Assessment and Validation of Land Surface Temperature Retrieval Algorithms Using Landsat 8 TIRS Data in Antarctic Ice-Free Areas

Abstract

This study evaluates the effectiveness of different Land Surface Temperature (LST) retrieval algorithms applied to Landsat 8 Thermal Infrared Sensor (TIRS) data in the ice-free regions of the Antarctic Peninsula. The primary objective is to determine the most accurate algorithm for LST estimation in these environments. Three algorithms, namely radiative transfer equation (RTE), single channel (SC), and mono window (MW), were utilised and compared to in-situ measurements at two locations in the northern part of James Ross Island (JRI), Antarctic Peninsula. The study considered various factors influencing LST accuracy, including land surface emissivity, atmospheric conditions, and sun elevation angles. The findings reveal that all three algorithms demonstrate significant sensitivity to emissivity. The MW algorithm emerged as the most suitable, showing the lowest root mean square error (RMSE) of 3.06 °C, followed by the SC and RTE algorithms with RMSE values of 3.68 and 3.98 °C, respectively. The study also underscores a strong positive correlation between LST retrieval accuracy and sun elevation angle, with more accurate results obtained from satellite images acquired in February, characterised by lower sun elevation angles. No significant relationship with water vapour content in the atmosphere was identified during the investigated period.