Spatial relations between lianas and trees in Yasuni National Park

Different non-stationary Poisson models were developed in order to search for the most optimal models to explain the spatial distribution of lianas based on the spatial distributions of tree taxa (as covariates). The data used come from liana and tree censuses conducted in 2008 and 2002−2003, respectively, in an area of 20 m × 500 m (1 ha) in the permanent vegetation plot of 50 ha in Yasuní National Park. Independent analyses were performed at the taxonomic level of tree species, genera and families, and also with different diameter classes of both lianas and trees. To assess the effect of different diameter classes on the quality of a model, different diameter classes for lianas (≥1.0, ≥1.5, ≥2.0 y ≥2.5 cm) and trees (≥1.0, ≥2.0, ≥3.0, ≥4.0, ≥5.0, ≥6.0, ≥7.0, ≥8.0, ≥9.0 y ≥10.0 cm) were combined in different models, resulting in 40 models per taxonomic level (total 120 models). Each model was evaluated by two indices that measure model performance: the Nash-Sutcliffe Efficiency (NSE) and the Akaike Information Criterion (AIC). In addition, a Spearman correlation was used to assess to what extent the results of a model can be correlated with liana prevalence data as observed in the field and reported in the literature (18 studies between 1984 and 2020). To select the most optimal model from the 40 models run for each taxonomic level of trees, the level of significance of the ρ coefficient of the Spearman correlation was used as a criterion. At the species level, Spearman correlations were not significant because only a few Yasuní tree species were found with liana prevalence data reported in the literature. At the genus level, the models of the tree diameter class ≥7 cm were the only that produced highly significant correlations, in particular the model that resulted when the diameter class of lianas ≥1 cm was considered (ρ = 0.718, P = 0.017). At the family level, the model with the most significant correlation was the one for the diameter class of trees ≥7 cm and the diameter class of lianas ≥2 cm (ρ = 0.628, P = 0.006). In these two models, selected for their resemblance to real liana prevalence data, estimates of the parameter θ_t, designated as θ ̂_t, were calculated for each tree taxon and their statistical significance under H0: θ_t= 0 was assessed. This estimate indicates whether a taxon shows a tendency to be or not surrounded by lianas (θ ̂_t<0 to indicate relative absence of lianas nearby the individuals of the tree taxon, and θ ̂_t>0 to indicate association of the individuals of the tree taxon with lianas). The anatomical characteristics of those taxa with a θ ̂_t significantly different from zero were evaluated in order to explain the possible ecological mechanisms behind the statistical results. In general, it was observed that those tree taxa with θ ̂_t<0 (statistically significant) do indeed display liana-deterring characteristics, while those tree taxa with θ ̂_t>0 (statistically significant) do indeed display features that may facilitate liana colonization.

• Autore principale: Chandi Villarroel, Franz Alexander (author)
• Natura: bachelorThesis
• Lingua: eng
• Pubblicazione: 2022
• Soggetti: Modelo de Poisson; Distribución espacial; Lianas; Árboles; Parque Nacional Yasuní (Ecuador); Poisson model; Spatial distribution; Trees; Yasuni National Park (Ecuador)
• Accesso online: http://repositorio.yachaytech.edu.ec/handle/123456789/510