Initialization with solid files fails.

[smithsg84]

Hi All-

During a short course we are holding here at Mines we discovered a bug in recent versions of ParFlow. The bug is in initialization with solid files that are vertically smaller than the computational domain; domains get huge initial pressures. This should have been caught by the ‘crater 2d’ regression test but somehow wasn’t. I’ve attached the script and solid file and have examples of correct and incorrect output. Nick and I had some prior versions of PF on our laptops (this is replicated on Ubuntu Linux too) and were able to narrow this down to about version 713. Looking at the check ins post 713, some of real space Z stuff was added and that might be the issue. Can we take a look at this and add the 1-d.overland as a test case?

let me know if you have more questions / need more information. Thank you!

Reed

version 713 and earlier
Correct output:
1 0.0
2 0.02928065179838431
3 0.10014436117193659
4 0.1826876221729191
5 0.2512343882653141
6 0.2959960006531399
7 0.3207496872250761
8 0.33297304580994486
9 0.3385580459293718

version somewhere 713 and later
Incorrect output:
1 451.7911349432279
2 36.46400911410956
3 6.671697940939478
4 2.085836404414654
5 0.9600865289004944
6 0.5913155802692779
7 0.44863761458028484
8 0.38901842943816844
9 0.363707749764536

files.zip

[smithsg84]

Problem occurs between 713 and 735. We get the issue if the DZ=0.05 (as set in the script) and a constant value of 15 (total domain size) minus 1/2*DZ is getting added to the hydrostatic IC. If we make the DZ=1 then we don’t get any differences between the versions. The Crater2D has a DZ=1 (I think this is why it passes in the regression test).