Woody biomass in tropical trees contributes significantly to global carbon stocks; however, these stocks are increasingly affected by climate and land-use changes. Understanding the growth mechanisms driving woody biomass production is essential for assessing the short- and long-term contributions to carbon stocks and dynamics in tropical forests. Trees accumulate biomass by increasing their size (wood volume) and/or tissue density (wood density). However, estimates of tree biomass production are often based solely on size increment through measurements of stem diameter growth, overlooking the potential spatial and temporal variation in wood density within trees. Tree-ring analysis can be applied to reconstruct past tree volume growth and wood density variations, allowing the quantification of their relative contributions when reconstructing past woody biomass production. Here, we studied trees of the widespread Neotropical genus Cedrela (C. fissilis, C. odorata, and C. montana) along an environmental (climate and soil) gradient to address two key questions: 1) How does temporal variation in tree diameter growth and wood density affect biomass production? 2) To what extent do these relationships vary along the environmental gradient? We examined long-term (ontogenetic, variation from pit to bark) and short-term (annual, interannual variation) variations in diameter growth and wood density, covering eighteen sites in the Amazon rainforest, Atlantic Forest, Cerrado savanna, and Caatinga dry forest. We found that diameter growth and wood density drive short- and long-term biomass production dynamics. Interestingly, diameter growth patterns predominantly explained short-term variability in biomass production at all sites, whereas wood density explained ontogenetic biomass patterns mainly at humid sites. These results highlight the importance of accounting for both short- and long-term variation, including climatic and ontogenetic drivers, to increase the accuracy of biomass estimations in tropical trees, particularly in humid forest ecosystems such as the Amazon. 

Diameter growth is an important and good indicator of forest carbon production. However, size-related changes in wood density, which are usually neglected, are critical for accurate short- and long-term carbon assessments, especially in tropical humid sites.

Sample collection

Samples were collected from eighteen forest sites across Brazil and Peru, covering a range of forest types and environmental characteristics. The study sites included national, state, and municipal parks and private forests located in tropical and subtropical regions. These sites represent ecosystems from the Amazon Rainforest, Atlantic Forest, Brazilian Cerrado savanna, and Caatinga dry forest. Samples from sites in the Amazon were collected in sustainable-use conservation units (National and State Forests), where legal forest management operations are conducted. In these areas, samples were obtained by collecting cross-sectional wedges from felled trees as part of authorized logging activities. In the remaining areas, samples were collected from state, municipal, and private forests where forest management is not legally authorized. In these cases, non-destructive coring methods were used to extract wood samples. Each population was sampled in different years for various dendrochronological applications. We included only samples that spanned from the pith to the bark, contained the pith, or were close enough to estimate the distance to the pith. At each location, the number of trees we used ranged from eight to eighteen, due to the availability of samples with complete pith and sufficient material for wood density analysis.

Sample preparation and wood density analysis

We selected one radius from each tree to perform the analysis. All the cores were previously polished and scanned at high resolution (1,200-2,400 dpi), and tree rings were marked following the marginal parenchyma bands associated with semi-porous or porous xylem. To measure the growth rate and wood density, cores were sectioned transversely into radial sections of 0.5-2 cm wide and 1.5 mm thick, while wedges were cut into sections of 2 cm wide and 1.5 mm thick. Wood density was determined following the protocol established by Quintilhan et al. (2021) and Tomazello et al. (2009). Radial sections were conditioned in a climatic chamber at 20°C and 60% relative humidity until they reached a stable moisture content of 12%. All samples were scanned in an X-ray densitometry chamber (Faxitron X-Ray, Illinois, USA) equipped with a measurement unit scale and a cellulose acetate calibration wedge for wood density calibration. Digital X-ray images of the radial section samples (96 dpi, .tiff) were analyzed with WinDendro Density 2017a® software (Regent Instruments Inc., Canada) to obtain the tree-ring width (TRW) and tree-ring wood density at 12% moisture content (WD_U12%) per year. Sample preparation, X-ray imaging, and wood density measurements were performed at the Tree-Ring and Wood Anatomy Laboratory of the University of São Paulo (Piracicaba, São Paulo, Brazil). Data processing and data analysis were conducted at the Dendroecology and Wood Biology Laboratory at the State University of São Paulo (Campinas, São Paulo, Brazil).

Diameter reconstruction and biomass estimation

For each tree, the cumulative diameter (D) and cumulative basal area (BA) during the year of ring formation were reconstructed by summing the previous tree-ring widths, assuming a circular growth pattern. For tree-ring samples without a visible pith, we estimated the distance based on the concentric circle method, using CDendro and CooRecorder® software (Cybis Electronic, 2013).

To calculate the aboveground biomass production per year, we first reconstructed the total historical aboveground biomass per year (B) using the pantropical equation described by Chave et al. (2014) (equation 1). Not all sites had tree height data available; therefore, we used the pantropical equation that included a diameter–height allometric model. This model has a stress factor (E, Table S2) that represents the environmental conditions of the site where the trees grow. Chave et al. (2014) demonstrated that factor E is strongly dependent on the diameter-height allometry relationship and is particularly useful for improving model accuracy in the absence of tree height data.

For each tree and year, biomass was estimated based on D and a tree-level weighted average wood density (WD_WA). The WD_WA value was obtained from the weighted average of the WD_U12% values by the basal area increment of each year (Equation 2). This calculation was designed to prevent the influence of the wood proportion near the pith (Williamson and Wiemann, 2010). Before weighing, WD_U12% was transformed into basic wood density (WD) by multiplying WD_U12% by 0.838 (Vieilledent et al., 2018). The absolute ABP was obtained by subtracting the total biomass from the present year minus the total biomass of the previous year (ABP = B_t-B_t-1). We also calculated the relative change in biomass per year by dividing the ABP from the present year by the total biomass from the previous year (ABP_P = ABP_t/B_t-1) to reduce the size-ABP relationship bias.

Equation 1:

B  = exp[-18.03 – 0.976E + 0.976 ln(WD_WA) + 2.673 ln(D) – 0.0299 [ln(D)²]

Where B is the historical aboveground biomass (kg); D is the total diameter of each tree per year reconstructed by the accumulated tree-ring width increment (cm); WD_WA is the average basic wood density-weighted by basal area increment (g/cm³); E is the environmental stress factor unitless for each site obtained from the raster file available in Chave et al. (2014). 

Equation 2:                                                                                       

WD_WA  = (BAIt * WDt)/BAt

Where WD_WA  is the weighted average wood density  (g/cm³); BAIt is the basal area increment in year t (m²); WD  is the basic wood density (g/cm³), obtained from WD_U12% * 0.838 in year t; BAt is the total basal area in year t (m²).