Ocean color product validation is vital to improve the algorithms and to secure subsequent applications. China coastal region is one of the most highly turbid waters in the world. The precision of NASA's standard MODIS ocean color products are poor for this region. However, the validations have not been made due to the shortage of high quality in-situ data. In this paper, in-situ data collected in the Yellow Sea and the East China Sea in the spring and autumn of 2003 are used to validate the main parameters of MODIS Aqua standard ocean color product, i.e., the normalized water-leaving radiance L_wN(λ) (converted to remote sensing reflectance R_rs(λ)) and chlorophyll-a concentration (Chl.a). The resolution of 3×3 pixel box and ±0.5h are used as the sampling window, and spatial homogeneity criteria are used to screen data. Based on the analysis of match-ups (in low and middle turbid waters), primary conclusions are obtained: (1) MODIS-derived R_rs(λ) can assure certain precisions but can not satisfy the expected error of 10%, and has obviously poor results in 412nm bands. There is good linear relationship between satellite images and modeled images. The overall root mean square error is 0.0030sr^-1 and the median of absolute values over the relative error is about 25%, respectively. Derived R_rs(λ) is similar with in-situ ones in spectrum shape, and is mostly underestimated especially in high value region. Besides, R_rs(λ) at 412 and 667nm bands are mostly underestimated with even negative values in low value region. The reasons for this phenomenon are due to the influences of water-leaving radiance in the near-infrared bands and improper aerosol models used in the atmospheric correction process. (2) MODIS derived Chl.a is very poor. The satellite derived values are systematically overestimated. And there is no obvious linear relationship between satellite derived and in-situ Chl.a, which the root mean square error is about 1.56mg/m³, the median absolute relative error about 103.25%, and the median satellite to in-situ ratio about 2.03. The major reason for this phenomenon is the OC3M model (coefficients) unsuitability.