Landscape-scale variation in structure and biomass of Amazonian seasonally flooded and unflooded forests
2.2. Methods 1. Study area
2.3.1. Stem identification and wood density
We sampled a total of 12,721 stems ≥ 10cm DBH (terra firme - TF: 6,389; várzea - VZ:
6,332) across the 200 plots (= 20 ha), from 191 genera in 55 families (TF: 152 genera, 50 families; VZ 126 genera, 44 families). Despite low levels (18.4%) of identification to species (TF: 9.9%; VZ: 26.9%), over three quarters (78.1%) of all stems were
successfully identified to at least the level of genus (TF: 79.9%; VZ: 76.3%) and we unambiguously identified 97.6% of all stems to at least the family level (TF: 96.9%; VZ:
98.4%). The WSG values assigned to each stem showed that the variation in wood density was significantly lower within genera than between genera (ANOVAs: GWDD F573,1430 = 8.85, p < 0.001, Jari F186,210 = 1.92, p < 0.001, Mamirauá F107,18 = 4.35, p <
0.001).
39 Figure 2.2. Density distribution of (a) stem density (stems ha-1), (b) forest basal area (m2 ha-1) and (c) aboveground forest biomass (Mg ha-1) for terra firme (white curve) and várzea (black curve) forests.
2.3.2. Forest structure and biomass
Stem density was similar in terra firme and várzea forests with both forest types dominated by smaller stems (Appendix 2.3) although large emergent trees (> 100cm DBH) had a disproportionately large influence on plot basal area, particularly in várzea forests (Figure 2.2). As a result, mean plot basal area was greater and more variable in várzea than in terra firme forest (Figure 2.3; TF mean ± SE: 32.4 ± 0.9 m2 ha-1; VZ:
40 37.6 ± 1.2 m2 ha-1; t-test: t = -3.411, p < 0.001). AGB estimated using the simplest allometric equation, based on DBH only, was similar across forest types. However, mean WSG per plot was significantly lower in várzea forest (TF: 0.67 ± 0.003 g cm-3; VZ: 0.58 ± 0.003 g cm-3; t = 20.085, p < 0.001), where canopy height rarely exceeded 30 m. Employing more complex allometric equations incorporating both WSG and tree height significantly lowered AGB estimates for várzea forest plots compared to those in terra firme (TF: 358.4 ± 14.4 Mg ha-1; VZ: 281.9 ± 12.0 Mg ha-1, t = 4.077, p < 0.001).
The relative difference between forest types was even more apparent when considering the structural conversion factor (TF: 10.7 ± 0.2 Mg m-2 basal area; VZ: 7.3 ± 0.1 Mg m-2 basal area; t = 18.154, p < 0.001), reinforcing the notion that várzea sites were
predominantly comprised of light-wooded tree species.
Figure 2.3. Mean values per forest plot of (a) stem density (stems ha-1), (b) forest basal area (m2 ha-1), (c) wood specific gravity (g cm-3), (d) aboveground biomass (Mg ha-1) from DBH-only equation, (e) aboveground biomass (Mg ha-1) from equations also including wood specific gravity and tree height, and (f) the structural conversion factor (Mg m-2 basal area) for terra firme (open boxes) and várzea forests (solid boxes).
Horizontal bars indicate medians; boxes indicate interquartile ranges; whiskers indicate minimum and maximum values; and circles indicate outliers (observations 1.5 times higher or lower than 1st and 3rd quartile, respectively).
41 The AGB values for plots in both terra firme and várzea forest were significantly
positively related to basal area (TF: R2 = 0.92, p < 0.001; VZ: R2 = 0.88, p < 0.001) and, to a lesser degree, to plot-scale WSG (TF: R2 = 0.06, p = 0.009; VZ: R2 = 0.09, p = 0.002) (Figure 2.4). There was also a significant positive relationship between AGB and stem density in terra firme forests but not in várzea forests (TF: R2 = 0.05, p = 0.013;
VZ: R2 = 0.01, p = 0.170), and between basal area and stem density in both forest types (TF: R2 = 0.17, p < 0.001; VZ: R2 = 0.04, p = 0.026) (Figure 2.4). WSG, however, was unrelated to both stem density (TF: R2 = 0.0006, p = 0.31; VZ: R2 = 0.003, p = 0.26) and basal area (TF: R2 = 0.011, p = 0.15; VZ: R2 = 0.015, p = 0.12) (Figure 2.4.)
Figure 2.4. Pairwise relationships between plot-scale mean wood specific gravity (WSG, g cm-3), stem density (SD, stems ha-1), basal area (BA, m2 ha-1), and
aboveground biomass (AGB, Mg ha-1) for 200 forest plots in terra firme (open circles, dashed line) and várzea forests (solid circles, solid line). Lines represent linear models;
grey shading represents 95% confidence intervals.
42 2.3.3. Landscape predictors of AGB
2.3.3.1. Water stress
The top-ranking model predicting AGB across both forest types on the basis of
landscape variables had a low Akaike weight of 0.24 (Table 2.3) suggesting uncertainty in identifying a single best model and supporting the adoption of a multi-model
approach. Twelve alternative models comprised the 95% set of models (cumulative ωi ≥ 0.95). The single best model contained only the variable forest type, which appeared in 10 of the 12 models with a cumulative Akaike weight of 0.86, confirming the lower aboveground biomass values across várzea forest compared to terra firme forest. The next most important variable was terrain elevation, with a positive influence on AGB across all plots, although this is mostly explained by elevation differences between forest types (Table 2.3, Figure 2.5). We therefore examined the potential effects of elevation and other landscape variables further within forest types by constructing models in the same fashion for terra firme and várzea forest separately.
Indicators of water stress had contrasting influences in each forest type, and greater importance in várzea forest. Elevation had a weak positive effect on AGB in terra firme, in contrast to a strong positive effect of greater flood duration in várzea forest (Table 2.3, Figure 2.5). The positive effect of distance to the nearest stream on AGB in terra firme is at odds with the negative effect in várzea forest, but low cumulative Akaike weights in each case show the low relative importance of this variable in the models (Table 2.3, Appendix 2.4). Most strikingly, flood duration (on the basis of ALOS ScanSAR flood mapping) had a positive effect on AGB in várzea forest.
2.3.3.2. Logging accessibility
In addition to water stress, many of the 95% set of models for each forest type contained the variables describing the historical accessibility of forest to selective timber
extractors. Distance to the nearest community was particularly prominent across models and notably was positively related to AGB in várzea forest, in contrast to a negative relationship in terra firme forest (Table 2.3, Appendix 2.4).
43
Table 2.3. Summary of multi-level mixed effects models of mean aboveground forest biomass within 200 biomass plots in both terra firme (TF) and várzea (VZ) forests, and for each forest type separately. All top-ranking models within 95% of the cumulative Akaike weight (ωi) are shown. Variables included in each model are shaded grey. Model averaged Akaike weights for each variable are shown in the first line.
No. models in 95% set
Model no.
Water stress Logging accessibility IC ∆IC ωi
Intercept Forest
44 Figure 2.5. Relationships between SRTM-measured elevation (m) and
ScanSAR-measured flood duration (months) with (a) wood specific gravity (WSG, g cm-3) and (b) aboveground biomass (AGB, Mg ha-1) for 200 forest plots in terra firme (open circles, dashed lines) and várzea forests (solid circles, solid lines). Curves represent smoothed means; grey shading represents 95% confidence intervals.