Description

The 'Change Detection Analysis' crashes when you run a probabilistic DoD calculation based on a multi-method input mask that has data extents greater than that of the DEM for which it was derived. The problem is described briefly in the video below.

Status/Trigger

If you attempt to load a multi-method mask that has NoData extents that extend beyond that of your DEM the mask is for, everything works fine in the survey library (including calculation of multi-method error surface), but at the time of calculating change detection by probablistic thresholding (which uses propagated errors), there is a crash. We think the crash is caused by the software attempting to calculate a T-Score where there is no DoD value (i.e. the extents with NoData in DEM but data in the propagated errors).

Workaround(s)

Simply make sure that your multimethod polygon is clipped down to the NoData extents of your DEM.

Plans

1. Add in to the survey library a routine that checks the multi-method mask polygon when being loaded to see if it covers all the cells in the DEM and checks to see if it extend beyond the DEM. For areas extending beyond the DEM, they are clipped out by intersection with the actual data extents of the DEM. For areas inside the DEM, they are added and flagged as 'Uncategorized'.
   
2. At the stage of calculating an error grid from a multi-method mask, the raster is checked to have a mask the same as the data extents of the DEM. In all cases, the created raster must be made concurrent with the DEM it is associated with.
3. At the stage of calculating propagated errors, the input error surfaces are checked for concurrency and similar data extents.