- Split View
-
Views
-
Cite
Cite
Barry Gardiner, Bryan Marshall, Alexis Achim, Rex Belcher, Colin Wood, The stability of different silvicultural systems: a wind-tunnel investigation, Forestry: An International Journal of Forest Research, Volume 78, Issue 5, December 2005, Pages 471–484, https://doi.org/10.1093/forestry/cpi053
- Share Icon Share
Abstract
We report on a wind-tunnel study with 1 : 75 scale model trees designed to examine the influence of canopy structure on the formation of turbulent gusts above forests. This was to test the hypothesis that more irregular canopy structures produce less intense gusts because the change in wind speed with height at canopy top is less severe. Measurements were made of wind speeds and turbulence within and above the model forests and of the wind loading on model trees in four different silvicultural systems. The systems tested were even-aged, single-tree selection, shelterwood/group selection and strip felling. The measurements showed that the profiles of different mean wind and turbulence characteristics above the forests are remarkably similar when vertical heights are normalized by the height of the tallest tree but differences do exist within the canopy. The wind loading measurements indicated no difference between the systems in terms of stability except possibly for the shelterwood/group selection. In the shelterwood/group selection system the presence of smaller sub-canopy trees appears to reduce the loading on the main canopy trees either by providing support and increasing damping or by absorption of energy from the canopy-penetrating gusts.
Introduction
There is increased interest in Britain in the use of continuous cover forestry to meet a range of management objectives, including improvements in landscape quality and biodiversity (Mason et al., 1999). However, wind damage is a serious limiting factor to commercial forestry in many parts of Great Britain, particularly in the exposed north and west (Quine, 1995) and, therefore, it is important to understand the vulnerability of such systems to the wind.
Continuous cover forestry uses a number of silvicultural systems that will promote irregular structures such as the use of single-tree and group selection (Cameron et al., 2001) and there have been claims that irregular-structured forests are more stable than uniform forests (for a comprehensive discussion on the subject, see Mason, 2002). The possible benefits of irregular forests are a spreading of the risk within a single area and the encouragement of adaptive growth of the dominants to withstand the wind (Mattheck, 1991). Furthermore, there is evidence from studies of turbulence over forest canopies that less uniform canopies will not produce as strong or frequent gusts as a very uniform canopy (Finnigan and Brunet, 1995). These gusts which form in the region of the inflexion in the wind speed profile at the top of canopies are known to be responsible for wind damage in forests (Gardiner, 1995). However, it needs to be remembered that predicting the vulnerability of particular types of forest to damage is extremely complicated because of the number of compounding factors (Savill, 1983).
To investigate the hypothesis that less-regular forests will reduce the intensity of damaging gusts, a series of wind-tunnel studies have been carried out. The experiments were based on the wind-tunnel experiments on a uniform model forest previously carried out at Oxford University by Stacey et al. (1994) and Gardiner et al. (1997), which was an extension of wind-tunnel work by Fraser (1964) and Papesch (1984). By carrying out experiments in the wind tunnel it is possible to reduce the number of compounding factors and concentrate primarily on the influence of canopy structure.
Methods
Model trees
The model trees (1 : 75 scale) used in the experiments were of necessity a compromise because the same model trees were to be used in a variety of silvicultural systems. The model trees were designed to represent Sitka spruce (Picea sitchensis (Bong.) Carr.) growing at Yield Class 14 at an initial spacing of 1.7 m × 1.7 m (3460 trees ha−1) and with intermediate thinning which represents a typical regime for upland forestry in Great Britain (Edwards and Christie, 1981). This also allows for the increased spacing between trees growing in single-tree or shelterwood/group selection forests. Five different model tree heights were used, representing 7.5–22.5 m at full scale (Table 1) with the canopy elements (i.e. foliage and branches distributed so as to be both physically realistic and to give the correct dynamic behaviour (appropriate scaled natural frequency and drag) according to Gardiner (1989) and Stacey et al. (1994). The compromise in model trees means that they will tend to be more like trees grown under even-aged conditions and consequently diameters will tend to be slightly smaller and crowns less deep than might be expected of trees grown in uneven-aged stands. The effects on the measurements are difficult to assess absolutely but are almost certainly minor compared with the consequences of differences in tree height and between-tree spacing. Previous studies by Stacey et al. (1994) found minor differences in measured bending moments when additional lower crown elements were added to edge trees.
Tree height (full scale) (m) . | Tree height (model) (m) . | Diameter at breast height (full scale) (cm) . | Diameter at breast height (model) (cm) . | Sway frequency (full scale) (Hz) . | Target sway frequency (model) (Hz) . | Actual model sway frequency (Hz) . |
---|---|---|---|---|---|---|
7.5 | 0.1 | 11.25 | 0.15 | 0.85 | 12.7 | 12.4 |
11.25 | 0.15 | 15.0 | 0.20 | 0.51 | 7.5 | 8.7 |
15.0 | 0.2 | 16.5 | 0.22 | 0.33 | 4.9 | 4.9 |
18.75 | 0.25 | 22.5 | 0.30 | 0.27 | 4.1 | 4.3 |
22.5 | 0.3 | 30.0 | 0.40 | 0.25 | 3.8 | 4.8 |
Tree height (full scale) (m) . | Tree height (model) (m) . | Diameter at breast height (full scale) (cm) . | Diameter at breast height (model) (cm) . | Sway frequency (full scale) (Hz) . | Target sway frequency (model) (Hz) . | Actual model sway frequency (Hz) . |
---|---|---|---|---|---|---|
7.5 | 0.1 | 11.25 | 0.15 | 0.85 | 12.7 | 12.4 |
11.25 | 0.15 | 15.0 | 0.20 | 0.51 | 7.5 | 8.7 |
15.0 | 0.2 | 16.5 | 0.22 | 0.33 | 4.9 | 4.9 |
18.75 | 0.25 | 22.5 | 0.30 | 0.27 | 4.1 | 4.3 |
22.5 | 0.3 | 30.0 | 0.40 | 0.25 | 3.8 | 4.8 |
Tree height (full scale) (m) . | Tree height (model) (m) . | Diameter at breast height (full scale) (cm) . | Diameter at breast height (model) (cm) . | Sway frequency (full scale) (Hz) . | Target sway frequency (model) (Hz) . | Actual model sway frequency (Hz) . |
---|---|---|---|---|---|---|
7.5 | 0.1 | 11.25 | 0.15 | 0.85 | 12.7 | 12.4 |
11.25 | 0.15 | 15.0 | 0.20 | 0.51 | 7.5 | 8.7 |
15.0 | 0.2 | 16.5 | 0.22 | 0.33 | 4.9 | 4.9 |
18.75 | 0.25 | 22.5 | 0.30 | 0.27 | 4.1 | 4.3 |
22.5 | 0.3 | 30.0 | 0.40 | 0.25 | 3.8 | 4.8 |
Tree height (full scale) (m) . | Tree height (model) (m) . | Diameter at breast height (full scale) (cm) . | Diameter at breast height (model) (cm) . | Sway frequency (full scale) (Hz) . | Target sway frequency (model) (Hz) . | Actual model sway frequency (Hz) . |
---|---|---|---|---|---|---|
7.5 | 0.1 | 11.25 | 0.15 | 0.85 | 12.7 | 12.4 |
11.25 | 0.15 | 15.0 | 0.20 | 0.51 | 7.5 | 8.7 |
15.0 | 0.2 | 16.5 | 0.22 | 0.33 | 4.9 | 4.9 |
18.75 | 0.25 | 22.5 | 0.30 | 0.27 | 4.1 | 4.3 |
22.5 | 0.3 | 30.0 | 0.40 | 0.25 | 3.8 | 4.8 |
The model stems were injection moulded (Cademuir Tooling, Selkirk, UK) in Nylon-66 and the branches in low-density polyethylene (LDPE). The model trees are illustrated in Figure 1. The canopies for the 0.1 and 0.15 m trees used the same canopy elements as the original 0.2 m trees (Stacey et al., 1994). For the canopies of the 0.25 and 0.3 m trees the canopy mould was modified to double the weight of the canopy elements. Reasonable agreement was found between the target and actual sway frequencies of the trees, although the 0.15 and 0.3 m model trees have frequencies 16 per cent and 26 per cent too high, respectively (Table 1).
Comparison of silvicultural systems
The different experiments reported in this paper were designed to simulate an even-aged system regenerated by clearfelling, a single-tree selection system, a two-storied shelterwood system, and a variant of the even-aged system, which is regenerated by strip clearfelling. The regeneration phase of a group selection system was also simulated within the shelterwood system. These experiments were chosen to represent a range of generic silvicultural systems which are applicable to British commercial conifer forests. The following section describes each system individually:
1 Even-aged. Forest constructed out of only 0.2 m trees at a spacing of 0.0231 m (1874 trees m−2) to simulate an even-aged stand of 15 m height and a spacing of 1.73 m. This is identical to the forest used in previous wind-tunnel experiments and therefore allows comparison with previously obtained results. This will be referred to as ‘EA’ in the remainder of the text. Results for a similar forest from which 50 per cent of the trees were removed are also presented. These were measured and presented in Stacey et al. (1994). It will be referred to as ‘EA-50 per cent thinned’.
2 Single-tree selection. The forest was made up of a pseudo-random mixture of all five tree sizes (0.1, 0.15, 0.2, 0.25 and 0.3 m) and gaps to represent seedlings. This arrangement was designed to simulate a forest managed under a single-tree selection system. In the ideal case, a forest managed under a single-tree selection system should have a classical reverse J distribution of tree sizes and each size class should occupy an equal area of ground (Nyland, 1996). However, a limitation of the model forest was that the spacing between the trees could not be changed. Equal numbers of trees of each size had to be used in order to have equal areas of ground occupied by each tree size. This system will be referred to in the paper as ‘ST’.
3 Shelterwood/group selection. The forest was constructed with equal numbers of trees of two heights (0.1 and 0.2 m) to simulate a canopy with an understorey, which will replace the high canopy after felling. The pattern was a regular repetition of both sizes. Although it may differ slightly from the true characteristics of a forest managed under this system, the simulation was designed to represent a shelterwood system where the forest canopy would eventually consist of trees from two different size classes. In practice, two-storied stands would often be expected to have a greater difference between the heights of the two size classes (Nyland, 1996). However, this arrangement was chosen to allow direct comparison with the EA-50 per cent thinned stand in order to determine any stability benefit due to the presence of the understorey trees. The system will be referred to as ‘SW’. To represent a group selection system, small gaps of different sizes (equivalent to 0.01–1 ha) were then created within the forest by removing the overstorey to simulate different sizes of felling coupe.
4 Strip (narrow and wide). The forest was constructed of strips of even-sized trees with the strips increasing in height downwind. First there was an open gap (representing seedlings) followed by three heights of strip (0.1, 0.2 and 0.3 m) with the pattern repeating through the depth of the forest. Strips of 15 trees (‘narrow’ = 26 m) and 30 trees (‘wide’ = 52 m) width were tested. They will be referred to as ‘NS’ and ‘WS’, respectively. Such systems have been attempted in New Zealand to reduce wind damage (Somerville, 1989) and are typically at least 100 m wide. Such widths could not be accommodated in this study because of the room available in the wind tunnel. The WS system represents the closest to actual practice.
The wind tunnel was run at a uniform flow of 6 m s−1, which represents a 30 m s−1 storm at full scale. Two different sets of conditions upwind of the model trees were simulated:
1 Infinite forest. Roughness elements (bricks and plastic cups) were used to simulate an infinite upwind stretch of even-aged forest. By replicating forest canopy flow upwind the flow over the model forest, which occupied an area of 2.4 m2 (full size = 3.3 ha), adjusts more quickly and the measurements can be treated as occurring in an extensive forest of the particular type being simulated.
2 Forest margin. Smaller roughness elements were used to simulate an upwind stretch of open farmland/grassland. This allowed the investigation of edge effects.
Wind-tunnel measurements
A series of standard measurements were made on each of the forest designs:
1 Bending moments. Extreme and mean bending moments experienced by each height of model tree were measured using the bending moment balance designed and constructed by Oxford University (Stacey et al., 1994).
2 Turbulence. Turbulence profiles were measured within and above each model forest. Measurements were made of the mean, variance, skewness and kurtosis of the streamwise and vertical velocities. Measurements were made also of the streamwise turbulence intensity
\((\mathrm{{\sigma}}_{u}/{\bar{u}})\)and the Reynolds stress\((\overline{u{^\prime}w{^\prime}}),\)which is a measure of the drag exerted on the flow by the forest.3 Gust frequency and structure. Identification of the frequency and intensity of gusts at canopy top was obtained with a laser Doppler anemometer which simultaneously measures horizontal and vertical velocities (Marshall et al., 1999).
The bending moments on trees of different heights for the same wind speed will vary because the wind acts on moment arms of different lengths. Therefore, it is necessary first to determine the inherent differences in bending moment we would expect due simply to differences in tree height. In Table 2, the bending moments for individual trees of different heights placed in a uniform non-turbulent flow are given as a ratio of the bending moment for a 0.2 m tree (e.g. for the 0.3 m tree the value in Table 2 = Qm(0.3)/Qm(0.2)). These ratios represent the differences we would expect in the mean and extreme bending moments for conventional forests (EA type stand) with trees of different heights. It is clear that small changes in height and canopy depth can make a big difference in bending moment so that the 0.25 m tree has almost double the bending moment for the same wind speed as the 0.2 m tree. It follows that if the bending moment ratio for a particular size of tree in a specific forest simulation is lower than the bending moment ratio in Table 2, the forest can be said to provide stability benefits to that size of tree compared with conventional (EA) forestry.
Tree size (m) . | Tree bending moment/0.2 m tree bending moment . |
---|---|
0.1 | 0.34 |
0.15 | 0.58 |
0.2 | 1.00 |
0.25 | 1.93 |
0.3 | 2.41 |
Tree size (m) . | Tree bending moment/0.2 m tree bending moment . |
---|---|
0.1 | 0.34 |
0.15 | 0.58 |
0.2 | 1.00 |
0.25 | 1.93 |
0.3 | 2.41 |
Tree size (m) . | Tree bending moment/0.2 m tree bending moment . |
---|---|
0.1 | 0.34 |
0.15 | 0.58 |
0.2 | 1.00 |
0.25 | 1.93 |
0.3 | 2.41 |
Tree size (m) . | Tree bending moment/0.2 m tree bending moment . |
---|---|
0.1 | 0.34 |
0.15 | 0.58 |
0.2 | 1.00 |
0.25 | 1.93 |
0.3 | 2.41 |
This paper concentrates on the results of the bending moment and turbulence profile measurements and the consequences for the stability of different silvicultural systems. The results of the analysis of gust structure over the forest canopy were presented in a separate paper (Marshall et al., 2000). Full details of the experimental arrangements and measurement techniques can be found in Marshall (1998).
Results
Bending moments
Figure 2 presents the mean and extreme bending moments in SW, ST, EA and EA-50 per cent thinned. (Note that the results for the EA stands were measured for 0.2 m trees and derived from Table 2 for other tree heights.) All bending moments are given as bending moment coefficients as described in equation 1. In the SW system, the extreme bending moment experienced by 0.1 m trees is lower than that of the EA stand, indicating a sheltering effect from the overstorey. The extreme bending moment experienced by a 0.2 m tree is slightly higher than the values obtained with the EA stand but is much less than in the EA-50 per cent thinned stand tested previously. The presence of the 0.1 m trees between the taller trees appears to reduce the bending moment experienced by these taller trees. Within the ST model forest, the 0.1, 0.15, 0.2 and 0.25 m trees all benefited from substantial shelter provided by the 0.3 m trees, but the 0.3 m trees were subjected to higher extreme loading than in an EA forest. Similar results were observed for mean bending moments, except for 0.1 m trees which experienced slightly higher mean bending moments in the SW and ST systems than in the EA. The ratios between mean and extreme bending moments increased with increasing tree height in the simulation and were very similar for the 0.2 m trees in the EA and SW forests (Table 3).
System . | Model tree height (m) . | Extreme/mean bending moment ratio . |
---|---|---|
ST | 0.1 | 4.11 |
ST | 0.15 | 4.90 |
ST | 0.2 | 5.79 |
ST | 0.25 | 9.62 |
ST | 0.3 | 12.79 |
SW | 0.1 | 6.14 |
SW | 0.2 | 11.48 |
EA | 0.2 | 11.70 |
System . | Model tree height (m) . | Extreme/mean bending moment ratio . |
---|---|---|
ST | 0.1 | 4.11 |
ST | 0.15 | 4.90 |
ST | 0.2 | 5.79 |
ST | 0.25 | 9.62 |
ST | 0.3 | 12.79 |
SW | 0.1 | 6.14 |
SW | 0.2 | 11.48 |
EA | 0.2 | 11.70 |
System . | Model tree height (m) . | Extreme/mean bending moment ratio . |
---|---|---|
ST | 0.1 | 4.11 |
ST | 0.15 | 4.90 |
ST | 0.2 | 5.79 |
ST | 0.25 | 9.62 |
ST | 0.3 | 12.79 |
SW | 0.1 | 6.14 |
SW | 0.2 | 11.48 |
EA | 0.2 | 11.70 |
System . | Model tree height (m) . | Extreme/mean bending moment ratio . |
---|---|---|
ST | 0.1 | 4.11 |
ST | 0.15 | 4.90 |
ST | 0.2 | 5.79 |
ST | 0.25 | 9.62 |
ST | 0.3 | 12.79 |
SW | 0.1 | 6.14 |
SW | 0.2 | 11.48 |
EA | 0.2 | 11.70 |
Edges and gaps
The new measurements were able to reproduce the original measurements for the EA forest made by Stacey et al. (1994) for different positions relative to the forest edge (Figure 3; htallest = height of the standard 0.2 m tree). This gave confidence that the new data could be usefully compared with previous experiments.
The extreme and mean bending moments at different distances downwind from the forest edge for an SW forest are very similar to the values obtained from the EA forest for a tree height of 0.2 m, but the values for the 0.1 m trees are substantially less (Figure 4). The relative magnitude of bending moment coefficients for each tree size in the ST forest remains relatively consistent at any distance from the edge (Figure 5). There is a group of shorter sheltered trees (0.1, 0.15 and 0.2 m), whereas the 0.25 m trees and, particularly, the 0.3 m trees are more exposed.
For all forest structures, there is a measurable increase in extreme bending moment on the taller trees after five tree heights back from the edge while the mean values continue to decline (Figures 3 and 4). This results in a sharp increase in gust factor (ratio of extreme to mean bending moments) for the tallest trees in the canopy as illustrated in Figure 6. Conversely, the shorter trees in the ST and SW simulations continue to receive shelter from the tallest trees (Figures 4 and 5). This means that the gust factor for the 0.1, 0.15 and 0.2 m trees only increases slowly back from the edge in the ST forest, whereas the gust factor on the 0.25 and 0.3 m trees increases sharply at four to five tree heights downwind from the forest edge (Figure 7).
The bending moments of the trees of each successively taller strip in the wide-strip (WS) system (Figure 8) are very similar to the values found for the same heights of trees back from the edge of the ST forest (Figure 5). The patterns of bending moments back from the edge of each strip are almost identical to the patterns back from a completely exposed edge including the rapid initial decrease in magnitude and subsequent increase a few tree heights back into the strip. Therefore, each strip appears to behave like an individual forest edge with no obvious stability benefit arising from the presence of any upwind strips. An identical pattern was observed in the narrow (NS) system.
There was little change in the mean bending moment acting on the most-exposed 0.1 m tree at the downwind side of different-sized gaps created by removing the 0.2 m trees in the SW system (Figure 9). The sheltering effect of the overstorey rapidly diminishes so that the extreme bending moment reached a maximum at a gap size of 1.5htallest. This suggests that gusts were propagating down to the smaller trees before the mean flow had increased at this height.
Airflow
Differences in the vertical profiles of mean wind speeds for the EA, SW and ST forests were observed (Figure 10a). This indicates that the shear (gradient of wind speed with height) at canopy top was significantly modified by changes in forest structure. The SW and EA forests had very similar profiles above the canopy. However, below the canopy (z/h < 1), the wind speed in the SW forest was reduced by the presence of the 0.1 m trees. This results in a more gradual reduction in the wind speed near the top of the canopy (z/h ≈ 0.75). In the ST forest, the most rapid change in the wind speed profile occurs near the top of the 0.3 m trees (z/h ≈ 1.5). The wind speed above the ST forest canopy is lower than for the other forests up to a height of 3.5h and the shear at canopy top (z/h = 1.5) is slightly lower than over the EA forest (z/h = 1) (Figure 10a). However, when the velocity profiles were normalized by the height of the tallest trees (0.2 m for the EA and SW forests and 0.3 m for the ST forest) the profiles were found to be very similar, particularly in the vicinity of canopy top (0.75 < z/htallest < 2) (Figure 10b).
Similarly, vertical profiles of turbulence intensity and the Reynolds stress were very similar for all the forests when the data were normalized by the height of the tallest trees (Figure 11). Note that both the turbulence intensity and the Reynolds stress
Discussion
The hypothesis tested in this study was that more irregular forests would reduce the intensity of shear at canopy top and reduce the frequency and severity of damaging gusts. The wind profile data show that some modification to the wind profile below canopy top does occur, resulting in less intense shear but that the turbulence characteristics over the different forests are similar if profiles are plotted against height normalized by the height of the tallest trees. Marshall (1998) and Marshall et al. (2000) have confirmed that the canopies show extremely similar wind speed and turbulence profiles if normalized by canopy height, the height of the inflexion in the wind profile (hi) or a shear scale factor (Ls = u(hi)/[du(hi)/dz]) suggested by Raupach et al. (1996). This is supported by the fact that the tallest trees in all the simulations have very similar extreme to mean bending moment ratios. This suggests that gusts are produced close to canopy top but the intensity is hardly affected by the vertical distribution of the canopy. The ratio of extreme to mean bending moments is an important determinant of windthrow risk because trees are damaged by extremes but are believed to grow in response to the mean climate (Gardiner et al., 1997; Nicoll and Dunn, 2000).
The bending moment data illustrate how trees of different sizes are affected by the canopy structure. Trees fall into two basic categories, exposed or sheltered with the tallest trees in each simulation being exposed and the shorter trees being sheltered. This shows up most clearly in the extreme to mean bending moment ratios which are consistently high (>11) for the tallest trees and consistently low (<8) for the shorter trees. Only in the region between the forest edge and a distance of five tree heights downwind are these gust ratios lower for the tallest trees. This suggests that no matter how the forest is manipulated the tallest trees are subjected to very much the same conditions and only the shorter trees below the main canopy benefit. This is contrary to what was expected because it was thought that gusts would still penetrate below the canopy top while at the same time the mean flow would be substantially reduced so that the gust factors for the shortest trees would be higher than for the tallest trees.
The strip system appears to offer no stability benefits with the wind loading on the trees in the belts of different heights being virtually identical to what they would experience at the edge of the forest. This confirms suggestions that strip systems simply increase the number of edges within the forest rather than deflecting the wind over the top of the forest (Somerville, 1989). The single-tree selection forest provides shelter for all the trees shorter than the tallest 0.3 m tree. However, in the course of normal management operations these tallest trees will eventually be removed. It would be useful to determine how the shelter benefits to the 0.25 m trees are affected if the 0.3 m trees were removed. The risk of damage to these trees may increase following such an operation because they would not be adapted to these changed conditions but rather to the more sheltered conditions they previously experienced.
The shelterwood/group selection forest appears to offer the best possibilities for improved stability. Even though the bending moments acting on the 0.2 m trees were slightly higher than those experienced by trees of the same height in the even-aged forest, they were substantially less than those for an even-aged forest (EA 50 per cent thinned) of 0.2 m trees growing at the same density as the 0.2 m trees in the shelterwood/group selection forest. The presence of the 0.1 m trees appears to reduce the loading on the taller trees probably by increasing their damping and reducing swaying. A potential drawback of this strategy is that, as in the case of the single-tree selection forest, the tallest trees would eventually be removed in the course of normal forest management operations, leaving the 0.1 m trees subjected to more exposed conditions. The practical importance of this change on the risk of windthrow may depend on the timing of the operation. If, as in the simulations, there is a considerable size and age difference between the overstorey and understorey, the increased risk may be of little significance. This is because shorter, younger, and hence more flexible, trees are known to have the capacity to dissipate high levels of wind energy (e.g. Savill, 1983).
The wind loading on trees in different forests cannot be disassociated from the ability of the trees to resist breakage or overturning. Unlike normal engineering structures, trees modify their growth in response to their wind environment (Mattheck, 1991; Telewski, 1995; Nicoll and Ray, 1996) and, therefore, the resistance of a tree growing in a windy environment will be greater than one growing under calm conditions, all other things being equal. Unfortunately at present we are unable to relate the magnitude of tree response to wind intensity exactly; we do not have the dose-response curves. If the mean wind strength is the critical parameter then the ratio of extreme to mean bending moments becomes extremely important and those trees with low ratios will be more stable. Gardiner et al. (1997) showed that trees at wider spacings have lower bending moment ratios but they argued that wider spaced trees were more likely to be overturned by the wind. However, this was based on the extrapolation of relationships obtained from tree pulling data obtained at close spacings which may not be relevant to wide-spaced trees. Achim et al. (2005) used a winch and cable system to apply loads to trees growing at different spacings and found some benefits of wider spacing on resistance to windthrow. However, more work is required to investigate, in detail, the growth response of trees to wind loading and the resistance of wide-spaced trees to breakage and overturning. In particular, the response of tree roots and consequent anchorage to the wind environment and changes in spacing require further work.
Conclusions
All conclusions have to bear in mind that the experiments in this paper were conducted with model trees in a wind tunnel and of necessity are a simplification of a real forest. In some instances, therefore, the simulations can only provide a rough approximation to reality. For example, we were unable to simulate fully the distribution of tree diameters in a single-tree selection forest. Furthermore, this paper does not provide the answers to how to manage forests to maximize tree stability, but it does provide measurements of the differences in wind loading on trees of different sizes under different simulated silvicultural systems. Such measurements can be very valuable in developing and testing the modelling tools that are now becoming available (e.g. Ancelin et al., 2004; Cucchi et al., 2005) to evaluate the risk to the whole population of trees in stands of complex structure.
A key conclusion that we can make is that there is no simple solution which optimizes stand stability. Any change to the forest structure due to clear-felling, thinning, group selection, etc. will initially have a negative impact on the stability of the stand. Understanding the change in wind loading on the stand following any adjustment in its structure is necessary to calculate the immediate change in wind risk and how that risk will subsequently change as the trees adjust to their new circumstances. However, it is clear that changes which occur ‘little and often’ are better than those which occur ‘a lot and infrequently’. The key is the magnitude of any change in the forest. The smaller the change the less the increase in risk and the more chance the forest will have to recover. No silvicultural system is ideal for stand stability but those that employ a more subtle approach are likely to be less vulnerable overall. At the same time it must be remembered that all forests are at risk if the wind is strong enough. Balancing the risk of wind damage over the life time of the stand and the time required to meet the objectives set out for the stand is key to developing silvicultural strategies for different site and climatic conditions.
The authors wish to thank the Forestry Commission and the Scottish Forestry Trust for providing the funding to support this project. Particular thanks to all the people who helped put the model trees together, particularly Mrs G. Favager. We would also like to thank the two anonymous reviewers who provided detailed and helpful reviews.
References
Achim, A., Ruel, J.C. and Gardiner, B.A.
Ancelin, P., Courbaud, B. and Fourcaud, T.
Cameron, A.D., Mason, W.L. and Malcolm, D.C. (eds)
Cucchi, V., Meredieu, C., Stokes, A., de Coligny, F., Suarez, J. and Gardiner, B.
Edwards, P.N. and Christie, J.M.
Finnigan, J.J. and Brunet, Y.
Fraser, A.I.
Gardiner, B.A.
Gardiner, B.A.
Gardiner, B.A., Stacey, G.R., Belcher, R.E. and Wood, C.J.
Lieblien, J.
Marshall, B.J.
Marshall, B.J., Marwood, R., Belcher, R.E. and Wood, C.J.
Marshall, B.J., Gardiner, B.A., Belcher, R.E. and Wood, C.J.
Mason, W.L., Kerr, G. and Simpson, J.M.S.
Nicoll, B.C. and Dunn, A.
Nicoll, B.C. and Ray, D.
Papesch, A.
Quine, C.P.
Raupach, M.R., Finnigan, J.J. and Brunet Y.
Stacey, G.R., Belcher, R.E., Wood, C.J. and Gardiner, B.A.
Somerville, A.