@article{Dell'Aversana2026, 
author = {Paolo Dell'Aversana},
title = {Adaptive self-aware joint inversion of multiphysics geophysical data for volcanic monitoring: A methodological study},
year = {2026},
journal = {AIMS Geosciences},
volume = {12},
number = {2},
pages = {540-569},
keywords = {volcanic hazard, geophysics, data integration, joint inversion, self-learning},
url = {https://www.sciopen.com/article/10.3934/geosci.2026021},
doi = {10.3934/geosci.2026021},
abstract = {Volcanic risk monitoring and assessment are fundamental components of geohazard management, especially in the context of densely inhabited and socio-economically vulnerable regions. Traditional surface-based monitoring techniques, while valuable, often fall short in detecting deep, subsurface processes that precede eruptive events, particularly in areas where social, infrastructural, and logistical constraints hinder the deployment of high-resolution seismic surveys. These limitations are especially critical in volcanic settings where communities are exposed to multiple layers of risk, not only from natural hazards but also from structural inequalities and uneven access to protective measures. To address these challenges, we introduce a novel framework based on Self-Aware Joint Inversion, an adaptive, learning-driven method for the integration of multi-physics geophysical data. By combining seismic and non-seismic geophysical approaches, such as electrical resistivity tomography, gravity, and electromagnetic methods, within a self-optimizing joint inversion technique, our approach enables the dynamic and high-resolution imaging of subsurface processes that are directly linked to magma and fluid migration. Unlike conventional models that often remain shallow or static, this methodology offers time-lapse capabilities and deeper investigation, thereby enabling early detection of phenomena critical to volcanic risk forecasting. The method was tested on synthetic case studies simulating realistic volcanic conditions, enabling rigorous evaluation of resolution, adaptability, and potential operational performance. Results indicated that this approach can reconstruct subsurface anomalies associated with pre-eruptive activity, even in scenarios where classical monitoring frameworks would fail. Beyond its scientific contribution, this methodology is explicitly designed to be non-invasive and scalable, making it particularly suitable for application in sensitive inhabited zones.}
}