Hydraulic fracturing of horizontal wells is an essential technology for the exploitation of unconventional resources, but led to environmental concerns. Fracturing fluid upward migration from deep gas reservoirs along abandoned wells may pose contamination threats to shallow groundwater. This study describes the novel application of a nonlinear autoregressive (NAR) neural network to estimate fracturing fluid flow rate to shallow aquifers in the presence of an abandoned well. The NAR network is trained using the Levenberg–Marquardt (LM) and Bayesian Regularization (BR) algorithms and the results were compared to identify the optimal network architecture. For NAR-LM model, the coefficient of determination (R²) between measured and predicted values is 0.923 and the mean squared error (MSE) is 4.2 × 10⁻⁴, and the values of R² = 0.944 and MSE = 2.4 × 10⁻⁴ were obtained for the NAR-BR model. The results indicate the robustness and compatibility of NAR-LM and NAR-BR models in predicting fracturing fluid flow rate to shallow aquifers. This study shows that NAR neural networks can be useful and hold considerable potential for assessing the groundwater impacts of unconventional gas development.

Hydraulic fracturing for natural gas extraction from unconventional reservoirs has not only impacted the global energy landscape but has also raised concerns over its potential environmental impacts. The concept of “features, events and processes” (FEP) refers to identifying and selecting the most relevant factors for safety assessment studies. In the context of hydraulic fracturing we constructed a comprehensive FEP database and applied it to six key focused scenarios defined under the scope of FracRisk project (http://www.fracrisk.eu, last access: 17 August 2018). The FEP database is ranked to show the relevance of each item in the FEP list per scenario. The main goal of the work is to illustrate the FEP database applicability to develop a conceptual model for regional-scale stray gas migration.

Predicting and characterising groundwater flow and solute transport in engineering and hydrogeological applications, such as dimensioning tracer experiments, rely primarily on numerical modelling techniques. During software selection for numerical modelling, the accuracy of the results, financial costs of the simulation software, and computational resources should be considered. This study evaluates numerical modelling approaches and outlines the advantages and disadvantages of several simulators in terms of predictability, temporal control, and computational efficiency conducted in a single user and single computational resource set-up. A set of well-established flow and transport modelling simulators, such as MODFLOW/MT3DMS, FEFLOW, COMSOL Multiphysics, and DuMu^X were tested and compared. These numerical simulators are based on three numerical discretisation schemes, i.e., finite difference (FD), finite element (FE), and finite volume (FV). The influence of dispersivity, potentially an artefact of numerical modelling (numerical dispersion), was investigated in parametric studies, and results are compared with analytical solutions. At the same time, relative errors were assessed for a complex field scale example. This comparative study reveals that the FE-based simulators COMSOL and FEFLOW show higher accuracy for a specific range of dispersivities under forced gradient conditions than DuMu^X and MODFLOW/MT3DMS. FEFLOW performs better than COMSOL in regard to computational time both in single-core and multi-core computing. Overall computational time is lowest for the FD-based simulator MODFLOW/MT3DMS while the number of mesh elements is low (here < 12,800 elements). However, for finer discretisation, FE software FEFLOW performs faster.