Michael Fuller

Modeling the Impact of Logging on Forest Development

As a postdoctoral fellow with Sean Thomas, in the Faculty of Forestry, University of Toronto, I was asked to build a forest harvest simulator that could be used for computer simulations, or "in silico" experiments, to investigate the effects of different harvest systems on post-harvest stand development. Here, I chronicle the results of those experiments.

Background

In harvesting timber, forest managers must decide which of a range of logging methods, or harvest systems, will best meet their harvest, and post-harvest, goals. In mixed hardwood temperate forests, managers generally prescribe one of two selection management approaches: group selection or single tree selection. In group selection, logging is restricted to a set of small, preselected patches within a stand. A portion of the marketable stems found within each patch are removed, creating moderate sized gaps in the forest canopy. In single tree selection, individual stems are targeted for removal, generating much smaller gaps in the canopy, but a greater number of gaps.

The impact of the gaps generated by logging on the development of the stand depends in part on the mix of species present. Tree species differ in the amount of light they require for optimum growth. As an aid to management, tree species are classified as being either shade-tolerant, and thus able to survive and grow under low light conditions, or shade intolerant, growing best under high light conditions. Over short periods, larger gaps created during a harvest favor shade-intolerant species, whereas smaller gaps favor shade-tolerant species. However, the growth response of the trees growing at the edges of gaps, and seedlings in the understory, can rapidly alter light conditions, as growth fills in the canopy openings.

clear cut photo
For many years, forest management practices focused on improving the yield of commercially valuable species, and therefore clear-cutting (photo above) was a common practice. More recently, there has been a rise in multi-use mandates, which strive to provide benefits to wildlife and human recreation, in addition to commercial harvesting.

A keystone of multi-use management is the preservation of biological diversity. As a consequence, forest managers have altered harvest systems in an attempt to favor a wider range of tree species. In particular, by moving away from clear cutting, toward selection-harvest systems, managers have tried to promote biodiversity. Yet, a great deal of uncertainty exists over whether selection harvest systems actually promote species diversity. Temperate forests generally require two to three decades to recover from logging, and thus many years must pass before the consequences of a given harvest system are known.

Harvest Simulations with SORTIE-ND

Using the C++ programming language, I built a flexible harvest simulator designed to interface with an existing forest growth modeling platform, called SORTIE-ND. The SORTIE-ND platform is an agent-based model developed by Charles D. Canham based in part on the work of Stephen Pacala.

SORTIE-ND simulates the growth, reproduction, and death of individual trees within a given set of environmental conditions. The harvester I developed works in concert with SORTIE-ND to simulate group-selection and single-tree selection harvest systems (other systems can also be simulated). The harvest simulator allowed us to investigate the effects of these harvest systems on the post-harvest development of the forest over multiple decades.

Shown above right are selection harvest patterns produced by the simulator. In the figure, circles represent individual trees. Black circles represent trees that were chosen for removal (i.e. trees that were "cut"). The trees removed represent the same total basal area in each panel, but the clumping of logged trees, and hence the size of the canopy gaps, differs.

Simulated Harvest Patterns

Model Projections

To model stand-level changes, we used data on the relative abundance and biological characteristics of tree species recorded at the Great Mountain Forest reserve of northwestern Connecticut. The Great Mountain forest supports a mix of mid-tolerant (in the middle range of light tolerance) and shade-intolerant species. We simulated the impact of stand growth under two conditions: no-harvest and single-tree selection harvesting. We used these two types of simulation to forecast the reponses of the different species to harvesting. For harvest protocols, we followerd the guidelines published by the Ontario Ministry of Natural Resources.

The figure at right shows the projected changes in proportional stand basal area of four different shade-tolerant species, under conditions of no harvest (blue line) and harvest (yellow line). Basal area for each species is calculated as the sum of the basal area of all of the individual stems of that species found in the stand. For each stem, the basal area is the area of a cross-section of the stem, measured at 3.5m from the ground. The proportional basal area for each species is a measure of it's contribution to the total biomass of a stand.

The simulations revealed that different species can differ dramatically in their response to logging. For example, three of the species shown at right exhibited weak to moderate response to harvesting, whereas the Eastern Hemlock (top right panel) increased dramatically in basal area following the simulated harvest. The simulations revealed that, although logging opened short-term gaps in the forest canopy, the gaps were quickly filled-in by subsequent growth, reducing overall light levels and generating conditions favorable for Eastern Hemlock.

The white dashed and dotted lines in the figures represent the upper and lower 95 percent confidence intervals of the simulations.

shade-tolerant results

The figure below right shows simulated responses of mid-tolerant species. Mid-tolerant species are generally of lower stature relative to shade-tolerant species. Therefore, they often occur below the crowns of the larger shade-tolerant species. The simulations revealed that even without harvesting, the long-term growth of shade-tolerant species can exert a suppressing influence on mid-tolerants. Over time, the cumulative basal area of the mid-tolerant species declines. But the rate of decline is projected to be much faster following a harvest (yellow lines).

Our results indicate that, although single-tree selection harvesting is intended to promote species diversity, it may actually hasten the decline of mid-tolerant species. Thus, our investigation suggests that current harvest practices for northeastern forests may not yield their intended effect, and may actually accelerate the natural loss of biodiversity that occurs in the absence of disturbance.

Summary

The harvesting simulator I built for these experiments has helped us to better understand the complex changes to forest ecosystems that can arise in response to logging. Our simulations indicate that single-tree selection harvesting can lead to:

  • Accelerated declines in biodiversity.
  • Reduced basal area of mid-tolerant species.
  • Substantial changes during the first 15-30 years post harvest.

Caveats

The simulations on which these figures are based consider the effect of light conditions on the growth of individual trees. The algorithms used to simulate forest growth take into consideration many of the processes known to influence the growth of trees. Moreover, a great deal of effort was made by the developers of SORTIE-ND to acquire accurate measurements of species-level parameters, for reliable projections of forest growth and development.

However, no model can perfectly reproduce the behavior of a system as complex as a natural forest. Although we believe the gross character of our projections are reasonably accurate, they should be interpreted with caution. SORTIE-ND is known to underestimate light attenuation in the understory, and the actual rate at which light levels decline post harvest may actually be more rapid than predicted. In addition, our simulations make several unrealistic assumptions about environmental conditions. For example, conditions were assumed to not vary from year to year. Natural forests can show a great deal of annual variation in average rainfall and temperature, and this variation can slow or accelerate conditions that influence changes in species abundance.

Nevertheless, we beleive our simulations have uncovered a potentially important downside of selection management systems. Our research suggests that current harvest systems are not optimal. Yet I believe that further work with in-silico experiments could lead to improvements in harvest practices.

mid-tolerant results

Epilogue

While working as a postdoctoral fellow at the University of Toronto, I had many opportunities to perform field work in northern forests, including the boreal forests of the Yukon Territory. My research on forest harvest systems deepened my understanding of forest dynamics, and helped me to appreciate the potential influence of human activities on managed systems.

Yukon photo