Conclusions
Sentinel-2 (10m) based mapping aids small scale (>1 ha)
vegetation monitoring and management, even Landsat (30m) is useful in
small scale mapping of vegetation and changes herein. By using off the
shelf images and methods 60-70% accuracy was achieved for single
images, while stacking year-round images yielded 86% accuracy.
Accuracies of 60-70% are already good enough to inform, prioritize and
decide in vegetation management. End-users need time to adjust and trust
a new data-source, it’s not only about the accuracy numbers. Inclusion
of Google Earth Engine simplified building prototypes and an operational
platform. Developing the tool in close cooperation with end-users helped
to quickly identify unanticipated uses, useful features and facilitated
trust in the tool.