Multispectral imagery from UAVs to classify and monitor vegetation change in semi-natural grasslands

The use of multispectral imagery achieved from fixed-wing unmanned aerial vehicles (UAVs) has the capability to deliver high resolution imagery that can be used to monitor vegetative composition and change in semi-natural grasslands. Using a multispectral camera mounted on a fixed wing UAV with autonomous flight capability, this study is classifying and monitoring vegetation at three sites, each >70 ha, in mid-Wales, across a three-year period. Acquired imagery is stitched and georeferenced orthomosaics produced. Vegetation classification is achieved through a supervised pixel-based classification using the random forest classifier, accessed from the open-source python interface software package RSGISLib. The relatively low cost of this method has the potential to deliver a wide range of environmental and agricultural services on differing types of grasslands and wider vegetative communities; whether it be increasing the accuracy and speed of habitat surveys in categorising vegetation (compared with e.g. Joint Nature Conservation Committee Phase 1 surveying), understanding how different livestock grazing species utilise different areas and patches of vegetation, or investigating how management techniques such as grazing, cutting or burning alter the structure of a vegetative community.