The eight neighborhood algorithm (ENA) was developed for estimating the variance of suspended sediment concentration (SSC) distribution in a pixel based on the current knowledge about the SSC retrieval models' scaling effect. Taking advantage of the results of ENA, further study developed the scaling correcting methods for linear quantitative models, exponential quantitative models and logarithmic quantitative models. Applied these methods to the results of the Landsat/TM imagery and the field spectral experiments in Tai Lake, it was found that the scale effect errors are strongly correlated with the SSC predicting models. The scale effect could result in 16% relative errors for reflectance measurement in complicated Case Ⅱ water bodies such as Tai Lake.