An Uncertainty-Driven Decision Tree Approach Guiding Feasibility Decisions of Shallow Geothermal Systems in Complex Subsurface Settings

Aquifer thermal energy storage (ATES) is a way to use the groundwater to heat and cool buildings, with very low CO₂ emissions. It classifies as a shallow geothermal technology, and it is gaining popularity worldwide because of its sustainability, efficiency and cost-effectiveness. While its potential has been extensively proven in traditional homogenous, productive sandy groundwater layers, investing in more complex subsurface settings has greater financial risk. This is related to uncertainty about the (hydraulic) project feasibility and (thermal) efficiency of the system. Basically, we cannot directly look underground, so it is uncertain to what extent our subsurface model correctly represents reality. Even though this subsurface uncertainty leads to a great globally untapped potential for thermal energy storage, it is often neglected in feasibility studies. To move new ATES developments forward in complex subsurface settings, we present an uncertainty-driven sound scientific method to make investment decisions. Uncertainty in subsurface models is recognized by using a stochastic approach. The model predictions are then processed with clustering and global sensitivity analysis. This allowed to define criteria on critical subsurface properties that guarantee project (in)feasibility. For edge-cases, uncertainty is quantified to determine the probability of project feasibility from a risk-taking or risk-averse decision-maker perspective. Additionally, this approach quantified the potential of changing operational parameters (flow rate, well spacing, design injection temperature) to enhance project feasibility. All results are summarized in an easy-to-interpret decision tree that guides go/no-go decisions for new ATES projects. Importantly, the decision-tree can be followed prior to carrying out costly field tests. To illustrate, the uncertainty-driven decision tree approach is applied to a low-transmissivity aquifer for ATES, which represents a subsurface setting at the limit of ATES suitability. In conclusion, our approach effectively handles uncertainty while also focusing on improving clear communication to investors about the probability of project feasibility. As such, it could be an example study on how to handle model uncertainty for predictions of aquifer thermal energy storage systems in the future.