Comparing ecological networks along gradients

Abstract. There has been a recent explosion not only of food web data – as there has been across network science – but of food webs along gradients: data sets consisting of tens or even hundreds of food webs of the same type, across gradients like latitude, altitude, or characteristics of the habitat. These large data sets of relatively small networks (roughly 10-1000 nodes) demand new automated methods of comparing networks. This brings together two traditionally distinct fields: statistical inference, and network comparison. Work on network comparison traditionally focuses on comparing two networks at a time, whereas these gradient data sets demand comparison of tens or hundreds of networks all against one another. Traditional work in statistical inference works on inferring one or more numbers, given some other numbers; in our setting, we want to infer not a number from a gradient value, but a network structure from a gradient value. This working group brings together two food web ecologists and two computer scientists to develop such network comparison techniques, and apply them to real food web data.