SEARCH

SEARCH BY CITATION

Keywords:

  • climatic gradient;
  • ForClim;
  • gap model;
  • long-term inventory data;
  • thinning;
  • uneven-aged forest management;
  • validation

Summary

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgements
  9. References
  10. Supporting Information

1. The empirical study of forest ecosystem dynamics is difficult because of the longevity of trees. Many types of models were developed to assist with this problem, all of them with advantages and disadvantages. The strengths of gap models are that they are able to simulate forest dynamics under changing climatic conditions and are therefore suitable for exploring future forest dynamics.

2. Most temperate and boreal forests are managed, making it important to incorporate harvesting functions depicting a wide range of silvicultural practices into the models and to test them under different climatic conditions. This is a necessary prerequisite to the application of these models under climatic change scenarios. Most gap models, however, do not feature such submodels, which disqualifies them as decision support tools.

3. We implemented a management submodel in the gap model ForClim that is able to simulate a wide range of cutting and thinning techniques, including continuous cover forestry (‘plentering’). We tested the new submodel against long-term data (72–111 years) from eight growth and yield research plots across climatic conditions ranging from warm-dry to cold-wet.

4. Stem numbers were simulated accurately in nearly all cases, basal area showed a good fit on Quercus-dominated plots, but an over/underestimation on Fagus sylvatica-dominated and Picea abies-dominated plots. The diameter distributions simulated for the time of the most recent inventory did not differ significantly from empirical data except for two cases. Harvested basal area and stem numbers mostly agreed well with empirical data, but showed the same deviation from reality as simulated basal area.

5. Simulations run with an accurate management plan taken from foresters’ reports for the plots yielded nearly the same results as those run with a generic management setting.

6.Synthesis and applications. We have demonstrated that (i) the management submodel adequately depicts silvicultural treatments, including continuous cover forestry; (ii) a generic harvesting setting can be substituted for a very detailed one, thus eliminating a major source of uncertainty in assessments of future forest dynamics; and (iii) as the new version of ForClim is able to deal with widely differing current climates, it can be employed with reasonable confidence to simulate future management strategies under climatic change. Overall, this modelling work is a major step towards the use of succession models as decision support tools in forest management.


Introduction

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgements
  9. References
  10. Supporting Information

Projecting forest growth from stand initiation to the final cutting is at the heart of forest science, going back to the first yield tables in the late 18th and early 19th centuries. In recent decades, dynamic models have been replacing static yield tables and a wide variety of forest growth models exists today, ranging from highly aggregated models that focus on even-aged, single-species stands (e.g. Grote 1998) to individual-based models that incorporate a detailed consideration of the local environment of every single tree (e.g. Pacala et al. 1996). This reflects the increasing tendency of modern forestry towards mixed-species, uneven-aged stands and the fact that climatic conditions are changing.

Accurately projecting the dynamics of managed forests under strongly changing management and climatic drivers remains a challenge. In the past 20–30 years, dynamic forest growth models have increasingly been used for this purpose (Pretzsch et al. 2008). The formulation of these models is based on past experience embodied in empirical data sets, thus making it difficult to use them under a rapidly changing future climate. Purely mechanistic models, however, tend to be quite complex and are calibrated for a few sites only, are not widely and freely available, or they are costly to calibrate and not easy to use (Kimmins et al. 2005).

The concept of forest gap models (Shugart 1984; Bugmann 2001) deviates strongly from that of forest growth models in the sense that they are formulated more generally and usually do not depend on site-specific parameterizations. Taylor, Chen & VanDamme (2009) postulated that models used to simulate adaptive management ideally should run with data that are readily available to foresters, such as growth and stand inventory information, and Stage (2003) suggested that model complexity should be no greater than that essential to represent the effects of proposed actions, with planning horizons of 50–200 years (Davis et al. 2001). Thus, it appears that gap models are well suited to handle the complexity of rapid changes in management and climatic conditions.

To date, gap models have mainly been used to simulate the dynamics of unmanaged forests (e.g. Botkin, Janak & Wallis 1972; Bugmann 1996; Shao, Bugmann & Yan 2001). The gap models with harvesting options include KIAMBRAM (Shugart et al. 1980), ZELIG (Garman et al. 1992), FORSKA-M (Lindner, Sievanen & Pretzsch 1997), FORMIX 3-Q (Ditzer et al. 2000), JABOWA-3 (Grinter 2001), 4C (Lasch et al. 2005), LINKAGES (Ranatunga et al. 2008) and PICUS (Seidl et al. 2008). However, few detailed descriptions have been published so the assessment of these management functions is difficult. In KIAMBRAM, JABOWA-3 and FORMIX 3-Q, single trees are removed; ZELIG eliminates percentages of the stand; and LINKAGES allows the removal of both whole trees and boles only, leaving branches, bark and leaves behind. In PICUS, harvesting regimes are simulated by reducing the number of trees in one or more of five diameter classes, and in FORSKA-M and 4C trees are removed based on a Weibull function defining the diameter distribution of the parting trees. Typical tests of these functions have covered only short time spans and one single site, which is insufficient to establish their credibility for simulations under global change.

In addition, what is missing in all these models is a function explicitly describing plentering, which is becoming increasingly important, as many forest agencies today promote the transformation of traditional silvicultural practices towards ‘near-natural’ forest management (Gadow, Nagel & Saborowski 2002). There are examples of gap models simulating this specific silvicultural technique, but they rely on approximations, e.g. by simulating several thinnings from above in the early development stages followed by a transition to target diameter harvesting (PICUS, Seidl et al. 2008) or by simulating thinnings in different canopy layers (4C, Kint et al. 2009). There are, however, some individual tree growth simulators capable of simulating uneven-aged forest management, notably SILVA (Pretzsch, Biber & Dursky 2002), MOSES (Hasenauer, Kindermann & Steinmetz 2006) and PrognAus (Ledermann 2001). PrognAus, for example, simulates uneven-aged forest management by using target diameter harvesting combined with structural thinning following de Liocourt (1808) (Sterba & Ledermann 2006). These functions yield good results when compared with desired uneven-aged forest structures (Hanewinkel & Pretzsch 2000).

In this article, we, therefore, evaluate whether a detailed plentering function can be utilized in a forest gap model to depict uneven-aged forest management and whether the same model is able to adequately simulate a wide range of management scenarios subject to different climatic conditions over multi-decadal to centennial time scales. Proof of the latter is a necessary prerequisite for modelling studies with climate change, because no trust could be placed in results from such studies if a model is not able to cope with different current climate conditions. Lastly, we explore whether a generalized harvesting setting can accomplish the same as a detailed one, thus reducing the uncertainty for simulations into the future.

Materials and methods

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgements
  9. References
  10. Supporting Information

ForClim description

ForClim is a gap model designed to incorporate simple yet reliable formulations of climatic influences on ecological processes, while using only a minimum number of ecological assumptions (Bugmann 1996).

The model consists of three submodels (Fig. 1, right): PLANT simulates establishment, growth and mortality of 30 European species on small patches of land. Tree establishment rates are determined from light availability on the forest floor, growing season temperature, soil moisture, minimum winter temperature and browsing pressure. Growth is modelled based on the carbon budget approach by Moore (1989), modified by Risch, Heiri & Bugmann (2005) and Didion et al. (2009). In this approach, the species’ optimal growth rate is decreased based on the degree to which environmental factors (nitrogen availability, growing season temperature and soil moisture) and crown size are at suboptimal levels. Tree mortality is modelled as a combination of an age-related and stress-induced component. The input data for these processes are provided by the submodels WEATHER and WATER, which calculate minimum winter temperature, growing season temperature and soil moisture based on long-term weather data and the stand-specific soil water holding capacity. For a detailed description of the model see Bugmann (1996), Bugmann & Solomon (2000), Risch, Heiri & Bugmann (2005) and Didion et al. (2009).

image

Figure 1.  Structure of the ForClim model with submodels management, plant, weather and water.

Download figure to PowerPoint

Management submodel

We implemented functions describing the following harvesting techniques (Fig. 1, left) in the latest version of ForClim (Didion et al. 2009), resulting in ForClim v2.9.8. Unless otherwise stated, the definitions are taken from Leibundgut (1949).

• Thinning: Reduction in stand density of trees primarily to improve growth. Called ‘thinning from below’ if the individuals are removed from lower crown classes and ‘thinning from above’ if they derive from (co)dominant crown classes (Helms 1998).

• Clear cutting: All trees are removed on a certain area.

• Target cutting: Removal of trees that have reached a certain diameter.

• Group selection (‘Swiss femel’): Gaps are cut into the forest and slowly extended in all directions over several decades.

• Strip felling: Parts of the stand are removed periodically, starting at one end and moving against the main wind direction.

• Shelterwood felling: The main forest body is removed step by step, leaving larger trees to protect the soil and regeneration.

• Continuous cover forestry (plentering): In an uneven-aged forest, basal area is held constant by removing the surplus ingrowth in each class, mainly from the highest diameter classes.

For a detailed description of the management submodel, see Appendix S1, Supporting Information.

Data used for model evaluation

Long-term forest trial plots

The study plots were located in Switzerland because there are a variety of climatic conditions, from warm-dry in the bottom of the central Alpine valleys to cold-wet at high elevations. Also, 14% of all Swiss forests are managed in a ‘near-natural’ way (Brändli 2010), enabling us to test the plentering model function in detail. Data from eight forest growth and yield research plots were obtained from the Swiss Federal Institute for Forest, Snow and Landscape Research (WSL). We chose the plots for their different species composition, environmental conditions, management regimes and sizes (Table 1, Fig. 2), selecting the largest ones where there was more than one plot in a specific stand. The plots extend from the colline to the upper subalpine zone (Ott et al. 1997) and were inventoried at intervals ranging from 1 to 13 years, starting when the stands were between 19 and 43 years old. The surveys include all trees on the plot with a diameter at breast height of at least two (Horgen), three (Aarburg, Galmiz, Hospental, Winterthur, Zofingen) or 8 cm (Morissen, St. Moritz).

Table 1.   Growth and yield research plots used in this study, their location, elevation, area, main species, simulation details on estimated bucket size (BS) or water holding capacity, available nitrogen (N), model patch size, overall number of patches used in the simulation, number of patches representing the plot area once and the simulation period and number of observations (n) available for comparisons
SiteLocation (°N, °E)Elevation (m a.s.l.)Area (ha)Main speciesBS (cm)N (Kg/ha*yr)Patch size (m2)Patch numberPatches/plotSimulation period (n)
Aarburg (ID_41024)47·3 7·94750·25F. sylvatica108083315031890–1994 (18)
Galmiz (ID_42018)46·9 7·14750·3Q. ssp. F. sylvatica128075020041925–1999 (12)
Horgen (ID_02021)47·3 8·66300·5F. sylvatica P. abies1010083330061907–1999 (16)
Hospental (ID_1002)46·6 8·614750·4Pabies, L. decidua, P. cembra108080025051933–2005 (10)
Morissen (ID_1012)46·7 9·216300·5Pabies, Pcembra105083330061929–2002 (10)
St. Moritz (ID_1033)46·5 9·918101·0Pabies, Pcembra, L. decidua1060833600121921–1999 (10)
Winterthur (ID_42005)47·5 8·75050·5Q. ssp. F. sylvatica910083330061928–2001 (11)
Zofingen (ID_41018)47·3 8·05100·25F. ylvatica1010083315031890–2001 (17)
image

Figure 2.  Location of the eight study sites in Switzerland.

Download figure to PowerPoint

Management data

Generally, each time an inventory was made, a silvicultural intervention took place. For each plot, the condition and age of the stand was recorded together with an account of the silvicultural interventions undertaken. These qualitative data are complemented by single tree data, most importantly indicating the time at which they were removed. Thus, we were able to calculate the intervention intensities (fraction of basal area removed) and to determine the targeted species.

Climate and site data

We obtained monthly data for mean temperature and precipitation sum from the database of the Land Use Dynamics Research Group at WSL, spanning the period 1960–2006. The data are interpolated spatially using DAYMET (Thornton, Running & White 1997) to a grid with a cell size of 1 ha. To derive long-term means of the variables mentioned previously, we chose data series from the grid cell directly covering the plot, plus those of its eight neighbours. The daily data from these cells were averaged, and from the resulting series we calculated averages, standard deviations and cross-correlations of monthly temperature and precipitation as required by ForClim. This allowed us to reduce the potential bias associated with using single grid cell data (M. Didion, unpublished).

The site-specific parameters needed for ForClim, available nitrogen [kg/ha*yr] and bucket size [cm], were estimated from the descriptions available for each plot. Beyond these parameters, no other site parameters were adjusted for the simulations.

Simulation experiments

Model initialization

For each plot, patch size in the model was set to a value close to 800 m2 so that it equalled plot size when multiplied by an integer number (Table 1); patch sizes thus varied from 750 to 833 m2, an unproblematic margin, as it can vary from 400 to 1500 m2 without significantly affecting the results (L. Rasche, unpublished). Wehrli et al. (2005) showed that 50 runs is sufficient to reduce stochastic noise in ForClim; hence, we used the single tree information (species, diameter) of the first inventory of each plot to populate the patches representing one evaluation unit and subsequently used this unit 50 times. For a more detailed description of this method see Wehrli et al. (2005) and Didion et al. (2009). Initial leaf area indices of the cohorts were derived from Breuer, Eckhardt & Frede (2003).

Simulation settings

For the simulation of forest management, we used the MANAGEMENT submodel with two different parameter settings: First, we kept to the empirical records, i.e. we let the management submodel intervene in the years the actual interventions had taken place with the same intensity and targeting the same species (below called ‘specific management’). Secondly, to evaluate whether a generic setting leads to the same results as the detailed one, we calculated the mean number of years between interventions and the mean intensities of the treatments and made all species present on the stand eligible for harvesting. Where more than one kind of intervention had taken place during the observation time, we adopted the one used most of the time, or, when they were equally abundant, the one with the highest intensity (Table 2). This is referred to as ‘generic management’.

Table 2.   The management regimes used in the simulations (TB: thinning from below; TA: thinning from above; P: plentering; Te: tending), as recorded for the specific years (n)*, as simplified to one regime, the intensity used for the generic thinning function as percentage of standing basal area to be removed and the interval at which interventions take place in the generic setting
SiteSpecific management (n)Generic managementMean intensityMean Interval (a)
  1. *When management periods do not coincide with inventory periods (Table 1), stands were inventoried without management taking place simultaneously.

AarburgTB (1895–1994)TB0·136
GalmizTA (1931–1999)TA0·167
HorgenTA (1911–1989)TA0·126
HospentalP (1940, 1951–75), TB (1933, 1945, 1985–1995)TB0·138
MorissenP (1929–1940), TA (1945–65), TB (1975–92)TB0·138
St. MoritzP (1927–1999)P9
WinterthurTA (1928–2001)TA0·167
ZofingenTe (1892–1945), TA (1954–2001)TB0·077

The duration of the simulations was determined by the number of years between the first and the last inventory. As no distinction was made in the data between trees removed because of natural mortality, damage caused by hazards and damage/mortality caused by other disturbances such as beetles or fungi (all were referred to as ‘incidental usage’) and given that their numbers were relatively low, we decided to include those trees in the management plan and, in turn, switch off the natural mortality function in ForClim.

The records showed that except in Aarburg and St. Moritz, nearly no establishment took place on the sites; hence, in the simulations tree establishment was allowed solely there. For detailed simulation settings, see Appendix S2 Supporting Information.

Results

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgements
  9. References
  10. Supporting Information

Specific management

In the Aarburg plot, simulated stem numbers of Fagus sylvatica L. agreed well with those of the empirical data (Fig. 3). Basal area was overestimated from 1895 (first thinning) onwards. The gap widened between 1902 and 1913, when the heaviest thinning occurred and diminished again after 1969. The shape of the simulated diameter distribution at the end of the experiment corresponded to the distribution of empirical measurements, but the main peak of the simulation was located in diameter class 54, four diameter classes larger than the direct measurements. Over time, the diameter distributions (Fig. 4) diverged steadily from each other, indicating an exaggerated simulated growth rate of F. sylvatica.

image

Figure 3.  Species-specific basal areas (left panels), stem numbers (middle panels) and overall diameter distributions in the final observation year (right panels) for the eight study sites. Years in brackets: only inventory took place, no management. Grey area in right panel: 2·5th and 97·5th percentile of simulated data.

Download figure to PowerPoint

image

Figure 4.  Diameter distributions for three points in time for the sites Aarburg (upper panels), Winterthur (middle panels) and St. Moritz (lower panels). Grey area: 2·5th and 97·5th percentile of simulated data. Initial simulation years: 1890 (Aarburg), 1928 (Winterthur), 1921 (St. Moritz).

Download figure to PowerPoint

In the Zofingen plot, simulated basal area increased consistently as long as thinning from below was applied (Fig. 3), similar to Aarburg, but it declined sharply with the switch to thinning from above in 1954. Simulated stem numbers, however, corresponded well to the measured data. The overstorey in Zofingen was also similar to Aarburg in experiencing an exaggerated simulated growth rate: the leftmost peak of the simulated diameter distribution preceded the empirical measurements by three classes. The understorey trees, however, lagged behind in their development leaving tree numbers in the medium diameter classes (42–66 cm) under represented.

In Winterthur, stem numbers of Quercus ssp. L. followed the course of the empirical measurements well except for slight overestimates in 1943 and 1948. Basal area was overestimated from 1937 to 1963 and underestimated thereafter (Fig. 3). Simulated basal area and stem numbers were continuously underestimated for F. sylvatica. The simulated diameter distribution corresponded to empirical measurements in overall shape, but there was a substantial underestimation of large trees (classes ≥68 cm) and an overestimation of small trees. This divergence developed particularly in the later years, i.e. after 1972 (Fig. 4). Thereafter, the model not only underestimated the growth rate of the overstorey trees but also started to underestimate Quercus ssp. basal area.

In Galmiz, the simulated basal area and number of stems of Quercus ssp. were constantly overestimated (Fig. 3), whereas for F. sylvatica the opposite applied, in this case more pronounced than in Winterthur. The diameter distribution for Galmiz showed the same characteristics as Winterthur, with a lack of large (classes 70 and 74 cm) and medium trees (classes 18–34 cm) and an overestimation of small trees.

In the mixed stand at Horgen simulated basal areas of all three species (F. sylvatica, Picea abies H. Karst. and Abies alba Mill.) tallied well with empirical measurements, although F. sylvatica basal area was slightly underestimated after 1964 (Fig. 3). Simulated stem numbers matched the empirical data also very closely, although there was a slight underestimation of F. sylvatica stem numbers in 1911. The diameter distribution showed that tree numbers in the medium diameter classes (18–54 cm) were underestimated, while the number of small tress (10–14 cm) was overestimated. The number of trees in the overstorey (classes 58–66 cm) matched those of the empirical data quite well.

Simulated basal area of all species in St. Moritz (P. abies, Pinus cembra L., Larix decidua Mill.) closely mimicked the empirical data (Fig. 3) although L. decidua was slightly underestimated from 1975 onwards. Nevertheless, stem numbers did not closely match the empirical data, there was an underestimation of stem numbers of P. abies and an overestimation of P. cembra. The diameter distribution showed that this overestimation was because of an overabundance of small trees (classes 6–22 cm), whereas the rest of the distribution corresponded quite well to the empirical distribution. This was also true for earlier years (Fig. 4).

In Hospental and Morissen, plentering was used in combination with thinning and as a consequence, the simulated basal areas did not match the empirical data quite as well as in St. Moritz (Fig. 3). All basal areas were underestimated, with the exception of L. decidua at Hospental, which was overestimated particularly from 1975 onwards. An avalanche destroyed parts of the stand in 1975, which is most likely the reason for this trend. Tree numbers and diameter distributions corresponded to the empirical data much better. Nevertheless, the simulated main peaks lagged behind the empirical data by some diameter classes: the number of small and medium trees was overestimated, while very large trees were underestimated.

A quantitative description of model accuracy regarding basal area and stem numbers can be found in Appendix S3 in Supporting Information.

Generic management

The overall shape of the diameter distribution simulated with the generic management setting in Aarburg (Fig. 5) was very similar to the one produced with the specific management setting and also to the measured one. The generic thinning setting was, however, in its intensity somewhat harsher and removed trees up to diameter class 42 cm, whereas the specific management setting only removed trees up to class 34. The loss of basal area via generic thinning was therefore unsurprising (Fig. 6a), although it did produce a greater number of stems overall (Fig. 6b).

image

Figure 5.  Diameter distributions in the final observation year for the eight study sites generated with the generic harvesting setting. Grey area: 2·5th and 97·5th percentile of simulated data.

Download figure to PowerPoint

image

Figure 6.  Comparison of simulated (specific/generic setting) and measured (a) basal area and (b) stem numbers for the eight study sites with standard deviation.

Download figure to PowerPoint

Generic management in Zofingen yielded unsatisfactory results because the mean thinning intensity executed at mean intervals was obviously too severe, resulting in virtually no trees in the medium diameter classes (18–70 cm). The remaining overstorey trees grew too quickly in the simulation and the understorey trees (class 14) did not grow enough. Not surprisingly, the simulated basal area (Fig. 6a) and the number of stems (Fig. 6b) were therefore severely underestimated with the generic management setting.

As with Aarburg, both simulated diameter distributions in Winterthur resembled each other closely (Fig. 5); although the simulated basal area and stem numbers were strongly underestimated under generic management (Fig. 6a, b). The same applied to Galmiz (Figs 5 and 6a, b).

In Horgen, the diameter distributions produced by the specific and generic thinning settings were very similar (Fig. 5), except for tree numbers in diameter class 14, where more trees were left behind under the specific management setting. Overall, the generic thinning setting underestimated basal area more than the specific one (Fig. 6a), whereas stem numbers were similar and reflected reality (Fig. 6b).

The generic management setting in St. Moritz yielded a diameter distribution that corresponded more closely to the empirical data than that obtained with the specific management setting, because tree numbers in the smaller diameter classes and overall stem numbers were lower (Figs 5 and 6b). Basal area did not differ greatly, neither from empirical values nor from those generated with the specific management setting (Fig. 6a).

In Hospental and Morissen, the diameter distributions closely matched the shape of the empirical ones (Fig. 5), but it was obvious that the generic thinning setting targeted the medium diameter classes (Hospental: 18–42 cm, Morissen: 18–34 cm) far more strongly than the specific setting. Basal area and stem numbers were underestimated in both approaches (Fig. 6a, b).

Comparison of harvested stem numbers and basal area

There were no systematic differences between the generic and the specific harvesting setting in the sum of basal area and stem numbers removed (Table 3). Irrespective of the approach, on average, the total number of stems removed per observation period corresponded more closely to the empirical number of stems removed than that of basal area removed (stems were misjudged by ±14% on average, basal areas by ±23%). In the conifer mountain forests, the harvested basal area was underestimated by 35% on average, whereas in the lowland F. sylvatica forests, it was overestimated by 39%. The number of trees harvested from the Quercus ssp. stands and from the mixed stand in the simulations reflected the empirical data quite accurately (±6%).

Table 3.   The sum of harvested stems (N) and harvested basal area (G) over the observation time for the yield research plots, the specific management and the generic management setting. (%): Percentage of simulated in regard to the measured numbers
SiteMeasuredSpecific managementGeneric management
∑N (# ha−1)∑G (m2 ha−1)∑N (# ha−1)%∑G (m2 ha−1)%∑N (# ha−1)%∑G (m2 ha−1)%
Aarburg232084·53103134109129258511166·078
Galmiz214151·2218810246·691224610548·795
Horgen832868·169188369·710269348371·6105
Hospental116565·9118610238·358131811345·369
Morissen132870·0138310449·471134710145·164
St. Moritz51645·63226237·2821843622·249
Winterthur189457·7199510558·2101200110658·9102
Zofingen516859·9520910186·9145517010085·8143

Statistical comparison of diameter distributions

Comparisons of the simulated and empirical cumulative diameter distributions under both management settings with a Kolmogorov–Sminrov test revealed no significant differences except at Horgen (P-values = 0·0141/0·0021 for specific and generic setting, respectively), Winterthur (P-values = 0·0063/0·0006) and Zofingen (P-value 0·0063, generic setting). None of the generically simulated diameter distributions differed significantly from the distributions simulated with the specific management setting. For details see Appendix S4 in Supporting Information.

Discussion

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgements
  9. References
  10. Supporting Information

Experimental setup

To our knowledge, this is the first time that a management submodel has been tested (i) against inventory data from so many different forest types along such a wide climatic gradient, (ii) for such long time periods, (iii) for both a specific and a generic setting, and (iv) without any site-specific calibration of species parameters. Growth models, for example, are usually tested for shorter periods (e.g. 15 years in Mette et al. 2009) or for mono-species stands (e.g. Matala et al. 2003) and need to be calibrated beforehand to the site conditions. Gap models that incorporate management options, e.g. LINKAGES, have been compared to inventory data from four Eucalyptus-dominated plots over a time period of 40 years in terms of biomass and basal area (Ranatunga et al. 2008); Seidl et al. (2005) used data from two long-term observation sites in Austria to compare simulated and observed growing stock and diameter distribution over a time span of 20 years; and Lasch et al. (2005) employed data from one Pinus sylvestris L. stand in Brandenburg (Germany) to evaluate model performance with regard to various measured properties over 61 years.

With eight multi-species sites at very different locations and simulation periods of 72–111 years, our testing regime was much more rigorous and extensive than has been undertaken previously, and it was not self-evident that ForClim would meet this challenge. The model performed well, tracking the development of measured basal area, stem numbers and diameter distribution closely in most cases. This has demonstrated (i) that ForClim adequately embodies the prescribed harvesting techniques, including plentering, and (ii) that dealing with widely different climatic conditions did not negatively influence model performance. Although we acknowledge that this does not prove the model’s applicability under future climate scenarios, the capability of handling widely different current climates undisputedly is a prerequisite for such applications.

There are, however, some exceptions with regard to forest type. In F. sylvatica-dominated stands, the growth rate of the overstorey trees was overestimated at all three sites (Aarburg, Horgen, Zofingen). The trees in the understorey are heavily shaded and consequently inhibited in their growth; this happens in reality, but to a lesser extent than in the model. This is most probably caused by the way light availability is simulated in ForClim, as in reality even in closed stands diffuse light reaches the forest floor from the side (Canham et al. 1990), whereas in ForClim, the leaf area of each tree is distributed homogeneously over the whole patch, not allowing any light from the side.

The overestimation of growth rates of overstorey trees also leads to an overestimation of basal area. Álvarez-González, Zingg & Gadow (2010) showed that F. sylvatica trees in Switzerland experience enhanced growth in basal area after thinning, but the strong increase in F. sylvatica basal area after thinning from below in ForClim is unrealistic. It may, therefore, be advisable to reevaluate the parameterization of the growth rate of F. sylvatica (cf. Heiri 2009). Another option would be to adapt the height growth function, since Lindner, Sievanen & Pretzsch (1997) reported an exaggerated diameter growth rate in simulations of F. sylvatica-dominated plots in Bavaria for the FORSKA model, which decreased to more realistic values after implementing a modified growth function.

On Quercus-dominated sites, the growth rate was under rather than overestimated, especially for Quercus ssp. itself. The empirical data on development at Winterthur indicated that the Quercus trees on this plot grew faster through the diameter classes than F. sylvatica. Schütz (1979) suggested that no foreign yield table was able to capture the growth rate of Quercus ssp. on richer sites in Switzerland and clearly ForClim experienced difficulties as well: the growth rate simulated for Quercus ssp. was similar to that for F. sylvatica, leading to a substantial underestimation of large Quercus trees in the later years. In Galmiz, this problem was not as clear, but the basal area of F. sylvatica was nevertheless underestimated, owing not to the growth rate, but to two other mechanisms: (i) F. sylvatica trees are found mainly in the understorey, leading to the shading problem mentioned previously, and (ii) the medium diameter classes they mainly occupy were selected for thinning most often in the simulation, thus unduly sparing the larger Quercus trees.

On the P. abies-dominated site at St. Moritz, the management submodel captured the nature of the uneven-aged forest management (cf. Zingg et al. 2009). In contrast to the F. sylvatica-dominated lowland forests, however, growth rates in these conifer mountain forests were underestimated by the model, leading to an underestimation of harvesting numbers. Seidl et al. (2005) suggested that precise soil and climate data are a prerequisite for accurately simulating P. abies growth rates, noting an underestimation of growth rates in simulations of a P. abies-dominated colline site in Austria not present elsewhere. Thus, it may be that ForClim does not capture all the factors determining tree growth rates at these elevations. And even though the model considers the degree-day sum – the most influential factor concerning growth rates at these sites (Ott et al. 1997) – more subtle mechanisms like the formation of tree clusters, i.e. collectives of conifers that are clearly separated from their surroundings, narrowly spaced and commonly featuring a near-zero bole height, cannot be captured by ForClim because of the nonspatial nature of the model. This may lead to lower light availability than in reality and thus to lower growth rates.

Simplifying complex silvicultural interventions to generic ones

The comparison between the two different simulation scenarios showed that it is possible to substitute variable intensities and intervals of thinning by mean intensities and average harvesting intervals. There is no significant difference between diameter distributions at the end of the fixed time-span and the deviations of simulated harvesting numbers from measured levels mostly occur within the same error margin. This suggests that when simulating forest dynamics into the future, one can be reasonably confident that a generic harvesting setting will yield results similar to a detailed one without introducing an additional source of uncertainty.

However, the generic management setting tends to underestimate basal area more than the specific setting. The reason for this is the way thinning intensities are defined, i.e. by removing a certain percentage of the growing stock at every intervention. If at any point in time, the average intensity defined in the generic setting is higher than the one that was used at this time in reality, basal area is permanently reduced, and the next percentage to be removed is calculated based on the overly decreased value. Another point of relevance is the mechanism underlying the thinning function: the Weibull distribution function draws the diameters of the trees to be removed based on parameters describing the actual diameter distribution. If the stand is altered because of differing thinning intensities, stand characteristics change as well and different diameter classes are selected for thinning, thus altering the characteristics even more.

Conclusions

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgements
  9. References
  10. Supporting Information

We conclude that the new management submodel in ForClim depicts real management scenarios adequately and that generic settings of the management function can replace detailed ones. Thinnings as well as plentering were executed in a way that reflects reality well. Basal area and stem numbers matched the empirical data reasonably well, and diameter distributions were also captured to a satisfying extent. In terms of removed basal area and stems, the simulation was acceptable for many sites, with some deviations of basal area in F. sylvatica- and P. abies-dominated stands. However, the simulations show that the model is quite sensitive to the thinning intensity employed, probably attaining better results with a cautious estimation of intensities.

Based on these encouraging results, we propose that ForClim v.2.9.8 can henceforth be used as a flexible tool to analyse future management scenarios under climate change and also that it may be not only a valuable tool for researchers but also for decision support in practical forestry.

Acknowledgements

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgements
  9. References
  10. Supporting Information

We are grateful to Petra Lasch of the Potsdam Institute for Climate Impact Research for supplying us with parts of the source code of 4C and to Hanspeter Portner for providing the interpolated climate data from the WSL database. We thank past and present members of the Forest Ecology Group at ETH Zürich for their input and the editors, Antoni Trasobares and two anonymous reviewers for valuable comments on the manuscript. This research was funded by the Swiss State Secretariat for Education and Research under the COST Action FP 0603.

References

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgements
  9. References
  10. Supporting Information
  • Álvarez-González, J.G., Zingg, A. & Gadow, K. (2010) Estimating growth in beech forests: a study based on long term experiments in Switzerland. Annals of Forest Science, 67, 307319.
  • Botkin, D.B., Janak, J.F. & Wallis, J.R. (1972) Some ecological consequences of a computer model of forest growth. The Journal of Ecology, 60, 849872.
  • Brändli, U.-B. (2010) Schweizerisches Landesforstinventar: Ergebnisse der Dritten Erhebung 2004–2006. WSL, Birmensdorf.
  • Breuer, L., Eckhardt, K. & Frede, H.-G. (2003) Plant parameter values for models in temperate climates. Ecological Modelling, 169, 237293.
  • Bugmann, H. (1996) A simplified forest model to study species composition along climate gradients. Ecology, 77, 20552074.
  • Bugmann, H. (2001) A comparative analysis of forest dynamics in the Swiss Alps and the Colorado Front Range. Forest Ecology and Management, 145, 4355.
  • Bugmann, H. & Solomon, A.M. (2000) Explaining forest composition and biomass across multiple biogeographical regions. Ecological Applications, 10, 95114.
  • Canham, C.D., Denslow, J.S., Platt, W.J., Runkle, J.R., Spies, T.A. & White, P.S. (1990) Light regimes beneath closed canopies and tree-fall gaps in temperate and tropical forests. Canadian Journal of Forest Research, 20, 620631.
  • Davis, L.S., Johnson, K.N., Bettinger, P.S. & Howard, T.E. (2001) Forest Management: To Sustain Ecological, Economic, and Social Values, 4th edn. McGraw-Hill, New York.
  • Didion, M., Kupferschmid, A., Zingg, A., Fahse, L. & Bugmann, H. (2009) Gaining local accuracy while not losing generality - extending the range of gap model applications. Canadian Journal of Forest Research, 39, 10921107.
  • Ditzer, T., Glauner, R., Förster, M., Köhler, P. & Huth, A. (2000) The process-based stand growth model Formix 3-Q applied in a GIS environment for growth and yield analysis in a tropical rain forest. Tree physiology, 20, 367381.
  • Gadow, K., Nagel, J. & Saborowski, J. (2002) Continuous Cover Forestry. Kluwer Academic Publishers, Dordrecht.
  • Garman, S.L., Hansen, A.J., Urban, D.L. & Lee, P.F. (1992) Alternative silvicultural practices and diversity of animal habitat in Western Oregon - A computer-simulation approach. Proceedings of the Summer Computer Simulation Conference (ed Luker, P.), pp. 777781. The Society for Computer Simulation International, Reno.
  • Grinter, L.E. (2001) Modeling old-growth forest management in upland oak stands in the Ridge and Valley, Blue Ridge, and Piedmont, USA. Proceedings of the Society of American Foresters, 2001 National Convention (ed Clark, T. & Staebler, R.), pp. 387389. Society of American Foresters, Denver.
  • Grote, R. (1998) Integrating dynamic morphological properties into forest growth and modelling II. Allocation and mortality. Forest Ecology and Management, 111, 193210.
  • Hanewinkel, M. & Pretzsch, H. (2000) Modelling the conversion from even-aged to uneven-aged stands of Norway spruce (Picea abies L. Karst) with a distance-dependent growth simulator. Forest Ecology and Management, 134, 5570.
  • Hasenauer, H., Kindermann, G. & Steinmetz, P. (2006) The tree growth model MOSES 3.0. Sustainable Forest Management: Growth Models for Europe (ed H. Hasenauer), pp. 6470. Springer, Heidelberg.
  • Heiri, C. (2009) Stand Dynamics in Swiss Forest Reserves: An Analysis Based on Long-Term Forest Reserve Data and Dynamic Modeling. Swiss Federal Institute of Technology, Diss. No. 18388, Zürich.
  • Helms, J.A. (1998) The Dictionary of Forestry. Society of American Foresters, Bethesda.
  • Kimmins, J.P., Welham, C., Seely, B., Meitner, M., Rempel, R. & Sullivan, T. (2005) Science in forestry: why does it sometimes disappoint or even fail us? The Forestry Chronicle, 81, 723734.
  • Kint, V., Lasch, P., Lindner, M. & Muys, B. (2009) Multipurpose conversion management of Scots pine towards mixed oak-birch stands - A long-term simulation approach. Forest Ecology and Management, 257, 199214.
  • Lasch, P., Badeck, F.W., Suckow, F., Lindner, M. & Mohr, P. (2005) Model-based analysis of management alternatives at stand and regional level in Brandenburg (Germany). Forest Ecology and Management, 207, 5974.
  • Ledermann, T. (2001) PrognAus - ein Waldwachstumsmodell zur Fortschreibung von Inventurdaten. Institut für Bodenkultur, Wien, pp. 152.
  • Leibundgut, H. (1949) Grundzüge der schweizerischen Waldbaulehre. Forstwissenschaftliches Centralblatt, 68, 257291.
  • Lindner, M., Sievanen, R. & Pretzsch, H. (1997) Improving the simulation of stand structure in a forest gap model. Forest Ecology and Management, 95, 183195.
  • de Liocourt, F. (1898) De l’amenagement des sapinières. Bulletin Trimestriel, Société forestrière de Franche-Comté et Belfort, 4, 396409.
  • Matala, J., Hynynen, J., Miina, J., Ojansuu, R., Peltola, H., Sievänen, R., Väisänen, H. & Kellomäki, S. (2003) Comparison of a physiological model and a statistical model for prediction of growth and yield in boreal forests. Ecological Modelling, 161, 95116.
  • Mette, T., Albrecht, A., Ammer, C., Biber, P., Kohnle, U. & Pretzsch, H. (2009) Evaluation of the forest growth simulator SILVA on dominant trees in mature mixed Silver fir–Norway spruce stands in South-West Germany. Ecological Modelling, 220, 16701680.
  • Moore, A.D. (1989) On the maximum growth equation used in forest gap simulation models. Ecological Modelling, 45, 6367.
  • Ott, E., Frehner, M., Frey, H.-U. & Lüscher, P. (1997) Gebirgsnadelwälder: Ein Praxisorientierter Leitfaden Für Eine Standortgerechte Waldbehandlung. Verlag Haupt, Bern.
  • Pacala, S.W., Canham, C.D., Saponara, J., Silander, J.A., Kobe, R.K. & Ribbens, E. (1996) Forest models defined by field measurements: estimation, error analysis and dynamics. Ecological Monographs, 66, 143.
  • Pretzsch, H., Biber, P. & Dursky, J. (2002) The single tree-based stand simulator SILVA: construction, application and evaluation. Forest Ecology and Management, 162, 321.
  • Pretzsch, H., Grote, R., Reineking, B., Rötzer, T. & Seifert, S. (2008) Models for forest ecosystem management: a European perspective. Annals of Botany, 101, 10651087.
  • Ranatunga, K., Keenan, R.J., Wullschleger, S.D., Post, W.M. & Tharp, M.L. (2008) Effect of harvest management practices on forest biomass and soil carbon in eucalyptus forests in New South Wales, Australia: simulations with the forest succession model LINKAGES. Forest Ecology and Management, 255, 24072415.
  • Risch, A.C., Heiri, C. & Bugmann, H. (2005) Simulating structural forest patterns with a forest gap model: a model evaluation. Ecological Modelling, 181, 161172.
  • Schütz, J.-P. (1979) Production de jeunes peuplements de chênes en relation avec la station. Mitteilungen der eidgenössischen Anstalt für forstliches Versuchswesen, 55, 1178.
  • Seidl, R., Lexer, M.J., Jäger, D. & Hönninger, K. (2005) Evaluating the accuracy and generality of a hybrid patch model. Tree physiology, 25, 939951.
  • Seidl, R., Rammer, W., Jäger, D. & Lexer, M.J. (2008) Impact of bark beetle (Ips typographus L.) disturbance on timber production and carbon sequestration in different management strategies under climate change. Forest Ecology and Management, 256, 209220.
  • Shao, G., Bugmann, H. & Yan, X. (2001) A comparative analysis of the structure and behavior of three gap models at sites in Northeastern China. Climatic Change, 51, 389413.
  • Shugart, H. (1984) A Theory of Forest Dynamics: The Ecological Implications of Forest Succession Models. Springer, New York.
  • Shugart, H.H., Hopkins, M.S., Burgess, I.P. & Mortlock, A.T. (1980) The development of a succession model for subtropical rain forest and its application to assess the effects of timber harvest at Wiangaree State Forest, New South Wales. Journal of Environmental Management, 11, 243265.
  • Stage, A.R. (2003) How forest models are connected to reality: evaluation criteria for their use in decision support. Canadian Journal of Forest Research, 33, 410421.
  • Sterba, H. & Ledermann, T. (2006) Inventory and modelling for forests in transition from even-aged to uneven-aged management. Forest Ecology and Management, 224, 278285.
  • Taylor, A.R., Chen, H. & VanDamme, L. (2009) A review of forest succession models and their suitability for forest management planning. Forest Science, 55, 2336.
  • Thornton, P.E., Running, S.W. & White, M.A. (1997) Generating surfaces of daily meteorological variables over large regions of complex terrain. Journal of Hydrology, 190, 214251.
  • Wehrli, A., Zingg, A., Bugmann, H. & Huth, A. (2005) Using a forest patch model to predict the dynamics of stand structure in Swiss mountain forests. Forest Ecology and Management, 205, 149167.
  • Zingg, A., Frutig, F., Bürgi, A., Lemm, R., Erni, V. & Bachofen, H. (2009) Ertragskundliche Leistung in den Plenterwaldversuchsflächen der Schweiz. Schweizerische Zeitschrift für Forstwesen, 160, 162174.

Supporting Information

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgements
  9. References
  10. Supporting Information

Appendix S1. Detailed description of management submodel.

Appendix S2. Table with further simulation settings.

Appendix S3. Calculation of relative bias and root mean square error (RMSE) of simulated basal area and stem numbers.

Appendix S4. Results of the Kolmogorov–Smirnov test for relative cumulated diameter frequencies.

As a service to our authors and readers, this journal provides supporting information supplied by the authors. Such materials may be re-organized for online delivery, but are not copy-edited or typeset. Technical support issues arising from supporting information (other than missing files) should be addressed to the authors.

FilenameFormatSizeDescription
JPE_2014_sm_AppS1-S4.pdf116KSupporting info item

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.