home

 

 

 

Detailed project description

  1. Introduction

  2. State of the art

  3. objectives

  4. working program

         Observational line

         model development line

                     Forest-Wood Chain

                     Environmental

                    Ecosystem

         simulation line

Detailed project proposal

1 Introduction                              

In recent years, the forestry sector has undergone important changes on a global scale. As the importance of forests for sustaining biodiversity, sequestering carbon emissions from fossil fuel use, bio-energy production and social benefits are acknowledged, the forestry sector needs to balance between multiple economic, environmental and social goals (Englin and Richard 1991; Schlamadinger and Marland 2000). In Flanders, the fragmented and limited forest area is under particularly heavy pressure, which increases their recreational, ecological and social importance.
There is a substantial need throughout the forestry sector and related industries for a better understanding of the forest ecosystem and of the forestry-wood chain. More specifically, there is a growing demand for decision-making tools able to predict the consequences of management actions and global climatic change, not only on the quantity but also on the quality of the produced wood, and on the economic, social and environmental impacts at different stages along the life cycle of the forestry-wood chain.
Flanders has considerable arrears in the field of using and integrating existing knowledge into operational decision-making tools about sustainable forest management. Decision support tools for forest management in our region are, even more than in other regions or countries, very premature and disintegrated. The reason is that all actors of the strongly parcelled forest resource (forest managers, policy makers, wood industries, environmental experts, scientists) use their own specific expert tools for each problem at hand. In many cases these tools are imported from other regions without being adapted to environmental conditions in Flanders, e.g. all empirical yield tables commonly used to predict forest production (such as Jansen et al. 1996). As a result, each of these tools fails to address today’s complex forestry problems, ranging from multipurpose management to prediction of wood quality. For effects of climatic change empirical tools fail per definition, while the need for a more fundamental, mechanistic simulation approach is increasing (Bergh et al. 2003).
We firmly believe that the only possible way to create a truly informative, realistic and predictive model for forests – complementary to the existing tools and interconnecting them – is to base the model on existing and new understanding of the principles underlying tree growth, down to the level of wood formation. Such a model will need to incorporate the variability within and between trees, and within the forest microclimate, to be able to predict not only quantities but also wood quality and stand structure. To accomplish this, major and innovative experimental research has to be carried out.
Clearly, different end-users have very different requirements, such as to know the monetary value of the sawn timber, the total carbon(C)-sequestration in the forest ecosystem or the effect on the soil water table. Hence, a simulation model will only be used as a management tool if it is user-friendly, if it yields results without including unnecessary detail or demanding unavailable inputs, and if it can give information on the degree of uncertainty.
 

2 State of the art   
Unlike agricultural crop research, experimental forest and tree research is greatly hindered by the long rotation periods inherent in forestry (20-250 years). Historically, forest management has been based on the sustained yield paradigm, implemented with empirical yield tables, reporting standing stock and increment as a function of tree species, age and site class. Recently, much progress has been made in refining such empirical models to include more forest and tree characteristics as well as some measure of wood quality (Leban et al. 1996; Ikonen et al. 2003; Soares and Tome 2003).
A few accounting models have been expanded to cover the entire forest-wood-chain and the total carbon-budget related to the full life-cycle (Gorcam, Schlamadinger et al. 1997; Efiscen, Liski et al. 2001) and some empirical models simulate the value of the produced wood (Carino and Biblis 2003). Nevertheless, these mathematical models do not have a mechanistic basis and will therefore never be able to predict beyond the range of their input data which are normally generated from empirical yield models. Therefore, simulations for uneven-aged mixed forest stands of changes in management practice (e.g. changes of thinning regime, shift towards natural regeneration), site productivity (e.g. by nitrogen deposition) and climate (e.g. global change) cannot be forecasted by the use of empirical models or by the use of accounting models.
To improve on the drawbacks of empirical models, many attempts have been made to simulate tree and forest growth from the bio-geochemical and physical mechanisms on which their functioning is based. Important insight into the structure and functioning of forests has been gained by these mechanistic models, yielding useful predictions at a global scale (BIOME BGC model, Pietsch et al. 2003; Ecosys model, Grant 2004), stand scale (Biomass model, McMurtrie and Landsberg 1992; Bergh et al. 2003; Secrets model, Sampson and Ceulemans 2000) and individual tree scale (Pretsch et al. 2002; Fritts and Dean 1993). However, these models have generally failed to give appropriate results to allow their use in forest-management applications, although recently serious efforts have been done to combine existing models into more user-friendly platforms (such as CAPSIS model, De Coligny et al. 2003). Overall, the use of mechanistic forest models is still mainly academic.
One of the main reasons explaining why they failed in convincing the forestry sector is that most existing models only predict the average volumes and biomass of the forest or the different tree compartments: only the quantitative yield is predicted. The few models including wood quality (Lemieux et al. 2001; Constant et al. 2003; Ikonen et al. 2003; Mäkelä and Makinen 2003) consider only average trees without detailing variability nor include management simulations. Wood and tree quality and its variability, however, determine the decision-making of the entire forestry-wood chain, from the management of the forest resource to the end-use of wood products.
Obviously, new models should be based on existing algorithms and modules, extending them with new expert knowledge on the link between tree ecophysiology and wood formation. Although several studies concerning this link have been published recently (Savidge et al. 2000; Steppe 2004), the current understanding of ecophysiological processes is not yet strong enough to allow its incorporation into a mechanistic model. However, some interesting first steps towards integration of all existing knowledge have already been taken in Flanders by the members of this consortium. In the PBO project (KULeuven-UA), the mechanistic model Secrets was coupled to the carbon accounting model Gorcam and to the catchment model Swat to predict carbon sequestration in ecosystem and wood products as well as its impacts on the water cycle. In the Afforest project (KULeuven, EU 5th framework), the C&W and Nucsam models were integrated in a spatial decision support system in order to predict effects of farmland afforestation on carbon sequestration, groundwater recharge and nitrate leaching. In the EU project Mefyque (QLKS-CT-2001-00345 – http://www.efi.fi/projects/mefyque/), a preliminary wood quality simulation model at the stand level has been developed by the UA in cooperation with Forest Research (Alice Holt, UK), which is partly based on experimental work from UGent.
Concerning the integration of knowledge, expertise and demands from different stakeholders, the Canadian Model Forest Program is inspiring. In this project, model forests at different sites throughout Canada are being studied with an emphasis on sustainable management and with consideration for the different forest functions. There is also a strong emphasis on including the ideas of the stakeholders at all levels into the project and on optimal dissemination of research results. We envisage to adopt a similar thorough validation and valorisation strategy.
 

 3. Objectives
The strategic objective of the consortium is to develop a physiological forest-wood chain model which is able to compare and evaluate different sustainable forest management strategies with respect to their impact on wood quality, ecosystem functioning and forest structural development (which is of importance in maintaining ecosystem diversity). The existing physiological model Anafore (Analysis of Forest Ecosystems) – derived from the Mefyque project - is used as a starting point and will be gradually elaborated to meet the following main purposes:
1. Characterisation of site impact on external tree structure and variability; phenology, site-variability, phenotypical features, management, competition for light and nutrients [WP1-WP2].
2. Increased understanding of wood formation and variability of wood structure in relation to climate, phenology, species-dependent behaviour, biomechanics [WP3-WP4].
3. Development, validation and verification of a physiological forest-wood chain model; develop three different front-ends that will emphasize different aspects of the forest-wood chain [WP5-WP9].
4. Case studies, to illustrate he capacities and power of the model as a decision-support system for different forestry applications [WP10].

4. Working programme
The work for the consortium will be developed along three main lines: an observational line that investigates the forest and tree variability (WP1-WP4), a model development line that integrates/links both existing knowledge and the new information from the experimental work into a validated and operational model with different front-ends (WP5-WP9) and a simulation line that illustrates the capacities and power of the model in three case studies (WP10).
 

4.1 The observational line
The experimental work will focus on the four main tree species in Flanders being indigenous oak (Quercus robur, Q. petraea), beech (Fagus sylvatica), Scots pine (Pinus sylvestris) and poplar hybrids (Populus x euramericana and P. x interamericana). These are the main species, not only in terms of covered area (an estimated 75 % or over 100.000 hectares), but also in terms of produced wood volume (over 600.000 m³ of commercial round wood) or importance of the (semi-) natural ecosystems they represent (ranging from dry and acid spodosols with Querco-Betuletum on the Campine plateau over fresh and mesotrophic cambisols with Fago-Quercetum, Milio-Fagetum and Querco-Carpinetum of the silt belt to the rich euthrophic alluvial soils with Alno-Padion in the river valleys).
The experimental work will take place in typical growth sites for these species: Meerdaalwoud and Zoniënwoud for oak and beech, the Dijle en Demer valleys for poplar, and some Campine public forests for Scots pine. In contrast to conventional forest inventory exercises, our study will not follow a random or stratified random sampling strategy but will look for two study areas per tree species (in the order of magnitude of 50 ha maximum) with a big variety in land suitability (also called site class) for the considered species (as a consequence of a gradient in soil texture, drainage class, humus type, topography, aspect or any other abiotic site factor). Five distinct site classes will be defined per study area, in each of which ministands will be selected consisting of five proximal model trees, belonging to two developmental stages (thicket stage where stem quality is being formed as a consequence of genetic information, competition and management; young tree stage where stem quality is formed and valuated by e.g. thinning) and two stem quality classes, and their surrounding competing trees of the same species (homogenous stands) or of other species (mixed stands). This yields a total of 200 model trees (50 of each species) and approximately 600 to 800 surrounding trees. Areas with background information on the used genetic material (in terms of clones, varieties or provenances) are preferred. Genetic quality is commonly known for poplar but often unknown for the other species. In WP1 to WP4, site and tree characteristics will be thoroughly inventoried, both concerning the determining factors of microclimate, soil, genetical background, management and the response factors of stand structure, external and internal tree architecture and wood properties. These detailed data are required for our generic modelling approach (WP5-WP8) but will not necessarily all be required as input when using the simulation front-ends (WP9-WP10), since a number of model variables will have fixed effects or be derived from allometric or stochastic relations.
Four experimental levels of characterisation will be distinguished. Using a top-down approach, these are the landscape level (forest, WP1), the plot level (forest stand, WP1), the individual tree level (tree architecture, WP2) and the within-tree level (wood structure, WP3). Ecophysiological processes and cambial activity will be monitored and analysed (WP4) to allow for a better understanding of wood formation and of the variability of wood structure. In this respect, the choice of the four species is justified from the functional wood anatomical point of view, since they are representative of the three main types of stem wood anatomy: oak is a ring-porous species (R-type), beech and poplar are diffuse-porous species (D-type) and pine is a coniferous species (C-type). These three wood anatomical types represent specific growth-strategies in terms of phenological behaviour, cambial activity and hydraulic safety and efficiency. Moreover, the four selected species have different light requirements and play a different role in forest succession. Scots pine and poplar impose high demands on light due to their pioneer character, whereas beech, being a climax species, is able to grow under low light conditions. Oak has intermediate features. Concerning reaction wood, the three broadleaf species (beech, oak and poplar) form tension wood, whereas the coniferous species Scots pine produces compression wood. Recent publications on reaction wood and tree biomechanics will serve to orient the investigations in WP4, such as those from the EU Compression Wood-project and doctoral dissertations on tension wood in poplar, e.g. Jourez (2002) and Badia (2003).
 

4.1.1 Work Package 1: Site mapping at the stand and the catchment level
Lead applicant: KULeuven
For this WP we use a Precision Forestry approach, which starts from the hypothesis that it is possible to detect micro-spatial differences in ecological features and in site quality within a forest stand and to take these differences into account to adapt and optimise the forest management to the local environment and, hence, to reach higher quality production.
The unit of management in forestry is the forest stand, as it is the parcel in agriculture. In forestry practice, it has been recognized since long time that the growth conditions within forest stands are not fully homogeneous and that management practice should take these spatial variations into account. However, nowadays such spatial adaptation in management is done on a purely empirical and observational basis. The principles of Precision Forestry will be applied in this WP, quantifying the aforementioned variations by detailed mapping of micro-spatial differences in site quality, taking advantage of the growing experience with Precision Farming and predictive mapping (also called regionalisation, cfr. Franklin, 1995; Heuvelmans et al . 2004).
Site conditions (elevation, aspect, slope, soil texture, soil depth, pH, base saturation, humus type, a number of soil water holding variables, a number of microclimatic variables) will be determined in each study area (forest units of 50 ha) and in particular in the ministands. For each variable the quickest and cheapest method with sufficient accuracy will be selected. In order to create well-interpolated maps, a large number of measuring points per study area are required. In addition to a sampling of each model-tree (see above), a systematic sampling in a 1-hectare grid will be performed. Sample points for environmental variables (grid points as well as mini-stands) will be mapped with a recently developed instrument for precision forestry, called FieldMap, which integrates GPS, compass, laser, digital calliper, clinometer, Bitterlich Relaskop and field computer. The data will then be used for location-based predictive mapping using geostatistical methods (Parajka et al., 2005). Semivariograms will be calculated and subsequently kriging interpolation will be performed.


4.1.2 Work Package 2: External characterisation at the tree-level
Lead applicant: KULeuven
This WP will focus on the architectural, external traits of the model trees growing at the centre of the mini-stands.
For each model tree, exact location, diameter at different heights (taper), branch-free stem length, crown dimensions and height will be measured and logged with the FieldMap instrument (cf. WP1). The 3-D mapping of standing tree architecture (stem shape, bow, lean, branch or whirl distribution) will be complemented with a laser scanner. At a later stage, these maps may be further refined using a 3-D log scanner, on logs of model trees selected for destructive sampling (WP3). A new methodology to perform such 3-D mapping tasks has been proposed for standing poplar trees and corresponding stem logs by Badia (2003). Furthermore, measures of phenotypical quality, such as habitus, social class or vitality, will be defined and estimated, to be added to the FieldMap database as essential attributes of each model tree. For the competing trees of each mini-stand, a rough dendrometrical description will be given comprising age, diameter, height, crown extent and position relative to the model tree. Measurements of LAI (leaf area index) will be performed by means of hemispherical digital photography, corrected for clumping based on the fractal dimension of the canopy, on model trees as well as competing trees.
 

4.1.3 Work Package 3: Characterisation of within-tree variability in wood structure and properties
Lead applicant: UGent
The tasks carried out in this WP will focus on the variability of the wood structure and of selected physical wood properties within the stem of model trees. The wood properties will not primarily be studied to model "classical" wood technological parameters but rather to elucidate the biological control of wood formation. The internal structure of roots and branches (crown wood) will not be studied in detail; these two compartments will be considered as black boxes.
Wood structure and properties will be studied from a functional wood anatomical point of view. The anatomical pattern of mature stem wood of the three basic types (R, D and C) is fixed but nevertheless subject to spatial and temporal variations which result in a quantitatively different anatomical structure. Thus, wood structure depends on the azimuthal, radial and height position in the stem, which corresponds at any height with a distinct cambial age of the wood. The anatomical variations are paralleled by variations of the physical wood properties, in particular of density and mechanical strength.
A semi-destructive sampling approach will be adopted, which will consist in extracting radial bark-to-pith cores with large-diameter increment borers (preferably 1 cm diameter). Internal structure will be studied at breast-height in all model-trees, by measuring ring-width and earlywood-latewood widths and by performing detailed quantitative anatomical analyses at the intra-ring level along four different radii (N, E, S and W) in all model-trees. Relative proportions of (vasicentric) tracheids, (normal, tension or compression wood) fibers, vessels, axial and ray parenchyma will be quantified. Specific attention will be attributed to the quantitative analysis of the hydraulic system of the model trees with respect to the absolute and relative sapwood-heartwood proportions and the number and (transversal) dimensions of vessels in the broadleaved species considered. These quantitative analyses will be facilitated by use of a LINTAB® and digital image analysis techniques (Visilog scripts). Digital input for image analysis will be provided by an ultrahigh-resolution CT-scanner.
Similarly, wood density and strength variations will be evaluated. Semi-destructive Resistograph® drillings will be performed on all model trees, using a 3 mm-diameter needle, from bark to pith along the four main radii (N, E, S and W) at breast-height and, if necessary, at other positions in the trees, to obtain drilling resistance profiles which are indicative of the radial density variations and possible internal defects. Drilling resistance also gives a fairly reliable measure of radial micro-mechanical strength variations (after correction for wood moisture variations). The internal variations of the physical wood properties will be evaluated also in relation to juvenile wood, reaction wood and defects (knots, cracks, fouling). Within the proposal's budget and time limits, it will not be possible to perform stress-grading analyses. However, the model output concerning lumber quality will allow simulating wood products sorted according to stress grades available on the present-day markets.
Especially the methods for measuring the mechanical strength of massive wood require samples of bigger size. Therefore, at least a part of the model-trees will be sampled more extensively, which implies destructive sampling (i.e. felling). The correct assessment of the variations of the internal structure and of the wood properties with position in the stem will require extensive sampling at different azimuthal, radial and height positions, in a representative number of model trees. This should allow establishment of allometric or physiological, age-dependent and tree-architectural relations between the properties measured at a reference position at breast-height and at other positions in the stem. The destructive sampling will be planned carefully in order not to compromise the research tasks described in this and other WP.
 

4.1.4 Work Package 4: Ecophysiological analysis of seasonal wood formation
Lead applicants: UA-UGent
In this WP, wood formation will be studied and analysed in connection with the seasonal cycles of climate and physiological activity (principal limiting factors are temperature, light and water availability), through sampling and measuring at well-defined time-intervals and at selected positions in a limited number of model trees (five per studied species).
It is widely known that stem circumference changes during the growing season, due to cambial growth (xylem and phloem formation) and variations in water content and water tension which induce stem shrinkage or swelling (Zimmermann 1983). Both cambial growth and shrinkage-swelling behaviour follow diurnal and seasonal cycles. However, cambial growth results in an irreversible increase in stem diameter. As to swelling-shrinkage behaviour, which is reversible, xylem diameter variations can be said to follow the dynamics of canopy evapotranspiration, while sugar transport affects the pattern of diameter variations in the phloem (Sevanto 2003). To be able to distinguish between xylem and phloem increment on the one hand, and swelling or shrinkage variations on the other, two inter-dependent experimental tasks are proposed for this WP:

  • task 4a: continuous high-resolution recording of over-bark circumferential variations at short time-intervals (15 min) by the use of electronic band dendrometers (strain-gauge type) which will be permanently fixed at breast-height on a select number of model trees;
  • task 4b: semi-destructive sampling of newly formed xylem and phloem along four cardinal directions (N, E, S and W) at breast-height, by means of a puncher-type increment borer. This will be performed at weekly (small samples of 5 mm diameter) and monthly intervals (big samples of over 10 mm diameter), depending on the physiological status of the tree, along an oblique line at each cardinal position. During cambial reactivation in spring or towards the onset of dormancy in autumn, for instance, small samples will be taken at weekly intervals. This experimental work will be done on all model trees.
  • A third task (4c) consists in the recording of temperature, light intensity, precipitation and relative air humidity, on a diurnal time-scale (every 15 min), in the vicinity of the studied trees. These recordings should preferably be linked to those monitored in task 4a.
  • In addition, leaf phenology (timing of bud break and LAI development in broadleaved species) will be monitored at weekly or monthly intervals (task 4d), closely linked to the sample timing of task 4b.

The data obtained in tasks 4c and 4d will be used to calculate the potential and actual evapotranspiration of the model trees. The dynamics of wood formation and of the circumferential variations will be analysed at different heights in one tree of each species studied by this consortium.
In the samples extracted in WP 4b, the transversal distribution (2-D) and the relative proportions of (vasicentric) tracheids, fibers, vessels, axial and ray parenchyma will be quantified. The development of reaction wood will be monitored as well. As in WP 3, special attention will be attributed to the quantitative analysis of the hydraulic system and the analyses will be facilitated by the use of digital image analysis techniques (Visilog scripts). As previously discussed (WP3), the imaging will be performed by an ultrahigh-resolution CT-scanner. It should be emphasised that this allows to study samples directly and in 3D, and it does not require the time-consuming sample preparation steps of older quantitative wood anatomical methods, as has been demonstrated by Steppe (2004).
 

4.2 The model development line
The new model will be based on the framework of the existing Anafore model.
The Anafore model was developed by the Mefyque consortium (see 1.2), including applicants 1 and 2 of this proposal, and is currently being validated for coniferous and deciduous forests in Europe. Anafore has unique features that allow its expansion to a full, physiological, forest-wood chain model. Anafore is fully modular and all code has been developed in a consistent and transparent way to allow modifications and expansions in a relatively simple way. The model has been coded in Fortran ‘90.
The main current characteristics of Anafore are:

  • Stand-scale model, but individual trees or tree categories are simulated (uneven aged or mixed stands and competition can be simulated).
  • Works at a daily time step concerning C and N allocation (growth).
  • Includes wood formation as development of the composing tissues (fibers, tracheids, vessels and parenchyma) at a daily time step.
  • Includes detailed photosynthesis (based on Farquhar et al. 1980) and evapotranspiration (based on Dewar 2003) modelling, at a halfhourly timestep, for sunlit and shaded fraction of leaf layers (up to 100 horizontal layers can be simulated within a canopy).
  • Includes branching and crown development as influenced by competition.
  • Includes tree lean predictions, though this module is not fully functional.
  • Includes responses of the forest to environmental change s.a. temperature and CO2 increase, and ozone damage.
  • Includes detailed soil simulator, based on an improvement of the Thornley soil model. The soil model simulates C, N and H2O of the soil dynamically, based on tree and micro-organism functioning (including separate simulation of Mychorriza), over 1 to 10 soil layers.
  • Has been linked to a total carbon-budget model (Efiscen model, Liski et al. 2001) and to sawing algorithms (from the Building Research Establishment BRE, UK, unpublished data) but these simulations are not hardwired into the Anafore model.
     

The main issues that need further development and/or improvement are:

  • Improve model run efficiency by improved data-management and memory allocation.
  • Allow for simulation of mixed stands.
  • Allow for within-stand site variation.
  • Include the total carbon-budget and the total forest-wood chain simulations in the model, from existing knowledge.
  • Include the variability between and within trees, from the experimental data of the consortium. Hereby the model output will move from predicting only the average output to yielding a realistic distribution.
  • Improve the simulation of lean and branching.
  • Include simulation of compression and reaction wood.
  • Improve the simulation of yearly tree phenology (bud burst, switch from early to latewood development, environmental effects on wood cell and C allocation between tissues, changes in specific leaf area and photosynthetic capacity of leaves).
  • Include calculations on the developing forest structure for sustaining forest diversity.
  • Include multiple criteria decision making (MCDM) tools
     

The basic model (version 1) will be used to develop at least three different front-ends (version 2), that will emphasize different aspects of the forest-wood chain; more concrete definition of the front-ends will depend on input from the stewardship councils (see 2.4.1 for details on the stewardship councils):
 

1. Forest-Wood Chain SimForTree
The main target groups for this front-end are forest managers, practitioners and forestry policy informers. The emphasis lies on the entire life cycle of the wood from the forest to the final product. Output includes costs and benefits and environmental effects, such as total carbon-budget, yield, economic value, carbon-sequestration, structural diversity, sustainability, impact on diversity and soil water. Additional output (compared to front-end 2) includes the evaluation of log and wood quality for different wood processing options, s.a. sawing, veneering, wood panel production and new applications like bio-energy production. Input options include detailed management (thinning, pruning…), market and sawmill description. Species parameters and climate scenarios are fixed.
 

2. Environmental SimForTree
Here we target environmental specialists, policy makers, landscape planners and the educational community. This will be a relatively simple version of the model with reduced input requirements (list of species, list of soils, list of climate scenarios and list of management options). Output emphasises on a global overview of costs and benefits and environmental effects at a yearly time-step, such as total carbon-budget, yield, economic value, carbon-sequestration, structural diversity, sustainability, impact on diversity and soil water.
 

3. Ecosystem SimForTree
Developed for the scientific community, this front-end will emphasise on the understanding of the ecosystem functioning. This version of the model will include full options in input (such as physiological parameters of each species can be changed) and output (daily and halfhourly values of photosynthesis, respiration, evapotranspiration, wood formation, density profiles, branching, carbon storage). This version has all options, but at the cost of longer runtime and increased input requirements.
The three resulting models will be almost identical concerning their functioning, though some detailed calculations can be omitted when they are not of interest to the user, to decrease running time of the model (for example halfhourly calculations can be omitted in the first two applications).
 

4.2.1 Work Package 5: Integrating state-of-the-art knowledge
Lead applicant: UA – input from KULeuven and UGent
Starting from the existing forest stand model Anafore, state-of-the-art knowledge will be incorporated in the model. The efficiency of the model will be improved concerning data-management and memory allocation. The main issues to be added are the full carbon-budget and economy of the whole forest as well as the within-stand site variation and species mixture.
A first step is to include the harvesting and sawing and/or pulping of the woody biomass allowing user-defined sawing options (market driven or other), and the end-use of the wood products (sawn timber, pulp, wood-based panels, biomass for bio-energy…). The main difficulty here is to ensure that we integrate all necessary options for the end-users to allow the model to simulate all current and future strategies of wood use.
For the choice of the options to be included, we will depend upon the stewardship councils. In a second step, all carbon-costs for harvesting the trees, for transport, sawing etc. up to recycling and/or disposal will be included.
The third step is to include some market values and the possibility to optimise sawing and management towards maximal gain.
Finally, the model will be parameterised, validated and used for the Flemish situation (for poplar, oak, pine and beech) yielding insight into issues such as:

  • The effects of management and wood use on the total carbon-sequestration.
  • The possibilities of using biomass for bio-energy (such as from short rotation coppice).
     

4.2.2 Work Package 6: Including new expert knowledge towards simulating tree variability
Lead applicant: UA – input form KULeuven and UGent
It is clear that the first version of SimForTree (WP5) will only be able to predict global and large-scale effects on forest growth. To yield a realistic simulation of a specific stand, new knowledge needs to be included in the model. Obviously, since we cannot predict what the experiments will show, we do not know the processes that will be incorporated into the model. The emphasis will be on several modules of the forest model:

  • The phenology module needs improvement concerning earlywood to latewood transition (currently only soil water potential and day length are used as triggers), carbon storage through the year and changes in allocation and leaf characteristics (photosynthesis, N, specific leaf area) through the year.
  • Intra- and inter-tree variation needs to be integrated on a sound mechanistic basis, governed by differences in microclimate and genetic background (yielding a distribution for species parameters).
  • The issue of reaction wood needs to be included based on existing literature (studies have shown the physical loads in different parts of a tree trunk, e.g. Badia (2003) and on new insights from the consortium's experiments (even straight trees can have large amounts of reaction wood and this has not yet been functionally explained).
  • The prediction of lean needs improvement, as it will interact with the reaction wood formation. Additionally effects of gust wind needs to be included. Also, the current Anafore model doesn’t include the effect of soil composition and rooting depth on the tree stability. In this step of the modelling work, the stewardship council will be solicited to provide their expert knowledge on the aspects that seems most relevant for inclusion in the final model.
     

4.2.3 Work Package 7: Integration of multiple criteria decision making (MCDM) tools
Lead applicant: KULeuven
The ultimate goal of the project is to develop tools capable of taking management or policy decisions in multi-interest problems. Multiple criteria decision making (MCDM) tools offer an appropriate solution for this purpose. MCDM refers to making decisions in the presence of multiple, usually conflicting, alternatives, criteria or choices of action, which may be expressed in different units (e.g. Van Elegem et al. 2002). The tools AHP (Analytic Hierarchy Process), ELECTRE (Elimination Et Choix Traduisant la Réalité) and PROMETHEE (Preference Ranking Organization Method For Enrichment Evaluations) are examples of currently available MCDM methods that have been used successfully by applicant 3, e.g. in the EU AFFOREST project (Gilliams et al. 2005a, 2005b). AHP compares alternatives pair-wise, finds a complete ranking of the alternatives and provides an overview of the complex relationships between decision elements (i.e. criteria and alternatives) by structuring them into hierarchies. The ELECTRE and PROMETHEE approaches, on the other hand, are based on building a valid outranking relation. To obtain a ranking or hierarchization of the alternatives, decision weights have to be assigned to each attribute and/or preference functions must be established which express how important the relative differences between alternatives for a certain criterion are for the decision maker or which reflect the degree of advantage of one alternative over another, along with the degree of disadvantage that the same set of criteria has with respect to the other alternative. Usually, the weights are normalized to add up to one (Gilliams et al. 2005a).
Using this MCDM approach, we intend to involve the stewardship councils in the objective ranking of different management scenarios and, thus, in the development of the integrated decision-making platform that should be capable of solving every multi-interest or multi-function problem presented to the front-ends (see 2.4, Valorisation strategy). First of all, the stakeholder representatives seating in the stewardship councils will be solicited to select the variables to be modelled and to predefine specific model assumptions. Furthermore, by answering specific questionnaires which present different management scenarios and subsequent ranking of these scenarios using either AHP ELECTRE or PROMETHEE, the experts will express their individual preferences and an integrated choice or decision will result. This way, the economic, environmental as well as social performance of the forest can be evaluated in relation to the chosen management alternative. In addition to the conceptual validation of the model (WP 8), it also allows for an implicit validation of the practical utility of the front-ends, in particular for users of Forest-Wood Chain and Environmental SimForTree.
The objective decision-making process described above will be incorporated as an MCDM user interface into the model. However, for the MCDM tool to be able to cope with future changes in preferences of the stakeholders, which are yet unknown, not all weights, ranks or preference functions may be predefined. Hence, they can not all be hardcoded as such into the model. Therefore, the MCDM interface will offer predefined decision parameters - i.e. default values as they had been determined by soliciting the experts in the stewardship councils - but will allow the user also to enter self-chosen decision variables. We will be able to test and validate the decision-making performance of the interface through its implementation in the case-studies (WP 10).

4.2.4 Work Package 8: Validation and verification of the model
Lead applicant: UA
After the initial development, parameterisation and conventional statistical validation of the model by comparing simulated to measured results at different levels (within tree variation to stand level), and this for different parameters (anatomical, physiological, biochemical),, the impact of variations in the parameters on the model outcome will be analysed. On the one hand, model output uncertainty caused by uncertainty of model input variables and parameters will be estimated (uncertainty analysis). Monte Carlo simulations can propagate input uncertainty throughout the model. Particular attention should be paid to the determination of the uncertainty distributions of the inputs (probability density functions). On the other hand, a sensitivity analysis can determine the parameters contributing most to the model output and the related output uncertainty. The Monte Carlo technique will be used in combination with multiple linear regression to rank the parameters for uncertainty. Not only will this conceptual validation of the model be used to detect flaws and evaluate its robustness, the outcome should also yield key parameters and insights into species-dependent effects on tree growth and allow detecting links between specific parameters (such as photosynthesis, transpiration) and tree phenology or competitiveness.
Model verification needs to be done using new data, unrelated to the data used for parameterization. Therefore, 1/3 of the data collected during this project, not used for model calibration, will be set aside for this purpose. In addition, data from recent European projects in which the UA is involved (Mefyque, Euroface, Popface, Casiroz) will be used. Once the model has been validated concerning not only total yield of a stand but also yield quality and stand structural development, its use in case studies (WP10) is justified.
 

4.2.5 Work Package 9: Development of three different tools
Lead applicant: UA
In the past, only simple (and therefore often with low predictive value) models have been widely used. Although more sophisticated models are available, they fail to attract users because of the large input requirements and of the overflow in unrequired output. We believe the allocation of substantial manpower to ensure a professional and easy interface is of major benefit to the final outcome of the project.
In parallel to the development of the core model, constant manpower will be allocated to the development of several user-friendly front-ends (including a user-interface and an illustrated help function) that will emphasize specific uses of the model. This work will be performed by an informatics specialist, working for applicant 1. Important input from a stewardship council and from relevant end-users will be sought, to ensure their full involvement in creating tools that are of practical use. The user will be able to choose from lists of input (species list linked to species parameters), ensuring easy and fast runs of the model with standard output (menu driven choice of graphs, tables and some 3D images). However, for the more advanced user, full options of input and output (for instance a density profile of the stem of a specific tree at a specified height or the daily evolution of carbon-pools in the ecosystem) will be available.
Besides standard graphic output (including 3D tree growth) and data tables, full linkage to Excel format (and possibly other spreadsheets) will be ensured to allow users to customize the presentation of the output they require. For each model, a complete manual will be made available to all users.
We envisage developing three different front-ends - Forest-Wood Chain SimForTree, Environmental SimForTree and Ecosystem SimForTree (for details: see above) - of which especially the first two will be capable of taking management or policy decisions in multi-interest problems.
 

4.3 The simulation line
4.3.1 Work Package 10: Case studies
Lead applicant: KULeuven
The capacities and power of the model as a decision-support system for different forestry applications will be illustrated with three different case studies using the three front-ends. These can be seen as first attempts towards a vast number of model applications.
The exact choice and design of the case studies will be subject to discussion with the respective stewardship councils (for case studies with Forest-Wood Chain SimForTree and Environmental SimForTree), and with the international scientific community (for the case study with Ecosystem SimForTree). As a first approximation we propose the following three case studies:

  • Predict changes in wood production quality and quantity as a result of forest management practice, for three Flemish wood-processing enterprises (cf. 2.3.1.1).
  • Predict the ecosystem fluxes of carbon and water for specific forest management and climatic change scenarios, in a representative Flemish forest.
  • Long-term prediction of ecosystem fluxes in a mixed uneven-aged and structurally complex forest, at the stand and landscape scales.