Normalized water-leaving radiance spectra nLw(λ) at the red, near-infrared (NIR), and shortwave infrared (SWIR) are quantified and characterized in highly turbid waters of the western Pacific using 3 yr (2009–2011) observations from the Moderate Resolution Imaging Spectroradiometer on the satellite Aqua. nLw(645; red), nLw(859; NIR), and nLw(1240; SWIR) were higher in the coastal region and river estuaries, with SWIR nLw(1240) reaching up to ∼ 0.2 mW cm−2 µm−1 sr−1 in Hangzhou Bay during winter. The NIR ocean-reflectance spectral shape represented by the ratio of the normalized water-leaving reflectance ρwN(λ) at the two NIR bands ρwN(748) : ρwN(869) is highly dynamic and region-dependent. The NIR spectral feature associated with the sediment source from the Yellow River and Ancient Yellow River is noticeably different from that of the Yangtze River. There are non-negligible SWIR nLw(1240) contributions for waters with the NIR nLw(859) > ∼ 2.5 mW cm−2 µm−1 sr−1. Estimation of the NIR ocean reflectance with iterative approaches might only be accurate for turbid waters with nLw(859) < ∼ 1.5 mW cm−2 µm−1 sr−1. Thus, the SWIR atmospherics correction algorithm for satellite ocean-color data processing is indispensable to derive accurate nLw(λ) for highly turbid waters. Current existing satellite algorithms for chlorophyll a, diffuse attenuation coefficient at the wavelength of 490 nm (Kd(490)), total suspended matter, and inherent optical properties (IOPs) using nLw(λ) at the red band for coastal waters are limited and can only be applied to turbid waters with nLw(859) < ∼ 1.5 mW cm−2 µm−1 sr−1. Thus, the NIR nLw(λ) measurements are required to characterize water properties for highly turbid waters. Based on the fact that pure water absorption is significantly larger than other absorption components in the NIR wavelengths, we show that it is feasible to analytically derive accurate IOP data for turbid waters with combined satellite-measured visible-NIR nLw(λ) spectra data.