WaveDINO: Learning-Based Atmospheric Correction of Unwrapped InSAR Interferograms Validated by GNSS: Results at Laguna del Maule and Campi Flegrei Volcanoes
Researchers have developed WaveDINO, a novel wavelet-based denoising framework for InSAR interferograms, which are often corrupted by atmospheric phase delays. This learning-based method utilizes a hybrid training strategy combining synthetic deformation with real atmospheric noise, conditioned on DINOv3 foundation-model features and terrain information. WaveDINO demonstrated superior performance compared to existing methods, improving agreement with GNSS measurements by up to 19% and surpassing weather-model-based corrections in tests conducted at volcanic sites in Chile and Italy. AI
IMPACT This research demonstrates a novel application of foundation models for improving geophysical data processing, potentially enhancing the accuracy of volcanic deformation monitoring.