Projections of wood resources have been needed for decades by both public authorities and public and private stakeholders. Projections are requested to formulate and assess forest policy and management strategies for the coming years to decades at regional, national, and international scales. The French NFI provides information for describing the forest resources and their spatio-temporal dynamics. These data contribute to initializing two large-scale dynamic models for forest resource projection that have been developed in France and implemented since the 1980s. The decision to use diameter-class or the age-class matrix models depends on the type of forest to be simulated. Both approaches provide robust projections from short to medium time scales and at spatial scales ranging from regional to national.

Recent projections carried out with these models have highlighted the potential for French forests to remain an important carbon sink until 2030, while fellings can be increased by two-thirds. This matrix-model approach is associated with interdisciplinary research programs that aim at developing environment-dependent representations of forest dynamic processes such as growth, density-dependent mortality and forest regeneration, and also coupling of forest dynamics with forest sector models. Model validation is also essential, highlighting the major importance of current continuous forest inventory, and making forest inventory data comparable over time following changes in mensuration protocols.

1 Introduction

French forests are characterised by a high diversity in various aspects that must be considered when developing models for projecting wood resources and potential wood supply . We first provide this necessary background information and then present objectives, data and projection methods, and perspectives for further model development.

1.1 Key Figures on Forests and Forestry in France

Forests cover 16.8 million ha of the French metropolitan territory (30.8% of the area) and 8.2 million ha in oversea territories, mainly in French Guiana. Temperate and Mediterranean forests of the metropolitan territory thus represent two-thirds of the total national forested area. Intensive and systematic forest monitoring via the National Forest Inventory (NFI ) has been carried out over the metropolitan territory since the late 1950s, enabling continuous analysis of forest features and wood resources . NFI -like monitoring is not conducted in the French tropical forests.

French metropolitan forests have developed in a variety of climates. Oceanic and sub-oceanic climates in western and north-central France favour the development of sessile (Quercus petraea) and pedunculate (Quercus robur) oaks. Maritime pine (Pinus pinaster) grows especially in the oceanic climate of the south-west. Semi-continental climate in the north-eastern and mountainous areas (Alps, Jura, Pyrenees, Massif Central and Vosges) are suitable for European beech (Fagus sylvatica), Norway spruce (Picea abies), silver fir (Abies alba), Scots pine (Pinus sylvestris) and European larch (Larix decidua). In the southern Mediterranean range, evergreen oak (Quercus ilex), Aleppo (Pinus halepensis) and Corsican (Pinus nigra ssp laricio var Corsicana) pines grow. As a consequence, French forests encompass all but the “boreal forests” type (13 types) according to the European Forest Types classification (EFT, Barbati et al. 2014), and rank first in Europe for the number of forest types covered. In addition, 18 tree species are found predominant over >200,000 ha (IFN 2010). Many other species are found in various locations, such as pubescent oak (Quercus pubescens), chestnut (Castanea sativa), hornbeam (Carpinus betulus), Douglas fir (Pseudotsuga menziesii), ashes (Fraxinus spp.), aspen (Populus spp.), etc. Mixed species forests hence form one half of the national resource, both in terms of forest area and growing stock (Morneau et al. 2008). Stand structure also shows great variations, including pure even-aged high forests of broadleaved and coniferous species mainly in the northern plains and in the Aquitaine region, uneven-aged mixed species forests in the mountainous domain, and coppices and coppices with standards in the plains.

The heterogeneity in species and structure of French forests originates from ecological factors. Coniferous species are mainly located in the mountainous domain, broadleaved species are found in the plains, and Mediterranean species are located in southern France. It also results from historical socio-economic factors. In the plains, forests are dominated by broadleaved species, with sessile and pedunculate oak species ranking first for the growing stock (566 million m3 representing 23% of the total national growing stock , IFN 2010). These species have been favoured since medieval and even Roman times, because of their ability to re-sprout from stumps after cutting. In the eighteenth and nineteenth century oak coppices with standards stands were managed to produce roundwood and fuelwood for domestic and industrial purposes. Later, following the introduction of coal and the need to produce timber, many of these forest types were either naturally or actively progressively converted into high forests, with high forests on stumps as a first structural step (IGN 2013).

Conifer plantations were promoted in the nineteenth century to drain the wetlands of Gascogne and Sologne and to protect fragile soils for purposes such as preventing erosion in the Alpine range or fixing dunes on the coast range. The National Forest Fund (Fonds Forestier National: FFN) policy framework was launched in 1947. It contributed to the development of a new conifer forest resource over marginal agricultural lands, and was instrumental in securing the wood supply for the country’s industries (Gadant 1987). The FFN is estimated to have contributed 75% of the increase in coniferous forest plantation area, i.e. 1.5 million ha of new forests with Norway spruce (Picea abies), Douglas fir (Pseudotsuga menziesii), pines (Pinus sylvestris, Pinus nigra) and European larch (Larix decidua) as the main afforestation species (Pardé 1966; Dodane 2012).

The area of forests available for wood supply is 15.7 million ha in 2013, of which 11.8 million ha (74.8%) are private forests, 1.4 million ha (9%) are state-owned forests and 2.5 million ha (16.2%) are other public forests including forests owned by territorial communities and public authorities (IGN 2014a). Over the past decades, the decline of agricultural activity in low productive areas has led to further natural afforestation by pioneer species which may, however, neither correspond to demand on the wood market nor be situated in easily harvestable locations (IGN 2012). As a result of both afforestation and natural colonization, the French Forest area has more than doubled since 1788 (Brénac 1984). This reflects the typical pattern of ‘forest transition’ (Mather 1992) that has been experienced by most developed European countries following the industrial revolution and urbanization processes.

Concern for nature and biodiversity conservation has produced legal constraints (e.g. Natura 2000 network) that partially exclude some French forest areas from traditional forest management and harvest schemes. The impact on wood supply can be locally significant. In the Vosges Mountains and on the Lorraine plateau, 4% of the total forest area is not available for wood supply , and 8% of the forests available for wood supply are subject to management for species conservation (Colin and Lambert 2012; Thivolle-Cazat et al. 2012). Physical constraints also significantly hamper forest management in France, and cause a decrease in the profitability of harvest operations. One-third of the forest area available for wood supply is estimated to be very difficult to access, and 1 million ha of the forested area has a forwarding distance greater than 1 km, mainly in south-eastern France (IFN 2010).

Total annual fellings amount to 64 million m3 of total aboveground volume over bark (IFN 2011c). Industrial roundwood removals (sawlogs, pulpwood, etc.) represent about 36 million m3 over bark per year (Agreste 2011) and fuelwood about 20 million m3 over bark per year (FAO 2011), while the remaining part corresponds to harvest residues . Total removals have increased slowly over the last 20 years and result from both an increase of removals from coniferous stands afforested between 1950 and 1980 (FFN) and a regular decrease in hardwood removals . Forest chip removals for energy have increased rapidly in recent years; the removed volumes have doubled every year since 2006, and reached 1.8 million m3 in 2013 (round wood equivalent).

For several decades, French forests have experienced an increasing gap between total drain and increment. Hence, growing stock , assessed as bole volume up to a top-end diameter of 7 cm, has increased by about 25 million m3 per year over a 26-year period. This increase represents an additional 650 million m3 (IFN 2011b) for a total of 2.52 billion m3. There is no sign of a future shift in this structural trend. Recent and severe damages caused by storms Lothar and Martin in 1999 (140 million m3 of damage , Doll 2000 and IFN 2003) and Klaus in 2009 (43 million m3 of damage , IFN 2009) represent anecdotal drains compared to the trend in growing stock on a national scale. The current expansion of the French forest area will further contribute to an increase in the growing stock in future decades. Thus, the French forest is far from a stationary state of development.

1.2 Motives and Objectives for Projection

The needs for projections on forest resources arise from two categories of stakeholders:

  • Public authorities, at both national and regional levels, require projections to formulate and assess public policies, for strategic guidance and planning of the industrial roundwood and fuelwood sectors, and to report the contribution of forests to carbon sequestration as required by international policy processes on Green House Gases (GHG ) emissions reduction.

  • Public or private stakeholders of the forest-wood sector, including the French Forest Service (Office National des Forêts, ONF) that manages the public forests, saw and veneer-mills, pulp and paper, panel and fuelwood companies, and professional organisations (e.g. associations of wood producers and processers, inter-branch organisations) require estimates for the wood availability and quality of a given species in a given area to formulate or support their economic strategies.

Inventory, assessment, analysis and projection of the forest resource, including wood volume, biomass or carbon, form the missions of the French NFI which is carried out by the French National Institute for Geographic and Forest Information (Institut national de l’information géographique et forestière: IGN). The French NFI consists of a repeated and spatially-systematic inventory of metropolitan forests with the objective of providing statistics at the regional and national levels (Vidal et al. 2005). It is thus well-suited for contributing to decision making and policy formulation at these scales. By contrast, it is not suitable for supporting forest management at a local scale.

Illustrations of recent wood resource projection studies based on NFI data follow:

  • 2009, national level: potential wood availability for industrial and energy uses (study funded by the national renewable energy agency and the ministry of agriculture, Colin et al. 2009; Ginisty et al. 2011);

  • 2010, intra-regional level: impact of the 2009 storm on the wood resource of the Landes de Gascogne maritime pine forest (study funded by the ministry of agriculture, Colin et al. 2010).

  • 2012, regional level: potential wood availability in Lorraine and Aquitaine until 2025 (studies co-funded by regional and infra-regional public bodies, as well as professional organizations for pulp, paper and fuelwood, Colin and Lambert 2012; Thivolle-Cazat et al. 2012; Colin et al. 2012).

  • 2014, national level: GHG emissions of the French forest until 2030 in the context of the development of renewable energies (study funded by the ministry of environment, Colin 2014).

2 Data Sources

2.1 NFI Data Used for Wood Resource Projections

Each year, the French NFI inventories the entire metropolitan forested area. The NFI is carried out in public and private forests, regardless of whether they are available for wood supply . Since 2004, the NFI design has featured a systematic sampling grid with a 10% systematic sample of the grid measured each year. Each year sampling is conducted in two phases. In the first phase, approximately 80,000 photo plots are interpreted to assess land cover and land use. In the second phase, approximately 7000 temporary ground plots are established at a subsample of first phase photo plot locations that have forest land use. All living trees on the sample plots are cored at 1.30 m height to measure radial increment over the past 5 years. Trees with diameters less than 7.5 cm are not measured. NFI statistics are calculated using a post-stratification process using the French forest map (IFN 2008) and aerial photo-interpretation of land cover/land use for the 80,000 photo plots for a given year. Typically five annual NFI samples are combined to calculate statistics for the forest resource.

NFI data used for projections include:

  • Contextual data: a biophysical classification of the forest area into “sylvo-eco-regions” (IFN 2011a), administrative regions, forest land ownership , availability of forests for wood supply , topographic position, elevation, physical accessibility , soil characteristics, etc.;

  • Plot level data: species composition and forest structure, stand age , number of stems, top height and site index (Seynave et al. 2005), etc.;

  • Tree level data: tree species, circumference at 1.30 m height, height, wood quality classes, radial increment, bark thickness, growing stock , mortality , fellings , etc.

2.2 Additional Data for Spatial Contextualization of the Forest Resource

When projecting wood resources , the territory is usually divided into strata, often derived from GIS information. Both the wood resource and silvicultural scenarios are described at this level. In such studies, potential wood availability estimates and actual fellings often show a gap, due to repeatedly acknowledged physical, socio-economic, and environmental constraints (Puech 2009). Physical restrictions such as slope, presence of forwarding tracks and forwarding distance, prevail in mountain areas. Socio-economic restrictions result from the private ownership structure in France with a majority of small forest estates (Agreste 2013). Environmental restrictions arise from conservation and protection measures (soils, freshwater springs, etc.).

The initial state of the wood resource and projection scenarios can be refined using external GIS information to identify forest types with specific management features. For example, maps of large private forests with forest management plans allow identification of areas with more intensive management opportunities. Legal restrictions for environmental protection also have an impact on harvest costs or management regimes, and the areas where silviculture is legally not permitted (e.g. biological reserves with strong protective rules) are sometimes excluded from wood projections. Specific forest management practices are also implemented in areas with harvesting limitations for fauna/flora conservation. For example in the Vosges Mountains, silvicultural practices for the conservation of the Western Capercaillie (Tetrao urogallus major) habitat are encouraged (MEDDE 2012).

2.3 Newer NFI Data to Quantify Felling Regimes

Since 2010, the French NFI has provided direct measurements of fellings based on a partial re-inventory of temporary plots inventoried 5 years earlier. These estimates are much more accurate than previous estimates based on consumption surveys or stump inventories (IFN 2003). Productivity, mortality and felling statistics are now fully consistent with variations in the growing stock (IFN 2011c). These new data can contribute to the calibration of silvicultural and mortality scenarios used for projections, and especially the Business-As-Usual (BAU) scenario (Colin 2014). The silvicultural scenario derived from these measurements is relevant for BAU projections, because it takes into account implicitly all the current management and harvest opportunities and constraints (physical, economic, environmental and social).

These new data can also contribute to virtually doubling the number of NFI sample plots available for estimating the forest resource for a given year, because they can be used to update the wood resource of the forest plots inventoried 5 years ago. This practice leads to more precise NFI estimates, which finally allows more spatially accurate NFI results or more detailed results for a given spatial domain (Colin et al. 2012).

2.4 Data on Wood Assortments of Standing Trees and Removals

The French NFI identifies three classes of roundwood quality for each standing tree, according to potential uses such as veneer, sawn timber, pulpwood and fuelwood. In addition, market fluctuations and changes in transformation practices lead to implementation of a complementary expert-driven approach, based on analysis of data such as historical removals data.

Removal assortments are estimated from two surveys for which representativeness and statistical accuracy /bias are not assessed: (1) Since 1948, the national annual survey on commercialized wood removals (Agreste 2011) has provided harvested volumes over bark by species and assortment, at administrative-department level; and (2) the national survey on energy consumption by households (CEREN 2008) provides a raw estimate for fuelwood volume at regional level. The survey is carried out every 5 years. Finally, estimates of historical removals can be compared to wood availability estimates .

2.5 NFI Data on Extreme Climatic Disasters

The flexibility granted by the NFI design makes it possible to quickly update the state of forest resources for purposes including estimation of the extent of damages caused by severe windstorms such as Klaus in 2009 (IFN 2009). Every NFI plot was remeasured just after the storm to evaluate both volume of windthrows and remaining standing volumes . Regularly updated initial states of forest resources can hence be made available for wood projection studies as was done in Aquitaine (Colin et al. 2012).

2.6 Auxiliary Data on Technical/Economic Factors of Production

With respect to wood availability assessment and scenario simulation , independent economic data and specific expertise can also be used. Some examples are: harvesting system costs as a function of accessibility , cut types and stand characteristics, stumpage prices as a function of species and stand characteristics, and roadside wood prices as a function of species and assortments .

3 Tools and Simulators

3.1 Main Purposes and Specification of Projections

Five main forest policy issues are raised at national and regional levels: (1) the match between wood supply and demand by industries, (2) the opportunity for the development of new energy-wood industries, (3) the trends in the national timber sector, (4) the contribution of the French forest sector to GHG emissions reduction, and (5) as a matter of increasing concern, the impact of climate change on future forest patterns under different economic and climate scenarios .

Fundamental target variables for estimation include future quantity and quality of growing stock , the availability of wood for different uses, and projections of forest carbon stocks and fluxes.

The typical temporal horizon for forest resource projections ranges between 10 and 30 years with a usual simulation time-step of 5 years. Three spatial scales are typical: (1) a given tree species in a given regional domain, (2) all the broadleaved/coniferous species in a given regional domain, and (3) a comprehensive approach including the entire forest resource of a regional or national territory.

3.2 Types of Models and Their Implementation

In France, specific growth models are not available for many tree species (Pérot and Ginisty 2004), and when these models do exist (typically stand-scale models such as for abundant species including sessile oak, beech or maritime pine), they are usually not appropriate for large-scale projections due to French forest diversity in forest types and ecological conditions, and their non-stationary state. This context has fostered the development of large-scale forest resource models and simulators , as well as a specific expertise based on forest inventory data (Pignard 1993; Buongiorno et al. 1995; Houllier 1995; Wernsdörfer et al. 2012). The description of the current state of forest resources (area, growing stock , etc. distributed by tree species etc.) and estimation of model parameters associated with forest dynamic processes (growth increment, natural mortality and recruitment) are based on NFI data.

Two types of large-scale forest dynamics models are currently used in France for forest resource projections: an age -class based approach for forests where stand age can be assessed, and a diameter-class approach for forests where this is not the case, or not meaningful. Both are currently developed at IGN in close association with the French NFI program, and both are used to address public and private needs (Sect. 13.1.2). The diameter-class approach is also associated with research programs (Wernsdörfer et al. 2012).

Diameter and age -class models are used in projection studies depending on the types of forests to be simulated. The age -class model applies to even-aged forests only (e.g. mono-species conifer plantations, poplar plantation, oak or beech high forests, managed coppices) while the diameter-class model can be used regardless of whether the stand is even-aged or uneven-aged , pure or mixed . Both models are implemented in specific simulation tools after being calibrated and initialised interactively on data extracted from the NFI database. They are run by an expert user .

4 Model Description

4.1 Forest Stratification

Forests are usually partitioned into homogenous subsets or ‘strata’ (Wernsdörfer et al. 2012), and projections are independently performed for each stratum. Stratification factors usually include geographic region or a categorization of the study area (biophysical classifications based on soil and climate such as the ‘sylvo-eco-regions’, IFN 2011a) with a view to encompass variations in growth conditions. Tree species composition is another mandatory stratification factor (e.g. single species, broadleaf-dominated, conifer-dominated and broadleaf-conifer mixed stands). Forest ownership category may be used as a stratification factor because management practices often differ between public and private forests and potentially lead to quite different stand characteristics for the different categories. Stand structure (high forest, coppice forest and high-forest with coppice) can also be used as a stratification factor when the available information is sufficiently accurate. Many other environmental factors can also be taken into account, depending on the context of the study. The number of strata is subject to statistical constraints including a minimum number of NFI plots per stratum to ensure an acceptable sampling error for within-stratum estimates of forest attributes.

4.2 Consideration for Environmental and Silvicultural Factors

In both the age -class and the diameter-class models , forest dynamics are not currently explicitly linked to environmental conditions such as site fertility. Spatial variations in growth conditions are implicitly accommodated in strata-specific estimates of model parameters . In addition, the models do not consider the impacts of climate and environmental changes. Last, density-dependence of forest dynamic processes is also implicit in forest stratification which may be sensitive to non-stationary contexts. This becomes very important when alternative silvicultural scenarios differing from the BAU scenario (Sect. 13.2.3) are simulated. For these reasons, the projection length should be limited to about 30 years.

4.3 Diameter-Class Model

4.3.1 Aim of Model Development

The tree diameter-class model was originally developed to assess future growing stock and wood availability for the medium-term (up to 30 years) within heterogeneous forested areas, i.e., uneven-aged and/or mixed-species stands . The model is also relevant for pure even-aged stands so that, unlike the age -class model , it allows for comprehensive projections for an entire territory (region, country) with various stand types. This model is particularly well-adapted to project forest resources in contexts of valuation of new afforested areas which are rarely even-aged, or where the forests are often heterogeneous such as in mountainous and Mediterranean regions. The diameter-class model is called MARGOT (MAtrix model of forest Resource Growth and dynamics On the Territory scale) and is described in Wernsdörfer et al. (2012). The next model development steps will be to incorporate competition and site as factors affecting forest dynamics.

4.3.2 Modelling Approach

Forest dynamics over time are described by within-stratum tree-diameter distribution and are governed by a Markov transition matrix with constant recruitment. The lower limit of the first diameter class is usually equal to the minimum diameter threshold (7.5 cm) of the French NFI . Three stratum-specific groups of parameters are defined: (1) the proportion of trees within a given diameter class that move up to the next larger diameter class (growth parameter ), (2) the proportion of trees within a given diameter class that are removed through felling and natural mortality (felling and mortality parameters ), and (3) the number of new trees that grow into the lowest diameter class (recruitment parameter ) over a time-step . Further output variables such as merchantable tree volume, biomass or carbon content can be provided by multiplying the number of trees in each diameter class by the respective output variable value estimated for the tree of mean size (i.e. mean merchantable tree volume, biomass or carbon content, respectively).

4.3.3 Assumptions

The current model relies on four assumptions: (1) diameter distributions within different strata change over time independently of each other; (2) forest area , site and competition conditions are constant over time within each stratum; (3) the state of a tree at the end of one time-step depends only on its state at the beginning of that time-step (Markov hypothesis); and (4) no tree can move up more than one diameter class during a time-step , or regress towards a lower diameter class (Usher hypothesis, Vanclay 1994). Some of these assumptions such as constancy of site and competition will be relaxed in the near future .

4.4 Age -Class Model

4.4.1 Aim of Model Development

The stand age -class model applies to even-aged forests, mostly single-species conifer and poplar plantations, high forests of broadleaved species in the plains regions, and managed coppices where intensive management practices are usually implemented. In such homogeneous stands, age is an efficient proxy to estimate growth and to apply forest management practices. The model was described for the first time by Alvarez-Marty (1989).

4.4.2 Modelling Approach

Forest dynamics over time are described by a within-stratum age distribution of the forested area. Each stand cohort is represented by an age class and its associated area, growing stock per hectare, and volume production per hectare. The model simulates the lifespan of each cohort, from its birth (i.e., natural seedlings or plantation) to its death (i.e., clear cut of the last area). The growing stock per hectare of each cohort is recalculated at the end of each time-step of the simulation by subtracting the total volume of thinnings over the period from the initial growing stock per hectare. The mean production per hectare of the age class is finally added to this new growing stock . The area of a given cohort remains the same during the simulation , until it reaches the final cut age . From there, it is progressively reduced over time until zero, thus representing the variety in forest management practices within the stratum, and also the variety in site conditions. Silvicultural operations are defined by two parameters : for age classes up to the final cut age , a thinning rate is applied as a percentage of the growing stock per hectare; for age classes corresponding to the final cut age and beyond, a clear-cut rate is applied as a percentage of the surface area of the age -class. Importantly, reforestation and afforestation within a given stratum are represented by a net influx of new areas in the first age -class.

4.4.3 Assumptions

The model relies on two assumptions: (1) the site and density conditions are constant over time within each age class; and (2) the state of a cohort at the end of one time-step depends only on its state at the beginning of that time-step and not on any previous state. Contrary to the previous approach, the total forest area under study is not fixed and afforestation and land-use issues can be addressed .

4.5 Model Evaluation

Both the diameter-class and the age -class models provide projections that are consistent with historical trends observed in NFI data. A first sensitivity analysis based on NFI data for the years 2006–2008 was carried out by Wernsdörfer et al. (2012) for the diameter-class model. It focused on the sensitivity of forest dynamics to various stratification factors, showing that both varying growth conditions represented by nine large forest regions, tree-species composition (broadleaf-dominated, conifer-dominated and broadleaf-conifer mixed stands ) and stand structure (high forest, coppice forest and high-coppice forest mixture) clearly affect forest dynamics at the larger-scale. Additional analyses on model evaluation are part of the future work envisaged in the frame of further model development.

5 Conclusions and Perspectives

Both large-scale age -class and diameter-class forest dynamics models are used for projecting wood resources in France, and they primarily target forest structures associated with specific silvicultural systems. While age -class models are of immediate advantage for simulation of forest areas of homogeneous stands and mapping their age -structure, diameter-class models encompass forest heterogeneity. However, the use of age as a proxy for developmental stage is questioned, because any historical factor may be confounded in this temporal proxy, including forest management practices, and environmental changes that affect soil fertility over time and create a shift in age -size relationships (Bontemps et al. 2009).

Diameter-class models also directly facilitate volume and biomass prediction using volume and biomass models. Quantification of available roundwood for different uses such as fuelwood, pulpwood and timber still requires further refinement. Recent developments include total aboveground volume models for the main tree species in France (Vallet et al. 2006; Loustau 2010). A new generation of adaptive volume and biomass models has also been developed in the recently completed research project Emerge, funded by the French Research Agency (e.g. Genet et al. 2011; Dassot et al. 2012; Longuetaud et al. 2013).

Accurate projections of wood resources over time require that BAU felling regimes and initial state conditions are correctly quantified. In the past, this was difficult due to only approximate estimates of felled tree volumes resulting from the temporary-plot approach, and the occurrence of major disturbances caused by severe windstorms (e.g. Lothar and Martin in 1999). The forest inventory design implemented since 2004 (Hervé et al. 2014) is a source of major improvements: (1) plots in forest areas affected by severe storms can be quickly remeasured (e.g. in 2009), and (2) felling levels can be accurately estimated by remeasuring plots that were measured 5 years ago (IFN 2011c). Responsiveness in forest resource projections after major disturbances highlights the major role of a continuous and flexible inventory design.

Newer information sources such as aerial images and laser data should also be combined with NFI data in the near future. Multi-source inventory using maps of forest attributes as auxiliary information can improve the precision of forest resource estimates, and finally of wood availability and forest resource projections.

There is increasing demand to account for changes in the socio-economic environment in forest resource projection studies, for instance, to include a cost-approach to account for the feasibility of harvesting operations. This suggests an approach whereby forest characterisation with respect to physical and socio-economic factors is included as an input to such simulations . Additional GIS data sources may contribute to better assessments of the socio-economic context, e.g. the cadastral (ownership ) data may be used to stratify projections according to ownership structure (IGN 2014b), and maps of harvest costs based on physical conditions may help to better identify the economic constraints (Clouet et al. 2010).

All models implicitly take growth conditions into account by varying growth estimates according to geographic domain or strata. In the face of ongoing environmental change impact onto forest growth as assessed from the French NFI (Charru et al. 2014), including climatic change, a future challenge will be to include environmental forcing into forest dynamics models using impact models of growth, mortality , and regeneration . Explicit approaches should rely on the environmental monitoring conducted on NFI plots (e.g. soil pH prediction using plant bioindication, Gégout et al. 2003, or mapping of soil water capacity, Piedallu et al. 2011), highlighting the interplay between forest inventory and environmental monitoring (FAO 2011).

Projections of wood resources are often conducted on a medium-term scale (about 20–30 years), either for regional strategic planning, or for national carbon accounting. In the context of global change, demand is increasing to perform simulations over the longer-term, e.g. 2050 up to 2100. Such applications should, however, not be encouraged as long as slow-developing processes such as changes in the forested area, in the growing stock , in tree species composition and site conditions are not covered at a minimum. Incorporating forest area dynamics into the diameter-class model approach is therefore needed. This also points to the importance of making forest inventory data comparable over time following changes in mensuration protocols.