The use of in situ gamma spectrometry is increasing nowadays thanks to its fast (immediate) results, good detection limits (comparable with laboratory results), and a good representativity of the average of the activity levels over the field under study. The principle is to place a Ge detector above the site under study and, from the measured spectra, to identify radionuclides in the soil, and estimate their activities and their associated dose rates. Despite the performance of this technique, it has been found that at some intercomparisons, there have been discrepancies between results from the in-situ analysis and the target values. In this work, Monte Carlo simulation has been used in order to understand these discrepancies. The Monte Carlo study has been separated in two stages: the detector model and the model of the photon flux impinging on the detector. The development permitted to understand the origin of the in-situ spectra, the main parameters affecting the spectra shape, the origin of the detected photons as well as the different contributions from radioactivity. This study permitted the discrepancies found during the intercomparisons. The model developed allows to improve the analysis by creating a spectra corresponding to different hypothesis i.e. allowing the localisation of point sources or the identification of the distribution of radionuclides in soil from the spectral shape.