Pest outbreaks, harmful algal blooms, and population collapses are extreme events with critical consequences for ecosystems, highlighting the importance of deciphering the driving ecological mechanisms underlying extreme events. By combining the generalized extreme value (GEV) theory from statistics and the hypothesis of a resource-limited metabolic restriction to population abundance, we evaluated theoretical predictions on the size-scaling and variance of extreme population abundance. Phytoplankton data from the L4 station in the English Channel showed a negative size scaling of the expected value of maxima, whose confidence interval included the predicted metabolic scaling (a = –1). We showed a humped pattern in variance with maxima at intermediate sizes. These results are consistent with the bounded abundance of small-sized populations that are subjected to strong grazing and with the expected decrease in variance towards large sizes. This approach provides unbiased return times, thereby improving the prediction accuracy of the timing of bloom formation, and describes a coherent framework in which to explore extreme population densities in natural communities.