Phytoplankton blooms are complex ecological events that emerge from the dynamics of an entire ecosystem. Increasing efforts to forecast blooms are hampered by inconsistent definitions of what constitutes a bloom event, from both conceptual (mechanistic) and empirical (quantitative) perspectives. By clarifying definitions of blooms using temporal system dynamics, we propose to target modeling and forecasting methods to appropriate settings, and generate testable ecological hypotheses into the underlying processes fueling bloom development. Here, we present a general bloom definition that highlights both growth and loss processes, and identify quantitative metrics of time-series structure associated with several subclasses of blooms. We hypothesize ecological processes that are consistent with these time-series structures, and that suggest promising approaches for forecasting different types of blooms.