Wellbore defects are a major engineering risk for geological carbon storage, creating high-permeability pathways for CO2 migration across confining boundaries. Despite growing interest in Microbially Induced Calcium carbonate Precipitation (MICP) as a remediation approach, systematic evaluation under varying reservoir permeability and pressure conditions remains lacking. This study develops an axisymmetric coupled multiphase-flow and biogeochemical model to quantify leakage behaviour and assess MICP injection strategies for defective wellbores. Simulation results reveal that leakage is controlled by the interplay between reservoir pressure dissipation and defect transmissivity. In high-permeability reservoirs, each order-of-magnitude reduction in defect permeability produces a corresponding decrease in leakage flux, making targeted sealing the most effective first intervention. In low-permeability reservoirs, pressure accumulation dominates and should be managed prior to sealing; otherwise, substantially greater reductions in defect permeability are required. Among the evaluated MICP strategies, injection from within the defect zone provides superior sealing performance by promoting the formation of a hydraulically effective low-permeability barrier at the defect inlet. Under the present axisymmetric configuration, leakage mitigation is strongly associated with localised permeability reduction along the leakage pathway, allowing nonuniform cementation to achieve comparable mitigation with reduced treatment volume. Under the simulated conditions, all strategies reduce leakage ratios below 0.1%, with defect-zone injection yielding superior performance. The results provide a process-based framework for prioritising sealing versus pressure management under site-specific geological conditions.
In this study, a process-based methodology for the assessment of susceptibility and hazard of rainfall induced shallow landslides is proposed. Landslide probability conditional to triggering and antecedent conditions is estimated with a Bayesian approach. The methodology is demonstrated with the example of Partenio Massif, an area of about 170 km2 in Campania (Italy), characterized by shallow pyroclastic soil over karstic limestone bedrock, where rainfall-induced landslides frequently occur. Besides rainfall, the triggering of such landslides is favored by hydrological conditions predisposing slopes to failure, mainly represented by root-zone soil moisture. The area is characterized by fairly homogenous geomorphology, lithology, vegetation, soil characteristics and climate, which allows considering landslide susceptibility controlled only by slope inclination and soil thickness. Based on these features, nine slope classes were defined. For each slope class 1D Richards' equation model was run to simulate soil saturation at hourly resolution for a period of 22 years (2002−23), using rainfall data from nine rain gauges operating in the area. Then, the factor of safety (FS) was calculated under the infinite slope hypothesis to assess landslide occurrence for each slope class. To account for the spatial variability of rainfall in the area, the rain stations were categorized into three groups based on the probability distribution of the observed rain events. The landslide probability assessed for the three groups from the calculated series of FS was used to map landslide susceptibility for all the slope units of the area. The obtained susceptibility map well agrees with the map of inventoried landslides. Landslide probability, conditional to current rainfall and antecedent root-zone soil moisture, was then calculated for all three groups and the nine slope classes, allowing dynamic landslide hazard assessment. Landslide hazard maps were obtained, based on rain gauge measurements and ERA5-Land root-zone soil moisture estimates on specific days, reliably identifying location and timing of reported landslides. This result shows the potential of the proposed approach for operational landslide hazard assessment.
