Reconstruction of regular time series from bi-monthly temperature data in the Yellow Sea and the northwestern East China Sea

Reconstruction of evenly-spaced, regular time series from routine survey serial data was investigated for precise analysis of spatio-temporal variations in a temperate sea at mid-latitudes where the seasonality dominates the interannual variability. Considering that the annual range of sea surface temperature in the Yellow Sea and the northwestern East China Sea can be as large as 15–20 °C, temperature data collected bi-monthly in these temperate seas were used for the assessment of reconstruction methodology. The cubic spline interpolation with a sampling interval of 0.5 months reconstructed the regular time series closest to the in-situ measurements among various interpolation schemes. Also, two computation methods for the interannual anomaly were compared; the residual method that the long-term monthly means are subtracted from the monthly serial data and the filtering method that high-frequency variations are removed using a low-pass filter. In that the high-frequency variations in frequencies greater than 1 cycle per year are comparable in magnitude to the interannual variation, the residual method proves inadequate in the temperate sea. High-frequency noises can be effectively removed through the use of a proper low-pass filter with bell-shaped weights.