Use of satellite information and application of assisted learning algorithms for CO2 injection and storage and its implications for groundwater in Colombia

The successful implementation of Carbon Capture, Utilization, and Storage (CCUS) technologies relies on the rigorous characterization of geological formations to ensure long-term geomechanical stability and containment integrity. In regions governed by complex hydrogeological dynamics, establishing a robust geomechanical baseline is paramount for effective early-stage site selection. While conventional site assessment typically depends on localized, high-cost geophysical surveys, regional-scale screening requires a more cost-effective methodology for evaluating baseline crustal deformation. This study evaluates the pre-feasibility of CCUS within the Bogotá and Middle Magdalena Valley basins in Colombia, utilizing an evaluation framework that integrates multiscale remote sensing with a systematic review of secondary hydrogeological and geophysical datasets.

To address the requirement for a regional monitoring framework, this investigation employs Sentinel-1 InSAR time-series (processed via MintPy) to identify millimeter-scale surface displacements. These observations are correlated with GRACE and GRACE-FO terrestrial water storage anomalies, which were statistically downscaled through a Random Forest regression—utilizing FLDAS and land-cover predictors—to achieve spatial alignment with the displacement grid resolution. To enhance the interpretation of these signals, the study incorporates a preliminary review of secondary well-log records and pressure data, specifically targeting the identification of pore pressure anomalies and reservoir overexploitation zones that could compromise storage security.

To systematically organize these diverse datasets, an exploratory assisted learning framework was implemented to categorize regional suitability based on stability proxies and hydrogeological response. Preliminary findings from the Bogotá basin, substantiated by Mann-Kendall trend analysis, indicate pronounced subsidence rates reaching up to 9 cm/year in critical sectors. Quantitative analysis demonstrates a spatial consistency exceeding 70% between groundwater level drawdown and InSAR-measured deformation trends, supported by high statistical significance (p-value < 0.05). The identification of these consistent patterns facilitates the effective filtration of seasonal hydrological noise, thereby establishing a baseline to differentiate between elastic soil responses and long-term subsidence risks. These findings establish a structured and robust workflow for the initial pre-feasibility of CCUS projects in geologically complex sedimentary environments where primary data is inherently limited.