Exploring multitemporal remote sensing for subterranean feature detection

This study introduces a remote sensing methodology using multitemporal satellite imagery (Sentinel-2, 2015-2025) to detect a 52-kilometer subterranean water pipeline by analyzing soil marks as surface proxies. The approach involves extracting pixel-level RGB and Near-Infrared (NIR) data from selected areas along and across the pipeline, applying normalization and enhancement techniques to accentuate temporal variations. Results indicate that enhanced analysis of transect datasets successfully highlighted localized indicators of the pipeline's presence. Temporal analysis revealed distinct patterns, including annual periodicity in statistical metrics (RMS, standard deviation) linked to vegetation cycles, and differential pipeline signal discernibility (e.g., clearer post-2020 in Dataset 2; consistent and non-seasonal in Datasets 3 and 4 after enhancement). High-resolution drone imagery (DJI Mavic 3, 3 April 2024) provided preliminary validation, confirming linear soil discolorations (20–25 m wide) aligned with satellite-derived patterns and temporal dynamics. The findings demonstrate the potential of systematic temporal analysis for monitoring underground infrastructure, although detectability is influenced by local pipeline-soil interactions and environmental factors, underscoring the need for in-situ validation. This work aims to form a basis for future systematic pipeline monitoring for leakages and structural integrity.