The long-held belief of stable old-growth tropical forests is challenged by recent observations of changes in productivity, species composition, and functional trait composition. These changes are likely driven by climate change (warming, droughts, increased CO2) and disturbances (logging, hurricanes, fires). Understanding forest responses to multiple drivers is crucial for predicting future changes. However, disentangling the effects of these drivers is difficult because they can have opposing effects and operate at different timescales. Most studies focus on shorter timescales (decades), while tree canopy recruitment and successional changes occur over centuries to millennia. This study addresses this gap by assessing tropical forest responses at millennial timescales using a functional trait approach. This approach links changes in pollen taxon associations (representing species abundance) to their functional traits, creating a more mechanistic understanding of vegetation change than traditional palynological studies. Previous research applying this approach demonstrated the impact of climate change and human disturbance on trait composition over 7000 years in a lowland Peruvian forest. This study expands upon this work by examining eight diverse South American tropical forest landscapes across a wide elevational gradient (0-3200 m asl). Elevational changes result in marked decreases in water availability, temperature, and pressure, leading to variations in species and trait composition. Lowland forests might respond most strongly to rainfall changes, while highland forests may be most sensitive to temperature changes. The influence of past human activities and the varying degrees of disturbance across different forest ecosystems further complicate this understanding. Forests with longer disturbance histories might be more resilient. This study examines how temporal changes in climate (precipitation, temperature anomalies, El Niño frequency) and disturbances (fire activity, general disturbances) affect community-mean trait values (adult height, leaf area, seed mass, wood density) across these eight Neotropical forests over the past 2000-12000 years. The study hypothesizes that increased rainfall will increase the abundance of drought-vulnerable taxa with specific trait combinations, while high disturbance intensity will impact species composition in varying ways. It is also hypothesized that precipitation and El Niño effects are stronger at low elevations, whereas temperature effects are stronger at high elevations.

The literature review extensively cites previous research demonstrating changes in tropical forest productivity, species composition, and functional traits. Studies examining the effects of climate change and disturbances on these forests over various timescales are reviewed. The limitations of short-term studies in capturing the full dynamics of forest responses are highlighted. The traditional palynological approach and its limitations are discussed, leading to the introduction of the functional trait approach used in this study. Several studies focusing on short-term (decadal) forest responses to environmental change using community-weighted mean trait values are mentioned. The authors' previous work examining millennial-scale shifts in functional trait composition based on species abundance changes from fossil pollen records is also detailed. The importance of considering elevation gradients in understanding the relationship between environmental drivers and trait composition is emphasized, citing research showing how distributions of lowland and montane tree taxa are influenced by different environmental factors such as rainfall and temperature, respectively. The varying degrees of human occupation and associated disturbances in tropical forests across different time periods are discussed, linking human activity to fire occurrence and forest resilience. Existing literature highlights the potential for both increased resilience and drastic shifts in ecosystem states under various disturbance regimes.

This study utilized data from eight South American tropical forest sites, four lowland and four highland, spanning an elevational range from 0 to 3200 m asl. Fossil pollen records from lake sediments provided data on pollen taxonomic composition, charcoal concentration (fire activity), and Cecropia pollen abundance (general disturbance) for at least the past 2000 years (up to 12,000 years for some sites). The pollen data were linked to four key functional traits: wood density (WD), leaf area (LA), adult height (H), and seed mass (SM). Community-mean traits were calculated for each pollen sample, weighted by ln-transformed tree taxon abundance. Independent climate proxies for precipitation, temperature, and El Niño frequency were obtained from external sources. The charcoal data were standardized using z-transformations within records. Cecropia pollen abundance was used as a proxy for general disturbances. The effects of climate and disturbance variables on community-mean traits were analyzed using linear mixed models. Each model included climate and disturbance proxies as fixed effects, site as a random intercept, and year as a random slope per site. Interactions between elevation and other predictor variables were also tested. Site-specific analyses were conducted using generalized least squares regression models, incorporating a temporal autocorrelation structure to account for potential autocorrelation in the data. All response and predictor variables were standardized before analysis.

The analysis revealed significant relationships between community-mean traits and environmental variables, with effects often varying depending on elevation. General vegetation disturbances were the most important predictor, increasing tree height and leaf area, but decreasing seed mass and wood density across both lowland and highland sites. Temperature was the most important climatic predictor, increasing leaf area and tree height at high elevations, but decreasing wood density and tree height at low elevations. The effects of El Niño frequency and fire activity varied considerably between sites, suggesting that site-specific responses to these drivers are context-dependent. Precipitation increased community tree height, while temperature decreased height, with El Niño frequency tending to decrease height. The temperature effect on height was positive at high elevations and negative at low elevations. Fire activity negatively affected height at high elevations but tended to have a positive effect at low elevations. Disturbances had a positive effect on height only at high elevations. Leaf area increased with precipitation, temperature, and general disturbance. El Niño frequency decreased leaf area at high elevations but increased it at low elevations. Seed mass was negatively correlated with general disturbance and fire activity at high elevations but positively at low elevations. Wood density decreased with temperature and general disturbance, with the latter effect being particularly strong at low elevations. Adult height and leaf area were generally more responsive to climate and disturbances than wood density and seed mass. The overall model showed 35% of relationships are significant, while site level analysis showed 34% of relationships are significant. This suggests that temperature and disturbance are strong drivers of trait change. The results support the hypotheses that water availability influences acquisitive trait values (tall trees, large leaves), while temperature has opposing effects at different elevations. Disturbances generally increase acquisitive traits except in high elevation forests where fires seem to result in higher drought and fire resistance.

This study provides a millennial-scale perspective on the responses of Amazonian and Andean forests to climate change and disturbances. The findings highlight the importance of considering both climatic variables and disturbance regimes in predicting future forest dynamics. The most significant driver, general disturbance, likely represents a wide range of factors including human land use, indicating the substantial role of human impacts. The contrasting effects of temperature at different elevations underscore the importance of elevation gradient in determining the trajectory of forest response to warming. The variation in El Niño and fire effects across sites points to the role of local factors in mediating responses to these events. The differences in trait response suggest varying sensitivities between traits such as adult height and leaf area versus seed mass and wood density. The higher sensitivity of traits related to plant size to environmental change suggests that these traits should be targeted for future studies.

This research offers valuable insights into long-term forest dynamics in response to climate change and disturbances. The study demonstrates that temperature and disturbances are key drivers of functional trait composition shifts in Neotropical forests, with distinct responses depending on elevation. Future warming may lead to increased height in highland forests and decreased height in lowlands, with a general increase in soft-wooded species with large leaves. These shifts have implications for biodiversity and ecosystem functioning, potentially affecting carbon sequestration. Further research could investigate the influence of specific disturbance types and fire intensity on forest responses, refine local-scale analyses, and build predictive models that account for regional climate variability.

The study relies on pollen data as a proxy for forest composition, which may not perfectly reflect the actual species abundances and trait values. The use of genus-level averages for traits could introduce some uncertainty. The climate and disturbance proxies used were derived from multiple sources, each with its own uncertainties. The study does not explicitly measure the effect of time and does not control for the specific type and severity of disturbance. Further, this study does not capture the absolute changes in functional composition but rather focuses on relative changes over time. Future work may use improved methodologies for data collection and analysis to overcome these limitations.