Zooplankton predator-prey network relationships indicates the saline gradient of coastal lakes. Machine learning and meta-network approach

The main purpose of the research was to determine the differences in the network relationships among zooplankton predator and prey taxa biomasses as a premise for dividing coastal lakes according to the brackish-freshwater gradient. Furthermore, the significance of these divisions in the overall interspecies interactions of the zooplankton networks of these waters was demonstrated. To solve problems with multidimensional and highly interactive character of the data, we used machine learning and meta-network analysis models. The data analysis methods comprised the support vector machines (SVM) multidimensional modelling, structural equation modelling (SEM), and graph network analysis. The zooplankton taxa biomasses showed high accuracy with lake classification on mesohaline, oligohaline and freshwater classes. We showed two interspecies relationships in coastal lakes in the zooplankton taxa predator-prey Cladocera and Copepoda - Rotifera assemblage, which variability of taxa biomass indicated more saline or more freshwater lake classes. Biomass increases in the predators Cyclopoida copepodites, B. longirostris, D. cucullata, and the prey K. cochlearis were associated with the freshwater class. However, increased Calanoida nauplii and his prey K. cochlearis tecta biomasses suggested that they belong to the more saline lake class. In the case of decreases in the biomass of these taxa, the preference would shift to the opposite of the above. Lake salinity classes differed in the role they had on the connections of the zooplankton taxa generalized meta-network of the South Baltic coastal lakes. The freshwater and mesohaline classes were the most important for the organization of the interspecies network, while the oligohaline class had the poorest connections among taxa and showed a marginal role in the integrity of the coastal lakes' zooplankton meta-network.

Automated summary

The research aimed to determine differences in network relationships among zooplankton predators and prey, showing the importance of dividing coastal lakes according to a brackish-freshwater gradient in overall interspecies interactions. Machine learning and meta-network analysis models were used to solve issues with multidimensional data, with high accuracy in lake classification based on zooplankton taxa biomasses. The study highlighted two interspecies relationships in coastal lakes and demonstrated the role of lake salinity classes on connections within the zooplankton meta-network.