Ocean remote sensing is the only effective method for monitoring the marine ecosystem on a global scale in the long term. This method has played a crucial role in the research of carbon cycle, global changing, and ocean acidification and its biological impacts. The validation of ocean color products is quite important for the improvement of the algorithm and the reliability of further application. Given the complex and highly dynamic optical marine characteristics, especially for case Ⅱ water, the evaluation of accuracy and the expression of remote sensing products have become difficult, and considerable research has focused in this field. This study aims to create a scientific evaluation method of the representativeness of ground stations and the heterogeneity of remote sensing pixels of sea surface temperature (SST) verification. The temporal and spatial variation parameters, including the statistical standard deviation of time series, the spatial variation index, and the frontal index, are designed. The analytic hierarchy process for determining the weight of different indices and the method for regionalizing the interval of a spatiotemporal variation. Verification datasets are classified using a distribution map of the spatiotemporal variation of space level division. Result of the experimental data shows that spatiotemporal variation is a direct cause of errors. Considering the strong spatiotemporal variation, numerous verification errors are introduced in the verification process, which uses data from different variation grade divisions, and the relative error of the verification result can reach more than 13.08%. The spatial distribution of spatiotemporal variation is large in the Yellow and Bohai Sea regions and small in the East and South China Seas. The seasonal variation is strong in winter and spring and weak in summer and autumn. In this study, we propose models and methods to reasonably select representative validation datasets and verify the scientific nature of the method from the experimental data. In the region where the spatiotemporal variability is intense, the verification errors are large. These errors are not completely remote sensing product errors. The results of verification are not representative and cannot really reflect the error characteristics of remote sensing products. For SST ocean remote sensing product verification, spatiotemporal variability and its contribution to the validation error must be considered. Moreover, a reasonable selection of evaluation data set, scientific method, and representative test area must be verified.