Quantification of ground ice is crucial for understanding permafrost systems and modeling their ongoing degradation. The volumetric ice content is however rarely estimated in permafrost studies, as it is particularly difficult to retrieve. Standard borehole temperature monitoring is unable to provide any ice content estimation, whereas non-invasive geophysical techniques, such as refraction seismic and electrical resistivity measurements can yield information to assess the subsurface ice distribution. Electrical and seismic data are hereby complementary sensitive to the phase change. A petrophysical joint inversion was recently developed to determine volumetric water, air, ice and rock contents from electrical and seismic data using a petrophysical model, but was so far only tested on synthetic data and one proof-of-concept field example. In order to evaluate its applicability on different types of permafrost materials and landforms (bedrock, rock glacier, talus slope), we apply this petrophysical joint inversion scheme to five profiles located in the northwestern Alps. The electrical mixing rule (Archie's second law) was hereby identified as a source of model uncertainty, as it applies only when the electrolytic conduction is the dominating process. We therefore investigate and compare four petrophysical models linking the electrical resistivity with the ground constituents: Archie's law, Archie's law with an additional surface conduction factor, a model considering only surface conduction, and the geometric mean model. In most cases, the three first resistivity relations yield largely comparable results, whose reliability is discussed. The geometric mean model better resolve high ice content, as it is less influenced by the ice-rock ambiguity. We perform a systematic analysis of the regularization parameters and then evaluate our results with validation data including thaw depths and ice contents derived from borehole measurements. Geophysical surveys have generally a lower resolution than borehole data, but have the advantage of providing spatio-temporal information in 2D or 3D. The joint inversion results are in relatively good agreement with the validation data for all sites from ice-poor to ice-rich conditions, when choosing the most adequate resistivity model and porosity initial value. Additional forcing constraints (e.g., porosity range constraint) based on site knowledge can improve the model parameter estimation.