I'm trying to use the Forest-based Classification and Regression tool to develop a model to predict (to Raster) the distribution of different tree species from Multi-Spectral Imagery. Thus far, I have found the results encouraging, however, I cannot seem to get the tool to output the uncertainty raster that I would like to use to thin out the results to show only higher confidence outputs.
When preparing to run the tool, the option to Calculate Uncertainty is available. However, upon entering my training data, which is of the categorical variety, the option disappears. Once I run the tool and view the parameters in the process window, the option to calculate uncertainty does appear to remain 'TRUE', however, the 'Output Uncertainty Raster Layers' field is empty, and when the tool completes, no Uncertainty Raster appears to be created.
The check-box always disappears from the window after I enter my categorical training data, does Uncertainty only work with non-categorical data? I don't recall seeing that anywhere in the documentation, but I certainly may have overlooked it.
If calculating uncertainty is not possible in this case, any suggestions to achieve this goal with a different method or work-around would be appreciated.