Contribution of microcosm experiments to conservation science.

Microcosms, or miniature experimental systems, have been used to develop models and theories in ecology. However, their contribution to conservation science is unclear. We explored the application, design, and impact of microcosms in conservation science from 469 systematically identified articles published from 1986 to 2023. We used generalized linear modeling to compare cumulative citations over time for each microcosm article with those of 71,738 nonmicrocosm articles in conservation science. We also surveyed the proportion of microcosm articles and nonmicrocosm articles focused on conservation science that were cited in policy documents. Two types of microcosms were used in conservation research: generalized microcosms (i.e., simplified analogies of systems used to test general theories and mathematical models) and specialized microcosms (i.e., recreations of specific ecosystems or species assemblages that test specific hypotheses). Microcosms were used to study biodiversity, invasive species, extinction, pollution, and climate change and were applied to a broad array of ecosystems and species. Microcosm experiments tended to be small (systems were liters in size or smaller) and conducted over short periods (weeks or months); could monitor study species for up to hundreds of generations; and had high sample replication. On average, microcosm studies were cited up to twice as often as nonmicrocosm studies 25 years after publication. Microcosm articles and nonmicrocosm articles focused on conservation science were cited in policy documents at similar rates to each other. We recommend that conservation science, which often focuses on urgent topics and rare or threatened ecosystems and species, may benefit from the manipulability and replicability that microcosms offer. Microcosm experiments may also be low risk for the study systems involved. Future uses of microcosms include providing experimental evidence and testing of conservation theories, models, and hypotheses.