Many predictive models are used to map the spatial distribution of surface processes and landforms, especially in rough terrain or limited accessibility areas. In the present study a statistical approach was used to identify the areas with highest probability for the occurrence of block streams in alpine area of Retezat Mountains. The aim of this approach is to improve the semi-automated digital mapping and to reduce the field work related with the quantification of these landforms. A multiple linear regression analysis and GIS-based technology was used to identify the areas with possible block streams occurrence. In the study area, 82 block streams were mapped in the field using a differential GPS. The sampling strategy was focused on uniform coverage of the alpine area. Using block streams contours the explanatory variables were determined on fieldwork, or extracted from a digital elevation model. Statistical analysis emphasized the most useful variables for the multiple linear regression equation. The accuracy of the models was statistically and spatially calculated using a validation dataset. Validation dataset was randomly selected from the mapped field samples. The resulted models achieved an accuracy ranged from 65 % to 80 % and the predicted area for block streams occurrence ranged from 26.3% to 39%.