Food web robustness depends on the network type and threshold for extinction

Species loss in ecological communities can trigger cascading extinctions, the extent of which likely depends on network type and extinction thresholds. Traditionally, network responses to node removal are analysed using unweighted food webs, ignoring interaction strengths and extinction thresholds. Here, we examine how food web robustness varies with network type (unweighted versus weighted), extinction thresholds, and species removal sequences, and explore how network properties - connectance and relative ascendency both unweighted and weighted - predict robustness. First, our results show that network robustness, measured by the R50 index, can be up to 40% lower in weighted networks compared to unweighted ones. Additionally, incorporating extinction thresholds reveals a consistent reduction in robustness when species deletions proceed from the highest to the lowest species degree or sum of link weights. This suggests that measures of robustness that do not include extinction thresholds overestimate of ecological network robustness. Furthermore, it highlights that species with high energy through-flow are crucial for maintaining energy pathways and network integrity in weighted food webs, emphasizing their importance in a conservation context. Second, relative ascendency emerged as the strongest predictor of food web robustness, providing the clearest temporal and ecological signals related to changes in energy fluxes. This metric reflects both link distribution (skewness) and pathway architecture (energy flux constraints), underscoring the importance of these network properties in assessing food web stability. Therefore, these properties should be considered in ecosystem management recommendations.