As part of a multi-laboratory team, we are carrying out two in situ thermal tests — the single Heater Test and Drift Scale Test, in an underground facility at Yucca Mountain, Nevada, USA, the proposed site for a high-level nuclear waste repository. Our objective in these tests is to gain a more in-depth understanding of the coupled thermal–hydrological–mechanical–chemical processes likely to exist in the fractured rock mass around a geological repository . These coupled processes are monitored continuously by numerous sensors emplaced in boreholes, while cross-hole radar tomography , neutron logging , electrical resistivity tomography, and interference air-permeability tests all serve to measure moisture change in the rock mass. Thermal–hydrological processes for both tests have been simulated (using a 3-D numerical model) and compared to the extensive data set. In this paper, we present examples to illustrate how an iterative approach requiring close integration of modeling and measurements enables us to track the complex coupled processes we seek to understand. The main manifestation of coupled thermal-hydrological processes is in the time evolution of the drying and condensation zones. Good agreement exists between model predictions and measurements, specifically the decrease in air-permeability values within zones of increased liquid saturation in the fractures and the increase of radar velocity in cross-hole radar survey in zones of decreased matrix liquid saturation. A heat-pipe signature in the temperature data arising from liquid–vapor counter-flow occurs in both the measurements and simulated results. The good agreement between predictions from the numerical simulations and measurements in the thermal tests indicates that our basic understanding of the thermal-hydrological processes in a potential repository at Yucca Mountain is sound. However, detailed behavior is impacted by site-specific heterogeneity, in the form of discrete fractures that are not likely to be predictable a priori. One emphasis of the on-going Drift Scale Test is to build on the present understanding and to assess the impact of heterogeneity to the repository performance.