File:FoodWeb.pdf

In this study we reconstruct predator-prey relationships from biomass time series of a simulated system of interacting species. To overcome the shortcomings of a static food webs representation we introduce a new model which accounts for both population and interaction dynamics. It is a derived version of the light-cone model from special relativity theory. To identify the existence of predator–prey relationships in the system we quantify the notion of distance in a food web. We use known measures from information theory, namely mutual information and transfer entropy, and we introduce a new measure based on causal states of point and patch predictors. To evaluate our results we compare the distances measured with a minimum distance measure from the underlying food web, and examine the accuracy of the measures in inferring the existence of the actual predator–prey rela- tionships. First results show that our new measure based on causal states of point and patch predictors together with the transfer entropy measure outperform the mutual information measure in terms of distance accuracy. A threshold based method to estimate adjacent links shows similar results.