Remote sensing of coastal and inland waters requires sensors to have a high spatial resolution to cover the spatial variation of biogeochemical properties in fine scales. High spatial-resolution sensors, however, are usually equipped with spectral bands that are wide in bandwidth ( $50\mathrm{nm}$ or wider). In this study, based on numerical simulations of hyperspectral remote-sensing reflectance of optically-deep waters, and using Landsat band specifics as an example, the impact of a wide spectral channel on remote sensing is analyzed. It is found that simple adoption of a narrowband model may result in $>20\%$ underestimation in calculated remote-sensing reflectance, and inversely may result in $>20\%$ overestimation in inverted absorption coefficients even under perfect conditions, although smaller ( $\sim 5\%$ ) uncertainties are found for higher absorbing waters. These results provide a cautious note, but also a justification for turbid coastal waters, on applying narrowband models to wideband data.