Reconstruction of gap-free time series satellite observations of land surface temperature to model spectral soil thermal admittance

Abstract

The soil thermal properties (soil thermal conductivity, soil heat capacity and soil diffusivity) are the main parameters in the applications that need quantitative information on soil heat transfer. Conventionally, these properties are either measured in situ or estimated by semi-empirical models using the fractions of soil constituents. The use of such methods over large and heterogeneous areas, however, is often costly, timeconsuming and sometimes impractical. This thesis proposes and evaluates a new approach to estimate the soil thermal properties by inverse modelling of Spectral Soil Thermal Admittance (SSTA) which is determined using the time series satellite observations of Land Surface Temperature (LST) and soil heat flux (G0) over the entire Qinghai-Tibet Plateau (QTP) from 2008 to 2010. To calculate the soil thermal admittance, the amplitudes of G0 and LST at significant frequencies are required which needs consistent, continuous and long time series. The hourly FY-2C LST time series used in this study were often contaminated by missing data (gaps) and outliers. The HANTS algorithm and M-SSA were used to fill the gaps and remove the outliers in the LST time series. Then, the gap-filled hourly LST was used to identify the most significant periodic components over a three-year data. The amplitude of soil heat flux and LST were estimated at significant frequencies and then the soil thermal admittance at each frequency was determined over the study area. The SSTA, which is the variation of STA against frequency, contains information about the soil thermal properties of different soil layers. An inversion model was used to estimate soil thermal properties of different soil layers (assuming three-layer soil) over the Q-TP.