Abstract In the recent past, droughts and climate change-related impacts have led to declining vitality of forests in Central Europe. In combination with an increasing frequency of disturbances, this created a need for consistent large-scale monitoring of forest dynamics. Remote sensing and particularly new multispectral satellite missions now offer high-quality data streams that enable the development of new spatially continuous monitoring systems. However, current remote sensing-based operational forest monitoring systems typically suffer from either low spatial or temporal resolution and often also do not reach near real-time (NRT) capabilities. Here, we present a novel implementation of a forest monitoring system with high spatial and temporal resolution that is capable of providing NRT information on the state of forests for the entire federal state of Bavaria, Germany. Our monitoring system considers all relevant Sentinel-2 and Landsat-8/9 scenes since 2017 and utilizes the disturbance index to detect canopy cover loss on a biweekly basis for historic time series as well as in the form of NRT updates. For the latter, a computationally efficient data-driven method is implemented to detect disturbances for each new time step. An ancillary context layer is created to ensure robustness of NRT results. Our combined approach shows an overall accuracy of 0.85 with an F1 score of 0.81. Different spatial aggregation units reveal distinct hot spots of forest canopy cover loss across entire Bavaria with a total area of about 161 000 ha lost in a 7-year time period, including single districts losing over 20% of forest canopy cover. Our NRT component is able to contribute in closing the gap of timely information on canopy cover loss dynamics.
Jaggy et al. (Thu,) studied this question.