Six general ecosystem properties tend to be more intense in biogeochemical cycling networks than in trophic webs
Network analysis has revealed several whole-network properties hypothesized to be general characteristics of ecosystems. These include pathway proliferation, network non-locality, network homogenization, network amplification, network mutualism, and network synergism. Collectively, these properties characterize the impact of indirect relationships among the ecosystem elements that result from the direct transactions of energy and matter. While all ecosystem networks trace a thermodynamically conserved unit through the system, there appear to be several classes of ecosystem network models with distinctive features. For example, trophic networks (TRO) are built around a food web, usually trace energy or carbon, and are the most abundant models in the literature. Biogeochemical cycling networks (BGC) typically trace nutrients like nitrogen or phosphorus. BGC networks tend to have more aggregated nodes, more abiotic compartments, less dissipation, and more recycling than TRO. We tested the hypothesized generality of the properties in BGC networks and that they tend to be more strongly expressed in BGC networks than in the TRO networks due to increased recycling. We compared the properties in 22 biogeochemical and 57 trophic ecosystem networks from the literature using enaR. We also evaluated the robustness of these results using a Monte Carlo uncertainty analysis based on inverse linear modeling.
The results generally support the hypotheses. First, pathway proliferation, network non-locality, network homogenization, network amplification, network mutualism, and network synergism occurred in varying degrees in all 22 BGC models, while network mutualism occurred in 86% of the models. Further, these results were generally robust to a 50% uncertainty in the model parameters; only one of the 95% confidence intervals overlapped with the existence thresholds. Second, the average network statistics for the 6 properties tended to be greater in the BGC models than the TRO models. A non-parametric Wilcoxan rank sum test showed a statistically significant difference between BGC and TRO models for all metrics. These results (1) confirm the general presence of these properties in ecosystem networks, (2) highlight the significance of different model types in determining the property intensities, (3) reinforce the importance of recycling, and (4) provide a set of indicator benchmarks for future systems comparisons. Further, this work highlights how indirect effects distributed by network connectivity can transform whole-ecosystem functioning, and adds to the rapidly growing domain of network ecology.