split-window algorithm

Land Surface Temperature (LST) is one of the most important parameters in land-atmosphere energy exchange that is applicable to many sciences such as climatology, hydrology, agriculture, ecology, etc. One of the most significant limitations of using remote sensing for estimation of LST is the presence of clouds, which remarkably affects the energy reflected from the surface and disrupts the reading ability of the optical and thermal sensors. In the present study, 23 Landsat 8 images in 2015 were used as an annual time series to estimate LST in a part of the pistachio farms of Yazd, Iran. LST in the 23 images was estimated by generalized split-window algorithm. The results showed in the estimated (23 images) LST time series, the minimum, maximum, and mean missing data due to cloud cover were 17%, 28%, and 19%, respectively. SSA algorithm was used to solve the problem of missing data. Root Mean Square Error (RMSE) and Mean Absolute Error (MAE) between the original and reconstructed data at the data points in the studied LST time series were 3.4 and 2.5 K, respectively. Moreover, the gap-filling error was estimated by extracting five random images with four iterations of time series and comparing the reconstructed images with the extracted spatio-temporal images. RMSE and MAE were estimated to be 4.4 and 3.6 K in reconstruction of temporal artificial gaps and 3.7 and 2.8 K in the spatial artificial gaps, respectively. Based on the findings, SSA algorithm can be effectively used to fill the problem of missing data due to cloud cover in Landsat 8 LST time series.