Quantifying Hypoxia Impacts on an Estuarine Demersal Community Using a Hierarchical Ensemble Approach

In coastal marine ecosystems, hypoxia and anoxia are emerging as growing threats whose ecological impacts are difficult to ascertain because of the frequent lack of adequate references for comparison. We applied conventional and hierarchical ensemble analyses to evaluate the weight of evidence in support of hypoxia impacts on local densities of individual and groups of demersal fish and invertebrate species in Hood Canal, WA, which is subject to seasonal hypoxia in its southern reaches. Central to our approach was a sample design and analysis scheme that was designed specifically to consider multiple alternative hypotheses regarding factors that dictate local species' densities. We anticipated persistent effects of hypoxia (felt even when seasonal hypoxia was absent) on species densities would be most pronounced for sessile species, but that immediate effects (felt only when seasonal hypoxia was present) would dominate for mobile species. Conventional analysis provided strong evidence that densities of sessile species were persistently reduced in the hypoxic-impacted site, but did not indicate widespread immediate density responses during hypoxic events among mobile species. The absence of strong weights of evidence for hypoxia effects was partly a consequence of alternative hypotheses that better explained spatial-temporal variation in species' densities. The hierarchical ensemble analysis improved the precision of species-specific effect sizes, and also allowed us to make inferences about the response of aggregated groups of species. The estimated mean density reductions during hypoxic events (dissolved oxygen ~2 mg/l) ranged from 73 to 98% among mobile invertebrates, benthic, and benthopelagic fishes. The large reduction in benthic and benthopelagic species suggests substantial effects of hypoxia in Hood Canal even at oxygen levels that were marginally hypoxic. Understanding the full ecological consequence of hypoxia will require a greater knowledge on the spatial extent of distributional shifts and their effects on competitive and predator-prey interactions.