AUTHOR=Candela Thibault , Chitu A. G. , Peters E. , Pluymaekers M. , Hegen D. , Koster Kay , Fokker Peter A. TITLE=Subsidence Induced by Gas Extraction: A Data Assimilation Framework to Constrain the Driving Rock Compaction Process at Depth JOURNAL=Frontiers in Earth Science VOLUME=10 YEAR=2022 URL=https://www.frontiersin.org/journals/earth-science/articles/10.3389/feart.2022.713273 DOI=10.3389/feart.2022.713273 ISSN=2296-6463 ABSTRACT=

Surface movement can be induced by many human subsurface activities: production of natural gas, geothermal heat extraction, ground water extraction, phreatic groundwater level lowering, storage of natural gas and CO2. In this manuscript, we focus on subsidence caused by gas production. While geological interpretations, seismic campaigns and flow modeling often provide a relatively rich pre-existing knowledge, understanding of the driving mechanisms for production-induced subsidence is still poor and forecasts are often very uncertain. This is related to the multiple poorly constrained models that translate gas production to ground surface displacements. Currently, a biased constraint of these models is inferred by arbitrarily pre-selecting a subset of those. Here, we have devised and deployed an integrated approach of the entire chain of models from the flow simulations to the ground surface displacements which, for the first time, accounts for all our pre-existing knowledge in terms of processes and uncertainties attached to them. More specifically for the transfer between reservoir depletion to compacting volume at depth, four reservoir-rock compaction models are a-priori considered, ranging from linear elastic model to nonlinear time-dependent viscous-type model. After assimilation of the geodetic observations (i.e., the ground-surface displacements) with ensemble-smoother algorithms, we demonstrate that even when all the a-priori known complexities were present in all steps of the modelling chain, the model parameter uncertainties of each model could be reduced. Interestingly we demonstrate that one can discriminate which reservoir-rock compaction model driving subsidence is activated at depth. This identification of the activated compaction model at depth is crucial to build confidence in our subsidence forecasts. The predictive power of the integrated approach is demonstrated with an ensemble of synthetic but complex reservoir flow simulations mimicking all the characteristics and uncertainties representative for real gas fields in the north of the Netherlands.