1 Introduction

Forest coverage in the Netherlands reached its minimum at the beginning of the nineteenth century when only 2% of the land was covered with forest. In the first half of the nineteenth century, afforestation was started, first by private landowners and later, in the second half of the century, by the state. These afforestations aimed at improving the economic value of unproductive areas, mainly drift sand and heathlands. Afforestation continued in the twentieth century, but the main purpose then was to create employment in times of economic crisis (1930s). Later, when the large polders were created, plantations were established to shape the new landscape . At first, new forests in the polders were located only on soils that were not good enough for agriculture, but after 1975 they were also located on better/richer soils. In the last decades, some agricultural land was afforested as part of set-aside arrangements. As a result, forest area increased from 270,000 ha in 1900 to 340,000 ha in 2000, with the largest expansion registered between 1960 and 1980; currently, forest area is 373,480 ha (11% of the land area). In the same period unproductive areas decreased from 620,000 to 140,000 ha. Some forested areas have again been cleared to provide corridors between remaining fragments of heathlands, or to increase windiness in the last patches of drift sand. Additions to forest elsewhere compensate for these losses .

Almost half of the Dutch forest is public property, one third is private, whereas 19% is owned by private nature conservation organisations (Schelhaas et al. 2014). During the second half of the twentieth century, many private owners sold their forest. Because most forests are located on poor soils, the financial returns have been very small and in many cases even negative. Before the 1970s these forests were mostly sold to the state, whereas in later decades the buyers were mostly nature conservation organisations.

As in many European countries, the standing stock in Dutch forests increased enormously over the last 30 years, from 40 to 81 million m3 overbark. This increase is attributed to a combination of aging forest and a decrease in harvesting activities. The area of conifers, especially Scots pine and Norway spruce, is decreasing, while broadleaved forests are increasing fast, from 29% area in 1984 to 45% in 2013. As the consequence of a more nature-oriented management, the percentage of mixed forests has increased, as well as the percentage of uneven-aged forest (16% of the area in 2012–2013). Standing stock in trees outside the forest such as road plantings is estimated at an additional 7–10 million m3. In the nineties, the annual harvest was approximately 1.25 million m3, but decreased to approximately 1 million m3 in the early 2000s. Currently it is again at 1.3 million m3. Private owners in particular tend to harvest less. Many of them own only small patches, do not live close to their forest, and do not depend on the forest for their income; therefore, they are not very active.

Forest industry demands mainly coniferous and poplar wood and has an estimated capacity of approximately 1 million m3. The majority of the wood harvested is thus from conifers (64%). Part of the wood used in industry is imported due to limited year-round availability of Dutch wood. The main reasons for this increase in imported wood are limitations on domestic harvest due to laws for protecting flora and fauna (no summer fellings allowed), increased harvesting costs due to more nature-oriented forest management and greatly fragmented forest ownership . The industry also reported that increased domestic demand for biomass for bioenergy affects the wood market (Oosterbaan et al. 2007).

In 2005, the Dutch Ministry for Agriculture, Nature and Food safety requested a projection for demand and supply of wood for the period 2005–2025, aiming at mapping risks and opportunities for forest owners as well as the woodworking industries (Oosterbaan et al. 2007).

2 Data Sources

Basic data for the supply projections were derived from the fifth National Forest Inventory (NFI5), carried out in the period 2001-2005. NFI5 includes 3622 sample plots based on a 1 × 1 km unaligned systematic sampling design. Trees are measured and recorded on a circular plot. The plot radius is established so that each plot includes at least 20 trees, but with a minimum of 5 m and a maximum of 20 m. All trees with a diameter of at least 5 cm at a height of 1.3 m above the stump are measured, including dead and lying trees. Half of the plots are permanent and are the only plots for which tree coordinates are mapped. In addition to the tree measurements, characteristics of the plot and/or stand are assessed including ownership , stand size, forest type, soil type, and age . Forest type assessment is rather subjective and includes a mixture of appearance of the forest and management aim such as even-aged, uneven-aged, natural afforestation , lanes, nature-oriented management, and parks.

3 Methods

Two forest models were applied in the study by Oosterbaan et al. (2007), the individual tree-based model ForGEM (Kramer et al. 2010; Kramer and Van der Werf 2010) and the large scale scenario model EFISCEN (Schelhaas et al. 2007). The advantage of EFISCEN is that it was designed for this type of study and can be easily parameterised using NFI data. However, EFISCEN is especially suitable for even-aged forests, but Dutch forests have tended to become more mixed and uneven-aged. Therefore, parallel to EFISCEN , the more suitable ForGEM model was used because it accommodates any mixture of species, age and tree size. However, ForGEM cannot readily use NFI data.

Both models were applied to the forest area where timber production could play a role (‘production forest ’), based on the forest type assessment. Production forest excludes forests where other goals are likely to be more important such as nature-oriented forests and parks. The area of production forest totals 240,000 ha. The remaining 120,000 ha of forest, classified as other forest, can in principle at least partly contribute to the market. Estimates for the latter forests were obtained under the assumption of an average increment of 6 m3/ha per year based on the simulated increment for oak forest. Otherwise, the estimated potential supply from trees outside forest was estimated using known information on area distribution over landscape types, and assumed densities and increment of trees in different landscape types, totalling 443,000 m3 per year. Two scenarios were studied, a low-supply and a high-supply scenario .

3.1 ForGEM

ForGEM is a process-based model that tracks the characteristics of individual trees over time. Trees produce biomass from intercepted light, which is subsequently allocated to different tree compartments. Trees produce seeds, which are then distributed over the area. When seeds germinate, they are treated as one cohort until they reach a user-defined height threshold (usually 1 m), after which they are treated as individuals. Trees can die due to old age , competition and late season frost (seedlings only). A range of management interventions are available including thinning from above or below, clearcut, shelterwood and selection system. For some processes the user has options regarding the level of simulation detail, depending on the question of interest. These options include, among others, the choice between the ray tracing method or the gap-type approach for light interception, and processed-based photosynthesis or the simpler light-use efficiency approach. Moreover, it is possible to make specific parameters dependent on the genetic structure of trees. This facilitates studies of the effects of natural or human selection on those parameters , and on the forest in general (Kramer and Van der Werf 2010). The version employed uses the gap-type approach for light interception with 20 × 20 m grid cells and the light use efficiency approach to convert radiation into biomass. The genetic module was not used. The time-step depends on the detail of the processes simulated; for our application a monthly time-step was used. The main drivers are temperature and radiation, which were derived from meteorological data from measurement station de Bilt, located centrally in the Netherlands, for the years 1975–2005 (KNMI 2007).

ForGEM is programmed in the language NSM (Nested Simulation Model), a language developed at Alterra and is written in C++. It can be run from a typical computer, but requires considerable memory capacity; a minimum of 2 Gb of RAM is recommended. Simulation times largely depend on simulation length, and level of detail required for the different processes. Typically the time ranges from several hours to several days, but can range up to weeks for long and detailed simulations . ForGEM has no specific user interface, but uses the general NSM interface.

The light use efficiency version of the model must be calibrated against local growth and yield tables , with different productivity classes represented by different light use efficiencies. In this case medium-productivity Dutch yield tables (Jansen et al. 1996) were used. After calibration on total productivity, results were visually compared to diameter and height development in the growth and yield tables . No further validation was done. ForGEM has been used mainly as a research tool for studies with emphases such as the effect of forest management system on wind damage (Schelhaas 2008), interacting effects of forest management and climate change (Kramer et al. 2008), and expected development of forest reserves (Schelhaas et al. 2005).

ForGEM can simulate an area up to a few hectares, but usually patches of 1 ha are simulated. Exact tree position, species, and stem and crown dimensions are needed for initialisation. Dutch NFI plots have radius between 5 and 20 m, but tree positions are recorded only for the permanent sample plots which comprise 50% of the total sample. Therefore, because ForGEM could not run for each sample plot individually, NFI plots belonging to production forests were grouped to produce a manageable number of forest types to be simulated by ForGEM . Grouping was done by dominant species, species mixture, and type of management (even-aged, uneven-aged, transition). In total, 19 representative groups were distinguished, with a minimum area per group of 3000 ha. Each forest type was simulated as a single stand for one single rotation, starting with an idealised, young, even-aged forest or a typical, uneven-aged situation for the group under consideration.

For each group, the low and a high supply scenario were implemented at harvest levels of 40% and 80%, respectively, of the basal area increment. These simulations represented the typical development of a 1 ha forest per group for the two scenarios for a full rotation. These idealised development curves as functions of age were then combined with the actual area distribution over age classes from the NFI to estimate the total potential of removals over the next 20 years. For example, if 20 plots were present in the forest type “Scots pine mixed with oak” in the age class 40–50 years, the removals for the simulation between ages 40 and 60 year of the corresponding forest type were used as an estimate of the removals on the corresponding 2000 ha for the next 20 years. ForGEM can provide output both at the tree level and the stand level for intervals of months to several years. For this application, the focus was harvest volume by diameter class per tree species. The main stochastic processes included in the model are seed dispersal, age -related mortality and selection of the simulation year to which the year of observed weather is applied. No replicates were used, because seed dispersal and age -related mortality are not important in these simulations , and weather had little influence on the results.

3.2 EFISCEN

The European Forest Information Scenario model (EFISCEN V3.1) simulates the development of forest resources at scales from provincial to European level (Sallnäs 1990; Nabuurs et al. 2007; Schelhaas et al. 2010; Verkerk et al. 2011). Forest resource analyses have been successfully conducted at the pan-European scale with the EFISCEN model for a range of applications.

Input data are usually obtained from NFIs in aggregated form. They can be stratified by province, tree species, site class and owner class, depending on the level of detail required and the size of the resulting groups, hereafter referred to as stand types. The input data are used to construct the initial age class distribution and growth as a function of age for each stand type. Each stand type is assigned a management regime defined in terms of the probability that a thinning or final harvest can be carried out as a function of stand age . For each 5-year time-step , the timber demand from the simulated area must be defined separately for thinnings and final fellings . This total demand is then obtained for the different stand types, according to the felling possibilities as defined by modelled age class distributions and the management regime.

Principal outputs of EFISCEN are age class distributions, growing stock volumes , harvesting levels and increment. Biomass Expansion Factors (BEFs ) for converting growing stock volume to biomass for different tree compartments and turnover rates are used in EFISCEN to estimate carbon stocks in living tree biomass and litterfall from those trees. The litterfall rates are used in the build-in YASSO model (Liski et al. 2005) to estimate soil carbon stocks.

EFISCEN has been validated on historical inventory data for Finland (Nabuurs et al. 2000) and Switzerland (Thürig and Schelhaas 2006). Accurate predictions were obtained at the national scale, but deviations occurred at the provincial and tree species levels due to differences in the distribution of harvest over the stand types. The model can reasonably be used for 50–60 year projections, but the projection horizon is commonly limited to 20–30 years.

The EFISCEN model was applied to the same NFI data again but only production forests were considered. For the EFISCEN simulations , the data were aggregated into eight groups based on dominant species, ignoring species mixtures and uneven-aged structures. Although a fraction of the plots is classified as uneven-aged, they still have a single age which is estimated from the time of regeneration of the original stand. The inputs for EFISCEN are the area, average volume and average increment per age class for each of the groups distinguished. Because no increment data were available in NFI5, we used the increment functions derived from the HOSP study and as implemented for the first European Forest Sector Outlook Study (Schelhaas et al. 2006). Simulations were done over the period 2005–2025.

For forests not specifically managed for wood supply (other forest in Table 20.1), potential wood supply is estimated using estimated increment and the specific scenario percentages, 40% and 80% of increment removed for low and high scenarios , respectively. Because no model projections were involved, removals from this area are assumed to be the same for both the ForGEM and EFISCEN projections. Further, the same 40% and 80% increment removal levels were applied to the potential supply from trees outside the forest.

Table 20.1 Total wood supply according to the projections by ForGEM and EFISCEN (average over the period 2005–2025), plus estimated supply from forest not managed for wood supply and other sources (1000 m3)
Full size table

4 Results

The low supply as projected by EFISCEN just overlaps with the high supply as estimated by ForGEM (Fig. 20.1; Table 20.1). The range between the low and high supply scenarios is much greater for the ForGEM model than for EFISCEN . Under a high demand scenario (+2% increase per year), EFISCEN would be able to supply enough wood under the low supply scenario until 2010, and until 2020 under the high supply scenario . The high supply scenario of ForGEM provides sufficient supply until 2017, while the low supply scenario stays well below the demand throughout the whole period.

Fig. 20.1
figure 1

Historical harvest (actual removals ), forest supply from production forest according to low and high supply scenarios for the two models and projected demand assuming an annual 2% increase (Reproduced from Oosterbaan et al. 2007)

Full size image

The potential supply from other forest and trees outside the forest is estimated at 300,000–600,000 m3 per year. This extra demand would help to compensate for the discrepancy between the supply from the production forest and the high demand scenario .

5 Discussion and Conclusions

5.1 Conclusions From the Report by Oosterbaan et al. (2007)

The projected developments of demand and supply were interpreted as opportunities for forest owners. Shortage in supply would mean increasing prices, which would make operations more profitable. Consequently, a larger amount of roundwood could be brought to market, especially from forests that are currently not harvested for economic reasons. According to industry, the quality of Dutch timber was adequate for their purposes.

Potentially, more timber can be harvested from the Dutch forests. For the 240,000 ha of production forest for which supply was estimated, possibilities for increased harvest are limited, because a high percentage of the increment is already harvested. Therefore, any harvest increase must be realised from the 120,000 ha of other forests where the management goals allow harvest of trees and from trees outside forest such as landscape forest and roadside plantations. The realisable harvest increase from these sources is estimated at 300,000–600,000 m3.

The timber market will be good for all conifer species. Also the supply of timber for broadleaved species will just exceed demand . Broadleaved species can be used to fulfil the anticipated greater demand for energy purposes, but might also be used for replacing species such as spruce and poplar which are likely to experience considerable future supply shortage.

5.2 Reflection

The differences in outcomes between the two models are quite large. For a large part, these differences can be attributed to the uncertainty surrounding the increment. ForGEM is calibrated with yield tables that are known to underestimate increment, whereas EFISCEN was calibrated with more recent data. Moreover, the increment level in ForGEM is influenced by the management during the simulation where low removal rates lead to lower increment. A preliminary comparison with the NFI6 increment values show that projections for conifers, except Scots pine, match best with those projected by EFISCEN , while broadleaves are generally closer to the projections by ForGEM . The overall increment (7.3 m3/ha per year) was closer to the projected value by ForGEM (6.8 m3/ha per year) than EFISCEN (8 m3/ha per year).

There was also considerable uncertainty in area estimates from the simulations due to missing information for a considerable number of plots. Furthermore, the division of the forest into “production forest ” and “other forest” was based on the NFI field assessments. However, these assessments are highly subjective and add more uncertainty . The NFI6 estimate of the area of production forest is 287,000 ha, 18% greater than in NFI5, although it is unlikely that much has changed in the meanwhile.

Altogether the outcomes of the study are very uncertain and can serve only as an indication for the possible supply. Meanwhile, NFI6 was finished, including a re-assessment of the permanent sample plots . This allows a much better assessment of the increment and how much of the area is really harvested. From the 1235 plots that were re-assessed, 38% showed no signs of harvest (Clerkx et al. 2015), indicating that harvest took place only on about 230,000 ha during the last decade. Reasons not to harvest on the other plots varied greatly, including priority to other functions (mostly nature) and the silvicultural state of the forest (Clerkx et al. 2015). A major problem remains in identifying the management goals of the owners, the reasons for acting or not acting, and how owners and managers could be stimulated to increase their harvest level. A future study should take owner behaviour into account, and should involve additional GIS analysis to better delineate areas where harvest is allowed and possible. Also, incorporating cost estimates would help to produce a more realistic picture. Furthermore, it would be beneficial to develop a model that allows each NFI plot to be simulated separately, rather than simulating an average development of a group of plots over an entire rotation. However, the most uncertain factor will remain the forest owners’ behaviour, especially how they will react to changes in prices and policies.