Tropical and subtropical ecosystems are believed to account for nearly 70% of all the carbon (C) sequestered by Earth's forests. However, estimates of tropical C uptake are largely based on studies of lowland ecosystems, and limited climatic variation in lowland tropical forests hampers our ability to extrapolate observed trends and drivers of C dynamics to areas with greater environmental heterogeneity and steeper climatic gradients, such as montane forests. Quantifying the contribution of tropical and subtropical montane forests to C uptake is essential for generating comprehensive estimates of global C cycling and for motivating the preservation of these forests and their ecosystem services. There is a pressing need to identify the ecological factors that drive large-scale changes in the amount of C stored in the living aboveground biomass (AGB) of tropical montane forests (AGC), to improve our predictive understanding of how these systems contribute to future C storage and cycling. The Andes, the world's longest mountain range and a biodiversity hotspot, have sparse and poorly quantified estimates of regional-scale C stocks. Many of South America's most populous cities are located in the Andes above 500 m asl, reflecting pre-Hispanic and post-Spanish colonization patterns. This has led to a long history of anthropogenic disturbances. Ongoing human activities, coupled with the instability of mountainous terrain, have created a heterogeneous mosaic of forests with varying disturbance levels. Global warming is causing directional changes in forest composition through upslope species shifts. These rapid changes raise the prospect of considerable C losses at the lagging edge of species' ranges due to elevated mortality of large adult trees in areas that become too hot and/or dry. This biomass loss may be partially offset by the increased recruitment and growth of more thermophilic species migrating upslope. This thermophilization process is expected to enhance forest disturbance, increase mortality, and drive net losses of AGC. Disentangling the relative importance of disturbance and climate change as drivers of AGC dynamics is crucial for understanding and predicting the role of tropical forests in global C cycling. AGC stocks and productivity in mature Andean forests are typically characterized as decreasing monotonically with elevation due to colder temperatures and harsher climates. However, patterns of AGC can be complex, impacted by multiple biotic factors. Belowground symbiotic root associations (SRA) are increasingly recognized as key drivers of forest dynamics and soil C stocks, but their role in Andean forests has been mostly evaluated with coarse-resolution global models. There is growing interest in the importance of evolutionary history and phylogenetic diversity (PDz) in determining patterns of AGC stocks and productivity, and specifically the increases in tree size and biomass that occur in many cold, high-elevation tropical montane forests. The evolutionary dimension of biodiversity may affect ecosystem functioning through niche complementarity and selection of functionally-redundant clades. Directly quantifying and disentangling the relative contributions of biotic and abiotic factors on AGC dynamics in Andean forests will enhance our ability to forecast the future composition and function of these forests.

Existing research on carbon sequestration focuses heavily on lowland tropical forests, with less attention paid to the significant role montane forests might play. Studies on Andean carbon stocks are limited in scope and regional quantification. While some research highlights the monotonic decrease of AGC with elevation due to harsher climates, other studies point to the complexity of this relationship, impacted by biotic factors such as belowground symbiotic root associations (SRA) and phylogenetic diversity (PDz). The influence of SRA on forest dynamics and soil carbon stocks has been established in other ecosystems, but Andean forests lack detailed, local-scale studies. Similarly, the relationship between phylogenetic diversity and AGC stocks and productivity in Andean forests requires further investigation. The limited existing literature highlights the need for a comprehensive continental-scale assessment of Andean forest carbon dynamics.

This study analyzes AGC dynamics using data from the Red de Bosques Andinos (RBA), a network of 119 forest-monitoring plots across five Andean countries. The AGC dynamics of each plot are characterized using annualized values of AGC mortality, AGC productivity, and AGC net change. Both structural equation modeling (SEM) and an Information Theory (IT) natural model-averaging technique are used to identify the dominant climatic and biotic drivers of AGC dynamics. Regional climate is characterized through Principal Component Analyses of temperature (PCAtemp) and precipitation (PCAprec). Biotic explanatory variables include the Thermophilization Rate (TR), symbiotic root associations (SRA), plant phylogenetic diversity (PDz), and a size-dependent mortality parameter (β). The extent of forest cover and the rate of forest loss in the Andes between 2003 and 2014 are estimated to assess the carbon balance. The study area comprises 119 forest inventory plots (73 tropical, 46 subtropical) across a latitudinal and elevational gradient. Climate variables were extracted from CHELSA bioclimatic rasters. Aboveground biomass (AGB) of each tree was estimated using an allometric equation by Chave et al. (2014), with wood density (WD) values assigned at the species, genus, or family level. Tree height (H) was estimated using allometric relationships between DBH and tree height developed for each plot. AGC dynamics were calculated from annualized values of AGC mortality, productivity, and net change. The thermophilization rate (TR) was calculated using herbarium collection records from GBIF, estimating thermal optima for tree species. Phylogenetic diversity (PDz) was calculated using a phylogenetic tree generated with Phylomatic, and SRA was assigned based on genus- or family-level designations. Forest cover and loss were estimated using Hansen et al. (2013) data via Google Earth Engine, focusing on areas with forest cover ≥70% and annual rainfall ≥700 mm. Bootstrapping was used to assess the mean and 95% confidence intervals of AGC stocks. Generalized additive models (GAMs) were used to evaluate latitudinal and elevational patterns of AGC stocks and dynamics. Structural Equation Modeling (SEM) and Information Theory (IT) were used to identify the most important explanatory variables of AGC dynamics. A natural model-averaging technique was applied to select the most important variables, considering both abiotic (PCAtemp, AGC1) and biotic (TR, SRA, PDz, β) factors. Partial standard deviations were used to standardize coefficients, accounting for multicollinearity.

The mean net change in aboveground biomass (AGB) across the 119 plots was 1.44 ± 0.18 Mg ha⁻¹y⁻¹, representing a net increase in AGC of 0.67 ± 0.08 Mg C ha⁻¹y⁻¹ (1.01 ± 0.13% increase). This net increase resulted from greater AGC productivity than mortality. SEM and IT analyses showed that PCAtemp2 (representing a south-north climate gradient) significantly explained AGC net change when considering only abiotic variables. When including biotic variables, the size-dependent mortality parameter (β) was a significant explanatory variable in both SEM and IT models. AGC net change was negatively associated with β, suggesting that much of the increase in AGC net change was due to post-disturbance growth of larger trees. AGC productivity was positively correlated with initial AGC stocks, abundance of AM fungal associations, and negatively correlated with phylogenetic diversity. AGC mortality was not significantly correlated with climate across all plots but was strongly influenced by β and, when β was excluded, by TR. The estimated mean AGC stock in 2003 was 3.83 Pg C (3.31–4.34 95% CI), increasing to 4.12 Pg C (3.43–4.81 95% CI) in 2014. AGC stocks decreased significantly with elevation. The total AGC forest cover weighted mean rate change (0.66 Mg C ha⁻¹ y⁻¹) was almost the same as the observed AGC net change at the plot level. This implies an estimated total AGC sink of 0.027 Pg C y⁻¹ (0.011–0.042 95% CI), despite a 4.2% reduction in total forest cover (estimated to have released 0.33 Pg CO₂ equivalent). The strong C sink in Andean forests (1.01% annually) is even stronger than that of mature lowland tropical forests in Amazonia, Africa, or Southeast Asia.

The findings indicate that Andean forests are acting as globally significant AGC sinks, even exceeding the rate of lowland Amazonian forests. The relationship between AGC mortality and net change with the size-dependent mortality patterns highlights the significant influence of disturbance and self-thinning on carbon dynamics. Around 30% of the net AGC uptake may be attributable to recovery from past disturbances, while the rest is likely due to factors like CO₂ fertilization and temperature increase. The negative correlation between AGC productivity and phylogenetic diversity in Andean forests contrasts with findings from lowland Amazonia. This suggests that selection effects and the conservation of large stature within key clades with different evolutionary histories play a significant role in driving AGC productivity. The positive association between AGC productivity and AM tree abundances highlights the importance of mycorrhizal associations in enhancing nutrient uptake, particularly at low temperatures. The increasing AGC stocks in remaining forests more than offset the estimated C emissions from deforestation, resulting in a net total uptake. The capacity of montane forests to gain AGC, along with expected long-term gains from upslope species migrations, indicates that forest recovery can substantially contribute to increased C storage.

Andean forests are globally significant AGC sinks with the potential to serve as important future carbon refuges. The declining strength of carbon sinks in lowland tropical forests increases the importance of montane systems for carbon management. Protecting remaining Andean forests and increasing restoration efforts are crucial for securing their contribution to global carbon storage and biodiversity conservation. Future research should focus on differentiating the drivers of biomass dynamics in complex tropical mountain forest ecosystems, particularly investigating the long-term effects of thermophilization and the role of mycorrhizal associations under changing environmental conditions.

The study acknowledges limitations in quantifying belowground carbon stocks and fluxes. Sampling intensity may affect the quantification of tree mortality and AGC dynamics, underscoring the need for larger sample sizes in future studies. The study's focus on mature forests might not fully represent the carbon dynamics of all Andean forest types. The use of country and pantropical H:DBH allometries in some cases could have homogenized AGB estimates along elevational gradients.