부산대학교 도서관

This paper determine the mean sea level of Ulleungdo, a representative island region with its own unique reference point, and recalculated the elevations of the reference point and national benchmark points based on this mean sea level. Additionally, we computed the ideal mean sea level for the surrounding areas of Ulleungdo using a geoid model. Furthermore, we developed a prediction model for the mean sea level using deep learning techniques, enabling the prediction of the annual mean sea level in the Ulleungdo region and the calculation of the height of the mean sea level at arbitrary times. We also applied the results of the prediction model to Dokdo and compared and analyzed them with the results for Ulleungdo.Through this research, a long-term (over 18.6 years) tidal observation dataset in the Ulleungdo region serves as the foundation for recalculating vertical datum performance, resulting in elevation accuracies compliant with international standards. A comparative analysis with the previous Ulleung benchmark elevation yields a difference of 1.68 cm. Subsequently, reciprocal leveling measurements are conducted in twelve sections (total one-way distance of 61.085 km) using the recalculated Ulleung benchmark elevation to reevaluate the elevation performance of national reference benchmarks.To compute the idealized geoid for mean sea level, the latest global gravity field model is analyzed, determining the optimal global gravity field model for the Ulleungdo region. The final gravity geoid model is then computed using the Remove-Restore technique. The results show an improvement in accuracy, approximately 2.3 cm, for the optimal global gravity field model, SGG-UGM-2.Through this study, we developed a model to predict the hourly variation of mean sea level from observational data. We compared the performance of two widely used deep learning techniques, LSTM and GRU models, and concluded that the GRU model exhibits slightly superior performance. The prediction of the annual mean sea level was conducted using the improved performance of the GRU model, and the predicted data were generated on an hourly basis for the years 2018 to 2021. The annual sea level rise was determined to be 0.52 cm/year, with an average RMSE of approximately 0.44 cm over the four-year period.The determined mean sea level values and model outcomes presented in the conclusion are evaluated based on actual observational data collected at the Ulleungdo tide gauge station. The mean sea level prediction model demonstrates exceptional performance, capable of predicting sea levels at hourly intervals with a high degree of accuracy, attributed to the extensive and highly periodic nature of the training data. This study anticipates its applicability not only to coastal regions like Ulleungdo but also to various tide gauge stations around South Korea for comprehensive research on national vertical datum systems and sea level rise.