Geoscience knowledge graphs (GKGs) formally represent geoscience knowledge in a way of directed graph and have strong capabilities in knowledge representation, openness and interconnectivity, and reasoning and prediction. GKGs have been one of the important infrastructures for the development of combining geoscience and artificial intelligence, thereby becoming one of the important research focuses in geoscience. Therefore, many international scientific organizations or groups have successively carried out studies in this domain, and constructed some representative GKGs. However, there is a lack of an in-depth study and analysis of these existing GKGs. To this end, this paper makes a systematic comparative analysis on their general information, construction methods, and main contents. On this basis, some enlightenments about future research of GKGs are discussed. In terms of the construction method, a unified representation framework for GKGs should be built, the source of geoscience knowledge should be enhanced by conflating multi-source and multimodal data, and methods for the representation and computation of geoscience knowledge should be studied. Regarding the contents of GKGs, complex spatiotemporal characteristics, relations, and reasoning rules should be considered. From the perspective of application, methods for assessing quality and making correction for geoscience knowledge should be developed, and application effects of GKGs should be improved.