There have been several studies that aim to determine the most superior Weibull parameter recovery approach of specifying a given forest stand’s Weibull diameter distribution, but no consensus has been made. The lack of agreement could be attributed to studies using different moments/percentile prediction models as well as using different plot size data. This study investigates how plot size and prediction model form affects the performance for moments, hybrid, and percentile Weibull parameter recovery approaches. Five plot sizes and three moments/percentile prediction models were used to determine their effects. Weibull parameters were calculated using each recovery method for each plot size and moments/percentile prediction model combination. Each combination’s diameter distribution was recovered and assessed using absolute error index. Results showed that plot size affected rank of precision for parameter recovery methods. Findings suggest that order statistics may be important in recovering Weibull distribution parameters from stand diameter summary statistics.

In previous studies it was determined that the fiber diameter distribution in a peripheral nerve could be estimated by a simulation technique known as group delay. These results could be further improved using a combinatorial optimization algorithm called simulated annealing. This paper explores the structure and behavior of simulated annealing for the application of optimizing the group delay estimated fiber diameter distribution. Specifically, a set of parameters known as the cooling schedule is investigated to determine its effectiveness in the optimization process. Simulated annealing is a technique for finding the global minimum (or maximum) of a cost function which may have many local minima. The set of parameters which comprise the cooling schedule dictate the rate at which simulated annealing reaches its final solution. Converging too quickly can result in sub-optimal solutions while taking too long to determine a solution can result in an unnecessarily large computational effort that would be impractical in a real-world setting. The goal of this study is to minimize the computational effort of simulated annealing without sacrificing its effectiveness at minimizing the cost function. The cost function for this application is an error value computed as the difference in the maximum compound evoked potentials between an empirically-determined template distribution of fiber diameters and an optimized set of fiber diameters. The resulting information will be useful when developing the group delay estimation and subsequent simulated annealing optimization in an experimental laboratory setting.

Genetic improvement has been an integral part of loblolly pine plantation forestry in southern United States for about 60 years with focus on improving timber yield, wood quality, and pest and disease resistance. Advances in techniques of genetic selection, breeding, and propagation of planting material have made it possible to achieve genetic gains that are likely to result in significant changes in loblolly pine stand dynamics. Height-age relationships, height and diameter relationships and distributions, and intraspecific competition were investigated in second generation open-pollinated, controlled-pollinated, and clonal loblolly pine with an objective of characterizing the nature and magnitude of changes in these characteristics due to genetic improvement and clonal forestry. Genetic improvement and/or clonal forestry had no practical effect on parameters of the height-age and height-diameter relationships beyond the effect on the asymptote parameter of the Chapman-Richards and Korf equations that were used to model these relationships. Genetic improvement resulted in an increase in the mean of height distribution without a corresponding increase in the mean of the diameter distribution, but had no effect the variance and skewness of the distributions. Thus, growth and yield models in which basal area is a function of height at a specific age (site index) are likely to over predict genetic gains in basal area growth and volume yield. Increase in stand density resulted in an increase in variance of the diameter distribution of non-clonal stands but had no effect on the variance of the diameter distribution of clonal stands. Thus, diameter distribution of clonal stands may differ from that of non-clonal stands after crown closure despite the distributions not being different before and during early stages of crown closure. This study also evaluated methods that may be used to predict height growth of new genetic varieties and those that may be used to asses intraspecific competition in forest stands. Mixed-model approach of calibrating a height-age relationship to a new loblolly pine clone gave biased estimates for clones that were at the extremes of the distribution of the groups. The use of maximum likelihood with simulated annealing (MLSA) to evaluate competitive interactions among trees in loblolly pine stands gave non-unique estimates of optimum competitor selection radius. A simpler technique that uses Pearson correlations was proposed and was found to work better than MLSA. / Ph. D.

Intresset för olika hyggesfria skogsbruksmetoder ökar men frågan är hur en skogsbruksplan kan anpassas till detta och vilken metod som kan användas. Denna studie inriktade sig på att mäta i fullskiktade skogar med enkla verktyg för att få ut beståndsdata till nytta för skötseln och att användas till en skogsbruksplan. För att få fram diameterspridningen klavades träden. Det krävdes minst 600 m2 för att få tillräcklig kvalitet i diameterspridning. För volymen jämfördes klavning, relaskopmätning och laserdata. Resultatet visade att det är svårt att mäta volym i fullskiktade bestånd beroende på tätheten och ojämnheten. Inte minst gäller detta relaskopmätning som undervärderade volymen. Klavning och laserdata fick ungefär samma precision. Studien visade också att tidsåtgången i fält för klavning jämfört med vanlig planläggning med relaskop endast skiljde några få minuter. Om klavning kombineras med att hämta volymen från laserdata kan till och med denna metod vara mer tidseffektiv.

A descrição de padrões, i.e., tendências ou arranjos não aleatórios em comunidades, possui um longo histórico em ecologia vegetal. Comumente, a estrutura e diversidade de comunidades vegetais são descritas a partir de sua distribuição em classes de tamanho (SDD), distribuição espacial (SSD) e de sua distribuição abundância de espécies (SAD). Isto porque há um pressuposto de que padrões existentes nestes descritores de comunidades são assinaturas de processos fundamentais na sua organização e funcionamento. Assim, a descrição de padrões é com frequência o primeiro passo para gerar ou testar hipóteses sobre esses processos que regulam a estrutura e diversidade de comunidades. Organizada em diferentes capítulos, esta tese teve como objetivo central descrever e comparar padrões em diferentes comunidades arbóreas Neotropicais, buscando gerar hipóteses sobre os processos que regulam sua organização e funcionamento. Para tanto, buscou-se utilizar uma abordagem de inferência baseada no ajuste e seleção de modelos, que foi realizado usando máxima verossimilhança estatística. Em todos os capítulos, os dados sobre as comunidades arbóreas são oriundos de diferentes parcelas florestais permanentes, quatro delas com 10,24 hectares, localizadas no Brasil, e outra com 50 hectares, localizada no Panamá. Além da introdução geral sobre os conceitos e técnicas utilizadas nesta tese (Capítulo 1), foram avaliados e comparados: (i) os descritores básicos da estrutura florestal (i.e. abundância, área basal e riqueza de espécies por sub-parcela - Capítulo 2); (ii) a SAD e como ela varia com o aumento da escala e tamanho amostral (Capítulo 3); (iii) a SDD e como ela se relaciona com a demografia das espécies (Capítulo 4.1 e 4.2); e, por fim, (iv) a SSD e como ela varia entre parcelas permanentes (Capítulo 5). Apenas a relação entre a SDD e a demografia das espécies foi realizada como os dados da parcela panamenha, enquanto que os demais capítulos se referem as quatro parcelas brasileiras. Cada capítulo utilizou métodos e modelos probabilísticos distintos para a descrição e comparação das variáveis de interesse. O Capítulo 2 mostrou que os descritores básicos florestais foram muito diferentes entre as quatro parcelas brasileiras. Estes descritores foram raramente normais nas escalas estudadas e as diferenças entre as parcelas foram mais evidentes quando a variância é considerada, fornecendo informações extras sobre os processos geradores de variabilidade dentro das parcelas. O Capítulo 3 mostrou que a mudança no formato da SAD com o aumento da escala é predominantemente um efeito indireto do tamanho da amostra. Assim, pode haver um efeito de escala, mas esse efeito é pequeno e parece depender do grau de similaridade de espécies entre amostras. No capítulo 4.1, foi necessária uma combinação de quatro diferentes distribuições de probabilidade para descrever a ampla gama de SDD, visto que os modelos candidatos raramente foram adequados para a maioria das espécies. No capítulo 4.2, verificou-se que o crescimento e recrutamento determinam o formato da SDD, o que não aconteceu com a mortalidade. No geral, curvas decrescentes de crescimento por diâmetro (i.e., maior crescimento juvenil) levaram à SDD menos íngremes, enquanto que taxas altas de recrutamento estiveram relacionadas à SDD mais íngremes. Apesar das previsões da teoria de equilíbrio demográfico terem apresentado relações positivas com as SDD observadas, houve muita variação, fazendo com que as previsões fossem pouco confiáveis. No capítulo 5, confirmou-se que a grande maioria das espécies se distribui de maneira agregada no espaço. No entanto, as parcelas apresentaram diferentes padrões de intensidade e tamanho de agregação. As diferenças nestes padrões entre parcelas foram, em geral, similares àquelas encontradas ao comparar populações de uma mesma espécie entre parcelas. Assim, as parcelas permanentes brasileiras apresentaram padrões bem distintos umas das outras, tanto em termos de estrutura quanto de diversidade de espécies, padrões estes que provavelmente foram determinados pelas condições ambientais as quais estas comunidades estão sujeitas. Diferentes padrões também foram encontrados em relação à distribuição espacial das espécies (i.e. frequência, intensidade e tamanho de agregação). Contudo, os resultados sugeriram que estes padrões estiveram mais ligados a heterogeneidade ambiental interna das parcelas do que com condições de clima e solo as quais estas comunidades estão sujeitas. Por outro lado, os padrões de distribuição espacial das espécies parecem ter influenciado os padrões de diversidade das comunidades. Não houve, entretanto, um número suficiente de parcelas sob as diferentes combinações de condições ambientais para testar estas sugestões ou para fazer generalizações para cada formação florestal, sendo necessário o confronto destas sugestões com outros estudos realizados em condições similares. Já na parcela panamenha, foi confirmada a expectativa teórica de que distribuições diamétricas refletem a demografia das espécies, em especial os padrões de crescimento e recrutamento. Mas, houve grande variação entre as espécies, dificultando a inferência precisa de padrões demográficos passados das espécies a partir de SDD atuais. Por fim, a abordagem analítica baseada no ajuste e seleção de modelos por máxima foi uma alternativa viável, flexível e apropriada, principalmente em relação à comparação simultânea de diferentes modelos e à busca de processos por trás dos padrões encontrados. Apesar de algumas limitações de cunho operacional, a abordagem baseado em modelos é uma alternativa adequada para a descrição de comunidades arbóreas, podendo ser utilizada de maneira consorciada com outras abordagens (e.g. testes de hipóteses) para descrever padrões e para gerar ou testar hipóteses sobre esses processos fundamentais que regulam a estrutura e diversidade destas comunidades. / Pattern description - search for trends or non-random arrangements in communities, has a long history in plant ecology. Commonly, the structure and diversity of plant communities are described based on their size class distribution (SDD), spatial distribution (SSD) and species abundance distribution (SAD). This is because there is an underlying assumption that the existing patterns in these community descriptors are signatures of key processes determining their organization and functioning. Thus, pattern description is often the first step to generate or test hypotheses about the processes governing community structure and diversity. Organized in different chapters, the main goal of this thesis was to describe and compare different patterns in Neotropical tree communities and to generate hypotheses about the processes that regulate them. To do so we used an approach based on model selection, which was performed using maximum likelihood. In all chapters the data on tree communities came from different permanent forest plots, four of them of 10.24 ha located in Brazil and another 50 ha located in Panama. In addition to the general introduction of key concepts and techniques used along the thesis (Chapter 1), it was evaluated and compared: (i) the basic forest descriptors (i.e. abundance, basal area and species richness per subplot - Chapter 2), (ii) the SAD and how it varies with increasing sample size and scale (Chapter 3), (iii) the SDD and how it relates to species demography (Chapter 4.1 and 4.2) and, finally, (iv) the SSD and how it varies between plots (Chapter 5). Only the relationship between SDD and species demography o was performed using the Panama plot data, while the remaining chapters relate the four Brazilian plots. Each chapter used different methods and probabilistic models for the description and comparison of the variables of interest. In Chapter 2, it was found that basic forest descriptors were very different between the four Brazilian plots. These descriptors were rarely normal at the studied scales and differences between plots were more evident when variance is accounted for, which seems to provide information on processes generating within-plot variability. Chapter 3 showed that the change in shape of the SAD due to increasing scale is predominantly an indirect effect of sample size. Thus, there may be an effect of the scale, but this effect is minor and seems to depend on the degree of species turnover between samples. In Chapter 4.1, it was shown that the combination of four different probability distributions was necessary to describe the wide range of SDD, since models were rarely appropriate for the majority of tree species. In Chapter 4.2, it was found that growth and recruitment, but not mortality, shape the SDD. On average, decreasing growth-diameter curves (i.e. higher juvenile growth) were associated to less steep SDD, whereas high recruitment rates were related to steeper SDD. Although the predictions of demographic equilibrium theory were positively related to the observed SDD, there was lots of variation, making predictions quite unreliable. In Chapter 5, it was confirmed that the great majority of species had clumped spatial distributions. However, the results of intensity and size of clumps showed that the patterns of aggregation were different among plots. Species shared between two plots generally showed patterns of spatial distribution that matched the patterns found for individual plots. Therefore, the Brazilian plots presented very distinct patterns, both in terms of structure and species diversity, which were most probably determined by the environmental conditions to which these communities are subjected. Different patterns among plots were also found in respect to species spatial distribution (i.e. frequency, intensity and size of aggregation). However, the results suggested that these patterns were more connected to within-plot environmental heterogeneity than with climate and soil conditions. On the other hand, the spatial distribution of the species seems to have influenced the patterns of diversity of communities. There was not, however, a sufficient number of plots under different combinations of environmental conditions to test these suggestions or to make generalizations for each forest type, being necessary to confront these suggestions with other studies conducted in similar conditions. In the Panamanian plot, it was confirmed the theoretical expectation that diameter distributions reflect the demographics of the species, in particular the patterns of growth and recruitment. But there was great variability among species, making it difficult to infer past demographic patterns from current SDD. Finally, the analytical approach based on model fit and selection by maximum likelihood was a viable, flexible and appropriate approach, particularly in respect to the simultaneous comparison of different models and to the search for mechanisms underlying patterns. Despite some more operational limitations, the model-based approach is an appropriate alternative for the description of tree communities and can be jointly used with other approaches (e.g. hypothesis testing) for pattern description and to generate and test hypotheses on the fundamental processes that determine the structure and diversity of these communities.

The structure of a forest stand is characterized by: (a) species composition, (b) age, (c) size (diameter and height), and (d) spatial (horizontal and vertical) arrangement of the trees. Depending on the species, site, and disturbance history, the stand structure varies with time, thus providing a snapshot of a particular development stage. Research on growth and stand structure has shown that the spatial distribution of trees is one of the key determinants of stand productivity. Forest inventories and ecological surveys carried out in British Columbia (BC) have shown that the structure of naturally established, unmanaged stands varies from simple (single-species, single-storied, and even-aged) to complex (multi-species, multi-storied, and uneven-aged). Only a few studies have quantitatively characterized this range of structural complexity, with nearly all studies focusing on old-growth stands. BC forest policy requires that harvested areas be regenerated with a mixture of tree species whenever a mixture is suited to the site. This policy is based upon the assumption that under appropriate conditions, increases in stand productivity, reliability, and/or biodiversity can be attained in mixed-species stands. This assumption has not yet been tested for forest ecosystems. One mechanism by which different tree species can reduce crown competition for light is through vertical separation (the development of multiple canopy strata). Canopy stratification is not easily recognized in mixed-species stands, particularly when species have similar shade tolerance and height growth patterns, and no quantitative methods have been developed to detect stratification. The diameter frequency distribution of two-storied stands have been characterized by inverted J-shaped as well as modal curves. Although it would be more appropriate to characterize stand structure by height frequency distributions, these distributions have not been developed. We suggest that (i) a stand is stratified if there are distinct, quantitatifiable modes in the size distribution; either diameter, height, or crown height, and (ii) height or crown height distributions will be the most sensitive measures. To characterize the structure of western hemlock (Tsuga heterophylla (Raf.) Sarg.) (Hw) and western redcedar (Thuja plicata Donn ex D. Don in Lamb.) (Cw) second-growth stands, and to investigate its influence on tree growth, we (1) described and compared size (diameter, height, and crown height) frequency distributions in single- and mixed-species stands, (2) determined whether mixed-species stands develop a stratified canopy, and (3) examined whether interactions between hemlock and redcedar affect tree growth.

Submitted by Jos? Henrique Henrique (jose.neves@ufvjm.edu.br) on 2017-05-03T14:57:17Z No. of bitstreams: 2 license_rdf: 0 bytes, checksum: d41d8cd98f00b204e9800998ecf8427e (MD5) petronio_henrique_alves.pdf: 952268 bytes, checksum: c672a103f4c27e5369547a6558985c5d (MD5) / Approved for entry into archive by Rodrigo Martins Cruz (rodrigo.cruz@ufvjm.edu.br) on 2017-05-17T12:03:12Z (GMT) No. of bitstreams: 2 license_rdf: 0 bytes, checksum: d41d8cd98f00b204e9800998ecf8427e (MD5) petronio_henrique_alves.pdf: 952268 bytes, checksum: c672a103f4c27e5369547a6558985c5d (MD5) / Made available in DSpace on 2017-05-17T12:03:12Z (GMT). No. of bitstreams: 2 license_rdf: 0 bytes, checksum: d41d8cd98f00b204e9800998ecf8427e (MD5) petronio_henrique_alves.pdf: 952268 bytes, checksum: c672a103f4c27e5369547a6558985c5d (MD5) Previous issue date: 2016 / Os objetivos deste trabalho foram avaliar a influ?ncia do desbaste e da fertiliza??o p?s-desbaste no crescimento de ?rvores de um povoamento clonal de eucalipto e comparar a idade t?cnica de desbaste (ITD), obtida atrav?s do M?todo dos Ingressos Percentuais, com a idade t?cnica de corte (ITC) de dois povoamentos de clones de eucalipto. Para o primeiro objetivo, os dados foram coletados em 30 parcelas permanentes, sendo 16 convencionais e 14 g?meas, em que foi analisado o efeito do desbaste (parcela convencional) e da fertiliza??o (parcela g?mea) sobre o crescimento das ?rvores. O desbaste foi realizado aos 81 meses, com intensidade de 20%, 35% e 50% de ?rea basal presente, eliminando os piores indiv?duos, e a fertiliza??o aplicada aos 99 meses. As vari?veis analisadas resultantes da totaliza??o dos dados das parcelas foram ?rea basal (B), di?metro m?dio (q), altura total m?dia (Ht) e volume total com casca (Vcc). A an?lise dos dados constituiu-se na compara??o do efeito de vari?veis em rela??o ? idade por meio de modelo de regress?o, e foi realizada aos 99 meses ap?s o desbaste e aos 81 meses ap?s a fertiliza??o. Para atender ao segundo objetivo, os dados foram coletados em 34 parcelas permanentes, instaladas em povoamentos com espa?amento de 3,0 x 2,5 metros, em dois s?tios diferentes. Essas parcelas foram medidas aos 24, 36, 48, 60 e 72 meses, durante o Invent?rio Florestal Continuo (IFC). O M?todo dos Ingressos Percentuais foi empregado para determinar a ?poca de realiza??o do primeiro desbaste. Para projetar as distribui??es diam?tricas, para idades superiores ? ?ltima medi??o, utilizou-se um modelo de distribui??o diam?trica, que consiste na recupera??o dos par?metros de uma fun??o de densidade de probabilidade (f.d.p) ao longo dos anos. Neste trabalho, adotou-se a fun??o Weibull, que foi ajustada pelo m?todo de m?xima verossimilhan?a. A partir das informa??es de n?mero de indiv?duos e valores estimados para os par?metros da fun??o Weibull, ? e ?, de ambas as distribui??es diam?tricas analisadas, calculou-se o ingresso percentual (IP) de ?rvores em classes de di?metro sucessivas ao longo do tempo. Para determinar a ITD, foi ajustado um modelo exponencial justaposto a um modelo linear simples. A ITC foi definida como o momento em que o incremento corrente ? igual ao incremento m?dio, sendo a produtividade volum?trica, ao longo do tempo, estimada pelo modelo de Clutter. O desbaste apresentou influ?ncia para as vari?veis ?rea basal, volume total com casca e di?metro m?dio, por?m n?o influenciou a altura total m?dia. A fertiliza??o realizada ap?s o desbaste n?o influenciou a produ??o das vari?veis analisadas. A idade t?cnica de desbaste de um povoamento de eucalipto localizado em um s?tio de maior capacidade produtiva ocorre antes, comparado com a idade de desbaste de um povoamento em um local de produtividade inferior. Esse mesmo comportamento ocorre para a idade t?cnica de corte. Para um mesmo s?tio, a idade t?cnica de desbaste ocorre ap?s a idade t?cnica de corte. / Disserta??o (Mestrado) ? Programa de P?s-Gradua??o em Ci?ncia Florestal, Universidade Federal dos Vales do Jequitinhonha e Mucuri, 2016. / The objectives of this work were to evaluate the influence of thinning and post-thinning fertilization on tree growth of a clonal eucalypt stand, and compare the thinning technical age (ITD), obtained by the Percentages Ingress Method, with the harvest technical age (ITC) of two clonal eucalypt stands. For the first objective, the data were collected in 30 permanent plots, 16 conventional and 14 twins, in which were analyzed the effect of thinning (conventional plot) and fertilization (twin plot) on tree growth. The thinning was carried out at 81 months, with intensity of 20%, 35% and 50% of present basal area, eliminating the worst individuals, and the fertilization applied at 99 months. The analyzed variables resulting from the summation of data from plots were basal area (B), quadratic diameter (q), mean total height (Ht) and total volume with bark (Vcc). The data analysis consisted in the comparison of variable effects, in relation to age, through regression model, and was held at 99 months after the thinning and 81 months after fertilization. To attend the second objective, the data were collected in 34 permanent plots, installed in stands with spacing of 3.0 x 2.5 meters, in two different sites. These plots were measured at 24, 36, 48, 60 and 72 months, during the Continuous Forest Inventory (IFC). The Percentages Ingress Method was employed to determine the epoch of realization of the first thinning. To project the diameter distributions for older ages after the last measurement, there was used a diameter distribution model, which consists in recovering the parameters of a probability density function (f.d.p) along the years. In this work, it was adopted the Weibull function, which was adjusted by the maximum likelihood method. From the information of number of individuals and estimated values for the parameters of the Weibull function, ? and ?, from both analyzed diameter distributions, it was calculated the percentage ingress (IP) of trees in successive diameter classes along the time. To determine the ITD it was adjusted an exponential model juxtaposed to a simple linear model. The ITC was defined as the moment that the current increment is equal to the average increment, in which the volumetric productivity along the time was estimated by Clutter model. The thinning presented influence for the variables basal area, total volume with bark and quadratic diameter, however did not affect the mean total height. The fertilization performed after thinning did not influence the production of the analyzed variables. The thinning technical age of an eucalypt stand located in a higher productive capacity site occurs earlier compared to the thinning age of a stand in a lower productivity site. This same behavior occurs for the harvest technical age. For the same site, the thinning technical age occurs after the harvest technical age.

A descrição de padrões, i.e., tendências ou arranjos não aleatórios em comunidades, possui um longo histórico em ecologia vegetal. Comumente, a estrutura e diversidade de comunidades vegetais são descritas a partir de sua distribuição em classes de tamanho (SDD), distribuição espacial (SSD) e de sua distribuição abundância de espécies (SAD). Isto porque há um pressuposto de que padrões existentes nestes descritores de comunidades são assinaturas de processos fundamentais na sua organização e funcionamento. Assim, a descrição de padrões é com frequência o primeiro passo para gerar ou testar hipóteses sobre esses processos que regulam a estrutura e diversidade de comunidades. Organizada em diferentes capítulos, esta tese teve como objetivo central descrever e comparar padrões em diferentes comunidades arbóreas Neotropicais, buscando gerar hipóteses sobre os processos que regulam sua organização e funcionamento. Para tanto, buscou-se utilizar uma abordagem de inferência baseada no ajuste e seleção de modelos, que foi realizado usando máxima verossimilhança estatística. Em todos os capítulos, os dados sobre as comunidades arbóreas são oriundos de diferentes parcelas florestais permanentes, quatro delas com 10,24 hectares, localizadas no Brasil, e outra com 50 hectares, localizada no Panamá. Além da introdução geral sobre os conceitos e técnicas utilizadas nesta tese (Capítulo 1), foram avaliados e comparados: (i) os descritores básicos da estrutura florestal (i.e. abundância, área basal e riqueza de espécies por sub-parcela - Capítulo 2); (ii) a SAD e como ela varia com o aumento da escala e tamanho amostral (Capítulo 3); (iii) a SDD e como ela se relaciona com a demografia das espécies (Capítulo 4.1 e 4.2); e, por fim, (iv) a SSD e como ela varia entre parcelas permanentes (Capítulo 5). Apenas a relação entre a SDD e a demografia das espécies foi realizada como os dados da parcela panamenha, enquanto que os demais capítulos se referem as quatro parcelas brasileiras. Cada capítulo utilizou métodos e modelos probabilísticos distintos para a descrição e comparação das variáveis de interesse. O Capítulo 2 mostrou que os descritores básicos florestais foram muito diferentes entre as quatro parcelas brasileiras. Estes descritores foram raramente normais nas escalas estudadas e as diferenças entre as parcelas foram mais evidentes quando a variância é considerada, fornecendo informações extras sobre os processos geradores de variabilidade dentro das parcelas. O Capítulo 3 mostrou que a mudança no formato da SAD com o aumento da escala é predominantemente um efeito indireto do tamanho da amostra. Assim, pode haver um efeito de escala, mas esse efeito é pequeno e parece depender do grau de similaridade de espécies entre amostras. No capítulo 4.1, foi necessária uma combinação de quatro diferentes distribuições de probabilidade para descrever a ampla gama de SDD, visto que os modelos candidatos raramente foram adequados para a maioria das espécies. No capítulo 4.2, verificou-se que o crescimento e recrutamento determinam o formato da SDD, o que não aconteceu com a mortalidade. No geral, curvas decrescentes de crescimento por diâmetro (i.e., maior crescimento juvenil) levaram à SDD menos íngremes, enquanto que taxas altas de recrutamento estiveram relacionadas à SDD mais íngremes. Apesar das previsões da teoria de equilíbrio demográfico terem apresentado relações positivas com as SDD observadas, houve muita variação, fazendo com que as previsões fossem pouco confiáveis. No capítulo 5, confirmou-se que a grande maioria das espécies se distribui de maneira agregada no espaço. No entanto, as parcelas apresentaram diferentes padrões de intensidade e tamanho de agregação. As diferenças nestes padrões entre parcelas foram, em geral, similares àquelas encontradas ao comparar populações de uma mesma espécie entre parcelas. Assim, as parcelas permanentes brasileiras apresentaram padrões bem distintos umas das outras, tanto em termos de estrutura quanto de diversidade de espécies, padrões estes que provavelmente foram determinados pelas condições ambientais as quais estas comunidades estão sujeitas. Diferentes padrões também foram encontrados em relação à distribuição espacial das espécies (i.e. frequência, intensidade e tamanho de agregação). Contudo, os resultados sugeriram que estes padrões estiveram mais ligados a heterogeneidade ambiental interna das parcelas do que com condições de clima e solo as quais estas comunidades estão sujeitas. Por outro lado, os padrões de distribuição espacial das espécies parecem ter influenciado os padrões de diversidade das comunidades. Não houve, entretanto, um número suficiente de parcelas sob as diferentes combinações de condições ambientais para testar estas sugestões ou para fazer generalizações para cada formação florestal, sendo necessário o confronto destas sugestões com outros estudos realizados em condições similares. Já na parcela panamenha, foi confirmada a expectativa teórica de que distribuições diamétricas refletem a demografia das espécies, em especial os padrões de crescimento e recrutamento. Mas, houve grande variação entre as espécies, dificultando a inferência precisa de padrões demográficos passados das espécies a partir de SDD atuais. Por fim, a abordagem analítica baseada no ajuste e seleção de modelos por máxima foi uma alternativa viável, flexível e apropriada, principalmente em relação à comparação simultânea de diferentes modelos e à busca de processos por trás dos padrões encontrados. Apesar de algumas limitações de cunho operacional, a abordagem baseado em modelos é uma alternativa adequada para a descrição de comunidades arbóreas, podendo ser utilizada de maneira consorciada com outras abordagens (e.g. testes de hipóteses) para descrever padrões e para gerar ou testar hipóteses sobre esses processos fundamentais que regulam a estrutura e diversidade destas comunidades. / Pattern description - search for trends or non-random arrangements in communities, has a long history in plant ecology. Commonly, the structure and diversity of plant communities are described based on their size class distribution (SDD), spatial distribution (SSD) and species abundance distribution (SAD). This is because there is an underlying assumption that the existing patterns in these community descriptors are signatures of key processes determining their organization and functioning. Thus, pattern description is often the first step to generate or test hypotheses about the processes governing community structure and diversity. Organized in different chapters, the main goal of this thesis was to describe and compare different patterns in Neotropical tree communities and to generate hypotheses about the processes that regulate them. To do so we used an approach based on model selection, which was performed using maximum likelihood. In all chapters the data on tree communities came from different permanent forest plots, four of them of 10.24 ha located in Brazil and another 50 ha located in Panama. In addition to the general introduction of key concepts and techniques used along the thesis (Chapter 1), it was evaluated and compared: (i) the basic forest descriptors (i.e. abundance, basal area and species richness per subplot - Chapter 2), (ii) the SAD and how it varies with increasing sample size and scale (Chapter 3), (iii) the SDD and how it relates to species demography (Chapter 4.1 and 4.2) and, finally, (iv) the SSD and how it varies between plots (Chapter 5). Only the relationship between SDD and species demography o was performed using the Panama plot data, while the remaining chapters relate the four Brazilian plots. Each chapter used different methods and probabilistic models for the description and comparison of the variables of interest. In Chapter 2, it was found that basic forest descriptors were very different between the four Brazilian plots. These descriptors were rarely normal at the studied scales and differences between plots were more evident when variance is accounted for, which seems to provide information on processes generating within-plot variability. Chapter 3 showed that the change in shape of the SAD due to increasing scale is predominantly an indirect effect of sample size. Thus, there may be an effect of the scale, but this effect is minor and seems to depend on the degree of species turnover between samples. In Chapter 4.1, it was shown that the combination of four different probability distributions was necessary to describe the wide range of SDD, since models were rarely appropriate for the majority of tree species. In Chapter 4.2, it was found that growth and recruitment, but not mortality, shape the SDD. On average, decreasing growth-diameter curves (i.e. higher juvenile growth) were associated to less steep SDD, whereas high recruitment rates were related to steeper SDD. Although the predictions of demographic equilibrium theory were positively related to the observed SDD, there was lots of variation, making predictions quite unreliable. In Chapter 5, it was confirmed that the great majority of species had clumped spatial distributions. However, the results of intensity and size of clumps showed that the patterns of aggregation were different among plots. Species shared between two plots generally showed patterns of spatial distribution that matched the patterns found for individual plots. Therefore, the Brazilian plots presented very distinct patterns, both in terms of structure and species diversity, which were most probably determined by the environmental conditions to which these communities are subjected. Different patterns among plots were also found in respect to species spatial distribution (i.e. frequency, intensity and size of aggregation). However, the results suggested that these patterns were more connected to within-plot environmental heterogeneity than with climate and soil conditions. On the other hand, the spatial distribution of the species seems to have influenced the patterns of diversity of communities. There was not, however, a sufficient number of plots under different combinations of environmental conditions to test these suggestions or to make generalizations for each forest type, being necessary to confront these suggestions with other studies conducted in similar conditions. In the Panamanian plot, it was confirmed the theoretical expectation that diameter distributions reflect the demographics of the species, in particular the patterns of growth and recruitment. But there was great variability among species, making it difficult to infer past demographic patterns from current SDD. Finally, the analytical approach based on model fit and selection by maximum likelihood was a viable, flexible and appropriate approach, particularly in respect to the simultaneous comparison of different models and to the search for mechanisms underlying patterns. Despite some more operational limitations, the model-based approach is an appropriate alternative for the description of tree communities and can be jointly used with other approaches (e.g. hypothesis testing) for pattern description and to generate and test hypotheses on the fundamental processes that determine the structure and diversity of these communities.