Climate change is shifting the timing of organismal life-history events. Although consequential food-web mismatches can emerge if predators and prey shift at different rates, research on phenological shifts has traditionally focused on single trophic levels. Here, we analyzed >2000 long-term, monthly time series of phytoplankton, zooplankton, and fish abundance or biomass for the San Francisco, Chesapeake, and Massachusetts bays. Phenological shifts occurred in over a quarter (28%) of the combined series across all three estuaries. However, phenological trends for many (~29%-68%) taxa did not track the changing environment. While planktonic taxa largely advanced their phenologies, fishes displayed broad patterns of both advanced and delayed timing of peak abundance. Overall, these divergent patterns illustrate the potential for climate-driven trophic mismatches. Our results suggest that even if signatures of global climate change differ locally, widespread phenological change has the potential to disrupt estuarine food webs.

We compiled and filtered long-term data for fishes, zooplankton, phytoplankton, temperature, and salinity from multiple monitoring surveys in the San Francisco, Chesapeake, Massachusetts Bays. For San Francisco Bay, we obtained fish and water quality data from the California Department of Fish and Wildlife’s San Francisco Bay Study (CDFW 2023a), and zooplankton and phytoplankton data from the Environmental Monitoring Program (CDFW 2023b). For Chesapeake Bay, we obtained fish and water quality data from the Virginia Institute of Marine Science’s Chesapeake Multispecies Monitoring and Assessment Program (Latour et al. 2023); and zooplankton, phytoplankton, and associated water quality from the Chesapeake Bay Program (CBP 2023). For Massachusetts Bay, we obtained phytoplankton, zooplankton, and water quality data from the Massachusetts Water Resources Authority’s Water Column Monitoring Program (MWRA 2023). All data were monthly, spanned at least 10 years, and were examined for quality.

Analyses were performed using R. The code is provided.