Geomorphic Change Detection Software

Error Propagation

Background Theory on Error Propagation

Error Propagation Theory Based on Minimum Level of Detection Logic

How does GCD calculate +/- Error Volumes?

The +/- error volumes are very conservatively estimated with error propagation on a cell-by-cell basis. For every change detection that is done based on error surfaces in the inputs, a properror.tif is calculated. This is nothing more than:

PropError = √((ErrorDEM New)^2 + (ErrorDEM Old)^2)

This is just error propagation on a cell by cell basis to get an estimate of propagated vertical error. Note that when using a minimum level of detection that is spatially uniform, the MinLoD is effectively the propagated error so it is not calculated separately.

The error volume is estimated just like change detection values are estimated: by multiplying vertical value (in this case propagated DEM error instead of elevation change) by the cell area, which is just the square of cell resolution.

Cell Error Volume = Cell-Resolution^2 * PropError

For the summary tabular statistics, these are simply summed up for erosion cells and deposition cells independently the same way that volumetric estimates of change are. In the video below, illustrate how GCD calculates error volumes under different thresholding techniques:


Further Reading on Error Propagation