Missing data imputation for an ocean research station based on multi-layer perceptron neural network

Missing data from ocean research stations pose significant challenges for accurate oceanographic modeling and analysis. Such data gaps frequently occur during summer typhoon seasons and often persist for extended periods due to severe weather conditions. This study introduces a multi-layer perceptron neural network for missing data imputation, using reanalysis datasets as inputs. Reanalysis products were used as reference data to provide contextual information on potential ocean conditions during the missing periods. The model was trained on periods with available observations, learning to leverage reanalysis data as supplementary inputs while aligning with observed patterns. The trained network was then applied to the missing periods, effectively using reanalysis data to reconstruct gaps caused by typhoon-related disruptions. The imputation results indicate that the proposed model performs effectively in filling long-term data gaps, demonstrating robustness and reliability. By leveraging reanalysis data for gap imputation, the method achieves high accuracy in reconstructing missing values and significantly improves the completeness and utility of datasets from ocean research stations for both scientific and operational purposes.