren, 2004).
CWT ¼ P=8 1=2ðP=16 C=dÞ
1=2
ð1Þ
where
P ¼ 2ðR þ TÞ
ð2Þ
and, R and T are the radial and tangential tracheid diameters, respectively (Jones et al., 2005). 2.1. Chronology development Each tree-ring series was cross-dated with respect to characteristic annual ring width patterns (Stokes and Smiley, 1968). The series cross-dating was verified using COFECHA (Holmes, 1983), and annually resolved chronologies were developed for each species and wood property identified (Table 1). Changes made to the time series during the ring width cross-dating process, such as the
L.J. Wood et al. / Forest Ecology and Management 361 (2016) 81–89
83
Fig. 1. Sampling sites (triangles): site A and B (Hudson Bay Mountain), site C (Blunt Forest Service Road), site D (Gullwing Forest Service Road), site E (Owl Creek Forest Service Road). Climate stations (circles) with longest records in closest proximity to sampling sites: Smithers, Fort St. James, Agassiz.
deletion of a false ring or adjustment of dates due to broken core segments, were applied over all of the measured wood property chronologies from the same samples to ensure consistency. The time-series created in COFECHA were transformed to ARSTAN master chronologies to eliminate non-climatic variation (Cook and Holmes, 1986). A negative exponential curve was applied to series showing age-related growth trends (Fritts, 1976). Following this, a smoothing spline was applied to the RW chronologies, with a frequency–response cut-off set to 67% to remove non-climate impacts (Cook and Kairiukstis, 1990). Previous research has shown growth trends due to factors such as inter-tree competition are not pronounced in densitometric tree-ring chronologies (Conkey, 1986). Consequently, only a first-order detrending was carried out on density and fibre property series, using negative exponential or straight line fits. To eliminate autocorrelation issues, only the pre-whitened residual ARSTAN chronologies were used. 2.2. Modelling The standardized and detrended master chronologies were compared to the climate data from the station closest to the sampling location (Fig. 1). Instrumental climate data were obtained from the Adjusted Historical Canadian Climate Data website (http://www.cccma.ec.gc.ca/hccd/) for the Smithers (station #1077500, Lat 54°82’N, Long 127°18’W, 522 m asl), Fort St. James
(station #1092970, Lat 54°45’N, Long 124°25’W, 686 m asl), and Agassiz climate stations (station #1100120, Lat 49°25’N, Long121°77’W, 15 m asl) (Fig. 1). The average summer temperatures (1895–present) recorded at the Agassiz climate station ranges between 20 and 27 °C, while those at the Fort St. James climate station (1895–present) range between 9 and 17 °C. Initial climate–tree growth relationships were established with a response function in PRECON version 5.17B (Fritts, 1999). Correlations were identified between temperature and precipitation data from the nearest climate stations, and indexed chronologies for the most recent half of the available climate data (1950 to present). Where significant relationships existed regressions were carried out to hind-cast wood property measurements using climate data as a proxy for wood properties. Wood property data from 1950 to present were used as the dependent variable in the model, and climate data from as early as possible (1938 or 1895) was used as the independent variable. Temperature and precipitation were investigated in their relationships with the wood attributes presented here, but only models which used temperature to predict the various wood parameters were considered to be reliable for presentation. Wood property models were restricted in length to expressed population signal (EPS) values of P0.85 with only one decade of this time permitted to drop to an EPS value of P0.80 (Wigley et al., 1984). Once successful models of wood properties were made, they were verified using correlation to the known wood measurements for the last 100 years, and using split
84
L.J. Wood et al. / Forest Ecology and Management 361 (2016) 81–89
Table 1 Master chronology statistics for BC wood quality sample collection sites. Sx = Engelmann white spruce hybrids, Fd = Interior Douglas-fir. RW = ring width, MXD = maximum density, MD = mean density, MND = minimum density, XCWT = maximum cell wall thickness, MCWT = mean cell wall thickness, NMFA = minimum microfibril angle, MMFA = mean microfibril angle, XMFA = maximum microfibril angle, NRD = minimum radial diameter. Chronology type
Species
Site name
Number of cores (trees)
Chronology length
Interseries correlation
Mean sensitivity
Expressed population signal P 0.80
RW RW RW RW RW MXD MD MND MXD MD MND MXD MD MND MXD XCWT XCWT XCWT MCWT NMFA NMFA NMFA MMFA XMFA NRD
Sx Sx Sx Fd Fd Sx Sx Sx Sx Sx Sx Fd Fd Fd Fd Sx Fd Fd Fd Sx Fd Fd Fd Fd Fd
A B C D E A A A B B B D D D E A&B D E E A&B D E E E E
23 (16) 17 (13) 19 (17) 18 (14) 18 (15) 19 (13) 20 (13) 16 (11) 19 (14) 20 (14) 18 (13) 7 (7) 17 (11) 14 (8) 8 (7) 19 (18) 7 (7) 13 (12) 11 (11) 13 (12) 8 (8) 8 (8) 10 (10) 10 (10) 8 (8)
1598–2006 1680–2006 1726–2006 1754–2006 1672–2006 1610–2006 1670–2006 1670–2006 1716–2006 1716–2006 1716–2006 1810–2006 1810–2006 1810–2006 1638–2006 1686–2006 1810–2006 1688–2006 1688–2006 1710–2006 1810–2006 1690–2006 1705–2006 1705–2006 1689–2006
0.448 0.493 0.507 0.500 0.555 0.438 0.465 0.352 0.633 0.536 0.401 0.499 0.461 0.354 0.406 0.473 0.474 0.503 0.443 0.364 0.453 0.377 0.401 0.374 0.330
0.165 0.195 0.169 0.205 0.216 0.074 0.054 0.073 0.106 0.068 0.081 0.076 0.069 0.077 0.114 0.109 0.085 0.083 0.067 0.074 0.102 0.094 0.065 0.102 0.056
1786–2006 1805–2006 1853–2006 1855–2006 1780–2006 1791–2006 None None 1764–2006 1764–2006 1814–2006 1912–2006 1912–2006 1970–2006 None 1782–2006 1912–2006 1787–2006 1787–2006 1870–1920 1940–2006 1775–1850 1940–2006 None None
verification, where the latter 50% of the time series was used to calibrate the models. To ensure that the observed correlations were true and not caused by autocorrelation, the Durbin–Watson statistic was used; values of 2.0 were considered to have no autocorrelation, and values equal or less than 1.0 were deemed positively autocorrelated. 3. Results 3.1. Chronologies In total, 20 chronologies were created: five acceptable RW chronologies; ten density chronologies of which seven were acceptable; four cell wall thickness chronologies of which three were used; five microfibril angle chronologies; and, one radial diameter chronology that did not meet the acceptable standards for EPS (Table 1). Only chronologies displaying reliable EPS values (Wigley et al., 1984) and with minimal autocorrelation according to the Durbin–Watson statistic are discussed. 3.2. Sites A and B The site B MXD and site A and B XCWT chronologies positively correlated to mean May, July, and August temperature records from the Smithers station, and the site A and B NMFA chronology was negatively correlated to mean summer temperature (average of June, July, August) from Smithers (Tables 2 and 3). Multivariate regression models predicted MXD, XCWT, and NMFA based on these variables (Fig. 2). The Durbin–Watson statistic (DW) suggested some positive autocorrelation in the MXD and XCWT reconstructions, but not to a level of concern (Table 3). Partial correlation tests were carried out to ensure that the mean monthly temperatures were all separately influencing the predicted variables. Each month remained significantly correlated to MXD, XCWT or NMFA, even when controlling for the others. Fig. 2 shows strong correspondences between the predicted models and the
measured values for Sx MXD and Sx XCWT; these relationships are also reflected in the Pearson’s correlation coefficients (Tables 3 and 4). Fig. 2 illustrates that monthly average summer temperature has a close correspondence to the maximum density of the cells and the thickness of tracheid cell walls at this northern interior site. The reconstructed values approximate the measured data, except for in 1977, a year which shows a much lower measured value than was predicted by either model. Fig. 2 also shows time periods, for example from 1982 to 1986, where the XCWT model underestimated the measured values. During periods when the model did not accurately predict the wood anatomical values, we assumed that another climate or environmental factor primarily influenced the development of these characteristics. Fig. 2 shows the correspondence between the predicted values of Sx NMFA and the measured values obtained from the SilviScan analyses. Proxy mean summer temperature data from the Smithers climate station approximates the minimum microfibril angle of spruce growing in the area, except for certain years. For example, in 1977 the actual measured data shows a high NMFA value not captured by the model. The 1977 data point shown in this model corresponds to the overestimated data point in 1977 in the MXD and XCWT models. While the measured MFA value for 1977 was high, the RW reconstruction in Fig. 2 shows a low measured ring width value for 1977 as was true for the models for XCWT and MXD discussed above. The NMFA reconstruction in Fig. 2 displays data that overestimates the measured values prior to 1950. The distinct appearance of model overestimation prior to 1950 could indicate a non-linear relationship through time, which may be due to a flaw in the data standardization process to remove growth-related trends.
3.3. Site C Ring width from site C was positively correlated with mean June temperature from Smithers (r = 0.476, p = 0.01) and a regres-
85
L.J. Wood et al. / Forest Ecology and Management 361 (2016) 81–89
Table 2 Pearson’s correlation coefficients illustrating significant relationships between the chronologies presented and climate station data. Where data is not reported, no significant relationship was identified. Note that chronologies from sites A, B, and C were compared to Smithers climate station data, chronologies from site D were compared to Fort St. James climate station data, and chronologies from site E were compared to Agassiz climate station data. NMFA = minimum microfibril angle, XCWT = maximum cell wall thickness, RW = ring width, MXD = maximum density, MMFA = mean microfibril angle, NRD = minimum radial diameter.
Avg May temp Avg June temp Avg July temp Avg August temp Avg summer temp Avg spring temp Total spring precip Total July precip Total August precip
Site A and B NMFA
Site A and B XCWT
Site B RW
Site B MXD
Site C RW
Site D RW
Site D MXD
Site D XCWT
Site E MMFA
Site E NRD
Site E XCWT
0.386 0.320 0.468 0.538
0.447 0.345 0.401 0.436
– 0.542 0.356 –
0.419 0.311 0.452 0.544
– 0.476 0.430 –
– – – –
0.343 0.413 0.548 0.302
0.350 0.453 0.516 0.342
0.289 – 0.369 0.311
– – 0.601 0.441
0.318 0.497 0.486 0.626
0.645
0.580
0.517
0.634
0.527
–
0.586
0.612
0.351
0.585
0.733
– –
0.456 –
– –
0.324 –
– –
– 0.466
– 0.331
– 0.308
0.476 –
– –
0.538 –
– –
– –
– –
– –
– –
– –
0.422 0.336
0.373 0.395
– –
– –
– –
Table 3 Statistics describing proxy wood quality reconstructions for specific sites in interior BC (⁄ = p < 0.05). NMFA = minimum microfibril angle, XCWT = maximum cell wall thickness, RW = ring width, MXD = maximum density, NRD = minimum radial diameter. Reconstruction
Interior spruce MXD
Interior spruce XCWT
Interior spruce NMFA
Interior spruce RW
Douglas-fir MXD
Douglas-fir XCWT
Douglas-fir XCWT
Douglas-fir MMFA
Region
North-Smithers
North-Smithers
North-Smithers
Chronology collection site Proxies
B
A and B
A and B
NorthSmithers C
North-Babine Lake D
North-Babine Lake D
SouthPemberton E
SouthPemberton E
May, July, August temp Smithers 0.691 0.477 1.036 0.411⁄ 0.261⁄
May, July, August temp Smithers 0.620 0.385 1.118 0.325⁄ 0.244⁄
Mean summer temp Smithers 0.645 0.416 1.476 0.451⁄ 0.044
June temp
Mean summer temp Fort St. James 0.586 0.343 1.359 0.331⁄ 0.614
Mean summer temp Fort St. James 0.612 0.374 1.336 0.364⁄ 0.573
June, July, August temp Agassiz 0.761 0.579 1.590 0.523⁄ 0.997
Mean spring temp Agassiz -0.476 0.227 1.170 0.151⁄ 0.137
Climate station Pearson’s R R2 Durbin–Watson RE calibration RE verification
Table 4 Correlation between measured wood properties averaged over a chronology and the reconstructed wood property based on climate variables.
Site Site Site Site Site Site Site Site Site
B MXD A and B XCWT A and B NMFA C RW D XCWT D MXD D RW E XCWT E NRD
Measured vs. reconstructed (Pearson’s R)
P value
0.613 0.546 0.500 0.368 0.334 0.284 0.237 0.418 0.343
0.001 0.001 0.001 0.002 0.001 0.003 0.013 0.001 0.001
sion model was created to predict RW. The r2 value for this model was 0.226 (p 6 0.001). DW was found to be 1.275, suggesting only slight positive autocorrelation (Fig. 2). Mean June temperature also under-predicted RWs of spruce from Site C after 2000. 3.4. Site D The XCWT and MXD chronologies from site D were significantly correlated to mean summer temperature from the Fort St. James climate station (Table 3; Fig. 3). Fig. 3 shows Douglas-fir XCWT and MXD, reconstructed from mean summer temperatures in Fort St. James. The models in Fig. 3 under-predict XCWT and MXD from 1937 to 1954, indicating that a factor other than the summer temperature variable may have influenced the cell wall deposition process during this time period.
Smithers 0.476 0.226 1.275 0.266⁄ 0.056
3.5. Site E At site E, the XCWT chronology was correlated to June, July, and August temperatures from the Agassiz climate station. These three temperature variables were independently used in a multivariate regression model to predict Douglas-fir (Fd) XCWT (Table 3). Fig. 4 shows that the reconstructed XCWT of Fd from site E is a close representation of the actual measured XCWT. This finding indicates that Agassiz June, July, and August temperatures, used independently in a multivariate model, provide a robust predictor of XCWT development in Douglas-fir for this region. The only time periods where the reconstruction does not correlate well with the measured record are short durations from 1900 to 1910, and from 1941 to 1947 when the model under-predicts the measured XCWT. The site E MMFA chronology was significantly negatively correlated to average spring (March–May) temperatures from Agassiz climate station. Fd MMFA was then hind-casted based on average spring temperature. The DW value displayed slight positive autocorrelation (Table 3, Fig. 4). 3.6. Model statistics The reconstructed wood properties based on climate proxies correlate back to the original measured wood properties and provide validation of the models (Figs. 2–4). The results of the split verification method used were mixed (Table 4). The predicted site B MXD and site A and B XCWT passed the verification test with reduction of error (RE) values significant at 95% when compared
86
L.J. Wood et al. / Forest Ecology and Management 361 (2016) 81–89
Fig. 4. Measured (dotted lines) vs predicted (solid lines) mean microfibril angle and maximum cell-wall thickness in Douglas-fir from southern interior BC (site E). MMFA modelled from average spring temperature (March, April, May) from Agassiz climate station. XCWT modelled from mean June, mean July, and mean August temperatures from the Agassiz climate station, and again using only July mean temperature from the Agassiz climate station.
Fig. 2. Measured (dotted lines) vs predicted (solid lines) maximum ring density, maximum ring cell-wall thickness, minimum ring microfibril angle, and ring width in interior spruce from northern interior BC. MXD and XCWT modelled from mean May, mean June, and mean August temperatures from the Smithers climate station. NMFA modelled from mean summer temperature (June, July, August average) from the Smithers climate station. RW modelled from mean June temperature from the Smithers climate station. Sites where trees were sampled are listed on the figure.
95% in the verification stage of the test (Table 3). This outcome may be due to a lack of similarity between the first half of the time series and the last half. This was especially notable in the site A and B NMFA reconstructions, where summer temperature from Smithers over-predicted NMFA prior to 1946, and in the site D MXD and XCWT reconstructions where summer temperature from Fort St. James under-predicted the wood property measurements prior to 1954. The site E reconstruction using MMFA also only showed significant correlation with the measured values after 1950. We also suspect that sample depth was responsible for poor verification outcomes, as larger sample sizes generally created more robust chronologies with which to compare to climate variables. For instance the reconstructions from sites A and B, and the XCWT reconstruction from site E were the most reliable, with P17 cores used in the chronologies from sites A and B, and 13 cores were used for the XCWT chronology from site E. The EPS values NMFA from site A and B were low, and only reached a value P0.80 for a short section of the time series (Table 1), indicating that the sample size was not large enough to generate a stable time series curve. A larger sample would reduce the variability among the MFA values for each year, and may eliminate the inaccuracy of the MFA models prior to 1950.
4. Discussion Fig. 3. Measured (dotted lines) vs predicted (solid lines) maximum ring density and maximum ring cell-wall thickness in Douglas-fir from northern interior BC (site D). MXD and XCWT modelled from mean summer temperature (June, July, August average) from the Fort St. James climate station.
to the actual measured data. The remaining reconstructions, site A and B NMFA, site C RW, site D XCWT, site D MXD, site D RW, site E XCWT, site E MMFA, and site E NRD all passed the calibration part of the split verification, but did not display a significant RE value at
4.1. Reconstruction of wood properties The strong relationship between tree growth and climate at the northern interior sample sites permitted successful construction of RW, MXD, XCWT, and NMFA chronologies using measured temperature variables from nearby climate stations. The models (Sx MXD, XCWT, NMFA, RW and FD XCWT and MXD) were not, however, reliable in all years. This outcome was particularly apparent in 1977 when the measured MXD, XCWT, and RW values were lower
L.J. Wood et al. / Forest Ecology and Management 361 (2016) 81–89
than average, and the measured NMFA value was higher than average. Given that mean August temperatures in 1977 were more than two standard deviations above average (Smithers, 1938–present), it may be that the trees in this region were unable to sequester excess carbon in their cell walls resulting in lower overall average density and cell-wall thickness. If this was the case, it may indicate that as August temperatures become warmer, wood density and quality attributes may decrease in northern coniferous trees. It is also possible, however, that reduced ring widths in 1977 were a lagged response to colder-than-average previous-year temperatures or greater than average precipitation (Chavardès et al., 2013). Increased precipitation would allow for cell radial expansion and the earlywood portion of the ring to become wider, thus increasing the width of the annual ring (Wimmer and Grabner, 2000; Tardif et al., 2001). For example, there are several higherthan-average values in the measured RW record prior to 1949 that were not captured in the model, suggesting that June temperature alone is not an effective predictor of spruce RW prior to 1949. Regardless of the source of this error, the inability of temperature variables to capture some of the variation in the wood quality attributes examined indicates that a more complex, multivariate model is required. Future research should identify soil moisture indices and snow pack data to test whether these variables contributed to the deviations observed here. In years of low ring density and narrow cell-wall thickness, it is not uncommon to see high MFA values, which explains the synchronization between models. There are few studies that investigate the relationship between MFA, cell-wall thickness, and density, so how these variables are related is not clear (Hein and Brancheriau, 2011; Xu et al., 2012; Wood and Smith, 2014). Hiller (1964a,b) and Evans et al. (2000) report significant correlations exist between density, cell-wall thickness, and MFA, but other researchers were unable to discern definitive relationships (Bergander et al., 2002; Schimleck and Evans, 2002; Lin and Chiu, 2007). High MFA values lead to high longitudinal shrinkage in dimensional lumber, and are also correlated with low strength and stiffness (Cave, 1968). Both MFA records have measured values lower than were predicted by the reconstructions models prior to 1950 (see Fig. 2). Microfibril angle is a trait that is extremely variable depending on age. Juvenile wood shows much higher fibril angles than mature wood (Haygreen and Bowyer, 1996; Burdon et al., 2004). This agerelated trend should not be reflected in the reconstructions presented because growth-related trends were removed from the series using a negative exponential curve in ARSTAN. It is possible that a variation on the negative exponential curve used for standardizing RW is required for MFA data. The overestimation of NMFA in this study indicates that measured and standardized minimum microfibril angles were lower than predicted by the climate proxy. Similar nonlinear behaviour was reported by Xu et al. (2012) who reported a stronger, more significant relationship between temperature and MFA from 1987 to 2009. The XCWT attribute model generated from the Agassiz climate station for trees sampled near Pemberton under-predicted the measured values from 1900 to 1910. This outcome could be the result of higher-than-average July temperatures in the Agassiz record not reflected in the model average of June, July, and August temperatures. To test this hypothesis, a regression of XCWT for site E was performed using only July temperature from Agassiz (Fig. 4). The statistical significance of this simple linear regression is not as strong as the multivariate model (r = 0.486, r2 = 0.236, p 6 0.001), but the predicted values visually follow the measured data very well for the 1900–1910 period that was previously underpredicted. This observation indicates weather events during a growing season, such as an interval of higher-than-average temperatures, may be very influential on wood anatomical
87
development. In the case of Douglas-fir from site E, July temperatures play a larger-than-normal role in the overall XCWT development from 1900 to 1910 than during the remainder of the time series. Furthermore, it should be noted that at this southern site, extreme July temperatures resulted in higher XCWT and MXD in Douglas-fir, where as in the north, the higher-than-average temperature values in August were observed to have the opposite effect. This differential climate-growth response to high summer temperatures emphasizes that trees with a higher percentage of latewood are usually more tolerant to heat and drought than those with proportionally more earlywood, a response believed related to the structural complexity of latewood cells and their decreased dependence on turgor pressure (Kozlowski, 1979; Barnett and Jeronimidis, 2003; de Luis et al., 2011). The MMFA predictions at the southernmost site (site) were not as strong as the other models developed in this study; however, they follow the long-term trends of the measured values despite the evident high-frequency variation (Fig. 4). MMFA was negatively correlated to mean spring temperatures, which means that when spring temperatures increased, MMFA decreased; warm springs produced conditions where mean microfibril angles were low in Douglas-fir. There is support for MFA–climate relationships (Xu et al., 2012). Xu et al. (2012) found significant correlations between July and August temperatures and spring precipitation and MFA measured in Picea crassifolia sampled from northeastern Tibet. The difference in climate response in this study is likely due to regional environmental differences. While the model presented here over-predicts the measured values prior to 1946 and for a span in the early 1970, it is included here as an example of the range of types of wood properties that can be investigated through the use of climate proxies. 4.2. Influence of climate forcing mechanisms Large regional climate forcing mechanisms, such as are described by the Pacific Decadal Oscillation (PDO), are presumed to be influential in the long-term growth trends of trees found in northwestern North America interior BC (Chavardès et al., 2013; Pitman and Smith, 2013; Rood et al., 2013). The PDO underwent a large-scale regime shift from a ‘‘cool phase” to a ‘‘warm phase’ between 1976 and 1977, and this shift is synchronous a climategrowth anomaly in the northern attribute reconstructions. Cool PDO phases have been correlated with growth-inhibiting climates off the coast of Alaska and growth-enhancing climates in the southwest USA, with warm PDO phases having the opposite influence (Mantua, 2000). Given these relationships, a PDO shift from a cool phase to a warm phase could result in the climate within interior BC transitioning from one that is growth-inhibiting to one that is growth-enhancing. Support for this hypothesis rests with the 1946–1947 PDO transition from a warm to a cool phase (Mantua, 2000), a regime shift that may be responsible for the under prediction NMFA values at sites A and B prior to 1946, and the decoupled of the measured and predicted records at site E (XCWT, MMFA and NRD) at this time. 4.3. Climate, wood development and wood quality The ability to correlate and model wood properties based on a single climate factor, such as temperature, provides forest managers and wood producers with a non-invasive methodology for describing forest stand quality attributes that are not otherwise discernible. While this approach may be effective for elucidating fluctuations in some of the attributes that contribute to wood quality, its application is limited in settings where compounding environmental factors influence tree growth. Trees that demonstrate variable inter- and intra-seasonal relationships to different
88
L.J. Wood et al. / Forest Ecology and Management 361 (2016) 81–89
limiting factors will require more complex models for predicting wood quality attributes. Relationships between certain attributes and climate could be combined with site index measurements for productivity to evaluate the overall impact that changing climates will have on productivity and quality within a forest stand. It is well known that a vast array of environmental and climate parameters interact to produce the ecosystem dynamics within which a tree grows (Kozlowski, 1979); however, accounting for variation in all of these factors requires a complex modelling system. The advantage of the modelling method presented in this study, however, is its ability to derive stand-level wood quality attributes from climate station variables. Through investigation of the general correspondence between climate variables and wood anatomical variables, we draw some preliminary conclusions regarding certain wood quality attributes of the stands studied. Correlation analysis demonstrated that most cell wall thickening, and consequently, density variables for both Douglas-fir and spruce stands in the interior of BC were positively correlated to increases in mean summer temperatures. There were a few instances where it was evident that another variable was more important to radial development, but summer temperature was the primary variable involved in wood formation in these stands over the last 50–100 years. Therefore, we conclude that wood tissue grown when summer temperatures are warmer for longer intervals will be higher in density and lower in MFA, and should demonstrate the greater strength and stiffness properties sought out for solid wood production (Haygreen and Bowyer, 1996; Wimmer and Downes, 2003; Ansell, 2011). A caveat for this general rule may be if summer temperatures reach an extreme threshold, at which time, trees in the north may be ill-adapted to the conditions and produce some attributes of poorer quality wood (lower density and CWT). In globally-changing climates conditions, this caveat may prove to be a very important factor for forest managers to assess. It was demonstrated that early summer temperatures have a proportionally greater influence on ring width development in interior hybrid spruce stands than do late summer temperatures. It could be expected then hybrid spruce forests within interior BC that experience warmer spring temperatures combined with average to cooler summer temperatures will produce wood with lower density and strength properties that is best suited for pulp production. In contrast, hybrid spruce trees growing at sites that experience warmer-than-average spring and summer temperatures are likely to produce the type of uniformly wide, high density rings that are optimal for solid wood products; provided that excessively warm temperatures do not lead to moisture deficits (Kozlowski, 1979; Wimmer and Downes, 2003). Lastly, it was demonstrated that warm spring temperatures in the southern BC interior result in the production of wood cells in Douglas-fir trees with a low mean microfibril angle. This finding indicates that as temperatures warm earlier in the growing season, the wood produced may have a lower MFA and, therefore, higher strength and stiffness properties. Wood with lower MFA is ideal for solid wood production (Jozsa and Middleton, 1994; Ansell, 2011).
5. Conclusion Radial tree growth is influenced by environmental factors that combine to produce variable annual ring increments. Regional climate variables interact with local site conditions to produce individual rates of photosynthesis, and localized adaptations to climate within each tree. This behaviour translates into an allocation of energy for radial growth and wood anatomical development. In this study, average summer temperatures were shown to have the
greatest effect on maximum density, cell-wall thickness and microfibril angle development in the forest stands investigated. This finding was reflected in consistent correlations between these variables, and also between the observed and reconstructed values for MXD, CWT, and MFA based on mean summer temperatures. Since these wood properties are important wood quality predictors, this information is significant. This study aimed to identify whether or not climate can be used as a reasonable indicator of wood properties and certain wood quality attributes. Based on the chronologies created for specific regions of BC, it was shown that distinct wood property characteristics can be modelled by various temperature variables. Obviously many environmental components contribute to the formation and strength of wood, however, targeting a few limiting factors with which to model wood quality within a stand provides valuable insight from a simplified method requiring less investment by individual forest managers. Climate is a major contributing factor to growth and development of wood cells, and therefore using it as an indicator is both valid and useful to the prediction of certain wood quality attributes in future forest stands. Fibre characteristics including density, cell-wall thickness, and microfibril angle were estimated based on local mean summer temperatures in this study. These traits fluctuate through time to create certain wood attributes when warm temperatures prevail, such as increased cell-wall thickness, which is a desirable wood quality attribute for dimensional lumber. Increased strength and stiffness properties from increased cell-wall thickness and decreased microfibril angle can be expected in warming BC climates, but only up to a threshold. Northern trees may react negatively to temperatures over a certain point, whereas southern interior trees may flourish. Soil water availability also becomes an important consideration at high temperatures. Reconstructions, such as those produced here, help to elucidate the years in which climate had a negative impact on certain wood quality attributes; for example, in extreme climate years when microfibril angle is high. This study demonstrates a non-invasive method useful for identifying climate-related wood quality attributes, and has potential for developing an understanding of the changes to wood quality likely to occur within BC’s interior forests as the climate warms over the next century. Acknowledgements The authors thank Leslie Abel, Bethany Couthard, Aquila Flower, Lynn Koehler, and Branden Rishel for their field assistance, and to Kyla Patterson for her data preparation and technical support. Financial support for this research was provided by Natural Science and Engineering Research Council of Canada (NSERC) awards to Wood and Smith, and a Canadian Foundation for Climate and Atmospheric Science (CFCAS) award to the Western Canadian Cryospheric Network (WC2N). References Adams, S.T., 2014. The Impact of Changing Climates on Tree Growth and Wood Quality of Sitka Spruce. Unpublished PhD thesis, University of Glasgow. Ansell, M.P., 2011. Wood – a 45th anniversary review of JMS papers. Part 1: the wood cell wall and mechanical properties. J. Mater. Sci. 46, 7357–7368. Barnett, J.R., Jeronimidis, G., 2003. Wood Quality and Its Biological Basis. CRC Press LLC, Boca Raton, USA. Bergander, A., Brändström, J., Daniel, G., Salmén, L., 2002. Fibril angle variability in earlywood of Norway spruce using soft rot cavities and polarisation confocal microscopy. J. Wood Sci. 48, 255–263. Brown, H.P., Panshin, A.J., Forsaith, C.C., 1949. Textbook of Wood Technology, vol. 1. McGraw-Hill Company Inc., New York, NY. Burdon, R.D., Kibblewhite, P., Walker, J.C.F., Megraw, R.A., Evans, R., Cown, D.J., 2004. Juvenile vs. mature wood: a new concept, orthoganol to corewood versus outerwood, with special reference to Pinus radiate and P. taeda. For. Sci. 50, 399–415.
L.J. Wood et al. / Forest Ecology and Management 361 (2016) 81–89 Cave, I.D., 1968. The ansiotropic elasticity of the plant cell wall. Wood Sci. Technol. 2, 268–278. Chavardès, R.D., Daniels, L.D., Waeber, P.O., Innes, J.L., Nitschke, C.R., 2013. Unstable climate-growth relations for white spruce in southwest Yukon, Canada. Clim. Change 116, 593–611. Creber, G.T., Chaloner, W.G., 1984. Influence of environmental factors on the wood structure of living and fossil trees. Bot. Rev. 50 (4), 357–448. Conkey, L.E., 1986. Red spruce tree-ring widths and densities in eastern North America as indicators of past climate. Quater. Res. 26, 232–243. Cook, E.R., Holmes, R.L., 1986. User’s manual for the program ARSTAN. In Tree-ring chronologies of western North America: California, eastern Oregon, and Northern Great Basin with procedures used in the chronology development work including users manuals for computer programs. COFECHA and ARSTAN. In: Holmes, R.L., Adams, R.K., Fritts, H.C. (Eds.), Chronology Series VI. Laboratory of Tree-ring Research, The University of Arizona, Tucson, pp. 50–65. Cook, E.R., Kairiukstis, L.A. (Eds.), 1990. Methods of Dendrochronology: Applications in the Environmental Sciences. Kluwer, Dordrecht. D’Arrigo, R.D., Jacoby, G.C., Free, R.M., 1992. Tree-ring width and maximum latewood density at the North American tree line: parameters of climatic change. Canad. J. For. Res. 22, 1290–1296. Davi, N., D’Arrigo, R., Jacoby, G., Buckley, B., Kobayashi, O., 2002. Warm-season annual to decadal temperature variability for Hokkaido, Japan, inferred from maximum latewood density (AD 1557–1990) and ring width data (AD 1532– 1990). Clim. Change 52, 210–217. ˇ ufar, K., 2011. Climate de Luis, M., Novak, K., Raventós, J., Gricˇar, J., Prislan, P., C factors promoting intra-annual density fluctuations in Aleppo pine (Pinus halepensis) from semi-arid sites. Dendrochronologia 29, 163–169. Deslauriers, A., Rossi, S., Anfodillo, T., 2007. Dendrometer and intra-annual tree growth: what kind of information can be inferred? Dendrochronologia 25, 113– 124. Evans, R., Stringer, S., Kibblewhite, R.P., 2000. Variation of microfibril angle, density and fibre orientation in twenty-nine Eucalyptus nitens trees. Appita J. 53, 450– 457. Fonti, P., von Arx, G., García-González, I., Eilmann, B., Sass-Klasen, U., Gӓrtner, H., Eckstein, D., 2010. Studying global change through investigation of the plastic responses of xylem anatomy in tree rings. New Phytol 185, 42–53. Fritts, H.C., 1976. Tree Rings and Climate. Academic Press, New York. Fritts, H.C., 1999. PRECON Version 5.17B. A Statistical Model for Analyzing the TreeRing Response to Variations in Climate.. Grabner, M., Wimmer, R., Gierlinger, N., Evans, R., Downes, G., 2005. Heartwood extractives in larch and effects on X-ray densitometry. Canad. J. For. Res. 35, 2781–2786. Haygreen, J.G., Bowyer, J.L., 1996. Forest Products and Wood Science, third ed. Iowa State University Press, Ames, Iowa. Hein, P.R.G., Brancheriau, L., 2011. Radial variation of microfibril angle and wood density and their relationships in 14-year-old Eucalyptus urophylla S.T. Blake wood. Bioresources 6, 3352–3362. Hiller, C.H., 1964a. Correlation of fibril angle with wall thickness of tracheids in summerwood of slash and loblolly pine. Tappi J. 47, 125–128. Hiller, C.H., 1964b. Estimating size of the fibril angle in latewood tracheids of slash pine. J. For. 62, 249–252. Holmes, R.L., 1983. Computer assisted quality control in tree-ring dating and measurement. Tree-Ring Bull. 43, 69–78. IPCC, 2013. Climate Change 2013: the physical science basis. Contribution of working group I to the fifth assessment report of the intergovernmental panel on climate change. In: Stocker, T.F., Qin, D., Plattner, G.-K., Tignor, M., Allen, S.K., Boschung, J., Nauels, A., Xia, Y., Bex, V., Midgley, P.M. (Eds.), Cambridge University Press, Cambridge, United Kingdom and New York, NY, USA, 1535pp. Jensen, W.B., 2007. The origin of the Soxhlet extractor. Chem. Educat. Today 84, 1913–1914. Jones, P.D., Schimleck, L.R., Peter, G.F., Daniels, R.F., Clark, A., 2005. Non-destructive estimation of Pinus taeda L. tracheid morphology characteristics for samples from a wide range of sites in Georgia. Wood Sci. Technol. 39, 529–545. Jozsa, L.A., Middleton, G.R., 1994. A Discussion of Wood Quality Attributes and Their Practical Implications. Forintek Canada Corporation, Vancouver, BC, Canada, 42p. Kozlowski, T.T., 1979. Tree Growth and Environmental Stresses. The University of Washington Press, Seattle, USA.
89
Larsen, JB., 1993. Testing winter desiccation resistance for species and provenance selection at timberline. In: Alden, J., Mastrantonion, J.L., Ødum, S. (Eds.), Forest Development in Cold Climates. Plenum Press, New York, NY, pp. 59–64. Lenz, O., Schar, E., Schweingruber, F.H., 1976. Methodological problems relative to measurement density and width of growth rings by X-ray densitogrames of wood. Holzforschung 30, 114–123. Lin, C.J., Chiu, C.M., 2007. Relationships among selected wood properties of 20-yearold Taiwania (Taiwania cryptomerioides) trees. J. Wood Sci. 53, 61–66. Lundgren, C., 2004. Cell-wall thickness and tangential and radial cell diameter of fertilized and irrigated Norway spruce. Silva Fenn. 38, 95–106. Mantua, N., 2000. The Pacific Decadal Oscillation. Joint Institute for the Study of the Atmosphere and Ocean. University of Washington.
Contents lists available at ScienceDirect
Forest Ecology and Management journal homepage: www.elsevier.com/locate/foreco
Predicting softwood quality attributes from climate data in interior British Columbia, Canada Lisa J. Wood a,⇑, Dan J. Smith a, Ian D. Hartley b a b
University of Victoria Tree-Ring Laboratory, Department of Geography, University of Victoria, Victoria, British Columbia V8W 3R4, Canada University of Northern British Columbia, Ecosystem Science and Management Program, 3333 University Way, Prince George, British Columbia V2N 4Z9, Canada
a r t i c l e
i n f o
Article history: Received 18 June 2015 Received in revised form 26 October 2015 Accepted 1 November 2015
Keywords: Wood quality Wood density Cell wall thickness Microfibril angle Climate Dendroclimatology
a b s t r a c t Ongoing and future climate changes are expected to result in fundamental shifts in forest productivity and wood quality over wide regions of interior British Columbia. This study was conducted to investigate the relationships between climate and wood property attributes within Douglas-fir and spruce forests, and to use those relationships to develop a non-invasive approach wood quality attribute prediction. Historical climate station data was correlated to tree-ring samples collected at five sites and climate–tree growth relationships were established to measure and predict wood density, cell-wall thickness, and microfibril angle attributes. Time series models were developed to reconstruct the measured wood properties, and a strong correspondence between the predicted and measured wood attributes was verified. The results confirm that climate parameters provide a useful index for assigning wood quality attributes to forest stands at many sites in BC’s interior. The findings of this research provide insight into the impact future climates may have on wood quality characteristics and could be applied elsewhere to investigate the impacts of climate change on forest fibre supplies. Ó 2015 Elsevier B.V. All rights reserved.
1. Introduction and background Over the next century, mean annual air temperatures in British Columbia (BC) are expected to increase from 1 to 4 °C, and be accompanied by precipitation events of increased intensity and severity (IPCC, 2013). These climate changes are certain to have an impact on BC’s forests (Kozlowski, 1979; Larsen, 1993; Wang et al., 2006) and are expected to initiate long-term alterations in cellular wood structure that could result in fundamental shifts in productivity and wood quality over wide regions (Williamson et al., 2009; Adams, 2014). To accommodate the accompanying social and economic impacts of these shifts, BC needs to develop informed stand management plans that take into account any climatically-induced wood quality changes in future forests (Spittlehouse, 2007). The term wood quality is often used in discussions concerning the contributions of cell growth and maturation to the radial development of a tree. Higher wood densities are most often equated with wood structure and properties of superior quality for the production and manufacturing of wood products (Haygreen and Bowyer, 1996). For example, wood with long tracheids and lower
⇑ Corresponding author. E-mail address: [email protected] (L.J. Wood). http://dx.doi.org/10.1016/j.foreco.2015.11.004 0378-1127/Ó 2015 Elsevier B.V. All rights reserved.
lignin content is preferred for producing most paper types (Taylor et al., 1982). Likewise, high-density wood with a high percentage of latewood in each annual ring, or a high cell wall-to-cell lumen diameter ratio, is often desired by solid wood and pulp and paper manufacturers because of its increased strength properties and high fibre yield (Haygreen and Bowyer, 1996). Fundamentally, every anatomical characteristic of a wood fibre, including the tracheid diameter, the lumen diameter, the cell-wall thickness, and the microfibril angle (MFA), controls the mechanical property of wood and the quality of any fibre-related products (Burdon et al., 2004; Vahey et al., 2007). Functional properties of wood products, such as axial stiffness and longitudinal shrinkage, are characteristics that rely heavily upon MFA in the S2 layer of the secondary cell wall (Ansell, 2011). As MFA increases, longitudinal shrinkage increases exponentially and stiffness in the axial direction decreases. Cave (1968) notes that cell-wall stiffness increases five-fold with a decrease from 40° to 10° in mean MFA. Microfibril angles greater than 35° lead to shrinkage as high as 5% of the total board length. For solid wood of good quality, it is important then that MFAs are as low as possible with respect to the longitudinal direction. Low stiffness leads to strength issues and high shrinkage leads to warping of timber, which are both serious wood quality issues (Barnett and Jeronimidis, 2003). Climate conditions throughout the growing season have a large influence on the development of radial cells and related wood
82
L.J. Wood et al. / Forest Ecology and Management 361 (2016) 81–89
anatomical characteristics in temperature-limited coniferous trees (Tardif et al., 2001; Deslauriers et al., 2007). Annual tree ring growth begins with the production of large, thin-walled earlywood cells (Creber and Chaloner, 1984; Barnett and Jeronimidis, 2003). As environmental conditions change through the summer months, the cells transition to small thick-walled latewood with decreasing lumen size (Brown et al., 1949; Wort, 1962; Creber and Chaloner, 1984; Barnett and Jeronimidis, 2003). Consequently, where temperature is the primary limiting factor of growth (D’Arrigo et al., 1992; Davi et al., 2002; Watson and Luckman, 2002; Savva et al., 2010), developing an understanding of conditions at the time of ring formation may provide the insight needed to assess specific wood quality attributes in a specific area (Fonti et al., 2010). Climate variables such as air temperature and precipitation directly affect energy accumulation within a tree by influencing net photosynthesis, and thus influence the resources available for construction of tracheids and cell walls. During periods of low net photosynthesis, radial growth slows significantly and the cells produced are smaller and/or thinner-walled. For example, Tardif et al. (2001) reported on the radial growth of seven boreal tree species in response to increases in soil and air temperature. Periods of substantial and continuous radial diameter were positively associated with precipitation and soil temperature, and negatively associated with the photoperiod (Tardif et al., 2001). Because of the existing relationship between climate and cell formation, intraspecific tree-ring and tracheid variability can be used to determine variability in climate and ecosystem conditions (Fonti et al., 2010). By identifying trends in the relationship between climate parameters and wood anatomical characteristics over time, it should be possible to predict how climate variables will effect wood production in specific locations, and therefore certain wood quality attributes produced at that site in the future. These insights would be valuable for forest managers as they design management and rotation strategies tuned to future climates (Williamson et al., 2009; MFLNRO, 2013; IPCC, 2013). The objectives of this study were to investigate the relationships between climate and wood anatomical development, and to use the relationships identified to predict certain wood quality attributes in interior BC forest stands. Time series were developed to predict wood properties based on correlations to historical climate station data, and the predicted values were compared to measured wood attributes. For the purposes of this study, we focus solely on cellular wood quality attributes including density, cell wall thickness, and MFA. The findings of the research provide preliminary insight into whether changing climates are likely to have a positive or negative impact on wood quality attributes within future forests in interior BC.
2. Methods Tree-ring samples were collected from five sites: four sites in northern BC and one site in southern BC (Fig. 1). All of the sites contain mature, open canopy, mixed-species forests dominated by softwoods. Sites were located on well-drained, nutrient-rich sites found in close proximity to long-term climate stations. High-elevation or northern latitude sites were identified to ensure that climate was the primary factor limiting tree growth (Fritts, 1976). Samples were collected from three hybrid Engelmann spruce (Picea glauca (Moench) Voss x engelmannii (Parry)) stands in the Smithers area (sites A (NSx1), B (NSx2), and C (NSx3)) and from Douglas-fir (Psuedotsuga menziesii (Mirb.) Franco) trees located at sites near Babine Lake (site D (NDf4)) and Pemberton (site E (SDf7)) (Fig. 1). Two 5.2 mm cores from opposite and one 12 mm core were collected from each tree at breast height (Stokes and Smiley, 1968).
The 5.2 mm cores were allowed to air dry, glued to a grooved mounting board, and were sanded until the annual ring boundaries were clearly visible (Stokes and Smiley, 1968). Mounted cores were scanned with a high resolution Epson XL1000 flatbed scanner to create digital images, which were then used to measure the width of each annual ring to 0.001 mm using WindendroÒ software (Version 2006). Annual rings that were exceptionally narrow or unclear were measured to 0.001 mm using a VelmexÒ tree-ring measurement system equipped with a trinocular boom-mounted microscope and CCD video display. Following air drying, each 12 mm core was prepared for densitometric analysis by gluing it flush to the surface of a 2.5 cm-wide fibreboard block. Once dry, a 2 mm thick wood lath was cut (pith to bark) with a Waltech high precision twin-bladed saw to reveal the radial surface of the core (Haygreen and Bowyer, 1996). Resins add to wood’s structural mass and must be removed prior to wood density measurement (Lenz et al., 1976). In this instance, wood resins were chemically extracted using an acetone Soxhlet apparatus (Jensen, 2007). Acetone was placed in a round-bottomed flask resting on a hot plate, and the wood lathes were placed in the Soxhlet chamber sealed by a condenser cycling cold tap water. Acetone was cycled over the samples for five hours by consecutively evaporating and condensing the liquid to remove the resins (Schweingruber et al., 1978; Grabner et al., 2005). The samples were air-dried after resin extraction and mounted vertically on pins in an ITRAX scanning densitometer. To ensure accurate light attenuation by the laser scanner, care was taken to orient samples perpendicular to the X-ray beam. Scanned images were measured using WindendroÒ ITRAX density software (Version 2008b). Sectioned and extracted 12 mm 2 mm samples were sent to Dr. R. Evans at the Australian Commonwealth Scientific and Research Organization (CSIRO) for SilviScan analysis. The SilviScan system allowed for precise measurement of density, microfibril angle and tracheid radial and tangential diameters, and also calculates cell-wall thickness (Eq. (1)) (Vahey et al., 2007; Lundgren, 2004; Jones et al., 2005). The 12 mm radial width of the core was trimmed to a precise 7 mm prior to scanning with SilviScan. Various wood property measurements were obtained as treering series: total ring width (RW); maximum, mean, and minimum density (MXD, MD, MND); maximum, mean, and minimum microfibril angle (XMFA, MMFA, NMFA); maximum, mean, and minimum cell radial and tangential diameters (XRD, MRD, NRD and XTD, MTD, NTD); and maximum, mean, and minimum fibre coarseness (XC, MC, NC). Data for maximum, mean, and minimum cell-wall thickness (XCWT, MCWT, NCWT) were calculated by SilviScan as a function of the measured density (d), coarseness (C), and cell perimeter distance (P) (Eq. (1)); cell perimeter (P) is calculated using radial vs. tangential cell diameter dimensions (Eq. (2)) (Vahey et al., 2007; Lundgren, 2004).
CWT ¼ P=8 1=2ðP=16 C=dÞ
1=2
ð1Þ
where
P ¼ 2ðR þ TÞ
ð2Þ
and, R and T are the radial and tangential tracheid diameters, respectively (Jones et al., 2005). 2.1. Chronology development Each tree-ring series was cross-dated with respect to characteristic annual ring width patterns (Stokes and Smiley, 1968). The series cross-dating was verified using COFECHA (Holmes, 1983), and annually resolved chronologies were developed for each species and wood property identified (Table 1). Changes made to the time series during the ring width cross-dating process, such as the
L.J. Wood et al. / Forest Ecology and Management 361 (2016) 81–89
83
Fig. 1. Sampling sites (triangles): site A and B (Hudson Bay Mountain), site C (Blunt Forest Service Road), site D (Gullwing Forest Service Road), site E (Owl Creek Forest Service Road). Climate stations (circles) with longest records in closest proximity to sampling sites: Smithers, Fort St. James, Agassiz.
deletion of a false ring or adjustment of dates due to broken core segments, were applied over all of the measured wood property chronologies from the same samples to ensure consistency. The time-series created in COFECHA were transformed to ARSTAN master chronologies to eliminate non-climatic variation (Cook and Holmes, 1986). A negative exponential curve was applied to series showing age-related growth trends (Fritts, 1976). Following this, a smoothing spline was applied to the RW chronologies, with a frequency–response cut-off set to 67% to remove non-climate impacts (Cook and Kairiukstis, 1990). Previous research has shown growth trends due to factors such as inter-tree competition are not pronounced in densitometric tree-ring chronologies (Conkey, 1986). Consequently, only a first-order detrending was carried out on density and fibre property series, using negative exponential or straight line fits. To eliminate autocorrelation issues, only the pre-whitened residual ARSTAN chronologies were used. 2.2. Modelling The standardized and detrended master chronologies were compared to the climate data from the station closest to the sampling location (Fig. 1). Instrumental climate data were obtained from the Adjusted Historical Canadian Climate Data website (http://www.cccma.ec.gc.ca/hccd/) for the Smithers (station #1077500, Lat 54°82’N, Long 127°18’W, 522 m asl), Fort St. James
(station #1092970, Lat 54°45’N, Long 124°25’W, 686 m asl), and Agassiz climate stations (station #1100120, Lat 49°25’N, Long121°77’W, 15 m asl) (Fig. 1). The average summer temperatures (1895–present) recorded at the Agassiz climate station ranges between 20 and 27 °C, while those at the Fort St. James climate station (1895–present) range between 9 and 17 °C. Initial climate–tree growth relationships were established with a response function in PRECON version 5.17B (Fritts, 1999). Correlations were identified between temperature and precipitation data from the nearest climate stations, and indexed chronologies for the most recent half of the available climate data (1950 to present). Where significant relationships existed regressions were carried out to hind-cast wood property measurements using climate data as a proxy for wood properties. Wood property data from 1950 to present were used as the dependent variable in the model, and climate data from as early as possible (1938 or 1895) was used as the independent variable. Temperature and precipitation were investigated in their relationships with the wood attributes presented here, but only models which used temperature to predict the various wood parameters were considered to be reliable for presentation. Wood property models were restricted in length to expressed population signal (EPS) values of P0.85 with only one decade of this time permitted to drop to an EPS value of P0.80 (Wigley et al., 1984). Once successful models of wood properties were made, they were verified using correlation to the known wood measurements for the last 100 years, and using split
84
L.J. Wood et al. / Forest Ecology and Management 361 (2016) 81–89
Table 1 Master chronology statistics for BC wood quality sample collection sites. Sx = Engelmann white spruce hybrids, Fd = Interior Douglas-fir. RW = ring width, MXD = maximum density, MD = mean density, MND = minimum density, XCWT = maximum cell wall thickness, MCWT = mean cell wall thickness, NMFA = minimum microfibril angle, MMFA = mean microfibril angle, XMFA = maximum microfibril angle, NRD = minimum radial diameter. Chronology type
Species
Site name
Number of cores (trees)
Chronology length
Interseries correlation
Mean sensitivity
Expressed population signal P 0.80
RW RW RW RW RW MXD MD MND MXD MD MND MXD MD MND MXD XCWT XCWT XCWT MCWT NMFA NMFA NMFA MMFA XMFA NRD
Sx Sx Sx Fd Fd Sx Sx Sx Sx Sx Sx Fd Fd Fd Fd Sx Fd Fd Fd Sx Fd Fd Fd Fd Fd
A B C D E A A A B B B D D D E A&B D E E A&B D E E E E
23 (16) 17 (13) 19 (17) 18 (14) 18 (15) 19 (13) 20 (13) 16 (11) 19 (14) 20 (14) 18 (13) 7 (7) 17 (11) 14 (8) 8 (7) 19 (18) 7 (7) 13 (12) 11 (11) 13 (12) 8 (8) 8 (8) 10 (10) 10 (10) 8 (8)
1598–2006 1680–2006 1726–2006 1754–2006 1672–2006 1610–2006 1670–2006 1670–2006 1716–2006 1716–2006 1716–2006 1810–2006 1810–2006 1810–2006 1638–2006 1686–2006 1810–2006 1688–2006 1688–2006 1710–2006 1810–2006 1690–2006 1705–2006 1705–2006 1689–2006
0.448 0.493 0.507 0.500 0.555 0.438 0.465 0.352 0.633 0.536 0.401 0.499 0.461 0.354 0.406 0.473 0.474 0.503 0.443 0.364 0.453 0.377 0.401 0.374 0.330
0.165 0.195 0.169 0.205 0.216 0.074 0.054 0.073 0.106 0.068 0.081 0.076 0.069 0.077 0.114 0.109 0.085 0.083 0.067 0.074 0.102 0.094 0.065 0.102 0.056
1786–2006 1805–2006 1853–2006 1855–2006 1780–2006 1791–2006 None None 1764–2006 1764–2006 1814–2006 1912–2006 1912–2006 1970–2006 None 1782–2006 1912–2006 1787–2006 1787–2006 1870–1920 1940–2006 1775–1850 1940–2006 None None
verification, where the latter 50% of the time series was used to calibrate the models. To ensure that the observed correlations were true and not caused by autocorrelation, the Durbin–Watson statistic was used; values of 2.0 were considered to have no autocorrelation, and values equal or less than 1.0 were deemed positively autocorrelated. 3. Results 3.1. Chronologies In total, 20 chronologies were created: five acceptable RW chronologies; ten density chronologies of which seven were acceptable; four cell wall thickness chronologies of which three were used; five microfibril angle chronologies; and, one radial diameter chronology that did not meet the acceptable standards for EPS (Table 1). Only chronologies displaying reliable EPS values (Wigley et al., 1984) and with minimal autocorrelation according to the Durbin–Watson statistic are discussed. 3.2. Sites A and B The site B MXD and site A and B XCWT chronologies positively correlated to mean May, July, and August temperature records from the Smithers station, and the site A and B NMFA chronology was negatively correlated to mean summer temperature (average of June, July, August) from Smithers (Tables 2 and 3). Multivariate regression models predicted MXD, XCWT, and NMFA based on these variables (Fig. 2). The Durbin–Watson statistic (DW) suggested some positive autocorrelation in the MXD and XCWT reconstructions, but not to a level of concern (Table 3). Partial correlation tests were carried out to ensure that the mean monthly temperatures were all separately influencing the predicted variables. Each month remained significantly correlated to MXD, XCWT or NMFA, even when controlling for the others. Fig. 2 shows strong correspondences between the predicted models and the
measured values for Sx MXD and Sx XCWT; these relationships are also reflected in the Pearson’s correlation coefficients (Tables 3 and 4). Fig. 2 illustrates that monthly average summer temperature has a close correspondence to the maximum density of the cells and the thickness of tracheid cell walls at this northern interior site. The reconstructed values approximate the measured data, except for in 1977, a year which shows a much lower measured value than was predicted by either model. Fig. 2 also shows time periods, for example from 1982 to 1986, where the XCWT model underestimated the measured values. During periods when the model did not accurately predict the wood anatomical values, we assumed that another climate or environmental factor primarily influenced the development of these characteristics. Fig. 2 shows the correspondence between the predicted values of Sx NMFA and the measured values obtained from the SilviScan analyses. Proxy mean summer temperature data from the Smithers climate station approximates the minimum microfibril angle of spruce growing in the area, except for certain years. For example, in 1977 the actual measured data shows a high NMFA value not captured by the model. The 1977 data point shown in this model corresponds to the overestimated data point in 1977 in the MXD and XCWT models. While the measured MFA value for 1977 was high, the RW reconstruction in Fig. 2 shows a low measured ring width value for 1977 as was true for the models for XCWT and MXD discussed above. The NMFA reconstruction in Fig. 2 displays data that overestimates the measured values prior to 1950. The distinct appearance of model overestimation prior to 1950 could indicate a non-linear relationship through time, which may be due to a flaw in the data standardization process to remove growth-related trends.
3.3. Site C Ring width from site C was positively correlated with mean June temperature from Smithers (r = 0.476, p = 0.01) and a regres-
85
L.J. Wood et al. / Forest Ecology and Management 361 (2016) 81–89
Table 2 Pearson’s correlation coefficients illustrating significant relationships between the chronologies presented and climate station data. Where data is not reported, no significant relationship was identified. Note that chronologies from sites A, B, and C were compared to Smithers climate station data, chronologies from site D were compared to Fort St. James climate station data, and chronologies from site E were compared to Agassiz climate station data. NMFA = minimum microfibril angle, XCWT = maximum cell wall thickness, RW = ring width, MXD = maximum density, MMFA = mean microfibril angle, NRD = minimum radial diameter.
Avg May temp Avg June temp Avg July temp Avg August temp Avg summer temp Avg spring temp Total spring precip Total July precip Total August precip
Site A and B NMFA
Site A and B XCWT
Site B RW
Site B MXD
Site C RW
Site D RW
Site D MXD
Site D XCWT
Site E MMFA
Site E NRD
Site E XCWT
0.386 0.320 0.468 0.538
0.447 0.345 0.401 0.436
– 0.542 0.356 –
0.419 0.311 0.452 0.544
– 0.476 0.430 –
– – – –
0.343 0.413 0.548 0.302
0.350 0.453 0.516 0.342
0.289 – 0.369 0.311
– – 0.601 0.441
0.318 0.497 0.486 0.626
0.645
0.580
0.517
0.634
0.527
–
0.586
0.612
0.351
0.585
0.733
– –
0.456 –
– –
0.324 –
– –
– 0.466
– 0.331
– 0.308
0.476 –
– –
0.538 –
– –
– –
– –
– –
– –
– –
0.422 0.336
0.373 0.395
– –
– –
– –
Table 3 Statistics describing proxy wood quality reconstructions for specific sites in interior BC (⁄ = p < 0.05). NMFA = minimum microfibril angle, XCWT = maximum cell wall thickness, RW = ring width, MXD = maximum density, NRD = minimum radial diameter. Reconstruction
Interior spruce MXD
Interior spruce XCWT
Interior spruce NMFA
Interior spruce RW
Douglas-fir MXD
Douglas-fir XCWT
Douglas-fir XCWT
Douglas-fir MMFA
Region
North-Smithers
North-Smithers
North-Smithers
Chronology collection site Proxies
B
A and B
A and B
NorthSmithers C
North-Babine Lake D
North-Babine Lake D
SouthPemberton E
SouthPemberton E
May, July, August temp Smithers 0.691 0.477 1.036 0.411⁄ 0.261⁄
May, July, August temp Smithers 0.620 0.385 1.118 0.325⁄ 0.244⁄
Mean summer temp Smithers 0.645 0.416 1.476 0.451⁄ 0.044
June temp
Mean summer temp Fort St. James 0.586 0.343 1.359 0.331⁄ 0.614
Mean summer temp Fort St. James 0.612 0.374 1.336 0.364⁄ 0.573
June, July, August temp Agassiz 0.761 0.579 1.590 0.523⁄ 0.997
Mean spring temp Agassiz -0.476 0.227 1.170 0.151⁄ 0.137
Climate station Pearson’s R R2 Durbin–Watson RE calibration RE verification
Table 4 Correlation between measured wood properties averaged over a chronology and the reconstructed wood property based on climate variables.
Site Site Site Site Site Site Site Site Site
B MXD A and B XCWT A and B NMFA C RW D XCWT D MXD D RW E XCWT E NRD
Measured vs. reconstructed (Pearson’s R)
P value
0.613 0.546 0.500 0.368 0.334 0.284 0.237 0.418 0.343
0.001 0.001 0.001 0.002 0.001 0.003 0.013 0.001 0.001
sion model was created to predict RW. The r2 value for this model was 0.226 (p 6 0.001). DW was found to be 1.275, suggesting only slight positive autocorrelation (Fig. 2). Mean June temperature also under-predicted RWs of spruce from Site C after 2000. 3.4. Site D The XCWT and MXD chronologies from site D were significantly correlated to mean summer temperature from the Fort St. James climate station (Table 3; Fig. 3). Fig. 3 shows Douglas-fir XCWT and MXD, reconstructed from mean summer temperatures in Fort St. James. The models in Fig. 3 under-predict XCWT and MXD from 1937 to 1954, indicating that a factor other than the summer temperature variable may have influenced the cell wall deposition process during this time period.
Smithers 0.476 0.226 1.275 0.266⁄ 0.056
3.5. Site E At site E, the XCWT chronology was correlated to June, July, and August temperatures from the Agassiz climate station. These three temperature variables were independently used in a multivariate regression model to predict Douglas-fir (Fd) XCWT (Table 3). Fig. 4 shows that the reconstructed XCWT of Fd from site E is a close representation of the actual measured XCWT. This finding indicates that Agassiz June, July, and August temperatures, used independently in a multivariate model, provide a robust predictor of XCWT development in Douglas-fir for this region. The only time periods where the reconstruction does not correlate well with the measured record are short durations from 1900 to 1910, and from 1941 to 1947 when the model under-predicts the measured XCWT. The site E MMFA chronology was significantly negatively correlated to average spring (March–May) temperatures from Agassiz climate station. Fd MMFA was then hind-casted based on average spring temperature. The DW value displayed slight positive autocorrelation (Table 3, Fig. 4). 3.6. Model statistics The reconstructed wood properties based on climate proxies correlate back to the original measured wood properties and provide validation of the models (Figs. 2–4). The results of the split verification method used were mixed (Table 4). The predicted site B MXD and site A and B XCWT passed the verification test with reduction of error (RE) values significant at 95% when compared
86
L.J. Wood et al. / Forest Ecology and Management 361 (2016) 81–89
Fig. 4. Measured (dotted lines) vs predicted (solid lines) mean microfibril angle and maximum cell-wall thickness in Douglas-fir from southern interior BC (site E). MMFA modelled from average spring temperature (March, April, May) from Agassiz climate station. XCWT modelled from mean June, mean July, and mean August temperatures from the Agassiz climate station, and again using only July mean temperature from the Agassiz climate station.
Fig. 2. Measured (dotted lines) vs predicted (solid lines) maximum ring density, maximum ring cell-wall thickness, minimum ring microfibril angle, and ring width in interior spruce from northern interior BC. MXD and XCWT modelled from mean May, mean June, and mean August temperatures from the Smithers climate station. NMFA modelled from mean summer temperature (June, July, August average) from the Smithers climate station. RW modelled from mean June temperature from the Smithers climate station. Sites where trees were sampled are listed on the figure.
95% in the verification stage of the test (Table 3). This outcome may be due to a lack of similarity between the first half of the time series and the last half. This was especially notable in the site A and B NMFA reconstructions, where summer temperature from Smithers over-predicted NMFA prior to 1946, and in the site D MXD and XCWT reconstructions where summer temperature from Fort St. James under-predicted the wood property measurements prior to 1954. The site E reconstruction using MMFA also only showed significant correlation with the measured values after 1950. We also suspect that sample depth was responsible for poor verification outcomes, as larger sample sizes generally created more robust chronologies with which to compare to climate variables. For instance the reconstructions from sites A and B, and the XCWT reconstruction from site E were the most reliable, with P17 cores used in the chronologies from sites A and B, and 13 cores were used for the XCWT chronology from site E. The EPS values NMFA from site A and B were low, and only reached a value P0.80 for a short section of the time series (Table 1), indicating that the sample size was not large enough to generate a stable time series curve. A larger sample would reduce the variability among the MFA values for each year, and may eliminate the inaccuracy of the MFA models prior to 1950.
4. Discussion Fig. 3. Measured (dotted lines) vs predicted (solid lines) maximum ring density and maximum ring cell-wall thickness in Douglas-fir from northern interior BC (site D). MXD and XCWT modelled from mean summer temperature (June, July, August average) from the Fort St. James climate station.
to the actual measured data. The remaining reconstructions, site A and B NMFA, site C RW, site D XCWT, site D MXD, site D RW, site E XCWT, site E MMFA, and site E NRD all passed the calibration part of the split verification, but did not display a significant RE value at
4.1. Reconstruction of wood properties The strong relationship between tree growth and climate at the northern interior sample sites permitted successful construction of RW, MXD, XCWT, and NMFA chronologies using measured temperature variables from nearby climate stations. The models (Sx MXD, XCWT, NMFA, RW and FD XCWT and MXD) were not, however, reliable in all years. This outcome was particularly apparent in 1977 when the measured MXD, XCWT, and RW values were lower
L.J. Wood et al. / Forest Ecology and Management 361 (2016) 81–89
than average, and the measured NMFA value was higher than average. Given that mean August temperatures in 1977 were more than two standard deviations above average (Smithers, 1938–present), it may be that the trees in this region were unable to sequester excess carbon in their cell walls resulting in lower overall average density and cell-wall thickness. If this was the case, it may indicate that as August temperatures become warmer, wood density and quality attributes may decrease in northern coniferous trees. It is also possible, however, that reduced ring widths in 1977 were a lagged response to colder-than-average previous-year temperatures or greater than average precipitation (Chavardès et al., 2013). Increased precipitation would allow for cell radial expansion and the earlywood portion of the ring to become wider, thus increasing the width of the annual ring (Wimmer and Grabner, 2000; Tardif et al., 2001). For example, there are several higherthan-average values in the measured RW record prior to 1949 that were not captured in the model, suggesting that June temperature alone is not an effective predictor of spruce RW prior to 1949. Regardless of the source of this error, the inability of temperature variables to capture some of the variation in the wood quality attributes examined indicates that a more complex, multivariate model is required. Future research should identify soil moisture indices and snow pack data to test whether these variables contributed to the deviations observed here. In years of low ring density and narrow cell-wall thickness, it is not uncommon to see high MFA values, which explains the synchronization between models. There are few studies that investigate the relationship between MFA, cell-wall thickness, and density, so how these variables are related is not clear (Hein and Brancheriau, 2011; Xu et al., 2012; Wood and Smith, 2014). Hiller (1964a,b) and Evans et al. (2000) report significant correlations exist between density, cell-wall thickness, and MFA, but other researchers were unable to discern definitive relationships (Bergander et al., 2002; Schimleck and Evans, 2002; Lin and Chiu, 2007). High MFA values lead to high longitudinal shrinkage in dimensional lumber, and are also correlated with low strength and stiffness (Cave, 1968). Both MFA records have measured values lower than were predicted by the reconstructions models prior to 1950 (see Fig. 2). Microfibril angle is a trait that is extremely variable depending on age. Juvenile wood shows much higher fibril angles than mature wood (Haygreen and Bowyer, 1996; Burdon et al., 2004). This agerelated trend should not be reflected in the reconstructions presented because growth-related trends were removed from the series using a negative exponential curve in ARSTAN. It is possible that a variation on the negative exponential curve used for standardizing RW is required for MFA data. The overestimation of NMFA in this study indicates that measured and standardized minimum microfibril angles were lower than predicted by the climate proxy. Similar nonlinear behaviour was reported by Xu et al. (2012) who reported a stronger, more significant relationship between temperature and MFA from 1987 to 2009. The XCWT attribute model generated from the Agassiz climate station for trees sampled near Pemberton under-predicted the measured values from 1900 to 1910. This outcome could be the result of higher-than-average July temperatures in the Agassiz record not reflected in the model average of June, July, and August temperatures. To test this hypothesis, a regression of XCWT for site E was performed using only July temperature from Agassiz (Fig. 4). The statistical significance of this simple linear regression is not as strong as the multivariate model (r = 0.486, r2 = 0.236, p 6 0.001), but the predicted values visually follow the measured data very well for the 1900–1910 period that was previously underpredicted. This observation indicates weather events during a growing season, such as an interval of higher-than-average temperatures, may be very influential on wood anatomical
87
development. In the case of Douglas-fir from site E, July temperatures play a larger-than-normal role in the overall XCWT development from 1900 to 1910 than during the remainder of the time series. Furthermore, it should be noted that at this southern site, extreme July temperatures resulted in higher XCWT and MXD in Douglas-fir, where as in the north, the higher-than-average temperature values in August were observed to have the opposite effect. This differential climate-growth response to high summer temperatures emphasizes that trees with a higher percentage of latewood are usually more tolerant to heat and drought than those with proportionally more earlywood, a response believed related to the structural complexity of latewood cells and their decreased dependence on turgor pressure (Kozlowski, 1979; Barnett and Jeronimidis, 2003; de Luis et al., 2011). The MMFA predictions at the southernmost site (site) were not as strong as the other models developed in this study; however, they follow the long-term trends of the measured values despite the evident high-frequency variation (Fig. 4). MMFA was negatively correlated to mean spring temperatures, which means that when spring temperatures increased, MMFA decreased; warm springs produced conditions where mean microfibril angles were low in Douglas-fir. There is support for MFA–climate relationships (Xu et al., 2012). Xu et al. (2012) found significant correlations between July and August temperatures and spring precipitation and MFA measured in Picea crassifolia sampled from northeastern Tibet. The difference in climate response in this study is likely due to regional environmental differences. While the model presented here over-predicts the measured values prior to 1946 and for a span in the early 1970, it is included here as an example of the range of types of wood properties that can be investigated through the use of climate proxies. 4.2. Influence of climate forcing mechanisms Large regional climate forcing mechanisms, such as are described by the Pacific Decadal Oscillation (PDO), are presumed to be influential in the long-term growth trends of trees found in northwestern North America interior BC (Chavardès et al., 2013; Pitman and Smith, 2013; Rood et al., 2013). The PDO underwent a large-scale regime shift from a ‘‘cool phase” to a ‘‘warm phase’ between 1976 and 1977, and this shift is synchronous a climategrowth anomaly in the northern attribute reconstructions. Cool PDO phases have been correlated with growth-inhibiting climates off the coast of Alaska and growth-enhancing climates in the southwest USA, with warm PDO phases having the opposite influence (Mantua, 2000). Given these relationships, a PDO shift from a cool phase to a warm phase could result in the climate within interior BC transitioning from one that is growth-inhibiting to one that is growth-enhancing. Support for this hypothesis rests with the 1946–1947 PDO transition from a warm to a cool phase (Mantua, 2000), a regime shift that may be responsible for the under prediction NMFA values at sites A and B prior to 1946, and the decoupled of the measured and predicted records at site E (XCWT, MMFA and NRD) at this time. 4.3. Climate, wood development and wood quality The ability to correlate and model wood properties based on a single climate factor, such as temperature, provides forest managers and wood producers with a non-invasive methodology for describing forest stand quality attributes that are not otherwise discernible. While this approach may be effective for elucidating fluctuations in some of the attributes that contribute to wood quality, its application is limited in settings where compounding environmental factors influence tree growth. Trees that demonstrate variable inter- and intra-seasonal relationships to different
88
L.J. Wood et al. / Forest Ecology and Management 361 (2016) 81–89
limiting factors will require more complex models for predicting wood quality attributes. Relationships between certain attributes and climate could be combined with site index measurements for productivity to evaluate the overall impact that changing climates will have on productivity and quality within a forest stand. It is well known that a vast array of environmental and climate parameters interact to produce the ecosystem dynamics within which a tree grows (Kozlowski, 1979); however, accounting for variation in all of these factors requires a complex modelling system. The advantage of the modelling method presented in this study, however, is its ability to derive stand-level wood quality attributes from climate station variables. Through investigation of the general correspondence between climate variables and wood anatomical variables, we draw some preliminary conclusions regarding certain wood quality attributes of the stands studied. Correlation analysis demonstrated that most cell wall thickening, and consequently, density variables for both Douglas-fir and spruce stands in the interior of BC were positively correlated to increases in mean summer temperatures. There were a few instances where it was evident that another variable was more important to radial development, but summer temperature was the primary variable involved in wood formation in these stands over the last 50–100 years. Therefore, we conclude that wood tissue grown when summer temperatures are warmer for longer intervals will be higher in density and lower in MFA, and should demonstrate the greater strength and stiffness properties sought out for solid wood production (Haygreen and Bowyer, 1996; Wimmer and Downes, 2003; Ansell, 2011). A caveat for this general rule may be if summer temperatures reach an extreme threshold, at which time, trees in the north may be ill-adapted to the conditions and produce some attributes of poorer quality wood (lower density and CWT). In globally-changing climates conditions, this caveat may prove to be a very important factor for forest managers to assess. It was demonstrated that early summer temperatures have a proportionally greater influence on ring width development in interior hybrid spruce stands than do late summer temperatures. It could be expected then hybrid spruce forests within interior BC that experience warmer spring temperatures combined with average to cooler summer temperatures will produce wood with lower density and strength properties that is best suited for pulp production. In contrast, hybrid spruce trees growing at sites that experience warmer-than-average spring and summer temperatures are likely to produce the type of uniformly wide, high density rings that are optimal for solid wood products; provided that excessively warm temperatures do not lead to moisture deficits (Kozlowski, 1979; Wimmer and Downes, 2003). Lastly, it was demonstrated that warm spring temperatures in the southern BC interior result in the production of wood cells in Douglas-fir trees with a low mean microfibril angle. This finding indicates that as temperatures warm earlier in the growing season, the wood produced may have a lower MFA and, therefore, higher strength and stiffness properties. Wood with lower MFA is ideal for solid wood production (Jozsa and Middleton, 1994; Ansell, 2011).
5. Conclusion Radial tree growth is influenced by environmental factors that combine to produce variable annual ring increments. Regional climate variables interact with local site conditions to produce individual rates of photosynthesis, and localized adaptations to climate within each tree. This behaviour translates into an allocation of energy for radial growth and wood anatomical development. In this study, average summer temperatures were shown to have the
greatest effect on maximum density, cell-wall thickness and microfibril angle development in the forest stands investigated. This finding was reflected in consistent correlations between these variables, and also between the observed and reconstructed values for MXD, CWT, and MFA based on mean summer temperatures. Since these wood properties are important wood quality predictors, this information is significant. This study aimed to identify whether or not climate can be used as a reasonable indicator of wood properties and certain wood quality attributes. Based on the chronologies created for specific regions of BC, it was shown that distinct wood property characteristics can be modelled by various temperature variables. Obviously many environmental components contribute to the formation and strength of wood, however, targeting a few limiting factors with which to model wood quality within a stand provides valuable insight from a simplified method requiring less investment by individual forest managers. Climate is a major contributing factor to growth and development of wood cells, and therefore using it as an indicator is both valid and useful to the prediction of certain wood quality attributes in future forest stands. Fibre characteristics including density, cell-wall thickness, and microfibril angle were estimated based on local mean summer temperatures in this study. These traits fluctuate through time to create certain wood attributes when warm temperatures prevail, such as increased cell-wall thickness, which is a desirable wood quality attribute for dimensional lumber. Increased strength and stiffness properties from increased cell-wall thickness and decreased microfibril angle can be expected in warming BC climates, but only up to a threshold. Northern trees may react negatively to temperatures over a certain point, whereas southern interior trees may flourish. Soil water availability also becomes an important consideration at high temperatures. Reconstructions, such as those produced here, help to elucidate the years in which climate had a negative impact on certain wood quality attributes; for example, in extreme climate years when microfibril angle is high. This study demonstrates a non-invasive method useful for identifying climate-related wood quality attributes, and has potential for developing an understanding of the changes to wood quality likely to occur within BC’s interior forests as the climate warms over the next century. Acknowledgements The authors thank Leslie Abel, Bethany Couthard, Aquila Flower, Lynn Koehler, and Branden Rishel for their field assistance, and to Kyla Patterson for her data preparation and technical support. Financial support for this research was provided by Natural Science and Engineering Research Council of Canada (NSERC) awards to Wood and Smith, and a Canadian Foundation for Climate and Atmospheric Science (CFCAS) award to the Western Canadian Cryospheric Network (WC2N). References Adams, S.T., 2014. The Impact of Changing Climates on Tree Growth and Wood Quality of Sitka Spruce. Unpublished PhD thesis, University of Glasgow. Ansell, M.P., 2011. Wood – a 45th anniversary review of JMS papers. Part 1: the wood cell wall and mechanical properties. J. Mater. Sci. 46, 7357–7368. Barnett, J.R., Jeronimidis, G., 2003. Wood Quality and Its Biological Basis. CRC Press LLC, Boca Raton, USA. Bergander, A., Brändström, J., Daniel, G., Salmén, L., 2002. Fibril angle variability in earlywood of Norway spruce using soft rot cavities and polarisation confocal microscopy. J. Wood Sci. 48, 255–263. Brown, H.P., Panshin, A.J., Forsaith, C.C., 1949. Textbook of Wood Technology, vol. 1. McGraw-Hill Company Inc., New York, NY. Burdon, R.D., Kibblewhite, P., Walker, J.C.F., Megraw, R.A., Evans, R., Cown, D.J., 2004. Juvenile vs. mature wood: a new concept, orthoganol to corewood versus outerwood, with special reference to Pinus radiate and P. taeda. For. Sci. 50, 399–415.
L.J. Wood et al. / Forest Ecology and Management 361 (2016) 81–89 Cave, I.D., 1968. The ansiotropic elasticity of the plant cell wall. Wood Sci. Technol. 2, 268–278. Chavardès, R.D., Daniels, L.D., Waeber, P.O., Innes, J.L., Nitschke, C.R., 2013. Unstable climate-growth relations for white spruce in southwest Yukon, Canada. Clim. Change 116, 593–611. Creber, G.T., Chaloner, W.G., 1984. Influence of environmental factors on the wood structure of living and fossil trees. Bot. Rev. 50 (4), 357–448. Conkey, L.E., 1986. Red spruce tree-ring widths and densities in eastern North America as indicators of past climate. Quater. Res. 26, 232–243. Cook, E.R., Holmes, R.L., 1986. User’s manual for the program ARSTAN. In Tree-ring chronologies of western North America: California, eastern Oregon, and Northern Great Basin with procedures used in the chronology development work including users manuals for computer programs. COFECHA and ARSTAN. In: Holmes, R.L., Adams, R.K., Fritts, H.C. (Eds.), Chronology Series VI. Laboratory of Tree-ring Research, The University of Arizona, Tucson, pp. 50–65. Cook, E.R., Kairiukstis, L.A. (Eds.), 1990. Methods of Dendrochronology: Applications in the Environmental Sciences. Kluwer, Dordrecht. D’Arrigo, R.D., Jacoby, G.C., Free, R.M., 1992. Tree-ring width and maximum latewood density at the North American tree line: parameters of climatic change. Canad. J. For. Res. 22, 1290–1296. Davi, N., D’Arrigo, R., Jacoby, G., Buckley, B., Kobayashi, O., 2002. Warm-season annual to decadal temperature variability for Hokkaido, Japan, inferred from maximum latewood density (AD 1557–1990) and ring width data (AD 1532– 1990). Clim. Change 52, 210–217. ˇ ufar, K., 2011. Climate de Luis, M., Novak, K., Raventós, J., Gricˇar, J., Prislan, P., C factors promoting intra-annual density fluctuations in Aleppo pine (Pinus halepensis) from semi-arid sites. Dendrochronologia 29, 163–169. Deslauriers, A., Rossi, S., Anfodillo, T., 2007. Dendrometer and intra-annual tree growth: what kind of information can be inferred? Dendrochronologia 25, 113– 124. Evans, R., Stringer, S., Kibblewhite, R.P., 2000. Variation of microfibril angle, density and fibre orientation in twenty-nine Eucalyptus nitens trees. Appita J. 53, 450– 457. Fonti, P., von Arx, G., García-González, I., Eilmann, B., Sass-Klasen, U., Gӓrtner, H., Eckstein, D., 2010. Studying global change through investigation of the plastic responses of xylem anatomy in tree rings. New Phytol 185, 42–53. Fritts, H.C., 1976. Tree Rings and Climate. Academic Press, New York. Fritts, H.C., 1999. PRECON Version 5.17B. A Statistical Model for Analyzing the TreeRing Response to Variations in Climate.
89
Larsen, JB., 1993. Testing winter desiccation resistance for species and provenance selection at timberline. In: Alden, J., Mastrantonion, J.L., Ødum, S. (Eds.), Forest Development in Cold Climates. Plenum Press, New York, NY, pp. 59–64. Lenz, O., Schar, E., Schweingruber, F.H., 1976. Methodological problems relative to measurement density and width of growth rings by X-ray densitogrames of wood. Holzforschung 30, 114–123. Lin, C.J., Chiu, C.M., 2007. Relationships among selected wood properties of 20-yearold Taiwania (Taiwania cryptomerioides) trees. J. Wood Sci. 53, 61–66. Lundgren, C., 2004. Cell-wall thickness and tangential and radial cell diameter of fertilized and irrigated Norway spruce. Silva Fenn. 38, 95–106. Mantua, N., 2000. The Pacific Decadal Oscillation. Joint Institute for the Study of the Atmosphere and Ocean. University of Washington.