The GOTHIC code was validated using three experiments carried out in the PANDA facility in the framework of the OECD/NEA HYMERES project. These tests addressed the mixing of an initially stratified atmosphere by means of a vertical jet in the presence of on obstacle (circular plate). This paper reports on the simulations of three experiments, and complementary, quasi-steady-state tests without stratification, where the flow structure above the impingement plate could be observed by means of particle image velocimetry (PIV) velocity measurements in a region larger than that considered in the transient experiments. Moreover, simulations of similar tests without obstacle conducted during the OECD/SETH-2 project are also discussed. The reference, best-estimate model used for the analyses of the three experiments with different flowrates and initial and pressure boundary conditions was built on the base of a multistep approach. This was based on mesh and modeling sensitivity studies mostly performed for the complementary tests, to assess the capability to represent the flow structure produced by the jet–plate interaction with different meshes around the plate. Generally, the results show that the use of a coarse mesh and the standard k–ε turbulence model permits a reasonable representation of the erosion process, but with a systematic over prediction of the mixing time. The results with the reference model were more accurate for two experiments with two flowrates and same initial conditions and all complementary tests. For the third test with different initial and boundary conditions, however, poor results were obtained with the reference model, which could only be improved by further refining the mesh. These results indicate that a model "qualified" for certain conditions could be inadequate for other cases, and sensitivity studies are necessary for the specific conditions considered in the analyses.