Ch 4: Insurance Effects Of Tree Diversity In Tropical Forest Restoration

Insurance Effects Of Tree Diversity In Tropical Forest Restoration: Survival And Growth During The First Decade Of The Sabah Biodiversity Experiment Context In chapters 2 and 3, I dealt with the biological mechanisms that may explain the effects of plant diversity on ecosystem functioning observed thus far. But the experiments and modelling in the previous chapters are still at a level of abstraction that makes them difficult to relate directly to real-world ecosystems. Field-scale experiments will help to elucidate which mechanisms may operate at the landscape level.

Efforts to capture greater realism in biodiversity-function research should aim to capture the biological and environmental complexity that characterises most ecosystems in nature. Such complexity makes study more challenging, but it may also provide the conditions necessary to describe the functional role of biodiversity in realworld ecosystems.

Here, I present results from a field-scale experiment that studies the mechanisms and applications of biodiversity-functioning in selectively logged tropical forests. I propose a mechanism by which plant diversity might enhance ecosystem functioning throughout this complex landscape.

I also show how this mechanism can inform the management and restoration of these human-altered forest ecosystems. Chapter Summary Swathes of tropical forest were degraded by selective logging before we could understand how they would be affected and whether they can recover. In SE Asia, enrichment planting with dipterocarp tree seedlings may help restore the forest’s structure and functioning if enough survive to aid canopy closure.

The role of tree diversity in tropical forest restoration is poorly known but planting more species-rich mixtures of dipterocarps into the forest may help ensure the seedlings avoid recruitment failure on one hand and potentially wasteful self-thinning on the other.

Get quality help now

Doctor Jennifer

Verified

Proficient in: Environmental Science

5 (893)

“ Thank you so much for accepting my assignment the night before it was due. I look forward to working with you moving forward ”

+84 relevant experts are online

Hire writer

We 4.80 analyse seedling mortality and growth over the first decade of the Sabah Biodiversity Experiment, to understand whether planting more diverse mixtures may provide such an insurance effect. Mortality was high overall but varied widely among species; their differences in survival were consistent over time. Those with high survival had slow growth and vice versa. Slow-growing species with high survival had more dense wood.

Mortality and growth varied spatially across our 500 ha experiment but species responded to changing conditions in independent ways, a key ingredient for an insurance effect. As expected, given the wide spacing between seedlings, species performance did not differ between monocultures and mixtures. But their independent preferences for site conditions may provide an insurance effect of diversity: sites planted as monocultures of poorly surviving species may fail to restore stem basal area and canopy closure—planting mixtures may ensure there is always a species that can perform. Long-term monitoring is needed to determine whether mixing species with different vital rates can insure against this recruitment failure while avoiding potentially wasteful self-thinning in the best performing monocultures. Introduction After twenty years of debate there is now broad consensus that biodiversity has a positive effect on the functioning and stability of ecosystems (Cardinale et al.

2012; Tilman et al. 2014). However, this consensus is founded on a first generation of research from grasslands and other easily manipulated systems, which are often shortterm, small-scale and highly controlled experiments. We need a next generation of experiments to quantify how biodiversity affects ecosystem functioning in more natural and applied situations, including habitat restoration (e.g. Bruelheide et al.

2014). Study of the relationship between biodiversity and the functioning of tropical systems has begun (Ewel & Bigelow 1996; Haggar & Ewel 1997; Lovelock & Ewel 2005; Potvin & Gotelli 2008), but the rich diversity of these systems plus their provision of important ecosystem services means further study is required. To help fill this knowledge gap for SE Asian forests we established the Sabah Biodiversity Experiment in Malaysian Borneo (Hector et al. 2011b). The project—a collaboration between ecologists, tropical foresters and a carbon offsetting scheme—tests the effects of tree diversity on the restoration of selectively logged forests which were enrichment planted with once-harvested species to return fully functioning ecosystems.

There are over 400 million hectares of logging estates in the tropics, and at least 20% of existing tropical forests were selectively logged between 2000 and 2005 (Edwards et al. 2014). These recently logged forests now cover larger areas of land than primary forest in most regions (Edwards et al. 2014; Laurance et al. 2014).

In SE Asia primary forest exists in inaccessible uplands but little remains in lowlands due to unprecedented rates of deforestation (Sodhi et al. 2010; Gaveau et al. 2014). In Sabah Malaysia, conversion to oil palm agriculture has driven forest extent from 86% in 1953 to below 50% (McMorrow & Talip 2001). Premature harvesting of previously logged areas has also been common (Reynolds et al.

2011). This forest loss and degradation is threatening many SE Asian plants (Sodhi et al. 2010), including members of the Dipterocarpaceae who dominate these forests and are a valuable timber source (Maycock et al. 2012). But a growing body of evidence is showing that selectively logged forest harbours greater ecosystem value than agricultural land, and even fragmented primary forest within an agricultural landscape, so long as they are not too degraded (Berry et al.

2010; Putz et al. 2012; Edwards et al. 2014). Some are calling to protect these vast areas from further land conversion, and maximise their conservation value by replanting with dipterocarps (Kettle 2010). Enrichment planting is the practice of replanting seedlings into residual stands of selectively logged forest that are poorly stocked with the harvested species.

The primary aims are to restock target species, either permanently or for future harvests, while rehabilitating the degraded ecosystem. Tropical tree species are often naturally found at low density so replanting logged species may help to supplement natural regeneration and overcome recruitment limitation. This might be particularly necessary for dipterocarps whose reproductive biology (irregular masting reproduction, low dispersal, no seedbank, and vulnerable seedling banks) may jeopardise regeneration. Enrichment planting aims to overcome dispersal and recruitment limitation, speeding the return to tall, complex canopies. However, evidence for the effectiveness of enrichment planting is incomplete despite its widespread implementation since the 1960s (Chan et al.

2008). Success will depend on how many natural seedlings remain and whether enough planted seedlings survive to recreate the pre-logging canopy structure. Improvements in enrichment planting techniques have helped to boost success, but progress is far from complete (Chan et al. 2008). The effectiveness of enrichment planting can only be assessed once key evidence gaps have been filled.

One key issue is whether effectiveness is hampered by planting at low diversity; single species, or mixtures of a few species, are typically enrichment planted over large areas. The vital rates (survival and growth) of commonly planted species and their environmental preferences are not well known, limiting the ability to match species with favourable planting sites. Whether speciessite matching is at all feasible is unclear since survival may vary over such fine spatial scales that its implementation is unrealistic. The role of tree diversity and how species combine in mixed-species plantings has received even less study. The dipterocarp forests of SE Asia therefore provide a unique situation where enrichment planting schemes can be used to investigate the role of tree diversity for forest functioning and restoration.

The Sabah Biodiversity Experiment, part of a network of tree diversity experiments (Verheyen et al. 2015), addresses these questions by manipulating the identity, composition and diversity of enrichment-planted dipterocarps to assess how effectively this practice can restore the functioning and stability of these selectively logged lowland rainforests (Hector et al. 2011b). Because of the wide spacing of planted seedlings (3 seedlings per 10 m of planting line) and the background vegetation left after logging, we did not expect to see biodiversity effects based on complementary species interactions this early in the experiment. Even so, enrichment planting does provide the potential for an insurance effect based on intrinsic differences in seedling mortality and growth.

The usual practice in enrichment planting schemes is to stock large areas with low-diversity mixtures, often monocultures of seedlings available from nurseries. Monocultures run the risk of recruitment failure if the planted species turns out to be a poor survivor under the given circumstances; tree density may become so depleted that the planting does nothing to supplement natural regeneration (Hector et al. 2011b; Saner et al. 2012). Planting more diverse mixtures might provide a simple insurance against such recruitment failure.

Diverse mixtures might also provide a more efficient use of seedlings by avoiding wasteful levels of selfthinning of species with high survival rates. Here, we present the mortality and growth of the first cohort of seedlings during the first decade of the project. We propose and test a potential insurance effect of tree diversity in replanting schemes, in which mixtures avoid the potential two-fold cost of monoculture planting: recruitment failure of the worst surviving species and wasteful self-thinning of the best. Methods Data collection The Sabah Biodiversity Experiment (Hector et al. 2011b) covers 500 ha in the Malua Forest Reserve, a region of selectively logged forest bordering primary forest at Danum Valley Conservation Area located in the Malaysian state of Sabah, Borneo.

Malua Forest Reserve, part of the Yayasan Sabah Forest Management Area forest concession, was logged in the late 1980s. Malua was first logged between 1984 and 1986 and, except our experiment site, again in 2007. The anticipated harvest cycle is 50–60 years, the estimated time needed to achieve a species composition similar to unlogged forest (Yamada et al. 2013). The Yayasan Sabah (Sabah Foundation) concession also includes the 30,000 ha Innoprise-FACE Foundation Rainforest Rehabilitation Project (INFAPRO).

The Sabah Biodiversity Experiment replicates INFAPRO’s enrichment-planting techniques where possible to maximise the potential for practical recommendations. The experiment contains 124 four-hectare (200 x 200 m) plots, split between two blocks that are north and south of a logging road (see figure 1 in Hector et al. 2011b or Supplementary Material SA1). There are 60 plots in the north block and 64 in the south block. Seedlings were planted 3 m apart along parallel planting lines in a stratified randomised design.

Each plot contains 20 planting lines spaced 10 m apart. Ninety-six of these plots comprise a diversity gradient treatment. The remainder are comprised of 12 unplanted controls, and another 16 sixteen-species mixtures that were given enhanced climber cutting (explained below). Only the 96 diversity gradient plots are analysed here. The diversity gradient manipulates species richness using a factorial design, including replicated species compositions within species richness levels (1, 4 and 16 species).

Species compositions within the four-species mixtures provide a gradient of generic richness and are designed to produce a range of canopy structures once the planted seedlings mature. Each species richness level has 32 plots. In the enhanced climber-cutting treatment, climbers are cut throughout the whole plot, not just along the lines as in standard enrichment line planting—this is said to improve recovery time during restoration. As with standard enrichment-planting practice, following early mortality, the initial planting cohort of seedlings (cohort 1 planted 2002–2003) were supplemented with a second replanting cohort (cohort 2 planted 2009–2011). Across both cohorts, a total of 96,369 seedlings have been surveyed.

Due to the scale of the experiment each full survey took two years to conduct (see Supplementary Material for histograms of seedling age). Although, to complement this large-scale but time-consuming monitoring, a subset of plots have been more intensively sampled—six extra surveys to date (Philipson et al. 2014). Therefore in 12 years there have been two full surveys of all seedlings. The first survey (Nov 2003–May 2005) included only the first cohort of seedlings whilst the second survey included both cohorts (Nov 2011–Sep 2013).

Here we analyse survival and growth of the first cohort using both full surveys (survival of the second cohort can only be assessed after the next survey). We recorded survival and size for every seedling each time they were visited. We measured basal diameter (2 cm) and, when they were tall enough, diameter at breast height (1.3 m). During early surveys we took ancillary measurements, such as descriptions of each seedling’s local environment. Study species The 267 species of dipterocarp known to occur in Malaysian Borneo belong to nine diverse genera—and roughly half of these species belong to one genus, Shorea (Ashton 2004).

The sixteen species we planted are Dipterocarpus conformis Slooten, Dryobalanops lanceolata Burck, Hopea ferruginea Parij, Hopea sangal Korth., Parashorea malaanonan (Blanco) Merr., Parashorea tomentella (Blanco) Merr., Shorea argentifolia Sym., Shorea beccariana Bruck, Shorea faguetiana Heim., Shorea gibbosa Brandis., Shorea johorensis Foxw., Shorea leprosula Miq., Shorea macrophylla Ashton, Shorea macroptera King, Shorea ovalis Korth., and Shorea parvifolia Dyer (Supplementary Material ST1). All species except D. conformis are members of the Shoreae tribe—though Dipterocarpus is sister to Shoreae and there is mixed support for the monophyly of Shoreae within this clade (Kajita et al. 1998). Shorea, Parashorea and Hopea form a polyphyletic group.

Several sections within Shorea, covering multiple commercial timber types, are represented within our species (Ashton 1982). Our species were selected because they (i) represented those found in the surrounding forest, (ii) cover a range of traits and ecological strategies, and (iii) were sufficiently available as seedlings when we planted. The seedlings initially planted were sourced from INFAPRO; a dedicated project nursery was later set up to cultivate newly collected seedlings for the second cohort. Other species that were too scarce for the main experiment have been studied in smaller associated experiments manipulating light and water (Saner et al. 2010; Paine et al.

2012; Philipson et al. 2012; O’Brien et al. 2013, 2014, 2015), producing data on a total of 28 species. SE Asian dipterocarps are emergent, shade-tolerant trees concentrated in aseasonally wet evergreen lowland forest on well-drained soils. They are mostly found below 800 m altitude, and their abundance and diversity declines above 400 m.

They produce seeds during mast fruiting events. If these seeds do not germinate quickly they die due to heavy browsing (Hautier et al. 2010) or recalcitrance (O’Brien et al. 2013). Surviving seeds produce a seedling bank.

Although dipterocarps are one latesuccessional functional group there is evidence for a trade-off, particularly at the juvenile stage, between growth and survival (Philipson et al. 2014). Dipterocarps reach peak biomass, density and species richness on yellow or red lowland soils, where they comprise >50% aboveground tree biomass and >70% of emergent individuals (Ashton 2004). It is the dipterocarps that give these forests their exceptionally high biomass (Banin et al. 2014).

In the 1980s, dipterocarps provided 25% of tropical hardwood supply worldwide, and 80% of this share came just from Shorea (Ashton 1982). Juveniles are easily disturbed during logging, undermining their regeneration; they may not return for centuries in heavily degraded soils (Ashton 2004). Palaeoecological work has shown that SE Asian tropical forests often take centuries to fully recover from disturbance—longer than any other tropical region (Cole et al. 2014). Data analysis Every seedling had its survival and size recorded in each survey (1 = alive, 0 = dead).

For cohort 1 there are two surveys of all seedlings, with median age of 752 days at survey 1 and 3691 days at survey 2. Growth and survival of seedling cohort 2 cannot be analysed here as they require a second measurement, which we will have once we conduct a third full census (one of the programme’s next priorities). We could not analyse growth and survival at the seedling level because there was spatial correlation within 200 m that would undermine our inference. We aggregated data to the specieswithin-plot level to remove within-plot spatial correlation (see Supplementary Material SA2). On average, these species-within-plot aggregations were based on 80 survival observations.

This left us with 1336 plot-level observations and a minimum of three replicates for any species within a species composition. Survival and growth were modelled by fitting two linear mixed-effects models. Survival was assessed as the proportion of planted seedlings remaining in a plot in a given survey. Growth was assessed as change in average log-transformed basal diameter of those surviving seedlings. We kept explanatory variables as consistent as possible to help compare survival and growth: species-within-plot was fitted as a random effect (one variance for a factor with 672 levels), and the fixed effects were a three-way interaction between species identity (16 levels), species composition (33 levels), plus a representation of survey time that differed between models (as species do not occur in every composition, there are 96 species-within-composition levels).

For survival, survey time was a factor with two levels, giving the average proportional survival since planting for each survey (survey 1: 0–2 years since planting; survey 2: 0–10 years since planting). For growth, instead of treating surveys as a factor, survey time was continuous (number of days since planting). The slopes of change in log size between the two surveys gave our estimated growth. Growth was therefore analysed using a subset of the survival data, using only seedlings alive at both surveys (1122 plot-level observations). Both models estimated 193 parameters: one additional variance component and 192 fixed effects.

For each of the 96 species-withincomposition levels, the survival model estimated two intercepts, whereas the growth estimated one intercept and one slope. These models were fitted using lme4 v1.1-7 (Bates et al. 2014) in R v3.2.1 (R Core Team 2014). Their model formulas were, meanSurvival ~ species*spComposition*surveyNumber + (1|plot:species) meanLogBasalDiameter ~ species*spComposition*meanSeedlingAge + (1|plot:species) To quantify the overall performance of each species, among species compositions, we took the average of their population-level predicted values. These species-level estimates of growth and survival in each survey were used to assess how strongly species differ, whether their ranking in survival remains consistent over time, and whether they trade off survival against growth.

We correlated survival and growth with wood density and specific leaf area, which were estimated from previous experiments within our site using the same seedling cohort (Philipson 2009; Philipson et al. 2012; O’Brien et al. 2013, 2014). Spatial variation in species survival was quantified using predictions from the random effect—plot-level deviations from the overall survival for each species. By tracking the relative effect of every 16-species mixture plot upon each species we could show whether species were responding differently to the same conditions.

We tested whether species were truly responding differently to plot conditions by fitting two non-nested models: one allowing species-specific responses to plot conditions and another assuming species respond equally (species-specific responses, (1|plot:species), were compared with (1|plot)). We compared these models by seeing how much AIC improved when species-specific responses were allowed (Pinheiro & Bates 2000). Finally, we summarised overall plot-level performance, averaging across species, as the density of surviving seedlings; this was plotted against species richness, and then broken down into specific compositions, to assess whether a spatial insurance effect might confer an advantage to planting more diverse tree mixtures. Results Seedling survival and growth varied widely among species, after two and ten 4.91 years since they were planted (Figure 4.1 top panel; for estimates see Supplementary Material SA2). The proportion of first-cohort seedlings that survived overall was low (0.36 after two years and 0.12 after ten years).

Species ranking in survival was consistent over the two surveys (Figure 4.1 bottom panel; Pearson’s r = 0.79). After ten years the seedlings had grown to an average apex height of 1.25 m (max. = 12 m) and average basal diameter of 1.6 cm (max. = 28 cm). There was a trade-off between survival and growth among species—though this fades over time as mortality mounts and proportional survival shrinks (0–2 years, r = -0.63; 2–10 years, r = -0.43).

Figure 4.1. Growth trades off with survival. Growth and survival during the two survey intervals (2002–3; 2008–11). Above: Proportional survival (at 2 and 10 years since planting) versus growth rate (change in log basal diameter between survey 1 and 2) for the 16 species, showing a negative trade-off. Below: The proportion of seedlings that survived 0–2 years versus 2–10 years since planting are positively correlated, showing consistent ranking in survival over time.

Species codes are shown in Supplementary Material ST1. Grey regression lines show overall trends. Figure 4.2. Survival spatially varies and species respond to environmental variation in different ways. The species identity by environment interaction for seedling survival in the 16-species mixtures.

Points represent the average survival of a species in a plot relative to the average of that species over the whole experiment (taken from the plot:species random factor)—so positive values show plots with better-than-average survival. Grey lines join particular plots, illustrating the varying performance of different species in the same conditions. The red line gives one example: while this plot shows above-average survival for some species (e.g. D. conformis shows its highest survival) other species experience below-average survival.

We correlated survival and growth after ten years with traits that may link to ecological strategies or insights for planting practice. Wood densities for all species (excluding H. ferruginea whose high mortality prevented trait estimation) positively correlated with survival after ten years (r = 0.78) and negatively correlated with growth (r = -0.50). Specific leaf area weakly correlated with survival (r = 0.06) and growth (r = -0.14). A buffering effect of increased tree diversity may occur if species show varied responses to spatial variation and respond independently or asynchronously to one another.

All species showed substantial spatial variation in survival and growth across the 500 ha experiment, though some more than others (Figure 4.2). The species that showed the most variable survival across the experiment were not necessarily those that showed the most variable growth (see Supplementary Material SA2 to compare growth with the survival in Figure 4.2). Species survival also responded to plot-level conditions in different ways, so the most favourable location for one species could be one of the least favourable for another (follow the red line in Figure 4.2). When speciesspecific responses to plot conditions were allowed, AIC and BIC both reduced by ~19, suggesting species truly respond differently to plot conditions. Figure 4.3.

Seedling density is less variable in more diverse mixtures. Density of surviving firstcohort seedlings as a function of plot species richness. The data, showing the number of seedlings per ha within each plot, are summarised with box and whiskers: boxes show the 25th, 50th and 75th percentile density, and whiskers extend to the most extreme density values within 1.5 times the inter-quartile range. While the median density remains constant, variation among plots decreases as species richness increases, particularly after two years (survey 1). Figure 4.4.

Mixtures avoid detrimentally low and wastefully high densities. Density of surviving first-cohort seedlings after ten years, in 16-species mixtures and monocultures. Small points show densities in each plot and large points are the means. The grey band shows the 95% confidence interval of the 16-species mixture mean. The confidence interval for the probability of survival, p, was obtained using the Wilson method (Held & Sabanés Bové 2014), then expressed as the number of surviving trees per ha, (p∙n)/4.

Many monocultures show extreme densities compared to mixture mean. We cannot assess this insurance effect conclusively due to the early stage of the experiment and the lack of data for the second seedling cohort. However, at the first survey the highest and lowest densities of surviving seedlings were seen in monocultures (Figure 4.3 and Figure 4.4). The species mixtures are no different from the monocultures on average, just as we expected, because interactions between 4.95 seedlings are not yet strong. The variability in density, however, does decrease as species richness increases, particularly after two years (Figure 4.3).

And the replicated monocultures of a given species were often more variable than what we saw among the 16-species mixtures (Figure 4.4). Planting more diverse mixtures did initially buffer the density of surviving seedlings after two years, but mortality continued over the following eight years and average density decreased within all species richness treatments. Whether there is a long-term insurance effect of diversity on forest restoration will depend on the immediate and long-term survival of both seedling cohorts. Discussion Despite the early stage of the Sabah Biodiversity Experiment, several clear results emerge from our analysis of survival and growth during its initial decade. Firstly, we found a clear life-history trade-off between survival versus growth and consistent differences amongst our sixteen dipterocarps in their positions along this trade-off (Figure 4.1).

Secondly, not only did species differ on average but they also responded differently to spatial variation, suggesting they specialise on different conditions (Figure 4.2). Thirdly, as expected given the wide spacing of the planted seedlings, there is no evidence of complementary species interactions in mixtures yet (Figure 4.3). Finally, the most extreme high and low seedling densities are found in particular monocultures (Figure 4.4). We discuss each of these points in turn before considering their relevance for enrichment planting schemes and the potential insurance effect of tree diversity in forest restoration. The trade-off between survival and growth The results of our more general analysis here support the conclusions of an earlier, more detailed analysis that identified a trade-off between growth and survival (Philipson et al.

2014). Our earlier work showed that these dipterocarps trade off survival against growth generally, irrespective of the light conditions they are exposed to: all species were affected by light, but their ability to grow or survive relative to others remained unchanged. This follows work in other tree communities showing that the growth-survival trade-off is a major axis of life-history variation (e.g. Grubb 1977; Pacala et al. 1996; Wright et al.

2010). While species differences in survival rates were consistent over time in our study, species estimates of survival are not completely consistent with other studies. When comparing our survival estimates with those at the nearby INFAPRO enrichment-planting sites, the same species observed over a similar timescale experienced unrelated levels of mortality (Godoong et al., unpublished data). Among the species found in both our experiment and the INFAPRO site, those that have shown the best survival so far in our experiment have not been the best survivors at INFAPRO. For example, D. lanceolata was clearly the best survivor at INFAPRO after 13 years, with approximately 30% survival—twice the survival rate shown by any other species at the time. However in our experiment, S. ovalis, also planted at INFAPRO, attained higher survival than D. lanceolata. These differences between our experiment and INFAPRO could be due to numerous factors, including age of seedlings and site-specific conditions (see below). Trait-mediated trade-offs Various authors have hypothesized links between demographic rates and plant traits, in particular wood density and specific leaf area (King et al. 2006; Sterck et al. 2006; Kraft et al. 2010), although some are more cautious (Anten & Schieving 2010; Larjavaara & Muller-Landau 2010; Paine et al. 2015). Both the results of this analysis and of our earlier work support the link of the survival versus growth trade-off with wood density, such that species with denser wood have higher survival but lower growth rates. On the other hand, both our current analysis and earlier work found no association with specific leaf area. We did find that average survival of species were positively correlated with both total biomass and root mass ratio of the initial sample of harvested seedlings, as is often found (Chan et al. 2008). Related experiments at the same site (with 8 dipterocarps including 7 of the species used here) have shown that individuals and species with higher levels of non-structural (soluble) carbohydrates survive longer under extreme drought—a major cause of tropical tree mortality that may be exacerbated by climate change (O’Brien et al. 2014, 2015). Extending this work on non-structural carbohydrates and drought survival to the Sabah Biodiversity Experiment is a next priority for the project. Spatial heterogeneity and species-by-environment interactions One strategy to improve enrichment planting survival rates may be to plant species in sites that will optimise their growth and survival based on their known ecology (Kettle 2010). If species respond to large-scale differences amongst sites in differing ways then there may be potential. However, if site conditions are fine-grained they may be at too small a scale to make species-site matching practical. Spatial variation in mortality over the 500 ha Sabah Biodiversity Experiment site was 4.98 substantial. Elevation is generally highest in the most northerly and southerly areas, decreasing towards the road separating the north and south blocks. There were no discernible effects of the road, nor the river (Supplementary Material SA2). Within distances of 200 m or less (within plots), percentage survival commonly varied ±10% from average and more extreme fluctuations were twice this in magnitude. Within plots, seedlings were planted with 3-m spacing along parallel lines 10 m apart. Survival tends to be more similar within lines than among them as would be expected given the shared conditions along lines (e.g. conditions when the line was planted; damage by elephants that use lines to move through the forest; canopy openness and light levels). So, while our results do show that species respond to site conditions differently, the most relevant conditions may vary at too fine a scale for species-site matching to be practical. This is supported by studies relating seedling survival to micro-topography and associated differences in soil moisture (Born et al. 2015). Lack of interactions between species in mixtures As we expected, we found no evidence for an effect of plot species richness (or composition) on growth or survival. This is because there is limited scope for interactions between trees during the early stages of the experiment given their size relative to the planting density (pre-mortality) of 3 x 10 m. However, while the average seedling height to apical meristem in 2013 was only about 1 m (including the younger second seedling cohort, see Supplementary Material SA2) some of the larger survivors from the first cohort were already approaching sizes (12 m) where they may interact with neighbours, especially along planting lines. Regular measurement of survival and growth will allow us to detect when enrichment-planted seedlings start to strongly interact. Enrichment planting We found high mortality for the first seedling cohort, with only 35% remaining after two years and 12% after a decade. Rapid mortality is typical for enrichment planting—hence the replanting—but levels in our experiment are higher than some rates reported elsewhere (Chan et al. 2008) and for the nearby INFAPRO (~50% at 3 years; MS and SWY) and INIKEA projects (~30-60% at 10 years). Intensive maintenance after planting improves survival rates (Ådjers et al. 1995) so it is possible that some enrichment planting schemes may achieve better survival through this route. The state of the planted seedling stock also impacts survival and growth, so it will be interesting to compare the mortality reported here with the second cohort, which came from different stock. A new survey that includes measurement of the second cohort is therefore a priority for the project. One caveat when comparing our results with the wider literature is that our seedling densities are based strictly on the enrichment-planted seedlings, whereas other projects may inadvertently or deliberately also include naturally-occurring seedlings. Potential insurance effect of tree diversity in forest restoration Due to the small seedling size relative to the planting density we knew there would be limited scope for interactions between species in mixtures during the initial stage of the experiment. However, we did anticipate that species differences in survival rates could provide the basis for an insurance effect of tree diversity, in which species mixtures avoid the potential recruitment failure of monocultures with low survival and the potentially wasteful self-thinning in stands of species with the highest survival. Our results show how the survival rates of these species are variable and susceptible to spatial variation, which could generate such an insurance effect. However, it is too early to predict the eventual densities of different monocultures and mixtures, or what levels of self-thinning and recruitment failure will result. In comparison, the INFAPRO project’s original goal was to reach a density of 15–30 mature harvestable (>60 cm dbh) dipterocarps per ha to replace the trees that logging operations removed (the INFAPRO area has since been protected from commercial logging). The trade-off between survival and growth means that these two contributions to stem area tend to cancel out, producing some plots with a higher density of smaller trees and others with a lower density of larger trees. However, in the long term we expect fast-growing species (lower survival and wood density) to be replaced by slow-growing species (higher survival and denser wood). Regular long-term monitoring will be essential to identify the role of tree diversity in the functioning of these ecosystems and its underlying biological mechanisms. Next Page – Ch 5: The Effects Of Organic Farming On Biodiversity Previous Page – Ch 3: How Do More Diverse Plant Communities Increase Ecosystem Functions