I am working on a tree detection problem, with a study area spanning tens of thousands of kilometers. The soils across this area vary greatly in appearance, with no section (that I know of) possessing soils that reflect the full range of the larger study area. For this reason, I would like to use raster bands from multiple locations across the study area in training my unsupervised ISO Cluster model so I can use the same signature file to classify the whole area. However, leaving the environment extent as "Default" throws "ERROR 010568: Invalid extent," while setting extent to "Union of Inputs" throws "ERROR 010112: The number of classes (0) is less than 1." I am using the "Iso Cluster Unsupervised Classification" tool with number of classes set to 40, all other settings default. Is it possible to accomplish what I am trying to do, and if so, what am I doing wrong?
