Atmospheric correction is a crucial procedure to derive physical parameters from satellite images. Since water reflectance is typically low, inland water studies require an efficient removal of the atmospheric signal for consistent estimates of water-leaving reflectance. In this paper, we contribute to the quality assessment of atmospheric correction approaches applied to Landsat 8 OLI (Operational Land Imager) scene. Three approaches were assessed over Amazon floodplain aquatic systems: 6SV model using MODIS (Moderate Resolution Imaging Spectrometer) atmospheric products, Acolite model and Landsat surface product of LASRC (Landsat Surface Reflectance Code). In general, the results show that satellite surface reflectance agrees well with field measurements for all approaches, with the correlation coefficients (R) ranging from 0.491 at blue bands to 0.907 at red bands. For comparison, the corrected data using MODIS products as input in 6SV model has a better agreement than Acolite model, and quite similar to that of LASRC code. Therefore, new MODIS product shows reliability as input data to support radiative transfer models. The results also show a fair agreement between Acolite and LASRC with in situ data. In particular, Acolite model present the benefits of pixel-by-pixel correction and image-based approach to overcome limitations imposed over regions without atmospheric information. For LASRC approach, high quality of surface reflectance dataset contributes to Landsat time series analysis and routinely monitoring of water dynamics. Therefore, all approaches presented satisfactory correlations with in situ data. Finally, required accuracy level for surface data relies on the specificities of each application, for example, detection of sediment plumes requires less attention with inaccuracies from atmospheric correction than the quantification of chlorophyll-a concentrations.