Seeing the Trees in the World’s Forests: An Extension of the Forest Transition Concept

The forest transition – or forest-area transition – has been put forward as a land-use concept by A.S. Mather in 1992 (The forest transition. Area 24, 367-379), to describe the historical trend generally observed in the forest area of developed countries, embodied in a V-shaped curve of the forest area over time, and that may serve as a paradigm to understand and anticipate deforestation in the developing world. Well in line with a geographical approach to forests, forest transition has thus been defined as one-dimensional, forest area being the reference state variable. From a forestry perspective, the analysis appears to be reductive, as forests are described by many other state variables than area, including forest growing stock, composition in tree species, or stand structure. Whether the drivers of forest transition (population dynamics, economic modes of production and consciousness, as classified by Mather) also impact these other forest state variables in a general way thus comes forth as a logical issue.From a deductive analysis of forest transition drivers, and from forest trends brought to light in Europe, France, and at other places in the world, we here argue that the forest transition concept can be extended to a multi-dimensional space of forest attributes, characterized by typical ideal dynamics. Cumulative impacts onto forests and irreversible losses in forest biodiversity over a forest transition are hence highlighted. Global change, as a parallel consequence of countries’ developing process, further appears as one additional albeit less coupled dimension of forest transition, as it modifies forest productivity and vitality over time. Since forest ecosystem services and forest profitability primarily depend on such attributes, we argue that the extension of the forest transition concept has significance for land-use change and forest protection issues. A prospect on future changes in the forests of developed countries with the European space as a benchmark is finally proposed that leads to extend the temporal significance of forest transition. Though poorly described, returning forests on abandoned agricultural lands are significant, and deserve greater attention.


Introduction
The forest (-area)

transition framework, and the omission of trees
The concept of forest transition (Mather, 1992) refers to the observed inversion in the longterm -often pluri-secular -trend of the forest area of a country that shifts from negative to positive (a typical V-shaped curve) over time, as an integrative consequence of the many drivers that come along with country's development (see also Grainger et al., 1995). In spite of the difficulty to dispose of relevant and compatible forest areal data over the longer term (e.g. Audinot et al., 2020), this phenomenon and its variations have been initially exemplified from the forest histories of several developed European countries, including Denmark, France, Scotland, Hungary , Mather, 1992, 2004) and other places in Europe (Mather, 2001). For most countries, the forest (-area) minimum is attested to have been reached in the 19th and 20th centuries (Meyfroidt and Lambin, 2011), a period time where national forest inventory programs were inexistent or in emergence (first inventories in Nordic European countries; Tomppo et al., 2010) and associated forest maps or area statistics hardly available. Taking countries as initial reference support units for forest transition analysis has been justified by the major historical role of national policies and markets in land-use change dynamic, despite some drivers may be common to a continent (Mather, 2001  Socio-economic drivers of the forest transition have been identified that provide a generic and mechanistic understanding of the phenomenon Mather, 2001).
These have been classified into three categories, each typical of distinct phases of the forest transition (Mather et al., 1999;their Figure 2), including population, mode of production and consciousness. These drivers demonstrate tremendous variations across countries, amplified by the irreversibility of the energetic substitution from wood to fossil fuels, the growing urbanization of populations (Rudel et al., 2005) or economic globalization (Lambin et al., 2001) and deforestation exportation (Meyfroidt and Lambin, 2009). Yet, both already draw attention onto the dual need for land area and wood resources. Population refers to the demographic dynamics of a country and the associated impacts onto forests induced by the needs for both agricultural surfaces, and for energy or construction wood at its early stages.
The negative relationship between rural population and forest cover changes has hence been evidenced on a global scale (Mather and Needle, 2000). Mode of production refers to the economic organization of agricultural and industrial production, and the general reduction of pressures exerted i) on forest area following technological development in agriculture (agricultural, or 'green revolution'; van Zanden, 1991) and facilitation of regional economic exchanges, and ii) on forest resources, owing to energetic transitions (Gales et al., 2007).
Consciousness refers to either strategic perceptions of wood resources, leading to forest protection and productive policies enforced by law, and to perception of the protective role of forests on threatened lands (against erosion or landslide; Mather, 2001), or more recently their support as recreational spaces to urban populations (Konijnendijk, 2003). Obviously, forests deliver wood, and not only space resources (Kauppi et al., 2006). While the forest transition concept has arisen from a geographic perspective on forests and has as such placed focus on the primary dynamics of forest area and its interplay with land-use change, this initial view has therefore remained reductive.
Forestry sciences moreover envision forests through many other attributes than forest area (McElhinny et al., 2005), most likely to be modified by the drivers of a forest transition, including volume of the growing stock, tree species composition and richness (Grainger, 2005), or forest structure. Last, managing forests in a sustainable way also increasingly calls for embracing its provision for species habitats, and interplay with associated wildlife and plant communities (Bremer and Farley, 2010). Adopting a forestry perspective on the forest transition concept hence comes forth as more inclusive approach of forest changes, and leads to ask whether and which forest attributes may undergo generic and concomitant changes in addition to forest area, and what their implications for forest and environmental policies may be.

Synthesis objectives and outline
The aims of this synthesis contribution were therefore threefold. First, we review the evidences for extending the forest transition concept to an enlarged set of forest attributes, including forest growing stock, diversity in forest trees and accompanying species, and forest structure, as new dimensions of the forest transition. The analysis was based on forest inventory statistics and historical perspectives in Europe, in France (that served as an initial support for Mather's theory), and in different regions in the world, and was used to infer generic forest transition trajectories. Anthropogenic environmental changes and their impact onto forest vitality are also discussed. Second, we show that this multidimensional perspective on the forest transition not only has a cognitive interest, but also provides a renewed framework in analyses of forest value and land-use change. Third, we attempt to infer and discuss possible future generic post-transitional forest changes with the European continent as a benchmark, by highlighting current and future trends in forest dynamic and management, in conservation ecology policies, and in land-use change. Section 1. We show that forests primarily provide not only space, but also wood and timber resources that both vary in a qualitatively similar way over a forest transition. Section 2. The mechanisms of forest exploitation or restoration are shown to modify tree species composition and forest structure, through preferential species exploitation or selection for afforestation, and implementation of efficient silvicultural approaches. Some of these processes have irreversible impact on species diversity. Section 3. We emphasize that global change, like forest transition, is an outstanding consequence of countries' economic development, and also has impact on forests, as it has modified several abiotic resources of forest growth. Section 4.
Implications of this multi-dimensional perspective on forest transition, with cumulative and irreversible forest alterations, are discussed as regards the valuation of forest lands in land-use change perspective. Section 5. We draw a prospect on future possible changes in the forests of developed countries, based on the current situation of forests in Europe, and we suggest that 'forest transition' may gain achieved temporal significance for attributes other than area over the longer term.  FAO, 1948), places a primary emphasis on both forest area and volume of the growing stock (either total volume, or average volume per unit of area for which total forest area matters). These variables have definitions that may vary across countries and/or that have been harmonized internationally. Hence, the international definition of a forest is inclusive and assumes a minimum area of 0.5 ha, with a minimum width of 20 m, a minimum ground vegetation cover of 10%, and a vegetation potential to reach a 5 m height (FAO, 2000). The definitions for growing stock are more variable, but always refer to part of aerial volume of the living trees (timber wood in the tree stem up to a given diameter limit, e. g. of 7 cm), and beyond a minimum individual diameter threshold at conventional "breast height", e. g. 10 cm at 1.30 m (Gschwantner et al., 2019), that makes these definitions much more restrictive.

Shifting to a 2-dimensional representation of forest transitions
From a forestry perspective, the acknowledgement that forests actually represent both space and wood resources suggests a two-dimensional description of forest transition (total forest area and total growing stock of a given territory). In a global analysis of the recent evolution of countries' forests, Kauppi et al. (2006) introduced the concept of 'forest identity' that separates four nested variables describing forests, including forest area (ha), density of growing stock (m 3 /ha), biomass ratio (tons of biomass/m 3 ) and carbon concentration (tons/ton). Country's forests evolution was hence analysed in a two-dimensional plane defined by rates of change in forest area and density of growing stock, therefore highlighting the relevance of such analytic approach. Nevertheless, a forest transition analysis based on the average growing stock per hectare may be less intuitive than by considering total growing stock (Rautiainen et al., 2011). Among obvious process examples, deforestation for newer agricultural lands lets the density of growing stock unchanged, but not the total growing stock. Also, active afforestation generates early-stage and low-stocked forest stands, leading to a transient decline in the density of the average growing stock (example of Viet-Nam, Kauppi et al., 2006). Second, the forest transition concept focuses on temporal changes in the forests, and thus on trajectories of forest attributes over time, more easily captured in a state space than in a momentum space (the space being defined by the rates of change of the different forest variables under consideration; Kauppi et al., 2006). We thus adopt the former convention in the following.

Area and growing-stock trajectories along the forest transition
In the initial reduction phase of the forest transition, a double need for both newer agricultural areas and energy/construction wood has been stressed. The need for agricultural surfaces implies clear-cutting forests, and thus a sudden change in the forested area. Deforestation of these new surfaces certainly comes along with a joint satisfaction of wood demand, so that the area and growing stock dynamics may be qualitatively similar. However, the need for energy/construction wood for domestic or industrial purposes may rather imply a progressive depletion in the forest growing stock of the concerned forest territories, resulting -or not -in deforestation. Severe depletions of the growing stock are hence attested in forests that have persisted, e.g. in France, including the Tronçais forest to provision forges for energy wood in the 18th century (Roy, 1969), in the Chaux forest to fuel neighbour forges, saltworks of Salins and Arc-et-Senans, and glass factories from the 16 th to the 18 th century (Plaisance, 1966). A scarcity in fuelwood was also reported in forests of Lorraine region that provisioned tin and glass factories and saltworks, some of which even ceasing their activities by the end of the 18th century (Badré, 1992). Growing stock depletion in areas still afforested remains therefore ignored in the forest transition analysis, despite being strongly associated to the developing processes of a country. The pattern has been general in Europe at the preindustrial era (Mather et al., 2001).
The same distinction remains operative in the forest expansion phase, where forests can expand: i) either naturally on abandoned agricultural or grazing lands following rural depopulation (Rudel, 1998;, or ii) artificially when resulting from active plantation programs aimed to restore and secure wood resources (Rudel, 1998), with an attested global success along the 20 th century (Sedjo, 1999). In the first case, a returning forest state will be reached following the progressive development of transient vegetation forms (lands with shrub-type vegetation) that will meet the forest definition far before they represent a non-zero growing stock. In this case, the growing stock expansion will be strongly delayed, currently over several decades. In the second case, planting for developing new forest resources will quite simultaneously add growing stock to these new forest areas, as technical options will be implemented to favour rapid forest development (soil preparation, use of nursery trees, and of fast-growing often coniferous tree species; Savill et al., 1997), despite these options may depend on socio-economic factors at a country level. In addition to these afforestation processes, non-deforested areas that were depleted or managed at a low density of growing stock (typical coppice forests to produce fuelwood) will also contribute to increase the total growing stock (e.g. Bontemps et al., 2020, in France), either through active conversion policies, or by natural development (section 2.2).
Theoretical trajectories featuring these different dynamics are summarized in Figure 1, both over time and in the two-dimensional forest state-space defined by area and growing stock.
These illustrate the relevance of this two-dimensional perspective on forest transition. These dynamics are also illustrated in Figure 2. aa. Representation in the state space defined by A and G:  forest depletion caused by a deforestation process leads to a decrease in both A and G. This decrease is linear as soon as the forest is assumed to be homogeneous in the density of growing stock (growing stock per unit area, or D),  forest depletion caused by overexploitation for timber of fuelwood exhibits a different footprint in this state space, where only G decreases. In situations of exacerbated depletion tending to zero G, gaps in the forest area may appear and cause a further decrease in A,  forest expansion caused by natural afforestation on abandoned agricultural lands. G will progress slowly as ground vegetation will precede a forest state, while definitions for G most often require a minimum countable threshold on tree diameter at 1.30 m height,  forest expansion due to active plantation programs aimed at securing a future timber and/or fuelwood energy resource. In this case, G will increase sooner than for natural afforestation as plantation engineering facilitates the quick establishment of forest stands, and it will also increase faster as the productivity of introduced species will a higher density of growing stock,  forest replenishment in G in areas where the growing stock was initially depleted, but that kept their 'forest' state (A does not change). Replenishment can occur naturally when forests turn under-exploited, or can result from active management, e.g. conversion of coppice forests into high forests. Under the hypothesis of homogenous forests, D increases proportionally. Progression in G under this process should be more regular than in forest expansion situations, because of a pre-existing forest state. Representation of the associated temporal dynamics in A and G. The same five situations are represented, and the density of growing stock (D, Kauppi et al., 2006) is added for the sake of comparison to the dynamics of total growing stock G. Vertical positions are arbitrary.  Under deforestation process, A and G decrease proportionally when the forest is assumed homogeneous in D. D however remains constant, which forms the footprint of pure deforestation.  Under an over-exploitation process, A remains constant and the decrease in D causes the decrease in G. Over-exploitation ultimately causes a decrease in A when forest gaps appear, which may slower the rate of decrease in D.  Under a natural afforestation process, the increase in A will be delayed as the definition of a forest requires a minimum size, ground cover, and vegetation able to reach a height of 5 m. The increase in G will take a longer time as the definition for growing stock requires a minimum tree diameter. Accordingly, D will initially remain stable, and will then decrease as a consequence of increasing A with a zero G. Only after G is expanding, D will meet a minimum and will expand again with forest reconstitution.  Under an active plantation policy, A will increase immediately, causing an immediate decline in D. G will increase sooner than under natural afforestation and will maintain a higher rate of progression. Note that under both processes of forest expansion, D will decrease. This early variation may seem counter-intuitive at first glance and it is further qualitatively similar to that of an over-exploitation, leading to discard D to evaluate growing stock.  the replenishment of forests with a continuous forest state leads to a fast increase in G, as the initial growing stock has a higher level than in previous situations to satisfy a forest state. In these forests, A is unchanged. Consequently, the dynamics of D and G are parallel.

Changes in forest composition
Forest composition in tree species present in the forests is prone to change during both the reduction and expansion phases of a forest transition. While the selective exploitation of forests before transition certainly has such impact, it remains far less documented than that of afforestation programs or natural regrowth (Cramer et  The most obvious driver of species compositional changes however prevails in the forest expansion phase, when the emergence of a forest productivism (Mather et al., 1999) has resulted in active forest plantations (Sedjo, 1999) on either formerly non-forested areas, or to convert ancient forests, over typically several decades. Over the past century, economic and industry-driven considerations have most often led to favour monocultures of exotic or native fast-growing coniferous tree species including pine and spruce genus, or Douglas fir (Pseudotsuga menziezii) species (Kuusela, 1994 in Europe), causing a shift in species composition in areas where broadleaved species prevailed (see illustration in Figure

Changes in forest structure
Forest structure refers to "the physical and temporal distribution of trees in a forest stand" (Oliver and Larsson, 1990), and it includes the horizontal and vertical distribution of trees, their size and age, and species, implying a geometrical description of forest stands (Stone and Porter, 1998). Heterogeneity in forest structure is generally greater in natural or semi-natural forests than in managed ones, as natural disturbance regimes favour spatial heterogeneity in tree size, stand density or species composition (Huston, 1994;Franklin et al., 2002;Huang et al., 2003). Conversely, forest structure tends to be homogenized for the practical purpose of forest management. Therefore, changes in forest structure are a key aspect of secondary forest definition (Chokkalingam and De Jong, 2001).
In managed forests, forest structure results from management systems that can be discretized in ideal types. Coppices correspond to trees growing on stumps (of tree species able to regenerate by sprouting) and coppiced over short rotations (one to two decades) in order to produce fuelwood with rapid regeneration. High forests are formed of adult trees developing over decades in order to produce timber wood. Such forests can be regular or irregular depending on the homogeneity of tree size/age, and therefore on the mode of regeneration of stands (clearcut with spatio-temporal rotation against selective cutting and gap regeneration).
They can also be homogeneous or mixed in tree species. The practice of setting aside high forests in reserve in feudal or royal forests ("defensa"), to secure timber wood production, originates in the early middle age in Europe (Hüffel, 1926). In France, it was generalized with the "Quart en reserve" (one fourth of the forest area in high forest) in the Water and Forest law of 1669 inspired by Minister of Finance Colbert. Coppice-with-standards correspond to two-horizontal layered forest stands where adult trees are maintained over several coppice rotations to further secure provision of timber wood, and hence form an alternative management option to that of separating coppice and high forest systems over space.
All over non-deforested plains of Europe (see regional references in Baeten et al., 2009), coppices of the main oak species have formed a prevailing forest system since the Roman and medieval periods (Haneca et al., 2005) for a fast and regular renewal of fuelwood resource (Peterken, 1993, chapter 3). In European mountain ranges including the Pyrénées, the Alps, the Appennine ranges (Ciancio et al., 2006;Coppini et al., 2007), common beech coppices were also widespread (common beech is able to grow from stump beyond a minimum altitudinal level; Boppe, 1886) that currently served to provide renewable fuelwood to rural populations living from cattle breeding.
The energetic transition (Ben Gales et al., 2007) from fuelwood to fossil fuels in the 18th and 19th centuries in Europe (mode of production of Mather, 1992) and the progressive abandonment of agricultural land in mountainous contexts has contributed to lower the pressure on the forest growing stock. Obviously, it has also impacted forest structure over large areas, with coppices either freely developing as, or being actively converted to, high forests, a process known as 'coppice conversion' (Peterken, 1993  This change may be one major driver of growing stock accumulation in Europe, as evidenced by Bontemps et al. (2020). In France, a large conversion momentum of oak coppices was initiated by Lorentz in the mid 19th century (Lorentz, 1837, cited by Hüffel, 1926), following the enactment of the forest code in 1827 and the conviction that "the role of State forests was to product timber of large dimension, not fuelwood nor money income". Conversion started in North-eastern France and then propagated to the whole country. It formed a major challenge, as an estimated 64% of the French forest area was managed as coppice or coppice-withstandard forests (Degron, 1998). Conversion also met some resistance as the predominating and so-called 'immediate conversion' method required the evolution from coppices to high forest on stumps, able to produce a natural regeneration, and implied transient profitability losses. The major conversion effort was paid during the period 1860-1888 after introduction of fossil fuels, with a rate of ongoing conversion of 74% by 1876 (Degron, 1998).
Nevertheless, achieving conversion has taken long and sometimes failed (Lafouge, 1964;Vannières, 1983). Ancient forest statistics as compared with modern forest inventory statistics over one century have shown this shift from coppice to high forests being lasting up to recently (Audinot et al., 2020). Accordingly, current estimates (Dubourdieu, 1991) still indicate a current forest area of 1.3 Mha of coppice stands, i.e. 8% of the total forest cover.

Forest transition trajectories of these state variables
Forest transition thus comes along with important changes in forest composition and structure. Forest composition in tree species strongly changes through favouring specific silvicultural systems, either in the depletion (e.g. coppices of sessile oak) or the expansion (e.g. plantations of coniferous species) phases. Tree species diversity is also likely to decrease, as a general pattern in transition from primary to secondary forests, despite the issue is of much more concern in the tropics (Chokkalingam and De Jong, 2001). An essay at idealizing theoretical trajectories for state variables describing forest composition (proportion of coniferous forests and tree species diversity) and structure (high forests, coppice forests, and even-aged forests) is provided in Figure 4.  represented as a temporal reference. Black lines correspond to forest areas where both deforestation/exploitation of forests and natural/plantation afforestation may play a role in the dynamics of forest area (see Figure 1). Grey lines correspond to a theoretical forest transition where only deforestation (need for area) plays a role in the dynamics of forest area, the maintained forest area remaining untouched. Vertical positions remain arbitrary.  Forest composition described by two state variables including the proportion of coniferous species occupying the forest area (C) and the diversity in tree species (Dv). In the temperate and tropical forest zones, the need to secure timber resources lead to implement afforestation programs and develop intensive forestry that will both contribute to increase A and C (see section 2.1). This evolution does not apply in the boreal zone where coniferous species prevail. Following human occupation and exploitation of forests, Dv should decrease and stabilize early in the forest transition process. The introduction of non-native tree species for afforestation, as well as natural afforestation following agricultural land abandonment should contribute to increase in Dv over the area, though at a much lower level than in the primary forests. In the deforestation-driven forest transition (light grey), Dv actually decreases, though at a much slower rate, due to variations in the distribution of species and habitats (Rozenzweig, 1995).

Returning forests on abandoned agricultural lands: significant, but unknown
 Forest structure described by three state variables including the proportion of high forests (HF), coppice forests (CF), and even-aged forests (EAF) in the forest area. (a) Following early human occupation and exploitation of forests, high forests should be altered by initial removal of timber wood and evolve toward fast-regenerating coppice forest systems, leading to an early increase in CF and equivalent decrease in HF in the forest area, that should then stabilize before the end of the forest area reduction phase. Along with coppice extension and exploitation of forests, initially low EAF, not common in primary forests, should also increase and stabilize. In the forest expansion phase, active afforestation and conversion of coppice forests into high forests should reverse the trends in HF and CF, as well as contributing to a newer progress of EAF over the forest area. (b) Theoretical forest transition only driven by deforestation. Non deforested areas remained unmodified, with high forest as an exclusive and permanent forest structure. Initially low EAF will remain unchanged, unless active afforestation contributes to forest expansion and favours even-aged plantations. Especially since they extend on physically abandoned lands, these naturally returning forests remain poorly described. They further hardly fit inventory classification categories of national forest inventory (NFI) programs or inventory thresholds (census diameter of a few centimetres, or measurable canopy cover rate), and have drawn restricted attention from the forest and ecology research communities to date. For instance, 1.2 million hectares of forest out of 17 million hectares remain of undocumented composition in France. Their exact contribution to shifts in forest composition or structure therefore remains largely unknown, despite it may lead to substantial inflection in the afore derived forest trajectories. In absence of historical analogues, the future dynamics of these forests, developing on pasture or agricultural lands, is highly uncertain (Schnitzler, 2014). economic value also depends on forest ecosystem processes that remains neglected to date, as not all functions ensured by forests are marketed (Pearce, 2001). In addition to timber and non-timber forest products (NTFPs), the socio-economic services provided by forests (tourism, recreation), the indirect values generated by watershed protection or carbon storage (MacDonald and McKenney, 2020), and option or existence values (i.e. acknowledgement of possible future uses and benefits, or will to conserve) will contribute to forest value (Pearce, 2001), and will again primarily depend on forest area, growing stock, structure, composition and species diversity. Shifting to forest ecosystem services as an extended framework for forest economics is therewith turning a matter of increasing attention (Kant, 2003). This inclusive apparaisal has also been recently translated into forest transition theory, with forest ecosystem-service transition curves put forward as a means to assess effects of forest transition processes on ecosystem services and their trade-offs (Wilson et al., 2017). Last, and beyond a service-driven analysis of forests, acknowledgement of the existence value of forests has arisen for more than a century, with development of ideas in environmental ethics providing substantial support to the intrinsic value of forests, as whole living systems worth of respect, and as sources of aesthetic and metaphysical experience for the human mind (HD Thoreau, RW Emerson, and A Leopold as early influential thinkers; Callicott, 2000).
Aesthetic value has also been shown to depend on structural forest attributes (e.g. Gobster, 1999). These aspects suggest that the present multi-dimensional perspective on forest transition does not only have an interest in widening the concept per se, but also deserves greater recognition in a land-use change perspective on forests.

Cumulative and irreversible impacts of a forest transition on forests: the need for integrative approaches
First and obviously, there are couplings in the temporal dynamics of these different attributes over a forest transition, and forest areas or resources generally undergoing depletion will come along with reductions in forest area, growing stock and tree stature, and diversity in tree species and habitats, and also subsequent services and values attached to these attributes (4.1).
Conversely, the expansion of forest resources following the forest transition will be accompanied with restoration of growing stock, forest structure and productivity, and a more forests (present at the forest minimum) and recent forests following forest expansion have however been found. They suggest that species dispersion processes may strongly limit the possibility of plant composition recovery (Hermy and Verheyen, 2007), and highlight the role of the landscape matrix in these dynamics (Chazdon, 2003). In addition, changes in soil properties following exploitation (Chazdon, 2003)  The speculative course of late forest transition trajectories as regards forest area, growing stock, and forest composition and structure are represented in Figure 6 and are referred to in the following.

How will the V-shaped area curve extend?
Forest transition has been experienced by many countries across the globe, and have been occurring from around two centuries ago up to recently (Meyfroidt and Lambin, 2011 and associated Fig 1). In Europe, forest area is logically expanding faster in late-transitioning countries than in the others, but is nevertheless increasing in all European countries, with a As aforementioned, current forest reporting does not distinguish between afforestation and natural forest expansion as factors of forest development (MCPFE, 2011, Fig. 42). Future trends regarding the contribution of afforestation programs may also be hard to formulate, as they remain largely subjected to national forest strategies and the associated support funds allocated to these. Agricultural land abandonment, however, is a clear pan-European trend, and hence forms a much more certain contribution to future forest expansion. Also, the new European forest strategy intends to foster forest cover in view of promoting ecosystem service delivery (Köstinger, 2015) Massive abandonment of marginal agricultural lands in Europe has so far been caused by   Forest area (A) and total growing stock (G). A has still experienced recent increase, and it may keep on expand during several decades due to land abandonment in marginal agricultural areas. G has experienced the second strongest relative increase in the world, and benefits, though with a delay, from the expansion in A. It may further profit from a level of felling lower than that of natural increment. However, severe disturbances may slower this progression over the longer-term (not represented). Figure 6 (caption continued)  Forest composition described by two state variables including the proportion of coniferous species occupying the forest area (C) and the diversity in tree species (Dv). C, in areas where it has some relevance, may decrease following a will to return to more natural, disturbance-resistant, and biodiversity-oriented secondary forests, known in Europe as 'forest conversion'. Forest conversion (*) may correspond to a late transitional event in the whole forest transition course. Dv should keep on increase, though at a slow rate, as some losses may be irreversible.
 Forest structure described by three state variables including the proportion of high forests (HF), coppice forests (CF), and even-aged forests (EAF) in the forest area. HF/CF are assumed to keep on increasing/decreasing, respectively, following late CF conversion effort and the ongoing trend in land abandonment. The development of short-rotation coppice forestry to produce fuel wood may however set coppice forests to a minimum proportion. Following two distinct phases of increase (early exploitation of forests with associated harvesting methods, and late development of even-aged forestry following the forest transition), EAF may start decrease following return to more natural forests, defining a second aspect of forest conversion (*).

Growing stock: current dynamic, environmental hazards, and carbon issues
Growing stock accumulation currently shows no sign of saturation With an average 156 m 3 /ha, Europe has the third largest regional growing stock in the world after South America, and Western/central Africa. Important increases in the growing stock of with increasing trust in this dependence (Cornwall, 2016). In an attempt to disentangle the respective contribution of climatic changes and forest changes on disturbance-related forest losses in Europe, Seidl et al. (2011) reported that these drivers were comparable in magnitude.
Drought and heat wave risks are also acknowledged but will rather affect forest productivity and mortality in a less sudden way (Allen et al., 2010 these strategies provides no easy assumption as regards evolutions of the growing stock. However, it may be noted that the future of Kyoto protocol, and the place of carbon sequestration in forests, faces substantial uncertainty and debate following closure of the first commitment period in 2012 (Plantinga and Richards, 2008 for the forest sector). By contrast, the valuation of energy wood as part of the future 'energy mix' may turn a more robust policy baseline in the EU, in a context where it ranks first as an energy importer and energy prices increase (EU, 2012). In the latter case, part of the strategy may also rely on highly-productive short-rotation coppices (e.g. of Populus species), possibly set on former agricultural lands (Kauter et al., 2003). Also, the more recent European strategy for bioeconomy (EU, 2015) intends to foster the use of 'green resources', including industrial and timber wood use for  (Zerbe, 2002). Forest structure, in turn, is prone to shift from even-aged to irregular in age and size, with a view to approaching a more natural forest structure, to reducing habitat fragmentation and soil erosion ('continuous cover forestry'), and susceptibility to windthrow (Buongiorno, 2001;Mason, 2002). Continuous-cover forestry also favours carbon sequestration in forests (Seidl et al., 2008). Mixed conversion strategies including structural and compositional aspects are also considered (Hanewinkel, 2001).
Toward initiating a full forest transition in structural and compositional forest attributes?
After decades of promoting conifer plantations as a privileged forest production system, it may thus happen in the future that the proportion of such species be reduced, in forest areas where broadleaved species show ability to grow. While high forest structure should persist, it may also shift from even-aged to more irregular structures. When shifts in both attributes are crossed, the occurrence of uneven-aged mixed forests may logically increase in the forest resource, hereby moving closer to the structure of primary forests, yet of lower growing stock through active forest management. Hence, the term 'forest transition' would start to more largely apply to forest compositional and structural attributes, though strongly delayed in time with respect to forest area and other attributes ( Figure 6).

Returning forests and European 'wilderness': the ultimate step in forest transition?
Returning forests on physically abandoned farmlands (sections 2.4 and 5.1) are a significant contribution to expansion of natural forests in Europe. Their free development may be comforted by resurging concern and strive for a European 'wilderness' (defined by the IUCN as 'a large area of unmodified or slightly modified land […] retaining its natural character and influence […] which is protected and managed so as to preserve its natural condition').
This request has been materialised in a recent resolution of the European Parliament lands may define a more radical pathway to naturalness than the 'conversion' of current plantation forests towards more natural forest ecosystems. In this sense, they would contribute to the full completion of a 'forest transition' in some areas.

Conclusions
 The forest(-area) transition theory (Mather, 1992) has arisen from a geographer's perspective onto forests, and has generated a profuse literature over the past twenty years, in the context of tropical deforestation. By contrast, restricted attention has been paid to forest attributes other than forest area, with a discrete exception as regards forest growing stock.
 From a forestry perspective, several forest attributes are shown to undergo generic transient changes, concomitant to the forest transition, which include: 1) the growing stock that undergoes a parallel V-shaped trajectory, 2) the composition and diversity of forests with an increase in the proportion of conifers in temperate and tropical areas, and a loss in species diversity that may take long to recover, and 3) the structure of forest stands, with a further development of even-aged forestry, and progressive abandonment of coppice systems in favour of high forests. These aspects turn obvious as soon as forest cover is also seen as a manifestation of the tree populations that compose it.
 Noteworthy, returning forests on physically abandoned farmlands significantly contribute to forest expansion in developed countries. These forests remain largely ignored to date, and their influence to the dynamics of former attributes is unknown.
Research efforts are needed to describe naturally returning forests and their possible dynamics.
 A later -and so far speculative -trend may arise in the end of forest transition dynamics in developed countries. It consists in promoting more natural forests in terms of composition and structure, and assumes a more gradual management. Such changes, known in Europe as 'forest conversion', are viewed as a means to develop more disturbance-resistant, biodiversity-beneficial and environment-oriented forests, with increasing growing stock, climatic change and growingly urban populations as contextual background. They may form a sub-transition per se in the whole course of forest transition.
 Forests returning naturally on abandoned farmlands may provide ultimate significance to the forest transition concept in some areas. In Europe, they may benefit from growing concern for fostering 'wilderness', and associated policies in the future. Such changes are called to play an increasing role in shaping tomorrow's forests, because of rapid changes in ecological and societal conditions.
 Acknowledging the multidimensionality of the forest transitions actually has not only a cognitive interest. It also highlights the many values and services provided by forest ecosystems in their connexion to the biologic, abiotic and human environment, beyond wood resource aspects. As such, it needs more attention in the socio-economic analyses of forest-related land-use change.