GRSG Conference 2022: Orbit to Outcrop
Title: Detection of geothermal anomalies using thermal remote sensing data from ECOSTRESS sensor
Author: Agnieszka Soszynska
Remote sensing techniques can significantly enhance early exploration of geothermally active areas. Until now, researchers mostly used either daytime or nighttime imagery from Sun-synchronous satellite sensors, which is not optimal for detection of geothermal anomalies, due to thermal inertia of surfaces and inhomogeneity in solar irradiance in hilly terrain. ECOSTRESS sensor installed on the International Space Station (ISS) can acquire imagery at different times of day and night, creating a great opportunity for remote sensing of geothermal anomalies.
The special orbit of the ISS enables acquisition of images at optimal times, which minimizes the effects of thermal inertia differences. In our research, we tested the feasibility of ECOSTRESS data to detect geothermal anomalies in Olkaria, Kenya. We used nighttime Land Surface Temperature data for anomaly detection, and validated the results with data obtained from field work, as well as auxiliary datasets.
We developed a kernel method with statistics-based thresholding. The proposed method allows an automated adaption of the threshold if a geothermal anomaly comprises bigger part of the kernel. Thus, the method is adaptable to different geographic regions and characteristics of geothermal anomalies. The spatial patterns in our detection map mirror the spatial patterns of the validation datasets, which suggests reliability of the method. In the validation process, our method achieved 78.4\% overall accuracy.
The main sources of errors for geothermal anomaly detection lay in image quality, and thermal inertia of the surfaces. Noise, image artefacts, sharpness of the image, as well as georeferencing errors contribute to omission and commission errors. Additionally, differences in heat decay of surfaces can cause false positive detections, and further investigation is needed in this subject.