1 | INTRODUCTION
No ecosystem has escaped human alteration and a poor understanding of factors leading to species decline can result in inefficient or ineffective species restoration (Craig et al ., 2017; Roni et al ., 2018). For riverine ecosystems, major anthropogenic alterations occurred in the 20th century through dam construction, concomitant flow regulation, and establishment of nonnative fishes; all which contributed to the imperilment of many native fishes (Burkhead, 2012; Kominoski et al ., 2017; Lamothe and Drake, 2018). In certain cases, the primary cause of imperilment can be directly linked to nonnative fish predation (Vander Zanden et al . 1999; Pelicice and Agostinho, 2009; David, 2017). In other cases when there has a poor understanding of ecosystem dynamics prior to severe habitat and species alteration, researchers may find it difficult to identify initial causes of native species declines, determine the extent to which they can be reversed, and assess trade-offs in recovery efforts (Cochran-Biederman et al ., 2015; Thurstan, 2022). Such knowledge gaps can also lead to extensive management efforts such as habitat manipulation (e.g., increasing structural complexity, mimicking variable flows, and reestablishment of floodplain connection) and nonnative species control that may not necessarily restore ecosystem processes needed to reestablish imperiled species if efforts are not sufficient to address proximate causes of population declines (Booth, 2016; Albertson, 2018).
Limited trophic resource availability is a potential proximate cause of some species’ declines but changes to food-web dynamics can be difficult to isolate and quantify (Naiman et al ., 2012; Albertson et al ., 2018; Booth, 2016; Thoms and DeLong, 2018). The difficulty arises because changes in food-web dynamics can be mediated by synergistic indirect abiotic (habitat homogenization and fragmentation) and biotic (reduced prey diversity and abundance or competition with nonnative species) effects from river regulation and fish community compositional changes (Naiman et al ., 2012; Booth, 2016; Haubrock, 2018). Nevertheless, trophic resource use studies have the capacity to increase our understanding of changes in riverine food-web dynamics although there is a scarcity of unaltered systems for comparison (Turner et al ., 2015; McManamay et al ., 2018; Vander Zanden, 2003). In certain instances, species or habitat compositional gradients can be used to make comparisons but these studies are not able to detect a departure from conditions prior to severe alterations, which might identify trophic resources that supported a more stable ecological system (Haubrock et al ., 2018; Rogosch and Olden, 2020; Pennocket al ., 2021). However, with the availability of museum specimens, use of stable isotopes, and advances in Bayesian statistical analyses, methods exist to use a hypothesis-driven framework to detect departures from a historical condition and changes to food-web dynamics that may underly the continued imperilment of native fishes (Turneret al. , 2015; Alp and Cucherousset, 2022).
The Colorado River Basin is a well-known example of a riverine ecosystem whose 20th century alteration resulted in the imperilment of many native fishes, including the system’s historical top predator Colorado Pikeminnow (Ptychocheilus lucius ) (Tyuset al ., 1982; Minckley et al. 2003). In the San Juan River subbasin, Colorado Pikeminnow was nearly extirpated by the 1990s (Ryden and Ahlm, 1996; Ryden, 2003). However, 20 years of conservation management that included mimicking a more-natural hydrograph from the most upstream reservoir, intensive removal of large-bodied nonnative fishes, and a hatchery augmentation program that stocked over 5 million juvenile fish, reestablished Colorado Pikeminnow (Franssen et al ., 2016). Yet, the current Colorado Pikeminnow adult population is relatively small because age-specific survival rates of stocked fish are low (Clark et al ., 2018). A significant contribution to these low survival rates seems unlikely to be driven by nonnative fish predation as neither Colorado Pikeminnow nor more common fishes, responded positively to intensive nonnative fish removal (Franssen et al ., 2014); possibly because the most abundant large-bodied nonnative fish, Channel Catfish (Ictalurus punctatus ), is not highly piscivorous (Hedden et al ., 2021; Pennock et al ., 2021; Heddenet al ., 2022). Additionally, attempts to mimic a more-natural hydrograph have been hampered by increased aridity in the San Juan River basin (Pennock et al ., 2022).
Given piscivorous Colorado Pikeminnow likely benefit from a robust fish prey base (Vanicek and Kramer, 1969), some of the slow and limited reestablishment of the species to the San Juan River could be due to trophic deficiencies caused by river regulation and species turnover (Franssen et al ., 2007). To assess this hypothesis, we used Bayesian methods to conduct stable isotope analyses of the river’s fish community prior to and after river regulation that was coincident with species compositional changes. To isolate the effects of species turnover, we quantified changes to community-wide trophic structure metrics (diversity of assimilated basal resources, food chain length, as well as overall resource use diversity, similarity, and distribution) with and without the inclusion of extirpated and new invasives. We then calculated changes in trophic niche overlap for fish species present before and after river regulation to explore potential changes in resource competition. To better understand any changes in both community-wide trophic structure metrics and niche overlap, we quantified changes in trophic dispersion (niche breadth). Finally, we inferred specific changes in trophic resource use of Colorado Pikeminnow between time periods using an isotopic diet mixing model.