The potential for theory to outstrip data has always been a risk for food web research. Early advances in the field were criticised for the poor quality of the empirical data on which they were based. In particular, arguments were raised over the effects of sampling on patterns observed in the architecture of food webs. In response to these critiques, research focused on improving the data and a second-generation of more resolved empirical food webs have been constructed.
However, concerns about data quality have not disappeared. Studies where an attempt has been made to resolve every possible link reveal that even in what are regarded as among the best sampled food webs, the diets of many of the species are only partially resolved, for example the number of samples of a consumer species is often found to be a good predictor of its diet breadth (Woodward et al 2010). Such under-sampling could have significant implications for the patterns observed amongst food webs, for example the relationship between the size and complexity of a web.
Results/Conclusions
In order to investigate the potential effects of sampling on food webs we use a simulation that samples computer-generated webs of differing characteristics. We can then examine how sampling effort might affect perceived web structure, so as to address whether i) the relationship between connectance and species richness is due to sampling effects ii) we can differentiate possible web structures in the face of incomplete information.
We find that size-complexity relationship is highly sensitive to sampling effort, and at realistic sampling efforts one is unable to distinguish between a scenario with constant connectance with increasing species richness and one with declining connectance. However examining the sampling curves might provide us with an indirect way of inferring the underlying relationship. This work highlights the importance of considering sampling effects when analysing patterns among food webs.