Eelgrass (Zostera marina) is an ecologically significant and fragile species of seagrass common in Denmark and the Northern Hemisphere. The depth limit of the eelgrass populations is used to evaluate the ecological condition of coastal waters, and coverage is used for evaluation of ecosystem health. Satellite remote sensing has the potential to improve the cost effectiveness of the analysis significantly. Based on a review of existing reported methods, this thesis used Sentinel-2 imagery with object based image analysis and various machine learning algorithms to classify submerged aquatic vegetation at Roskilde Fjord. An ecological model of eelgrass stress parameters was applied to the classification output to produce an empirical classification of eelgrass coverage. The results indicate that Random Forest is the most suitable machine learning algorithm for submerged aquatic vegetation classification, and a scale parameter of 10 produces image objects that obtain the highest classification accuracy. Water column correction and multi-temporal analysis are demonstrated as techniques to improve classification accuracy. The thesis concludes Sentinel-2 imagery may be used for mapping submerged aquatic vegetation but not for the specific identification and analysis of eelgrass.