Thesis defense // Saint-Martin SAINT-LOUIS

THESIS

Development of infiltrometric tools and methods for the characterization of urban soils.

Abstract

The hydrodynamic characterization of soils is a preliminary and essential step in understanding water fluxes and pollutant transfers within the unsaturated zone. Soil hydrodynamic properties are thus used to predict water infiltration, surface runoff, groundwater recharge, and the associated contamination risks. They play a decisive role in the planning of urban and agricultural areas, as well as in the management of water resources and stormwater in urban environments. As part of this thesis, we investigated the Beerkan method and the BEST algorithms, which enable the processing of infiltrometric data and the characterization of soil hydraulic properties. We specifically focused on two variants of the Beerkan method: the traditional manual approach and an automated version based on an infiltrometer prototype specifically developed prior to this work. We also investigated new mathematical approaches to data processing, adapted to particular configurations and specific physical processes (notably soil hydrophobicity). This manuscript is structured around three main axes: (i) a comparative study between the classical and automated Beerkan methods, (ii) an analysis of the influence of the infiltrometer radius on data quality, (iii) an application to a large number of experimental sites within the same urban fabric, allowing us to test the automated version and validate the series of BEST algorithms, including their recent developments. The results provide several insights. The automated infiltrometer presents numerous advantages over the classical method, particularly in terms of accuracy, reproducibility, and ease of use. However, the manual method remains highly relevant in contexts with limited resources. Furthermore, the size of the infiltrometer proves to be crucial: a larger radius promotes the activation of macropore networks and thus better accounts for preferential flow. Finally, deployment across numerous sites validated the approach, with satisfactory results in the majority of cases. This study also highlights the need to continue developing alternative algorithms to better represent specific flow processes, particularly in hydrophobic soils.

Thesis director

Laurent LASSABATERE (ICTPE Engineer, ENTPE)