Efeitos da inclusão de grupos ecológicos e alométricos em índices de competição para expressar o crescimento diamétrico em floresta ombrófila mista
Resumo
Resumo: A pesquisa aborda o desafio de modelar o crescimento de florestas naturais, destacando a complexidade de lidar com dados de florestas com alta diversidade de espécies. Para isso, frequentemente são utilizados modelos a nível de árvore individual que estimam o crescimento das árvores a partir de índices de competição, medidas essas que resumem a influência das árvores vizinhas sobre uma árvore alvo. Esses modelos também são comumente utilizados em conjunto com a formação de grupos de alométricos ou de espécies, que tentam gerar conjuntos de observações mais homogêneas entre si. No entanto, o efeito competitivo que cada grupo exerce sobre outro ainda é pouco explorado, e poucas são as tentativas de inserir diretamente a informação do grupo no índice de competição. A pesquisa destaca os potenciais ganhos em realizar tal operação por meio das hipóteses de que a inclusão dos grupos melhora o desempenho dos índices de competição e fornece poder interpretativo para o efeito, e testa também se os grupos possuem diferentes relações ou efeitos entre si. O objetivo da pesquisa foi quantificar a competição individual entre árvores por meio de índices de competição dependentes e independentes da distância adaptados para incluir informação de grupos de espécies. A pesquisa foi conduzida na área experimental situada no campus Jardim Botânico da Universidade Federal do Paraná. A área total do fragmento de Floresta Ombrófila Mista (FOM) possui 12,96 hectares, porém foi selecionada somente a porção central, a fim de evitar efeito de borda nos índices de competição, o que totalizou uma área efetiva de 6,25 hectares de floresta. Toda a área foi alvo de censo, com medição inicial em 2006 e remedida a cada três anos, onde se mensurou o diâmetro a 1,3 m acima do solo (DAP) das árvores igual ou maior que 10 cm e coordenada das árvores. Foram utilizados os índices de competição independente de distância Basal Area in Larger Trees (BAL) e os índices dependentes de distância de Hegyi e de Tomé e Burkhart. Foram testados também dois métodos de agrupamento. O primeiro foi feito a partir da classificação ecológica das espécies em Pioneira, Secundárias e Climácicas, com a inclusão de um grupo para as espécies exóticas presentes no local, especialmente Hovenia dulcis Thunberg, Ligustrum lucidum W. T. Ait., e Eriobotrya japônica (Thunb.) Lindl. O segundo foi realizado a partir de uma relação alométrica entre a árvore alvo e as vizinhas a partir do diâmetro quadrático médio das árvores circundantes. A inclusão desses grupos nos índices de competição foi feita a partir da decomposição do índice via combinação linear, e a cada grupo foram atribuídos pesos obtidos a partir da maximização da correlação entre índice de competição ponderado e incremento diamétrico. A maximização foi realizada sob restrição, o que garante interpretação dos pesos encontrados. Após a ponderação, o diâmetro futuro foi estimado como função dos índices de competição utilizando um modelo de variâncias heterogêneas e uma função de ligação log do inverso do diâmetro. Os resultados mostraram os grupos podem ter diferentes níveis de influência na vizinhança. O índice independente de distância é mais afetado pela proporção de ocorrência dos outros grupos na vizinhança, ou seja, o grupo mais presente possivelmente resultará no mais influente. Há efeito também de árvores de diferentes dimensões na vizinhança. As menores árvores são pouco influenciadas pelas árvores maiores e vice-versa. O modelo heterocedástico foi oportuno para capturar a variabilidade entre grupos, e os índices ponderados utilizados foram capazes de predizer os diâmetros erro médio de 23%. Abstract: The research addresses the challenge of modeling the growth of natural forests, highlighting the complexity of dealing with data from natural forests. For this, models at the level of individual trees are often used, which estimate tree growth based on competition indices, measures that summarize the influence of neighboring trees on a target tree. These models are also commonly used in conjunction with the formation of allometric or species groups, which attempt to generate more homogeneous sets of observations among themselves. However, the competitive effect that each group exerts on another is still little explored in the field, and there have been few attempts to directly insert group information into the competition index. The research highlights the potential gains in performing such an operation through the hypotheses that the inclusion of groups improves the performance of competition indices and provides interpretative power for the effect, and also tests whether groups have different relationships or effects among themselves. The objective of the research was to quantify individual competition between trees through competition indices dependent and independent of distance adapted to include species group information. The research was conducted in the experimental area located at the Botanical Garden campus of the Federal University of Paraná. The total area of the Mixed Ombrophilous Forest (MOF) fragment is 12.96 hectares, but only the central portion was selected to avoid edge effects on competition indices, totaling an effective area of 6.25 hectares of forest. The entire area was surveyed, with initial measurements taken in 2006 and remeasured every three years, where the diameter at 1.3 m above ground level (DBH) of trees equal to or greater than 10 cm, and tree coordinates were measured. The competition indices independent of distance Basal Area in Larger Trees (BAL) and the dependent distance indices of Hegyi and Tomé and Burkhart were used. Two clustering methods were also tested. The first was based on the ecological classification of species into Pioneer, Secondary, and Climactic, with the inclusion of a group for exotic species present on-site, especially Hovenia dulcis Thunberg, Ligustrum lucidum W. T. Ait., and Eriobotrya japonica (Thunb.) Lindl. The second was based on an allometric relationship between the target tree and its neighbors based on the mean quadratic diameter of the neighborhood. The inclusion of these groups in competition indices was done by decomposing the index via linear combination, and weights were assigned to each group obtained from maximizing the correlation between weighted competition index and diameter increment. The maximization was performed under constraint, ensuring interpretation of the weights found. After weighting, future diameter was estimated as a function of competition indices using a heterogeneous variances model and a log inverse diameter linking function. The results showed that groups may have different levels of influence in the neighborhood. Distance-independent indices are more affected by the proportion of occurrence of other groups in the neighborhood, meaning that the most present group will possibly result in the most influential one. There is also an effect of trees of different dimensions in the neighborhood. Smaller trees are minimally influenced by larger trees and vice versa. The heteroscedastic model was suitable for capturing variability between groups, and the weighted indices used were able to predict diameters with an average error of 23%.
Collections
- Teses [387]