Experimental and numerical investigation of soil-atmosphere interaction

The soil-atmosphere interaction was investigated by conducting physical model test in the laboratory and by performing numerical analysis using data from an experimental site. e physical model test, a large scale environmental chamber was developed. This chamber was instrumented by various sensors allowing the soil suction, volumetric water content, and temperature to be monitored at various depths. In the zone occupied by air, temperature, relative humidity and air rate (wind speed) were monitored. The soil surface temperature was also measured by an infrared thermometer. The soil sample was prepared by compaction and wetted from the soil surface before being dried under controlled conditions of relative humidity, temperature and flow rate of air. A camera fixed above the chamber allowed monitoring the development of cracks. The evaporation rate was calculated based on the temperature and relative humidity data at the inlet and the outlet of the chamber, allowing assessment of the performance of the chamber to simulate the evaporation phenomena. The results show that evaporation is a heat consuming process: in the test condition, both air and soil were cooled by the evaporation process. Cracks developed during soil drying. Analysis of the ‘cracking surface ratio’ and ‘weighted width’ showed that the evolutions of these two parameters are similar: thus, in practice, only one is needed for evaporation analysis. e numerical analysis, a two-dimensional model of soil-atmosphere interaction was developed and implemented in a fully coupled thermo-hydro-mechanical (THM) code. The soil surface boundary conditions were determined from meteorological data using water balance and energy balance equations. The settlement due to soil-atmosphere interaction in an experimental site was simulated. Comparison between calculation and measurement showed that the THM codes can satisfactorily predict the soil settlement, provided that appropriate evapotranspiration models are implemented.