Abstract: Objectives: With the drastic changes in the global climate, it is increasingly likely that the Arctic Passage will become fully navigable. However, the current research on Arctic Passage faces the following challenges: fragmented and heterogeneous data integration, as well as the lack of a systematic indicator system for complex influencing factors. To solve these issues, we propose a knowledge graph construction method for evaluating the summer navigability of Arctic Passage. Methods: The method is mainly derived from the classic knowledge graph, aiming to better adapt to polar application scenarios. First, multimodal and multi-source data are preprocessed based on the relevant ontology model of the Arctic Passage Knowledge Graph. Retrieval augmented generation (RAG) technology combined with a large language model (LLM) is used for ontology-guided data cleaning and verification. Entities and relationships are then extracted to obtain triples with high confidence. At the same time, we build knowledge graph chain rules, reasoning models that meet decision-making needs, store and manage the graph chain system through the Neo4j graph database. Then we combine the relevant standards and specifications provided by experts in the polar field to retrieve the knowledge graph library, so as to achieve calculation and reasoning of the summer navigability of Arctic Passage based on knowledge graphs. Results: The experimental verification is carried out on the example of the Northeast Passage from 2017 to 2021 and the international ship trajectory data in 2021.The results show that the summer navigability of the Arctic Passage has shown a continuously improving trend, with the sea areas of high-risk and nonnavigable decreasing year by year, but experiencing a rebound in 2021. Conclusions: This method can achieve reliable transformation from multi-source and multi-modal data to decision oriented summer navigability assessment knowledge.