I'm modeling wind stress toward a sloping shoreline. I get a blowup error at the same time and place that the water temperature spikes to hundreds of degrees Celsius, negative and positive. This problem sounds somewhat like the "Trouble in shallow water" discussion of Bristol Bay back in 2003, but since that was five years ago I though I'd start a new topic.
The temperature spike happens at the shallow water where storm surge is wetting the formerly dry land, but not up in the "draped" area. There is turbulence here: u and w are on the order of 0.1 meters per second. DT=2.0 seconds and DCRIT=0.1 meters. 10 vertical levels. I may have a CFL violation but I didn't expect to boil the water.
Can anyone suggest what might be going on here, and how to fix it?
Carl
Water temperature from -492 to 602 degrees Celsius?
Re: Water temperature from -492 to 602 degrees Celsius?
My Bristol Bay issues were much more recent than 2003. I found that things behave better with dcrit of 0.5, but then my whole domain is much bigger and deeper than yours. If you were to hit a CFL problem, that's exactly what would happen - things grow exponentially. You can get negative salinities which is what ultimately makes the model blow up.
-
drews
- Posts: 35
- Joined: Tue Jun 19, 2007 3:32 pm
- Location: National Center for Atmospheric Research
- Contact:
Re: Water temperature from -492 to 602 degrees Celsius?
Kate -
Thanks for the suggestions. (Sorry, I was looking at your Joined date, not the Bristol Bay Posted date.)
I'm using a horizontal resolution of 100 meters, with a strong wind stress of 2.73 N/m2 blowing toward shore. Steep gradients in the bathymetry (1:20) blow up sooner than shallow gradients. Increasing DCRIT from 0.1 to 0.5 meters appears to help somewhat, but only postpones the blowup error by about 30 hours.
I haven't written a CFL scanner yet, but my spot-checks indicate that CFL conditions are still okay. Decreasing DT from 2.0 to 1.0 seconds causes the blowup to happen 3 hours SOONER. Decreasing the number of vertical levels gives a small improvement, and doubling NDTFAST makes no difference.
Total kinetic energy peaks at about 40 hours, then drops just as the temperature anomalies start to build up. Can this be runaway convection, given that I am putting a lot of kinetic energy into the system at the surface? There is downwelling near shore and strong upwelling out to sea (at the domain boundary). The initial potential temperature is a uniform 20 degrees, but the later profiles show hot water under cold water. It sounds like Hernan's description from the "Blow up" thread on May 27, 2004: "a lot of convection takes place with huge vertical velocities."
I tried #defining MY25_MIXING, using Jamie Pringle's suggestion from May 17, 2008 "Left side of the upwelling example". All the problems go away! The potential temperature remains a uniform 20 degrees. The circulation changes somewhat, but still looks reasonable to my eye. GLS_MIXING also fixes the problem, and the only small difference is the behavior at the domain boundary. I'll go read up on the MY25_MIXING and GLS_MIXING options some more, but if anyone wants to comment further, please feel free.
Carl
Thanks for the suggestions. (Sorry, I was looking at your Joined date, not the Bristol Bay Posted date.)
I'm using a horizontal resolution of 100 meters, with a strong wind stress of 2.73 N/m2 blowing toward shore. Steep gradients in the bathymetry (1:20) blow up sooner than shallow gradients. Increasing DCRIT from 0.1 to 0.5 meters appears to help somewhat, but only postpones the blowup error by about 30 hours.
I haven't written a CFL scanner yet, but my spot-checks indicate that CFL conditions are still okay. Decreasing DT from 2.0 to 1.0 seconds causes the blowup to happen 3 hours SOONER. Decreasing the number of vertical levels gives a small improvement, and doubling NDTFAST makes no difference.
Total kinetic energy peaks at about 40 hours, then drops just as the temperature anomalies start to build up. Can this be runaway convection, given that I am putting a lot of kinetic energy into the system at the surface? There is downwelling near shore and strong upwelling out to sea (at the domain boundary). The initial potential temperature is a uniform 20 degrees, but the later profiles show hot water under cold water. It sounds like Hernan's description from the "Blow up" thread on May 27, 2004: "a lot of convection takes place with huge vertical velocities."
I tried #defining MY25_MIXING, using Jamie Pringle's suggestion from May 17, 2008 "Left side of the upwelling example". All the problems go away! The potential temperature remains a uniform 20 degrees. The circulation changes somewhat, but still looks reasonable to my eye. GLS_MIXING also fixes the problem, and the only small difference is the behavior at the domain boundary. I'll go read up on the MY25_MIXING and GLS_MIXING options some more, but if anyone wants to comment further, please feel free.
Carl