Application of random forest model in Google earth engine to enhance land surface temperature estimation: a machine learning approach

Abstract

Land Surface Temperature (LST) is a crucial parameter in Earth observation and environmental studies due to its significance in various fields [1]. Understanding the temporal and spatial variations in LST is essential for surface modeling processes. This knowledge is crucial for various applications, including soil moisture estimation, wildfire monitoring, and urban climate change mitigation. Consequently, satellite-derived time series of land surface temperatures (LSTs) with high spatial resolution offer valuable insights for monitoring the dynamics of the land-atmosphere interface across diverse landscapes [2]. The purpose of this research is to enhance LST mapping by using the Random Forest algorithm in relates with various spectral indices derived from Landsat 8 imagery. By integrating these indices with the Random Forest machine learning model, the research aims to improve the accuracy of LST estimation and mapping across different landscapes.