The parameter set up of mass flow models decides on the quality of the results. We present a novel parameter optimization procedure, building on the comparison of model results and observed deposition areas. As it is evident that natural conditions cannot be represented by one constant value for larger areas we suggest using input value ranges rather than discrete values. The proposed procedure consists in a stepwise optimization of the input parameter values and ranges by using the precedent results as input for the next step. The procedure is applicable to any type of single-value input parameter and any type of mass flow propagation model. We use two GIS-based open source modelling tools: (i) the conceptual r.randomwalk and (ii) the complex physically-based r.avaflow. Both tools are capable to process value ranges of input parameters. The still ongoing study aims to derive reliable guiding values for various process types to be used for forward calculations. We demonstrate the procedure using two rock avalanche events (Acheron and Black Rapids). The first outcomes are promising, but more work is necessary to confirm and extend the validity of the results.