In remote sensing (RS), use of single optical sensors is frequently inadequate for practical Earth observation applications (e.g., agricultural, forest, ecology monitoring) due to trade-offs between spatial and temporal resolution. The advent of spatiotemporal fusion (STF) of RS images has allowed the production of images with high resolution at both spatial and temporal scales. Despite the development of more than 100 STF models in the past two decades, many of these models have not been practically applied due to the possibility of limited understanding of the models. Therefore, this study aims to provide a comprehensive review of STF methods, including their conception, development, challenges, and applications. This study focuses primarily on deep learning-based STF models, which achieved superior performance and significantly increased the number of STF models. This review can guide the selection and design of STF models, as well as proposes future directions for STF modeling. The findings of this review facilitate further STF research to improve the accuracy and application of fused RS images in the field of agriculture, forestry, and ecological monitoring.