How to assess properties of large ecological networks with partial data?

: The links-species relationship, its relevance and biases

Résumé

Ecological networks are mathematical tools allowing represent the interspecific interactions taking place in ecosystems. They consist of S nodes (species) and L links (interactions between species). These two fundamental metrics are bound by the so-called "link-species relationship": when S increases, so does L. This approach was built on previous work: in 1990, Warren suggested that the L~S relationship remains the same in a network despite variations in time and space. If these hypotheses are correct, it could lead to a simplification of sampling methods, as some properties of large networks could be inferred from smaller ones. In this paper, we tested and confirmed these two hypotheses by analyzing 32 real ecological networks and 1,163 subwebs. We also showed how to incorporate bias induced by sampling into the L~S relationship. Lastly, we compared various methods with these results to build large metaweb from small subwebs and show how they impact the metaweb connectance (probability for a link to exist), which is a fundamental property that drives several other properties (e.g. modularity or robustness). Taken together, our results show that the b parameter is a pertinent and useful fundamental property of ecological networks that could be used to infer properties of larger networks if biases around its approximation are taken into account.

la date de réponse	2024
langue originale	Anglais
L'institution diplômante	Universite de Namur Université Catholique de Louvain
Superviseur	Camille Carpentier (Promoteur) & Frederik De Laender (Promoteur)