Land Surface Temperature Retrieval From Landsat-9 Thermal Infrared Sensor Data

Abstract

Land Surface Temperature (LST) is a key variable governing land-atmosphere energy exchange and plays a fundamental role in global environmental monitoring. The launch of the Landsat 9 satellite, equipped with the second-generation Thermal Infrared Sensor 2 (TIRS-2), offers enhanced capabilities for thermal monitoring due to reduced stray light and improved radiometric stability. This study proposes and validates a Split-Window (SW) algorithm based on global coefficients for LST retrieval using Landsat 9 thermal bands 10 and 11. Validation was conducted using a time series of in-situ radiometric measurements acquired at the Barrax experimental site (Spain). The results demonstrated the high accuracy of the proposed algorithm, yielding a RMSE of 1.22 °C and a bias of -0.35 °C. When compared to the official USGS Level-2 operational product, the proposed algorithm significantly improved accuracy by correcting the systematic overestimation (+1.15 °C) and reducing the overall error exhibited by the official product (RMSE = 1.99 °C). Additionally, residual analysis confirmed the robustness of the algorithm against variations in columnar atmospheric water vapor content, overcoming the limitations of single-channel methods under variable humidity conditions.