Savannas are tropical formations in which trees and grasses coexist. According to the "honeycomb rippling model", inter-tree competition leads to an effect of trees growing and dying due to competition, which, at fine spatial scale, would resemble honeycomb rippling. The model predicts that the taller the trees, the higher the inter-tree distances and the evenness of inter-tree distances. The model had been corroborated in arid savannas, in what appears to be caused by uneven distribution of rains, but had not yet been tested in seasonal savannas, such as the cerrado, which could be caused by the irregular occurrence of fire.A basic assumption of the model is that strong inter-tree competition affects growth (estimated by height) and mortality (estimated by inter-tree distances). As a first step towards testing this model in the cerrado, we tested this assumption in a single cerrado patch in southeastern Brazil. We placed 80 quadrats, each one with 25 m², in which we sampled all shrubs and trees. For each individual, we measured its height and the distance to its nearest neighbour - the inter-tree distance. We did not find correlations between tree height and both inter-tree distances and evenness of inter-tree distances, refuting the honeycomb rippling model. Inter-tree distances were spatially autocorrelated, but height was not. According to our results, the basic assumption of the model does not apply to seasonal savannas. If, in arid savannas, rainfall events are rare and unpredictable, in seasonal savannas, the rainy season is well-defined and rainfall is considerable. We found horizontal structuring in the community, which may be due to soil nutrient heterogeneity. The absence of vertical structuring suggests that competition for light among adult trees is not as important as competition for nutrients in the soil. We tested the basic assumption of the model in a single patch and at a single moment. To test the model effectively, we suggest this assumption to be tested in many patches over time.
competition; honeycomb rippling model; patch-dynamics model; savanna; spatial autocorrelation
