http://www.subsim.com/radioroom/showthread.php?t=124973
Sonoboy's tests showed a link between detecting platform depth and detection range, when the contact was at the surface and the detecting platform was below the layer in a surface duct SSP. My interpretation of his data is that DW increases modeled transmission loss at higher velocity points on the SSP. I doubt this is only applicable to cross-layer detections.
I have accidentally attributed this data to Dr. Sid in the past, so I apologize to Sonoboy.
EDIT: I can't say I fully understand Dr. Sid's data on the shadow zone (the half on the SNR vs range is simple enough though) but my best guess is that it shows a "longer" shadow zone at depths closer to the layer--that the shadow zone extends closer to the detecting platfrom at shallower depths. That is completely consistent with Sonoboy's data showing a significant correlation between depth and detection range.