The Initial Impact of a Hydroelectric Reservoir on the Floristics, Structure, and Dynamics of Adjacent Forests in the Southern Amazon

Abstract

This study assesses whether the rise in water level—following three years of reservoir filling at the Teles Pires Hydroelectric Plant (135.6 km² water surface) in Southern Amazonia—has affected the floristic composition, structure, and dynamics of adjacent forests. We established 62 permanent plots (2000 m² each) across a topographic gradient from the reservoir margin and conducted annual tree inventories for individuals with DBH ≥ 10 cm from 2014 to 2017. A total of 6322 individuals were recorded, representing 322 species, 210 genera, and 61 families. Fabaceae was the most abundant family, and the ten species with the highest importance value index (IVI) before reservoir filling remained dominant afterward. The forests exhibited high species richness and were characterized by a few common and many rare species. Mortality rates were highest within 10 m of elevation from the maximum reservoir level, indicating possible hydrological impacts, although no abnormal dieback or sharp shifts in floristic structure were observed. These results suggest limited short-term effects on species composition, but subtle changes in vegetation dynamics underscore the importance of long-term monitoring.

1. Introduction

The Amazon Basin is the largest watershed in the world, encompassing approximately 7 million km²—about 40% of South America’s land area. It harbors the planet’s most extensive and biodiverse tropical rainforest, composed of multiple forest types that reflect distinct biogeographic provinces [1]. The region’s forest ecosystems are shaped by a vast and intricate drainage network which, together with other ecological variables, governs the distribution of tree species and biomass [2]. These features make the Amazon one of the world’s most significant hydrological and biological systems, which plays a critical role in global climate regulation.

Although the Amazon Basin is the largest drainage basin in both area and water volume, it does not yet contain the world’s highest concentration of hydroelectric plants—that distinction belongs to China’s Yangtze River Basin. Nevertheless, hydroelectric development in the Amazon has expanded rapidly, with reservoirs now covering an estimated 15,000 km² of surface water. Despite this growth, knowledge remains limited regarding the impacts of reservoir formation on adjacent forests. This is particularly concerning given that groundwater levels are a key determinant of forest structure in the region [2].

The Teles Pires Hydroelectric Plant (TPHP), located along the middle and lower stretches of the Teles Pires River in northern Mato Grosso, is one of Brazil’s strategic energy initiatives [3]. Hydroelectric infrastructure in the region alters the physical environment in multiple ways: elevating the local water table, reshaping riverbanks, reducing seasonal hydrological variability, and altering nutrient dynamics. These changes can disrupt both ecosystem connectivity and function [4]. Hydroelectric projects in the Amazon have long been at the center of ecological and social controversies [5]. The TPHP, situated on the border between the states of Mato Grosso and Pará, has an installed capacity of 1820 MW and is frequently referenced in environmental impact assessments [6].

This study is part of the TPHP Flora Monitoring Program, which tracks vegetation succession in areas affected by the dam’s construction and operation, with particular emphasis on the newly designated Permanent Preservation Areas (PPAs) along the reservoir margins. The study analyzes whether long-term ecological changes have occurred in arboreal vegetation around the reservoir, focusing on potential effects related to topographic elevation and water level fluctuations. It also aims to identify species most sensitive to these changes. By assessing floristic composition and species distribution along environmental gradients, we aim to determine whether reservoir filling has triggered shifts in forest structure. Monitoring tree community responses in such altered landscapes is essential for improving our understanding of floristic dynamics in the region [7] and for addressing existing gaps in the literature, particularly the lack of information on how forest structure and species composition respond to landscape fragmentation and edge effects in the open rainforest ecosystems of the southern Amazon.

The Amazon is renowned for its exceptional tree species richness [8]. Despite recent advances in species classification and the modeling of distribution patterns, an estimated 30% of Amazonian tree species remain undescribed [9]. In light of the limited data on vegetation types, high levels of endemism, and the transformative effects of hydroelectric development, documenting and analyzing the structural and floristic composition of these forests is crucial. This study evaluates the initial impacts of the TPHP reservoir on species richness, diversity, and horizontal forest structure in the Southern Amazon.

We specifically analyze vegetation dynamics to assess changes in species composition, structure, and ecosystem function in response to the filling of the Teles Pires reservoir. This monitoring effort aimed to evaluate recruitment, mortality, and growth dynamics during the first three years of operation of the Teles Pires Hydroelectric Reservoir, located in the middle Teles Pires River basin.

Based on these research questions, we propose the following hypotheses:

–

Tree mortality will be higher in low-elevation areas due to prolonged exposure to flooding and soil hypoxia.

–

Tree recruitment will be lower in inundated zones, whereas regeneration may be enhanced in higher-elevation areas due to shifts in moisture availability and soil composition.

–

The growth of surviving trees will be influenced by changes in water availability, exhibiting distinct patterns along the elevational gradient.

To address these questions and test the proposed hypotheses, this study aims to:

–

Quantify tree mortality, recruitment, and growth across topographic zones affected by the reservoir.

–

Assess the influence of reservoir-induced environmental changes on tree survival, focusing on factors such as soil moisture.

–

Identify species-specific responses of trees to the altered hydrological and topographic conditions.

2. Materials and Methods

2.1. Study Area and Experimental Design

The Teles Pires Reservoir is located along the middle course of the Teles Pires River, with the dam of the Teles Pires Hydroelectric Plant (TPHP) situated precisely on the border between the municipalities of Paranaíta (Mato Grosso) and Jacareacanga (Pará), Brazil (Figure 1). The Teles Pires River—also known as the São Manoel River—joins the Juruena River to form the Tapajós River, a major clearwater tributary of the Amazon. The Teles Pires Basin covers approximately 142,000 km², with most of its drainage area in Mato Grosso and a smaller portion in Pará. The dam axis is located at coordinates 9°21′1.47″ S and 56°46′38.48″ W. The impoundment created a reservoir covering 160.81 km², with a water surface area of 147.16 km² at the maximum normal operating level of 220.44 m above sea level, spanning both municipalities.

According to the Köppen climate classification, the region has an Am (tropical monsoon) climate, typical of northern Mato Grosso [10]. Annual precipitation ranges from 1296 to 2493 mm [11], and the mean annual temperature is approximately 26 °C [12]. Two forest types were identified in the study area: Submontane Dense Ombrophilous Forest, which is predominant, and Alluvial Dense Ombrophilous Forest, which occupies smaller areas. These classifications follow the Brazilian Vegetation Technical Manual [13] and the environmental impact assessments conducted for the TPHP [6].

We established 62 permanent plots of 2000 m² each (50 × 40 m), totaling 12.4 hectares, in accordance with the guidelines of the environmental agency responsible (IBAMA—the Brazilian Institute of the Environment and Renewable Natural Resources) for monitoring the impacts in areas most influenced by the TPHP reservoir. The plots were grouped into modules located across three distinct environments: the main channel of the Teles Pires River (MB), lateral tributaries (LBs), and reservoir islands (RIs).

2.2. Monitoring of Permanent Plots

All plots were georeferenced and physically marked to facilitate long-term monitoring. The baseline inventory was conducted in November 2014, prior to reservoir filling in January 2015. The second inventory occurred in February 2015, followed by semiannual assessments, with the seventh and final monitoring campaign carried out in November 2017. Data from each campaign were used to evaluate temporal vegetation changes.

Within each plot, all trees with a diameter at breast height (DBH) ≥ 10 cm were measured and tagged with numbered aluminum labels. Recorded attributes included species name, family, DBH, total height, and the height of the morphological inversion point (PIM). For individuals with buttress roots, DBH was measured 1.3 m above the base of the main trunk. In multi-stemmed individuals, each stem was measured separately.

2.3. Overall Reservoir Vegetation Analysis

Floristic and phytosociological surveys were conducted through analysis of the sampled communities and supplementary collection of botanical specimens from the surrounding area. Botanical material was processed following the methodology described by Fidalgo and Bononi [14] and deposited as voucher specimens at HERBAM (Herbarium of Southern Amazonia, Alta Floresta, MT, Brazil). Species were identified in the field when possible; otherwise, samples were taken for later identification using the specialized literature or submitted to herbaria for taxonomic confirmation.

We adopted the APG IV classification system [15], consulted the relevant literature, compared specimens with those in the herbarium, and utilized online databases. Taxonomic nomenclature was verified using the Brazilian Flora List [16].

The following phytosociological parameters were estimated: relative and absolute density, frequency, dominance, importance value index (IVI), and coverage value, following the methodology of Mueller-Dombois and Ellenberg [17]. For the sampled communities, we calculated the Shannon–Weaver diversity index (H′), Simpson’s diversity index (C), Pielou’s evenness index (J′), the Jackknife richness estimator, and Sørensen’s similarity coefficient. All phytosociological metrics at the family, species, and community levels were calculated using Microsoft Excel and Mata Nativa 4.03 software [18].

2.4. Vegetation Analysis Along Elevation Gradients

We assumed that variation in terrain elevation correlates with differences in groundwater depth [2], which we did not measure directly due to the high cost and logistical complexity of installing over 60 wells across the study area. However, direct measurements were deemed unnecessary, as forest responses to groundwater variation were inferred from inventory-based metrics such as mortality, dominance, and species richness.

To evaluate floristic dissimilarity among plots at different elevations relative to the reservoir water level, we applied the UPGMA clustering method using the Sørensen/Bray–Curtis dissimilarity indexes based on species abundance data. In the main river body (MB), plots were established at 50, 150, 250, 350, and 450 m from the reservoir edge. In the lateral bodies (LBs) and island bodies (IBs), plots were located at 50 and 100 m, respectively (Figure 2). These analyses were performed using functions from the Vegan Package Version 2.9 [19] within the R statistical environment [20].

We also applied species richness estimators, including Jackknife 2 and Bootstrap. For Jackknife 2, estimates were based on the proportion of rare species relative to the total observed richness, using indicators such as singletons and doubletons (species represented by one or two individuals), or uniques and duplicates (species occurring in only one or two samples). For the Bootstrap estimator, observed richness was combined with the inverse of the proportion of plots in which each species occurred [21]. All estimates were computed in R.

We performed simple and multiple correlation analyses using the PerformanceAnalytics package in R. Differences in variables and demographic rates among environments were assessed using Tukey’s test at the 5% significance level.

2.5. Vegetation Dynamics

Tree dynamics were evaluated based on annual mortality and recruitment (ingress) rates, as well as annual gains (increments) and losses (declines) in basal area, using exponential models [22] and the Mata Nativa 4.03 software [18]. According to the cited methodology, calculations must account for community size over time, assuming that changes are proportional to the initial population size.

For correlation analyses, we used the following vegetation parameters: M = mortality, R = recruitment, N = number of individuals, and S = species diversity, combined with physical variables describing module location and elevation (average plot elevation relative to the reservoir’s maximum water level of 220.44 m). These variables were analyzed using Principal Component Analysis (PCA). Specifically, we applied a correlation-based PCA using Spearman’s method, due to the differing units and high variance among variables. All variables were standardized (n − 1, mean = 0, standard deviation = 1) to ensure equal contribution to the total variance. Highly collinear variables were excluded, as their near-linear relationships introduced redundancy into the principal components. All statistical analyses were conducted in the R environment [20].

3. Results

3.1. Overall Reservoir Vegetation

Across all plots, we sampled 6322 tree individuals, representing 322 species across 210 genera and 60 families. The families with the highest species richness were: Fabaceae (788 individuals), with 60 species recorded in November 2015 and 59 in the other inventories; Moraceae (714 individuals), with 20 species; Annonaceae (174 individuals), with 17 species initially, increasing to 18 in later inventories; Sapotaceae (213 individuals), with 14 species; Burseraceae (1224 individuals), with 12 species; Malvaceae (393 individuals), with 12 species; Apocynaceae (114 individuals), with 11 species; and Euphorbiaceae (246 individuals), with 11 species across all inventory periods (Table S1).

Vegetation structure changed slightly between 2014 and 2017 in terms of tree density. Before the reservoir filling in 2014, tree density was 480.7 individuals per hectare; by 2017, it had decreased to 460.9 individuals per hectare—a 4.1% decline. In the 2014 inventory, species with densities above two individuals per hectare (55 species) accounted for only 17.2% of total species richness but represented approximately 68.5% of the overall individual density. In 2017, this pattern remained stable: 17.3% of species accounted for about 68.9% of the total density. While part of this 4.13% decline in tree density may be attributed to natural processes such as senescence, pathogen activity, or competition for resources, the spatial concentration of mortality near the reservoir margin, as revealed by the PCA and topographic analyses, suggests an additional effect associated with hydrological changes following reservoir filling. These include prolonged soil saturation, hypoxic conditions, and edge effects, which are less likely to occur uniformly across the landscape in the absence of the reservoir.

Notably, all ten species with the highest importance value index (IVI) before reservoir filling remained within the >2 individuals/ha group in the final inventory, although shifts occurred in their relative rankings. The most dominant species in the initial inventory was Protium altissimum (IVI = 11.1%). It was followed by Helianthostylis sprucei (2.38%), Bertholletia excelsa (2.35%), Hevea guianensis (2.15%), Theobroma speciosum (2.01%), Toulicia subsquamulata (1.96%), Pseudolmedia laevigata (1.84%), Metrodorea flavida (1.75%), Astronium lecointei (1.56%), and Rinoreocarpus ulei (1.56%). In the final inventory, P. altissimum remained the most dominant species. H. guianensis rose to second place (2.32%), followed by H. sprucei (2.29%), B. excelsa (2.25%), T. speciosum (2.01%), T. subsquamulata (1.98%), M. flavida (1.91%), P. laevigata (1.71%), A. lecointei (1.62%), and R. ulei (1.62%).

The forest structure showed a high degree of stability, with dominant species maintaining their importance values over time. Two species, Guatteria sp. and Inga sp., were newly recorded during the monitoring period, while four species—Cassia leiandra, Loreya sp., Trema micrantha, and Vismia gracilis—were no longer observed, resulting in a 4.13% reduction in the total number of individuals recorded. Species richness estimated by Jackknife 2 and Bootstrap ranged from 318 to 321 species across the 62 plots. Jackknife 2 consistently yielded higher estimates than Bootstrap, with a slight divergence observed in November 2015, after which the estimates converged.

3.2. Vegetation Dynamics (Mortality vs. Recruitment)

Our results indicate no significant changes in floristic composition or species abundance between the first and final inventories, as demonstrated by the low variation across elevation gradients. Similarity analyses revealed the formation of nine floristic groups in the initial inventory (November 2014) and ten groups in the final inventory (November 2017).

Plot modules located in upstream island bodies (RIs) exhibited particularly high similarity, likely due to their spatial proximity and shared position within alluvial zones—areas most affected by water level fluctuations following reservoir impoundment. No sampled plot showed signs of flooding, even temporarily, as the reservoir’s water level is regulated by the hydropower plant.

After the reservoir filling in December 2014, the mean annual tree mortality rate between November 2014 and November 2017 was 3.9%, while the mean annual recruitment rate was 2.2%. Mortality ranged from 3.24% to 4.56%, and recruitment from 2.14% to 3.71%.

Principal Component Analysis (PCA) showed that the first axis (PCA1) explained 32.51% of the total variance, and the second axis (PCA2) explained 23.40%, totaling 55.91% of cumulative variance. In PCA1, the most influential variables were species diversity (0.83) and number of individuals (0.78). In PCA2, elevation (−0.63) and mortality (0.63) were the strongest contributors, but in opposite directions (Figure 3).

Mortality was highest in plots located within 0–10 m of elevation difference from the reservoir’s maximum water level. This pattern aligns with localized rises in the water table and edge effects caused by reservoir impoundment. Recruitment rates were generally lower than mortality across the monitoring period but showed a slight upward trend over time.

Following the reservoir filling in December 2014, the average annual mortality rate between November 2014 and November 2017 was 3.9%, while the average annual recruitment rate was 2.2%. Throughout the monitoring period, annual mortality rates fluctuated, reaching a maximum of 4.56% and a minimum of 3.24%. Recruitment rates also varied during this period, ranging from a maximum of 3.71% to a minimum of 2.14%. In the April 2016 inventory, an increase in the average density of dead trees was observed across all plots (21.21 individuals per hectare). However, this loss was partially offset by a relatively high number of recruited individuals in the same period (17.1 individuals per hectare).

The highest annual mortality rates following reservoir filling were recorded in plots closest to the reservoir margin. Specifically, mortality reached 6.3% on islands (IS), 5.8% in lateral arms (LB), and 4.2% in the main body (MB), all within 100 m of the water’s edge. In contrast, the lowest mortality rates were found in plots located downstream of the dam (DS) and in areas more than 100 m from the main reservoir body (MB > 100), with rates of 2.5% and 2.3%, respectively (Figure 4;

Table 1. Demographic parameters and annual dynamics of tree communities by environment in the influence area of the Teles Pires Hydroelectric Plant, from November 2014 to November 2017. LB_100, IS_100, MB_100, MB > 100, and DS represent the sampled environments. DC = elevation difference in the sampling unit from the reservoir’s water level (220.44 m); NT = total number of individuals sampled; S = species diversity; NM = number of dead individuals; NR = number of recruits (ingress); NM/ha = dead individuals per hectare per year; NR/ha = recruits per hectare per year; TM = annual mortality rate (%); TR = annual recruitment (ingress) rate (%).

Environments	DC	NT	S	NM	NR	Annual Periodic Increment (API)
						NM/ha	NR/ha	TM	TR
LB_100	4.1	1669	196	212	86	22.1	9.0	5.8	2.0
IS_100	0.7	693	129	96	34	22.9	8.1	6.3	1.9
MB_100	4.6	673	149	77	41	21.4	11.4	4.2	2.3
MB > 100	39.8	2368	240	163	146	11.8	10.6	2.5	2.3
DS	36.9	919	187	59	54	9.5	8.7	2.3	2.1
Total	10.4	6322	322	607	361	16.3	9.7	3.9	2.2

). In these environments, annual recruitment rates showed little variation, ranging from 1.9% to 2.3%—values considered typical for natural forest dynamics in the region.

Recruitment and mortality values for all inventories are expressed as the number of individuals per environment (LB_100, IS_100, MB_100, MB > 100, DS). In total, 361 recruited individuals were recorded, representing 104 species. Mortality was observed for 607 individuals, encompassing 171 species. Figure 5 presents the ten species with the highest recruitment and mortality during the monitoring period.

The species with the highest number of dead individuals was Protium altissimum (Aubl.) Marchand, with 86 recorded deaths. This species also showed the highest recruitment, with 50 individuals, and maintained the highest IVI (%) across inventories. Mortality among species was consistently concentrated in areas adjacent to the reservoir (LB_100, IS_100, MB_100), while recruitment was more uniformly distributed across all sampled environments.

The species with the highest mortality, in decreasing order, were Helianthostylis sprucei Baill. (28) Theobroma speciosum Willd. ex Spreng. (19), Pseudolmedia laevigata Trécul (18), Rinoreocarpus ulei (Melch.) Ducke (16), Inga alba (Sw.) Willd. (16), Bauhinia ungulata L. (15), Toulicia subsquamulata Radlk. (14), Dialium guianense (Aubl.) Sandwith (11), and Metrodorea flavida K.Krause (10).

As for recruitment, the species with the highest number of recruited individuals were Metrodorea flavida K. Krause (20), Rinoreocarpus ulei (Melch.) Ducke (12), Senegalia polyphylla (DC.) Britton & Rose a(11) Tachigali chrysophylla (Poepp.) Zarucchi & Herend (11), Hevea guianensis Aubl. (10), Inga alba (Sw.) Willd. (10), Trichilia quadrijuga Kunth (10), Zanthoxylum djalma-batistae (Albuq.) P.G.Waterman (10), and Cecropia sciadophylla Mart. (9).

4. Discussion

The UPGMA Sørensen/Bray–Curtis similarity values indicate that the tree community has remained relatively stable in terms of species composition, suggesting that reservoir filling did not significantly alter species diversity over the three-year monitoring period—even near the reservoir margin. The formation of subgroups (factor k) in the UPGMA analysis reflects some variation in species composition, but no clear spatial patterns related to plot location or topography. The high similarity among plots across topographic gradients and modules supports the notion of environmental uniformity [23].

Despite some localized variation, the overall similarity in species composition among plots across different modules and topographic zones suggests limited spatial heterogeneity at the landscape scale. This relative uniformity may reflect a shared disturbance history, consistent soil conditions, or ecological filtering processes that select for species with broad environmental tolerances [24,25]. The absence of distinct floristic clusters aligned with elevation or geographic location reinforces the interpretation that the forest in this area functions as a relatively cohesive ecological unit. This low spatial turnover in species composition provides important context for the observed floristic stability over time, as a more heterogeneous landscape might have exhibited greater temporal shifts in response to environmental changes following reservoir formation.

In this context, the Jackknife 2 estimator—recognized for its higher precision and lower bias compared to the Bootstrap method in datasets with high species abundance variation [26]—proved especially informative. Given the observed compositional stability and subtle structural changes, Jackknife 2′s capacity to account for sample heterogeneity likely provided more accurate species richness estimates.

Species accumulation curves (or sampling effort curves) showed no signs of reaching saturation during the study period, reflecting the high heterogeneity in species composition and structure across the landscape. This is a common pattern in tropical forests, where species richness is high and complete inventory of the species pool is rarely achieved [27,28].

The predominance of Fabaceae throughout the monitoring period was expected. The richness of this family is attributed to its wide array of growth forms—including trees, shrubs, herbs, and lianas [29]—as well as its adaptive capacity in dystrophic environments, facilitated by the presence of numerous nitrogen-fixing species [30]. Several studies have similarly identified Fabaceae as the most species-rich family in Amazonian forests, especially in nutrient-poor soils [31,32]. The families recorded in this study are also frequently reported in other tropical forest communities [33].

Between 2014 and 2017, despite relative compositional stability, structural changes were observed in the forest, particularly a 4.13% decline in tree density, from 480.73 to 460.88 individuals per hectare. This is consistent with values reported for undisturbed terra firme forests in the Amazon [34], such as Paragominas (496.34 ind.ha⁻¹) and Caracaraí (525 ind.ha⁻¹) in the Central Amazon. These studies also reported low overall diversity but high dominance by a few species—a pattern of hyperdominance broadly described for the Amazon Basin by ter Steege et al. [35]. Diversity indices in terra firme forests typically range from 3.58 to 4.76 nats.ind⁻¹ [36,37]. The Shannon index obtained in our study falls within this range, reflecting high species richness, low dominance, and relatively uniform spatial distribution of species—consistent with the similarity clusters identified by the Sørensen/Bray–Curtis index.

The forest adjacent to the Teles Pires reservoir exhibited high biological diversity, marked by a prevalence of rare or infrequent species and relatively few dominant taxa. Differentiating between rare and common species is essential in tropical forest studies, as it improves ecological understanding and informs conservation strategies [38,39]. Despite some variation in species composition among plots, no significant differences were observed in relation to module location, elevation, or time, indicating overall compositional stability. These findings support the idea that, so far, the reservoir has not significantly altered local species diversity or dominance patterns. The observed changes mainly reflect the loss of low-density species—those represented by only one or two individuals—highlighting the importance of spatial and environmental factors in shaping species distributions [40,41].

Conversely, following reservoir filling in December 2014, the mean annual mortality rate reached 3.9%, while recruitment remained at 2.2%. These values reflect forest turnover rates, which are key indicators of forest dynamics—particularly under post-disturbance conditions. Mortality rates are especially relevant near hydroelectric reservoirs, where they help elucidate the environmental impacts of large-scale infrastructure. The mortality rates observed here are higher than those typically reported for undisturbed Amazonian forests, such as 1.03% in the Central Amazon [42] and 1.14% in the Southern Amazon [43]. In contrast, our mortality rates align with those recorded in impacted areas such as FLONA Tapajós (3.7%) [44] and Fazenda Rio Capim in the Eastern Amazon (2.97%) [45], suggesting a potential influence of the reservoir in increasing tree mortality—with possible long-term consequences for forest composition and structure.

The PCA results showed that mortality was strongly associated with lower elevations, with the highest rates occurring in plots located within 10 m of the reservoir’s maximum water level. This pattern is consistent with hydrological changes such as rising groundwater levels and edge effects. Similar trends have been observed in other hydroelectric projects, such as the Jirau Reservoir in Southern Amazonia (Rondônia State, Brazil), where elevated water levels led to increased tree mortality in both várzea and terra firme forests [46]. In some areas, water retreat exposed bare soils, disrupting vegetation structure and ecological succession.

In addition to elevation alone, other environmental variables such as soil type, microtopography, and subsurface heterogeneity may also influence the relationship between elevation and groundwater dynamics. Variations in soil texture and structure can affect water retention capacity, infiltration rates, and root aeration, thereby altering the extent and duration of waterlogging in low-lying areas [47,48]. Microtopographic depressions, even within plots at similar elevations, may concentrate water and exacerbate anoxic conditions, increasing stress on tree roots. Likewise, heterogeneity in subsurface geology—such as the presence of compacted layers or differing sediment compositions—can modulate local water table responses to reservoir fluctuations. These factors, acting in concert with elevation, likely contribute to the spatial variability in tree mortality observed near the reservoir margin.

The upward trend in both mortality and recruitment observed over time at Teles Pires may reflect vegetation shifts and structural reconfiguration induced by altered hydrological conditions. Artificial flooding can impose more intense environmental filters than those in natural floodplains, especially in terra firme forests, which lack flood-adapted traits [49]. This supports the hypothesis that hydroelectric reservoirs can significantly affect forest structure and dynamics, although some changes may take more than three years to become fully evident.

The combined effects of deforestation, land use change, and large infrastructure projects (e.g., roads and dams) exacerbate ecological impacts in the Amazon [50], underscoring the need for long-term ecological monitoring. Understanding recruitment and mortality dynamics is crucial for assessing forest resilience to anthropogenic disturbance [51,52,53]. Ultimately, our findings contribute to advancing knowledge of tropical forest ecology and support the development of conservation and adaptive management strategies in landscapes affected by hydropower development [28]. Detecting the dynamics of rare species is especially important to ensure biodiversity conservation and the long-term integrity of Amazonian forest ecosystems.

While the overall species composition and floristic structure remained relatively stable throughout the monitoring period, the short temporal window may have limited the detection of more gradual ecological shifts. The consistent dominance of Fabaceae and the persistence of the ten most influential species (based on IVI) suggest resilience in the arboreal community. However, subtle fluctuations—particularly among rare species—underscore the potential for long-term compositional changes that may only emerge with extended monitoring.

At the same time, the reservoir’s influence on demographic dynamics was more immediately apparent, particularly in plots located near the water’s edge. Elevated mortality rates at lower elevations reflect hydrological pressures such as groundwater rise and edge effects. The PCA findings emphasize that these topographic and environmental gradients are key drivers of vegetation response. Although recruitment showed signs of recovery over time, the demographic imbalance, where mortality exceeded recruitment, suggests a transitional phase in forest structure. The presence of elevated mortality in more distant plots further points to the role of external climatic stressors, such as strong winds, in shaping vegetation dynamics beyond the direct influence of the reservoir.

Although our results indicate a relatively stable floristic composition over the three-year monitoring period, it is important to acknowledge that tropical forest responses to disturbance often follow nonlinear trajectories and may exhibit significant temporal lag effects [54,55]. Species turnover, shifts in dominance, and recruitment bottlenecks can arise years or even decades after the initial disturbance, particularly in mature forest systems characterized by complex species interactions and long-lived individuals [56]. Therefore, the apparent compositional stability observed thus far should be interpreted with caution. Continued, long-term monitoring is essential to detect gradual or delayed ecological responses—especially those involving rare species, altered successional trajectories, or changes in functional composition that are unlikely to be captured in short-term assessments [57]. Recognizing this limitation strengthens the interpretation of our findings and underscores the importance of longitudinal datasets to fully understand forest community dynamics under hydrological stress.

5. Conclusions

This study offers a detailed assessment of forest dynamics following the filling of the Teles Pires reservoir, examining tree mortality, recruitment, and species composition over a three-year period. A total of 6322 individuals were recorded across all plots, representing 322 species, 210 genera, and 60 families. While species richness exhibited only minor fluctuations, the overall forest structure remained stable, with dominant species maintaining their relative rankings. Fabaceae (788 individuals) and Burseraceae (1224 individuals) were the most abundant families, underscoring their ecological prominence in the study area.

Tree density declined by 4.1% between 2014 and 2017, from 480.7 to 460.9 individuals per hectare. Mortality rates consistently exceeded recruitment during the monitoring period, averaging 3.9% per year compared to an annual recruitment rate of 2.2%. Mortality was highest within 10 m of the reservoir’s maximum water level, indicating a strong influence of elevated water tables and edge effects. Principal Component Analysis (PCA) revealed a spatial relationship among elevation, species diversity, and mortality, highlighting the localized impacts near the reservoir margin.

A key strength of this study lies in its multi-year monitoring approach, which enabled a detailed analysis of tree demographic trends and ecological responses to hydrological changes. The examination of floristic groupings and PCA associations also provided valuable insights into forest structural stability. However, a limitation of the study is its relatively short observation window, which may not fully capture delayed successional processes or long-term species turnover. Despite this constraint, the study establishes a solid foundation for long-term ecological monitoring and contributes to the broader understanding of forest resilience in the context of hydroelectric reservoir development, thereby informing future conservation strategies.