Within the widely investigated field of forest disturbance monitoring, the detection of forest storm damages with remote sensing techniques gained rather low attention in the last years. This work aims to fill this gap. The project of storm damage detection, focusing on spruce forests, was initiated by the Norwegian Forest and Landscape Institute ‘Skog og Landskap’. The triggering event for this investigation was the storm ‘Dagmar’ from December 2011. This storm event and its impact on spruce forests on Norway´s west coast are investigated to develop a semi-automatic storm damage detection model. For detecting storm damages, primarily the question of adequate data pre-processing of Landsat 7 ETM+ is discussed. In the pre-processing stage, haze reduction, image-to-image registration, atmospheric and topographic correction are applied. The ‘Wide Dynamic Range Vegetation Index’ (WDRVI) is analysed and evaluated for its applicability when detecting forest storm damages. Pixel information from known storm areas is extracted, and compared with a focus on data distribution and the trend behaviour for different damage categories. A correlation was detected between the data trend of the WDRVI and the increasing damage percentages in the forest, showing an increase in WDRVI values for increasing damage percentages in the observed forest stands. Therefore, the WDRVI provides the best possibilities to detect storm damages in the study area. Through a non-linear regression analysis and ‘Partitioning Around Medoids’ classification (PAM), thresholds are derived from the WDRVI change image. Implementing those thresholds in an ERDAS 2013 spatial model, a tool is developed, which detects forest changes without the requirement of further user input. The only requirements are pre-processed Landsat 7 images before and after the storm, and a defined area of interest data (AOI), e.g. a vector-mask of spruce forests. Testing and evaluating the semiautomatic detection model on a larger AOI (covering almost a whole Landsat 7 scene) achieved an overall accuracy of 96.3% (Cohen’s KAPPA of 0.94). With very good detection results, this investigation contributes to forest management and a faster response to storm damaged forest areas.