Spatial land cover pattern analysis

In the previous chapters we introduced land cover classifications, fractional cover and time-series analysis. All these approaches aimed to extract ecological relevant information based on the spectral signal. However differentiating a tree plantation (spatially regularly planted trees of same species, age, height) from a natural forest based on the spectral signal only might be quite challenging since the spectral signals might be quite similar but their spatial heterogeneity is different. A tree plantation will not have a high spatial variation in its spectral signal due to the same age and height of the trees while a natural forest will have different tree heights with casting shadows or even tree fall gaps, hence such a forest will show up with a higher spatial variation. Such information can be retrieved using texture metrics based on remote sensing data sets e.g. the NDVI.