These data sets serve as training and testing data for modelling the extension of temperature field emanating from one groundwater heat pump. There are simulated with Pflotran and saved in h5 format.

The data set for training is called "dataset_medium_k_3e-10_1000dp". It contains 999 simulation runs, each simulated until a near steady state is reached. Each data point measures 320 m x 6400 m with 64 x 1280 cells. The varying parameters of the data set are pressure and permeability. Both are constant within the domain, but vary across the data set.

A longer data set for testing is called "dataset_long_k_3e-10_7dp". It contains only 7 simulation runs with a larger domain of 320 m x 10.240 m with 64 x 2048 cells.

A third data set, with varying permeability within the domain, is called "dataset_medium_100dp_vary_perm". It contains 100 simulation runs. Each data point measures 320 m x 6400 m with 64 x 1280 cells.

Other parameters that define the data set, such as porosity, are chosen to be as close as possible to reality and defined in the simulation input files "inputs/pflotran.in". Source: "Die hydraulischen Grundwasserverhältnisse des quartären und des oberflächennahen tertiären Grundwasserleiters im Großraum München", Geologica Bavarica Volume 122. Generated with scripts from https://github.com/JuliaPelzer/Phd_simulation_groundtruth (commit a3777510c) with arguments given in inputs/args.yaml.