[en] This research thesis reports the development of an analytic method based on a Monte Carlo simulation to quantify radionuclides present in soils by means of in-situ gamma spectrometry, to understand physical phenomena involved before and after detection, and to improve and complement results after spectrum analysis. The first part describes the evolution of in-situ gamma spectrometry: sensor development, in-situ measurement principle, evolution of the analysis principle. The second part introduces the Monte Carlo simulation and describes the used models (sensor model using the stripping method, development of a new simulation model for the incident flow). The third part discusses the understanding of an in-situ spectrum with the localization of the origin of incident photons and the identification of measurement parameters. Modelling results are then presented, as well as the development of spectrum de-convolution method, and the calculation of dose factors. Finally, the use of the 'Peak-to-Valley' method completed by the Monte Carlo simulation results is explained and used to localize a source depth and to define the exponential distribution of Cs-137 in Orsay