Abstract. Geophysical methods are widely used to investigate the influence of climate change on alpine permafrost. Methods sensitive to the electrical properties, such as electrical resistivity tomography (ERT), are the most popular in permafrost investigations. However, the necessity to have a good galvanic contact between the electrodes and the ground in order to inject high current densities is a main limitation of ERT. Several studies have demonstrated the potential of refraction seismic tomography (RST) to overcome the limitations of ERT and to monitor permafrost processes. Seismic methods are sensitive to contrasts in the seismic velocities of unfrozen and frozen media and thus, RST has been successfully applied to monitor seasonal variations in the active layer. However, uncertainties in the resolved models, such as underestimated seismic velocities, and the associated interpretation errors are seldom addressed. To fill this gap, in this study we review existing literature regarding refraction seismic investigations in alpine permafrost permitting to develop conceptual models illustrating different subsurface conditions associated to seasonal variations. We use these models to conduct a careful numerical study aiming at a better understanding of the reconstruction capabilities of standard and constrained RST approaches. Our results demonstrate, that the incorporation of structural constraints in the inversion and the usage of constrained initial models help to better resolve the geometry and the velocity structure of the true models. Moreover, we present the successful application of this extended constrained approach for the inversion of refraction seismic data acquired at Hoher Sonnblick (Austria) by incorporating complementary information obtained from the modelling of ground-penetrating radar (GPR) signatures. In conclusion, our study shows the potential of an extended constrained RST to improve the reconstruction of subsurface units and the associated seismic velocities in a permafrost environment, permitting to reduce the uncertainties in the interpretation of the imaging results.