Continuous Forest Inventory Research Articles

Eucalypt production in different rotations, sites and spatial arrangements

This work aimed to evaluate the productivity of eucalypt rotations in different site indexes and spatial arrangements. The database came from continuous forest inventories executed during the first and second rotations of 718 clonal stands (AEC 144®, AEC 224® and AEC 1528®), implanted in the 8.4 m2 plant−1 spacing and distributed in two arrangements: 3.0 × 2.8 m (rectangularity = 1.07) and 6.0 × 1.4 m (rectangularity = 4.29). Height, basal area, and volume differed between rotations as site quality improved and stands aged. In absolute terms, stands with lower site index, and installed under a larger rectangular arrangement showed less volume divergence between rotations. At 84 months of age, the first rotation of sites with the highest productive capacity tended to show 42.63 and 38.67 m3 ha−1 year−1 in smaller and larger rectangularity arrangements, respectively. The volumetric declines between the first and second rotations averaged 6.8 and 41.3 % at 84-month reference point, for arrangements of 6.0 × 1.4 m and 3.0 × 2.8 m, respectively. It is concluded that the less rectangular spatial arrangement (3.0 × 2.8 m) proved to be more productive for conducting a single rotation (high forest). Arrangement 6.0 × 1.4 m showed higher productive potential in stands managed with two rotations within the same cultivation cycle.

Site index curves using the generalized algebraic difference approach – Gada in planted forests of Pinus taeda L. in the midwest of Santa Catarina, Brazil

One of the most widely used tools for assessing the productive potential of planted Pinus taeda L. forests is to generate site index curves. Site index is a measure of the productive potential of forests and is influenced by various factors such as soil, water, light, among others, representing the relationship between the dominant height of trees and age at a specific location. This relationship produces a curve and can be used for site classification, consequently estimating the productivity of the forest stand. The objective of this study was to evaluate the performance of dynamic equations, the Algebraic Difference Approach (ADA), and its generalization (GADA). The data were obtained from permanent plots of Continuous Forest Inventory (CFI), collected in forest stands, through a sequence of annual remeasurements from 5 to 17 years of age. Dominant height was estimated as a function of age using nonlinear equations, and subsequently, the ADA and GADA methods were applied using the Chapman-Richards basic equation. The results of the GADA method provided a substantial reduction in the root mean square error (RMSE), less than half when compared to the traditional guide curve method, and an increase from 0.95 to 0.99 in R2adj. Therefore, the GADA methodology proved to be suitable and efficient for site index classification, with noticeable flexibility and reliability on its estimates.

CARBON STOCK AND DYNAMICS IN NATURAL REGENERATION IN MANAGED FOREST IN THE CENTRAL AMAZON

Around 2% of the living biomass above ground is found in the natural regeneration of tree species. Despite being a relatively low value, diagnosing the lower stratum of the forest is important in the current context of climate change. The aim of this study was to evaluate the effects of forest management on the carbon stock and dynamics of natural regeneration 20 years after logging in a forest in the Central Amazon. Data collection was carried out in three Annual Production Units (APU) that received exploration in the years 1996, 1997 and 1998. Data from continuous forest inventories carried out in 41 sampling units were used, with four measurement periods: a pre-exploratory and three other post-forest exploration. The carbon stock was estimated in trees with a diameter at breast height between 5 and 15 cm and the Annual Periodic Increment in carbon (APIc) and the carbon entry and mortality rates were calculated. The carbon stock ranged from 10.80 ± 1.27 Mg ha-1 to 14.86 ± 0.81 Mg ha-1. The APIc ranged from 0.18 ± 0.03 Mg ha-1 year-1 to 0.59 ± 0.25 Mg ha-1 year-1, with an entry rate of 0.00% to 3.26% and mortality between 3.35% and 13.84%. The carbon stock and APIc were not significantly influenced by the forestry management applied to the forest, with only time as a determining factor in its dynamics. Considering natural regeneration, it is concluded that the managed forest is capable of recovering the carbon stock lost during the forest management process.

A New Forest Growing Stock Volume Estimation Model Based on AdaBoost and Random Forest Model

Forest growing stock volume is a crucial indicator for assessing forest resources. However, contemporary machine learning models used in estimating forest growing stock volume often exhibit fluctuating precision and are confined to specific tree species, lacking universality. This limitation impedes their capacity to provide comprehensive forest survey services. This study designed a novel model for predicting forest growing stock volume named RF-Adaboost. The model represented the inaugural application of the Adaboost algorithm in estimating forest growing stock volume. Additionally, the authors innovatively refined the Adaboost algorithm by integrating Random Forest as its weak learner. To substantiate the model’s effectiveness, the authors designed three data combination schemes at different scales and conducted regression estimation using the RF-Adaboost model, traditional Random Forest, and Adaboost models, respectively. The results indicated that the RF-Adaboost model consistently outperforms others across various data schemes. Furthermore, utilizing a combined data scheme of remote sensing and Continuous Forest Inventory, the RF-Adaboost model demonstrated optimal performance in estimating forest growing stock volume (R2 = 0.81, RMSE = 7.08 m3/site, MAE = 3.36 m3, MAPE = 8%). Finally, the RF-Adaboost model exhibits greater universality, eliminating the need for strict differentiation between tree species. This research presented an efficient and cost-effective approach to estimate forest growing stock, addressing the challenges associated with conventional survey methods.

Wood density estimation using dendrometric and edaphoclimatic data in artificial neural networks

Forestry measurement is aimed at volumetric production of wood; however, for the pulp processing industry, the main interest is productivity in wood biomass and, to know this variable, it is necessary to determine the basic wood density (BWD) beforehand. Artificial neural networks (ANN) have been used in the forestry sector quite successfully to describe the dynamics of forest characteristics, such as estimating wood volume. In this context, the objective of this study was to assess the accuracy of the basic wood density estimates by means of ANN’s with Continuous Forest Inventory (CFI) and edaphoclimatic input variables. The database consisted of 3,797 data, from permanent plots of the CFI conducted in Eucalyptus sp stands and edaphoclimatic data from the planting sites. The five best ANNs were selected and the analysis of the estimates was carried out through the correlation between the estimated and BWD, the relative root mean square error (RMSE%) and graphical information. It was observed that both the CFI, edaphoclimatic information and the combination of both are potential and present similar results for the basic wood density estimate, and the errors associated with the estimates are between 3.9% to 3.5%. The ANNs based only on the CFI information presented higher RMSE. The use of ANN’s is feasible for estimating BWD and allows for excellent accuracy statistics.

Simulation and prediction of changes in tree species composition in subtropical forests of China using a nonlinear difference equation system model.

Changes in tree species composition are one of the key aspects of forest succession. In recent decades, significant changes have occurred in the tree species composition of subtropical forests in China, with a decrease in coniferous trees and an increase in broad-leaved trees. This study focuses on Zhejiang Province, located in the subtropical region of China, and utilizes seven inventories from the National Continuous Forest Inventory (NCFI) System spanning 30 years (1989-2019) for modeling and analysis. We categorized tree species into three groups: pine, fir, and broadleaf. We used the proportion of biomass in a sample plot as a measure of the relative abundance of each tree species group. A novel nonlinear difference equation system (NDES) model was proposed. A NDES model was established based on two consecutive survey datasets. A total of six models were established in this study. The results indicated that during the first two re-examination periods (1989-1994, 1994-1999), there was significant fluctuation in the trend of tree species abundance, with no consistent pattern of change. During the latter four re-examination periods (1999-2004, 2004-2009, 2009-2014, 2014-2019), a consistent trend was observed, whereby the abundance of the pine group and the fir group decreased while the abundance of the broad-leaved group increased. Moreover, over time, this pattern became increasingly stable. Although the abundances of the pine group and the fir group have been steadily declining, neither group is expected to become extinct. The NDES model not only facilitates short-term, medium-term, and even long-term predictions but also employs limit analysis to reveal currently obscure changing trends in tree species composition.

Size of sampling units aproppriate for management level forest inventory in Eastern Amazon, Brazil

This study aimed to indicate the ideal sample size for continuous forest inventories in the tropical forest in Amapá. Data were obtained from a monitoring inventory started in 2010 and completed in 2016, in five permanent sampling units of 100m x 100m with an inclusion level of DBH ≥ 10cm. Randomly allocated sample subunits with five different dimensions ranging from 400 m2 to 1600 m2 were simulated within these units. The measurement times of each sampling unit, from the first to the last three, were timed. Dimensions were analyzed for their accuracy and relative efficiency. Given the above and depending on the purpose of the inventory, the standard sampling unit of 1 ha (100 m x 100 m) is recommended for the inventory variables. The 20 m x 20 m (400 m2) square sampling unit was the most suitable for monitoring growth variables (annual increment in diameter, basal area, and volume), indicating that these sampling units for continuous inventories can become an essential alternative for use in monitoring inventory in the region.

Predicting eucalyptus plantation growth and yield using Landsat imagery in Minas Gerais, Brazil

Remote sensing applications in forest inventories allow for a detailed view of areas with extensive forest cover while potentially lowering costs. In this context, the objective of this study was to propose a methodology for using Landsat data to predict and project eucalyptus forest growth and yield, as well as to analyze forest growth dynamics. Three scenarios with varying sample sizes were proposed to assess the possibility of cost savings. The estimation was performed using an artificial neural network (ANN) and the random forest (RF) algorithm. Wall-to-wall prediction and projection were used, and projection maps were used to calculate the mean and current monthly increments. The relative root mean square error (RRMSE) values for the wall-to-wall prediction and for the projection ranged from 7.9% to 14.5% and 6.8% to 11.8%, respectively. All of the estimated means fell within the forest inventory confidence intervals. The mean and current monthly increment values projected by the ANN model did not differ from those obtained by the continuous forest inventory. Considering that the sampling intensity of traditional continuous forest inventories normally is one plot every 10 ha, the methodology presented here can be used to reduce sampling intensity to one plot for every 83.3 ha, without significant loss of accuracy, which represents a relevant cost reduction.

Multistage Sampling and Optimization for Forest Volume Inventory Based on Spatial Autocorrelation Analysis

It is important to achieve estimates at the minimum cost, with no greater uncertainty than that which is appropriate for the objectives of the inventory. The aim of this study was to estimate the forest volume efficiently and accurately by sampling and analyzing the existing forest survey data, which is also a technical challenge. In this work, we used the spatial statistics tools in the ArcGIS software to analyze spatial autocorrelations with the data from the sixth to ninth continuous forest inventories (CFI) of Sichuan Province from 2002, 2007, 2012, and 2017. Based on the sampling framework of the CFI, we divided the sampling units into five groups using different methods to create the second-stage samples. Combined with the spatial autocorrelation analysis results, we selected certain samples from the collection of second-stage samples through stratified sampling to form the third-stage sampling units. We applied the sampling ratio, sampling accuracy, workload, and costs as the evaluation indexes for the sampling efficiency analysis. The main results are as follows: Before conversion, the forest volume density had a positively skewed distribution. There was substantial positive spatial autocorrelation, and its intensity was affected by the distance scale, especially at 187.3 km, where the spatial processes of clustering were most pronounced. At the significance level of α = 0.01, the high-volume stands were mainly concentrated in the Aba Prefecture, Garze Prefecture, and Liangshan Prefecture, while the low-volume stands were mainly concentrated in the Sichuan Basin region. The heterogeneous gatherings were staggered between the high-volume areas and low-volume areas, while the transition zone between the three prefecture regions and basin region was randomly distributed. With 95% reliability, the average estimation accuracy of the systematic sampling, random sampling, and cluster sampling in the second stage was 94.09%, which is less accurate than the CFI estimation accuracy. The mean correlation coefficients (R) between the estimated value of the forest volume and the observations of the systematic sampling, random sampling, and cluster sampling in the second stage were 0.95, 0.98, and 0.96, respectively. The relative differences (RD%) were −0.52, −0.39, and −0.36, respectively. The spatial stratified sampling in the third stage, which is based on spatial distribution pattern information, significantly reduced the sampling ratio to 1.68 per 10,000, compared with the average ratios of the CFI sampling and second-stage sampling, which were 13.73 per 10,000 and 2.75 per 10,000, respectively. With 95% reliability, the mean accuracy of the spatial stratified sampling in the third stage was 93.05%, the R was 0.94, and the RD% was −0.09. Spatial stratified sampling is more in line with the actual work conducted in annual surveys because it effectively reduces the sample size using prior spatial information, which can better meet the requirements of the annual output.

ALGORITHM FOR THE PROJECTION OF FOREST GROWTH AND PRODUCTION

The modeling of forest growth and production is an essential tool for forestry management because it allows us to perform simulations and project forest biometric variables in the future, thus assisting in stock planning and economic analyses. In this work, a growth and production model by diameter distribution was proposed with the application of the Weibull function based on the recovery of parameters through simplified functions between the forest attributes and the parameters of the Weibull function. The algorithm was developed in Excel’s VBA language. Validation was performed with data from the Continuous Forest Inventory (CFI) in a stand of Khaya grandifoliola and in rows of Eucalyptus spp. in the ILPF system, which were ordinarily organized into seven date combinations, from the most distant from to the closest to the projection date. The results were evaluated by the percentage standard error (SE%) applied to the projected and observed volumes and by the Kolmogorov‒Smirnov test applied to the diameter distributions to verify adherence. It was possible to identify an exact relationship for parameter c of the Weibull function as a function of the percentiles and for parameter b, improving the parameter recovery method. Another methodological improvement was the use of maximum diameter and maximum height for age to adjust the hypsometric function. The algorithm presented results for total volume with errors up to 20% in 85% of the tests.

Evolução da severidade do cancro e qualidade de fuste no crescimento do mogno-africano

Abstract The objective of this work was to investigate the influence of the qualitative traits canker severity and stem quality on the mean annual increment (MAI) in diameter, total height, and stem height of an African mahogany (Khaya grandifoliola) stand. The study area is located in the state of Minas Gerais, Brazil, where continuous forest inventories were conducted over six years for data collection. Given the nonparametric characteristic of the data, an aligned rank transform analysis of variance was performed for each factor, and the averages were compared by Dunn’s multiple test. There was no significant effect of the interaction between the classes of canker severity and stand age. Increased canker severity has a significant effect on MAI values, causing reductions of 13% in diameter, 11% in total height, and 21% in stem height in trees from the highest canker severity class, when compared with the healthy ones. Age interacts significantly with stem quality, and MAI values tend to be similar in younger trees, differing over time in the oldest ones. In addition, the highest MAI values are recorded in trees with the best stem classification.

Site classification for Eucalyptus sp. in a tropical region of Brazil.

The aim was to determine the productive capacity of a forest site by applying different methods of fitting, combined with geostatistical techniques, to stands of Eucalyptus sp. in a tropical region of Brazil. Data were collected from 845 plots of a continuous forest inventory over four years. The classification of local production capacity was performed using growth curves obtained by the guide curve (GC) method, algebraic difference approach (ADA) and generalized algebraic difference approach (GADA) methods and ordinary kriging through the spherical, exponential, and Gaussian models to determine the spatial dependence of the variables site, geographical boundaries of site index classes, and their respective areas for each hectare in an annual production unit (APU). The modified Chapman-Richards model, fitted by the generalized difference approach method (GADA), provided the best statistical results, an improvement of 12.23% and 39.80% on the ADA and GC methods, respectively. The exponential model selected to express the spatial distribution of dominant height showed a high degree of spatial dependence.

Carbon stock and sequestration as a form of payment for environmental services in a sedimentary basin humid forest refuge in brazilian semiarid

Forests function as carbon reservoirs since they act in its sequestration and storage, playing a fundamental role in global climate change mitigation. Payments for this kind of environmental service have emerged as an important means for combating deforestation. This study evaluated the potential of a Sedimentary Basin Humid Forest refuge in a Semiarid Brazilian region (Chapada do Araripe, southern Ceará state) to receive payments for environmental services (PES) for carbon (C) assimilation and storage. The biomass quantification was performed by the non-destructive method and the determination of the C content was carried out using a LECO carbon analyzer to correlate carbon production in different litter components with climate variables. The carbon, carbon increment and stored carbon values were obtained by information collected from a continuous forest inventory. The average carbon content of each litter component and the volume of wood stored in the forest indicated that the fragment has 27.78 t ha−1 of carbon stored in its living biomass and an annual increment of 1.26 t ha−1 year. The carbon sequestered annually totaled 3.99 t ha−1 [carbon incorporated in the litter (2.73 t ha−1) + average annual increment of carbon in the commercial volume (1.26 t ha−1)] indicating that the area sequesters an average of 102.02 t ha−1 CO2e. Of the three studied compartments, only the leaves component showed a significant correlation with any climatic variable (rainfall). Based on amounts paid per ton of carbon sequestered, it is estimated that the area can earn € 2583.79.ha−1 should it participate in a program of PES for carbon sequestration and storage. This value serves as an incentive for the conservation of biodiversity, promoting environmental benefits and financial advantages compared to other forms of land use.

Predicting teak tree (Tectona grandis Linn F.) height using generic models and artificial neural networks

The continuous monitoring of dendrometric variables provides estimates that assist in conducting fast-growing stands. In this study, we aimed to investigate the performance of generic models and artificial neural networks to estimate total height of Tectona grandis in a forest stand in the Eastern Amazon. Continuous forest inventory was performed in this population, where total height and diameter at breast height were measured. The variables such as age and the square root of the average diameter (dg) of the plots were used to compose the methods adopted to estimate the height of the trees. The accuracy of these methods was assessed using the residual standard error of the estimate, the coefficient of correlation, and the graphical analysis of residues. The aggregated difference and ANOVA were calculated to compare the methods. The independent variables mentioned were able to describe the height behavior of individuals. We concluded that the methods have good residual dispersion, normal distribution of errors and little tendency to overestimate height. It was found that the generic models and the ANNs do not differ significantly from each other and are efficient to estimate the height of individuals. We also concluded that the ANNs, especially those that included dg, presented superior statistical indicators

ESTIMATION OF THE WEIBULL DISTRIBUTION FORM PARAMETER AS A FUNCTION OF THE SCALE PARAMETER BY THE PERCENTILE METHOD

Among the probability density functions that estimate and project diametric distributions, the Weibull function is one of the most studied, and the percentile method is an alternative approach for estimating its parameters, with different deduction methods for calculating the location, scale, and shape parameters; in addition, of these parameters, some require iteration between formulas. In this study, the percentile method was used, and a function was established between the shape parameter, the asymptotic percentiles, and the scale parameter. This function was tested with continuous forest inventory (CFI) data of a Khaya grandifoliola stand and Eucalyptus spp. in integrated forest-livestock-forest (IFLF) systems, and the function was also validated based on distributions designed by the application of a recovery methodology for determining forest growth and production (FG&P) parameters. The results showed an exact correlation when compared with the parameters calculated by the percentile method. This function using the percentile method, due to its practicality, may be an alternative for obtaining the shape parameter in FG&P methodologies.

A new paradigm for Continuous Forest Inventory in industrial plantations

In contrast to the conventional approach of using measured yields for independent variables in a double sampling estimator, a new estimator is presented for Continuous Forest Inventory (CFI) with partial replacement that is based on projected yields for independent variables. This proposed approach requires a modern growth and yield system for prediction and projection at the stand-level. Other attributes, such as trees per hectare and basal area per hectare are indirectly derived from the Sampling with Partial Replacement (SPR) estimator, together with a consistent stand-table of tree frequency at the diameter class level. The method is extended to stratified sampling and can accommodate cases in which the sample size of permanent plots for a stand is less than 10. Recommendations are provided about the plot size and plot configurations for temporary plots with the aim of establishing procedures that are fast, easy to measure, and inexpensive. It is demonstrated that the proposed sampling estimator can also be extended to some well-known plantation species where the silvicultural practice of thinning is common.

Effects of sustainable forest management on tree diversity, timber volumes, and carbon stocks in an ecotone forest in the northern Brazilian Amazon

The tension between the large global demand for tropical timber, and ecological sustainability and local socioeconomic development in the fragile natural ecosystems of the Amazon region has challenged many generations. In this case study, carried out in an ecotone forest in the northern Brazilian Amazon, we seek to demonstrate how forest management can became more sustainable through silvicultural prescriptions that reduce forest degradation and soil erosion. We evaluated the effects of the sustainable forest management (SFM) on tree diversity, timber volumes, and carbon stocks. A continuous forest inventory was carried out in nine 1-ha permanent plots (900 subplots of 100 m²), distributed in three treatments with three repetitions: T1 = control without selective logging, T2 = SFM, and T3 = SFM + silvicultural treatment of thinning release by tree girdling and poisoning. SFM was carried out with an average of three commercial timber trees harvested per hectare with DBH (diameter at breast height) ≥ 50 cm, corresponding to a timber volume of 20 ± 6 m3 ha-1, equivalent to the removal of 7 ± 2 Mg C ha-1 (5% of the original C). The formation of clearings (gaps) arising from the falling of exploited trees (0.12 ± 0.09 ha) resulted in greater impacts on the forest structure than the construction of 4 m skidder trails (0.06 ± 0.04 ha), with an average of 7 ± 4 damaged trees ha-1 and 5 ± 3 dead trees ha-1. Although six trees species became locally extinct, tree diversity was little altered, and timber volumes and forest carbon volume stocks remained essentially constant at this average logging intensity. Based on our results, sustainable forest management can become an efficient model for land use in the Amazon when harvesting is carried out using this average logging intensity. However, long-term monitoring studies using permanent plots in ecotone forests in the Brazilian Amazon will still be necessary. Quantifying, evaluating, and reporting impacts related to forest management will enable the formation of an empirical basis to support sustanaible forestry practices and for updating environmental legislation, contributing to local socioeconomic development, and maintaining the environmental services provided globally by tropical forests.

Configuration of the Deep Neural Network Hyperparameters for the Hypsometric Modeling of the Guazuma crinita Mart. in the Peruvian Amazon

The Guazuma crinita Mart. is a dominant species of great economic importance for the inhabitants of the Peruvian Amazon, standing out for its rapid growth and being harvested at an early age. Understanding its vertical growth is a challenge that researchers have continued to study using different hypsometric modeling techniques. Currently, machine learning techniques, especially artificial neural networks, have revolutionized modeling for forest management, obtaining more accurate predictions; it is because we understand that it is of the utmost importance to adapt, evaluate and apply these methods in this species for large areas. The objective of this study was to build and evaluate the efficiency of the use of a deep neural network for the prediction of the total height of Guazuma crinita Mart. from a large-scale continuous forest inventory. To do this, we explore different configurations of the hidden layer hyperparameters and define the variables according to the function HT = f(x) where HT is the total height as the output variable and x is the input variable(s). Under this criterion, we established three HT relationships: based on the diameter at breast height (DBH), (i) HT = f(DBH); based on DBH and Age, (ii) HT = f(DBH, Age) and based on DBH, Age and Agroclimatic variables, (iii) HT = f(DBH, Age, Agroclimatology), respectively. In total, 24 different configuration models were established for each function, concluding that the deep artificial neural network technique presents a satisfactory performance for the predictions of the total height of Guazuma crinita Mart. for modeling large areas, being the function based on DBH, Age and agroclimatic variables, with a performance validation of RMSE = 0.70, MAE = 0.50, bias% = −0.09 and VAR = 0.49, showed better accuracy than the others.

Temporal stability of stratifications using different dendrometric variables and geostatistical interpolation

Stratifying a forest results in more precise and cheaper inventories. This study aimed to select the stratifying variable that estimates more precise and stable inventory over the years for a eucalyptus plantation in Minas Gerais state, Brazil. The continuous forest inventory was performed annually from 2.7 to 6.8 years, and based on the field measurements, arithmetic mean diameter (d), height (h), dominant height (Hdom), basal area (G), volume (V), and mean annual increment in volume (MAI) were calculated. Semivariograms were generated and the exponential, spherical and Gaussian models were fit for each stratifying variable for each measurement date. The models were assessed by the reduced mean error and its deviation, being the exponential model selected. Maps showing the spatial distribution of all variables were generated for each measurement age, using ordinary kriging. Next, the study area was divided in four strata based on each stratifying variable for each measurement age. The stability of each stratifying variables for each measurement age were assessed by: 1) coincident strata area; 2) stability of total strata area; 3) plot permanency on each stratum; and 4) inventory error using stratified random sampling procedures. All variables in all ages presented spatial dependence structure. G and Hdom were the stratifying variables that generated the most and the least coincident strata area over the years, respectively. G and height (h and Hdom) were the stratifying variables yielding the least and most plot stratum changes, respectively. The same trend was observed for the total strata area stability. Stratifying based on MAI and V yielded the smaller inventory error, and h and Hdom yielded the largest. G was selected as the best stratifying variable because it yielded small inventory errors and was the most stable variable in terms of coincident strata area, total strata area and plot stratum changes over the years.

Evaluation of different silvicultural management techniques and water seasonality on yield of eucalyptus stands

Appropriate management of factors that influence forest development is essential to increase yield of forest plantations. The objective of this study was to evaluate the influence of water seasonality, nutritional management, and uniformity on yield of eucalyptus plantations and estimate the potential, attainable, and observed yield of adult eucalyptus stands. We evaluated Eucalyptus clonal stands in six regions of Minas Gerais State, Brazil, in a system of first and second rotation across four age classes using the twin-plots method (TP). In the study, 142 TPs were allocated alongside 142 plots in a continuous forest inventory (CFI) network of a private enterprise. The CFI received operational fertilization and additional fertilization was carried out in the TP. The trees were measured at the beginning of the experiment and at every six months to quantify the production in the wet (PW) and dry (PD) periods of the year. Uniformity of plantations was evaluated using Pvar 50 % and the optimal uniformity index. Potential, attainable, and observed yields were estimated using the average annual increase at seven years of age. The PW showed higher plant yield. There was an effect due to water availability and nutrient level on the yield of the stands. The driest semester of the year produces approximately 30 % of the current annual increase. The stands under the second rotation tend to have less uniformity than the in the first rotation. Potential yield varied depending on rainfall intensity where wetter regions had the highest yield.