CARBON RETENTION BY REDUCED-IMPACT LOGGING
MICHELLE AMY PINARD
A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY
UNIVERSITY OF FLORIDA
This dissertation is dedicated to John P. and Florence H. Pinard.
I am indebted to many people for their support and contributions to this project. Francis E.
"Jack" Putz first recognized the potential for developing reduced-impact logging techniques in Sabah
as a carbon offset. My mentor, Jack, provided me with encouragement and support throughout my
doctoral program and I am grateful to him for his time, energy, enthusiasm, and concern. Wendell P.
Cropper helped me to construct and evaluate the simulation model included in this dissertation.
Wendell's patience, consistent support, and thoughtful questions helped me to develop my ideas and
progress through this dissertation. John J. Ewel's critical review of this dissertation was instructive;
I am grateful to him for his questions, skepticism and encouragement. Kimberlyn Williams provided
suggestions on methods and emphasis; I thank her for her support. David A. Jones and the staff in
the Department of Botany have provided administrative support. I thank Thomas Sullivan at New
England Electric systems for his encouragement and for working to keep me with adequate funding.
I am grateful to Martin Barker for the unconditional support he provided throughout my doctoral
Danum Valley Field Centre and Tekala Logging Camp staff provided logistical and technical
support in Sabah. The Silviculture Unit of Rakyat Berjaya Sdn. Bhd. (Innoprise Corp.) under the
supervision of M. Rajin and J. Tay carried out the plot-based samples described in chapter 3.
Rangers with the RIL Project and foresters with the Queensland Forest Service (Australia)
contributed to my understanding of the implementation of harvesting guidelines. A. Aribin, D.
Kennard, S. Ducham, C. Alsaffar, A. Smith, and C. Chai assisted in the field. M. Barker, M.
Carrington, J. Cedergren, D. Dykstra, J. Gerwing, J. Harison, B. Howlett, D. Kennard, B. Ostertag,
V. Salzman, and L. Snook provided critical comments on various sections of this dissertation. New
England Electric systems. National Geographic Society, and the Garden Club of America provided
financial support for this project. I thank the Government of Malaysia for allowing me to conduct
research in Malaysia.
TABLE OF CONTENTS
ACKNOW LEDGM ENTS ......................................................... iv
A B ST R A C T .................................................................. v ii
1. THE REDUCED-IMPACT LOGGING PROJECT IN SABAH, MALAYSIA ...... 1
Introduction ............................................................ 1
Scope of D issertation .................................................... 2
Conventional Logging Practices in Sabah .... ............................. 3
The Reduced-Impact Logging Project ....................................... 5
RIL Harvesting Guidelines ................................................ 6
Training in Reduced-Impact Logging Techniques ............................. 11
M monitoring Damage ................................................... 13
The Cost of Reducing Logging Damage ..................................... 13
D discussion ................................................. ........ 14
2. SOIL DISTURBANCE FROM BULLDOZER-YARDING OF LOGS AND
POST-LOGGING FOREST RECOVERY ON SKID TRAILS ................. 16
Introduction .................... ...................................... 16
M ethod s .......................................... .................... 18
R esu lts . . . . . . . . . . . . . . .. 2 1
D iscu ssion ............................................................ 3 1
C conclusions ........................................................... 35
3. RETAINING FOREST BIOMASS BY REDUCING LOGGING DAMAGE ..... 36
Study Site ................................................. ........ 36
M ethod s .............................................................. 37
R results . . . . . . . . . . . . . . . 4 5
Discussion .............................. ............................. 60
4. A SIMULATION MODEL OF ECOSYSTEM CARBON STORAGE
FOLLOWING LOGGING ............................................... 71
Carbon Storage and Patterns of Recovery Following Logging .................. 71
Background and Basic Model Structure ..................................... 73
Methods for Simulations and Evaluation ................................. .. 85
M odel A applications ..................................................... 86
Results and Discussion .................................................. 88
Conclusions ......................................................... 107
5. REDUCED-IMPACT LOGGING AS A CARBON OFFSET .................. 110
Introduction .......................................................... 110
Criteria for Joint Implementation Projects .................................. 111
Valuation of the Carbon Offset Associated With RIL ........................ 115
C conclusions .......................................................... 119
A. HARVESTING GUIDELINES ......................................... 120
B. STEM VOLUME EQUATIONS AND WOOD DENSITIES ................. 122
C. SIMULATION MODEL CODE ....................................... 128
D. SIMULATION MODEL FLOW CHART ............................... 145
E. SENSITIVITY ANALYSIS RESULTS .................................. 150
R EFEREN C E S ............................................................... 155
BIOGRAPHICAL SKETCH ..................................................... 168
Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy
CARBON RETENTION BY REDUCED-IMPACT LOGGING
Michelle Amy Pinard
Chairperson: Francis E. Putz
Major Department: Department of Botany
Global concern over rising atmospheric concentrations of carbon dioxide is stimulating
development and implementation of policies aimed at reducing net greenhouse gas emissions by
enhancing carbon sinks. One option for reducing net emissions is to lessen damage to residual
forests during selective logging thereby retaining carbon in biomass. A pilot carbon offset project
was initiated in Malaysia in 1992 in which a power company provided funds to a timber
concessionaire to implement guidelines aimed at reducing logging damage; in doing so, the utility
gained potential credit towards future emissions reduction requirements. To quantify the reduction in
soil disturbance resulting from the implementation of harvesting guidelines, I measured soil
disturbance associated with ground-skidding in the two areas. To quantify the carbon retained due to
this effort, I compared the biomass both before and after logging of dipterocarp forests logged
according to reduced-impact logging guidelines with forests logged by conventional methods.
Prior to logging, the forest stored approximately 400 Mg biomass ha-1. High volumes of
timber were removed from both logging areas (mean
of the conventional logging area was covered by roads and skid trails; in contrast, 6% of the reduced-
impact logging area was similarly damaged. Skid trails in reduced-impact logging areas were less
severely damaged than those in conventional logging areas; the proportion of skid trails with subsoil
disturbance was less than half that in conventional logging areas. Forty-one percent of the
unharvested trees <60 cm dbh were severely damaged from logging in conventional logging areas in
contrast to 15% in reduced-impact logging areas. One year post-harvest, reduced-impact logging
areas held about 42 Mg C ha' more than conventional logging areas.
To investigate the consequences of reductions in logging damage for ecosystem carbon
storage, I constructed a model to simulate changes in biomass and carbon pools following logging.
Simulation results indicate that the relationship between fatal stand damage and ecosystem carbon
storage is not linear and, at 50-60% fatal stand damage, biomass recovery following logging is
severely limited. Reducing fatal damage from 40 to 20% is associated with a 20% increase in mean
carbon storage over 60 years.
THE REDUCED-IMPACT LOGGING PROJECT IN SABAH, MALAYSIA
Uncontrolled logging and rising atmospheric concentrations of "greenhouse" gases are
distinct problems with somewhat overlapping solutions. Many logging operations in the tropics
involve unregulated and unsupervised selective cutting; though only a small proportion of the trees
are harvested, a large proportion of the forest is damaged (e.g., Fox 1968a; Uhl & Viera 1989;
Johnson & Cabarle 1993). Without costly silvicultural interventions, heavily damaged residual
forests yield little timber and thus are at high risk of conversion to other types of land use. Open
canopies and heavy vine loads, typical of many heavily logged forests, increase forest vulnerability to
fire and further degradation (e.g., Uhl & Buschbacher 1985; Kauffman et al. 1988). Appropriate
timber harvesting methods exist but incentives to implement better practices are lacking in many
countries (Gillis & Repetto 1988). Policies aimed at reducing greenhouse gas emissions may provide
a financial incentive for better logging.
In 1992, 52 nations signed a resolution to adopt policies to mitigate climate change by
limiting emissions and enhancing greenhouse gas sinks and reservoirs (Framework Convention on
Climate Change, UNCED 1992). The convention supported cost-effective approaches to reducing
net emissions and recommended cooperation between nations such as in joint implementation
programs that allow greenhouse gas emissions in one nation to be offset by reduced emissions or
increased sequestration in another.
A wide range of opportunities exists for carbon offset programs in forestry. For example,
estimates of the potential for carbon sequestration have been published for the following activities:
preserving old growth forests (Harmon et al. 1990), controlling forest fires (Faeth et al. 1994),
creating plantations and reforesting degraded lands (Sedjo 1989; Schroeder 1992), increasing
rotation times in plantations (Cropper & Ewel 1987; Hoen & Solberg 1994) and reducing logging
damage (Putz & Pinard 1993). Forestry-based offsets increase terrestrial carbon storage either by
expanding forest cover or by maintaining or improving existing forest for carbon storage. This
dissertation explores the potential for increasing carbon retention in managed forests by reducing
avoidable logging damage. By improving harvesting practices, fewer trees are killed or damaged
during logging and more carbon remains in the forest in living trees. Furthermore, if residual stands
contain more trees of larger diameter than areas conventionally logged, future yields of timber are
also likely be higher.
Scope of Dissertation
The objective of this dissertation is to explore the potential of reduced-impact logging for
offsetting carbon emissions. My interests are principally with relevant ecological and biological
processes and, consequently, my treatment of related political, economic, and silvicultural issues is
superficial. The dissertation contains five chapters. This first chapter introduces the concept of
reducing logging damage as a carbon offset, describes conventional logging practices in Sabah, and
provides an overview of the harvesting guidelines upon which the Reduced-Impact Logging (RIL)
Project is based.
Soil disturbance caused by yarding with bulldozers is the subject of the second chapter.
After a comparison of soil disturbance associated with two logging systems, conventional and
reduced-impact logging, I explore the importance of soil disturbance for forest recovery based on a
study of woody stem densities and species richness across a chronosequence of old skid trails and
The third chapter details quantification of carbon retained in forest biomass due to
implementation of harvesting guidelines. In this chapter, I describe forest biomass, above- and
below-ground stores, before and after logging. I compare logging damage in forest logged by
conventional methods and in forest logged according to RIL guidelines. Finally, I quantify the
carbon retained in biomass due to implementation of the guidelines.
The simulation model of carbon dynamics in dipterocarp forest (Chapter 4) is intended to
simulate forest recovery following disturbance by logging. I use the model as a tool for organizing
information relevant to forest carbon storage and fluxes after logging. Simulation results are
evaluated through a series of sensitivity analyses and comparisons to field observations and
published data. I evaluate the effects of reductions in logging damage on forest carbon storage by
examining output from simulations.
In the final chapter, I discuss several policy issues related to international carbon offset
programs in forestry. I describe how the carbon retained due to the Reduced-Impact Logging Project
might be valued and end with general conclusions about the suitability of reduced-impact logging as
a carbon offset.
Conventional Logging Practices in Sabah
When commercial forests in Sabah are selectively logged (e.g., Kleine & Heuveldop 1993),
all mature trees (>60 cm dbh) of commercial species felled during the first harvest. Trees in the
Dipterocarpaceae represent 90% of the total volume of commercial timber extracted (Sabah Forestry
Department 1989). Sabah's silvicultural system is a modification of the Malayan Uniform System
(Wan Razali 1993); seedlings and saplings present at the time of logging are assumed to replace the
mature trees in a 60 yr logging cycle. Pre- and post-logging inventories are carried out, but the data
are not currently used to prescribe cutting limits or silvicultural treatments (Tang 1987). Tending of
the residual potential or future crop trees through poison-girdling of overstory competitors, though
initially part of the silviculture system, was discontinued because only about a third of the logged
forest retained an overstory (Chai & Udarbe 1977).
In a typical logging operation in Sabah, logs are skidded to the roadside or log landing (flat,
cleared area for storing logs) by bulldozers (a few high-lead cable yarding systems are also used).
On average 8-15 trees are felled per ha, representing 50-120 m3 of timber (Sabah Forestry
Department 1989). Damage to the forest is extensive; as much as 30-40% of the area is traversed by
bulldozers (Chai 1975; Jusoff 1991; Nussbaum et al. 1995), and 40-70% of the residual trees are
damaged (Fox 1968a; Nicholson 1979). These relatively high levels of damage are due to both high
timber volumes extracted and poor harvesting practices. Typically, little pre-harvest planning is
carried out, and the activities of fellers and bulldozer operators are not well-coordinated.
Current forest management practices in Sabah are not sustainable because the volumes of
timber extracted, the area logged each year, and damage to advanced regeneration are all too high
(Sabah Forestry Department 1989). A new forest management system is clearly needed in Sabah and
is presently under development by the Sabah Forestry Department (Kleine & Heuveldop 1993;
Udarbe et al. 1994). As is true for many tropical countries, however, lack of forestry department
staff and difficulties in enforcing regulations over large and dispersed tracts of forest can render even
the best regulations ineffective (Jabil 1983). Programs that provide concession holders with
incentives for better management practices may help stimulate change in the industry.
The Reduced-Impact Logging Project
In 1992, the Reduced-Impact Logging (RIL) Project was established between Innoprise
Corporation, a timber concessionaire in Sabah, Malaysia, and New England Electric system, a coal-
burning utility in Massachusetts, USA. New England Electric provided funds to Innoprise for
training staff and implementing harvesting guidelines (Appendix A) aimed at reducing logging
damage in 1400 ha of their concession (total concession area is z 1 million ha with annual logging of
about 20,000 ha). The carbon retained in the forest due to these efforts could be claimed by the
utility as a carbon offset. Contemporary conventional selective logging practices in the area provide
the baseline for comparison.
The 1400 ha experimental area dedicated to the project is divided between two commercial
forest reserves in southeastern Sabah, a 450 ha tract in Ulu Segama Forest Reserve (50'N,
11730'E, 150-750 m a.s.l.) and a 950 ha tract in Kalabakan Forest Reserve (425'N, 11729'E,
150-900 m a.s.l.). This study is based on data from Ulu Segama only. The project began in May
1992 when woody vines were cut in Ulu Segama; logging is expected to be completed in the second
tract, Kalabakan, by December 1995. The logging crews and forest rangers working in the
experimental area were trained by foresters from the Queensland Forest Service and expert fellers
from Sweden. The harvesting guidelines (Appendix A) were based on best management practices
recommended in Indonesia, Malaysia, and Australia.
The harvesting guidelines developed and adopted by the Reduced-Impact Logging Project
specify practices expected to reduce logging damage and thereby retain more carbon in living trees
and promote post-logging biomass increments. The focus of the remainder of this chapter is the
development and implementation of the Reduced-Impact Logging harvesting guidelines.
RIL Harvesting Guidelines
The reduced-impact logging guidelines were initially drafted from best management
practices recommended by the Queensland Forest Service (Australia) and the Smartwood
certification program of the Rainforest Alliance. The guidelines include specifications for pre-
harvest planning, vine cutting, felling, skidding, and post-harvest site closure. During the project's
first two years, the guidelines have been modified to increase operational efficiency and the
guidelines' applicability to the forest conditions and soils in Sabah. Refinement of the guidelines
involved input from field staff, an international advisory committee, environmental groups, and
representatives of local state and private sector forestry institutions. In the following sections, the
harvesting guidelines are outlined, and some of the issues that emerged during the first two years of
implementation are described.
Knowledge of both the terrain and the distribution of harvestable timber is central to
controlled selective logging. Topographic maps (1:50,000 scale) available for the RIL project area
are unreliable and commercial trees are unevenly distributed. Therefore, pre-harvest planning calls
for preparation of a 100% stock map (1:5,000 scale) of harvestable trees. The stock map also shows
stream and road buffer zones and sensitive areas to be excluded from logging; the map forms the
base for the harvest plan.
The value of these costly 100% stock maps was debated midway through the project. Costs
of stock map preparation represented about 16% of the total cost of implementing the guidelines
(about $53 ha'). As illustrated in a concurrent research project in the area (Cedergren et al. 1994),
without any prior knowledge of the terrain, trained rangers can locate trees to be felled on an ad hoc
basis as they mark extraction routes based on a simple spacing rule that considers tree heights, winch
cable lengths, and terrain. The rangers involved with the RIL project, however, argued that stock
maps are essential for proper planning. In the process of making the maps, they become intimately
familiar with the forest and the timber resource. All subsequent aspects of the harvest plan are
based on the stock map.
Efficiency of logging operations is greatly facilitated by sensible road routings, but prior to
the RIL project, stock maps were not made, and road locations were consequently often suboptimal.
Roads and skid trails generally are located on ridges to avoid steep grades, to facilitate uphill
skidding, to minimize skidding distances and stream crossings, and to reduce the amount of sidecast
soil entering streams. Main extraction routes and landing areas are located on the stock map and
then these locations are checked in the field and marked with paint. The end points of all skid trails
are also clearly marked.
Rangers mark trees to be felled with a record number and with a vertical paint blaze to
indicate the intended direction of fall. Also, potential crop trees of good form and larger than 20 cm
dbh are marked with a ring of blue paint if they are at risk of being damaged from felling or skidding.
Tree enumeration and marking for directional felling were not done simultaneously during the pilot
project, but combining these two activities will increase operational efficiency.
About one year prior to logging, all vines with stems > 2 cm dbh are cut. Figs are protected
and no cutting is done in buffer zones. Vine-cutting reduces felling damage because the tree crowns
are not tied together, and it reduces post-felling vine infestations because there are fewer fallen vine
stems to resprout (Fox 1968a; Putz 1991). The effects of vine cutting on arboreal animals and forest
biodiversity in general deserves attention from researchers but has not yet been addressed in the RIL
The need to cut vines before logging is primarily a problem for tropical forestry. In our
study area, density of woody vines > 2 cm dbh averaged 586 stems per hectare (SD = 211, N= 104
plots in 4 logging units, Chapter 3). The utility of vine cutting is debated, perhaps because the
impacts of vine cutting on logging damage are not always obvious and the cost is substantial (about
$US 25 per ha). Certainly part of the reduction in the number of trees uprooted during logging (a
decrease from 37% of the residual trees in areas logged conventionally to 13% in areas logged
according to the guidelines) is related to vine cutting. Several studies designed to measure the
decrease in logging damage due to vine cutting in Malaysia reported a benefit (Fox 1968a; Liew
1973; Appanah & Putz 1983; Cedergren et al. 1994).
Decisions about felling direction are based on feller safety, ease of skidding, avoidance of
damage to harvested and potential crop trees, and minimizing impacts on buffer zones. Trees are
felled and winched towards pre-marked skid trails. Directional felling reduces damage to potential
crop trees and facilitates skidding both by avoiding the need to reorient logs and by shortening the
overall extraction distance by up to the length of the log (up to 30 meters).
Recommended in the guidelines but not required is the use of plastic wedges during felling.
Wedges give fellers added control and help fellers discern small changes in direction of lean as the
tree is felled. Fellers working on the RIL project have not adopted the use of wedges and argue that
plastic wedges are unnecessary and their required use actually places fellers at risk. When needed,
fellers use wooden wedges made ad hoc at the tree to be felled.
Fellers consistently drop trees within 10 degrees of the marked direction (Project Records,
unpubl. data). Success in directional felling of huge trees with eccentric crowns on steep slopes is
impressive, but it should be pointed out that tree markers select the silviculturally optimal direction
from what they judge to be the possible range. Furthermore, during the first year of the project, the
number of harvestable trees felled in the experimental area was less in RIL areas than in comparable
areas logged using conventional methods, perhaps in part due to more frequent rejection of trees by
fellers, uncertain of their ability to fell the tree in the direction indicated. More training would
increase the skill and confidence of the fellers, thereby increasing the arc over which trees can be
felled and reducing the number of harvestable trees left standing. During the pilot phase, rangers
marked trees with the assistance of fellers because the rangers felt they had insufficient training to
determine possible felling directions. This process, however, is subject to undo influence by fellers.
Forest rangers need to be trained to so as to be able to select felling directions from the full range of
technically possible directions.
Winching and Skidding
Bulldozers are destructive machines that were not designed for skidding logs, but their utility
in logging heavily stocked primary dipterocarp forests cannot be ignored; in eastern Sabah, the
average log weighs 7-9 tons, and 50-200 cubic meters of timber typically are harvested from each
hectare. Under the RIL guidelines, main skid trails are constructed by logging crews following
rangers' paint blazes. During extraction activities, bulldozers are restricted to these main trails. The
guidelines call for extensive use of bulldozer-mounted winches to move logs from the stumps to main
skid trails. The weight of the 32 mm cables, however, precludes winching over distances greater than
about 15 meters. The use of a second small winch to pull out the heavy cable deserves investigation.
By restricting the practice of blading surface soil and sidecutting, the deleterious effects of
skid trails are reduced. In the first 450 ha to be logged according to the RIL guidelines, skid trail
area averaged 3.4% of the total area logged in contrast to 12% in adjacent areas logged by
conventional methods (Chapter 2). Further, the percentage of the skid trails with subsoil exposed
averaged 38% in the RIL areas in contrast to 87% in the conventionally logged areas. Many tropical
soils are highly erodible; the presence of a litter layer on the soil surface can reduce soil erosion
substantially (e.g., Ross & Dykes 1993).
Skidding logs with bulldozers is difficult, dangerous, and particularly destructive on slopes
greater than 15-20 degrees. In commercial forests elsewhere in the world, ground-based yarding is
restricted to slopes less than 17 degrees (30%, e.g., Dykstra 1994) because environmental damage
increases greatly as slope increases (Brady 1984). The RIL project guidelines limit bulldozers to
slopes less than 35 degrees (70%). Trees on slopes greater than 35 degrees can be felled only if they
can be winched and skidded from a position on a slope 35 degrees or less. The 35 degree cut-off
reflects a compromise between reducing soil damage and foregoing timber in a large portion of the
remaining commercial forest in Sabah. For example, over 20% of the first parcel dedicated to the
project (450 ha in Ulu Segama Forest Reserve) included slopes greater than or equal to 35 degrees;
the net area inaccessible due to the slope restriction was even greater as some less steep areas were
surrounded by steep areas. Consequently, the volume of timber removed from the first 450 ha
reduced-impact logging area was possibly 20% less than what might have been extracted by
conventional selective logging.
The issue of loss of timber harvested due to slope restrictions is the primary focus of current
negotiations about the future of the project. Arguments are being made by the concessionaire that
the slope restriction should be relaxed because overall damage to the forest can still be minimized by
careful planning of skid trail locations, directional felling, etc. An underlying assumption in their
argument is that damage to the soil is less important than damage to the residual stand. An
alternative solution would be to combine aerial with bulldozer yarding. The current proposal for
project expansion involves a combination of helicopter and bulldozer yarding and incorporates the
higher extraction costs associated with helicopter system into the cost of the carbon offset. Where no
incentive exists to protect the resource, the additional extraction costs associated with aerial yarding
are difficult for the concessionaire to justify.
The restriction against wet weather skidding, although certainly important for minimizing
soil damage (DeBonis 1986), slowed harvesting operations substantially in the RIL areas. The
delays experienced by the contractors increased the overall costs of extraction. Though comparative
financial assessments of selective logging in Sarawak, Malaysia (Marn & Jonkers 1981) and
Suriname (Jonker 1987; Hendrison 1990) suggest that reduced-impact logging costs less per cubic
meter of timber than conventional approaches, our experience with the RIL project in Sabah suggests
that damage-controlled logging may cost more than conventional logging.
Logging Area Closure
After logging is completed in a 40-60 hectare unit, the skid trails are closed through
installation of cross drains at specified intervals (e.g., < 20 meters on slopes 15-20 degrees). The
goal of the guidelines for skid trail marking, construction, use, and closure is to reduce overall
damage to the forest. If erosion is minimized, the same skid trail network should be utilizable when
the stands are logged in 30 to 60 years. Although the drain spacings recommended in the RIL
guidelines seem fairly standard, the field staff has argued convincingly that on some skid trails,
cross-drain construction would increase disturbance to soils. If surface soils are protected from
blading and skid trails are properly located, installing drainage structures may not be justifiable on
hydrological grounds. Inspection of skid trails on slopes of 15 20 degrees that were subjected to as
many as 30 bulldozer passes and three months of heavy rains revealed no signs of gullying. Skid
trails with an intact root mat and litter layer are uncommon in the conventionally logged areas (mean
of 4 logging units = 1.6%), but they represent 12% of the skid trails in the RIL areas.
Training in Reduced-Impact Logging Techniques
Successful implementation of the reduced-impact logging guidelines depends on substantial
technical expertise on the part of sawyers, bulldozer operators, and forest rangers. Traditionally,
Malaysian forest rangers are trained in mensuration and inventory methods, but their familiarity with
harvesting techniques is limited. Sawyers and bulldozer drivers receive no explicit training but
apprentice for several years before becoming operators. The Reduced-Impact Logging Project
sponsored training for representatives of several levels in the forest management hierarchy. One of
the first project activities was a visit by senior Innoprise staff and logging contractors to areas
managed by the Queensland Forest Service. Although it would have been better to have visited an
actively managed forest, seeing one that had been carefully logged was nonetheless valuable. Several
of the Australian foresters who hosted the ICSB visit then came to Sabah as advisors in
implementing the reduced-impact logging guidelines. Ten tractor drivers and fifteen ICSB field staff
worked with three experienced Australian foresters for three weeks. During this training period,
timber in a logging block of approximately 50 hectares was harvested.
Sawyers were trained by a Swedish specialist in directional felling during two 5-day training
programs. Although these programs undoubtedly increased the fellers' abilities to direct the fall of
trees, more training is clearly needed. Furthermore, forest rangers need to be trained so as to be able
to select felling directions from the full range of technically possible directions. These rangers may
themselves serve as future instructors, a situation from which considerable advantage will derive in
regard to effectiveness, cost, and ease of implementation.
The people most responsible for success of the Reduced-Impact Logging Project are the
Innoprise forest rangers, most of whom are high school graduates with one year of formal forestry
training. The rangers supervise and participate in stock mapping, vine cutting, tree marking for
directional felling, and skid trail planning, construction, use, and closure.
Compliance with the reduced-impact logging guidelines and verification of reductions in
logging damage are assessed by an independent team consisting of three foresters, one appointee of
New England Electric systems (a representative from Rainforest Alliance), one appointee of
Innoprise Corporation (a representative from the Forest Research Institute Malaysia), and one joint
appointee (a representative from the Department of Forestry, University of Florida). The team,
referred to as the Environmental Audit Committee, conducts 5-10 day site inspections twice per year.
During these inspections, the team walks through the logging area and evaluates adherence to the
guidelines and levels of logging damage. Also, the Committee meets with the field staff, loggers, and
researchers responsible for logging damage studies and the carbon calculations for the offset due to
reduced-impact logging. The Committee's involvement is anticipated to increase the project's
international credibility, critical for qualification as a carbon offset. The rangers' records of logging
damage provide data for monitoring the contractor's performance and for verifying compliance with
the guidelines. The data provided in this dissertation also provide baseline data for carbon offset
The Cost of Reducing Lo2gging Damage
As mentioned earlier, operational delays due to wet weather shut-downs increase extraction
costs. Also, as compared with conventional practices, felling times are slower when following the
RIL guidelines due to time spent marking and preparing trees for felling (Chua 1986a; Tay unpubl.
data). The additional planning, mapping, and monitoring activities also increase extraction costs as
compared to the conventional method.
Conversely, bulldozer maintenance costs are low in controlled logging sites, presumably
because of less side-cutting and blading, because the steep, rocky areas are avoided, and because the
total length of skid trails constructed is much reduced. Also, total skidding time is less when
following the RIL guidelines due to shorter skidding routes and less search time (Chua 1986b; Tay
unpubl. data). The denser stocking of potential crop trees in areas with reduced logging damage
eliminates the need for costly rehabilitation with enrichment planting and shortens felling cycles. It is
still too early to provide a comprehensive view of the costs and benefits of the project, but an
economic analysis is underway.
While reducing logging damage does not guarantee sustainability, it is a general prerequisite
for good management of selectively logged forest. Managing a forest sustainably makes economic
and ecological sense for long-term concession holders that want to stay in the timber business, but
the appropriate financial incentives seem to be lacking. More effective appear the incentives for
conversion of logged-over forest to non-forest uses (e.g., oil palm plantations). Management decrees
initiated by forestry departments, though frequently based on sound management principles, are often
rendered ineffective due to a lack of enforcement capacity. For example, in the Forestry Department
in Sabah one professional forester is employed per 93,000 hectares of commercial forest reserve
(Sabah Forest Department 1989).
Alternatives to bulldozer yarding on steep slopes need to be developed that are acceptable to
local loggers. For example, demonstration of successful, commercial operation of skyline yarding
systems in selectively logged tropical forests would help establish the viability of this method.
Training in designing, rigging, and operating skyline systems is also needed.
Though the RIL Project has received accolades in the press (e.g., Miller 1994), expansion of
reduced-impact logging carbon offset projects is predicated on acceptance of the concept of joint
implementation in both developed and developing countries. Several developing countries are
outspoken against some types of cooperative programs to abate climate change and are suspicious of
the motivations of developed countries. Furthermore, if developing countries that are signatories of
the Global Convention of Climate Change sell their inexpensive carbon offsets to outsiders, they will
be left trying to satisfy the terms of the Convention with more costly offsets such as radically
modifying their power-generating and fuel-consuming industries.
Alternately, if foreign utilities can produce appropriate financial incentives, concession
holders may be tempted to endure outside assessments of their forest management. Carbon offset
money could absorb the operational costs associated with altering harvesting systems and decreasing
extraction rates. A reduction in extraction rate associated with adoption of better harvesting
practices will move concessionaires toward sustainability and closer to qualification for timber eco-
certification. Once sustainability is within reach, the profit margin and other advantages of certified
timber may drive concessionaires further toward better management practices.
SOIL DISTURBANCE RESULTING FROM BULLDOZER-YARDING OF LOGS AND
POST-LOGGING FOREST RECOVERY ON SKID TRAILS
In East Malaysia, though only 8-15 trees are extracted per ha, typically 15-40% of the area
is traversed by bulldozer paths (Chai 1975; Jusoff 1991; Nussbaum et al. 1995). There are
alternative harvesting systems that cause less soil disturbance, for example skyline (Miller & Sirois
1986) or helicopter (Blakeney 1992) yarding, but these techniques are generally more expensive than
ground skidding on all but the most difficult terrain (e.g., Aulerich et al. 1974). One of the goals of
the Reduced-Impact Logging Project in Sabah was to reduce the area with soil disturbance while
using existing equipment and personnel; bulldozer and chain saw operators were trained in damage-
control techniques and harvesting guidelines were implemented in 1400 ha of dipterocarp forest
(Chapter 1; Pinard et al. 1995). In this chapter, I compare soil disturbance associated with ground
skidding in areas logged using conventional and reduced-impact logging techniques. To explore the
importance of minimizing damage to soils for forest recovery, I examine tree regeneration on
abandoned skid trails.
In the process of extracting logs from the forest with bulldozers, soil is disturbed in a
number of ways that affect forest recovery. First, topsoils are displaced by the bulldozer blade
during skid trail construction; displaced soil (hereafter, sidecast soil) is dispersed over slopes or
forms linear mounds along the edges of skid trails. Although total soil organic matter content may
not change across the entire logged area, its distribution does (Johnson et al. 1991), with bulldozed
areas losing, and sidecast mounds accumulating, soil organic matter (Gillman et al. 1985; Rab
1994). These localized losses in organic matter can have substantial effects on soil fertility (Gillman
et al. 1985; Zabowski et al. 1994) and tree seedling growth and survival (Nussbaum et al. 1995;
During ground-based log yarding operations, subsoils are exposed and churned by the tracks
of the bulldozer. Soil losses from these denuded areas can be substantial (e.g., Homrnbeck & Reinhart
1964; Ross et al. 1990). A hydrological study of recently logged dipterocarp forests in Sabah,
Malaysia, documented stream sediment loads 14 and 2.5 times that of a nearby unlogged catchment
during the first and second year after logging, respectively (Douglas et al. 1993); eroding roads and
gullied skid trails were identified as the principal sources of post-logging sediment. Installation of
proper drainage structures on skid trails, roads, and landings can reduce erosion substantially (Stuart
& Carr 1991; Wenger 1984).
Soil structure is also damaged due to compaction from loads applied by bulldozers and logs
skidded across the forest floor. As soils are compacted, soil porosity decreases, often causing
decreased water infiltration and increased surface runoff, as well as decreased soil moisture
availability, aeration and rooting space (Greacen & Sands 1980; Malmer & Grip 1990). During
heavy rains, seeds and seedlings may be washed away (Borhan et al. 1987; Pinard et al. 1996). Soil
bulk density values recorded in many post-logging habitats are within the range of values that
negatively affect tree growth (Greacen & Sands 1980; Rab 1994). In some forests, changes in soil
physical properties due to logging are apparent decades after logging (Congdon & Herbohn 1993;
Van der Plas & Bruijnzeel 1993).
The extent and degree of soil disturbance associated with bulldozer yarding are variable and
appear to be related to slope (Dymess 1965; Stuart & Carr 1991), soil texture (Daddow &
Warrington 1983 in Clayton 1990; Jusoff 1992), and soil moisture content at the time of logging
(DeBonis 1986; Jusoff 1992). Certain logging practices also influence soil damage, for example,
size of logs extracted (Dickerson 1968) and extent of bulldozer blade use (Miller & Sirois 1986).
Pre-harvest planning can increase the efficiency of log extraction and reduce the area damaged
(Froehlich et al. 1981). Prohibiting wet weather skidding, skidding on steep slopes, and use of the
bulldozer's blade can reduce further soil damage associated with logging.
To describe the reduction in soil damage achieved in the Reduced-Impact Logging (RIL)
Project area, I compared areas logged using conventional and RIL techniques in terms of extent and
degree of soil disturbance. To better understand the impacts of soil damage for forest recovery, I
studied a series of skid trails in areas logged using conventional methods in 1976, 1988, and 1991.
Short-term studies of pioneer tree establishment on skid trails and log landings suggest that,
during the first year after logging, tree establishment is limited by unfavorable site conditions, not by
seed availability (Pinard et al. 1996). Herbivore damage and trampling of tree seedlings on skid
trails are also commonly observed (pers. obs.; Moura-Costa & Lundoh unpubl. data). If skid trails
are unfavorable for tree regeneration, I expect sapling densities to be lower on skid trails than in
adjacent residual forest. If site conditions on skid trails become more favorable for tree
establishment over time, I expect that sapling densities on older skid trails would be more similar to
those in adjacent residual forest than those on younger skid trails.
The study was conducted within the Yayasan Sabah concession in Ulu Segama Forest
Reserve (50'N, 1 1730'E, 150-750 m a.s.l.). Prior to logging, the tall, diverse forest is dominated
by dipterocarps (see Chapter 3 for more details). Soils are orthic acrisols, nutrient rich in the upper 5
cm then dropping steadily in concentration through the soil profile (Nussbaum 1995); the upper
horizons have a loamy texture and are well-drained. The conventional timber harvesting system used
in Sabah, as well as the harvesting guidelines being followed in Reduced-Impact Logging (RIL)
areas, are described in detail in Chapters 1 and 3 of this dissertation. The key differences between
the two systems are as follows: 1) RIL follows a pre-harvest plan with locations of all skid trails
identified on a stock map of trees to be harvested, whereas conventional logging involves little or no
pre-harvesting skid trail planning; 2) RIL restricts bulldozers to slopes <35 degrees, whereas
conventional logging has no slope restriction; and 3) RIL restricts bulldozer blade use and
encourages the use of the winch cable, whereas conventional logging does neither.
To determine the extent and severity of soil disturbance associated with logging I mapped,
measured, and classified all soil disturbance associated with bulldozer activity in eight logging units
(approximately 50 ha each). Four units were selected randomly from a 450 ha experimental area
logged according to the RIL guidelines by trained crews and closely supervised by forest rangers.
For comparison, four units were selected randomly from an adjacent area logged by unsupervised and
untrained crews using conventional methods.
I used three broad disturbance categories: 1) roads and log storage landings; 2) bulldozer
paths (skid trails); and, 3) areas covered by sidecast soils. Roads and log landings generally are
leveled and graveled surfaces on subsoils. Skid trail surfaces are variable and were further classified
by degree of soil disturbance as follows: 1) subsoil exposed, either by blading or heavy bulldozer
churning; 2) churned but topsoil mixed with upper layers of subsoil; and, 3) compacted by bulldozer
passing over area but relatively little mixing of topsoil with subsoil. In the eight logging units, 100%
of the area was surveyed for soil disturbance caused by logging. I measured lengths and slopes of
roads and skid trails by sections; a section was a length of road or skid trail that was relatively
uniform in slope, width, and direction. Widths were measured every 10 to 15 m or, for more rapidly
changing sections, in the midpoint of each section. Contiguous areas of sidecast soils (e.g., linear
soil mounds or tips) were also measured; for large areas with sidecast soils adjacent to roads and skid
trails I measured the average slope and distance to the end of each soil mound (or slide). No effort
was made to measure areas crushed or scraped during the winching of logs to the skid trails. The
area of disturbed soil was calculated based on net loggable area per subblock (defined in Chapter 3).
T-tests performed on arcsine-transformed data were used to compare treatments.
Plant Regeneration on Skid Trails
To describe woody plant establishment on skid trails I sampled old skid trails in 1994 in
three logging coupes (1991, 1988, and 1976). Within each logging coupe, skid trails were located in
four logging units that were separated by at least 1 km. Main skid trails originating at log landings
or roads were selected in all cases. Skid trails were easily located in all three logging coupes. Often,
the edge of the skid trail was marked by an uneven soil surface, probably the result of side-cutting
with the bulldozer blade.
In each logging unit, I established 10 sampling points at 20 m intervals along a skid trail
with the first point located at a random distance (0-20 m) from the landing or road. At each
sampling point, three 2x2 m plots were established, one in the center of skid trail, another at the edge
of the skid trail, and a third 10 m into adjacent forest, following a line perpendicular to the skid trail.
The width of the skid trail was measured from edge to edge. Random numbers were used to
determine whether the edge and forest plots would be placed to the left or the right of the skid trail;
edge plots did not include skid trail surfaces though often sidecast soil was included. Within each
plot I recorded the following: canopy cover (above 1 m) using a spherical densiometer (Lemon
1957), number of woody stems (>1 m tall, <5 cm dbh), and number of species. Trees >5 cm dbh
were not included in the samples because the plot size was too small to adequately sample their
densities at this level of replication. All dipterocarps (i.e., commercial species) and colonizing tree
species (e.g., Macaranga spp.) were noted as such. For plots on the surface of the skid trail, 1
subplot (1 m2) was randomly selected for determination of above- and below-ground biomass. All
vegetation was clipped at ground level, weighed, and subsampled for dry weight determination.
Coarse roots (>5 mm diameter) were collected from a 50 x 50 x 50 cm pit located in center of
subplot; roots were washed, and live and dead roots were separated, weighed and subsampled for dry
For all analyses, logging units were considered replicates, and the plots within each unit were
considered samples. Analysis of variance followed by Tukey multiple comparisons was used to
compare stem densities, species richness, and canopy cover among the three logging coupes and the
three habitats within each coupe. To compare skid trail width and biomass in skid trails in the three
logging coupes, Kruskal-Wallis tests were used, followed by Tukey-type nonparametric multiple
comparisons (Zar 1984). In all cases, the significance level used to reject the null hypothesis was
A greater area of soil was disturbed in conventional units than in RIL units (t = 5.6, df= 6, P
= 0.001; Fig. 2-1; Table 2-1). Road area was similar in the two treatments (t = 1.04, df= 6, P =
0.34), but skid trail area was much less in the RIL units than in conventional units (t = 4.95, df= 6,
P = 0.003). Including only logged areas, mean skid trail density was much higher in conventional
units (mean = 199 m ha', SD = 35.8) than in RIL units (mean = 66.5 m ha', SD = 25.7; t = 6.0, df=
6, P < 0.001).
Total volume of timber extracted per logging unit was not statistically different between the
two methods (t = 1.88, df = 6, P = 0.11; Fig. 2-2); however, high variability and low replication limit
Figure 2-1. Diagrams of the eight logging units in which soil disturbance was measured. (CNV =
conventional logging areas, RIL = reduced-impact logging areas). Thick black lines represent
roads, thin black lines are skid trails, blackened areas are log landings, stipled areas are riparian
zones, and the hatched area is a landslide below a road.
Table 2-1. Soil disturbance in conventional and reduced-impact logging units (100% area); N =
4 per treatment. Skid trail area includes area covered with sidecast soil. Values are
mean percentages (SD) of logged areas.
Conventional Logging Units Reduced-Impact Logging Units
Total Area Disturbed (%)'" 16.6% (2.3) 6.8% (2.6)
Roads and Landings (%) 4.7% (0.8) 3.3% (2.5)
Skid Trails (%)"** 11.9% (2.7) 3.5% (2.1)
I I I
Timber volume extracted (m3 )
Figure 2-2. Total skid trail area (ha per logging unit) related to timber volume extracted (m3 per
logging unit) for reduced-impact logging areas (solid circles) and conventional logging areas
the power of this analysis. Excluding unlogged sections within units, mean volume extracted was
136 m3 ha- (SD = 29) in conventional units and 92 m3 ha' (SD = 40) in reduced-impact logging
units (Pacific Hardwoods Sdn Bhd, unpubl. data). Skid trail area (including sidecast mounds) per
timber volume extracted was greater in conventional units (mean = 8.8 m2 m3, SD = 0.56) than in
RIL units (mean = 4.6 m2 m3, SD = 3.04; U = 15, df= 1, P = 0.04). Including road area, soil
disturbance per harvested tree was 140 m2 tree' (SD =16) in conventional and 94 m2 tree' (SD = 28)
in reduced-impact logging areas. Skid trail disturbance was positively correlated with volume
extracted for conventional units (Pearson Correlation Coefficient = 0.97, P = 0.03) but not for RIL
units (Pearson Correlation Coefficient = 0.53, P = 0.54).
Within the area disturbed by skid trails, the severity of disturbance to the soil was greater in
conventional than in RIL logging units (Table 2-2). Skid trails with a bladed surface (or sidecut)
were predominant in the conventional units (mean = 87.2%, SD = 5.6%) whereas only 38% (SD =
9.9%) of the skid trails in the RIL units had a bladed surface. The most common surface condition
for skid trails in the RIL units was churned (i.e., the topsoil remaining in place but being mixed with
the upper layer of subsoil; Table 2-2). Skid trails with intact topsoil and litter layer were very
uncommon in conventional logging units but covered about 12% of the skid trail surfaces in RIL
units. In these compacted areas, saplings and vines resprouted soon after logging.
Plant Regeneration on Skid Trails
The width of skid trails surveyed in the three logging coupes ranged from 3.9 m to 6.0 m;
mean width in the '91 coupe (mean,, = 5.4 m, SD19 = 0.7) was greater than that in the '88 and '76
coupes meang = 4.1 m, SD88 = 0.2; mean76 = 3.9 m, SD76 = 0.2; F=13.2, df= 2,9 P = 0.002;
Tukey's Test P < 0.05). Both skid trail and forest habitats in the two older logging coupes had
nearly closed canopies at the time of sampling (Table 2-3). Skid trail tracks in the '91 coupe, had
more open canopies than edges or adjacent forest plots. For all three logging coupes, species
Types of soil disturbance recorded in conventional and reduced-impact logging units
(N = 4 per treatment) presented as mean percentage (SD) of total area logged.
Conventional Logging Units Reduced-Impact Logging Units
Area With Sidecast Soil(%)*** 2.1% (0.2) 0.4% (0.5)
Skid Trail Surface Area (%)'" 9.9% (2.7) 3.2% (1.6)
Bladed (%) 87.2% (5.6) 37.7% (9.9)
Churned (%) 11.1% (4.9) 50.2% (7.3)
Compacted (%) *1.6% (1.1) 12.1% (9.5)
** P < 0.01
Characteristics of vegetation in 1994 on skid trail tracks, skid trail edges, and
adjacent forest in three logging coupes. Species richness refers to all woody stems
>1 m tall and <5 cm dbh. All values are means (SD) per plot for four logging units,
each unit with 10 sampling plots (2 x 2 min). Different superscripted letters within a
row denote a significant difference (P < 0.05) between habitats within a coupe using
Tukey multiple comparisons following ANOVA.
Coupe Skid Trail Track Skid Trail Edge Forest
Canopy Cover '91 66% (14)a 83% (3)b 89% (2)b
'88 90% (3)' 92% (4)" 94% (2)"
'76 93% (2)b 90% (2)" 93% (2)b
Species Richness '91 1.8 (1.1)8 5.5 (2.2)b 7.0 (0.7)b
'88 1.1 (0.1). 3.8 (1.1)b 5.1 (1.0)b
'76 2.7 (0.9)a 5.2 (1.0)b 6.5 (1.0)b
richness (woody plants >1 m tall, <5 cm dbh) was lower on skid trail tracks than in edge or adjacent
forest habitat (Table 2-3).
Fewer saplings were found on skid trail tracks than on skid trail edges or adjacent forest in
all three logging coupes (Table 2-4). The '91 and '76 logging coupes had identical mean sapling
densities on skid trail tracks although variance was higher in the '91 coupe than in the '76 coupe
(Table 2-4). Forest habitats in all three coupes had relatively few pioneer tree saplings. Pioneer tree
saplings were relatively abundant on skid trails and edges in the '91 coupe. Edges in the '88 coupe
had more pioneer saplings than either the skid trails or adjacent forest. In the '76 coupe, pioneer tree
sapling density was similar in the three habitats.
Dipterocarp sapling density was less on skid trails than adjacent forest in the '76 and '91
coupes (Table 2-4). In the '88 coupe, the three habitats had similar densities ofdipterocarp saplings.
The observed densities of dipterocarp saplings on skid trail edges and in adjacent forest habitats were
similar to densities recorded for unlogged forest (mean = 430 saplings ha"', SD = 158; 1993 coupe)
though the diversity was quite low in the '76 coupe where Hopea nervosa was dominant.
Aboveground biomass on skid trail tracks was extremely low ( 1.1-2.2 Mg biomass ha') and
was similar for the three logging coupes (KW= 1.5, df= 2, P = 0.47; Table 2-5). Coarse root
biomass under skid trails followed a general pattern of more biomass under older skid trails. Under
skid trails in the '76 coupe, coarse root biomass was greater than it was under skid trails in the '91
coupe (Table 2-5; q = 4.16, P < 0.05). Coarse root biomass under skid trails in the '88 coupe was
intermediate and not statistically different from either the '91 or '76 year old skid trails (qy9 ,8 =
2.08, P > 0.05; q-6 ,,,, = 2.08, P > 0.05). Dead root mass under skid trails was much higher in the
'91 coupe (median,, = 6.6 Mg necromass ha-', N= 4) than in either the '88 or '76 coupes (median88
= 1.0; median76 = 0.6, Table 2-5). Total coarse root mass was not different among the three logging
coupes (KW=3.04, df= 2,P = 0.22). Woody roots >15 mm diameter made up about 43%, 50%,
Stem densities (>1 m tall, <5 cm dbh) in 1994 on skid trail tracks, skid trail edges,
and adjacent forest in three logging coupes. All values are means (SD) per ha;
densities were calculated from 4 m2 plots, 10 sample plots per unit, four units per
coupe. Different superscripted letters within a row denote a significant difference (P
< 0.05) between habitats within a coupe using Tukey multiple comparisons
following ANOVA. Similar results were obtained when habitats were compared
using frequency data in contingency table analyses.
Coupe Skid Trail Track Skid Trail Edge Forest
# Saplings and Vines '91 8,130 (4,880)a 22,060 (8,670)b 22,500 (2,280)"
'88 3,500 (710)' 12,630 (3,350)b 15,880 (2,950)b
'76 8,130 (2,070)' 18,380 (1,830)b 22,750 (4,410)"
# Pioneer Tree Saplings '91 2,630 (2,630)"b 4,690 (2,100)' 560 (800)"
'88 1,000 (890)b 3,560 (970)' 880 (720)b
'76 310 (320)' 440 (720)a 60 (130)'
# Dipterocarp Saplings '91 60 (130)a 560 (560).b 810 (130)b
'88 130 (250)' 250 (500)' 310 (470)2
'76 0 (0). 440 (330)b 1,560 (1,390)"
Table 2-5. Above- and below-ground biomass and necromass for three ages of skid trails.
Values are medians (Mg organic dry mass ha' with range noted parenthetically) for
N = 4 logging units (n = 10 samples per unit). Different superscripted letters within
rows denote a significant difference (P < 0.05) between ages in nonparametric
3 Year Old 6 Year Old 18 Year Old
Aboveground Biomass 2.2 (0.8- 7.9)a 1.1 (0.5-1.7)a 1.4 (0.8-1.6)a
Living Coarse Roots 0.3 (0.0- 1.0)' 1.7 (0.7-4.2)'b 4.6 (2.2-7.6)b
Dead Coarse Roots 6.6 (0.4-12.0)' 1.0 (0.2-2.1)' 0.6 (0.2-3.4)b
and 69% of total living coarse roots under skid trails in the '91, '88, and '76 logging coupes,
Soil Disturbance Conventional versus Reduced-Impact Logging
In sites logged according to the RIL harvesting guidelines, proportionally less area of soil
was disturbed than in sites logged by conventional methods. An inefficient layout of skid trails,
typical of unplanned, unsupervised operations, was apparent in conventional logging areas. Skid
trails in conventional logging units were often cross-linked and located within 10 m of each other,
whereas, in general, skid trails in RIL units were widely spaced and evenly dispersed across the
logged area. A time-motion study of skidding practices in the two treatment areas documented the
higher level of efficiency of yarding in RIL operations as compared with conventional logging, 1.98
US$ m3 and 4.51 US$ m3, respectively (J. Tay unpubl. data).
The harvesting guidelines adopted by the RIL project include specifications about road
location and construction, but the road in the RIL pilot project area was constructed before adoption
of the RIL guidelines by the concessionaire, compromising flexibility in locating skid trails. The
road was positioned low on slopes; this location was suboptimal and often forced downhill skidding.
There was no difference in road density for the two methods (Table 2-1), but the area covered by
sidecast soils, associated with the road, was less in RIL areas than in conventional logging areas,
even though the roads in RIL units were used for processing logs. This difference reflects the
attitudes of the operators working in the two areas; bulldozer operators in RIL areas worked carefully
and with the awareness that the project's goal was to reduce damage. The operators in conventional
logging areas were not similarly motivated.
The extent of soil damage associated with conventional logging in this study (mean = 17%;
range = 14-20%) was at the low end of the range of published values for unsupervised logging in
Malaysia (e.g., 43%, Fox 1968a; 17%, Borhan et al. 1987; 16%, Jusoff& Nik 1992; 30%,
Nussbaum et al. in press) and was similar to values for operations in Suriname (14.5% and 16.0%,
Hendrison 1990) and Indonesia (16%, Cannon et al. 1994). Skid trail area in RIL units was similar
to values obtained with planned operations in Suriname (5-7%, Hendrison 1990) and Australia (5%,
Crome et al. 1992). The large variation in values reported for dipterocarp forests in Sabah may be
due to differences in sampling methods, biases towards roadside locations, or differences in local
topographical conditions. I expect the results from this study are relatively free from sampling biases
because soil disturbance associated with logging was measured in 100% of the area of the eight
In general, damage to the residual stand is positively correlated with timber volume extracted
(Nicholson 1979). In this study, soil damage was positively associated with harvested volumes in
conventional logging areas but not in reduced-impact logging areas (Fig. 2-2). If main skid trails are
located to optimize efficiency of log extraction, bulldozers are restricted to main skid trails, and logs
are winched from the forest to the skid trail, one might expect that, after the whole area was rendered
accessible by the main skid trails, the proportion of area disturbed by logging would remain fairly
constant, regardless of the number of trees removed.
Unfortunately few studies of soil damage associated with logging in tropical forests include
information on the volume of timber extracted or express damage in terms of volume extracted.
Failure to include information about logging intensity makes it difficult to compare sites. One
exception is a study conducted in the Brazilian Amazon; Verissimo et al. (1992) found that 218 m2
of ground surface was scraped by bulldozers (roads and skid trails) for each harvested tree.
Comparable figures for this study are much lower (mean cNv = 140 m2 tree' and
mean IL = 94 m2 tree-'), perhaps reflecting differences in the size of harvested trees and number
extracted per hectare.
Skid trails in RIL units were, in general, less severely damaged than those in conventional
logging units, the proportion of skid trails with subsoil disturbance was less than half that in the
conventional logging areas. In part, this difference may be due to the fact that bulldozers did not
traverse slopes >35 degrees in RIL areas, so may have been less likely to require the use of the blade.
Blading is often considered essential on slopes >24 degrees to increase stability and control (Stuart &
Canrr 1991). But blading and side-cutting were not restricted to steep areas in the conventional
logging units; z87% of the skid trails had exposed and disturbed subsoils. The skid trails receiving
subsoil disturbance in RIL units (=38%) were typically main skid trails that received heavy traffic.
In conventional logging units, branch and main trails were not distinguishable in terms of soil
damage class. The restriction on wet-weather skidding in RIL areas also probably contributed to the
observed differences; in RIL units, all skid trails showing subsoil disturbance had been logged during
the wetter season.
Plant Regeneration on Skid Trails
Fewer sapling and pole-sized trees were found on abandoned skid trail tracks than in
adjacent, residual forest in '91, '88, and '76 logging coupes. This result suggests that, even 18 years
after logging, tree regeneration on skid trails is less than that in residual forest. Tree regeneration on
the edges of skid trails appears similar to that in adjacent forest in terms of sapling densities and
species richness. However, species composition is different in the two habitats, with pioneer tree
species being more common on skid trail edges than in residual forest.
Sapling densities in the '91 and '76 coupes are very similar, suggesting that conditions for
tree regeneration on older skid trails are no better than those on younger skid trails. The similarity in
biomass on skid trails from the three logging coupes also suggests little change. Immediately after
logging, 98% of the skid trail area in conventional logging units was bare of vegetation. The
quantity of aboveground biomass on the 3-, 6-, and 18-year-old skid trails was only slightly higher
than that recorded on one-year-old skid trails (0.3 Mg biomass ha'; SD = 0.38; Chapter 3). Living
coarse root biomass appears to be increasing with time since logging, as one might expect, but at 18
years after logging, coarse root biomass was 12% of the pre-logging value observed elsewhere in the
forest reserve (Chapter 3).
I interpret the results from this study with caution and recognize that pre-logging conditions
in the three logging coupes studied may have differed. Nevertheless, I am fairly confident that all
three areas were heavily logged (Pacific Hardwoods, unpubl. data) and that the coupes have not been
re-entered by heavy equipment after the initial selective cut. I chose three different-aged logging
coupes in order to look at the potential for recovery on skid trails over time, but comparisons among
habitats within a coupe involve fewer assumptions than do comparisons across the three coupes.
I expected that if soil disturbance favors pioneer trees over more persistent species, then the
density of pioneers on skid trails would be higher than in adjacent forest. This was supported by the
data from the younger areas, the '91 and '88 coupes, where pioneer sapling densities on skid trail
edges were higher than densities in adjacent forest. Densities on skid trail surfaces were not different
from densities in forest plots. Perhaps pioneer tree densities in forest plots were high relative to
undisturbed forest because of the inclusion of felling gaps, which may provide opportunities for
pioneer tree establishment, in some of the plots. Also, few pioneer saplings would be expected to
survive under the closed canopy observed in the '88 and '76 coupes.
Several studies in neotropical rain forest recorded vigorous tree seedling establishment along
the edges of skid trails and roads (e.g., Jonkers 1987; Verissimo et al. 1992; Guariguata & Dupuy
1995) two to three years after logging. It does not necessarily follow, however, that high densities of
saplings on skid trails will eventually develop into a stand of trees; unfavorable soil properties (e.g.,
compaction and low nutrient status) may continue to limit tree growth on skid trails for many years.
I attribute lower densities of saplings on skid trails as compared with adjacent forest to
unfavorable establishment conditions in those habitats. An alternative explanation for lower sapling
densities on skid trails is that crowns and root systems of residual trees occupy these areas and the
competition for resources on skid trails is greater than that in adjacent forest. Sapling densities in
these sites may have been lower than in adjacent forest prior to skid trail construction. A
manipulative study of tree establishment in these habitats that controlled for competition with
neighboring trees would help to elucidate the mechanisms driving differences in sapling densities.
Implementation of harvesting guidelines in a ground-based yarding system substantially
reduced the extent and degree of soil disturbance associated with logging. About 84% of the skid
trail area in conventional logging areas had subsoil disturbance. Distribution patterns in biomass,
species richness and sapling density across habitats in logged forest suggest that even 18 years after
conventional logging, areas with soil disturbance are less productive than areas without. In reduced-
impact logging areas, about 62% of skid trail area retained topsoil. Retention of organic matter in
these compacted areas may result in improved plant regeneration (Woodward in press), but for many
soils, most compaction associated with skidding happens with the first few passes of the bulldozer
(Dias & Nortcliff 1985; Koger et al. 1985). If damage to soil structure is to be minimized, reducing
the area traversed by bulldozers will be more important than reducing the traffic on any particular
RETAINING FOREST BIOMASS BY REDUCING LOGGING DAMAGE
A pilot carbon offset project, in which a power company provided funds to a timber
concessionaire to implement guidelines aimed at reducing logging damage, was initiated in Malaysia
in 1992; in doing so, the utility gained potential credit towards future emissions reduction
requirements. To quantify the carbon retained due to this effort, dipterocarp forests logged according
to reduced-impact logging guidelines were compared to forests logged by conventional methods, in
terms of above- and below-ground biomass both before and after logging. This comparison is the
focus of this chapter.
I have three objectives for this chapter. The first objective is to describe forest biomass
stores both before and after logging. The second is to compare logging damage in forest logged by
conventional methods and in forest logged according to reduced-impact logging harvesting
guidelines. The third objective is to quantify the carbon retained in biomass due to implementation
of the harvesting guidelines.
The experimental area in Ulu Segama supports primary dipterocarp forest, spectacular both
for its stature and its high density of big trees. Canopy height averages z45 m but emergent trees
reach heights of 70 m. The terrain consists of series of steep ridges; over 75% of the area occurs on
slopes exceeding 20 and generally <200-300 m long (Pinard, unpubl. data). Soils are varied but
primarily are Ultisols derived from Tertiary sediments (Ohta & Effendi 1992). The climate is only
slightly seasonal with a dry period centered on April. Mean annual rainfall is approximately 2700
mm and mean daily temperature is 26.7C (Danum Valley Field Centre Records, 1986-1993).
Forest biomass and stand structure before logging were measured to allow comparison of the
effects of logging treatment on carbon stores. Prior to logging, four logging units (30-50 ha each)
were randomly selected from the experimental area to be logged according to the reduced-impact
logging guidelines (hereafter RIL units); four additional units were randomly selected from an
adjacent area destined to be logged conventionally (Fig. 3-1). Units logged conventionally or by the
RIL guidelines were paired according to topography and logging schedule to reduce variability of
logging impacts on the residual stand due to differences in soil moisture content and slope. The
conventional logging units were harvested by crews not involved in the reduced-impact logging
project. A crew that was trained with funds from the power company and was experienced with
directional felling and proper log extraction techniques harvested the RIL units according to the
reduced-impact logging guidelines.
Within each unit 20 to 35 1600 m 2 plots (40 x 40 m for 6 units or 20 x 80 m for two units,
approximately 10% of each logging unit area) were located according to a stratified random design
(Fig. 3-1), avoiding areas within 20 m of permanent streams, within 10 m of a logging unit boundary
or a main road, steep rocky areas (slopes >45), and landslides. In the eight logging units, a total of
216 plots was established. No plots were established at 49 points dismissed due to exclusions listed
V ) q7
o0 0 U-0
-* CT3 0
cU 0 UL
2~ io .
so Xl 2
r xu o
-B- B- -B-
B- -B-B- -
-B- -B-B -
Each tree >60 cm diameter was tagged and its diameter measured at 1.3 m or above
buttresses (hereafter, dbh). Nested subplots were used for smaller trees and lianas (Fig. 3-1). All
commercial trees tagged in the plots were identified to species or timber species group. Stem and
bark damage were described, and any other tree characteristics that might be mistaken for logging
damage were noted. Lianas were tagged and measured only in the four units to be logged
conventionally because most of the lianas were cut prior to plot establishment in the units to be
logged according to the reduced-impact logging guidelines.
Aboveground tree biomass was estimated allometrically using tree inventory data and stem
volume diameter height relations calculated for 15 local species groups in the Ulu Segama Forest
Reserve (Forestal International Limited 1973; Appendix B) and a Biomass Expansion Factor (BEF)
developed for good hill dipterocarp forest in West Malaysia (Brown et al. 1989). The BEF for good
hill forest was selected over the factor developed for other Malaysian dipterocarp forest types
because the basal area for good hill forest (28.5 m2 ha"1 for trees >15 cm dbh) most closely matched
that of the study site. Wood densities were available for 120 of the 124 species or species groups
recorded in the plots (Burgess 1966). To convert wood densities determined at 12% moisture
content (air-dry weight) to density at dry weight, I applied a regression developed by Reyes et al.
(1992). For non-dipterocarp species whose wood density was not known, I used the arithmetic mean
of the known species that were not dipterocarps (0.503 g cm3, N= 48 species). For dipterocarp
species whose wood density was not known, I used the mean value of the known species within the
genus (or section of the genus, when applicable). Biomass of lianas >2 cm dbh was estimated from
basal area using a regression equation developed for Venezuelan liana species (Putz 1983).
To supplement the available stem volume equations, which I judged were inappropriate for
trees <10 cm dbh, I harvested 40 randomly selected trees 1-10 cm dbh, representing a mixture of
species and determined their aboveground biomass. Sampling was conducted in primary forest
within 1 km of the study sites. A regression equation was calculated with dbh as the independent
variable and tree biomass as the dependent variable. To determine total small tree biomass, I applied
the dbh-biomass equation to the trees (1-10 cm dbh) in the permanent plots.
Shrub, herb, palm, and herbaceous vine biomass was measured in 3 RIL and 3 conventional
logging units using 1 m2 circular clip plots (n = 15 per unit, N = 45 per treatment) randomly located
in a stratified random fashion using three topographical positions as strata. Each plot was considered
a sample and logging unit divisions were assumed to be inconsequential to the estimate as the
variation within a unit was much higher than that between units. In the clip plots, all above-ground
plant biomass (<1 cm diameter at base) was cut at the soil surface and weighed, and then a
subsample was oven-dried at 70 C to constant mass. For self-supporting species, only plants rooted
inside plots were included. For vines, all stems and leaves occurring over the clip plots were
collected regardless of the rooting site. Palms (primarily stemless rattans) that occurred in the plots
were also clipped and collected.
A conversion factor of 50% is frequently used to estimate carbon content of plant tissues
(e.g., Harmon et al. 1990; Hoen & Solberg 1994). To determine whether woody tissue in my site
was similarly about 50% carbon, I tested a small number of wood samples randomly collected from
fresh logging debris for carbon content. Twenty samples, approximately 45 cm3 each, were collected
from log and branch debris. The samples were split into small pieces, oven-dried, ground and sieved.
Carbon content was determined using a Carlo-Erba NA 1500 Carbon-Nitrogen Elemental Analyzer
(Isotope Ratio Mass Spectrometer, Department of Soil Science, University of Florida, Gainesville,
FL). Carbon content averaged 49.2% (SD = 1.1, N= 20; statistically different from 50%, t = 3.55,
df= 19, P < 0.005). I assume all plant tissues to be 49.2% carbon by dry weight though I recognize
that certain tissues often have carbon contents that are above or below this percentage (e.g., seeds
and fine roots, respectively; Golley 1969; Williams 1986).
Pre-logging root biomass was sampled in the eight logging units using a stratified random
design, traversing terrain typical for each unit (n = 10 pits per unit, total N = 40 pits per treatment;
logging unit divisions were disregarded in the analyses). Coarse roots (>5 mm diameter) were
sampled in 50 x 50 cm monoliths of soil extracted to 50 cm with a sledge-driven flat blade. Roots
were separated from the soil in the field and washed, live and dead roots were separated, and sorted
into four diameter classes in the lab (5-15, 15-50, 50-150, and > 150 mm diameter); and live roots
were weighed and subsampled for dry weight determination. Dead roots were weighed and
subsampled in a subset of the samples (N = 56). I did not sample deeper than 50 cm in the soil
profile for coarse roots and consequently underestimate carbon stored in coarse roots. Fine root (<5
mm diameter) mass was estimated using 5 cm diameter cores taken to 10 cm depth; two cores were
taken at each sampling site, combined, soaked in water, and agitated (n = 10 sample sites per unit,
total N= 40 per treatment; logging unit divisions were disregarded in analyses). Roots were then
separated from soil, oven dried, and weighed. Due to difficulties in confidently differentiating live
and dead roots, only total fine root mass values are reported. A few of the early samples were not
included as only live roots were dried and weighed.
To determine coarse root biomass directly beneath trees where core sampling was
impractical (hereafter, butt roots), 14 partially uprooted trees (20-130 cm dbh) along roadsides and
skid trails within the 1993 logging area were opportunistically sampled to establish the relationship
between butt root mass and dbh. Coarse roots >10 mm diameter within 1 m of the bole of the tree
were separated from the soil, cut into pieces <50 kg, washed, weighed, and subsampled for dry
weight determination. Butt root mass was log-transformed and used as a dependent variable in a
regression equation with dbh as the independent variable. To determine total root mass, I applied the
dbh butt root mass equation to trees in the permanent plots and calculated the mean butt root
biomass per ha across the eight logging units. Coarse and fine root biomass are expressed on a per
ha basis, the calculation of which excluded areas occupied by butt roots.
Damage Assessment and Necromass Production
Permanent plots were recensused for tree damage and survival 5-30 days after logging and
again 8-12 months later. All trees and vines were relocated and assessed for damage. Although
numerous damage classes were used in the field, here I compress them into the following: destroyed
(uprooted and crushed), snapped-off below crown, and other damage (includes crown, stem, bark, or
root damage of varying severity).
From the damage assessment data I estimated (by logging unit) the following parameters:
timber volume extracted; necromass produced from the branches, leaves, stumps and butt roots of
harvested trees; necromass produced from trees destroyed during harvesting; and, necromass
produced from damaged trees that died within the first 8-12 months after logging. Aboveground and
butt root biomass were included in these calculations.
The biomass of shrubs and herbs in logged forest was measured using 1 m2 clip plots (n =
15) randomly located along transects dispersed through each of seven logging units (3 RIL, 4
conventional logging) 1 year after logging; the sampling protocol was similar to that used for pre-
logging measurements. To determine the biomass of colonizers and resprouted plants in areas with
soil disturbance (e.g., skid trails, log landings and roads) 1 year after logging, I sampled skid trails in
the same seven logging units using 1 m2 clip plots (n = 10 per unit, N = 70) located randomly along
skid trails. Although skid trails and other areas with soil disturbance covered a relatively small
percentage of the total area (conventional logging areas, mean = 11.9%, SD = 2.7, N = 4; RIL areas,
mean = 3.5%, SD = 1.6, N = 4; Chapter 2), biomass in these areas was expected to be more variable
than that in the rest of the forest, so sampling intensity was higher. As with pre-logging shrub and
herb biomass measurements, logging unit divisions were disregarded in the analyses. Pre- and post-
logging measurements of shrub and herb biomass are not paired, as sampling points were located
randomly in logging units.
Coarse root mass (both living and dead) was measured 3 months after logging in four
logging units (2 RIL, 2 CNV) following the protocol described for the pre-logging measurements.
Coarse-root pits were located randomly on skid trails (10 pits per logging unit, N= 40) and in other
areas of disturbed forest (10 pits per unit, N= 40). The difference between mean coarse root
biomass before and 3 months after logging (corrected for proportion area in skid trails and disturbed
forest) was considered to have entered the necromass pool (if biomassfore > biomasser at a = 0.05).
As with understory biomass, logging unit divisions were disregarded thus pre- and post-logging
samples were not paired for statistical comparisons. I did not harvest fine roots after logging and
assume that fine root mass 1 year after logging is similar to mass before logging.
For all of statistical comparisons I use a significance level of 5% but report test statistics
when P values are between 0.1 and 0.05. T-tests are two-tailed using pooled variances unless stated
otherwise. For t-tests using separate variances, degrees of freedom were calculated following
Brownlee (1965 in Wilkinson 1990). For treatment comparisons based on the aboveground biomass
plots, rather than using a global analysis of variance, I use separate t-tests for each diameter class.
Nested subplot size was selected for sampling convenience, not to allow equal variances among the
diameter classes. The term biomass always refers to living plant material.
Stand structure in RIL units was similar to that in conventional logging units prior to logging
(Fig. 3-2). The mean number of harvestable trees per hectare (commercial species with dbh >60 cm
dbh) ranged from 14.4 to 26.9 in the eight logged units (mean = 19.0, SD = 3.88); densities in RIL
units did not differ from those in conventional logging units (t = 0.244, df= 6, P > 0.8). Total basal
area of trees 2 10 cm dbh in the eight logging units ranged from 24.9 to 33.1 with an overall mean of
27.5 m2 ha' (SD = 2.86). Tree densities for the two treatment areas did not differ for any diameter
class (t-tests, a = 0.05). In the conventional logging units, density of lianas >2 cm dbh averaged 586
stems per ha (SD = 211, N = 4), about 86% of which were <5 cm dbh.
Of the 6298 trees 10 cm dbh in the plots, 59.3% (representing 83.3% of the total basal
area) were identified to species or species group. Dipterocarpaceae was well-represented in the study
area, comprising 29.6% of the tagged trees 10 cm dbh and 67.9% of the basal area (Table 3-1).
The forest was dominated by two dipterocarp species, Parashorea tomentella and Shorea
johorensis, which together made up 20% of the stems 10 cm dbh and 47.8% of the basal area. The
10 most abundant species or species groups were represented similarly in the RIL and conventional
logging units. The 15 stem volume equations used in biomass calculations for trees (2 10 cm dbh)
along with the species allocated to each are presented in Appendix B.
Total biomass in the two treatment areas averaged about 400 Mg ha' with approximately
17% occurring below ground (Table 3-2). For each diameter class, aboveground biomass per ha was
equivalent for the two treatments (Table 3-2). Approximately 59% of the initial aboveground
biomass was in trees >60 cm dbh. Small trees (<10 cm dbh) contributed approximately 4% of total
aboveground biomass (Table 3-2; Fig. 3-3). Understory plant biomass contributed approximately
1% of total aboveground biomass and was similar in RIL and conventional logging units (Table 3-2).
3000 300 T 30
2500 250 25
S2000 200 20
1500 150 15
vaines. ) 00)
1000 100 10
500 50 5
1-5 5-10 10-20 20-40 40-60 >60
Diameter class (cm dbh)
Figure 3-2. Stem density (mean +/- SE) for logging units prior to logging (black bars units logged
according to the RIL guidelines, open bars units logged conventionally; Note different y-axes). Pre-
logging stem densities by diameter class did not differ for the two treatments (t-tests with pooled
variances, a = 0.05).
The 10 most common species or species groups of trees > 1 cm dbh based on
density and basal area (BA, m2 ha') before logging in the eight logging units (all
plots, N = 170). Merchantable species or species groups are marked with an
Species or % Stems Species or % BA
Species Group Species Group
Parashorea tomentella* 12.2 Parashorea tomentella* 24.1
Shorea johorensis* 7.8 Shorea johorensis* 23.8
Eugenia spp. 5.8 Annonaceae 5.4
Lauraceae 5.5 Shorea parvifolia* 3.5
Diospyros spp. 5.3 Eugenia spp. 3.4
Annonaceae 4.0 Diospyros spp. 3.4
Shorea parvifolia* 1.4 Shorea leprosula* 2.6
Shorea leprosula* 1.4 Dryobalanops lanceolata* 2.4
Dryobalanops lanceolata* 1.4 Lauraceae 2.2
Shorea section Shorea* 1.1 Shorea section Shorea* 1.7
TABLE 3-2. Above- and below-ground biomass for the two logging treatment areas before
logging. Values are means (Mg ha-'), with SD and N noted parenthetically. For
trees, vines, and butt root mass, SD describes variation among four logging units
and does not incorporate error in biomass equations. No significant differences
were detected between treatments (t tests, P < 0.05).
Before Logging Conventional Logging Reduced-Impact Logging
Trees >60 cm dbh 190 (35, 4) 190 (53, 4)
Trees 40-60 cm dbh 53 (20, 4) 46 (6.5, 4)
Trees 20-40 cm dbh 46 (2.5, 4) 46 (6.3, 4)
Trees 10-20 cm dbh 21 (2.7, 4) 23 (2.8, 4)
Trees <10 cm dbh 13 (2.0, 4) 12 (2.0, 4)
Vine Biomass 7.6 (3.8, 4) 7.6 (3.8, 4)b
Understory Biomass 2.87 (1.50,45) 2.94 (1.67, 45)
Butt Root Mass 26.8 (6.2, 4) 24.5 (5.7, 4)
Coarse Roots (Alive)c 35.9 (33.0, 40) 39.4 (38.7, 40)
Coarse Roots (Dead)' 1.6 (2.6, 30) 1.8 (3.5, 26)
Fine Root Mass 2.57 (1.30, 31) 2.74(1.43, 18)
Total Mean (SD) Biomass Before Logging 399 (40)' 394 (59)e
a Each logging unit considered a replicate and subsampled with 10-27 plots.
b Assumed to be equivalent to conventional logging units; no statistical comparison made between
c Log-transformed data used for statistical comparison.
d Not included as biomass.
SVariance for sum of means calculated using a weighted estimate: E1k ((k(wi)si2)/(n,)), where k = #
of components, wi = mean of component/sum of means.
2 4 6 8 10
Figure 3-3. Relationship between dbh and total biomass for small trees (1-10 cm dbh), all species
combined. The line represents the following regression equation: Loge (Dry Weight in kg) =
0.539 DBH 1.25 (R2 = 0.93, SEslop =0.021, SEntercept = 0.119, P < 0.001, N= 40).
Approximately 2% of the aboveground biomass in conventional logging areas was in vines; small
vines (2-5 cm dbh) contributed about 56% of the vine biomass.
Total belowground biomass averaged approximately 66 Mg ha' in the two treatment areas;
about 40% was in butt roots, and about 56% was in coarse roots (Table 3-2). Estimated biomass in
butt roots for trees 220 cm dbh increased with diameter according to the following relationship:
Loga(Dry Weight) = 0.014 DBH + 1.51 (Fig. 3-4). Application of the above regression equation
to trees 220 cm dbh in the plots used for aboveground biomass yields an overall butt root biomass
estimate of 25.6 Mg ha' (SD = 5.84, N= 8). Coarse root biomass (>5 mm diameter) was extremely
variable (overall C.V. = 95%; Table 3-2), owing to the presence of widely dispersed, large roots of
the canopy trees that may extend more than 35 m away from the tree's stem (see Baillie & Mamit
1983 for discussion). Coarse root mass in the two treatment areas were not different prior to logging
(Table 3-2). Mean fine root mass (<5 mm) in the upper 10 cm of soil also did not different in the
two treatment areas (Table 3-2).
Details of Logging
Logging started in July 1993 and ended in March 1994 (Table 3-3). The time required to
log a unit varied from 1 to 24 weeks. Logging in two of the RIL units was prolonged due principally
to wet weather. No dry season occurred during the study period (unpubl. data), and environmental
conditions during logging were fairly similar for all units.
A portion of each of the RIL logging units (mean = 44%, SD = 18.9, N= 4, range 12 to
63%) was deemed unloggable by the rangers due to steep terrain, unstable substrates, lack of
commercial trees, or inaccessibility. Because the principal comparison of this study involves
impacts of two harvesting methods, I eliminated these unlogged areas (and any influenced plots)
from the analysis. Difficulties arose when trying to identify these areas a posteriori but I used the
following criterion: if neither a skid trail nor a stump of a harvested tree was inside a plot or within
0.0 -. .-----.- .-.-. .-. I I I -
20 40 60 80 100 120 140
Figure 3-4. Relationship between dbh and butt root biomass for trees >=20 cm dbh. The line
represents the following regression equation: Log10 (Dry Weight in kg) = 0.014 DBH + 1.51
(R = 0.88, SEope = 0.001, SEtercept = 0.10, P < 0.001, N= 14).
Dates of logging and volumes of timber removed from reduced-impact logging units
(RIL) and conventional logging units (CNV) in Ulu Segama Forest Reserve.
Volume extracted is based on data from 1600 m2 plots distributed among the four
units, including total area and only loggable area.
Unit No. Dates Logged Volume Extracted Volume Extracted
Per Total Area Per Loggable Area
32 RIL 17 Jul'93 6Aug'93 49.7 99.3
41 CNV 17 Jul'93 -10 Sep'93 129.6 134.1
36 RIL 7 Aug'93 16 Aug'93 18.5 50.4
38 CNV 7 Aug'93 -21 Sep'93 175.9 175.9
30 RIL 10 Oct '93 4 Apr'94 97.6 178.3
23 CNV 10 Oct '93 -11 Nov'93 129.9 129.9
35 RIL 24 Nov'93 -21 Apr'94 67.7 85.3
39 CNV 24 Nov'93 -21 Dec'93 167.8 167.8
Mean RM = 58.4 Mean R = 103.3
SD L = 33.1 SD pL = 54.1
Mean CNv = 150.8 Mean cNv = 151.9
SD = 24.5 SD = 23.3
30 m of any plot boundary, the plot was considered to be within an unloggable area. By this
definition, 48 of the 114 plots in the RIL units were eliminated; none of the 104 plots in the
conventional logging units were eliminated.
Mean volume of timber extracted per total unit area ranged from 19 to 176 m3 ha-1 (Table 3-
3). If only loggable areas are included in the calculations, mean volume extracted was 152 (SD = 23)
in conventional logging and and 103 (SD = 54) in RIL areas. The two treatments did not statistically
differ in terms of volume removed from loggable areas (Table 3-3) or associated biomass converted
into logging debris (Table 3-4), but small sample sizes and large variances limit the power of this
analysis. The Pacific Hardwoods mill that converts the timber extracted from Ulu Segama into
lumber, veneer, and blockboard does so with about 50% efficiency (Eng W. H., pers. comm.).
Therefore, in addition to the biomass converted to necromass in the forest, I included 50% of the
biomass of extracted timber in the necromass pool. (Note: most scrap at the mill is burned to
produce electricity; Table 3-4.)
Damage Assessment and Necromass Production
For all dbh classes, proportionally more trees were damaged (all types of damage combined)
from logging in conventional logging areas than in RIL areas (one-tailed t- tests, arcsine transformed
data, ix = 0.05; Fig. 3-5). Proportion of residual trees damaged differed by dbh class (ANOVA on
arcsine transformed data, F= 3.45, df= 5,36, P < 0.02). Generally, proportionally more small trees
were damaged than large trees (Fig. 3-5). There was no interaction in the proportion of trees
damaged between logging method and dbh class (F= 1.4, df= 5,36, P = 0.25).
The percentage of trees destroyed during logging was higher in units logged conventionally
than in units logged according to the RIL guidelines for all dbh classes (Fig. 3-5, one-tailed t-tests,
arcsine transformed data, a = 0.05); the mean values by dbh class ranged from 17 to 57% in
Biomass converted into necromass. Values are means (Mg ha-'), with SD noted
parenthetically. SD describes variation among four logging units and does not
incorporate error in biomass equations.
50% of Extracted Timber'
Branches, Stumps, and Butt Roots of Extracted Treesb
Destroyed Trees (Uprooted and Crushed)
Damaged Trees Dead Within One Year After Logging
Understory Plant Deathc
Coarse Root Death (Excluding Butt Roots)c
Total Necromass Produced =
Mean (SD) Difference Between Two Logging Methods =
86 (43) Mg Necromass ha'1
" 50% of the extracted timber is assumed to be converted into wood products.
b Treatment comparison t-test with separate variances, t = 1.79, df= 3.8, P = 0.15.
c Represented as the difference between biomass before logging and biomass at 1 yr after logging.
d Variance for sum of means calculated using a weighted estimate: Elk ((k(w)sj2)/(n)), where k = #
of components, wi = mean of component/sum of means.
Proportion of trees
Uprooted and crushed
I -I Snapped-off below crown
Other types of damage
Figure 3-5. Mean proportion of trees completely destroyed, snapped-off, or otherwise damaged
(stem, bark, crown, or root) during logging in 4 units of each treatment. The proportion of trees
snapped-off or otherwise damaged did not differ for the two treatments (t-tests, arcsine transformed
data, alpha = 0.05). (Destroyed trees do not include harvested trees.)
I i I
conventional logging areas in contrast to 2 to 22% in RIL areas (Fig. 3-5). The biomass in these
destroyed trees was assumed to enter the necromass pool (Table 3-4).
The proportion of trees snapped-off (below crown) ranged from 3.5 to 10% across the dbh
classes (Fig. 3-5) and was higher in conventional logging than RIL areas for only one of the six
diameter classes, trees 10-20 cm dbh (t = 1.77, df= 6, 0.01< P < 0.05; one-tailed t-tests, arcsine
transformed data). Snapped-off trees were distinguished from other severely damaged trees because
I expected a proportion of these would resprout and would not enter the necromass pool.
The incidence of minor to moderate damage (e.g., crown or bark damage) was higher in
conventional logging units than in RIL units for three diameter classes, 40-60 cm dbh (arcsine
transformed data, t = 1.97, df= 6, P < 0.05), 10-20 cm dbh (t = 2.17, df= 6, P < 0.05), and 5-10 cm
dbh (t = 4.10, df = 6, P < 0.05); the other three diameter classes did not differ (one-tailed t-tests on
arcsine transformed data, c = 0.05; Fig. 3-5).
Sixty-seven percent of the vine stems were killed during logging in conventional logging
units, contributing an average of 4.68 Mg biomass per ha (SD = 0.18) to the necromass pool.
Mortality was evenly distributed across diameter classes (ANOVA, F= 0.49, df= 3,12 P > 0.6).
Vines in the RIL units were neither tagged nor measured prior to cutting. To estimate vine biomass
killed in the RIL areas I assume that vines cut were killed (87% of stems 2 2 cm dbh cut, F. E. Putz,
unpubl. data) and that they represented 87% of the total vine biomass (Tables 3-2, 3-4 & 3-5).
At 8-12 months after logging, many of the damaged trees were dead (Table 3-6). Overall,
18% of the trees (>5 cm dbh) snapped-off below the crown had not resprouted, so were considered
dead. In general, trees snapped off at a height > 10 m resprouted regardless of logging treatment.
The mortality rates for trees receiving other types of damage ranged by dbh class from 0 to 3% in
RIL areas and from 3 to 10% in conventional logging areas (Table 3-5). The percentage of these
u o -oa
r' o/~ Cl -' 6
0- 0C \0 0
^f f Cl -S
4) 4) 4
. E 0
0 S 2 0
Table 3-6. Percentage of trees dead at 8-12 mo after logging for each treatment (RIL =
reduced-impact logging, CNV = conventional logging). All trees in the four logging
units were pooled for each treatment to generate mortality figures; sample sizes (i.e.,
number of trees) are noted parenthetically. All trees uprooted or uprooted and
crushed were assumed dead.
Dbh Snapped-off Other Damage Undamaged
CNV RIL CNV RIL CNV RIL
14.3% (7) 28
22.0% (18) 42
21.7% (60) 12,
17.9% (84) 21.
23.8% (21) 22.
damaged trees that died during the first year after logging was higher in conventional logging areas
than in RIL areas for all diameter classes (Table 3-5). The two logging treatments were not
compared statistically because none of the damaged trees in many logging units had died. Although
many of the damaged trees were expected to die soon, only the proportions that died before the
recensus were incorporated into the necromass pool (Table 3-4).
Between the time the plots were established and the recensus (approximately 18 mo after
establishment), an average of 0.5% of the undamaged trees (>5 cm dbh) died. The mortality rates for
undamaged trees appears similar for the two treatments (Table 3-5).
Shrub and herb biomass 12 months after logging was less than before logging both on skid
trails (t- test using separate variances, t = 12.64, df= 133.5, P < 0.001; Tables 3-2 & 3-5) and in
otherwise disturbed forest (t = 8.97, df= 193, P < 0.001; Tables 3-2 & 3-5). Biomass on skid trails
was greater in RIL units than in conventional logging units (Table 3-5). Biomass in other
areas of disturbed forest did not differ for the two treatments (Table 3-5). The difference between
shrub and herb biomass before logging and at 12 months after logging was considered necromass
Three months after logging, coarse root biomass (exclusive of butt roots) on skid trails did
not differ between the logging treatments (Table 3-5) but was less than pre-logging levels (log-
transformed data, t = 15.2, df= 118, P < 0.001; Tables 3-2 & 3-5). This decline is probably due to
both root death and excavation and relocation from bulldozer activities. Dead coarse root mass on
skid trails did not differ between treatments (Table 3-5) and was similar to dead root mass before
logging in both conventional (log-transformed data, t = 1.07, df= 58, P = 0.29; Tables 3-2 & 3-5)
and RIL areas (log-transformed data, t = 1.64, df= 54, P = 0.11; Tables 3-2 & 3-5).
In disturbed forest (not skid trails) 3 months after logging, coarse root biomass did not differ
between treatments (Table 3-5) and was similar to pre-logging estimates (log-transformed data, t =
1.6, df = 118, P = 0.11; Tables 3-2 & 3-5). Dead coarse root mass in logged-over forest did not
differ from pre-logging mass (log-transformed data, t = 1.35, df 118, P > 0.18; Tables 3-2 & 3-5),
nor did treatments differ (Table 3-5). Because conventional logging units had proportionally more
area with disturbed soil or skid trails than did RIL units (approx. 12% and 3.5%, respectively;
Chapter 2), the calculated total standing stock of coarse root biomass in conventional logging units
was less than in RIL units (Table 3-5). My estimates of necromass produced from coarse root death
(biomassbefore-biomassafter) are associated with relatively large standard deviations (Table 3-4),
reflecting the large variance in the pre- and post- harvest biomass estimates.
One year after logging, forest areas logged by conventional methods and according to RIL
guidelines contained approximately 44% and 67% of their pre-logging biomass, respectively (Tables
3-2 & 3-5). The difference in necromass produced was 76 Mg ha-' (37 Mg C ha-1; Table 3-4). The
greater number of residual trees destroyed during logging in conventional logging areas was
responsible for approximately 62% of the difference between the two methods; the difference in
debris produced from trees felled accounted for approximately 25% of the difference. A large
proportion of the standard deviation associated with the estimate of the difference between the two
methods is due to variation in coarse root death.
Implementation of RIL harvesting guidelines substantially reduced logging damage. The
residual forest in the two treatment areas is dramatically different, hence each forest's potential for
both short- and long-term carbon storage also differ. In the following sections, I compare my
biomass estimates to other dipterocarp forests and briefly discuss estimation methods. I compare
levels of logging damage recorded at my sites with other selective cutting operations and discuss
ecological implications of reductions in damage for forest recovery. I also discuss the amount of
carbon retained due to implementation of the RIL guidelines, how it could be increased, and how it
relates to power plant emissions and other offset options. Finally I identify several issues relevant to
future efforts to offset carbon through reduced-impact logging and suggest topics needing further
Residual Forest Biomass
Pre-logging aboveground biomass estimates for my sites (291-400 Mg ha-'; mean = 330) are
higher than average moist forest biomass in southeast Asia (mean = 225 Mg ha', N = 204 stand
inventory data sets; Brown et al. 1991) but are comparable to estimates for unlogged forests in
Sarawak (280-405 Mg ha-'; Brown et al. 1991). Big trees (>60 cm dbh) made up about 59% of the
pre-logging biomass at my sites. Degraded forests tend to have few big trees and, consequently, have
much lower stores of biomass (see Brown et al. 1991).
I calculated tree biomass using published regression equations and conversion factors. Both
stem volume equations and biomass expansion factors (BEF) are associated with standard errors but
these errors were not incorporated into my estimates. I assume that the variance inherent in
calculated estimates apply equally to the two treatments. Stem volume equations used in this study
were generated from trees within the Ulu Segama area (Forestal International Limited 1973). The
BEF, however, was based on data taken from Peninsular Malaysia, Indonesia, Cambodia and Brazil;
I did not harvest trees to determine whether or not the selected BEF was appropriate for my site. I
also made no provision for hollow trees.
The estimates of necromass produced from logging were based on a relatively large
sampling area but did not incorporate the complete necromass pool. No effort was made to measure
necromass inputs from trees damaged but not killed (e.g., crowns of snapped off trees, or branches
from trees subjected to crown damage) making my estimate conservative. Also, trees snapped-off
below the crown which had resprouted at the 8-12 months census were considered alive, although
many of these trees will probably die within the second year post-harvest (Putz & Brokaw 1989).
Data published on belowground biomass in tropical moist forests are sparse and, generally,
based on few samples. For example, Edwards and Grubb (1977) excavated roots from 2 pits (10 x 5
m) to a depth of about 25 cm. Sim and Nykvist (1991) excavated roots from 7 pits (0.5 x 0.5 m) to
50 cm depth. My estimate (about 17% of aboveground biomass) falls close to the mean of reported
values for tropical moist forests (mean = 19%, range = 7-41%, N = 7; Ogawa et al. 1965; Hozumi et
al. 1969; Jenik 1971; Klinge & Rodrigues 1974; Edwards & Grubb 1977; Bullock 1981; Sim &
Nykvist 1991). Although I underestimate coarse root biomass by sampling only to 50 cm depth, a
more comprehensive root biomass study in dipterocarp forest on similar terrain in Sarawak found
most of the lateral coarse roots to be at 15-40 cm below the surface (Baillie & Mamit 1983). For
fine roots, my sampling of the upper 10 cm probably included 55-60% of total fmine root mass (Green
1993). Bias in my butt root measures are harder to predict. Uprooted trees along roads and skid
trails may not have complete root systems and do not represent a random sample from the
population; furthermore, I made no effort to separate live and dead sections of root.
In this study, there was no correlation between the proportion of stems fatally damaged and
timber volume extracted (R2 = 0.39, P = 0.37, N = 8; Fig. 3-6). This result is contrary to Nicholson's
(1979) finding that logging damage and volume extracted are positively correlated. Across the broad
range of volumes extracted in RIL units, fatal damage was less than 20% of the stand, lending
support to the conclusion that treatment differences in logging damage were due to logging technique
not harvesting intensity.
Relative to other selectively logged tropical forests, the amount of timber removed from my
study site was high, as was the level of logging damage. First cuts in Amazonian moist forest
I I I I I I
80 100 120 140 160 180
Timber volume extracted (m ha )
Figure 3-6. Mean proportion of stems (1-60 cm dbh) fatally damaged plotted against mean timber
volume extracted (inm3 ha- ; open circles conventional logging units, solid circles reduced-impact
generally take <50 m3 ha-' (Uhl & Vieira 1989; Thiollay 1992; Verissimo et al. 1992); in African
forests generally <30 m3 ha' of timber is harvested (Nwoboshi 1987; Ola-Adams 1987; Kio &
Ekwebelam 1987; Wilkie et al. 1992; White 1994). Even though the lack of standard methodologies
precludes direct comparisons of results, for four studies where logging damage was reported, damage
to residual trees >10 cm dbh averaged 11% (Gabon White 1994), 18% (Nigeria Ola-Adams
1987), 26% (Brazil Uhl & Vieira 1989) and 43% (Brazil Verissimo et al. 1992). The damage
recorded in my conventional logging areas (approximately 66%), though higher than the figures from
Amazonia and Nigeria, is similar to figures reported for other sites in Sabah (Fox 1968; Chai &
Udarbe 1977), Sarawak (Nicholson 1979; Marn & Jonkers 1981), and West Kalimantan, Indonesia
(Cannon et al. 1994).
Implementation of RIL guidelines in the study area was associated with a reduction in
damage to the residual stand, both in extent and severity. In reduced-impact logging areas, z=27% of
trees >10 cm dbh were damaged and = 19% were dead within the first year after logging, compared
with r=54% damaged and =46% dead in conventional logging areas. Efforts to control damage in
tropical moist forest in Suriname (Hendrison 1990) and Indonesia (J. G. Bertault & P. Sist pers.
comm.) also reduced damage by about half as compared with uncontrolled or conventional logging.
The slopes in my sites, on average, exceeded those recommended for ground-based skidding.
Switching to an aerial yarding system (e.g., skyline cable yarding), as is generally recommended for
slopes >25 degrees (Dykstra 1994), could further reduce damage, as might further training of fellers
and bulldozer operators.
RIL areas had about 25% fewer severely damaged residual trees (all dbh classes) than
conventional logging areas. Often severe damage (e.g., uprooted, crushed, or snapped-off) is
associated with skidding operations and felling trees laden with lianas (Fox 1968; Appanah & Putz
1984). Vine-cutting, planning skid trail locations, and controlling skidding operations may have
been instrumental in reducing severe damage in RIL areas. Reductions in less severe damage (i.e.,
crown and bark damage) in RIL areas may have been related to directional felling. Directing trees
onto skid trails or into gaps created by previously felled trees further reduced overall gap size and
felling damage (Hendrison 1990).
Implications for Forest Recovery and Carbon Storage
RIL areas contained nearly 100 Mg more biomass per ha than conventional logging areas 1
year after logging. If both forests were ultimately to recover pre-logging biomass stores, then
regardless of conditions immediately following logging, the net difference in stored biomass, at this
ending point, would be zero. Given that these are production forests, repeated cutting cycles or
conversion to plantations are their probable fates; they are unlikely to be abandoned for the time
needed to fully recover biomass. The timescale relevant to this discussion, therefore, may be through
the next cutting cycle (generally stated as 60 years but it undoubtedly will be shorter). During this
period, differences in growth and mortality rates and other responses to logging could increase or
decrease the difference between the two treatments in biomass stores. I expect biomass to continue
to decline in both areas for 2-6 years after logging. Following stabilization of mortality rates, I
expect biomass accumulation rates to be greater in RIL areas than in conventional logging areas.
The rationale behind my predictions is outlined below.
Mortality rates in logged forest are often relatively high for several years after logging
relative to pre-harvest levels (Wan Razali 1989). Elevated mortality rates may be due to any or all of
the following: a) damage incurred during logging; b) increased exposure and edge effects (e.g.,
Kapos 1989; Young & Hubbell 1991); c) increased incidence of mechanical damage from vines
(Putz 1991) and falling debris (Wan Razali 1989); and d) competition with fast growing trees and
vines (Fox & Chai 1982). Conditions in conventional logging areas (i.e. proportionally more
damaged trees and greater degree of crown exposure) are expected to be associated with higher
mortality rates (Korsgaard 1992). The difference in mortality during the first year of post harvest
observations supports this conjecture.
Growth rates in logged forest have been found to be correlated with crown exposure
(Korsgaard 1992; Daalen 1993) and, in general, increased growth rates are frequently observed in
residual trees following selective logging (Jonkers 1987; Wan Razali 1989), thinning operations
(Fox & Chai 1982; Korsgaard 1992), or natural gap formation (Brown & Whitmore 1992). Though
fewer in number, the undamaged residual trees in conventional logging areas may show larger growth
increments after logging than trees in RIL areas because of the more open canopy conditions after
conventional logging. Overall biomass accumulation, however, is expected to be greater in RIL area
than in conventional logging areas because of several characteristics of logged dipterocarp forest,
First, large canopy openings can lead to extensive vine and pioneer tree invasions (e.g., Chai
& Udarbe 1977; Cannon et al. 1994). Residual trees infested with vines or overtopped by pioneer
trees may experience reduced growth rates (Lowe & Walker 1977; Putz et al. 1984). Vine invasions
in RIL areas are expected to be less common than in conventional logging areas due to vine cutting
before logging and more closed canopy conditions after logging (Appanah & Putz 1984). Pioneer
trees may be more likely to colonize conventional logging areas because of more extensive canopy
openings and soil disturbance (Chai & Udarbe 1977). Pioneer trees, because of their low wood
densities and short life spans, may not accumulate as much biomass per unit area as similar-sized
persistent forest species (Jordan & Famrnworth 1980). Second, RIL areas contain more undamaged
trees and more trees in the larger dbh classes than conventional logging areas, so the residual trees in
RIL sites will probably have larger volume increments than residual trees in conventional sites.
Third, sites with scraped and compacted soils accumulate biomass more slowly than sites free of
heavy soil disturbance (e.g., Maycock & Congdon 1992), and proportionally more soil was severely
damaged in conventional logging areas. For the skid trails that were opened in RIL areas, higher
biomass 1 year after logging relative to skid trails in conventional logging areas may reflect less
severe soil disturbance due to controlled logging (e.g., restrictions on soil scraping and wet weather
logging). The effect of a larger input of nutrients from logging debris in conventional areas as
compared with RIL areas is difficult to predict. The input may stimulate tree growth but could lead
also to nutrient immobilization by microbes, decreasing nutrient availability for trees (Lodge el al.
To summarize, for some time after logging, I expect carbon stored in both RIL and
conventionally logged forests to decline from levels immediately following logging because of high
mortality rates and decay of logging debris. If carbon accumulation rates are higher in RIL areas due
to low mortality rates and small quantities of decaying logging debris, they will become net sinks for
carbon in fewer years after logging than the conventional logging areas.
Carbon Offsets Through Reduced-Impact Logging
In this pilot project I demonstrated that implementation of RIL guidelines in dipterocarp
forests that would otherwise be logged in an uncontrolled and destructive manner could result in the
short-term retention of, on average, about 42 Mg C ha-' at a cost of approximately U.S. $300 ha71 (J.
Tay, pers. comm.). If the carbon "savings" was considered through the next rotation (e.g., 40-60 yr),
the difference in carbon stored in RIL areas compared with conventional logging areas is expected to
be greater than 42 Mg ha1'. How policy makers will translate this effort into carbon credits is
uncertain (Dixon et al. 1993; USDOE 1994). Without doubt, however, the time profile of emission
reductions or carbon sequestration will be important for determining the consequences of the action
for climate change (e.g.. Price & Willis 1993).
Forestry-based carbon offset programs, like the Reduced-Impact Logging Project, can
supplement but not replace other efforts such as energy conservation, fuel switching, and increased
power plant efficiencies. For example, application of my estimate (43 Mg ha-') to the loggable
portion of the project area (66% of 1400 ha) yields 39,732 Mg C, equivalent to about 11% of the
annual emissions from a 200 MW coal-burning energy plant (Freedman et al. 1992). Given the
ubiquity of poor timber harvesting practices, considerable scope exists for application of reduced-
impact logging in other tropical, subtropical, and temperate-zone forests. This approach to offsetting
carbon may not be appropriate for forests with large proportions of their ecosystem carbon stored in
fallen logs and soil organic matter because harvesting operations can result in large net losses in
carbon over time (Harmon et al. 1990). A forest's potential for retaining carbon by altering
harvesting practices is primarily a function of the forest's biomass, the baseline to which the damage-
controlled site is compared, possibilities for damage reduction, and the volume of timber extracted.
In the pilot project in Sabah, about 36% (or 15 Mg C ha-1 ) of the additional carbon retained in RIL
areas was related to volume extracted and debris from felled trees (i.e. treetops, stumps, butt roots).
Reductions in net volume extracted of the magnitude observed in the project in Sabah are not
inherent to RIL operations but are related to areas in streamside buffer zones, terrain, expertise of
operators, and supervision of field operations. As the project expands in Sabah, I expect differences
related to number of trees felled per ha to disappear, as will this proportion of the carbon savings.
As policies supporting forestry-based carbon offset projects develop, so will a system for
evaluating potential projects, their credibility, reliability, and verifiability (Dixon et al. 1993).
Describing the costs and benefits of reduced-impact logging as a harvesting technique is complicated
by externalities and undervalued environmental services (Kramer et al. 1992). Assessment of the
cost-effectiveness of applying RIL techniques for offsetting carbon will require an even more
complex analysis. Reduced-impact logging carbon offset programs may be attractive to power
companies because, relative to many tree planting programs, the carbon benefits come earlier and
less risk is involved. The risk of losing the investment to pests, fire or disease are small relative to
that for trees in industrial plantations with rotations of 7-20 years.
Expansion of the RIL approach to carbon offsetting is predicated on international acceptance
of joint implementation. Hesitancy is coming from developing countries suspicious about the
motivation of wealthy countries. Also, as nations industrialize they will develop their own need for
reducing net emissions. Although it would not be sensible for developing countries to sell all of their
inexpensive offset options to the industrialized nations, poorly managed forests abound, and the
world's supply of forestry-based carbon offset options is not in jeopardy.
Policy makers will look to biologists and foresters to provide estimates of impacts of
forestry-based carbon offset programs. Particularly lacking are data on the biomass of very large
trees, including roots. For many tropical trees, little is known about the effects of mechanical
damage on growth rates, wood quality, fruit production, mortality rates, and pathogen attack.
Foresters promote vine cutting as a useful tool for reducing logging damage, but the implications of
vine cutting on wildlife species, particularly frugivores and foliovores should be investigated.
Logging stimulates leaf production in some species (e.g., Johns 1988) but few data exist describing
changes in fruiting phenology or fruit abundance following logging (but see Wong 1983; Johns
1988). The incidence of weed invasions in logged-over forest appear related to gap size, soil
disturbance, and pre-logging species composition. Research directed towards elucidation of these
relationships could be useful for predicting impacts of harvesting clusters of trees in comparison with
scattered individuals and trees growing in areas with climbing bamboo (Dinochloa spp.), and the
importance of minimizing soil disturbance. Further efforts to quantify the impacts of forest
management activities on carbon storage or sequestration rates through models (e.g., Cropper &
Ewel 1987; Dewar 1990; Dewar & Cannell 1992) and the validation of models will contribute to the
database from which proposed carbon offset projects can be assessed.
A SIMULATION MODEL OF CARBON DYNAMICS FOLLOWING LOGGING
Reductions in logging damage can result in increased carbon retention in forest biomass
(Chapter 3). In this chapter, I examine the effect of this biomass retention on long-term carbon
storage over a 60 year period in dipterocarp forest. I present a simulation model of dipterocarp forest
development based on FORMIX, a model developed by Bossel & Krieger (1991). My model tracks
carbon stored in forest biomass and necromass pools over time and is intended to simulate forest
recovery following logging. The amount of carbon stored in a logged or silviculturally managed
forest is influenced by factors and processes that are both internal to the system (e.g., species
composition, growth rates, decay rates) and external to the system (e.g., rotation times, logging
damage, timber volume extracted). The model provides a tool for organizing this information. I
evaluate the model using sensitivity analyses and comparisons with field observations and published
data on biomass and necromass stores in primary and logged dipterocarp forest. Finally, I use output
from simulations to evaluate effects of reductions in logging damage on carbon storage.
Carbon Storage and Patterns of Recovery Following LoggiMng
When timber is removed from a forest, total ecosystem carbon storage declines. Selective
cutting often involves harvesting only a few trees, but many others are usually damaged. As
damaged trees die and logging debris decomposes, total carbon stored declines further. Only when
carbon sequestration in growth and recruitment exceeds carbon losses in death and decay will total
carbon storage increase. Over time and in the absence of large-scale disturbance, ecosystem carbon
storage may approach an asymptote, the position of which may or may not be the same as before
Logging may influence a site's potential to store carbon (i.e., height of the asymptote) and
the rate at which the forest recovers and sequesters carbon. For example, soil compaction and
erosion, often a consequence of ground-based yarding, may decrease site productivity and,
consequently, decrease carbon storage potential. If, after selective logging, the residual stand
becomes dominated by vines, grasses and sedges, or pioneer trees, growth of persistent forest tree
species, many with high wood densities and large stature, may be suppressed for several decades1.
Changes in forest structure associated with selective logging operations in Sabah influence
environmental conditions within the forest and increase the forest's vulnerability to fire (Uhl &
Kauffminan 1990). An increase in fire frequency also reduces the forest's potential to accumulate
carbon in biomass.
The current state-mandated management plan for timber-producing dipterocarp forests in
Sabah calls for 60-year cutting cycles. Consequently, logging impacts that influence the rate of
carbon storage between logging and 60 years post-logging are of particular interest. The degree to
which total carbon stores decline during and after logging depends on many factors, including timber
volume extracted and how this volume is distributed among diameter classes, incidental damage to
the residual stand, and the degree to which the vegetation responds to opening. Recovery rates will
be influenced by site productivity, species composition, changes in necromass stores, long term
1 My concept of pioneer tree species includes species that, relative to the common
dipterocarp forest species, have low density wood (< 0.4 g cm3), short lifespans (10-40 years),
produce copious quantities of seeds that require relatively high light and temperatures for
germination and establishment, and do not maintain an understory seedling bank. I use "persistent"
forest species in reference to tree species that are able to establish in shade and that maintain a
seedling bank rather than a seed bank.
effects of nonfatal tree damage, the duration of elevated mortality rates following logging, and
impacts of soil damage on vegetation recovery.
In this paper I use a computer simulation model of carbon flow in dipterocarp forest
following logging to explore the potential influence of several factors on carbon recovery.
Specifically, I use output of simulations to address the following questions: 1) Over 60 years, how
does mean carbon storage in a logged forest change with reductions in logging damage?; 2) How do
changes in post-logging mortality rates affect mean carbon storage and the final biomass storage
over 60 years?; and, 3) How might a temporary post-logging shift in species composition affect
ecosystem carbon storage patterns over time?
Background and Basic Model Structure
Previous research has clarified some aspects of forest development and the carbon cycle in
dipterocarp forest. Primary productivity and organic matter dynamics were studied in a dipterocarp
forest ecosystem in the early 1970s, as part of the International Biological Program (IBP) in
Malaysia (synthesized in Kira 1978). The researchers presented a pool and flux model of ecosystem
carbon cycling for steady-state conditions (Kira 1987). Using a portion of the IBP data, Bossel and
Krieger (1991) developed a physiologically driven model of dipterocarp forest development and
natural treefall gap dynamics called FORMIX. FORMIX is useful for looking at forest growth and
structural development and, in combination with the IBP data, provides a base for modelling carbon
flow in dipterocarp forest. As originally published, however, FORMIX does not adequately
simulate forest recovery from logging with bulldozers because it does not incorporate community-
level and ecosystem changes fundamentally associated with soil disturbance and logging in Sabah.
Changes I have identified as potentially important to carbon storage are elevated post-logging
mortality rates, changes in seedling survival, and increased representation of pioneer trees among the
The model used in this chapter, which I refer to as C-REC (for carbon recovery), tracks
carbon stored in dipterocarp forest and is intended to simulate forest dynamics both before and after
logging (Appendices C and D). The basic system is scaled to 1 ha, uses annual time steps, and
includes carbon pools for aboveground biomass and necromass (Fig. 4-1). Carbon storage in the
pools is followed as trees grow, shed litter, die, and are replaced. The basic structure of C-REC is
identical to FORMIX, as are processes describing carbon gain through photosynthesis, transition
rates between layers, recruitment, and mortality rates. C-REC differs from FORMIX in that it
simulates carbon transfer from biomass to necromass through tree mortality and litterfall.
Necromass decomposition is simulated as proportional mass loss. Coarse woody, small %%ood\, and
fine debris decay include transfer of carbon to soil organic matter. Carbon is lost from the soil
organic matter pool at 5% mass loss per year (based on Yoneda et al. 1977; Kira 1978). Carbon
stored in roots, shrubs, herbs, vines, and in mineral soil below 50 cm is not included in the C-REC
As in FORMIX, I divided the forest into 5 canopy layers (Fig. 4-1) which correspond to the
following: Layer 1, canopy trees (>45 cm dbh); Layer 2, subcanopy trees (25-45 cm dbh); Layer 3,
pole-sized trees (10-25 cm dbh); Layer 4, saplings (1-10 cm dbh); and, Layer 5, seedlings (0-1 cm
dbh). Initial stem densities are entered for a hectare of representative unlogged forest. Input files
contain individual trees identified by a number and diameter at breast height (at 1.3 m, hereafter
dbh). Initially all trees are assumed to be persistent forest species characterized by attributes of the
Shoreajohorensis-Parashorea group of the Dipterocarpaceae (e.g., photosynthetic rates, allometric
sisuqluAsoIoqd uOJA UVD uoq.jD
relationships, wood density; Table 4-1); Dipterocarpaceae dominate this forest in terms of basal area
and tree stem density (Chapter 3). Using these data, stem, branch, leaf, and total biomass are
calculated for each tree using diameter-biomass regression equations (Kira 1978; Brown et al. 1989).
Layers are defined and tracked by total biomass and tree numbers.
Annual gross photosynthate production is calculated for each layer and is based on total
layer leaf area, incident solar radiation, light attenuation through the canopy, and photosynthetic
capacity (following a light response curve; Tables 4-1 and 4-2). Litter production and respiration by
fine roots, leaves, and stems are subtracted from gross photoproduction to yield net annual biomass
production per layer. A complete description of the basic model is found in Bossel & Krieger
Allometric relationships are used to calculate mean stem diameters and crown projection
areas for each layer (Table 4-3). When a layer's mean stem diameter exceeds the maximum diameter
set for the layer, a given proportion of the trees (and associated biomass) are transferred to the next
layer. Transition probabilities were calculated by Bossel and Krieger (1991). Each layer is
associated with two specific mortality rates, a standard rate and a higher rate which applies to
crowded conditions. Crowded conditions exist when the layer's canopy is completely closed as
determined by crown area/stem diameter ratio, mean stem diameter, and number of trees per layer
(Tables 4-1 and 4-3). Recruitment into the seedling layer is controlled by the number of trees >25
cm dbh; each mature tree contributes 1000 seedlings per year (Table 4-1); the base survival rate for
established seedlings is 50% per year.
Characteristics (with code name) of the 2 types of tree species used in the model.
Values that differ from those used in FORMIX (Bossel & Krieger 1991) are noted
by superscripts. Variables not defined here are defined in Table 4-2.
Persistent Forest Pioneer
P,. (g CO2 m hr"1) 1.5 2.5a
M (g CO hr-' W-') 0.015-0.025 0.04b
PR 0.50 0.35c
Photosynthetic Production for Litterfall(PSD; proportion) 0.10 0.10
Stemwood fraction (TR) 0.70 0.70
Wood density (G; g cm-3) 0.52d 0.33d
Crown diameter ratio (CD; m m-1) 25 32c
b Walters & Field 1987
d Burgess 1966
Table 4-2. Equations describing carbon gain (taken from Bossel & Krieger 1991). Subscripts
refer to specific layers that are defined in the text and Figure 4-1.
Solar Radiation Received By a Layer 11 = I ,+1 EXP (-K i+ LAI i+)
Gross Photosynthetic Production PS, = C*(PI./K, )*LOG,((1 + (M / P,,)*I ,)/(1+ (M/P,,J)*I -1))
Photosynthetic Production Adjusted for Crown Area PT = PS, AT,
Photosynthetic Production Adjusted for Leaf and Root Respiration PB, = PT i PR
Photosynthetic Production Adjusted for Stem Respiration Cgain = PB (R, B )
I = radiation above the canopy, 335 W m'2.
K = light extinction coefficient (values per layer in Table 4-3).
LAI = layer leaf area index (maximum values per layer in Table 4-3).
C = conversion factor from g CO2 m2 hr' to metric tons of oven dry mass per ha per yr.
M = initial slope of the light response curve (values in Table 4-1).
Pinax = maximum rate of photosynthesis at light saturation (values in Table 4-1).
AT = current crown fill ratio; represents crown cover per layer.
PR = leaf proportional energy use efficiency; accounts for leaf and fine root respiration.
R = biomass proportional energy loss rate; accounts for stem respiration, 0.06.
B = layer total biomass (Mg oven dry mass hal).
Variables describing the two species groups represented in the C-REC model, by
layer. Persistent species refers to tree species able to establish in shade and that
maintain a seedling bank rather than a seed bank. Pioneer species refers to tree
species that require relatively high light for seedling establishment and that do not
maintain a seedling bank.
Layer 1 Layer 2 Layer 3 Layer 4 Layer 5
>45 cm 25-45 cm 10-25 cm 1-10 cm Seedlings
Mortality Rate (mn ) 0.005 0.008 0.01 0.05 0.10
Crowding Mortality Rate (mc)a 0.10 0.15 0.20 0.50 0.50
Post-harvest Mortality Rate (ml)' 0.05 0.05 0.05 0.05 0.05
Maximum Leaf Area Index (LAI) 2.00 2.00 2.00 2.00 1.00
Transition Rate (TSi)D n/a 0.02 0.05 0.08 0.10
Mortality Rate (mni)8 0.01 0.01 0.05 0.05 0.10
Crowding Mortality Rate (mc,)3 0.25 0.25 0.25 0.50 0.50
Post-harvest Mortality Rate (mi)" n/a n/a n/a n/a n/a
Maximum Leaf Area Index (LAIP1) 2.00 2.00 2.00 2.00 2.00
Transition Rate (TSP,)a n/a 1.00 1.00 1.00 1.00
Common to Both Groups
Light Extinction Coefficient (K) 0.86 0.86 0.54 0.54 1.00
Form Factor (F) 0.38 0.42 0.44 0.45 0.50
Height-Diameter (HD; m inm-') 40 48 56 67 140
Maximum Diameter (DM1; m) n/a 0.45 0.25 0.10 0.01
a Expressed as proportions of individuals per hectare.
Necromass exists in five compartments: coarse woody debris (branches or logs with
diameter >15 cm); small woody litter (diameter ranging from 2 to 15 cm); fine litter (leaves, fruits,
twigs <2 cm diameter); and, soil organic matter. Dead roots are not included in the model. Initial
pool sizes and decay rates were taken from published data for Malaysian dipterocarp forests (Table
4-4). Annual inputs to the necromass pools include biomass from dying trees and photosynthetic
production that goes towards litterfall. Soil organic matter receives annual inputs of coarse woody
debris, small woody debris, and fine litter. A proportion of the soil carbon is evolved as CO2. Soil
carbon below 50 cm depth is assumed to be static and is not included in the model; this probably
represents about 40 Mg C ha' (Ohta & Effendi 1992). Root biomass also is not included in the
model; this probably represents about 20% of aboveground tree biomass (Chapter 3).
Impacts of logging on the forest are variable and depend, in part, on timber volume
extracted, the harvesting system used, and the extent of damage to the residual stand and to the soil.
Selective logging, as currently practiced in Sabah, removes a proportion of trees >60 cm dbh
(generally, 8-15 trees per ha), damages a portion of the residual stand, and generates logging debris.
The model incorporates the effects of logging in a sequence of steps.
First, timber volume extracted per ha is entered as a variable (m3). The value is converted
into biomass (Mg) using an mean specific gravity (Table 4-2) and is translated into number of trees
per ha based on the assumption that stem biomass represents 52.8% of total tree biomass (Biomass
Expansion Factor, taken from Brown et al. 1989). The biomass and number of trees felled are
subtracted from the top layer of the forest (trees >45 cm dbh). Non-stem biomass (i.e, branches,
leaves, stumps) enters the necromass pools (80% coarse woody debris, 10% small woody debris,
10% fine litter).
Variables (with code names) describing necromass stores and fluxes, initial
conditions listed with reference. O.D.M. refers to oven dry mass.
Coarse Woody Debris (qc)
Woody Litter Conversion to Carbon
Small Woody Litter (qswl)
Fine Litter (qfl)
Fine Litter Conversion to Carbon
Soil Organic Matter (qsoil)
Leaf Litter Decay Rate (fldk)
Leaf Litter to Soil (fltoS)
Small Woody Decay Rate (swldk)
Woody Debris Decay Rate (qcdk)
Woody Debris to Soil (qctoS)
CO2 Evolution From soil (seflx)
49.5 Mg O.D.M. ha'
50% carbon by mass
2.5 Mg O.D.M. ha'
2.4 Mg O.D.M. ha-1
46.9% carbon by mass
33 Mg C ha-1
71% mass loss yr"'
2.2% transfer to soil yr-1
50% mass loss yr-1
14.4% mass loss yr-'
4.6% mass loss yr'
5% loss yr-1
Yoneda et al. 1977
Burghouts et al. 1992
Burghouts et al. 1992
Ohta & Effendi 1992
Burghouts et al. 1992
Burghouts et al. 1992
Yoneda et al. 1977
Yoneda et al. 1977
Second, the proportion of trees receiving fatal damage during logging is entered; a single
value is used to describe fatal damage for all diameter classes. This proportion of each layer's
biomass and individual trees is transferred to the necromass pools (for allocation see Appendix C).
This input variable (mean proportion of trees fatally damaged across all layers) is then used to
represent the proportion of the 1 ha stand that will be colonized by pioneer tree species rather than by
persistent forest species during the first two years after logging.
Prior to logging, pioneer tree species are uncommon in the dipterocarp forests of Sabah
(Whitmore 1978; Comer 1988) but they are a dominant component of logged forests in Sabah, often
occurring as monodominant stands (Fox 1968b). Pioneer trees are incorporated into the model to
provide a way of exploring the impacts of a temporary shift in composition, a shift away from
dominance by relatively slow growing, persistent species to relatively fast-growing, colonizing
species with low wood densities. Pioneer trees are represented by a suite of physiological and
allometric attributes distinct from the trees that dominated the site before logging (i.e., the
dipterocarps; Table 4-2).
Bomrnean species of pioneer trees tend to establish in disturbed sites with open canopy.
Establishment patterns suggest that pioneer recruitment increases with some soil disturbance (Putz
1983; Kennedy 1991), but on compacted soils and subsoils typical of skid trails and log landings in
Sabah, pioneer tree recruitment is sparse (Pinard et al. 1996). In the model, pioneer tree seedlings
establish at 13500 seedlings per ha, equivalent to 1.25 Mg O.D.M. ha-' (Pinard et al. 1996;
The model tracks pioneer tree stand development separately from the residual forest.
Carbon gain and transitions within the pioneer tree stand subset follow the same processes described
earlier for the persisent forest species though specific parameters differ (Table 4-2). Layer transition
probabilities for pioneers are set to simulate development of an even-aged (i.e., one layer) stand.
Seedlings of persistent tree species begin to establish under the pioneer tree forest 5 years
after logging. Mature residual trees (>25 cm dbh) provide seedlings to the pioneer tree forest.
Generally, fruits of dipterocarp trees do not disperse far from parent trees (Ashton 1982) so both the
density and distribution of mature residual trees are important for seedling establishment under
pioneers. The model assumes that, as the area occupied by pioneer trees increases (i.e., proportion of
stand fatally damaged), the proportion of residual trees able to disperse seeds into the pioneer forest
decreases. The following equation describes the relationship used in the model to determine the
number of individuals contributing seedlings of persistent forest species under the pioneer stand:
NST = (N1 +N2) ((1- DAMF)2)
where NST equals the number of tree contributing seeds in a given year, N,, the number of trees in a
layer, and DAMF, the proportion of the stand receiving fatal damage.
In Ulu Segama Forest Reserve, observations of planted dipterocarp seedlings suggest that
seedlings on skid trails experience higher mortality rates than seedlings off skid trails in logged forest
(P. Moura-Costa pers. comm.). The relatively high seedling mortality rates on skid trails are due, in
part, to increased incidence of animal browsing and trampling (Pinard, unpubl. data). In the model,
survival of seedlings of persistent tree species in the pioneer stand is calculated using the following
survPD = basesurvival (1 AST)
where survPD equals persistent forest species seedling survival in the pioneer stand, basesurvival is
the baseline annual seedling survival rate, AST is the proportion of area with soil disturbance.
Although seedling growth rates are also affected by adverse soil conditions on skid trails (e.g.,
compacted soils or nutrient poor subsoils), the model does not incorporate any changes related to
carbon gain for trees on skid trails.
Although maximum lifespans of colonizing tree species are variable, the maximum lifespan
for the species of Macaranga that dominate the pioneers in Ulu Segama, is probably close to 30
years (Fox 1968b). To simulate senescence of pioneers, annual mortality rates of the pioneer trees
increase to 50% at 30 years after logging. The model continues to track the "pioneer" stand but,
after 35 years, the subset is predominantly trees of persistent species.
During logging, a proportion of the residual stand is damaged but some this damage (e.g.,
crown or bark damage) does not always cause immediate tree death. This damage is assumed to
influence growth rates of affected trees, simulated by removing 25% of the crown area of damaged
trees. Growth and yield plot studies in logged dipterocarp forest document an elevation in mortality
rates for 2 to 8 years following logging (Wan Razali 1989; Korsgaard 1992). These tree deaths may
be related to damage received during logging or may be related to changes in environmental
conditions in the residual stand. The model represents this phenomenon by uniform application of
5% mortality rates for five years following logging.
Methods for Simulations and Evaluation
Simulations were run under both no logging and logging scenarios. All carbon pools were
tracked over a 60 year period. Longer simulations (1000 yrs) were also performed to evaluate the
model's stability. As part of the model evaluation process, a selection of variables, constants and
parameters used in the model was increased by 15%, simulations were run, and output values of
response variables were recorded. The response variables used in the "sensitivity" analyses for a no-
logging scenario were as follows: mean total carbon storage over 20, 40, and 60 years, ending total
carbon storage at 20, 40 and 60 years, and ending total biomass in big trees (>45 cm dbh) at 60
years. Because a subset of the variables and parameters was applicable only to a logging scenario,
another set of "sensitivity" analyses was conducted assuming 125 inm3 of timber were extracted, 40%
of the residual stand fatally damaged, 20% of the area with soil disturbance, and 20% of the residual
stand nonfatally damaged. The response variables used in these logging scenario analyses included
those listed above but also included total biomass in pioneer-dominated forest at 20 years.
To evaluate the output of the no-logging scenario, I compared estimates of pre-logging
aboveground biomass and necromass stores from the study site with results from simulations run for
60 and 500 years. To evaluate the output of the logging scenario, I compared simulated densities of
pioneer trees at 6 and 18 years after logging to data from logged forest in Ulu Segama. Also,
simulated estimates of the amount of coarse woody debris 6 years after logging were compared with
measurements of detrital stores in logged forest in Ulu Segama.
To evaluate the impacts of timber volume extracted on mean carbon storage, I ran a series of
simulations in which damage was held constant and timber volumes were increased from 0 to 200 m3
in 25 m3 increments. Mean timber volume extracted from the study sites in Ulu Segama was about
125 m3 ha-1 (Chapter 3), so I used this value for all subsequent simulations.
The rationale for promoting reduced-impact logging as a carbon offset is based on the
assumption that more carbon is retained in forest biomass when logging damage is lessened. To
evaluate the importance of reduced logging damage for ecosystem carbon storage, I ran a series of
simulations holding constant the volume extracted (125 inm3), nonfatal damage (0%), and area in skid
trails (20%) but increased the proportion of residual stand killed in 10% intervals from 10 to 90%
In the study sites, about 22% of the individuals in the residual stands in both conventional
and RIL areas received damage that did not immediately result in tree death (Chapter 3). To evaluate
the potential importance of nonlethal damage to carbon storage, I ran two series of simulations in
which I varied mortality rates following logging. In the first series, the duration of elevated mortality
rates (0.01 for all layers) was increased in 1 year increments from 2 to 10 years. In the second series,
duration was set at 5 years, and post-logging annual mortality rates ranged from 1 to 12%. To
examine the impacts of reducing crown area for the proportion of trees receiving nonfatal damage, I
ran simulations reducing crown area of damaged trees from 80% to 10% of full crown. I also ran a
series of simulations in which the proportion of nonfatally damaged trees was increased in 10%
increments from 0 to 90%, holding volume extracted, area in skid trails, and fatal damage constant.
Conventional and reduced-impact logging, as described by the data in this dissertation
(Chapters 2 and 3), differ in terms of volume extracted, fatal damage, and soil damage. To compare
the integrated effects of these differences for carbon storage, I ran the model using values observed
for each logging method. For conventional logging, the input variables were 154 m3 ha' timber
extracted, 16.6% area with soil disturbance, 40% of the stand fatally damaged, and 20% of the stand
with minor damage. For reduced-impact logging, the input variables were 104 m3 ha' timber
extracted, 6.8% area with soil disturbance, 20% of the stand fatally damaged, and 20% of the stand
with minor damage.
I used mean total carbon storage over time as the response variable for simulations exploring
the effects of increasing volume extracted, fatal damage, nonfatal damage, and mortality rates.
Results from sensitivity analyses indicated that mean carbon storage was relatively insensitive to
small changes in parameter values. For the simulations, I used the following three time intervals: 60
years to represent one cutting cycle, 40 years to represent the NEES-ICSB project lifespan (Chapter
5), and 20 years to allow me to identify trends specific to a shorter time period.
Results and Discussion
Evaluation of Model Simulations
Over a 1000 year time span, simulated carbon stores in the unlogged forest fluctuate
between 200 and 265 Mg ha' (Fig. 4-2A); mean carbon storage over a 60 year simulation was 220
Mg ha-1 (SD = 11). Aboveground biomass ranged from 130 to 220 Mg C ha' and showed a mean
value of 166 Mg C ha-1 (SD = 19.5) over a 60 yr simulation, and 170 Mg C ha-' (SD = 23) over a
500 yr simulation. The distribution of biomass across diameter classes fell within the range of values
observed in Ulu Segama before logging (Table 4-5). Mean necromass store over a 60 year
simulation was 54 Mg C ha-1 (SD = 8.9; Fig. 4-2C). Coarse woody debris stores fluctuated between
10 and 60 Mg C ha-' and trends were negatively associated with fluctuations in total biomass stores
(Fig. 4-2, B and C). The mean quantity of coarse woody debris over a 60 year simulation (13.5 Mg
C ha-', SD = 7.5) was similar to the mean value recorded in our plots in Ulu Segama (mean = 12.2
Mg C ha-', SD = 2.3; unpubl. data).
Simulated biomass stores cycled with an approximate 100 year frequency (Fig. 4-2A).
Stand dynamics involving fluctuations of the magnitude observed in simulation results could be
expected in an area prone to regularly occurring large storms, droughts, or fires but would not be
expected if individual treefall gap dynamics were the principal structuring phenomenon in the forest.
However, one of the limitations of the C-REC model is that the entire hectare behaves as a unit.
When the overstory is "filled" with trees, overstory mortality rates switch to a higher rate, causing a
decline in biomass that affects the full hectare, similar to 1 hectare gaps. Natural forest canopy gaps
are generally much smaller, =0.02 ha. A spatially explicit model that incorporates a mosaic of
interconnected patches would simulate natural forest dynamics more realistically than the C-REC
model, meaning that the extreme fluctuations would be damped (e.g., FORMIX2, Bossel & Krieger
0 200 400 600 800 1000
Figure 4-2A. Results from simulation with no logging. Carbon stored in aboveground biomass,
necromass, and both (total) over 1000 years.
0 50 100 150 200 250 300
30 Coarse woody debris Soil
Small woody little
0- I I I I
0 50 100 150 200 250 300
Figure 4-2B and C. Results from simulation with no logging. B) Changes in carbon storage
in 5 canopy layers identified by dbh class. C) Changes in carbon storage in soil, coarse woody
debris, fine litter, and small woody litter.
Mean aboveground biomass (Mg C ha-1) for output from model simulations over a
60 year run without logging compared with mean biomass for 8 experimental
logging units before logging (Chapter 3).
Dbh class Model Results Dbh class Field Measurements
>45cmdbh 115 >40cm 120(28,8)
25-45 cm 25 20-40 cm 23 (3,8)
10-25 cm 17 10-20cm 11 (2,8)
1-10 cm 8 1-10 cm 7 (1,8)
Following selective logging, carbon storage dropped to a low of 97 Mg C ha-1, 7 years after
harvesting (Fig. 4-3A). Ecosystem carbon storage did not reach pre-logging levels (213 Mg C ha-')
within the 500 years after logging (Fig. 4-3A). Carbon storage peaked approximately 120 years after
logging (about 150 Mg C ha'), after which time cycling was similar to that seen in simulations
without logging. The mean carbon storage over 60 years after logging was 107 Mg C ha-' (SD =
9.9), about 52% of the level for the no-logging scenario. The small peak in carbon storage that
occurred about 30 years after logging was related to a peak in pioneer tree biomass and necromass
production associated with the death of the pioneer trees (Fig. 4-3 A, B, and C).
During the first 35 years after logging, only 23% of mean total forest biomass was in pioneer
trees even though the pioneer forest dominated 40% of the site (Fig. 4-3B). During the first 60 years
after logging, 84% of the mean biomass was in residual trees (Fig. 4-3B). Persistent tree species that
establish beneath the pioneer tree canopy increase in importance (for biomass storage) beyond 50
years after logging. Prior to this, these trees represent only 8% of the mean biomass stored per year
(7 Mg C ha'). Before logging, the model forest plot contained 41 trees >45 cm dbh per ha; 60 years
after logging, the layer contained 19.3 trees >45 cm dbh per ha.
Results from a simulation of logging (125 min3, 20% soil disturbance, 40% fatal damage, 20%
nonfatal damage), generated pioneer tree densities similar to those observed in logged forest in Ulu
Segama Forest Reserve (Chapter 2). At 6 years after logging, simulated pioneer tree density was
1603 stems ha1, with trees belonging to layer 4 (1-10 cm dbh). At 18 years after logging, simulated
density was 51 stems ha-1, at which time all pioneer trees were in the uppermost layer (>45 cm dbh).
Observed pioneer tree densities in logged forest, 18 years after logging, overlapped with simulated
values (pioneer trees >5 cm dbh, mean = 188, SD = 244) but few pioneers were found with dbh
greater than 45 cm.