Because of the averaging effect of focal statistics, the
resulting spatially autocorrelated error raster has a much lower
standard deviation than 1. The Zonal Statistics tool determines the standard deviation of the error raster and the Divide tool
produces a new error raster with a standard deviation of 1.
The new error raster is multiplied with the user-defined Root
Mean Square Error (RMSE) in vertical map units of the DEM.
The RMSE value can be altered to simulate the effect of small
and large errors. The final error raster is therefore a spatially
autocorrelated error with a mean of zero and a standard deviation
equal to the user-defined RMSE value. This error term is then
added to the original DEM; this new DEM becomes the input
into the stream delineation model.
The second part of the model, shown in Figure 2, creates the
stream network: sinks are filled, flow direction and flow
accumulation are derived from the DEM and streams are
delineated based on a constant flow accumulation threshold.