Assessing Risk in Long-Term CO2 Storage Under Uncertainty via Survival Analysis-Based Surrogates

We develop survival analysis-based surrogate models to assess leakage risk in geological CO2 storage under uncertainty. Long-term reservoir simulations are computationally expensive, while shorter simulations produce censored observations when leakage has not occurred before monitoring ends. Our framework accounts for this censoring to estimate leakage risk from limited simulation data.

Using scenarios with varied operating conditions and geological properties, we construct interpretable surrogates with methods ranging from Kaplan–Meier estimation to random survival forests. A saline aquifer case study demonstrates that these models can predict time to leakage in new scenarios without additional reservoir simulations, reducing computational cost while supporting risk-informed storage decisions.

Recommended citation: Gurwicz, A., Chen, J., Gutman, D. H., & Gildin, E. (2025). Assessing risk in long-term CO2 storage under uncertainty via survival analysis-based surrogates. SPE Journal, 30(5), 2837–2854. https://doi.org/10.2118/220737-PA