We explored the predictive value of common landscape characteristics for landslide vegetative stages in the Luquillo Experimental Forest of Puerto Rico using four different analyses. Maximum likelihood logistic regression showed that aspect, age, and substrate type could be used to predict vegetative structural stage. In addition it showed that the structural complexity of the vegetation was greater in landslides (1) facing the southeast (away from the dominant wind direction of recent hurricanes), (2) that were older, and (3) that had volcaniclastic rather than dioritic substrate. Multiple regression indicated that both elevation and age could be used to predict the current vegetation, and that vegetation complexity was greater both at lower elevation and in older landslides. Pearson product-moment correlation coefficients showed that (1) the presence of volcaniclastic substrate in landslides was negatively correlated with aspect, age, and elevation, (2) that road association and age were positively correlated, and (3) that slope was negatively correlated with area. Finally, principal components analysis showed that landslides were differentiated on axes defined primarily by age, aspect class, and elevation in the positive direction, and by volcaniclastic substrate in the negative direction. Because several statistical techniques indicated that age, aspect, elevation, and substrate were important in determining vegetation complexity on landslides, we conclude that landslide succession is influenced by variation in these landscape traits. In particular, we would expect to find more successional development on landslides which are older, face away from hurricane winds, are at lower elevation, and are on volcaniclastic substrate. Finally, our results lead into a hierarchical conceptual model of succession on landscapes where the biota respond first to either gradients or disturbance depending on their relative severity, and then to more local biotic mechanisms such as dispersal, predation and competition.
Additional publication details
Predicting landslide vegetation in patches on landscape gradients in Puerto Rico