Understanding patterns and mechanisms of species assemblages is a major goal in community ecology. Meta-community theory emphasizes the importance of interactions of local communities through the dispersal of potentially interacting species, motivating research to study a whole set of interacting communities across a landscape. However, it is often difficult to collect ecological data from a large number of communities at a short time scale. Recent development of environmental DNA (eDNA) techniques provide us a powerful tool to achieve this attempt at large scales. Here we used eDNA metabarcoding, a high-throughput parallel DNA sequencing of eDNA to assess species diversity and community structure of freshwater fishes in the Sado Island that is over 280 km in circumference. We sampled water from 121 independent streams (i.e., basins) that are located in three geographically different regions in the island. We show inter- and intra-regional variation in species diversity and spatial structure of freshwater fish communities to conclude eDNA metabarcoding is useful to elucidate large-scale and fine-grained community structure and dynamics of aquatic species.