Growth at the microscale: long term thinning effects on patterns and timing of intra-annual stem increment in radiata pine

Study sites

Dendrometers and monitoring growth

High resolution point dendrometers were used to monitor stem growth at 15 min sampling intervals on three trees per plot, mounted at 1.3 m above ground level. Data were finally averaged to a daily time step. The Mt. Gambier site was instrumented in Oct 2010, and in Oct 2011 instruments were installed at Flynn (HVP). Growth was monitored for two seasons in each case, although started a year apart due to practical issues which delayed the Flynn installation.

Different, but comparable, systems were used at two sites. The dendrometers at Mt. Gambier used custom-designed housings and mountings, with the sensor itself a Macrosensors AC LVDT (http://www.te.com). Sensors were connected to a Campbell Scientific (http://www.campbellsci.com) CR1000 data loggers via a multiplexor and Macrosensors signal conditioner. The systems at Flynn used dendrometers incorporating Midori Precisions “GreenPot” radial potentiometers (http://www.midori.co.jp) connected to a Campbell Scientific CR800 data logger. Every dendrometer was individually calibrated in the laboratory prior to installation in the field, and temperature sensitivity provided by the suppliers taken into account in the data logger program. Data were regularly uploaded into a MySQL database, where they were stored and processed further.

Weather data

Daily maximum and minimum temperatures, total daily rainfall and total evapotranspiration were obtained for the two study sites from the SILO interpolated data grid (Beesley et al. 2009). This facility uses actual weather data for each date from the nearest weather station/s and provides the best possible estimate of conditions at the particular site by taking into account effect topography and proximity. The approach relied on thin plate smoothing splines for interpolating daily climate variables except daily and monthly rainfall, for which ordinary kriging was used (Jeffrey et al. 2001).

Cambial sampling and analyses

Cambial micro-core samples were taken from the three trees per plot on which dendrometers were mounted the Flynn site on four separate occasions during the 2011–2012 growing season using a Trephor corer (Rossi et al. 2006a). These samples were immediately placed in FAA fixative solution (35% distilled water/50% ethyl alcohol/5% glacial acetic acid/10% formaldehyde). Samples were later reduced in size, mounted in resin, and sectioned to a thickness of 4 μm. These sections were mounted on glass slides, and images captured using a Zeiss Axioscope microscope (Zeiss, Oberkochen, Germany). The numbers of cells in the dividing (cambial) zone and subsequent stage of permanent enlargement were counted, based on increasing radial diameter and the onset of secondary thickening (determined by detecting birefringence in the cell wall) which was taken to indicate the end of enlargement.

Data analysis

Data management and analysis in this study was undertaken using a MySQL database and the R system for Statistical computing (R Core Team 2016). When there were temporary issues resulting from dendrometer failure, generally of short duration, a spline-based interpolation was used to fill in gaps for the purposes of calculating rates of change and other variables.

The onset and cessation of growth was determined not on the raw data, but on modelled annual growth data from a fitted Gompertz function to ensure the initial and final change points of the “main” annual growth trajectory were being captured (see, for example, van der Maaten 2013). The fit of the Gompertz equation against relative stem size increase in two very different trees is shown in Fig. 1. The Gompertz equation fitted the data well at the tree-level. Onset of growth was defined to be the point at which stem diameter exceed 2.5% of the minimum size for the season and cessation the point at which the stem reached 97.5% of the maximum diameter. The number of days of growth was defined to be all days within a monitored period on which a net increment (as opposed to zero change or a shrinkage) could be detected. The sum of these days divided by the total days monitored yielded an estimate of the growth proportion.

Growth proportion or days, onset and cessation of growth, average growth rates were tested between sites, treatments at sites and between years simply using a Student’s t test, in cases where there was a contrast between two cases (e.g. Flynn treatments), and in cases with more than one contrast (e.g. Mt. Gambier treatments), assessment of differences was done using analysis of variance, followed by a pairwise comparison of means using the Tukey HSD test.

Initial tree sizes

Weather conditions and patterns of growth during our study

Patterns of daily growth and overall increase in stem size over the periods for which dendrometers were installed at the study sites is shown in Fig. 2 and Fig. 3. The graphs illustrate the increase in stem radius relative to a zero start point, to make visual assessment of patterns easier, as well as the daily growth rate in stem basal area (mm²∙d^− 1). At both sites, but particularly at Mt. Gambier, there was a growth lull in the winter period from about May to July/August, although some growth was evident throughout the year. At Mt. Gambier, the highest growth rates were seen just after onset, in early Spring, with a general trend of decreasing growth rate thereafter. This pattern was not as evident at Flynn, although there was some indication that growth rates were highest later in the Spring, and declined somewhat thereafter.

At both sites, winter and early spring (when growth started) was slightly cooler in 2012 compared to 2011 (Fig. 4). At both sites, rainfall was lower in February 2012 compared to 2011, and markedly so at Mt. Gambier. At Flynn, Rainfall was lower in late summer months (Jan – April) in 2013 compared to 2012. At the same time, estimated evapotranspiration in late summer was higher in 2012 compared to 2011 and (at Flynn only) in 2013 compared to 2012 (Fig. 4). Taken in concert with the rainfall, mid-summer in 2012 can be considered to have been particularly dry compared to 2011, and 2013 was even drier than 2012 (at Flynn).

Onset, cessation and duration of growth

At both sites, there was evidence that the different thinning histories and final spacing had an effect on the timing of the cessation of annual growth but not the onset (Table 5). At Flynn, trees in the thinned plots stopped growing, on average, 26 days later (p = 0.013) than the unthinned plots. The overall duration of growth, however, did not differ (p = 0.147), partly because of higher variability in the timing of the onset of growth. At Mt. Gambier, the trees in the never-thinned treatment stopped growing about 22 days earlier (p = 0.070) than the other two treatments and had a significantly shorter growing season duration (241 d vs 293 d; p = 0.010).

Growth rates in our study

At Flynn, trees in the thinned plots exhibited positive growth on 174 more days (p = 0.002) (out of 623 monitored) than in the unthinned plots. Overall, the unthinned treatment at Mt. Gambier grew for about 60 fewer (p = 0.013) days (8% of the total monitored period of 727 d) on average than the other two treatments at that site.

If only days on which growth occurred were considered, however (at both Flynn and Mt. Gambier), the differences in growth rate were not significant (Table 6). There was no evidence of any difference in the growth rates of thinned and unthinned plots on days when positive growth was recorded between thinning treatments at Flynn (p = 0.253). At Mt. Gambier, there was slight evidence (p = 0.080) of slower rates of average daily growth, even on “growth days”, in the unthinned treatment compared to the other two treatments combined.

Short-term responses to rainfall events

The differences between short-term patterns and trends in growth were particularly marked between the thinned and unthinned treatments at the Flynn site. These effects are highlighted over one notable period, as seen in Fig. 5. Prior to the rainfall accumulated between mid and late February 2012, trees in the unthinned plots showed no overall growth from early January. Trees in the thinned plots, on the other hand, were growing actively over the same period. Note the very strong increment which was evident following rainfall events on 17th February, 27th – 28th February and 16th – 17th March. In the unthinned plots, these increments were followed by a period of lower rates of growth, even periods of shrinkage. The average magnitude of the daily shrinkage/expansion was significantly (p < 0.001) smaller in the unthinned plots than the thinned plots.

Cambial dynamics at Flynn

Over the course of the 2011–2012 growing season, trees in the thinned treatment had significantly (p < 0.001) more dividing cells (in the cambial zone) (14.9 ± 1.3 vs 7.0 ± 0.5) as well as enlarging cells (13.7 ± 0.9 vs 5.4 ± 0.5) (Fig. 6) compared to the unthinned plots. The ratio of cambial cells to enlarging cells was, however, larger in the unthinned plots (p = 0.026) (Table 7). There was no evidence of any difference, overall, in the number of cambial and enlarging cells between September, November, December and February sampling (p > 0.35 in all pairwise comparisons between months). There was, however, a significant (p = 0.030) declining trend overall from early to late in the growing season which was not the case in the thinned plots. Even in February, there was evidence of high levels of activity in the cambial zone, with a large number of newly formed cell walls clearly discernable and a wide zone of cambial cells (Fig. 6). This could not be detected in the unthinned plots.

Treatment	Date of sampling	Cambial zone cells (#)	Enlarging zone cells (#)	CZ cells/EZ cells
Thinned	2011/09/20	16.3 ± 2.9	14.3 ± 1.9	1.1 ± 0.05
	2011/11/03	10.0 ± 1.0	10.0 ± 2.0	1 ± 0.10
	2011/12/15	18.5 ± 0.5	15.5 ± 0.5	1.2 ± 0.07
	2012/02/10	14.3 ± 1.3	14.3 ± 1.3	1 ± 0.01
Unthinned	2011/09/20	8.0 ± 0.1	6.0 ± 0.0	1.3 ± 0.01
	2011/11/03	7.0 ± 1.5	5.3 ± 1.3	1.3 ± 0.08
	2011/12/15	7.0 ± 1.0	5.7 ± 1.3	1.4 ± 0.33
	2012/02/10	5.5 ± 0.5	4.0 ± 1.0	1.4 ± 0.23

Effects of thinning and stand density on growth dynamics

We postulated that the differences in cumulative growth seen between trees in the different thinning treatments could, in principle, be explained by differences in one or more of three main dynamics (Linderholm 2006). We conjectured that some combination of all three of these possibilities led to the observed differences between the two treatments at Flynn, and between the two thinned vs unthinned (but now at constant SPH) treatments at Mt. Gambier.

Some studies have found that larger, dominant trees retained a competitive advantage by starting to grow earlier than smaller and suppressed trees (Rathgeber et al. 2011; Vieira et al. 2015). Other studies (e.g. van der Maaten 2013 in European beech) have observed that thinning can stimulate earlier onset and a longer season duration. In our study, while we did not find evidence that the timing of onset of seasonal growth was a clear factor in determining differences in overall growth between thinned/unthinned plots, it was clear that thinning affected the timing of the cessation of growth in the major growing season: growth in trees in thinned plots ceased later in the season at both sites. It is particularly notable that even in the case of Mt. Gambier, where all stands were (for our study) at the same stand density, those trees which had previously been thinned ceased growing later in the season than those never thinned.

In addition, we found at both the Mt. Gambier and Flynn sites that the slower growing, never-thinned treatments had fewer days of detectable increment (within the season) than in previously thinned plots. Interpreting the occurrence of a detectable “over-bark” increment as measured by the dendrometers is, of course, complicated by water movement into expansible tissues, and other effects, and is not purely a function of the occurrence of cell production and irreversible expansion (Mäkinen et al. 2003; Zweifel et al. 2006; Drew et al. 2014) and ultimately “growth”. However, it is almost certain that cambial activity would be substantially reduced, if not stopped, when no growth was detectable (Abe et al. 2003; Gruber et al. 2010). Certainly, trees in unthinned plots maintained an overall smaller cambial zone than thinned plots. It is interesting, however, and consistent with previous findings in P. radiata by researchers like Skene (1969), that the sizes of these zones remained fairly constant throughout the season (although possibly a small decline was evident very late in the season in the unthinned plots). Potentially, the relatively more erratic growth during the season in the unthinned plots may have been more a function of variable meristematic activity in a constant-sized cambial zone than of constant expansion/reduction in the size of the meristematic population. The higher ratio between cambial and enlarging cells suggests that the unthinned plots also had a lower duration of enlargement than the thinned plots (cf. Drew and Pammenter 2007).

On days when increment did occur, rates did not differ at all at Flynn, and at Mt. Gambier, by a far smaller margin, between the thinned and unthinned treatments. That is, growth rates were relatively high, temporarily, on “growth days” in the unthinned treatments, but not maintained. These temporary, pulse-like (Zeppel et al. 2008; Drew et al. 2009; De Swaef et al. 2015) responses will be a combination of both a short-term hydration of flexible tissues, and a substantial, but temporary, enhancement of rates of cell division and irreversible tracheid radial/longitudinal expansion (Downes et al. 2009; Zweifel et al. 2014).

It is reasonable to conclude, from our data, that during and at the end of the season limitations from lack of resources became acute in cases where trees were competing more heavily. By contrast, the onset of growth was not affected by these resource limitations.

Seasonal conditions and the timing of growth

Early season temperatures have been shown by a number of authors, often in boreal conifers, to have an important effect in controlling the onset and ending of xylogenesis (Kramer et al. 2000; Rossi and Deslauriers 2007; Rossi et al. 2008; Moser et al. 2009; Bryukhanova and Fonti 2012). In our study, however, the onset of seasonal growth was not apparently affected by whether or not winter conditions were warmer or colder between years. In fact, it is notable that in onset of growth occurred right in the middle of winter, when temperatures were at (or close to) their lowest values, and day length was also not much longer than the winter solstice (about 21 June). Similar early (winter) onset of growth has been seen in radiata pine in places like New Zealand (Tennent 1986) where the conditions on the North Island are comparable to what we had in our study. These findings suggest very strongly that temperature in the very moderate and mild conditions at which the trees were growing in our study, in southern mainland Australia, is of no consequence as a determinant of the timing of growth in Pinus radiata. Certainly, this has been found in Australian P. radiata, in terms of general physiological activity, including bud elongation and internode extension in P. radiata, which occurred throughout the year (Cremer 1973). Similarly, Barnett (1971) found that P. radiata the cambium never became truly dormant. These attributes of P. radiata, growing in this relatively mild environment, may also be a reason why growth was not maximized later in the season. We did not find, as has been found in studies on northern conifers in colder regions (e.g. Rossi et al. 2006b) that maximum growth rates occurred around the time of the longest days. Rather, it would seem that growth rates are as high as ever, if not maximized (this was not always distinct), soon after growth onset, or (at the latest) towards the end of the Spring flush.

By contrast, it was evident that the cessation of growth did differ between years. Growth ceased, at both sites, earlier in the drier (based on both rainfall and evaporation) years. This was not surprising, as other studies have shown that trees growing under drier conditions have been shown to cease growth earlier than trees in less xeric conditions (Gruber et al. 2010). Through the effect of water availability on tree water relations, these conditions can be expected to cause a reduction in cambial activity, tracheid expansion and general phenology and metabolism (Hinckley et al. 1979; Kramer et al. 2000; Drew et al. 2014).

Other phenological changes, in concert with these differing conditions will likely also play a role. Timing of flowering has been shown to play in a role in P. radiata, for example (Fernández and Cornejo 2016) and cessation of growth may be in large part due to phenomena at the tree-level which are themselves linked to the drier conditions. Some work suggests that bud activity and shoot extension in P. radiata is also reduced or temporarily ceases in late summer/early autumn, probably also for drought-related reasons (Bollmann and Sweet 1979). Overall, however, our data would suggest that growing season in P. radiata growing under conditions such as in our study, is more determined by conditions in the late summer months, not temperature related per sé (except insofar as high temperatures may lead to drought-stressed conditions).

Data supporting fine-scale models of growth

Detailed fine-scale data on growth dynamics provide an opportunity to validate and calibrate models on new metrics, in addition to only yield. That is, process-based or hybrid models should predict not only that overall growth rates are higher in thinned stands, but also the changes in duration of season and number of growth days. This is particularly the case with fine temporal scale models such as CABALA (Battaglia et al. 2015) which are applied in predicting P. radiata growth and yield in Australia and internationally (Drew and Downes 2015; Drew et al. 2017). It makes particular sense to link fine-scale dynamics in stem radial growth and wood formation to morphological and phenological variables, in appropriately developed functional–structural models, such as the framework proposed by Fernández et al. (2011).