Abstract: Intertidal flats represent a type of coastal wetland characterized by significant dynamic variations. Accurately mapping intertidal flats is crucial for coastal ecosystem monitoring and resource management. However, the classification of intertidal flats from remote sensing images faces two major challenges: the sediment-water mixed environments complicate the collection of high-confidence samples; there is a lack of effective methods for integrating time series information, as traditional methods primarily rely on percentile-based feature extraction while failing to leverage tidal level information. Focusing on the coastal area in Zhanjiang City, Guangdong Province, this study proposed a high-precision intertidal flat mapping framework that integrates tidal level information and the LightGBM model. Specifically, high-confidence samples were extracted during stable low-tide periods and then classified based on single-scene images. The classification results were further optimized by incorporating elevation, cloud cover, and other environmental factors. Furthermore, a dynamic retrieval method for high- and low-tide periods was developed based on the correlation between intertidal area variations and tidal levels, thereby generating stable intertidal flat maps. A quantitative evaluation based on Shapley additive explanations (SHAP) feature analysis reveals that the brightness, slope, near-infrared (NIR), and short-wave infrared 2 (SWIR2) contributed significantly to intertidal flat classification, whereas water indices and elevation played a dominant role in water body classification. Experimental results demonstrate that the proposed method achieved an overall accuracy of 0.97, a Kappa coefficient of 0.96, and a stable F1 score of 0.97 across multiple validation sample areas. It exhibited superior capability in accurately capturing small-scale intertidal flat features and distinguishing intertidal flats from aquaculture ponds.