COS 201-4 - Meta-analysis of metabolic rates in three anadromous fishes reveals variation in scaling and functional forms across mass, temperature, and physical activity .

Abstract: Climate change and other anthropogenic stressors are increasing metabolic strain on fishes worldwide. Anadromous fishes are of particular concern as droughts, temperature shifts, and damming make seasonal migration upriver to spawning habitat increasingly difficult. Bioenergetics models of riverine fishes are useful tools for understanding energy budgets and are used within modeling frameworks for research, conservation and/or management purposes. Calibrated from experimental respirometry, these models predict energy used by fish to grow, migrate, reproduce, or otherwise interact with their surroundings. Recent work suggests higher diversity in metabolic scaling relationships across species and across life stage than previously thought, and metabolic data are not always sufficient for parameter calibration across all species and life stages of concern. These two issues can cause misleading results when researchers expropriate parameters or functional forms from one species/life stage to another. We performed a meta-analysis of three anadromous, ESA listed fishes in North America: Chinook salmon (Oncorhynchus tshawytscha), steelhead (Oncorhynchus mykiss) and green sturgeon (Acipenser medirostris) and used three modeling methods to predict energy consumption as a function of mass, swim speed, and temperature. We compared the Wisconsin (WI) Model with generalized linear models (GLM) and generalized linear mixed-effects models (GLMM) to test whether allowing for greater flexibility in metabolic relationships and accounting for variation across experimental study can improve the accuracy and precision of bioenergetics models across ontogeny. We find that the scaling parameter values, functional forms, and interaction terms vary across the best-performing models for each species investigated. In green sturgeon, all three model types were similar in precision and accuracy. However, in the two salmonid species, the GLM and GLMM outperform the WI Model in this regard, particularly in pre-smolts. The WI Model tended to overestimate metabolic rates for Chinook pre-smolts by about 128% and underestimated metabolic rates of pre-smolt steelhead by about 38%. The improved model performance we observe upon inclusion of interaction terms between parameters may reflect shifts in metabolic scaling across ontogeny. Moreover, the inclusion of a random effect for study helped to account for inherent variation across respirometry experiment setup and equipment. We show that the WI Model can introduce significant error in metabolic rate estimation when fit across the entire fish lifespan, but by accounting for variation between studies and across physiological shifts with ontogeny, bioenergetics models can be made a more accurate and precise tool to inform conservation and habitat restoration projects.