Reconstruction of the land surface temperature time series using harmonic analysis

Satellite remote sensing is an important approach for obtaining land surface temperature (LST) over wide temporal and spatial ranges. However, the presence of clouds generates numerous missing and abnormal values that affect the application of LST data. To fill data gaps and improve data quality, the Harmonic ANalysis of Time Series (HANTS) algorithm was employed to remove cloud-affected observations and reconstruct the Moderate Resolution Imaging Spectroradiometer (MODIS) LST data taken in the year 2005 for the Yangtze River Delta region of China. Analysis of MODIS data quality indicated that the yearly proportion of high-quality LST data in this region was less than 50% with numerous missing and low-quality data points. To reconstruct 8-day LST via the removal of cloud-contaminated observations, we applied pixel-by-pixel harmonic fitting to the time series and used fitted values to replace the missing and abnormal values in the original LST data. To evaluate the reconstruction performance, a simulated dataset was generated according to the percentage of cloud coverage in each 8-day period. Satisfactory validation results indicate that the harmonic method can effectively fit the NA Values caused by cloud cover and fill data gaps in the LST data, which can significantly improve the practical value of the MODIS LST dataset.