Wildfires increasingly threaten human health and infrastructure with consequences for forestry, agriculture, and biodiversity. Predictions show that climate change will likely increase the wildfire frequency and severity in the Alpine region. Providing high-quality data to estimate fire danger can improve resource planning of decision-makers and the timing and quality of early warnings for society. Forest fire danger forecasts are based on empirical or physical models which estimate the moisture levels of fuels as a function of weather conditions. These forecasts often use indices based on meteorological data, such as the Canadian Fire Weather Index (FWI). However, meteorological forecasts are typically only available at relatively coarse spatial resolutions (up to ca. 1 km), and therefore, of limited use in mountain regions with complex topography. Also, other factors, such as vegetation type and structural elements and the role of humans causing ignitions, are often not considered. Therefore, there is a need for an integrated wildfire danger assessment for Austria. The CONFIRM project, which started in December 2019 with funding from the Austrian Research Promotion Agency (FFG), tries to address this gap and develops a novel, high-resolution, and satellite-supported integrated forest fire danger system (IFDS) for Austria. For that purpose, radar and optical satellite data from the Copernicus Sentinel-1 and Sentinel-2 missions, airborne Laserscanning (ALS), socio-economic data, and topographic properties are used next to meteorological data. The project uses two independent methods: (i) an expert-based approach that allows a combination of various data layers with different weightings and (ii) a machine learning approach. Key stakeholders from national weather services, fire brigades, state forest administrations, and infrastructure providers are providing feedback on the prototype of the IFDS according to their needs and requirements. Here, we present the results of the machine learning approach for a study site covering the state of Styria (ca. 16 400 km ²). Several machine learning techniques have already proven suitable in similar studies (e.g. Random Forest and Maxent) are employed. We used satellite-derived moisture indicators and tree species classifications, ALS-derived vegetation structure parameters and irradiance, topographic and socio-economic data, and meteorological variables as input features to estimate fire danger. The predictors were trained using forest fire events from the Austrian forest fire database, which occurred between 2016 and 2021. The precision metrics used in the course of spatial cross-validation show that the best performing model manages to predict high fire danger for the majority of fire events.