Time-varying flow-ecology relationships for an endangered fish population: Longfin Smelt in the San Francisco Estuary

Major estuaries globally are experiencing steep changes in hydrology and ecosystem dynamics. However, connecting alteration of river flow regimes to fish population dynamics remains a challenge, partly due to the untested assumption that flow regimes, fish dynamics, and the resulting flow-ecology relationships are stationary (i.e., with no systematic changes in mean or variance over time). Here, we studied the endangered population segment of the Longfin Smelt (Spirinchus thaleichthys) in the San Francisco Estuary, which depends on seasonal river flows to reproduce. We used extensive biomonitoring data (1980-2020) and two time-series modeling techniques, Multivariate Autoregressive State-Space models (MARSS) and Dynamic Linear models (DLM), to understand how population dynamics respond to interannual flow variation, and whether flow-ecology relationships have changed over time. MARSS outputs showed that population trajectories are best explained by a combination of lateral and vertical dimensions of habitat structure, i.e., whether individuals were collected in channels vs. shoals, and in pelagic vs. benthic environments. In turn, DLM models revealed time-varying, but often positive effects of flow on young-of-the-year abundance in shallow channel and shoal habitats, but no consistent relationships for older individuals (age-1+), likely due other drivers of survival between age-0 and age-1+. Finally, we found that inferences from the two modeling approaches converged only in about 30% of the cases. Divergence in the sign and/or magnitude of flow effects suggests that time-averaged approaches may oversimplify non-stationary relationships between the environment and fish population dynamics. From a conservation standpoint, the gradually weakening but positive flow-ecology relationship (as opposed to a step change in the relationship) suggests it may still be possible to reverse the steep population declines of Longfin Smelt through a combination of flow and habitat restoration actions. While we focused on a particular endangered population, our quantitative approach is transferable to other taxa and geographies, and could help inform management of flow-dependent resources in systems strongly affected by non-stationarity. We contend that time-varying flow-ecology relationships are needed to capture ecological realism, and could help design more effective conservation strategies in fast-changing environments.