The organization of highly resolved, multi-habitat species interactions in an Eocene paleo-food web

Background/Question/Methods

Extending ecological research through deep time can provide an important way to understand the macroevolutionary context of the structure, function, and dynamics of today’s and tomorrow’s ecosystems. While many paleoecological studies focus on species diversity and distribution, there are also opportunities to analyze species interactions. Here, we present highly and evenly resolved food web data for the 49 million year old Eocene Messel Shale. The dataset includes species from a steep-sided maar lake and the surrounding paratropical forest, and represents a temporal window of ~10 to 100 thousand years. It is notable for the detailed resolution of terrestrial plant-insect interactions, multi-habitat representation, and high certainty of most interaction data. Using various network structure analysis techniques, including the recently developed probabilistic niche model, we assess the organization of this paleo-food web with comparisons to extant ecological networks.

Results/Conclusions

The full Messel Shale food web includes 700 trophically unique taxa and 6446 links. 54% of taxa are resolved to the genus or species level, and 82% to the family or better level. The taxa include 187 land plants, 326 invertebrates, and 143 vertebrates. The other 44 taxa include protists, fungi, and prokaryotes. 77% of links are rated as middle or high certainty. The web can be split into a terrestrial web with 633 taxa and 5551 links and an aquatic web with 94 taxa and 517 links. While levels of maximum generality are similar between the two habitats (i.e., there is a species in each habitat that feeds on ~30% of taxa), there is much greater trophic specialization in the terrestrial web as reflected in its low connectance (0.014) compared to the aquatic web (0.059). Probabalistic niche model (PNM) analysis indicates that the niche model structure of the aquatic web is within the range observed for extant webs, while the PNM performs poorly for the terrestrial web, even when it is strongly aggregated by trophic similarity. Adding a second niche dimension increases fit of the PNM for both webs. The results are very robust to the removal of low certainty links. Reasons for the poor PNM fit of the terrestrial web are considered, and potentially include factors such as its size, high resolution of specialized plant-insect interactions, differences between paleo and extant ecosystems, and methodology.