Introduction

Forest certification emerged in the 1990s as a market-based mechanism to promote sustainable management of forest resources (e.g., Auld and others 2008). While there is a great deal of indirect evidence suggesting that certification improves management practices in production forests, direct field measures of impacts are scarce (Blackman and Rivera 2010). This study helps alleviate this deficiency by providing data on some of the environmental impacts of selective logging in a certified and an adjacent uncertified forest concession in Gabon. It also highlights some of the inherent difficulties in achieving statistical robustness in evaluations of the impacts of forest management certification and other conservation interventions.

The principles and criteria by which responsible tropical forest management operations are assessed in certification programs, like the Forest Stewardship Council (FSC), address a variety of environmental, social, and economic factors (Burger and others 2005). Environmental benefits are derived from the short-term local impacts of planned and improved logging practices, including proper road design and construction, with additional positive effects at broader scales (e.g., avoiding unnecessary sedimentation of water bodies) and that extend into the social, economic, and policy dimensions of forest management. This study focuses only on the direct, short-term, biophysical impacts of selective timber harvesting based on field data. We compare a certified concession, in which reduced-impact logging (RIL) methods were applied, with an adjacent concession that was subjected to conventional logging (CL). We then discussed the results with emphasis on the long-term impacts of logging on residual stand damage and carbon emissions, keeping in mind that this is a non-replicated study of an intervention that was not assigned at random. RIL includes worker training and supervision along with preparation of detailed harvest plans with demarcated extraction paths (i.e., skid trails), directional felling to reduce collateral stand damage, and protection of riparian areas (e.g., Putz and others 2008a).

Selective logging, no matter how carefully conducted or low in intensity, changes forests. Because timber removal reduces standing stocks of carbon in forests that continue to satisfy the definition of forest, from a carbon sequestration perspective, even RIL causes forest degradation (FAO 2005). Logging also changes forest structure and species composition, but the reported magnitudes of these impacts vary among forest types, taxa, stand histories, logging intensities, and research protocols (e.g., spatial and temporal aspects of sampling); they presumably also vary with logging techniques, but these differences are less well-studied (White 1994; Putz and others 2012). The adoption of sound management practices, as indicated by voluntary third-party certification by reputable organizations, like the FSC is presumed to reduce these impacts to acceptable levels, which some authors refer to as sustainable forest management (SFM; Write and others 2002; Sheil and others 2004). Although credible certification is widely touted for its high environmental, economic, and social standards (Tikina and Innes 2008; Eba’a Atyi and others 2006; van Kuijk and others 2009; Sheil and Meijaard 2010; Cerutti and others 2011), the assumed benefits need to be evaluated with field measurements (Blackman and Rivera 2010).

All credibly certified forest operations include measures for biodiversity conservation and environmental protection in their management plans. Examples include the designation and protection of riparian zones and other areas of particularly high conservation value. Certified forest concessions are also expected to be logged in compliance with RIL guidelines that encompass careful planning of the road infrastructure, operational logging plans, directional felling, and other elements (Eba’a Atyi 2002). Several studies have shown that experimental use of RIL reduces harvest-induced carbon emissions (reviewed in Putz and others 2008a). In contrast, claims of increased biodiversity retention in certified forests (e.g., van Kuijk and others 2009; Zagt and others 2010) are mostly based on indirect indicators and secondary sources, with few studies of any sort conducted in Africa (but see Poulsen and others 2009; Cerutti and others 2011).

In Gabon, where this study was conducted, forest certification is promoted by the government and appears to have helped convince policy-makers to define new technical and legal standards for forest resource management (Eba’a Atyi 2004). In early 2012, four of 44 granted concessions (9.7 % of the area of the production forests) were FSC-certified, and two concessions (8.2 % of the concession area) were Keurhout-certified (de Wasseige and others 2009; WRI 2009; Nasi and others 2012; http://info.fsc.org). To determine whether forest certification serves to sustain timber yields, reduce carbon emissions, and preserve tree species diversity, we collected field data to compare an FSC-certified concession with an adjacent concession that was not certified.

Like most studies on the impacts of environmental interventions, the treatments (certified or not) were not assigned at random. A further deficiency in our experimental approach is that by treating sample plots in the two concessions as replicates, we are guilty of pseudo replication (Halme and others 2010). We also lack an unlogged control area and thus must assume that any changes in the logged stands are the result of the logging treatments to which they were subjected. These shortcomings notwithstanding, the study was carried out over a 2-year period during which no general environmental changes were detected, and the two concessions compared are well-matched on the basis of several criteria (Table 1). On the basis of these criteria, we pose the uncertified concession as a qualitative counterfactual for this case study that builds on explanatory approaches to evaluate an intervention with low replication (Andam and others 2008; Bamberger and others 2009; Pattanayak and others 2010; White and Phillips 2012).

Table 1 Descriptive characteristics of the CEB-FSC and SEEF-CL logging concessions in Gabon (A) and matching variables utilized to qualitatively construct a counterfactual for the comparison of impacts of certified and non-certified logging in two forest concessions in Gabon (B)
Full size table

Methods

Study Sites

FSC-Certified Site (CEB-FSC)

This 508-ha research site is located in the FSC-certified logging concession of Compagnie Equatoriale de Bois (CEB), a member of the Switzerland-based Precious-Woods Group (www.preciouswoods.com). The study was conducted in a cutting area in Milolé, Province of Ogooué-Ivindo, southeastern Gabon (0°15′S; 12°45′E; Fig. 1). The concession is bordered on the northwest by Ivindo National Park, on the north and southeast by the Corà Wood Gabon concession (CWG), and on the west by the Société Equatoriale d’Exploitation Forestière (SEEF) concession described below (Fig. 1). This mixed lowland tropical forest is characterized by the presence of Aucoumea klaineana (Burseraceae, okoumé), the principal source of commercial timber, along with Julbernardia pellegriniana (Caesalpiniaceae, béli) and Scyphocephalium mannii (Myristicaceae, sorro). The rich fauna includes large mammals such as forest elephants (Loxodonta africana cyclotis), gorillas (Gorilla gorilla), and chimpanzees (Pan troglodytes). Soils are mostly oxisols developed on megalithic mid-Precambrian rocks (Martin and others 1981; Leonard and Richard 1993). The climate is equatorial with a long rainy season (February–May), mean annual precipitation of 1,300–1,500 mm, and mean annual temperatures of 21–28 °C (Doucet 2003).

Fig. 1
figure 1

Locations of study sites in the CEB-FSC and SEEF-CL logging concessions in Gabon

Full size image

CEB started commercial logging in the region after obtaining its 616,700 ha concession in 1987; the first management plan was developed in 2000. Of this concession, 15,737 ha is allocated solely for conservation objectives and 18,588 for non-timber forest product harvesting and environmental service provision (http://info.fsc.org). Areas classified as high conservation value forests covering 2.6 % of the concession were demarcated in the concession, but none is located in our 508 ha study area (CEB 2007). The concession was FSC-certified by Bureau Veritas in 2008 after starting to implement RIL practices in 2007 when its management plan was approved by the Gabonese Ministry of Waters and Forests. In the entire forest concession, CEB harvests a total of 41 species, but A. klaineana constituted about 80 % of the annual (1996–2009) average production of 200,000 m3 (CEB 2007). Four tree species are protected for their biodiversity values and were flagged prior to harvest. Until the governmental log export ban in 2010, most logs were exported; since then, annual harvests declined to 160,000 m3, and the roundwood is processed in CEB’s own sawmill (G. Tokpa, 2010 pers. comm.).

We established twenty 1-ha permanent plots prior to logging (April–August 2010) in a previously unlogged 508-ha area separated from Ivindo National Park by a 5-km wide buffer zone (Fig. 1). The area was logged in October–December 2010 by trained fellers using Stihl 780 chainsaws with logs yarded along pre-planned skid trails with a Caterpillar 535 crawler-tractor. Logging damage was assessed in January–February 2011.

Conventionally Logged Site (SEEF-CL)

The conventionally logged study site (200 ha) was demarcated in a previously unlogged ha block in the 48,000 concession of SEEF in Milolé, southeastern Gabon (Fig. 1). (The concession is bordered on the west by CEB and on the south by CWG). As in the CEB-FSC site, the forest is dominated by A. klaineana, J. pellegriniana, and S. mannii; the fauna, soils, and climate are likewise similar (Martin and others 1981).

The logging system implemented by SEEF-CL, since 1998, employs tree finders to search for harvestable trees, principally of A. klaineana which constitutes 90 % of the timber harvested. Once found, the trees are measured, assessed for quality, and numbered. Tree diameter, estimated height to the first branch, stem quality, and approximate location are recorded by the team leader working without the aid of surveying equipment or a reference map. Access routes to the trees to be harvested are slashed with machetes and marked with sticks notched to indicate the number of trees to be logged. SEEF has a plywood mill where it processes only logs of A. klaineana; the 2010 log export ban, apparently, did not affect SEEF’s harvest volumes. Trees were felled using chainsaws (Stihl 070 and MS 880) and yarded with crawler-tractors (CAT 525) in January–February 2011. Logging damage was assessed in March 2011. Although CL practices were employed, the SEEF-CL fellers were trained in directional felling techniques by a worker from CEB, which means there was some potential for spillage of the certification treatment. Also, while the field study was underway, SEEF was reportedly in the process of developing a forest management plan as a step toward forest certification (T. Ricordeau 2009, pers. comm.), but that plan had not yet been implemented.

Tree Measurements

Prior to logging, we established 20 and 12 permanent 200 × 50 m (1 ha) plots at random locations in the CEB-FSC and SEEF-CL areas, respectively. Although these are technically sub-plots and not true treatment replicates (i.e., they are pseudo-replicates; Hurlbert 1984, 2004), the large size and substantial heterogeneity within each logging area at least partially mitigates this statistical design deficiency (e.g., Heffner and others 1996; Oksanen 2001; Paillet and others 2010). Within each plot, all trees ≥10 cm dbh were measured, tagged, mapped, classified according to stem quality and crown position (suppressed, side-lighted, sub-dominant, co-dominant, dominant; see Hall and others 2003), and assessed for the presence or absence of lianas. Trees were identified to species based on vegetative characteristics. Voucher specimens were collected from trees that could not be identified in the field for species determination at the Arboretum de SIBANG in Libreville, Gabon.

Above-Ground Biomass Estimates

Total AGB of trees in each plot was estimated using the allometric equation of Chave and others (2005) that was based on the destructive harvest of 1,501 tropical trees 5–156 cm dbh:

$$ { \ln }\left( {\text{AGB}} \right) \, = \, - 1.576 \, + \, 2.179\ln \, \left( D \right) \, + \, 0.198\left( {{ \ln }\left( D \right)} \right)^{2} - 0.0272\left( {{ \ln }\left( D \right)} \right)^{3} + \, 1.036\ln \left( \rho \right) $$

where D = dbh and ρ = wood density. For trees >156 cm dbh, AGB was extrapolated using this same equation. Species-specific wood density estimates were used when available (Zanne and others 2009). When there were multiple published wood density values for the same taxon, we used the arithmetic mean; in the absence of species-specific wood density data, we used generic or familial means; and, if none of these values were available, we used the plot mean weighted by the number of trees of each species.

Damage Assessment

Logging damage was assessed in the 1-ha sample plots using methods well-established in the literature (e.g., Johns and others 1996; Whitman and others 1997; reviewed by Putz and others 2008b). Crown damage was recorded as severe (>66 % crown loss), moderate (33–66 %), or minor (<33 %). Bole damage was recorded as severe (broken bole), moderate (>100 cm2 of bark removed), or minor (<100 cm2 of bark removed). Uprooted trees were recorded as such. Root damage was recorded as major (>10 % of surface roots injured) and minor (<10 % of surface roots injured). Crown, bole, and root damages were attributed to felling, skidding, or the construction and use of log decks and roads.

Canopy gaps opened by felling in the 1-ha plots were measured in the cardinal and sub-cardinal directions from a central point based on Brokaw’s (1982) definition (gaps = forest canopy openings >20 m2 that extend down through all foliage levels to <2 m above the ground). Log landings and machine maneuvering zones in felling gaps as well as the ground surface disturbed on skid trails were also mapped and measured in the entire 508 ha (FSC) and 200 ha (CL) study areas. Skid trails were classified on the basis of the number of passes by the skidders (which always yarded single logs): primary >10; secondary: 2–10; and tertiary: 1. Skid trail widths were measured at 10 m intervals. The number of skid trails that crossed streams was determined using maps drawn with ArcGIS Desktop10 using digitized hydrological maps in combination with field data. The number of avoidable stream crossings was determined by analysis of the locations of mapped tree stumps in relation to the streams and skid trails.

Damage to trees >10 cm dbh and soils that resulted from road construction and maintenance (primary = main road used throughout the year; secondary = used only during a single 4-month harvest season) was assessed in 5 × 100 m (0.05 ha) plots placed at 100 m intervals on alternating sides of roads through the 508 ha FSC and 200 ha CL study areas. Road widths were measured at 20 m intervals.

Logging-Induced Conversion of Above-Ground Biomass (AGB) into Necromass

To estimate the amount of AGB of trees >10 cm dbh that was converted into necromass (and therefore committed as emissions) during logging operations, we estimated the mass of each harvested and destroyed (i.e., uprooted or bole snapped) tree as described above. Stumps were measured and their mass estimated; crown biomass was calculated by subtraction of the mass of wood harvested from the total estimated tree mass after accounting for the stump mass. Due to lack of data on decomposition rates, emissions are assumed to be immediate. We also did not account for the residence time of carbon in products manufactured from the harvested wood.

Due to lack of data on the fates of damaged trees, we estimated committed emissions of carbon due to logging in two ways. For our conservative estimates, we assumed that all damaged trees survive and calculated emissions from just the harvested and destroyed (i.e., uprooted) trees. Given that many damaged trees will die or develop stem hollows within a few years after logging, we also present emissions as the sum of damaged, destroyed, and harvested trees.

Data Analysis and Presentation

Tree Species Density

Tree species densities in the 1 ha plots before and 1–2 months after logging were compared using EstimateS 8.2.0 (Colwell 2006) with sample-based rarefaction to compute expected species accumulation curves. A randomization procedure without replacement was run to compute the density estimators and diversity index (e.g., Fisher’s alpha). We compared species density before and after logging at each site with paired t tests of the outputs from 100 randomized iterations.

Statistical Analyses and Modes of Data Presentation

Analysis of variance (ANOVA) was used to test for differences in number of trees damaged per unit length of skid trail and road. All analyses were conducted on a per hectare basis as well as by the number of trees and volumes of timber harvested to account for differences in harvesting intensities between the FSC and CL sites. In all tests, the units of replication are skid trails and roads. If overall differences were detected, Tukey’s Honest Difference tests were employed to make pairwise comparisons of means.

Further analyses were conducted to compare logging-induced changes in species density, stem density, basal area, and AGB in the certified and uncertified concessions. Analysis of covariance (ANCOVA) was used to determine the relationships between the dbh of felled trees, felling gap areas, and the number of damaged trees. To detect shifts in tree species composition in the plots before and after logging at each site, we used a canonical correspondence analysis (CCA); magnitudes of change are expressed as Euclidian distances between the plots pre- and post-logging (Borcard and others 2011; Oksanen and others 2011). All analyses were conducted with R version 2.10.1 (R Development Core Team 2010).

Results

Forest Structure and Tree Species Density

Forest Structure

A total of 6,083 trees ≥10 cm dbh were inventoried prior to logging of the twenty 1-ha FSC plots and 4,649 trees ≥10 cm dbh in the twelve 1-ha CL plots; mean stem densities (Fig. 2), basal areas, and estimates of AGB were all lower in the FSC plots (Table 1). Of particular importance to the results of this study (see below) and the reason why we discuss logging impacts both per hectare and per cubic meter of harvested timber is that harvestable volumes of the principal timber species, A. klaineana, were substantially higher in the CL area. Liana-infested trees were relatively uncommon in both sites, especially among the large trees; only 3.8 and 1.9 % of trees >60 cm dbh were liana-infested in the FSC and CL plots, respectively.

Fig. 2
figure 2

Pre-logging stem densities by diameter (dbh) class in the twenty 1-ha plots in the CEB-FSC concession and twelve 1-ha plots in the SEEF-CL concession

Full size image

Changes in Tree Species Density and Composition

Prior to logging the FSC site, the twenty 1-ha plots supported 298 species of trees ≥10 cm dbh representing 169 genera and 45 families (40 unidentified morphospecies representing 13 % of inventoried trees are included in the species tally). In the twelve 1-ha plots at the CL site, inventoried trees represented 47 families, 140 genera, and 201 species including 22 morphospecies (11 % of the total inventoried trees). Stem density was lower in the FSC than in the CL sites with A. klaineana representing only 0.2 % of the total trees in the FSC plots and 0.9 % of the trees in the CL plots (Table 1). Tree species density declined after logging by only 1.4 % in the FSC plots and 1.9 % in the CL plots. During logging, similar proportions of protected trees were damaged in the FSC plots (nine of 131) and in the CL plots (four of 53; χ square d = 0.03, P = 0.89). Logging had no detectable impact on any of the measures we used for tree diversity (Table 2).

Table 2 Tree species diversity before and after logging in the CEB-FSC and SEEF-CL logging concessions in Gabon
Full size table

Short-term logging-induced changes in tree species composition were small in both FSC (Fig. 3a) and CL (Fig. 3b) sites. Nevertheless, the mean Euclidian distances between plots prior to and after logging were greater for the CL site (Fig. 3d; mean = 0.07, s.d. = 0.04, n = 12) than for the FSC site (Fig. 3c; mean = 0.04, s.d. = 0.03, n = 20; t = 2.25, p = 0.03).

Fig. 3
figure 3

Location of plots on the first axis derived from a canonical correspondence analysis (a, b) and Euclidean distances between plots (c, d) based on the number of trees per species before and after logging in the CEB-FSC and SEEF-CL study sites, respectively; distances were greater for the SEEF-CL site (n = 12) than for the CEB-FSC site (n = 20; t = 2.25, df = 16, P = 0.003)

Full size image

Above-Ground Biomass

Pre-logging total AGB estimates for the twenty 1-ha FSC plots ranged 215.2–636.8 Mg ha−1 with a mean of 380.0 ± 86.78 Mg ha−1, whereas for the twelve 1-ha CL plots, they ranged 306.7–547.6 Mg ha−1 with a mean of 387.3 ± 81.01 Mg ha−1 (Table 1). Tree species contributed to the AGB in somewhat different proportions in the two concessions. A. klaineana represented 1.7 % of the total AGB in the FSC plots and 9.5 % in the CL plots. Immediately after harvest, AGB declined by 7.1 % in the FSC plots and 13.4 % in the CL plots. Pre-logging AGB value of trees protected for their biodiversity was 58.4 % higher in the FSC than in the CL plots. After logging, AGB of these trees declined by 2.7 and 4.1 % in the FSC and CL plots, respectively.

Logging Intensities

A total of 198 trees ≥60 cm dbh (range: 60–148 cm dbh; 93.7 ± 19.57 (mean ± s.d)) were harvested at an intensity of 0.39 trees ha−1 (5.66 m3 ha−1) from the 508-ha FSC site and 153 trees ≥58 cm dbh (range: 58–148 cm dbh; 100.3 ± 21.32) with an intensity of 0.76 trees ha−1 (11.38 m3 ha−1) from the 200-ha CL site (Table 3). The mean AGB in logs harvested from the FSC and CL sites were 5.97 and 6.93 Mg ha−1, respectively. Of the twenty 1-ha FSC plots, six were harvested and yielded a total of ten trees, while 16 trees were harvested from nine of the twelve 1-ha CL plots (Table 4). Although there was no difference in stump heights when expressed per cubic meter harvested, biomass in tree stumps at the FSC site was slightly higher than in the CL site. In contrast, biomass left in tree crowns at the FSC site was substantially lower than at the CL site. Therefore, AGB converted to necromass from felled trees was lower at the FSC site than at the CL site expressed either on a per hectare basis or in relation to the timber volume harvested (Table 4).

Table 3 Numbers of trees and volumes of timber harvested from the 508 ha CEB-FSC and 200 ha SEEF-CL study areas in Gabon
Full size table
Table 4 Logging impacts in the entire 508-ha CEB-FSC and 200-ha SEEF-CL study areas in Gabon
Full size table

Logging Damage

At the FSC site, damage to trees in plots from which none was harvested resulted from felling of nearby trees (one plot), skidding (six plots), felling and skidding (two plots), skidding and road-related activities (one plot), and felling, skidding, and road-related activities (one plot); three plots were untouched by logging. Among the three CL plots from which no timber was extracted, one was damaged by both nearby felling and skidding, another suffered only skidding damage, and one was untouched by logging.

Felling Damage

At the FSC site, the felling that occurred in six of the twenty 1-ha plots damaged on average 9.1 trees ≥10 cm dbh per felled tree. In contrast, 20.9 trees ≥10 cm were damaged per felled tree at the CL site. Of the damaged trees, 19 % in the FSC site and 25 % in the CL site suffered severe crown damage, smaller proportions suffered severe bole damage (i.e., broken, FSC = 13 %, CL = 8 %) and others were uprooted (FSC = 20 %, CL = 8 %; Fig. 4a, b). The impact of felling on AGB was lower at the FSC site than at the CL site when expressed per hectare but somewhat greater when expressed per timber volume extracted (Table 5). An ANCOVA revealed that the number of trees damaged increased with the diameter (dbh) of the felled trees at both sites; when adjusted for dbh, that number was slightly greater at the CL site (Fig. 5a).

Fig. 4
figure 4

Percentage of trees by damage category resulting from felling and skidding in the a CEB-FSC and b SEEF-CL concessions in Gabon

Full size image
Table 5 Logging impacts in permanent plots in the 508-ha CEB-FSC and 200-ha SEEF-CL study areas in Gabon
Full size table
Fig. 5
figure 5

Relationships between harvested tree stem diameter (dbh) and a the number of damaged trees and b felling gap area in the CEB-FSC (508 ha) and SEEF-CL (200 ha) concessions in Gabon based on linear ANCOVA. a CEB: number of damaged trees = 0.09 × DBH − 0.05 (n = 10; R 2 = 0.21; P = 0.181); SEEF: number of damaged trees = 0.09 × DBH + 1.6 (n = 16; R 2 = 0.05; P = 0.38); b SEEF: felling gap area = 10.9 × DBH − 388.1 (n = 13; R 2 = 0.30; P = 0.05); CEB: felling gap area = 10.1 × DBH + 205.3 (n = 12; R 2 = 0.59; P = 0.008); dashed lines represent fitted line for the CEB-FSC site and dark solid lines represent fitted line for the SEEF-CL site

Full size image

Skidding Damage

Skid trails covered 1.6 % of the 508-ha FSC site and 4.5 % of the 200-ha of the CL site (Table 4; Fig. 6). The number of skid trails that crossed streams was also lower in the FSC than in the CL sites, with means of 1.2 and 2.0 crossings per 100 ha that seemed avoidable, respectively. Skidding operations damaged fewer trees per logged hectare in the twenty 1-ha FSC plots than in the twelve 1-ha CL plots, but trees damaged per timber volume extracted did not differ (Table 5). Of the damaged trees at the FSC site, 31 % were destroyed whereas 23 % were destroyed in the CL site. Root damage was suffered by 4 and 7 % of the trees at the FSC and CL sites, respectively (Fig. 4a, b).

Fig. 6
figure 6

Maps of skid trails and felled trees in the CEB-FSC (left) and SEEF-CL (right) concessions at two scales

Full size image

Skid trails of all orders were consistently narrower in the FSC than in the CL concession (primary: 4.2 vs. 5.8; secondary: 3.6 vs. 5.1 m; and tertiary: 3.4 vs. 5.1 m; Table 6). In the FSC concession, the number of damaged trees per length of skid trail differed between primary (0.08 trees m−1), secondary (0.08 trees m−1), and tertiary (0.05 trees m−1) skid trails (F = 4.31; P = 0.013). At the CL site, 0.18 trees m−1 were damaged on both primary and secondary skid trails and 0.15 trees m−1 along tertiary trails. The impact of skidding on AGB was 6.2 Mg ha−1 lower at the FSC site than at the CL site but did not differ when expressed on the basis of timber volume harvested (Table 5).

Table 6 Ground area disturbed and logging waste in the 508-ha CEB-FSC and 200-ha SEEF-CL study areas in Gabon
Full size table

Felling Gaps Opened and Ground Area Disturbed

In the twenty 1-ha FSC plots, the 12 felling gaps ranged 270–1794 m2 with a mean per felled tree of 787 m2. In the twelve 1-ha CL plots, the 13 felling gaps ranged 293–1565 m2 with a mean of 751 m2 (Table 5). Felling gaps increased with the size of the felled tree in both areas (Fig. 5a), but were relatively larger in the FSC than in the CL sites (Fig. 5b).

Log landings covered 0.12 % of the ground surface in the 508-ha FSC site and 0.13 % in the 200-ha CL site, while machine maneuvering disturbed 0.31 and 0.25 % of the ground area in the FSC and CL sites, respectively (Table 6).

Above-Ground Biomass Converted into Necromass

AGB of destroyed trees per logged hectare did not differ, but, in contrast, when expressed as per timber volume extracted, was 0.17 Mg m−3 higher in the twenty 1-ha FCS plots than in the twelve 1-ha CL plots (Table 3). In contrast, AGB of damaged trees was 15 Mg ha−1 lower at the FSC site than at the CL site, but did not differ when expressed as per timber extracted. When the AGBs of damaged, destroyed, and harvested trees are combined, CL site resulted in twice the losses as observed in the FSC site (Table 5). Overall, AGB converted into necromass (i.e., AGB of harvested plus completely destroyed trees) was less in the FSC site (11.2 Mg ha−1; 2.0 Mg m−3 or 2.9 % of the pre-logging total biomass) than in the CL site (24.6 Mg ha−1; 2.3 Mg m−3 or 6.3 % of the pre-logging total converted into necromass; Table 5).

Damage Due to Road Construction and Maintenance

The unique primary road covered 0.8 % (75.7 m2 ha−1) of the FSC site, while the two secondary roads covered 0.7 % (33.5 m2 ha−1) of the ground surface (Table 6). In the CL study area, there were no primary roads but the two secondary roads covered 5.4 % (539.7 m2 ha−1) of the ground surface (Table 3). Secondary roads in the CL site were substantially wider (66.6 m) than in the FSC site (15.1 m; Table 7). In the twenty 1-ha FSC plots, road construction and maintenance damaged and destroyed a total of 30 trees ≥10 cm dbh (Table 4) of which 9 % suffered severe crown damage and 55 % were uprooted. Whether expressed per tree or per cubic meter of timber harvested, the wider roads in the CL area covered nearly an order of magnitude more ground surface than in the FSC area; the number of trees damaged per length of road differed among the primary (0.16 trees m−1) and two secondary roads (0.18 and 0.14 trees m−1; F = 8.7; df = 610; P < 0.001). At the CL site, none of the twelve 1-ha plots suffered road-related damage; but in the entire 200-ha study area, construction and use of the two secondary roads damaged and destroyed an average of 0.49 trees m−1 (Table 5). Overall, road construction and maintenance resulted in 2.87 Mg ha−1 (0.24 Mg m−3) AGB loss (i.e., completely destroyed trees) higher in the CL site than in the FSC site (Table 6).

Table 7 Roads in 508 ha—CEB-FSC and 200 ha SEEF-CL—study areas in Gabon compared with FAO standards (FAO 2003)
Full size table

Discussion

Impacts of Certified and Conventional Logging on Forest Structure and Composition

FSC principle 6 specifies that forest management shall conserve biological diversity and its associated values by minimizing the deleterious impacts of logging. In our CL and FSC study sites, which share many environmental characteristics, selective timber harvesting had no apparent short-term effects on tree species density and little impact on composition primarily because logging intensities were light (Hall and others 2003; Ruiz Perez and others 2005; Nasi and Forni 2006). We expect that impacts will become apparent over time, especially in the most severely affected plots, where many of the damaged trees are likely to die and pioneer species are likely to proliferate.

Above-Ground Biomass Impacts and Conversion into Necromass

Due primarily to the low harvest intensities, the immediate losses of AGB were only 2.9 and 6.3 % at the FSC and CL sites, respectively. Logging intensity at the uncertified site was nearly twice that at the certified site, and therefore, the total impacts on AGB were also generally higher when expressed on a per hectare basis. Lower logging intensity in the FSC-certified concession was apparently related to the company’s response to the 2010 log export ban as well as to the lower stocking of A. klaineana and was not due to implementation of RIL techniques or compliance with certification standards. In particular, in the 508 ha study area in the certified concession, there was no foregone timber due to harvesting restrictions on steep slopes, in stream buffer zones, or in demarcated conservation areas.

The harvest intensity in the uncertified CL concession was higher than in the FSC certified concession even though more species were harvested from the latter. From the CL site, 90 % of the volume harvested was from Aucoumea klaineana, all of which was processed in the concession’s plywood mill and was thus not affected by the log export ban. The higher logging intensity was thus mostly related to the higher standing stocks of A. klaineana timber in the CL area. As a consequence, AGB converted into necromass (i.e., committed emissions) was much lower at the FSC than at the CL site when expressed per logged hectare; this difference declined but persisted when expressed per cubic meter of timber harvested. It is noteworthy that our necromass estimates include the biomass in the extracted logs, a portion of which will be converted into end forest products with long carbon retention.

Logging Damage

Damage Due to Felling

Felling caused somewhat more collateral damage to residual trees at the CL site than at the FSC site for a number of reasons. First of all, prior to the harvest, stem densities were higher at the CL site, particularly among small trees. Not surprisingly, therefore, more trees suffered severe crown and bole damage than at the FSC site. Determination of the fates of damaged trees awaits future plot measurements, but if the responses resemble those reported in other tropical forests (e.g., Durrieu de Madron and others 2000; Jonkers 2000; Nasi and Forni 2006; Sist and Ferreira 2007; Blanc and others 2009; Mazzei and others 2010), then carbon emissions from the CL site will be much greater than those from the FSC site where fewer trees were damaged. Further plot monitoring will be needed to reveal any long-term differences in the impacts, but biomass recovery rates are expected to be faster in the FSC site because it suffered fewer deleterious stand impacts during logging.

The small difference between the FSC and CL sites in short-term biomass impacts of felling may be a result of the training in directional felling received by workers in both concessions. This might also explain the small difference in stump heights, but does not explain why much more ostensibly marketable timber was wasted in the CL site due to poor bucking.

Damage Due to Skidding

The observed reductions in tree and soil damage due to skidding in the FSC site relative to the CL site were due mostly to skid trails being narrowed, presumably due to pre-felling skid trail planning, training, and supervision of skidder drivers at the FSC site (Table 7; Fig. 6). While there was no difference between the CL and FSC concessions in skid trail length per harvested tree, because skid trails in the CL area were 40 % wider, they covered 38 and 36 % more ground surface per tree and per cubic meter harvested, respectively. The lack of a difference in the number of avoidable stream crossings per unit area in the two concessions (Table 6; Fig. 6) may indicate that the company failed to demarcate and protect streamside buffer zones in the FSC site, as called for in the 2008 and 2011 audit reports by Bureau Veritas (Forest Certification System 2011; Système de certification 2011).

It is also important to note that although skidder drivers at the CL site received no formal training in RIL and had no maps of planned skid trails to follow, the tree finders that preceded them marked paths to the harvestable trees that avoided steep slopes, large trees, and otherwise facilitated log yarding. Perhaps these simple measures reduced the differences in biomass of trees damaged and destroyed by skidding when expressed on the basis of the volume of timber harvested.

Damage Due to Road Construction and Maintenance

Compared to roads in the FSC concession, roads in the CL concession were wider, longer, and caused more damage whether expressed per unit area, per number of trees harvested, or per harvested timber volumes (Table 6). The roadside clearings in the CL site that resulted in huge collateral stand damage were presumably opened to facilitate the drying of the road surface. Trafficability of the relatively narrow roads in the certified concession, in contrast, was maintained by assuring proper drainage and surfacing with lateritic gravel, as recommended by the FAO (2003).

Implications for Forest Management and Carbon Retention

Damage reduction in forest is an important step toward reduction in carbon emissions from forests, as part of REDD + mechanisms. Such procedures require the implementation of improved forest management techniques that deems a pre-requisite for forest certification. Certification is believed to have already improved tropical forest management, and it is expected that certified forests retain and sequester more amount of carbon and support greater biodiversity than uncertified forests (Putz and Nasi 2009).

The expected biophysical benefits of forest certification derive mostly from minimizing residual stand damage and protecting fragile areas in forest management units. Damage reduction results in reductions in carbon emissions, which contribute to climate change mitigation. The use of improved forest management techniques under FSC standards, therefore, constitutes a mechanism to reduce emissions from tropical production forests while preserving ecological integrity. In recognition of this synergy, achieving and maintaining FSC certification should make forest management units eligible for payments under a REDD + scheme that recognizes the roles of conservation, sustainable management of forests, and enhancement of forest carbon stocks.

If forest certification is considered as an indication of SFM (Cerutti and others 2011) and with the increased interest in the climate change mitigation potential of improved forest management (Putz and others 2008b), FSC criteria and indicators should be refined to capture the impacts of logging on AGB. Given that the principal challenge for certification in the Congo Basin is the cost of getting and remaining certified, incentives from REDD + programs should be provided to help cover these costs. For the Congo Basin region, where timber industries are socio-economically important and intact forests help regulate the climate, protect biodiversity, and provide cultural services, governments need to reinforce forestry laws necessary for sustainable forest management.

Conclusions

Recognition of the benefits of improved forest management in stand conditions, as indicated by credible certification, and the associated reductions in carbon emissions should lead to broader adoption of improved logging practices. Unfortunately, field-based evidence for these connections is sparse partially due to the difficulty in implementing proper experimental or quasi-experimental designs with sufficient replication. This study suffers from lack of replication but insights about sampling methods, nevertheless, can be derived from its case-study approach with a qualitatively selected counterfactual and a paired design. This comparison of the short-term environmental impacts of selective logging also provides some basis for the contention that forest management certification leads to reduced stand damage and reduced carbon emissions, but the differences were not large and varied with how the data were presented. The small RIL benefit compared to that reported from elsewhere in the tropics (reviewed by Putz and others 2012) is likely related to the low logging intensities and the specific conventional logging practices employed. In particular, the marking of extraction routes on the ground by tree finders may deliver most of the benefits expected from preparation of detailed harvest plans. The training that CL fellers received is also worthy of note as a probable spillover effect of certification.

Many of the stand damage reduction benefits ascribed to FSC certification derived from the opening of narrower haul roads. We believe this attribution of impact is correct, but recognize that the topic of road planning and construction is complex and needs to be studied at very large spatial and temporal scales. For example, analyses at larger scales would reveal whether the roads are likely to be used in the future, which influences their construction. Also, opening wide swaths of forest does serve to speed drying of road surfaces, which may be the preferred technique if there are no nearby sources of road gravel; this tradeoff deserves further attention when management plans are developed. Unfortunately, none of the other RIL vs. CL studies in the literature adequately consider the economic and engineering issues related to roads (Medjibe and Putz 2012).

Logging damage varies with logging intensity, harvesting techniques, and whether the data are expressed per cubic meter of timber harvested or on the basis of area. Although both modes of expression have different values, per area values vary with exactly how overall logging areas are measured. In particular, if areas that are not loggable due to steep terrain as well as areas in which logging is prohibited (e.g., due to proximity to water bodies) are not considered to be part of the cutting blocks, which is how it should be, then logging damage appears more severe. Expressing logging damage per cubic meter harvested has advantages insofar as it avoids this problem and accounts for vast differences in tree sizes and logging intensities.

Future research on the impacts of logging should be designed in ways that avoid the inefficiencies and logical limitations encountered in this study, particularly when logging regimes are to be compared. In particular, avoidance of the pseudo replication problem will require more efficient sampling of logging areas so that less effort is expended in areas that end up not being directly affected by logging. Randomly located permanent plots are suitable for many purposes (e.g., for monitoring post-logging forest dynamics), but miss a lot of the logging activities especially where harvesting intensities are low and spatially variable. On the positive side, the sampled plots, not subjected to logging during this particular cycle, can be used as a baseline in a properly designed before-and-after intervention study if they are eventually logged.

Despite its shortcomings, this study constitutes one of the first examples of a field-based study comparing some of the ecological impacts of certified with uncertified forestry in the tropics. As such, it provides some of the data needed to inform decision-making about investments in forest management and certification and serves to illustrate some impediments to proper impact evaluation.