For this study a total of 165 1-ha plots were selected from the Network of Permanent Plots in lowland Bolivia (Fig. 1). These plots were established in old-growth forests by various projects and forestry concessionaries (see Acknowledgements for more details). To date this network and its database has been coordinated and managed by the Instituto Boliviano de Investigación Forestal (IBIF). The plots are located between 10–18° S and 59–69° W, in upland forests (terra firme; the 5% of plots found in areas of seasonal flooding were also included in the analysis), generally on flat terrain (20% on sloping ground in hilly areas), and in an altitude range from 100 to500 m asl. These plots are distributed over the main environmental gradients of climate and soil and 52% of them have been affected by logging.
Figure 1. Variation in (a) average diameter (DGRavg) and (b) stand basal area (BAGRstand) growth rates of 165 permanent plots located in four departments (Pando, La Paz, Beni and Santa Cruz) of lowland Bolivia. The size of the symbols scales proportional with the growth rate. Potential forest cover of areas assigned to timber production is indicated in grey. The white areas in Pando pertain to floodplains, in Beni and north of La Paz to savannas, and in Santa Cruz to the Cerrado and Chaco vegetation.
Download figure to PowerPoint
Lowland Bolivia is characterized by two rainfall gradients: a south–north gradient where rainfall increases towards the equator with mean annual precipitation ranging from 1100 to 1900 mm and an east–west gradient where mean rainfall increases from 1600 to 2200 mm due to orographic uplift towards the Andes. However, the precipitation in individual years can vary from 600 to 3000 mm per year from the driest to wettest areas (based on at least 30 years data, 1970–2007, Servicio Nacional de Meteorología e Hidrología-SENAMHI, unpublished data). In general, the lowlands in Bolivia experience a 4–7-month dry season (with precipitation <100 mm per month) from about April to September, corresponding to the austral winter. Mean annual temperature is between 24.2 and 26.4 °C. Additionally, lowland Bolivia is characterized by differences in geomorphology and geological history (Montes de Oca 1997; Suárez-Soruco 2000), leading to strong gradients in soil characteristics. The soils vary considerably in fertility, from acid Acrisols in the Amazon forest in the north, via Acrisols and Luvisols in the centre, to Cambisols and Arenosols in the south (Gerold 2003).
Data Collection and Analysis
The 165 selected plots were all established between 1995 and 2005 and measurement periods varied between 2 and 11 years, with the last measurements taking place in 2007. Plots were typically square (100 × 100 m), with only 11 of them being rectangular (20 × 500 m). In each plot, every tree ≥10 cm diameter at breast height (DBH; measured at 130 cm, or higher when buttresses were present) was measured with diameter tape, painted at the measurement point, tagged and identified, following standard protocols (Alder & Synnott 1992; Contreras et al. 1999). Re-censuses were mostly carried out in the same season or month as the plots were established, thus minimizing the effect of intra-annual variation in DBH change. In most of the total of 85 plots that were affected by logging activities, logging started immediately after their establishment; 80 plots were not logged.
The annual diameter growth per individual was calculated as: (Df – Di)/t where Df is the final diameter and Di is the initial diameter at the start of the interval. Based on this diameter growth rate (hereafter DGR) we calculated five variables per plot, representing the growth rate at the individual level: average (DGRavg), median (DGR50), 90th (DGR90), 95th (DGR95) and 99th percentile (DGR99) of annual diameter growth. Values for the 90th and 95th diameter growth rate percentile are not included in the results because they were highly correlated with the DGRavg, DGR99 and between themselves (see Table S1 in Supporting Information). The DGR50 and DGR99 were calculated to provide information on median and upper levels of growth rate. Additionally, we calculated the basal area growth rate at the stand level (hereafter BAGRstand) as the net yearly basal area change per plot. The BAGRstand was calculated as: (BAf – BAi)/t, where BAf is the final total plot basal area and BAi is the plot basal area at the start of the measurement interval (for control plots) or just after logging (for logged plots). In both formulae, t is the time, in years, between the two measurement dates. Note that BAGRstand includes the effects of growth, recruitment and mortality, while DGR is based upon individuals that survived the whole monitoring period.
For each plot 20 soil samples were collected from the first 30 cm of soil with an auger, and a pooled sample of 500 g was analysed within a week after collection at the Center of Tropical Agricultural Research (CIAT-Santa Cruz, Bolivia). The analyses included 12 edaphic variables: percentage of clay, silt and sand, measured with the Bouyoucos hydrometer; exchangeable Ca, Mg, Na, K (in 1 M ammonium acetate at pH 7); cation exchange capacity (CEC, sum of cations plus acidity); acidity (in 1 M KCl); plant available phosphorus (P, Olsen method); organic matter (OM, Walkley-Black method) and total nitrogen (N, micro-Kjeldahl method). For each plot we obtained five climatic variables, interpolated from available data from 45 weather stations in the region, and 12 edaphic variables obtained from sampled soils. To summarize these environmental variables we performed two independent Principal Component Analyses, (PCAs, see Table S2 in Supporting information). The PCA was done for 220 1-ha plots that are part of the Network of Permanent Plots in lowland Bolivia, and included the 165 plots that are analysed here for their dynamics. The climatic PCA considered annual temperature, annual precipitation, precipitation of the three driest months, length of the dry period (# months <100 mm), and length of the drought period (# months <50 mm). The first two axes of the climatic PCA explained 94% of the variation. The first axis (65%) correlated positively with annual precipitation and negatively with dry period length (henceforth referred to as the rainfall axis). The second axis (29%) correlated positively with mean annual temperature and negatively with the precipitation of the driest months (hereafter temperature axis). The edaphic PCA considered the 12 edaphic variables. The first two axes of the edaphic PCA explained 68% of the variation. The first axis (48%) correlated positively with variables related with soil fertility (CEC, Ca, Mg, Na, K, P, OM and N), and negatively with acidity (hereafter soil fertility axis). The second axis (20%) represented variation in soil texture and correlated positively with clay and silt and negatively with sand (hereafter soil texture axis) (Table S2).
Four logging-related variables were used to describe forest disturbance in each plot. Two dummy variables were created: Logging Presence (LP) to describe whether logging occurred (1) or not (0) in the plots and Logging Impact (LI), which describes whether the impact was high (1) or low (0) in logged plots (based on the number and location of logged trees and number of additional trees that died due to logging operations). Other continuous variables representing logging disturbance were the Logged Basal Area (LBA, in m2 ha−1, based upon the number and diameter of the trees logged) and the Time After Logging (TAL, in years).
The four growth variables were first correlated (Pearson correlation) with the individual environmental variables to evaluate what components of these composite axes were most important. Each of the four growth rate variables were subsequently regressed on the four main environmental axes and the four disturbance variables (including interactions and their quadratic terms, when necessary) using a series of multiple backward regressions. We used the PCA axes for this regression analysis, rather than the original individual environmental variables, to avoid problems with multicollinearity and over fitting. Quadratic terms were included in the models because non-linear relationships between growth rates and predictors were observed in scatter plots. Because the increases in explained variation as a result of including interaction effects in the models were very low, we re-ran the analyses without the interactions and present only the latter results. For each variable the Kolmogorov–Smirnov test for normal distribution was applied and, if necessary, the data were logarithmic (log10), square root or arcsine-transformed. All statistical analyses were performed with SPSS 15.0 for Windows (SPSS Inc., Chicago, IL, USA).