The detection of human activity in isolated structures is crucial for improving surveillance and resource management in remote rural areas, where communication infrastructure and sensor networks are often nonexistent. This study investigates the possibility of using publicly available satellite data to infer thermal emissions associated with the presence of people or livestock in pig farms located in the Ebro Valley (the municipalities of Albalate, Ballobar, Granén, and Tamarite). Although the typical range for detecting human heat (8,000–14,000 nm) does not match the bands available on Sentinel-2, band 12 (SWIR, around 2,200 nm) provides spectral information that can be indirectly correlated with thermal variations inside buildings. To overcome the limitations of the spatial and radiometric resolution, machine learning techniques were applied to learn patterns of thermal emission from the spectral response of this band and meteorological data, particularly daily temperature collected between 2021 and 2023.
