Avalanche forecasting is an iterative process, where forecasters use weather data and snow observations in addition to previous assessments to conclude what forecast to publish. This project investigates how the forecasting process could be automated, using three seasons worth of data from 23 of Norway’s avalanche forecasting regions. Three scenarios were considered, using different amounts of input parameters based on what data would be available to the model in each respective scenario. For each scenario a machine learning model was trained, and a separate naïve model was constructed. The machine learning model could only beat the naïve model in the simplest scenario, using only weather data. In the other scenarios it was found that the data representation was lacking; highly intermittent snow observation data was structured as timeseries when a more preprocessed representation may have been more fruitful / Snow Models and Automatization in Geohazard-Forecasting