Research on Deriving Consistent Land Surface Temperature from the Geostationary Operational Environmental Satellite Series

Abstract

Geostationary Operational Environmental Satellite (GOES) have been continuously monitoring earth surface since early 1970. The frequent observations provided by GOES sensors make them attractive for deriving information on the diurnal land surface temperature (LST) cycle and diurnal temperature range. These parameters are of great value for the research on the Earth’s diurnal variability and climate change. Accurate extraction of satellite-based LSTs has long been an interesting and challenging research area in thermal remote sensing. However, derivation of LST from satellite measurements is a difficult task because surface emitted thermal infrared radiance is dependent on both land surface temperature and land surface emissivity (LSE), two closely coupled variables. Satellite LST retrievals have been conducted for over forty years from a variety of polar-orbiting satellites and geostationary satellites. Literature relevant to satellite-based LST retrieval techniques have been reviewed. Specifically, the evolution of two LST algorithm families, temperature and emissivity separation method (TES) and Split Window (SW) approach, have been studied in this work. This work also summarizes the LST retrieval methods especially adopted for geostationary satellites. All the existing methods could be a valuable reference to develop the LST retrieval algorithms for generating GOES LST product. The primary objective of this study is the development of models for deriving consistent GOES LSTs with high spatial and high temporal coverage. Proper LST retrieval algorithms will be studied according to the characteristics of sensors onboard the GOES series. A new TES approach is proposed in this study for deriving LST and LSE simultaneously by using multiple-temporal satellite observations from GOES 8 to GOES 12 series. Two split-window regression formulas are selected for this approach, and two satellite observations over the same geolocation within a certain time interval are utilized. This method is particularly applicable to geostationary satellite missions from which qualified multiple-temporal observations are available. Dual-window LST algorithm is adopted to derive LST from GOES M (12)-Q series. Instead of using conventional training method to generate optimum coefficients of the LST regression algorithms, a regression tree technique is introduced to automatically select the criteria and the boundary of the sub-ranges for generating algorithm coefficients under different conditions. GOES measurements as well as ancillary data, including satellite and solar geometry, water vapor, cloud mask, land emissivity etc., have been collected to test the performance of the proposed LST retrieval algorithms. In addition, in order to validate the retrieval precision, the satellite-based temperature will be compared against ground truth temperatures, which include direct skin temperature measurements from the Atmospheric Radiation Measurement program (ARM), as well as indirect measurements like surface long-wave radiation observations over six vegetated sites from the SURFace RADiation Budget (SURFRAD) Network. The validation results demonstrate that the proposed GOES LST algorithms are capable of deriving consistent surface temperatures with good retrieval precision. Consistent GOES LST retrievals with high spatial and temporal coverage are expected to better serve the detections and observations of meteorological phenomena and climate change over the land surface.