Attempting to classify several (n = 8) rasters into classes. The rasters have a cell size of 50 x 50 m and occur across a 200,000 km2 area!
We are attempting a landscape classification but note that conventional algorithms are struggling to cope with the size and multidimensional nature of the classification. Further, we are looking for an unsupervised and ultimately a hierarchical classification.
Any advice or reccomended workflows would be greatly appreciated.
If you have a high dimensionality in your data and mainly numeric variables (or you might convert them) it could be worth looking at t-SNE algorithms that should be able to help you find relevant clusters.
It is powerful for unsupervised and anomaly detection. You would have to do some interpretation work since results might not be initially straightforward.