Reconstruction of cloud-free time series satellite observations of land surface temperature (LST) using harmonic analysis of time series algorithm (HANTS)

Land surface temperature (LST) is an essential parameter in the energy exchange between the earth surface and atmosphere. It is widely used in various scientific fields, such as climatology, hydrology, agriculture, ecology, public health and environmental science where the time series analysis of LST is vital. One of the methods to estimate LST is to use thermal remote sensing technique and infra-red satellite imageries. But, the time series satellite data are commonly prone to miss data, outliers (spatially and temporally) due to clouds, aerosols, cloud masking algorithm malfunctioning and sensor errors. In this study, to solve the problem of missing data (gaps) and outliers Harmonic ANalysis of Time Series algorithm (HANTS) was used. The day and night MODIS LST products (MOD11A1) were used in 2015, with 1 kilometers and daily spatial and temporal resolution, respectively. The study area covers most part of Iran, Turkmenistan and the Caspian Sea, which belongs to an image frame that in the sinusoidal MODIS frame system has the horizontal and vertical number of 22 and 5 (h22v05), respectively. The quality evaluation of original data showed that on average 36.8 and 35.6 percentage of data was covered by a cloud by day and night time. The results of the HANTS algorithm illustrated that the Root Mean Square Error (RMSE) between the original and reconstructed data were 3.87 and 2.68 Kelvin during the day and night time. The results of this study indicate that HANTS algorithm can effectively solve the problem of gaps and outliers and improve the quality of data used in time series LST of MODIS.