The sensitivity of food web topology to temporal data aggregation

Abstract

There is a host of network measures provided by ecological network analysis in order to better understand food webs. A key question is which network indices to use in studying a particular problem, especially because the biological meaning of many indices is poorly understood. We argue that one aspect to consider when making this decision is the sensitivity of network measures to temporal data aggregation. Aggregating food web data in time is not a matter of choice: to some extent, every food web is temporally aggregated. We present a simple survey on how five global network measures behave when aggregating food web data in three time series. We study connectance ( C), average distance (AD) and the network centralization index for degree centrality (NCI DC), closeness centrality (NCI CC) and betweenness centrality (NCI BC). These indices describe the richness (C) and arrangement (the other four) of links in food webs. We focus on differences between indices, not between ecosystems or seasons. The first question is how variable the values of particular indices are in the time series. NCI BC is the most variable index, possibly reflecting changes in the relative role of species (components) during the year. AD is the most constant index suggesting not much change in the average speed of spreading indirect effects in the studied networks. Some indices, especially average distance (AD) are highly sensitive to temporal aggregation: the value of the aggregated (cumulative) network can be far from the range of seasonal values. Using an index like this for aggregated data is clearly misleading. High quality, comparable food web time series are not very common. There is still much work to be done in describing and collecting them in order to gain integrative and comparative knowledge on the behavior of particular indices during temporal food web aggregation. Doing so, progress in network analysis could compensate for the methodological difficulties of network description.