a model selection was conducted (Anderson, 2008). Linear mixed effects models (Pinheiro and Bates, 2000) were adjusted using the environmental variables: organic matter, interstitial salinity, fine sediment fraction, to each one of structural variables: mean height, canopy height, basal area and density, in all cases the estuary was used a random variable. The models that we build were (i) null, (ii) models with a single envi- ronmental variable, (iii) models with two variables (all combina- tions) and (iv) a model with the three variables. In models with two or three variables, test for interactions between independent var- iables were performed. We used the Akaike Information Criterion (AIC) and Akaike Information weight (AIC weight) to select the best model (Anderson, 2008), models with the lowest value of AIC were considered the best models but we also considered models with less than 50% of AIC weight of the best model as equiprobables. In the cases where more than one model were considered equiprob- able, they were subjected to stepwise deletion of variables to ach- ieve the significance of each term in increase the total deviance (Crawley, 2007), the minimum adequate mode models