This research took place in the Mbali River Local Communities Forest Concession, situated in the Bolobo territory of the Democratic Republic of Congo ([45]; [32]). This is an area managed for conservation by local communities, with leadership provided by the Congolese non-governmental organization Mbou-Mon-Tour ([18]; [32]) as shown in Figure 1. The study site is located at the far southwestern edge of the bonobo’s natural habitat and is the area nearest to the capital city, Kinshasa ([20]; [32]). The landscape in the Bolobo territory is a mixture of forest and savanna, approximately 60% tropical rainforest and 40% savanna ([46]; [32]). The rainforest areas include several types: forests dominated by Marantaceae plants, diverse mature forests, previously disturbed forests, and forests that are temporarily flooded ([48]). The savanna areas consist of grasslands and areas with scattered trees ([44]). The total rainfall for one year (May 2012 to May 2013) was 2387 millimeters, featuring a three-month dry season from June to August where monthly rainfall was less than 100 millimeters (This Study; [32]). The population density in the region is very low, with fewer than 5 people per square kilometer, and all villages are situated within the savanna areas ([49]). Five small traditional farms, each supporting one or two families, were located on the border of the Manzano Forest (This Study). The local Teke people follow a cultural prohibition against harming bonobos, whom they regard as being nearly human ([45]; [32]). Additionally, the Teke people tend to avoid the animals themselves, seeing a forest encounter as a bad sign, and they also steer clear of bonobo droppings ([18]; [32]).

Figure 1. Fallows identified at Nkala and Embirrima forests (Source: Réalisée par Wendy Mbombenga).

The Teke relies heavily on the forest for their subsistence, entering it nearly daily to practice shifting cultivation (growing crops like cassava and corn), to hunt, to fish, and to gather forest products ([45]; [32]). Starting in 2001, the village of Nkala, through its community leadership, chose to set aside a section of forest specifically to protect bonobos ([18]; [32]). A number of other villages later adopted similar conservation measures as Embirrima which forest is the most studied ([45]; [32]). In the forest patches included in this research, activities like hunting and farming are banned, but people are permitted to be present in the area for NTFP collection (This Study; [32]).

From the six villages in the forest concession, Nkala and Embirrima were chosen because of their seniority. In fact, they have been the first to be given for forest conservation to preserve Bonobo’s habitat. Fallow ages were determined during interviews with local former agriculture who used the land. To confirm their estimation, we selected randomly local peoples to have accuracy of the different fallow ages given. We selected randomly 10 fallows of at least 10 years hold in each forest concessions where information was accurate. In all, 20 plots that had been selected. We sat up fallows ages class ranged between 10 years of difference to meet tropical forest evolution without taking in count the number within each class.

For each selected fallow plot, a square study plot measuring 50 m/50 m was set up. Inside each plot, every tree was identified. Following standard scientific protocol, each tree with DBH ≥ 10 cm was measured. We also calculated basal area (ba), plant density (D) and estimated the heights.

For every plot, we recorded the elevation and geographic coordinates using a GPS device, and measured the slope direction with a compass. The above-ground biomass (AGB) for each tree was calculated using a standard mathematical formula from Chave et al., with other formulas used for comparison. To convert the biomass weight into carbon weight, we used the standard conversion factor of 0.47, as recommended by the IPCC. The wood density values needed for the calculations were obtained from existing scientific databases.

All statistical tests were performed using the R software program, version 3.3.1. Before running the tests, we checked that the data met the necessary statistical requirements for normality and equal variance. The fallow plots were sorted into six groups based on their age since abandonment: 10 - 19 years, 20 - 29 years, 30 - 39 years, 40 - 49 years, 50 - 59 years, and 60 years or more.

For each tree with a diameter at breast height (DBH) ≥ 10 cm, aboveground biomass (AGB) was computed at the individual level using the pantropical allometric model of [6], which was selected for its extensive validation across tropical forest types and its reliance on DBH, wood density, and height-parameters well-suited to the heterogeneous regrowth forests of the study area. Tree-level biomass values were then summed for each 50 × 50 m plot and scaled to a per-hectare basis by multiplying by four (the number of plots per hectare). For species with missing wood density values, the genus or familylevel average from the Global Wood Density Database was assigned to minimize estimation bias.

2,578 individuals across 170 species, 126 genera, and 59 families (APGIII) were collected in Nkala and Embirrima forest concessions of local communities. The Fabaceae group was the most represented family, accounting for 12% (with Mimosoideae at 3%, Faboideae at 4%, and Caesalpinioideae at 5%), followed by Annonaceae and Apocynaceae at 8% each, Euphorbiaceae at 7%, and Rubiaceae at 5%. The remaining families showed low representation. Nkala had a higher number of taxa (135) compared to Embirrima (114), though the difference between the two sites was not statistically significant (p-value 0.227 > 0.005).

According to Table 1, which presents spatial variation in biodiversity indices for the pre-forest fallows, the Simpson_1-D index indicates high species diversity at both sites, with values near 1 reflecting low dominance. Contrary to earlier findings, the Shannon index is slightly higher at Nkala (H' = 4.03) than at Embirrima (H' = 3.88), suggesting marginally greater diversity at Nkala, while Simpson values are nearly identical (0.9593 and 0.9599), confirming comparable overall diversity. Evenness (J') is around 0.82 for both sites, pointing to a balanced distribution of species abundances.

Table 1.Biodiversity indices fallow sites.

Figure 2 shows a right-skewed frequency distribution of tree heights, indicating a young or regenerating stand. The majority of individuals fall within the lower height classes [10 - 15] and [15 - 20], with frequency declining sharply as height increases from [20 - 25].

Figure 2. Distribution of individual trees across pre-defined height classes.

Figure 3 illustrates the increase in stand basal area with fallow age, showing a rapid accumulation of biomass in the first two decades of succession, followed by a gradual approach to an asymptotic maximum in older fallows.

Figure 3. Basal area recovery in secondary forest succession.

The distribution exhibits a strong reverse-J profile, a hallmark of active regeneration in tropical secondary forests. The majority of individuals are concentrated in the smallest diameter classes, with a pronounced peak in the 10 - 20 cm class (n = 350 stems/ha), reflecting high recruitment rates of pioneer and early-successional species following land abandonment. Stem frequency declines exponentially with increasing diameter, consistent with size-dependent mortality and thinning processes typical of tropical forest succession. The presence of trees in the largest classes (90 - 130 cm) indicates residual or emergent individuals from previous forest cycles, which play a critical role in seed dispersal, structural complexity, and rapid biomass recovery (Figure 4).

Figure 4. Size-class distribution of trees in the study area based on DBH.

This multi-panel figure (Figure 5) provides a direct graphical comparison of four published allometric models used to predict tree Aboveground Biomass (AGB) from DBH. The four allometric models compared ([10]; [33]; [6]) show a strong linear relationship between diameter at breast height (DBH, in cm) and aboveground biomass (AGB, in t/ha), as evidenced by high coefficients of determination (R²) (0.81 - 0.84) and correlations (r ≈ 0.91). These results confirm the relevance of using diameter as a robust predictor of biomass in the studied forest ecosystems. However, the structural parameters of the models vary considerably: the slope (a) ranges from 0.050 to 0.081, and the scaling exponent (b) from 0.78 to 1.4. These differences likely reflect variations in the sampled stands (species composition, site conditions, disturbance history) or in the modeling approaches. For example, the second model of Ngamanda, with the steepest slope and highest exponent b, may have been calibrated for larger-diameter trees or species with higher wood density.

Figure 5. AGB comparison of four allometric equations per diameter.

Unlike diameter-biomass models, the relationships between stand age (in years) and aboveground biomass (AGB, in t/ha) presented here show a lack of significant correlation (Figure 6). Indeed, the coefficients of determination (R²) are extremely low (0.01 to 0.09), indicating that age explains less than 10% of the variability in biomass. The correlation coefficients (r) are also close to zero and slightly negative (from −0.09 to −0.1), suggesting a negligible tendency for biomass to decrease with age in the datasets considered.

Figure 6.ABG variation by fallow ages.

The data points in each panel are widely scattered, showing high variability in biomass for stands of similar age. For instance, a 10-year-old fallow may be predicted to hold between 20 and 80 t/ha of AGB depending on the model and specific site conditions, and this range overlaps substantially with predictions for 20- or even 30-year-old fallows. This visual scatter is corroborated by quantitative metrics. The coefficients of determination (R²) for linear regressions between age and model-predicted biomass are exceptionally low, ranging from approximately 0.01 to 0.09. Statistically, this indicates that fallow age explains less than 10% of the total observed variance in AGB. The corresponding correlation coefficients (r) are marginally negative but effectively zero (≈−0.1), dismissing any meaningful linear trend—positive or negative—between these variables. Age-AGB relationships were tested using linear regression after visual inspection of scatterplots. No significant quadratic or logarithmic terms improved model fit. Residuals for all age-AGB models met normality assumptions (Shapiro-Wilk test, p > 0.05 for all); and Breusch-Pagan tests confirmed constant variance (p > 0.05 for all) (Figure 7).

Figure 7.Variation of AGB in different fallows ages according 4 equations.

The pronounced reverse-J diameter distribution (Figure 4) and right-skewed height distribution (Figure 2) are classic structural hallmarks of uneven-aged, regenerating tropical forests ([12]; [9]). The high density of stems in the smallest size classes (e.g., 350 stems/ha in the 10 - 20 cm class) indicates vigorous and continuous recruitment, a pattern consistently observed in Neotropical secondary forests ([38]; [34]). This structure is maintained by high rates of ingrowth from pioneer and early-successional species, which rapidly colonize abandoned fields ([8]). The presence of large residual trees (>90 cm DBH) within the fallows is particularly significant. These individuals, likely remnants from the pre-clearance forest, act as “biological legacies” that accelerate succession by providing seed sources, creating favorable microsites, and enhancing structural complexity ([7]; [17]). Their contribution to rapid basal area recovery in the first two decades (Figure 3) aligns with studies demonstrating the importance of remnant trees for biomass accumulation rates in tropical landscapes ([26]).

A central finding of this study is the stark contrast between the strong predictive power of tree diameter for biomass and the weak predictive power of fallow age. The four allometric models tested ([10]; [6]; [33]) all showed strong correlations between DBH and AGB (R²: 0.81 - 0.84; Figure 5), confirming that diameter is a robust, direct predictor of tree biomass, as established in pantropical allometry ([5]). However, the substantial variation in model parameters (e.g., exponent *b* ranging from 0.78 to 1.4) underscores a critical methodological consideration: model selection introduces significant uncertainty into carbon stock estimates ([40]; [47]). The second model [33], with its higher exponent, predicts substantially more biomass for large trees, highlighting how calibration datasets influence outputs as shown in the annexe.

In stark contrast, fallow age explained less than 10% of the variance in AGB (R²: 0.01 - 0.09; Figures 6-7). The near-zero correlation contradicts the common assumption of a simple, deterministic increase in biomass with time in chronosequence studies ([22]). This result strongly suggests that time since abandonment is a poor proxy for ecosystem recovery in this landscape. Similar weak age-biomass relationships have been found in other heterogeneous tropical landscapes, where prior land-use intensity, soil conditions, and stochastic recruitment events override the effect of age ([31]; [39]). The rapid initial rise in basal area (Figure 3) likely reflects the fast growth of a few early colonists and residual trees, but subsequent accumulation becomes highly idiosyncratic, dependent on local contingencies.

These results show that age alone is an unreliable predictor of AGB in these post-AIB fallows, and that other factors (e.g., past land-use intensity, soil conditions, species composition) likely drive biomass accumulation.

These findings have direct implications for forest carbon projects and policies like REDD+ (Reducing Emissions from Deforestation and Forest Degradation). The high structural variability and weak age-biomass correlation mean that carbon sequestration potential cannot be reliably estimated from fallow age alone. This challenges the feasibility of simple “carbon-for-age” payment schemes in similar complex landscapes. Effective carbon accounting must be based on direct measurement of forest structure through inventories, not assumed recovery curves ([16]; [37]).

Furthermore, the observed structural patterns—continuous recruitment alongside large residuals—indicate that these fallows are not merely carbon sinks but are developing complex, biodiverse ecosystems. This aligns with the concept of “forest transition” where fallows can evolve into high-value secondary forests if allowed sufficient time and protection ([9]). For the Teke communities of Bolobo, this suggests that protecting fallows beyond the typical rotational cycle could yield significant co-benefits: enhanced biodiversity, improved ecosystem services, and access to carbon finance, provided robust measurement protocols are used ([1]).

This study’s chronosequence approach, while practical, inherently assumes space-for-time substitution, which can be confounded by spatial variability in environmental factors ([22]). The limited number of plots per age class and the focus on fallows >10 years old also restrict analysis of the crucial early successional stages. Future research should incorporate direct measurements of prior land-use intensity, soil nutrients, and seed rain to disentangle the drivers of recovery variability ([21]). Longitudinal monitoring of permanent plots would provide the most accurate data on carbon accumulation trajectories. Additionally, integrating remote sensing with field-validated allometric models could enable cost-effective scaling of carbon stocks across the concession ([2]).