The percentile and K-means clustering methods are utilized to analyze the longitude and latitude information of Fujian ocean buoy, trying to solve the problem of position estimation of missing observations under the conditions of position drift and swing in the long sequence of ocean buoy data. On this basis, with the third-party data such as CLDAS and ERA5, the inverse distance weighted method and the ordinary Kriging interpolation algorithm are used to calculate when the timing and longitude/latitude positions are determined, and the corresponding temperature, pressure, relative humidity, wind direction and wind speed are estimated. First, the evaluation results of different third-party information and different interpolation algorithms indicate that the stability of data sequence and its length have great impact on the filling of elements, and the inverse distance weighted method performs better than the ordinary Kriging interpolation. Subsequently, based on the conclusions, a new data set is reconstituted by integrating the original observation data from 7 Fujian ocean buoys. A comparative evaluation and validation of the two data sets show that the Pearson correlation coefficients of various meteorological elements are generally above 0.9. The new data set not only significantly improves its completeness but also preserves the consistency and robustness of the original observation data. Local abnormal element values are also effectively controlled. During the validation process of typhoon cases from 2014 to 2021, the wind speed elements are evaluated by comparing them with the real-time typhoon tracks. The correlation coefficients of the wind speed elements in the new data set averagely increase by 0.03, and the root mean square error averagely decreases by 0.5 m[DK]·^s-1. The experiment shows that the data continuation technology can improve the operational applicability of ocean buoy observations.