Surrogate-based implementation of sewer network structures into numerical heat transport models: First results of the Basel-City case study

Adequate management strategies are critically required to increase the resilience and long-term availability of groundwater resources in the light of progressive climate change and accelerating urbanization. Here, robustly parameterized numerical models, designed for simulating water flow as well as solute and heat transport processes in the hydrogeological subsurface, are powerful and widely established tools supporting decision making and planning.

Among other applications related to more general quantity- and quality-related questions, these numerical tools can be also used, e.g., for investigating the current thermal state of the subsurface and occurring changes due to artificial and natural influences. Models designed for this very specific task should include at least all major artificial objects (e.g., underground car parks, tunnels, buildings, sewer networks) which thermally contribute to the overall groundwater heat regime. For instance, the heat exchange between the subsurface and sewer systems may significantly contribute to the subsurface urban heat island effect and should, therefore, be implemented. However, fully three-dimensional implementations of sewer networks (typically with hundreds of kilometers of pipes) are mostly out of question when applying such numerical models, since it would be associated with large computational demands and increasing numerical instabilities.

To overcome this limitation, the focus of our ongoing research is to evaluate the suitability of an adaptive surrogate method to be coupled to existing numerical heat transport models. This method is based on linking expected thermal exchange rates between small subsurface objects (e.g., sewer pipes) and their surrounding area, which depend on site-specific parameters (e.g., surface-groundwater table distance, pipe dimensions, shapes and materials), with the spatial elements of an existing model mesh, e.g., as area-averaged heat sources or sinks. Numerical heat conduction simulations performed on pipe scale while employing seasonally changing ambient and sewer conditions point towards the importance of considering both stationary (such as materials) and transient input datasets (such as temperature fluctuations) in this linking process. The collection and pre-processing of both dataset types is performed in separate workflows employing standardized geographic information system (GIS) software. Based on these input datasets, heat flux calculations can be done either employing the numerical code itself (if the model code allows user-defined calculations) or, again, in a GIS-assisted step (in order to further reduce the computational demand during the runs of the numerical model).

The conceptual workflow, first results as well as expected advances and limitations of this surrogate approach will be critically discussed using the example of the well-documented heat transport case study of ‘Basel-City’. Among others, the aforementioned stationary and transient input datasets, and based on that, processed vertical heat fluxes will be presented for selected areas of the Swiss canton.