Analyses of three-dimensional species associations reveal departures from neutrality in a tropical forest.

The study of community spatial structure is central to understanding diversity patterns over space and species co-occurrence at local scales. Although most analytical approaches consider horizontal and vertical dimensions separately, in this study we introduce a three-dimensional spatial analysis that simultaneously includes horizontal and vertical species associations. Using tree census data (2000-2016) and allometries from the Luquillo forest plot in Puerto Rico, we show that spatial organization becomes less random over time as the forest recovered from land-use legacy effects and hurricane disturbance. Tree species vertical segregation is predominant in the forest with almost all species that co-occur in the horizontal plane avoiding each other in the vertical dimension. Horizontal segregation is less common than vertical, whereas three-dimensional aggregation (a proxy for direct tree competition) is the least frequent type of spatial association. Furthermore, dominant species are involved in more non-random spatial associations, implying that species co-occurrence is facilitated by species segregation in space. This novel three-dimensional analysis allowed us to identify and quantify tree species spatial distributions, how interspecific competition was reduced through forest structure, and how it changed over time after disturbance, in ways not detectable from two-dimensional analyses alone.

Predicting the effect of habitat modification on networks of interacting species.

A pressing challenge for ecologists is predicting how human-driven environmental changes will affect the complex pattern of interactions among species in a community. Weighted networks are an important tool for studying changes in interspecific interactions because they record interaction frequencies in addition to presence or absence at a field site. Here we show that changes in weighted network structure following habitat modification are, in principle, predictable. Our approach combines field data with mathematical models: the models separate changes in relative species abundance from changes in interaction preferences (which describe how interaction frequencies deviate from random encounters). The models with the best predictive ability compared to data requirement are those that capture systematic changes in interaction preferences between different habitat types. Our results suggest a viable approach for predicting the consequences of rapid environmental change for the structure of complex ecological networks, even in the absence of detailed, system-specific empirical data.In a changing world, the ability to predict the impact of environmental change on ecological communities is essential. Here, the authors show that by separating species abundances from interaction preferences, they can predict the effects of habitat modification on the structure of weighted species interaction networks, even with limited data.