Abstract The Cooling the Tube Programme was set up to implement London Underground's intention to mitigate future warming of the underground system and to control tunnel temperatures. As part of this programme the use of groundwater from the Chalk in open-loop systems was considered for cooling a number of stations and a groundwater model was determined as a requirement by the Environment Agency to assess impacts on the Chalk aquifer. It seemed plausible that the plumes of injected, warmed water would interact hydraulically and possibly also thermally with each other at some locations. Furthermore, London Underground wished to investigate any longer-term warming effects and potential loss in cooling benefit. This paper provides an overview of the hydrogeological modelling approach that was adopted to assess the potential hydraulic and thermal effect of the proposed schemes and summarizes some of the findings that may be of broader interest. A staged approach was used to guide and refine the modelling. Initially, analytical solutions were used to investigate the processes of heat transport in the Chalk and to derive parameters from a field-scale test. The results indicated that, even for low fissure-to-matrix contact areas, heat is likely to be conducted far into the Chalk matrix. Over longer time scales, significant heat loss occurs into the under- and overlying formations and the available analytical solutions are of less use in scoping heat calculations for open-loop schemes at this scale. Distributed, finite-element, numerical models were constructed to simulate the interactions within groups of ground source cooling systems and to investigate specific operating conditions at single schemes. Extensive sensitivity analysis was carried out to assess the level of uncertainty in the model predictions, and the relative sensitivity to the various parameters. The greatest uncertainty was associated with detailed aspects of the conceptual model of heat flow. Model results were sensitive to the assumed vertical distribution of permeability in the Chalk. Additional uncertainty was associated with the selection of appropriate thermal boundary conditions and the extent of interaction with the under- and overlying formations. Uncertainty in the model hydraulic and thermal material parameters had less effect on the simulated performance of the systems. Operational conditions, such as the amount of heat injected and the separation of the abstraction and injection boreholes, were assessed and this paper provides some quantification of the significance of these factors. This work also demonstrates that equivalent MT3D and FEFLOW models give reasonable agreement and both approaches are feasible for modelling heat transport, as density-driven flow is not significant for the temperature differences associated with these cooling schemes.