Characterizing topology of ecological networks along gradients: The limits of metrics' standardization

Species interact in nature to form complex ecological networks. There has been a rising interest in recent years to characterize the topology of such networks along various gradients (e.g. successional, climatic, elevational) to better understand how they assemble in space and time. However, to compare structure of networks that vary in size, shape and connectance, topological metrics need to be standardized (as most metrics covary with such network attributes). Traditionally, this has been done by transforming network metrics into z-scores prior comparisons. Here, I show that such standardized metrics are not independent of basic network properties such as connectance. Instead, I found that there was a consistent tendency for z-scores to approach 0 when connectance progressively decreased and approached its minimal value. This is probably due to the reduced null space available for null models to randomize interactions at such low connectance. I discuss ways to circumvent the problem in future studies.