The discovery of anaerobic ammonium oxidation (anammox) process provides us not only a novel sub-pathway of global nitrogen cycle, but also a promising technique for sustainable autotrophic nitrogen removal, in particular with a granule-based partial nitritation-anammox (PNA) system. However, little is known whether size-fractionated granules shape different microbial communities, and the importance of granular sludge as a micro-level model system for microbial ecology has also been largely overlooked. This study aims to dissect microbial communities and compare diversity differences in size-fractionated anammox-based granules. Granules from a PNA reactor were classified into five different size samples (i.e. <0.2, 0.2-0.5, 0.5-0.8, 0.8-1.0, >1.0 mm) and characterized by using 16S rRNA gene high-throughput sequencing. Global microbial communities and unique species in the size-fractionated granules were revealed, with an emphasis on functional microorganisms involving in nitrogen conversion. The results showed, for the first time, that larger granules harbor a higher microbial biodiversity, compared to smaller ones. The sequencing indicated that the abundance of anammox bacteria (dominated by Candidatus Brocadia) increased with the increasing granule size (from <0.01% in the smallest granules to 19.4% in the largest granules). In contrast, the abundance of ammonia-oxidizing bacteria (AOB, Nitrosomonas) decreased from 10.6% to 4.7% with the increasing granule size. Moreover, larger granules harbor more diverse anammox bacteria, with 4 anammox genera found in the largest granules but only 2 in the smallest ones. It is implied that, to achieve a better functional stability, an engineered ecosystem (like PNA process) should hold more larger granules for both a higher microbial diversity and a greater functional redundancy. The findings suggest a granular size of >1.0 mm provided favourable niches for the consortium of AOB and anammox bacteria, enabling partial nitritation and anammox reactions in the same granules. They improve our understanding of granules as a micro-level ecological system and guide optimisation of the autotrophic nitrogen removal process, together with metagenomics, metatranscriptomics and metaproteomics to be presented later.