# Richards flow: numerical instability in calculation of velocity

**URL:** <https://discourse.opengeosys.org/t/richards-flow-numerical-instability-in-calculation-of-velocity/538>\
**Category:** Help & Support\
**Tags:** ogs5, ogs5-input-files, ogs5-keywords\
**Created:** [June 10, 2020, 5:18pm UTC](https://discourse.opengeosys.org/t/richards-flow-numerical-instability-in-calculation-of-velocity/538 "2020-06-10T17:18:41Z")\
**Posts on this page:** 1\
**Showing post:** 1

<div class="post-metadata">

**Author:** ![swamini.khurana](https://discourse.opengeosys.org/user_avatar/discourse.opengeosys.org/swamini.khurana/32/692_2.png) [@swamini.khurana](https://discourse.opengeosys.org/u/swamini.khurana)\
**Post date:** [June 10, 2020, 5:18pm UTC](https://discourse.opengeosys.org/t/richards-flow-numerical-instability-in-calculation-of-velocity/538/1 "2020-06-10T17:18:41Z")

</div>

Hello,

I am simulating Richards flow in OGS5 using the following conditions:

1. Initial condition - uniform pressure in the domain

2. Boundary conditions:

- Inlet (line)- constant source term (CONSTANT\_NEUMANN)
- Outlet (line)- constant pressure

1. Numerical solver settings:

"  
#NUMERICS  
$PCS\_TYPE  
RICHARDS\_FLOW  
$ELE\_UPWINDING  
0.5  
$ELE\_MASS\_LUMPING  
1  
$ELE\_GAUSS\_POINTS  
3  
$LINEAR\_SOLVER  
; method error\_tolerance max\_iterations theta precond storage  
3 6 1.e-015 5000 1.0 100 4  
$NON\_LINEAR\_SOLVER  
; method error\_tolerance max\_iterations relaxation  
PICARD 1.0e-15 5000 0.0  
"

The pressure and saturation calculations are stable over time in all tested cases, while the calculation of velocity at each node varies over time (varies up to 3-4% of expected value in some cases). I understand that this might be the case if the solver focuses on precision in the calculation of the primary variable (Pressure in this case), while it does not follow through on the calculation of the secondary variable (velocity in this case). Is this true for OGS5? Is there any way I can address this (maybe change the solver or other parameters in the .NUM file)?

Regards  
Swamini

---

_[View the full topic](https://discourse.opengeosys.org/t/richards-flow-numerical-instability-in-calculation-of-velocity/538)._
