EFFECT OF FRAGMENTATION AND FENCING ON THE SURVIVORSHIP OF BEILSCHMIEDIA TAWA SEEDLINGS AND REGENERATION TRAJECTORIES
Habitat fragmentation is one of the main processes responsible for the reduction of natural ecosystems to scattered isolated forest remnants (Iida and Nakashizuka 1995, Laurence 2000, Echeverria et al. 2006). There is considerable evidence that fragmentation negatively affects seed dispersal and seedling establishment, so preventing species from colonizing suitable sites (Laurance et al. 2006, Cramer 2007, Herrera and Garcia 2010, Zambrano et al. 2014). The effect of fragmentation on post-dispersal processes can be due to changes in soil composition, canopy cover, patch size, forest community structure, seed and seedling predation, altered microclimates, and continued degradation of the fragment (Bruna 2002, Laurance et al. 2002, Herrera and Garcia 2010, Ismail et al. 2014). As such, fragmentation is interrupting one of the most crucial processes in the plant life cycle. Reproduction is a crucial stage of any plant’s life cycle. A species must establish itself in enough numbers so as to permit the replacement of dying individuals. Any phenomenon that disrupts this replacement process over a sustained period of time can, ultimately, cause a species to become locally extinct (Sodhi et al. 2010).
Laurance et al. (2006) showed that the effect of fragmentation could be rapid (within 20 years) and could cause the population decline or extinction of tree species, especially those with large seeds. In other to avoid or mitigate the effect of fragmentation, land managers have used fencing as a strategy to help conserve forest fragments (Hayward and Kerley 2009). One of the most popular methods of fencing is the basic use of poles and wires which mainly act to exclude large herbivores. Other more sophisticated fencing methods, such as predator-proof fences, have proven to be very successful although their use is limited because of the high costs involved (Burns et al. 2012).
In general, the use of fencing is perceived as a restoration activity based on the belief that once the predators or herbivores are removed most of the post-dispersal processes will return to their pre fragmentation state (Reeves 2000, Spooner et al. 2002). However, herbivore exclusion can have variable effects (e.g. shrub recruitment, understory recovery) depending on the type of vegetation and past history of the fragment (Spooner et al. 2002, Burns et al. 2011). For example, Aronson and Handel (2011) found that in deciduous forest fragments in United Stated the exclusion of white-tailed deer (Odocoileus virginianus) was not enough to secure the recovery of forest fragments if invasive plant species were not eradicated too. On the other hand, Spooner et al. (2002) found that after fencing woodlands in south-eastern Australia, tree and shrub regeneration was improved together with promotion of perennial grasses and a decreased soil compaction.
In the Waikato region of the North Island of New Zealand, habitat fragmentation began at least a century ago (Gudex 1959). The Waikato region is home to some of New Zealand’s most fragmented forests, with approximately 5000 fragments each less than 25 ha (Burns et al. 2011). In general, these forest fragments are on private land (Ewers and Didham 2006), with minimal management undertaken to maintain their biodiversity (Burns et al. 2011). The most common management activity is fencing to prevent herbivory by exotic mammals (Burns 2012). Fencing can involve different approaches that range from preventing access by large exotic mammals (e.g. cattle (Bos taurus), sheep (Ovis aries) to predator-proof fences, which also exclude small mammals (e.g. rabbits (Oryctolagus cuniculus), ship rats (Rattus rattus), as is the case with the Maungatarutari Scenic Reserve (Speedy et al 2007). The former approach is the most commonly used both in New Zealand and around the world, mainly because predator-proof fencing is not economically feasible for most landowners.
In Chapter 2 I used a multivariate analysis to explore vegetation-environment relationships in a series of forest communities in the context of fragmentation and fencing. In this chapter I further assess the impact of fragmentation and fencing on podocarp-tawa forest fragments, in particular, whether fragmentation and fencing could affect Beilschemiedia tawa recruitment. We are particularly interested in the podocarp-tawa forest type because throughout most of its distribution this forest type is under exotic herbivore pressure, competing with exotic weeds and recruitment appears to be limited (Dodd and Power 2007, Smale et al. 2008). To accomplish these objectives I compared the survivorship and growth of natural and potted seedlings in unfragmented forests, and fenced and unfenced fragments. We also explored tree-by-tree replacement (Perry and Enright 2007) data to compare and contrast possible successional dynamics across sites.
The same study sites as described in Chapter 2 were used in this chapter (Section 2.2.1, Figure 2.1). They comprise nine sites near Cambridge in the Waikato region, and include three unfragmented forests, three fenced (for at least 20 years) fragments and three unfenced fragments. The vegetation at all nine sites is podocarp-tawa forest with Beilschmiedia tawa and Laurelia novae-zelandiae the dominant canopy tree species (Burns et al. 2011).
Seedling survivorship and growth
At each site I randomly selected three experimental areas to carry out growth and survivorship experiments (these sites were the foci of the vegetation data collection described in Chapter 2, Section 2.2.2). Each of these experimental areas was located beneath the canopy of a mature B. tawa tree. A cylindrical chicken-wire enclosure (1 m diameter × 1.1 m height) was constructed at the base of the B. tawa tree, and 10 potted B. tawa seedlings were placed inside it. An additional 10 potted seedlings were placed outside the enclosure within a 1 m radius (Figure 3.1). The chicken wire excluded small mammalian herbivores (e.g. ship rats and rabbits), but other herbivores (e.g. insects) could access the plants. A total of 540 seedlings (20 individuals × three areas × nine sites) were placed across the nine experimental sites. Seedlings were procured from the Taupo Native Plant Nursery, Mission Bush, South Auckland, New Zealand. The seedlings were of the same age and ranged in height from 42 – 44 cm at the start of the experiment. Two hundred and fifty seedlings came from seeds collected from Waipahihi Reserve (Taupo) in the Central North Island and 350 seedlings came from seeds collected at Battle Hill in the Wellington region (Taupo native Plant Nursery, personal communication). We randomly distributed the seedlings among the experimental sites.
The B. tawa seedlings were placed at the study sites in July 2011. Due to permitting requirements at some sites, they could not be planted and so all individuals remained in their pots at all sites. We marked each seedling on its main stem at a height 10 cm below the apical bud with typing correction fluid (e.g. Twink®, Liquid paper®). We also marked the seedlings at a height 5 cm above the ground with correction fluid as a guide for all diameter measurements (as per Carswell et al. 2012). Every two months from November 2011 to January 2013 I measured the height from the stem mark to the dominant apical bud and so derived height increments. Stem diameters were measured using callipers with the aforementioned 5 cm mark acting as a reference.
At each census seedlings were classified as being ‘alive’, ‘browsed’ or ‘desiccated’. We considered a seedling to be ‘alive’ if it had at least one green leaf; this category, therefore, included those seedlings that were attacked by insects and where the terminal bud was eaten but a new lateral shoot had become dominant. Those seedlings showing evidence of herbivory-induced mortality (i.e., with the stem completely missing) were classified as ‘browsed’ and were no longer measured. We classified the seedlings as ‘desiccated’ (dead) if the stem was present, but they did not have any green leaves or shoots (Figure 3.2).
Natural seedling regeneration
We assessed natural regeneration using 18 plots (1 × 1 m) per site (n = 162). The plots were placed near each enclosure using a randomly selected bearing and distance (Chapter 2, Figure 2.4b). Each 1 × 1 m plot was divided into four quarters within which I recorded the presence of seedlings of any subcanopy and canopy species classified into three height categories: 5 – 10 cm, 10 – 30 cm and 30 – 50 cm.
We tagged naturally occurring B. tawa seedlings (> 30 cm) in all of the experimental areas at each site (i.e. in each of three types of management). All seedlings within 5 m of the enclosures were tagged, checked for survival every six months, and any mortality recorded. In some plots the high density of seedlings made it impossible to tag them all, in which case I tagged and counted all individuals within randomly located 1 x 1 m sub-plots. The tagged forest individuals were all from Maungatautari north and Maungatautari south, as I could not find any natural seedlings near the study stations in Te Miro forest.
Tree replacement data
As noted in Section 2.2.2, I collected tree replacement data to construct tree-by-tree replacement tables (Horn 1975, Ogden 1983). We recorded the species most likely to replace the focal tree in the PCQ surveys described in Chapter 2. The replacement individuals were classified into different functional types (tree ferns, subcanopy trees, canopy trees and B. tawa). To further increase the sample size I randomly selected adult trees at each site (Table 3.1) and for each recorded the species of the two most likely replacement individuals (as per Perry and Enright 2007).
All the analyses described in this section were carried out using R-2.15.1 (R Core Team 2010), unless otherwise stated.
Seedling survivorship and growth
Tree seedling data from inside and outside the enclosures were analysed separately in order to determine the effect of this protection on herbivory. We calculated the mean height and diameter at the beginning and at the end of the study for each seedling. Survivorship was calculated from the number of individuals that fell into the alive, insect attack (but alive), desiccated (dead) or browsed (dead) categories at the end of the study. We therefore discriminated between individuals that died due to herbivory attack from those that dried out. The insect attack category was considered as part of the alive category because such attacks did not prevent the seedlings from growing (see Figure 3.2c).
We analysed seedling survivorship inside and outside the enclosures. We analysed the effect of enclosures on survivorship, without the presence of herbivory by large mammals (unfenced fragment data was not considered in this analysis). We also performed an analysis of the effect of fragmentation (unfragmented forest versus fragmented sites) and fencing (fenced versus unfenced fragments) on seedling survivorship. For this analysis I only used the survivorship inside the enclosures to avoid confounding with the effect of herbivory. In addition, I analysed the survivorship of unfragmented forest and fragmented sites using the data from inside and outside the enclosures together. We did the same to compare survivorship between fenced and unfenced fragment. These analyses were performed using a two-sided Fisher’s exact test (“stats” package).
We calculated the cumulative height increment between measurement periods for all the seedlings inside the enclosures that were alive at the end of the study. We calculated the cumulative mean increment of the three plots per site with three replicates per level of management (unfragmented forest, fenced fragment and unfenced fragment; n = 9). We did not analyse the diameter data as it did not show major changes; the mean (± 1 SD) diameter increment across all individuals was just 0.15 ± 0.38 mm (over 29 months; 0.06 mm per yr).
We analysed the data using a linear multi-effect model as proposed by Crawley (2007) with a Satterthwaite approximation for degrees of freedom. Because the data were not normally distributed I used a log-normal transformation. We considered log-normal cumulative increment on height as the response variable with the level of management as a fixed factor and seedling identifier as a random effect. We used the “lme4” (Bates et al. 2014) and “lmerTest” packages (Kuznetsova et al. 2014) in R for these analyses.
Tree replacement data
To depict the regeneration dynamics across the different levels of management I used the “sna” package (Butts 2014) to construct schematic diagrams (network graphs) of the different regeneration dynamics for each site based on transitions between the canopy species, subcanopy species and tree ferns. The main objective of this exploratory visual analysis was to determine which species were characterising the regeneration bank and which canopy species were enabling this process. We used the same species classification as described in Chapter 2, Section 2.2.3.
Seedling survivorship and growth
Of the 540 seedlings initially placed across the nine sites, 326 (60 %) were still alive at the end of the experiment 26 months later (Table 3.2). Of the 214 dead seedlings, 51 were eaten and 163 were desiccated, representing 24 % and 76 % of the dead seedlings, respectively. Irrespective of the type of management, survivorship inside the enclosures was higher than it was outside the enclosures (Table 3.2). We did not find a significant effect of the enclosures on the survivorship of the seedlings in the absence of herbivory by large mammals (p = 0.27). We did not find a significant difference in survivorship between unfragmented sites and fragmented sites (using survivorship inside the enclosure) (p = 0.45) but I found a significant difference between the fenced and unfenced fragments (p = 0.04) (using survivorship inside the enclosure). When I ran the analysis including survivorship inside and outside the enclosure I found significant differences between the unfragmented forest and the fragmented sites (p = 0.001), but we did not find a significant effect between fenced and unfenced fragments (p = 0.9).
Table 3.2 shows that mortality was higher in fragmented sites compared with unfragmented forest, especially in seedlings outside the enclosures. Enclosures had the effect of stopping mortality by herbivory and therefore mortality was reduced inside enclosures. Outside the enclosures, however, desiccation and vertebrate browsing were much higher. Herbivory was relatively unimportant for the seedlings that were outside the enclosures in the unfragmented forest (herbivory was reported only at the UF1 site, where fallow deer (Dama dama) are present at low densities (DoC 2014) and fenced fragments; herbivory-induced mortality accounted for 6 % of deaths in both cases. In the unfenced fragments mortality from herbivory was more frequent than desiccation (43 vs. 30 %, respectively). Fenced fragments had a higher rate of desiccation-induced mortality (See section 2.3.4) than did the two other levels of management.
The height growth of seedlings inside the enclosures was lower than outside the enclosures, except for the unfragmented forest sites where the growth was almost identical inside and outside the enclosures (Table 3.3).
The mean height increment inside the enclosures in the unfragmented forest was higher than the fragmented sites, 2.6 ± 0.8 cm and 2.4 ± 0.7 cm per year respectively. The mean height increment of the fenced fragments was lower than the unfenced fragments (Table 3.3).
Seedlings grew throughout the entire year (Figure 3.3), but the growth in both the fenced and unfenced fragments slowed between January 2013 and March 2013. This period corresponded to a severe drought across northern New Zealand (see Section 2.3.4). In the unfragmented forest growth was initially more rapid than at the other sites (November 2011 to May 2012), with the fenced fragment having the lowest initial growth. The growth in the fenced fragments was more consistent over time than in the unfragmented forest and unfenced fragments (Figure 3.3).
Over the unfragmented forests, fenced and unfenced fragments the lowest height growth was recorded at the Maungatautari south, FD2 and UD1 sites, respectively (see Chapter 2, Figure 2.1). The highest total mean growth recorded in an unfenced fragment was at the UD2 site (3.5 ± 0.75 cm annual increment). There was no change in the diameter of the seedlings over the course of the study (Table 3.3). Individuals in the fenced fragments had the lowest mean height increment rates of 2.3 ± 0.3 cm per year.
A linear multi-effect (LME) model showed no significant differences (at α = 0.05) in growth between the unfragmented forest and the fragmented sites (F2,7 = 4.26, MSE = 1.39, p = 0.07). Despite this result, the LME model suggested that seedlings in fragmented sites grew less than those in the unfragmented forest: 1.5 cm (log = 0.43) on average across the 29 months measurement period (Table 3.4). A second analysis did not find any significant differences (at α = 0.05) in growth between the fenced and unfenced fragments (F2,4 = 1.3, MSE = 0.5, p = 0.31) (Table 3.4).
Natural Seedling Regeneration
Based on the 1 × 1 m regeneration plots the most abundant canopy tree species in the seedling bank at most of the sites was L. novae-zelandiae, other than at the FD1 and FD3 fenced fragments (Fig. 3.4). where, respectively, Knightia excelsa and Litsea calicaris were most abundant. In general, seedlings of B. tawa were present in low numbers throughout the unfragmented forest. B. tawa seedlings were absent from two of the fenced fragments (FD1 and FD3) and one unfenced fragment (UD2). In one fenced fragment (FD2) the density of B. tawa seedlings was exceptionally high (18888 ± 14369 ind/ha), as was seedling density in general in this fragment (Figure 3.4). The most abundant subcanopy species were H. arborea and P. excelsum (Figure 3.4). H. arborea was present at all of the sites, but in one fenced fragment (FD2) was particularly abundant (65000 ± 29627 ind/ha) (Figure 3.4). Mean survivorship of tagged natural B. tawa seedlings was higher in the unfragmented forest (46.6 %) than in the fragmented sites (28 %). The fenced fragments had a lower mean survivorship (24.7 %) than the unfenced fragments (31.3 %) (Table 3.5). Mean survivorship (%) was significantly different between unfragmented forest and the fragmented sites (2 = 4.37, p = 0.04). However, the effect of fencing on survival was not significant (2 = 0.89, p = 0.42).
Tree replacement data
Graphical analysis of the tree-by-tree replacement data suggests that the unfragmented forests have similar replacement relationships across the various functional types (tree ferns, subcanopy trees, canopy trees and B. tawa) as the fragmented sites (Figures 3.5a, 3.5b).
However, the recruitment of tree ferns (Cyathea dealbata, C. medullaris and Dicksonia squarrosa) under parent tree ferns was absent from the fragmented sites. Although, the unfragmented forests and fragmented sites shared most of these replacement relationships, these relationships are weaker in the fragmented sites, meaning few individuals are recruiting (Figures 3.5a and 3.5b). The unfenced fragments had a larger number of relationships between the different functional types when compared to fenced fragments (Figures 3.5c and 3.5d). However, the relationships that exist in the fenced fragments have a higher number of recruiting individuals than in the unfenced fragment.
In unfragmented forests, the relationships between B. tawa recruiting under canopy trees (other than B. tawa) and subcanopy trees are the most important (Figure 3.4a). Even though this is the most frequently observed successional relationship, recruitment under canopy trees by other canopy trees species is also common. Subcanopy trees tend to recruit under canopy trees and, to a lesser extent under B. tawa trees.
The majority of the replacement pathways in the sites under other types of management (fenced and unfenced fragments) show that fewer individuals are recruiting under trees (dbh > 5) (Figures 3.5c and 3.5d). For example, recruitment of B. tawa under canopy trees decreases by 15 % in the fenced fragments and 17 % in the unfenced fragments, when compared to the unfragmented forest.
In the unfragmented forest the tree ferns (Cyathea dealbata, C. medullaris and Dicksonia squarrosa) are recruiting under all the functional groups but this relationship is completely absent from the fenced fragments. These results would suggest that there would be fewer or even no tree ferns present in the fenced fragments, but I found 47 tree fern individuals distributed in two of the fenced fragments (FD1 and FD3), which is comparable to the 46 tree fern individuals in the unfragmented forest over the three sites I considered.
The goal of this chapter was to determine the role of fragmentation and fencing in the survivorship of podocarp-tawa forest fragments. In particular, I wanted to determine whether fencing could help buffer the effects of recruitment problems faced by B. tawa.
The main factors that affected B. tawa seedling survivorship were desiccation and herbivory; and seedlings were more likely to die of these factors in fragments. Although I cannot be completely certain that the potted seedlings in the experimental areas died because of water stress, most of them started to die or had evidence of drought-stress after March 2013, which was just after a prolonged and intense drought period (NIWA, 2014a). This pattern suggests that protection against large herbivores (i.e. fencing) is not in itself sufficient to improve seedling survival, as water stress may have as large or even a larger negative impact than herbivory does (Table 3.2). PCA analysis of soil samples from the fragmented sites (see Chapter 2, Section 2.3.3) showed lower soil water content than in the unfragmented forests. Similarly, Zambrano et al. (2014) found that soil from forest fragments had significantly less nitrogen and water available to plants.
We did find a significant difference in the survivorship of the experimental B. tawa potted seedlings in the unfragmented forest when compared with fragmented sites. However, the single act of fencing fragments had no effect on survivorship. It is worth mentioning that seedlings in the unfragmented forest had a higher percentage of survival than those in the fenced forest fragments which was statistically significant, regardless of whether they were inside or outside the enclosures. The main cause of death in the fenced fragments was desiccation, which was higher than in the unfragmented forest or unfenced fragment. We would have expected to find evidence of lower desiccation mortality in the fenced fragments as I found evidence of a potential climate buffering effect of fencing in the analyses from Chapter 2 (Section 2.3.4).
Rates of death by desiccation in the unfragmented forest were lower than the unfenced fragments. The fact that survivorship was lower in the fenced fragments contradicts my initial expectation that seedlings in a fenced fragment would experience similar survival rates to those in unfragmented forest. On the other hand, the experimental enclosures did make a difference to the survival of the potted seedlings in the unfenced forest fragments where survivorship of seedlings outside the enclosures was greatly reduced due to herbivory, mainly by cows and rabbits. This was an expected result in an unfenced fragment. Burns et al. (2012) and Smale et al. (2005) suggest that pest-proof fencing can isolate ecosystems from exotic pests and allow landowners to make significant conservation gains.
Our data suggest that in terms of improving B. tawa seedling survival odds, unfragmented forests represent the ‘best’ management strategy and that fencing alone, without considering other abiotic stresses such as dehydration, may not suffice. The LME showed that B. tawa seedlings had a higher growth rate in the unfragmented forest in comparison with the fragmented sites, however these results were only marginally significant. Fencing did not have an effect on growth in comparison with the unfenced fragments (Table 3.4). In addition the potted B. tawa seedlings in the unfragmented forest grew faster than those in the other types of management (Table 3.3 and Figure 3.3).
However, at one unfenced site, UD2, seedlings grew at the same rate as in the unfragmented forest. This similarity in growth could be due to a variety of factors, including the local geography. The unfenced UD2 site was a natural gully formed by running water, which, in addition, provided some protection from wind exposure. Wind exposure was a difficult factor to control for, and although not always possible I strived to place all experimental areas as far from the edges of the fragments as possible. Also, due to the small size of the fragment, enclosures had to be placed near the centre of the fragment, i.e. near the gully and source of water. These seedlings were likely exposed to higher humidity due to these topographic effects. Unfortunately, the humidity data of this site was incomplete due to the loss of the sensors and I could not analyse this further.
In some, but certainly not all, fenced fragments naturally established seedlings, including B. tawa, L. novae-zelandiae and D. cupressinum seedlings were present in low numbers (Figure 3.4). In a similar study Burns et al. (2011) also found low numbers of B. tawa and L. novae-zelandiae in fenced fragments. Thus, the results obtained from the fenced fragments suggest that they are more similar to unfenced fragments than was initially expected. It is interesting to note, however, that at least one fenced site had very high numbers of seedlings of these late-successional species. The FD2 fenced fragment contained a relatively high number of B. tawa and L. novae-zelandiae seedlings, 23888 and 44444 individuals per ha, respectively. These high densities may indicate that factors other than the fencing itself may be at play, in contrast to what Smale et al. (2005) found for Dacrycarpus dacrydioides (kahikatea) forests where fencing itself was enough to increase the density of native species. Burns et al. (2011) suggested that a site’s history can be just as important or even more so, than protection against large herbivores, and the inconsistent response of fragments to fencing that I observed also suggests that site-level effects and contingencies are important.
While regeneration plots did not detect the presence of B. tawa in all the sites, seedling presence/absence data from PCQ plots from Chapter 2 (Section 2.3.1) showed that B. tawa was present in all types of management (unfragmented forest, fenced fragments and unfenced fragments). This discrepancy is expected as, when species densities are too low, abundance methods (regeneration plots) may not detect them (Joseph et al. 2006). Basically PCQ plots covered a bigger overall area, at a lower resolution, than the regeneration plots and the chances of detecting any one individual were higher. When I started this study I did not have any a priori information to guide the study design so I decided to apply both methods to detect the presence of the different species, especially B. tawa. In summary, the species that were not detected in the regeneration plots were not present or were present in low densities and were detected by the PCQ plots.
Survivorship of tagged natural seedlings was higher in the unfragmented forest than in the fragmented sites, which is consistent with the potted seedling survivorship results. In the fenced fragments survivorship was lower than in the unfenced fragments. However, the fenced fragments had the highest number of tagged individuals (169 ind. in total) in comparison with the unfragmented forest (143 ind. in total) and unfenced fragment (133 ind. in total), and all those seedlings came from the same fenced site (FD2) that had high numbers of other canopy species seedlings too (i.e. L. novae-zelandiae) (Figure 3.4).
Tree replacement data showed that most of the successional relationships between the functional types are present in the unfragmented forest and in the fragmented sites. However the relationships in the fragmented sites are weaker than in the unfragmented forest. In the fenced fragments some relationships are missing in comparison with the unfenced fragment. Even though fencing is not helping to improve the relationships among the different functional types (tree ferns, subcanopy trees, canopy trees and B. tawa) it is helping to improve the ones that are present when compared with the unfenced fragments. For example, the relationship that exists between B tawa recruitment and the parent trees and subcanopy trees is stronger in the fenced than in the unfenced fragments.
Based on the tree replacement data, it seems that canopy species and B. tawa will continue to dominate in the unfragmented forest. In the fragmented sites, dominance of canopy species and B. tawa will decrease; potentially favouring subcanopy species and tree ferns. In the fenced fragments canopy species will still dominate but B. tawa will become less prevalent; subcanopy trees will become more prevalent. In the unfenced fragment all canopy species, B. tawa and subcanopy species will become less prevalent potentially due the effect of herbivory.
In chapter 2 I discussed that the major caveat of this study was the low number of replication (n = 3). In addition to the reasons argued in the previous chapter (e.g. lack of suitable sites), we also faced logistical problems. Distributing more seedlings among more sites would have required more human resources than were available. Furthermore, it proved very difficult to procure a higher number of seedlings and the experiment as conducted involved purchasing the entire production of the only nursery that had sufficient seedlings to fulfil the specifications outlined for this experiment (e.g. age). Despite these limitations, I believe that the significant effects that I found in terms of survivorship showed that that effect was strong even with a small number of samples. In the case of B. tawa seedlings growth among the different management types had a marginal effect (p = 0.07) that would probably strengthen with a higher level of replication. In short, I believe that despite the small number of replications my results showed some trends that can help to better understand the factors affecting recruitment in forest fragments.
In conclusion, fragmentation appears to have a negative effect on survivorship, growth of potted B. tawa and on survivorship of natural B. tawa seedlings. In general, my results show that fencing does not appear to be an overwhelmingly important factor in survivorship and growth of B. tawa, although fencing did improve the abundance of seedlings of some species in some cases (FD2 site). Also, the potential climate buffer effect that was described in Chapter 2 seems to have little effect on the survivorship of B. tawa, especially under stressful conditions (i.e. drought). Clearly, the most important effect of fencing was diminishing herbivory by large mammals. As suggested by Burns et al. (2011) and Ewers (2006), and discussed in Chapter 2, it seems that the past history and the particular circumstances of each fragment are as important as fencing itself. We recommend future studies to research the importance of the past history and the current state of the fragments.
TABLE OF CONTENTS
1.2 Podocarp-tawa forest in northern New Zealand
1.3 Current management activities of podocarp–tawa forest fragments
1.4 Examples of restoration projects in New Zealand
2. EFFECTS OF FRAGMENTATION AND FENCING ON VEGETATION COMPOSITION, STRUCTURE AND ENVIRONMENT OF FOREST FRAGMENTS
3. EFFECT OF FRAGMENTATION AND FENCING ON THE SURVIVORSHIP OF BEILSCHMIEDIA TAWA SEEDLINGS AND REGENERATION TRAJECTORIES
4. IS THE LACK OF RECRUITMENT OF BEILSCHMIEDIA TAWA CAUSED BY AUTOTOXICITY?
5. MODELLING SPATIAL DYNAMICS OF PODOCARP-TAWA FOREST FRAGMENTS
5.2 Model description
5.3 Baseline analysis
5.4 Sensitivity analysis
6. A SCENARIO-BASED EVALUATION OF THE LONG-TERM DYNAMICS OF PODOCARP – TAWA FOREST FRAGMENTS
7. GENERAL CONCLUSIONS
7.1 Empirical Context
7.2 Environmental modelling
GET THE COMPLETE PROJECT