UNCERTAINTY IN SALINE AQUIFER CAPACITY AND LEAKAGE: IMPACTS OF HYDROLOGIC HETEROGENEITY IN THE ROCK SPRINGS UPLIFT, WYOMING

Geological storage of CO₂ into saline aquifers is an effective method to reduce release of anthropogenic CO₂ into the atmosphere and mitigate global warming. Many geological, geochemical and hydrological factors influence CO₂ sequestration. Of these heterogeneity in the saline aquifer can significantly affect design of injection wells, CO₂ injection rate, CO₂ plume migration, storage capacity, and potential leakage and risk assessment. This study applies geostatistic methods and a Markov chain model at the Rock Springs Uplift, Wyoming to simulate facies-based heterogeneous fields for porosity and permeability in the target saline aquifer (Pennsylvanian Weber sandstone) and surrounding lithostratigraphic units (Chugwater, Phosphoria, and Madison). We next use the multiphase flow simulator FEHM to model CO₂ sequestration into the target saline aquifer. Our results reveal that (1) heterogeneity seriously influences performance of both CO₂ injection wells and brine pumping; (2) CO₂ injection rates in different injection wells significantly change with local permeability distributions; (3) brine production rates in different pumping wells are also significantly impacted by the local permeability heterogeneity; (4) liquid pressure evolution during and after CO₂ injection in saline aquifer varies greatly for different realizations of random permeability fields, and this has potential important effects on hydraulic fracturing of the reservoir rock and the integrity of the cap-rock; (5) different distributions of porosity and permeability fields in the target saline aquifer lead to different CO₂ storage capacity estimates, ranging from 392 Mt to 540 Mt for a domain of 16km×16km with 50 years of CO₂ injection. Finally, we present estimates of uncertainty in leakage into overlying formations due to heterogeneity in both the target saline aquifer and surrounding formations. This uncertainty in leakage will be used to feed into risk assessment modeling.