A methodology combining molecular fingerprinting, pyrosequencing-based identification of phylotypes with PyroTRF-ID, and multivariate statistics was developed to investigate on the long run, with high resolution and low cost, the dynamics of bacterial populations involved in nutrient removal in aerobic granular sludge wastewater treatment systems. The bacterial community continuum comprised two major opponent clusters. Members of the first cluster, mainly related to Accumulibacter and Nitrospira, were favored under conditions leading to efficient nutrient removal. Members of the second cluster, mainly related to Competibacter, Sphingobacteriales, Cytophaga, and Intrasporangiaceae, proliferated during periods of low performances. The granular ecosystem was further composed of up to 30% of Xanthomonadaceae independently of operation conditions.