Due to its potential to enable global connectivity in remote locations, such as rural areas and islands, Space-Air-Ground networks have become an ambitious solution for terrestrial communication in the sixth generation (6G) wireless communication network. In this paper, we propose a novel structure of Space-Air-Ground-Sea integrated networks (SAGSINs) to study and derive the coverage probability (CP) of users who are annotated as surface stations (SSs) on the far-reaching ocean surface that is far away from the coastline. By incorporating different types of relays such as onshore stations (OSs), tethered balloons (TBs), high altitude platforms (HAPs), and satellites (SATs), communication links between the terrestrial core connected base stations (CCBSs) and SSs are established via one of the four types of relay stations. Considering practical scenarios with a random distribution of SSs, we model the channel using the point-to-area model, which is recommended by ITU (for OSs to SS), the Rician model (for TBs or HAPs to SS), and the Shadowed-Rician model (for SATs to SS). When the SS’s distance from the coastline continues to increase from zero, since different channel models are considered, different relay stations will result in specific received signal strengths at SSs. The most powerful relay station will be chosen as the relay at one time. Hence, as we move away from the coastline, the respective strengths of the different types of relay stations vary, and hence, the association preference (among HAPs, OSs, TBs, and SATs) of the SSs changes leading to a CP value high enough even at locations far away from the coastline into the ocean. We analyze the CP using tools from stochastic geometry. Comparisons of CP between the integrated system with four types of relay stations and the single relay station system (only one type of relay station available) are represented. Numerical results verified by Monte-Carlo simulations reveal insights into the applicability of SAGSINs.