Advancing spatio-temporal fusion with spectral super-resolution for enhancing remote sensing of marine debris and landfills

Monitoring environmental challenges such as landfill expansion and marine debris accumulation requires high-resolution satellite imagery with enhanced spectral, spatial, and temporal details. However, remote sensing systems often face trade-offs between these resolutions, limiting their effectiveness in detecting and analyzing dynamic environmental changes. To overcome this limitation, we propose a novel Spatio-Temporal Fusion (STF) framework that integrates Sentinel-2 and PlanetScope satellite datasets, improving image quality through spectral super-resolution. This study systematically evaluates deep learning-based STF methods, focusing on Conditional Generative Adversarial Networks (cGANs) and Multi-Stage Spectral-wise Transformers (MST++). Results indicate that instance normalization-based cGANs enhance spectral fidelity but introduce noise, while MST++ models offer superior structural reconstruction and generalization. By retraining models with an expanded dataset, their robustness significantly improved, with MST++ v2 emerging as the most reliable. The integration of STF methodologies enhances environmental monitoring by enabling more precise and frequent observations of critical issues such as plastic pollution, landfill management, and marine debris detection. The ability to generate high-resolution, multi-sensor fused images provides a powerful tool for sustainable resource management and environmental decision-making. This research highlights the potential of STF frameworks to advance Earth Observation applications and support global sustainability efforts, offering innovative solutions for addressing pressing environmental concerns.