Mercury (Hg) contamination of riparian communities and of environmental compartments of the Amazon can be directly related to the occupation of the territory. The objective of this study was to identify the characteristics of aquatic environments that are associated with high levels of Hg in ichthyofauna. Our research aimed at determining the influence of variables related to fish ecology, types of aquatic environment, fishing activities by local riparian populations, and watershed use on the levels of contamination of ichthyofauna. Six sites were sampled during two distinct periods of the hydrological cycle: at the beginning of descending waters and during low waters. We focused on ten dominant fish species representing four trophic levels: Curimata inornata, Geophagus proximus, Schizodon vittatum, Leporinus fasciatus, Anostomoides laticeps, Hemiodus unimaculatus, Caenotropus labyrinthicus, Hoplias malabaricus, Plagioscion squamosissimus, Acestrorhynchus falcirostris. The study sites, which included lotic and lentic habitats, are exploited year-round by local riparian communities. Spatial variations in Hg contamination in ichthyofauna were determined by factorial analysis of variance taking into account fish diets, seasons and sampling sites. Multiple regressions were used to check the influence of ecological and anthropogenic variables and variables related to watershed uses, on Hg levels in key species representing the four trophic groups. Each variable was checked independently. Next multiple regressions were used to verify the concomitant influence of selected variables. The models listed are only those with R2> 0.50, and those strongly correlated with variations in mercury concentrations. Independently of the study site and phase of the hydrologic cycle, fish Hg contamination followed the trend piscivores > omnivores > herbivores > detritivores. In all aquatic study sites, Hg levels measured in predatory species were often higher than the 500 ng/g fresh weight threshold. Mean Hg levels in key species were significantly higher during descending waters in lotic environments, and during low waters in lentic environments. Data from this study demonstrated that simple models based on watershed use and on easily obtained variables such as suspended particulate matter (SPM) load and SPM Hg, number of inhabitants, habitat types and hydrological cycle enable very good prediction of Hg levels in fish. Our cartographical data clearly showed that the watershed site with the highest aquatic vegetation cover (6% of the open water body) and by the lowest forest cover (62% of the land) corresponded to the highest Hg concentrations in fish. Conversely, the watershed site with 94% forest cover and 1% aquatic vegetation corresponded to the lowest levels Hg concentrations in fish. These results suggest that land uses of watersheds play a key role in the level of Hg contamination of local ichthyofauna.