Meteorological data exhibits distinct trends with strong temporal correlations; however, intermittent signals from instruments and sensors often result in gaps spanning multiple reporting periods. This paper outlines the application of Seasonal-Trend decomposition using LOESS (STL) to time-series data, followed by the filling of gaps using decomposed trends and seasonal and residual components using the Kalman filter, SARIMAX, and ARIMA methods. The effectiveness of these gap-filling techniques and their impact on model training are compared with those of conventional methods, including linear and spline interpolation. Our method reduces prediction errors and preserves the trend and seasonal variations of climate data, making the temporal completion of climate data more robust and providing an essential basis for improving the accuracy of climate models.