Automatic Classification of Tree Species in RapidEye Data Using existing Core Service Data

For the generation of the service “Derivation of Forest Functional Parameters” it is a prerequisite to incorporate a forest species map for the differentiation of coniferous from broadleaf. However, the currently available Core Service layer has a too coarse spatial resolution of 20m compared with the very high 0.5m resolution of LiDAR data. Thus, it is a need to generate a higher resolution forest species layer in order to be used properly for the LiDAR processing line.

In order to fulfil this demand a RapidEye image with 5m resolution was ordered and classified. Therefore, a methodology has been developed which uses the existing Core Service product CS-EL-4B from the project Geoland2 as training data for the classification of the RapidEye scene.

That means the classification process does not need the cost-intensive ascertainment of training data, which implies that the incorporation of Core Service data is saving time and money.

The improved forest species layer, in this example the broad-leaved fraction, with 5m spatial resolution has then been used for the confinement of coniferous vs. broadleaves (see Figure below) in the LiDAR processing line.

The figure is showing the broadleaf fraction of the classified RapidEye image (upper part) in comparison to the Core Service product CS-EL-4B tree species type from Image 2006 (lower part).