(1. 浙江大学 海洋学院，舟山 316021;
2. 自然资源部第二海洋研究所 自然资源部海底科学重点实验室，杭州 310012;
3. 上海交通大学 海洋学院，上海 200240)
摘 要: 海底多金属硫化物是未来可供开发利用的重要矿产资源。由于海底环境复杂，勘探成本巨大，利用成矿理论开展资源预测工作就显得尤为重要。本文综述了现有的海底多金属硫化物成矿远景区预测方法，分析比较了各预测方法的特点，借鉴陆地火山成因块状硫化物的资源预测方法，并结合卡尔斯伯格脊的应用实例，对多金属硫化物资源预测工作进行了探讨：多金属硫化物成矿预测方法需综合考虑研究区的勘探程度、数据资料的精度、覆盖范围等实际情况，并结合各方法的特点及其适用性进行合理选取；应用知识驱动与数据驱动的组合预测方法和深度学习算法解决已知硫化物矿床(点)不足、小样本、数据缺失、数据耦合、主客观误差等问题，提高预测的准确性和效率；通过综合比较基于不同原理的预测方法获得的结果进行验证，提高资源预测的可靠性。
关键字: 多金属硫化物；资源预测；预测方法；成矿远景区
（1. Ocean College, Zhejiang University, Zhoushan 316021, China;
2. Key Laboratory of Submarine Geosciences, Second Institute of Oceanography, Ministry of Natural Resources, Hangzhou 310012, China;
3. School of Oceanography, Shanghai Jiao Tong University, Shanghai 200240, China）
Abstract:The seafloor polymetallic sulfides are important mineral resources available for human development and utilization in the future. Due to the complex submarine environment and tremendous exploration costs, it is significant to use metallogenic theory for resource prediction. In this study, the exciting methods for predicting prospective area of polymetallic sulfide were reviewed. Through analyzing the characteristics of each method and prediction methods for volcanic-hosted massive sulfide resources, the prediction of polymetallic sulfide resources with application of Carlsberg Ridge was discussed. First, the prediction methods of polymetallic sulfide resources should be selected reasonably, based on the degree of exploration, the accuracy and coverage of data, as well as the characteristics and applicability of methods. Second, the combined prediction methods of knowledge-driven and data-driven, and deep learning algorithm can be considered for polymetallic sulfide resource prediction to improve the accuracy and efficiency by solving the problems of insufficient known deposits, small samples, data missing, data coupling, subjective and objective errors. Third, the reliability of polymetallic sulfide resource prediction can be evaluated by comparing the results of methods with different principles for predicting metallogenic prospective area.
Key words: polymetallic sulfide; resource prediction; prediction method; metallogenic prospective area
