Planning for marine conservation often requires estimates of the amount of habitat needed to support assemblages of interacting species. During winter in subtidal San Pablo Bay, California, the 3 main diving duck species are lesser scaup Aythya affinis (LESC), greater scaup A. marila (GRSC), and surf scoter Melanitta perspicillata (SUSC), which all feed almost entirely on the bivalve Corbula amurensis. Decreased body mass and fat, increased foraging effort, and major departures of these birds appeared to result from food limitation. Broad overlap in prey size, water depth, and location suggested that the 3 species responded similarly to availability of the same prey. However, an energetics model that accounts for differing body size, locomotor mode, and dive behavior indicated that each species will become limited at different stages of prey depletion in the order SUSC, then GRSC, then LESC. Depending on year, 35 to 66% of the energy in Corbula standing stocks was below estimated threshold densities for profitable foraging. Ectothermic predators, especially flounders and sturgeons, could reduce excess carrying capacity for different duck species by 4 to 10%. A substantial quantity of prey above profitability thresholds was not exploited before most ducks left San Pablo Bay. Such pre-depletion departure has been attributed in other taxa to foraging aggression. However, in these diving ducks that showed no overt aggression, this pattern may result from high costs of locating all adequate prey patches, resulting reliance on existing flocks to find food, and propensity to stay near dense flocks to avoid avian predation. For interacting species assemblages, modeling profitability thresholds can indicate the species most vulnerable to food declines. However, estimates of total habitat needed require better understanding of factors affecting the amount of prey above thresholds that is not depleted before the predators move elsewhere.