ISLAND LAND-COVER MAPPING
GSD has in house experts experienced in identifying land cover from remotely sensed data - satellite Earth Observation. GSD tackled this task over a multitude of spatial and temporal scales. At GSD we pride ourselves in keeping abreast with the latest mapping techniques, to stay on the cutting edge of remote sensing and provide the highest quality with upmost confidence in our results.
In the case of this project mapping in the Comoros Islands, the goals of this project were to create and deploy a mapping framework that could accurately map the technically challenging terrain to identify in particular both natural and disturbed forest types. This was to be achieved using a framework that was solely built on utilising freely available satellite data.
To achieve these aims, GSD consulted with the client to build a method from the ground up, discussing the land cover types they were aiming to map and how best this might be done at different scales. GSD then worked to build a training data library so that a machine learning classifier could be deployed for optimum identification of land covers. To overcome the unique challenges when mapping the tropical and high rainfall forests of the study area, GSD first explored the available satellite platforms, from optical, radar and lidar. Here GSD incorporated multiple sensors together, to provide a more consistent product that mitigated against single sensor inhibition. As well as this, GSD also developed seasonal compositing techniques to create high quality cloud free optical images. This satellite data was then used to train a high-performance machine learning classifier, to accurately identify forest classes and provide mapping products that could be used for client analysis.