Although nutrient supply often underlies long-term changes in aquatic primary production, other regulatory processes can be important. The Sacramento-San Joaquin River Delta, a complex of tidal waterways forming the landward portion of the San Francisco Estuary, has ample nutrient supplies, enabling us to examine alternate regulatory mechanisms over a 21-yr period. Delta-wide primary productivity was reconstructed from historical water quality data for 1975-1995. Annual primary production averaged 70 g C m Ϫ2 , but it varied by over a factor of five among years. At least four processes contributed to this variability: (1) invasion of the clam Potamocorbula amurensis led to a persistent decrease in phytoplankton biomass (chlorophyll a) after 1986; (2) a long-term decline in total suspended solids-probably at least partly because of upstream dam construction-increased water transparency and phytoplankton growth rate; (3) river inflow, reflecting climate variability, affected biomass through fluctuations in flushing and growth rates through fluctuations in total suspended solids; and (4) an additional pathway manifesting as a long-term decline in winter phytoplankton biomass has been identified, but its genesis is uncertain. Overall, the Delta lost 43% in annual primary production during the period. Given the evidence for food limitation of primary consumers, these findings provide a partial explanation for widespread Delta species declines over the past few decades. Turbid nutrient-rich systems such as the Delta may be inherently more variable than other tidal systems because certain compensatory processes are absent. Comparisons among systems, however, can be tenuous because conclusions about the magnitude and mechanisms of variability are dependent on length of data record.Phytoplankton primary productivity in lakes, estuaries, and the ocean plays an essential role in element cycling, water quality, and food supply to heterotrophs (Cloern 1996). Although we implicitly recognize primary productivity as a time-varying process, much of our effort to measure and understand this variability has focused on time scales of 1 yr or less. How much does annual primary production vary from year to year or over periods of decades, and what are the underlying mechanisms of variability at these longer time scales? These time scales are of particular interest from a practical viewpoint: they span the period over which we must separate anthropogenic influences from natural variability in order to understand the effects of our current use of water resources. Long-term studies of annual primary production in individual systems can also help us to understand