This thesis investigates network properties of natural food webs. In particular, it focuses on the e ect that external disturbances have on their substructures and robustness. The importance of a network-level methodology lies in its capacity to capture entangling species interactions and identify inter-connecting properties in heterogeneous food webs. The research rst analysed the responses of freshwater food webs under the stress of drought. A core/periphery structure was detected and its relative size was found to be unchanged after drought despite a signi cant biodiversity loss. Species extinction triggered extensive link rewiring and movement of species from the core to the periphery. These results showed that the robustness was maintained indicating that the redundancy in the core can e ectively mitigate species level perturbations. Secondly, the research further examined the e ects of Genetically Modi ed Herbicide Tolerant (GMHT) management on food web properties and robustness. Network analysis showed that such change in farming practice has no signi cant impact on the agro-ecosystems. However, crop switching, a common practice in agriculture, was found to pose much more significant changes on network properties and robustness when compared to GMHT crops. Thirdly, the research examined over 50 empirical food webs and demonstrated that the relative core size is a much more e ective indicator of food web robustness than the classical ecological measure connectance, as the latter was found to be insensitive to changes in the interaction patterns. Lastly, the research established the relationships between centrality measures and species ecological and/or functional role in food webs, and how they impact on network robustness.