Unfeasible expectations: why simple predictors outperform structural stability measures for understanding community assembly

Understanding what determines community assembly and disassembly in a changing environment is a core challenge for ecology. Recently a family of structural stability approaches that determine the range of intrinsic growth rates compatible with system feasibility have been gaining popularity as a measure of how likely a community is able to persist in fluctuating conditions. This offers a theoretical basis for understanding and predicting the assembly and stability of complex multi-species communities from only interaction network structures. However, here I show that the high sensitivity of calculations of the feasibility domain, coupled with empirical uncertainties inherent in estimated interaction strength, are likely to preclude the approachs reliable application to empirical settings as a metric to compare the stability of different communities. Across four reanalyses of previous empirical demonstrations of the structural stability approach, more parsimonious measures based on species connectance provide better explanations for patterns of community assembly or differences in stability. Calculation of structural stability metrics therefore appears to lose, rather than synthesise, information embedded in empirical interaction matrices. This success of simpler measures is good news for the purposes of prediction and emphasises the value of multiple-competing hypotheses in validation tests to demonstrate value-added associated with new approaches.