MANUSCRIPT

TITLE: Two different diversity patterns promote new insights into tree community assembly and diversity maintenance in a Central Amazonian Forest

Community structure, defined as species composition and abundance, is affected by at least three kinds of factors: species-dependent factors (SDF), trait-dependent factors (TDF) and stochastic factors (SF). SDF are species-specific factors like specialized pathogens, seed predators ou herbivores. These kind of factors are supposed to limit population density of the most abundant species, as described by the Janzen-Connell hypothesis. TDF are insensitive to species identity (as opposed to SDF), but act similarly in species sharing the same functional trait values. Examples of SDF are droughts, that may kill plants with broader or thinner leaves, climate, that may sort species with higher or lower potential growth rate, or even herbivory, that may affect more severely species with more nutritive and less tougher leaves. Finally, SF are those factors that are independent of species identity or traits, like debris fall damage, lightning or eventual attacks from natural enemies.

Several aspects of plant life-history were considered: leaf traits (specific leaf area, leaf size, leaf dry matter, leaf nitrogen and phosphorus content, leaf tensile strenght), a whole plant trait (maximum height), a stem trait (wood density) and regenerative traits (dispersal mode and seed size). Traits were collected following Cornelissen et al. (2003), with slight modifications in some cases (petioles and raquis were not considered when calculating SLA, LS and LDMC).

Stochasticity is implicit in most models.

• J-C model: species populations are not allowed to grow more than a certain limit (L). This is a stochastic model.
• Dominance-decay model: the greater the dominance of a species, the greater the chance of a invasion by a new species. This may be a stochastic (Power-Fraction model with k>1) or a deterministic model (original Dominance-Decay model).

• Competitively limited communities model: all plants use exactly the same resources; they may also use them similarly, in which case there will be a great competition among these plants. So, the more similar two species are, the less probable is their coexistence. This model predicts a community with great functional diversity.
• Environmental filtering model: habitat conditions lead to the selection of species with determined trait values. This model predicts a community with poor functional diversity.
• Mixed model: both environmental filtering and competition takes place at the same time.

• MacArthur-fraction model: total abundance is ramdomly distributed among species. This is a pure stochastic model. {nam_2.0.r|macfra}}

Bayesian approach: each model has a probability distribution for each species abundance ranking position; for example, model 1 has a normal probability distribution with mean = 20 and sd = 3 for the first position (most abundant species in the community), a normal probability distribution around 10 for the second position and so on. Given the observed data (abundance of of the first most abundant species, abundance of the second most abudant species and so on), we will calculate the likelyhood of each model being correct as the combined probability of each predicted ranking position abundance being correct.