The ecology, distribution, conservation and management of large old trees

I. INTRODUCTION

Large old trees are among the biggest and most long-lived organisms on earth (Table 1). They have an important place in the human psyche and have many human cultural and aesthetic values (Blicharska & Mikusinski, 2014) and symbolic significance (sensu Lester, 2010). Large old trees also play an array of ecological roles and have significant impacts on the distribution and abundance of many other entities ranging from water and nutrients to entire organisms including fungi, other plants (including other tree species), and numerous species of animals. Studies of large old trees provide novel perspectives on organism growth and development (Kartzinel, Trapnell & Shefferson, 2013; Koch et al., 2015) that are not always congruent with studies of aging and development in animals (Penuelas, 2005; Issartel & Coiffard, 2011) such as increased reproductive output with size (and therefore age) in some tree species (Thomas, 2011; Wenk & Falster, 2015). The longevity, size, and patterns of spatial distribution of large old trees also can lead to new insights into evolutionary origins (Tng et al., 2012), long-term environmental changes (Phillips et al., 2008), and historical disturbance regimes (D'Amato & Orwig, 2008).

Table 1. The height and girth dimensions of the some of the world's largest old trees

However, the very attributes that confer evolutionary advantage for features such as extreme height make large old trees vulnerable to environmental stressors like drought (e.g. Rowland et al., 2015), increased frequency of lightning strikes associated with climate change (Romps et al., 2014), and human landscape transformation (Laurance et al., 2000). Indeed, we argue that large old trees are among the most imperiled organisms on earth and that their protection demands innovative approaches to management and monitoring over unprecedented time frames. Particular challenges arise from: (i) the rarity and limited spatial distribution of large old trees; (ii) the requirement for prolonged periods of stability for growth and development coupled with their susceptibility to threats such as drought, dieback, and insect attack and relatively rare catastrophic events (such as severe fires and windstorms) that are very difficult to predict; (iii) the array of potentially interacting threats that affect large old trees but that manifest in ecosystem-specific ways and thereby must be countered in ecosystem-specific ways; and (iv) the difficulty of replacing large old trees and their numerous associated ecological and cultural roles once they have vanished from an ecosystem.

In this review, we provide new insights into the ecology, function, evolution and management of large old trees. We have attempted to achieve this through broad cross-disciplinary perspectives from literatures in plant physiology, growth and development, habitat value for fauna and flora, and conservation management. We also glean new perspectives by contrasting markedly different ecosystems spanning tropical, temperate and boreal forests to deserts, savannas, agro-ecological areas, and urban environments. By reviewing large old trees in such an array of ecosystems, our review transcends earlier perspectives on these keystone structures (sensu Tews et al., 2004) such as on protecting large old trees only within old-growth or primary forest.

II. APPROACH

We present here a narrative review of many aspects of the ecology, evolution, distribution and conservation of large old trees. We searched four major electronic databases [Web of Science (1945–present), Zoological Record Plus (1978–present), CAB Abstracts (1973–present) and SCOPUS (1960–present)] on 30 May 2015 using the following search terms: big trees, canopy trees, cavity using animals, cultural trees, drought, emergent trees, forest dynamics, forest fires, hollow-bearing trees, insect outbreaks, large trees, lianas, mega trees, old trees, snags, tree cavities, tree diseases, and tree mortality. We used different combinations of search terms based on the requirements or limitations of each database. No constraints on year of publication or language of publication were imposed on the database searches. Our search returned 1563 articles. We then read their titles and abstracts and retained 742 papers relevant to our study of large old trees. All retained articles were read in full and additional works cited in the reference lists of those articles also were read. The full set of articles was then added to an extensive library of papers on large old trees that had been assembled by both authors over the past 30+ years of research on the topic.

We elected not to conduct a formal systematic review or meta-analysis (sensu Lortie, 2014) of the literature for a range of reasons, but particularly because of the cross-disciplinary nature of our review and the limited number of empirical articles at the intersection of various key disciplines that related specifically to large old trees. Our narrative review subsequently spanned the following key topics in relation to large old trees: (i) definition; (ii) key characteristics; (iii) ecological roles; (iv) cultural roles; (v) distribution and abundance; (vi) evolutionary processes; (vii) population and other dynamics; (viii) processes threatening them; and (ix) conservation and management.

III. DEFINING LARGE OLD TREES

Defining large old trees is challenging because they are an ecosystem-specific phenomenon; a large tree in an African or Australian savanna (Williams et al., 1999; Vanak et al., 2011) is far smaller than a large redwood in California (Sequoia sempervirens) (Sillett et al., 2015). Similarly, different tree species in the same ecosystem will obviously have varying maximum sizes and longevities. Definitions of large old trees can vary even among authors working in the same general ecosystems (Nilsson et al., 2002) and change according to scientific versus legal requirements (Lindenmayer et al., 2015).

Rigorous ecosystem-specific approaches have been developed to distinguish large old trees from smaller trees based on morphology, such as their diameter, crown form, and loss of limbs. However, such classifications are confined to extensively studied environments like those dominated by Douglas fir (Pseudostuga menziesii) in eastern Washington, USA (van Pelt, 2008).

Large old trees can be characterized by extreme longevity, such as those beyond 500–8000 years for individuals of some species (Owen, 2008). This includes some temperate rainforest and tropical rainforest trees (Chambers, Higuchi & Schimel, 1998; Wood et al., 2010) and the giant gymnosperms of western North America (Sillett et al., 2015). However, we argue that the use of an age-specific criterion to define older trees (e.g. Nilsson et al., 2002) is problematic because age data are very challenging to acquire for an entire tree population, and for many individual tree species. Trees in seasonal environments often have distinctive growth rings but determining the ages of such trees requires them to be cored or cut, which is impractical in many instances (Sillett et al., 2015; Tsen, Sitzia & Webber, 2015) and culturally inappropriate in others (Blicharska & Mikusinski, 2014). As an example, efforts to core one of the world's oldest trees resulted in it being destroyed (Eveleth, 2012). In individuals of some species, heart-rot and other factors can obscure growth rings and complicate dating (Banks, 1993; Waring & O'Hara, 2006). There are also challenges in accurately aging very old trees in the tropics or aseasonal environments where growth rings are absent or less pronounced (Chambers et al., 1998; Martinez-Ramos & Alvarez-Buylla, 1998). This can make it necessary to use methods such as radiocarbon dating (Chambers et al., 1998) or annual growth-increment and tree-diameter estimates from long-term demographic studies (Laurance et al., 2004a) to infer tree age.

Rather than relying on tree age, for the purposes of this review, we suggest that the definition of a large old tree should be based on its relative size (both diameter and height) and on a species-specific basis. Two key steps will therefore be important. First, it is important to establish the typical minimum diameter of reproductively mature (flowering and fruiting) individuals. Second, one can then define ‘large’ trees as being above a certain percentile, for example, the top 5% of all reproductive trees. However, we are acutely aware that the largest individuals of any particular tree species may not always be the oldest (Chambers et al., 1998; Cecile, Silva & Anand, 2013). For example, fast-growing species such as some Populus spp. can attain large size relatively quickly but may not be old. Moreover, the oldest living trees are not always the tallest trees (Cecile et al., 2013; Moga et al., 2016) with strong evidence of tree height deterioration with prolonged age and canopy failure (Lindenmayer, Cunningham & Donnelly, 1997).

Despite these caveats, we believe that for any particular species and ecosystem type, it should be generally appropriate to define large and generally older trees. For example, using growth-ring analyses, tree diameter was strongly and generally linearly associated with tree age for 10 hardwood and coniferous tree species in temperate North American forests [with diameter explaining an average of 78% (range: 47–92%) of the variation in age for each species; Leak, 1985]. Similarly, tree diameter was strongly and generally linearly correlated with tree age for four species of montane coniferous and hardwood trees from Vietnam (Zuidema, Vlam & Chien, 2011), six species of tropical hardwoods from Bolivia (Brienen & Zuidema, 2006) and five species of tropical hardwoods from Cameroon (Groenendijk et al., 2014). Notably, such relationships become much weaker if age–diameter relationships are being assessed for an entire suite of species. For example, Chambers et al. (2001) found that tree diameter explained only 25% of the variation in estimated tree age (based on radiocarbon dating of tree cores for 44 trees in central Amazonia) but this was for 16 different tree species spanning 8 tree families.

IV. KEY CHARACTERISTICS OF LARGE OLD TREES

Large old trees have an array of key attributes in addition to extreme age, height and girth, and which are not characteristic of large young trees or small young trees. These attributes include: (i) extensive buttressing (Nolke et al., 2015), which is most prevalent and extensively developed in large old tree cohorts (Cushman et al., 2014); (ii) large and numerous cavities with extensive internal volume (Remm & Lohmus, 2011) that are significantly more likely to develop in large old trees (Fischer & McClelland, 1983; Gibbons & Lindenmayer, 2002; Cockle, Martin & Wesolowski, 2011); (iii) large, well-developed crowns (Brokaw & Lent, 1999; van Pelt & Sillett, 2008); (iv) large lateral branches (Franklin et al., 1981; Killey et al., 2010); (v) deeply fissured bark and/or extensive bark streamers (Lindenmayer et al., 2000; Pederson, 2010).

These attributes vary on a species-by-species basis as well as within species depending on disturbance history, local site conditions, growth form and other factors such as tree genotype (Jordan et al., 2000). For example, the large clumps of decorticating bark streamers that characterize large old mountain ash (Eucalyptus regnans) trees (the world's tallest angiosperm) (Lindenmayer et al., 2000) are not found on giant gymnosperms. Different species can exhibit inter-specific similarities in stature but marked divergence in growth dynamics. For example, mountain ash reaches top height more quickly but then also dies more quickly than do giant gymnosperms such as California redwoods (Sillett et al., 2010). Many species of large old trees continue to add diameter with age (Stephenson et al., 2014) but there appear to be limits to maximum height, likely as a result of hydraulic pressure constraints curtailing the movement of water to the canopy (Phillips et al., 2008; Sillett et al., 2010; Koch et al., 2015). Rates of tree diameter growth tend to be much lower in older (>200–400 year old) trees than in younger individuals (Ashton, 1975; Chambers et al., 1998).

Within species, growth forms can be profoundly influenced by whether a given individual grew in a dense stand or in an open environment such as an agricultural field, with the former tending to have tall branches and the latter branching at all vertical levels. In addition, the development of some of the key characteristics of large old trees may be strongly influenced by periodic events such as fires or windstorms (Inions, Tanton & Davey, 1989). This means their development may not be linearly related to time.

Many of the key attributes of large old trees exhibit spatially variable distribution patterns. For example, dead standing trees have larger average diameters in temperate and boreal forests than in tropical and sub-tropical forests (Gibbs, Hunter & Melvin, 1993). A global study revealed that the prevalence of large old trees with cavities varies with rainfall patterns and other factors such as the abundance of primary cavity nesters (Remm & Lohmus, 2011), although this does not hold on continents like Australia where groups such as woodpeckers are absent (Gibbons & Lindenmayer, 2002). Links between cavity prevalence and rainfall patterns may be allied with the fact that precipitation also influences the abundance of large old trees (Slik et al., 2013) (see Section VIII); links between tree size and the probability of cavity development are well established empirically (Lindenmayer et al., 1993; Remm & Lohmus, 2011).

It is ironic in an evolutionary context that some of the key attributes of large old trees also can contribute to their demise. For example, extreme height can trigger hydraulic deterioration and greater levels of mortality relative to smaller trees (Rowland et al., 2015). The presence of cavities which often characterize large old trees (Cockle et al., 2011) can weaken a stem and increase its risk of failure and collapse. Similarly, the long bark streamers of giant eucalypts (Fig. 1) can readily carry fire to the canopy with trees subsequently being killed by a crown-scorching conflagration. The reasons for the development of, and presumably continued selection for, many such features remain unresolved. However, recent work from Central America suggests that large old trees benefit from the extra nutrients brought by animals attracted to key features like cavities (Voight, Borissov & Kelm, 2015).

Changes in characteristics of large old trees over time also can make them susceptible to direct human removal. For example, increasing rot and/or the presence of cavities in ageing trees can result in them being perceived as being diseased or a human safety or fire risk. In New Zealand, the development of cavities in large old trees means they are sometimes targeted for removal because they are occupied by pest species such as the common brushtail possum (Trichosurus vulpecula) which are, in turn, carriers of bovine tuberculosis that can be spread to domestic livestock (Fairweather, Brockie & Ward, 1986).

There are substantial challenges in studying certain attributes of large old trees such as features that cannot be accurately quantified using ground-based methods. The presence and attributes of cavities is an example (Cockle et al., 2011). Destructive tree dissection to explore internal cavity dimensions (Mackowski, 1987) is inappropriate in very large, rare or ecologically and culturally important specimens. Destructive sampling also can damage particular features and hence preclude their effective study (e.g. the architecture of buttresses, crowns or large lateral branches) (Nolke et al., 2015; Sillett et al., 2015). Time-consuming methods such as climbing trees to complete above-ground measurements may be required to meet some of these challenges, although emerging technologies such as laser scanning and light detection and ranging (LIDAR) can make useful contributions in some contexts (Thomas et al., 2013; Nolke et al., 2015) and have helped to reveal huge individuals of existing tree species (Yard, 2013).

In summary, large old trees support a wide range of key attributes that are either poorly developed or absent from smaller, younger trees. Many of these attributes are relatively well documented such as tree height, diameter and the presence of cavities. Others are not and remain poorly studied and understood. The canopies of large old trees have often been considered as new frontiers for the discovery of huge numbers of species (Ozanne et al., 2003). But there are other likely highly species-rich environments associated with some species of large old trees that remain almost entirely unexplored. The extensive decorticating bark microhabitats of Australian eucalypts is one such example (Fig. 1).

V. KEY ECOLOGICAL ROLES OF LARGE OLD TREES

Large old trees play a huge array of key ecological and other roles (Table 2). These include having large influences on hydrological regimes, nutrient cycles and numerous critical ecosystem processes. For example, large old trees are vital in facilitating ecosystem recovery after fire (Lutz, van Wagtendonk & Franklin, 2009) and overgrazing (Fischer et al., 2009), developing and maintaining nutrient and biodiversity hotspots in deserts (Dean et al., 1999) and forests (Jayasekara et al., 2007; Voight et al., 2015), and facilitating connectivity via promoting animal movement under climate change (Manning et al., 2009). Large old trees also play critical roles in carbon storage and therefore in maintaining forest carbon stocks (Slik et al., 2013; Chen & Luo, 2015). Indeed, large old trees are a small proportion of the number of stems in a stand but large contributor to carbon biomass (Clark & Clark, 1996; Keith, Mackey & Lindenmayer, 2009). In some forest types, the largest and oldest trees continue to accumulate large amounts of biomass (Stephenson et al., 2014), including right up to the time of apical crown collapse, even though these trees are often also the most decayed individuals (Koch et al., 2015).

Large old trees also strongly influence the spatial and temporal distribution and abundance of individuals of the same species and populations of numerous other plant (Punchi-Manage et al., 2015) and animal species and profoundly influence the structure of entire communities of organisms (Martin, Aitken & Wiebe, 2004; Stahlheber et al., 2015). The importance of large old trees as habitat for animals and other plants is so considerable that they can act as ‘biodiversity hotspots’ by supporting far more species than elsewhere in the surrounding landscape (Dean et al., 1999). Indeed, some species almost exclusively use large old trees (Le Roux et al., 2015). For some animal species, dependencies are so strong that the occurrence of large old trees acts as a robust, cost-effective surrogate for animal presence and abundance (Lindenmayer et al., 2014). Some regions such as South Africa and Australia have disproportionate numbers of species closely associated with large old tree attributes (e.g. Roberts, 1971). For example, 42% of the mammal and 28% of the reptile fauna in south-eastern Australia depends on cavities in large old trees (Gibbons & Lindenmayer, 2002).

The ecological roles of large old trees extend well beyond the immediate zone of the individual stem and influence ecological processes and spatial patterns of biodiversity occurrence and abundance at multiple spatial scales, including across landscapes (Table 2). Indeed, some species occur in a given area only because of the presence of large old trees, even in highly modified environments such as those dominated by exotic tree plantations (Kavanagh & Turner, 1994) and urban settlements (Carpaneto et al., 2010).

Large old trees also have a significant temporal footprint, playing many key ecological roles that extend well beyond their often prolonged lifespans (Stahlheber et al., 2015). For example, the animal-habitat roles of large old trees can persist for many decades and even centuries after trees have died (Rose et al., 2001). Large old trees also act as critical germination substrates for other plant species for prolonged periods after their death (McKenney & Kirkpatrick, 1999) (reviewed by Harmon et al., 1986). The presence of large dead trees in young forests regenerating after disturbance can promote the persistence of many animal and plant species that might otherwise be lost from such areas, facilitating temporal continuity in long-term species occurrence (reviewed by Franklin et al., 2000). Similarly, the significant carbon-storage role of large old trees (Sillett et al., 2010; Stephenson et al., 2014) can be retained even following high-severity stand-replacing disturbances through the long-term maintenance of carbon in dead stems (Keith et al., 2014a). Changes in soil nutrients associated with a shift in carbon from pools in large old trees to the soil for prolonged periods after tree death are also well documented (Harmon et al., 1986).

Large old trees have so many key ecological roles that it is challenging even to list them (see Table 2). Many are well known, but others remain poorly understood such as the relationships among cavity use by animals, nutrient transfer, soil fertility and tree growth (Voight et al., 2015). Many of these key ecological roles interact, such as the inter-relationships among the occurrence of fauna, the dispersal of propagules, tree growth, and carbon storage (Bello et al., 2015). For instance, chronic overhunting of tropical mammals and birds that disperse large seeds (which are associated with large, densely wooded, shade-tolerant tree species that store large amounts of carbon) is expected to lead to substantial (2–12%) carbon losses in Neotropical, African and South Asian forest, where such large-seeded species are especially prevalent (Osuri et al., 2016). However, very few studies have examined such important interactions. More are urgently needed both to better quantify the roles of large old trees and changes in ecosystem and other functions when populations of large old trees are depleted or lost entirely.

VI. CULTURAL ROLES OF LARGE OLD TREES FOR HUMANS AND ANIMALS

Humans have strong cultural connections with large old trees, both in natural areas and urban environments. They often feature in paintings, oral histories and iconic books like those by Tolkein and Enid Blyton (the Ents and the Magic Faraway tree) as well as in recent films, such as Avatar. The long lifespans of trees is illustrated by historical paintings of large old trees that still exist today, albeit where the surrounding landscape has often changed profoundly over centuries (Fig. 2). Large old trees are also prominent in pagan rituals and animist beliefs as well as being part of many religions, such as sacred sites for some communities and faiths (e.g. the locations of burial sites of saints) (Blicharska & Mikusinski, 2014). For example, large old bayan (Ficus benghalensis) trees are holy sites for Hindus in India and elsewhere in Asia (such as the Indonesian island of Bali) and are visited by tens of millions of people annually.

Human attachment to such ‘charismatic’ organisms has meant that many individual large old trees have been given unique names such as Centurion, Methuselah, and General Sherman – unlike virtually any other plants and the vast majority of animals including charismatic vertebrate ‘mega-fauna’.

Large old trees can have significant socio-economic values. For example, the California redwoods are a major tourism asset in western North America, attracting more than 330000 visitors annually and generating more than $US20 m annually in non-local visitor spending (http://www.nps.gov/red/learn/news/national-park-service-releases). Elsewhere, such as south-western Australia and southern Tasmania, areas supporting large old trees have formed the basis of major tourism ventures that attract numerous visitors each year and have rejuvenated the economy of entire regions (e.g. Winfield & Svenson, 2009).

Perversely, strong human cultural connection to large old trees and their economic value also can have fatal consequences for trees. This can occur through, for example, excessive visitation leading to soil compaction and the spread of pathogens. Such trees also can be targets for vandalism, such as by those opposed to conservation activities. In some cases, the location of large old trees is kept secret to protect them (e.g. in the case of the ‘living fossil’ wollemi pine Wollemia nobilis in south-eastern Australia).

Some animals have strong apparent ‘cultural’ connections to large old trees and may even persist in particular locations where these trees used to occur but now no longer exist. The eclectus parrot (Eclectus roratus) of Australia is an example; this species may forego breeding and remain for prolonged periods in the vicinity of a collapsed nest tree (R. Heinsohn, unpublished data). Similar observations have been made for the Lumholtz's tree-kangaroo (Dendrolagus lumholtzi) in north Queensland, where individuals stayed with their trees even after they had been clear-felled by chainsaw and bulldozer (Newell, 1999). As in human societies, animal cultural connections to large old trees also can be ‘unlearned’. For example, the establishment of nest boxes in parts of the UK resulted in populations of the tawny owl (Strix aluco) changing nest-selection preferences away from large old trees (Newton, 1998). Other bird species such as the great tit (Parus major) are known to exhibit similar patterns of ‘behavioural unlearning’ from natural cavities to nest boxes (Drent, 1984).

In summary, humans and animals alike have strong cultural connections to large old trees that can take a range of different forms. In the case of humans, it will be interesting to determine if the trend for increasing urbanization and disconnection from nature (Louv, 2005) translates into a diminished appreciation of large old trees or if appreciation of them will be maintained via books, films and other media. We note, however, that a group of senior foresters in south-eastern Australia failed to recognize stands of large old mountain ash trees despite younger stands of this same species being the primary wood-production tree cut extensively elsewhere under their direction!

VII. EVOLUTIONARY PROCESSES AND LARGE OLD TREES

In terms of their life history, large old trees are among the more evolutionarily distinctive organisms on the planet. Their extreme longevity and long generation times create both opportunities and hazards in ecological and evolutionary terms. Large size must have selective advantages or it would not have evolved in the first place. Key factors that likely favour tall trees are their capacity to dominate nearby plants in competition for light (by overtopping them and via lateral crown expansion), soil nutrients, and water. Large trees may also be less sensitive to physical damage, such as might occur via tree- or branch-fall from nearby trees (Thomas et al., 2013). Given their extensive and often-deep root systems (many tropical rainforest trees tend to be shallow-rooted because most nutrients come from litterfall decomposing on the forest floor), taller trees might also be better adapted than are smaller species to low-nutrient environments (Poorter et al., 2008). However, the largest trees on continents like Australia are confined to areas with the most fertile soils. Tng et al. (2012) hypothesized that giant eucalypts in Australia evolved in environments subject to high-severity fire necessary for germination but also characterized by conditions promoting rapid growth without a need for extreme longevity. Finally, because large old trees can provide food and shelter for a large array of animal species and substrates for the growth of many other plant species (Table 2), they might function as ‘nutrient traps’ – gaining critical nutrients such as nitrogen and phosphorus from the waste of animals that feed on, nest in, or den inside the tree (Dean et al., 1999) or benefitting from boosted site productivity generated by nitrogen-fixing plants living on their trunks and lateral branches (Lindo & Whiteley, 2011). Recent work from Central America has revealed that faeces from bats using cavities of large rainforest trees add extra nutrients that are taken up from the soil by the mesh of fine roots of these trees (Voight et al., 2015).

The extreme longevity and size of some tree species result in important trade-offs in growth, metabolism and reproduction (Penuelas, 2005; Thomas, 2011). For example, whereas the fraction of a plant's metabolism dedicated to reproduction varies markedly among species (Wenk & Falster, 2015), reproductive allocation in many long-lived tree species does not asymptote but rather continues to increase throughout their lives (Hirayama et al., 2008). In another example, Issartel & Coiffard (2011) hypothesized that extreme longevity can be explained by particularly low rates of metabolism in the stems of large old trees, which contrasts with much higher rates of energy allocation and senescence in the tree's leaves. We further hypothesize that such trade-offs through the low metabolic rate of stems may ultimately set an upper limit on tree longevity.

A long lifespan provides large trees with many opportunities for reproduction (Thomas, 2011). This ‘storage effect’ (Chesson & Warner, 1981) means that adults can withstand long periods of unfavourable conditions for seed production or juvenile recruitment. For instance, some large tree species, such as Douglas fir, rose gum, and mountain ash rely on rare, stand-replacing fires to eliminate competitors and/or provide fertile ash beds for seedling recruitment (Franklin et al., 2002; Smith et al., 2013).

Extreme size and longevity also means that trees can attain a size and canopy stature that allows them to produce massive seed crops. Reproductive allocation generally increases with tree size (Thomas, 2011), and the biggest trees can thereby overwhelm recruitment at local and landscape scales. For instance, genetic fingerprinting has revealed that a few large remnant trees produced abundant seedlings that dominated recruitment across a fragmented landscape in Costa Rica (Nason & Hamrick, 1997). Because seed and fruit production is energetically expensive (Wenk & Falster, 2015), we hypothesize that tree fertility will be generally proportional to the volume of photosynthesizing foliage, especially that in the upper canopy and emergent layers where photosynthetically active radiation is most abundant.

Extreme size and longevity also bring evolutionary and ecological risks. Reproduction in large trees is often strongly dominated by a few individuals (Nason & Hamrick, 1997; Aldrich & Hamrick, 1998; Dick et al., 2008), potentially translating into a small effective population size (N_e). Species that have small N_e could be vulnerable to factors that increase mortality of the largest, most reproductively active individuals. For example, elasticity analyses with minimum viable population models demonstrate the critical role of adult survival in many long-lived plant and animal species (Bell, Bowles & McEachern, 2003). In this sense, clearcut logging systems that fail to retain large trees, or selective logging regimes that target the largest individuals, could have major negative impacts on genetic diversity (York, 2015).

The small effective population sizes of large trees mean that they may be particularly vulnerable to genetic bottlenecks. The impacts of such bottlenecks on genetic diversity are not just a function of bottleneck ‘size’ (i.e. population size), but also bottleneck ‘width’ (i.e. the number of generations in which the species experiences a small population size; Bouzat, 2010). Species that recover rapidly from population collapses can actually retain much of their genetic variation (Bouzat, 2010). This means that efforts to propagate and breed long-lived tree species that have become rare could play an important role in their genetic management. Such efforts may become essential because the forces that are increasing large-tree mortality (e.g. habitat loss and fragmentation, logging regimes, increasing droughts, novel pathogens and enemies, altered fire regimes; see Section X) are often chronic and long-term in nature. For these reasons it may be very difficult for large trees to recover and maintain genetic diversity without active interventions.

We suggest that large old trees offer a unique opportunity to study genetic and epigenetic changes in individuals over long time spans (Bräutigam et al., 2013). In large old trees, the genotype may vary between old tissues (adapted to earlier environmental conditions) and new tissues (that have accumulated random genetic changes over time, or that have accrued mutations because of environmental impacts). Conversely, new tissues may still have the same genotype, but a different phenotype and this epigenetic change may also alter tissue functions (Bräutigam et al., 2013).

Another disadvantage of extreme adult longevity is that the climatic niches (and regeneration niches; sensu Bell, 1999) of older individuals may differ from current conditions (Smith et al., 2016). This could create inherent lag-effects that are destabilizing for population persistence, especially when climatic conditions are changing rapidly. The giant Australian flowering plant, mountain ash, is a useful example of this problem. Bioclimatic modelling has demonstrated that the species occupies a relatively narrow set of climatic conditions with other tree species dominating adjacent environmental domains (Lindenmayer, Mackey & Nix, 1996). Simulations of projected future climatic conditions suggest that areas supporting suitable bioclimatic conditions for mountain ash will be rapidly depleted by 2050 (Lindenmayer et al., 1991b). Therefore, areas currently supporting the species will likely be unsuitable for germination in the future. Contemporary and future climatic changes could create major problems for the persistence of some large tree species. Species that have large propagules, typically dispersed by larger animals, gravity or water, may be limited in their capacity to disperse latitudinally or elevationally as climatic conditions change.

VIII. FACTORS INFLUENCING THE DISTRIBUTION AND ABUNDANCE OF LARGE OLD TREES

The distribution and abundance of large old trees can be influenced by combinations of multiple factors acting over multiple spatial and temporal scales (Vanak et al., 2011; Ikin et al., 2015; Lindenmayer et al., 2016; Moga et al., 2016). Key drivers at large spatial scales include rainfall, temperature and soil fertility (Slik et al., 2013). However, in a cross-taxon analysis, Tng et al. (2012) showed that species characterized by heights exceeding 70 m occupied a very broad environmental envelope (based on precipitation and evapotranspiration) that also contained many vegetation types that did not support such tall trees. We suggest that this indicates other factors play important roles in the evolution and development of large old trees. Individual species traits are correlated with large old tree abundance; for example, the large old trees that occur at the highest population densities in tropical forests, such as the dipterocarps in Southeast Asia, are wind-dispersed tree species with high wood density (Slik et al., 2013).

At landscape and local scales, factors such as slope, aspect, proximity to watercourses, topographic wetness, soil depth, and the prevalence of herbivores can be important determinants of the occurrence of large old trees (Lindenmayer et al., 1991a; Pederson, 2010; Vanak et al., 2011; Thomas et al., 2013; Ikin et al., 2015). In some cases, places naturally supporting large old trees will be those characterized by optimal conditions for tree growth (Jones, 1997), although these same highly productive places are also those often targeted for activities like logging.

The distribution and abundance of large old trees is also driven by events such as natural disturbances. Recurrent fire can reduce or eliminate populations of large old trees from particular areas (Barlow et al., 2003; Lindenmayer et al., 2012a) as can widespread insect attack (Kashian, Jackson & Lyons, 2011; Popkin, 2015) and dieback from pathogens or toxins such as acid rain (Palik et al., 2011). Conversely, floods, high-severity fire or periodic relief from high-intensity ungulate grazing can trigger regeneration cohorts that lead to recruitment pulses of trees (George, Walker & Lewis, 2005; Moe et al., 2009; Smith et al., 2013).

Human management is also a key driver of the distribution and abundance of large old trees; logging, clearing, prescribed fire and other activities such as prolonged livestock grazing strongly influence where large old trees are found (Nilsson et al., 2006; Kauppi et al., 2015; Moga et al., 2016). The economic value of land also can have substantial impacts on the occurrence and abundance of large old trees. For example, large old trees are likely to be absent or scare in economically valuable agricultural areas and commercial forests where land is used for commodity production.

Significant cultural reasons (beyond factors associated with natural resource management and human disturbance) can also underpin the occurrence of large old trees. For example, some very large trees occur in particular places because they have been preserved for religious or other cultural reasons (Blicharska & Mikusinski, 2014). For example, the thousands of Church Forests of Ethiopia (retained because the Eastern Orthodox religion believes their churches should be ‘havens for God's creatures’) are some of the most important forest refugia in the nation (Wassie, Sterck & Bongers, 2010).

Quantifying the factors influencing large old tree distribution and abundance is critical for management such as identifying refugia where such trees have the greatest long-term chances of persistence both now and in the future (see Section XII). Yet at the same time, we suggest that at least four inter-related problems can complicate analyses of large old tree distribution and pose non-trivial challenges for widely applied methods such as species distribution modelling (sensu Elith & Leathwick, 2009). First, different environmental factors can influence trees at different stages of their life cycle (Smith et al., 2016). The large old tree growth stage may occupy only a small subset of the overall climatic and environmental envelope for a given tree species (Mackey et al., 2002). Second, the extent of past and current land uses means that many trees have been removed from large areas of their former distribution (Crowther et al., 2015) and in some cases we may know little about their past distribution. This may mean that remaining large old trees are now confined to areas where past logging or land clearing did not occur and these places may not be representative of locations potentially suitable for such trees. Third, large old tree distribution and abundance can be affected by rare and episodic events such fires or floods, the timing of which are notoriously difficult to predict. Finally, some very long-lived tree species may have germinated under environmental conditions that were markedly different from those that currently exist at that location. Hence, it may be difficult to assign the current distribution of a given species of large old tree to contemporary environmental conditions.

Despite the challenges in accurately modelling and predicting the distribution and relative abundance of large old trees, improved understanding of where they occur and why they occur there is critical for better spatial and temporal targeting of management actions (Ikin et al., 2015). We suggest that increasing use of technologies such as remote-sensing LIDAR can help to detect and map emergent and sub-canopy large old trees better, particularly for forest environments (e.g. Thomas et al., 2013). In addition, there will be value in developing more powerful species distribution models (sensu Elith & Leathwick, 2009) that help quantify the subset of a tree species' potential niche where large old trees are more likely to occur (e.g. Smith et al., 2012).

IX. POPULATION AND OTHER DYNAMICS OF LARGE OLD TREES

Adult mortality and recruitment are two key life-cycle components that strongly affect the dynamics of populations of large old trees. Adult mortality, in particular, has a critical impact on all long-lived organisms (Fritz, Bininda-Emonds & Purvis, 2009), including large old trees (Gibbons et al., 2008b). In some ecosystems, however, a paucity of natural regeneration is also threatening the development of new cohorts of trees to replace existing populations of rapidly senescing large old trees (Weinberg et al., 2011; Manning et al., 2013). Such trends are evident in many forests in India, Australia and South Africa where the introduced Neotropical shrub Lantana camara grows so densely that natural tree regeneration is impossible (Bhagwat et al., 2012).

There is considerable temporal and spatial heterogeneity in mortality patterns of large old trees and this may give rise to different estimates of mortality, even for the same broad ecosystem [for instance, compare van Mantgem et al. (2009) with Acker et al. (2015)]. Such variation can be complicated by the fact that mortality is an outcome of: (i) chronic (background, non-catastrophic) processes (van Mantgem et al., 2009) with tree death influenced by many different factors (Stephenson et al., 2011), such as competition and age (Richardson et al., 2009) but also linked to other related attributes like tree height (Thomas et al., 2013) and diameter (Lindenmayer et al., 1990); and, (ii) episodic event-based processes resulting from catastrophes such as fires (Lindenmayer et al., 2012a), droughts (Nepstad et al., 2007; Rowland et al., 2015), windstorms (Webb, 1988; Platt, Doren & Armentano, 2000), insect attack (Shore, Brooks & Stone, 2003) and damage by large herbivores (Morrison, Holdo & Anderson, 2016).

In some ecosystems, such as tropical forests of Central America and temperate forests in New Zealand and western North America, mortality rates of large old trees appear to have remained relatively stable (Richardson et al., 2009; Thomas et al., 2013; Acker et al., 2015). Conversely, populations of large old trees are declining rapidly in other ecosystems. These include: (i) forests in western North America (Lutz et al., 2009), south-eastern USA (Jones, 1997), and south-eastern Australia (Lindenmayer et al., 2012a); (ii) agricultural areas in parts of Europe (Gibbons et al., 2008b) and southern Australia (Maron & Fitzsimons, 2007; Fischer et al., 2010b); (iii) savannas in Africa and Australia (Williams et al., 1999; Vanak et al., 2011); and (iv) urban environments in Europe (Carpaneto et al., 2010), among others. A wide range of factors underpin these declines including land clearing, fire, logging, removal for human safety in urban streetscapes, and abundant exotic or native herbivore populations.

Large old tree populations are increasing in other ecosystems such as boreal and hemi-boreal ecosystems in Finland and other forested ecosystems in eastern and western USA (Kauppi et al., 2015) as well as some tropical ecosystems in Africa and South America (Fashing et al., 2004; Lewis et al., 2009a). Land-use changes, particularly reduced logging, agricultural land abandonment and/or declining incidence of fire, appear to be drivers of increases in some large old tree populations (Kauppi et al., 2015). These positive trends may reflect recovery from very low baseline population sizes (such as in the north-eastern USA; D'Amato, Orwig & Foster, 2009). Nevertheless, we argue that positive increases in some ecosystems indicate that it is wrong to describe populations of large old trees as the ‘living dead’ (Janzen, 1986), particularly given prolonged adult longevity, high levels of seed production in some very large old trees (Clark & Clark, 1996; Thomas, 2011) and the potential for targeted management to limit mortality, promote recruitment or both (see Section XII).

We suggest that significantly more long-term data sets are required to better quantify key components of the temporal dynamics of large old trees, including the factors affecting rates of mortality. Statistical methods such as zero-inflated Poisson regression (Welsh et al., 1996) and match-case control borrowed from other fields like product manufacturing and medicine can be useful in analysing temporal dynamics of large old trees. Such approaches could help researchers to predict the responses of individual trees better to rare and/or catastrophic events including the risks of premature death and collapse. A related further challenge is a need to define ‘baseline’ values for the density of large old trees in ‘natural ecosystems’ as a benchmarking exercise and to create abundance and distribution targets for conservation and restoration programs (Nilsson et al., 2006). However, many environments have a prolonged history of human use and widespread and consistent tree removal (Crowther et al., 2015). This can make it difficult to develop benchmarks for large old trees, such as for most European forests (Nilsson et al., 2002) and for temperate woodland ecosystems in agricultural Australia (Gibbons et al., 2010).

X. THREATS AND THREATENING PROCESSES

Large old trees are susceptible to a wide range of interacting threatening processes. These include human land uses such as land clearing (deforestation), the establishment of human infrastructure (such as roads and houses; Forman, 2014), selective and clear-cut logging (Lindenmayer et al., 2016; Schiermeier, 2016), agriculture, and grazing. It has recently been estimated that more than 15 billion trees are cut annually (Crowther et al., 2015), thereby significantly reducing global populations of many tree species, including large tree species. Other aspects of human land use can have significant negative impacts on populations of large old trees. For example, some have argued that because large old trees grow more slowly than younger-aged individuals, they should be cut down and replaced with young regenerating stands that are believed to fix more carbon more quickly. However, recent research suggests that large old trees continue to accumulate large amounts of biomass throughout their lives (Stephenson et al., 2014; Koch et al., 2015). Actions to protect such trees can store significantly greater amounts of carbon and lead to substantially less carbon emissions than logging and regenerating forest (Keith et al., 2014b, 2015; MacIntosh, Keith & Lindenmayer, 2015). Social and institutional instability also can threaten populations of large old trees. For example, institutional and political change in Poland is opening long-reserved forests for logging that will trigger the loss of huge numbers of large old trees (Schiermeier, 2016).

Beyond direct human land-use effects, other factors threaten populations of large old trees including drought, fire, windstorms, populations of large native herbivores, and invasive species including pathogens (Table 3). Some of these factors interact (Vanak et al., 2011) and their impacts also can affect different stages of the life cycle of large old trees in different ways. For example, very old trees containing large amounts of dead wood can be destroyed by fire (Lindenmayer et al., 2012a); yet fire also can be essential for accelerating the development of key attributes like cavities (Inions et al., 1989) and stimulating the germination of new cohorts of fire-adapted trees (Franklin et al., 2002; Smith et al., 2013).

Just as large old trees can be susceptible to disturbances, they also can contribute to disturbances, such as promoting spatial expansion of canopy gaps in forests (Young & Hubbell, 1991) and acting as a source of sparks that spread fire across landscapes (Crowe et al., 1984) (Fig. 3).

XI. MULTIPLE INTERACTING THREATENING PROCESSES

There can be important interactions among factors threatening populations of large old trees. For example, land-use practices such as selective logging, clearcut logging, and forest fragmentation can interact to increase the vulnerability of some ecosystems to destructive fires (Cochrane & Laurance, 2008; Taylor, McCarthy & Lindenmayer, 2014). It has also been suggested that large-scale drivers, such as regional or global changes in climate, can interact with land-use change, particularly habitat fragmentation and land clearing (McAlpine et al., 2007; Laurance et al., 2014). The decline of large old American beech (Fagus grandifolia) trees in parts of the north-eastern USA was caused by a combination of a fungus and an invasive scale insect (McNulty & Masters, 2005). The impacts of introduced or native pests, pathogens or disease vectors could be increased by climatic change; for instance, milder winters have allowed many species of bark beetles to proliferate in western North American coniferous forests. This, coupled with widespread forestry practices that favour particular beetle-susceptible tree species, is thought to underpin major increases in tree mortality (Raffa et al., 2008). Large old trees, which are preferentially attacked by the beetles, are especially vulnerable (Kashian et al., 2011). As a final example, logged and fragmented forests can be highly susceptible to defaunation via overhunting and population isolation (Redford, 1992), which in turn can negatively impact tree populations by reducing animal seed dispersers and increasing density-dependent seed and seedling mortality (Wright et al., 2000; Dirzo et al., 2014).

In summary, populations of large old trees are susceptible to an array of threatening processes. A fundamental part of management must be to tackle these processes. However, this will often be challenging as large old trees in many ecosystems are subject to multiple threats (Vanak et al., 2011) that can interact in cumulative, additive or multiplicative ways (sensu Didham et al., 2007). In addition, different threats will manifest in different ways in different ecosystems and tackling those threats requires ecosystem-specific actions. Multiple threatening processes also may have different effects on different life stages of trees, with large old trees particularly at risk from some factors. We therefore suggest that predicting and managing risks must extend beyond traditional approaches such as population viability analysis because, for some tree species, it is not a risk of extinction per se that is of concern, but rather the functional extinction of a particular (large old tree) growth stage. To the best of our knowledge no formal viability analysis has yet been conducted that focuses specifically on large old trees. We nevertheless argue that it is critical to predict the impacts of multiple threatening processes better because losses of the key roles played by large old trees may render some ecosystems vulnerable to collapse. For example, the rapid decline in large old tree populations in Australian-mainland mountain ash forests led to these ecosystems being formally classified as Critically Endangered under the IUCN Red List ecosystem methodology (Burns et al., 2015). There is little doubt that many other large old tree species and associated ecosystems worldwide would be assigned similar status under this IUCN ecosystem criterion. The largest tree species in the Arabian Peninsula, Mimusops laurifolia, is but one of many examples (Hall et al., 2010).

We have developed a simple conceptual framework for helping to meet some of the challenges associated with mitigating the multiple and often ecosystem-specific manifestations of multiple interacting threats facing large old trees (Fig. 4). Our framework is simple so as to facilitate its application and is characterized by explicit links and feedback loops between monitoring population decline, diagnosing the reasons for that decline (including interactions between key threatening processes), prioritizing and then implementing management actions to mitigate the impacts of these threats, and then monitoring the effectiveness of management interventions.

XII. CONSERVATION AND MANAGEMENT OF LARGE OLD TREES

Beyond the problems created by multiple interacting threats, at least five other factors create major challenges in conserving populations of large old trees. (i) In broad terms, large trees are adapted for stability and longevity but these are increasingly rare commodities in the modern world. (ii) The attributes that confer evolutionary advantages for features such as extreme height make large old trees vulnerable to various threats. For example, the prolonged time until reproduction in some species of large old trees (Lindenmayer et al., 2011; Thomas, 2011) means they can easily be eliminated from particular areas as a result of frequent disturbances such as recurrent high-severity wildfire. (iii) Large old trees are obviously less mobile than animals; they may be growing in places that were suitable for germination 500–1000+ years ago but which are no longer suitable for new cohorts of trees. (iv) Traditional approaches to management learning and identifying drivers of decline such as intervention experiments and adaptive management can be inoperable in the case of large old trees because of their rarity, cultural importance and an array of impaired ecosystem functions if they are lost. (iv) Large old trees are strongly associated with old-growth forest ecosystems, but they are also critical organisms in desert, savanna, agricultural, and urban environments, each of which has different management needs, thereby creating enormous challenges in conservation of these keystone structures.

Given that large old trees are essentially irreplaceable structures in many ecosystems (Laurance et al., 2000), we assert that offset policies (Maron et al., 2015) for them will ultimately be flawed or at best highly limited in effectiveness. As an example, nest boxes and other kinds of artificial nest sites are sometimes touted as offsets for the loss of large old trees. Yet nest boxes provide for only one of the ecological roles of large old trees (cavity accessibility) and make no contribution to other roles such as carbon storage, pulses of flowering and seed dispersal, and providing food for a large array of fauna. In addition, many animal species do not occupy artificial cavities and the approach has not been employed for the majority of taxa of conservation concern. For instance, a survey we completed as part of this review showed that nest boxes have been deployed in the conservation of less than 1% of the world's ∼1350 endangered IUCN listed cavity-dependent bird species. Moreover, recent work in urban settings has shown that nest boxes are clearly most beneficial when they are attached to large old trees (Le Roux et al., 2016), indicating that some fauna are responding to other attributes of these trees beyond cavity availability. Strategies that attempt to accelerate the development of large old trees such as stand thinning to promote growth and vine cutting to reduce competition do not promote the development of other key features of large old trees that require prolonged periods of time to accrue (such as internal levels of trunk decay) (Gibbons & Lindenmayer, 2002).

Meeting the myriad challenges associated with the conservation of large old trees will demand novel management policies and strategies at scales ranging from individual trees through to landscape and regional levels as well as over prolonged periods of time. We touch on a small subset of possible approaches in the remainder of this section.

XIII. CONCLUSIONS