Numerical modelling of snow cover stratigraphy with, for example, the 1-D snow cover model SNOWPACK has the potential to increase the spatial and temporal resolution of snow stratigraphy information – data very much needed for avalanche forecasting. One of the key properties for interpreting snow stratigraphy in regard to stability is snow hardness. In manually observed snow profiles, differences in snow hardness between layers were found to be an indicator of instability. We improved the hardness parameterization implemented in the snow cover model SNOWPACK. Hardness is estimated from simulated snow density and grain shape. Using ordinal logistic regressions we calculated for the principal grain shapes the threshold density for all hardness steps (on a dataset of 14,521 manually observed layers). We thus implemented snow hardness as a discrete parameter in SNOWPACK. The structural stability index (SSI), and the threshold sum approach (TSA) were then used to detect potentially weak layers. Both indices strongly depend on the hardness parameterization. With the new hardness parameterization the agreement between measured and simulated snow hardness is fair. Furthermore, it does not require any further calibration and is thus more robust against future changes of the model. Potentially weak layers detected in simulated stratigraphy with either the SSI or the TSA corresponded in about half of the cases to observed CT failure layers. The correspondence improved if only sudden collapse fracture were considered. These preliminary results are promising as they suggest that stability information can be derived from simulated stratigraphy in particular after further improving the detection methods.