Model blow up

Discussion about modeling ice with ROMS

Moderators: arango, robertson

Post Reply
Message
Author
hqz
Posts: 10
Joined: Tue Dec 23, 2014 7:06 pm
Location: Ho Hai university

Model blow up

#1 Post by hqz » Sat Dec 24, 2016 1:27 pm

Deal all,
I run my model with ecmwf surface forcing(only contain sustr,svstr,shflux,swflux,swrad),tide and river,and the hmin=5m,hmax=5000m,and the initial condition come from hycom ,the boundary condition come from SOAD.my model blow up(fig 1) ,in the first time(dt=180s), it is a point with too large speed in ubar field,then i change the dt from 180s to 120s, and it blow up too ,and in the last record of rst file ,the zeta ,ubar,vbar, u, v, temp and salt is all nan in some region(around Korea)(fig 2 show the zeta).
Can someone give me some suggestions . Thank you very much.
Attachments
upwelling.h
(3.68 KiB) Downloaded 131 times
ocean_upwelling.in
(118.46 KiB) Downloaded 130 times
fig2.png
fig1.png
fig1.png (20.22 KiB) Viewed 5941 times

User avatar
kate
Posts: 3678
Joined: Wed Jul 02, 2003 5:29 pm
Location: IMS/UAF, USA

Re: Model blow up

#2 Post by kate » Sat Dec 24, 2016 4:05 pm

You could try a shorter timestep yet again. If the model blows up at the same model time with that, then you know it's something else, not the timestep. Once things have gone to NaN, it's hard to know what's the cause. You could run to a step or two before the blow up and plot those fields.

User avatar
kate
Posts: 3678
Joined: Wed Jul 02, 2003 5:29 pm
Location: IMS/UAF, USA

Re: Model blow up

#3 Post by kate » Sun Dec 25, 2016 6:20 am

This doesn't belong in ROMS-ice, does it? Are you running my branch with the sea-ice turned on?

hqz
Posts: 10
Joined: Tue Dec 23, 2014 7:06 pm
Location: Ho Hai university

Re: Model blow up

#4 Post by hqz » Sun Dec 25, 2016 6:41 am

This doesn't belong in ROMS-ice,maybe i post my problem in wrong topic.I have try a shorter timestep(60s),the model blows up at the same model time with a large speed point in ubar field(i=314,j=183).And i find the h+zeta near this point (i=314,j=184)with value -0.1243 small than zero.I think this would be the reason lead to model blow up.And i find the zeta value(Fig1 last record from -6.2 to 4.2 m) is strange to the value of soda monthly data(-0.1296 to 1.5208 m in the average of 1999 to 2010 in my research region).Could anyone tell me what may lead to my zeta field different from SOAD value.
Attachments
fig1.png
fig1.png (13.69 KiB) Viewed 5910 times

User avatar
kate
Posts: 3678
Joined: Wed Jul 02, 2003 5:29 pm
Location: IMS/UAF, USA

Re: Model blow up

#5 Post by kate » Mon Dec 26, 2016 5:03 am

You've got tides turned on. Are there large tides in that bay? If your model tides don't match the real ones it could be the bathymetry. What is the source of your bathymetry and how good is it? You could also turn on the WET_DRY option so that the model won't let h+zeta go negative. We have resorted to modifying the bottom drag to change the tides too.

hqz
Posts: 10
Joined: Tue Dec 23, 2014 7:06 pm
Location: Ho Hai university

Re: Model blow up

#6 Post by hqz » Mon Dec 26, 2016 5:55 am

I have add 10 tidal component(m2,s2,n2,k2,k1,o1,p1,q1,mm,mf come from Tpxo8)to my model,,and first i run the model with my tide forcing in 2D dimension about 200 day(test.h and test_tide.in,no other forcing and boundary add).The model never blow up.The bathymetry of my region come from General Bathymetric Chart of the Oceans(GEBCO),i set the hmin=5 and hmax=5000,then make some smooth by rgrd_smoothgrid.m(rfac=0.35).Then i add the initial from hycom, surface forcing from ecmwf,boudary from soda2.2.4 and river,the model blow up. Maybe i should try to adjust my min depth or add WET_DRY.
Attachments
rgrd_smoothgrid.m
(2.44 KiB) Downloaded 125 times
test_tide.in
(117.63 KiB) Downloaded 131 times
test.h
(1.5 KiB) Downloaded 122 times

User avatar
kate
Posts: 3678
Joined: Wed Jul 02, 2003 5:29 pm
Location: IMS/UAF, USA

Re: Model blow up

#7 Post by kate » Mon Dec 26, 2016 6:26 am

You added a lot of things all at once. The tidal run might have had high tides in that same place and simply not blown up because there was less going on. I would try the WET_DRY.

GEBCO is a coarse and smooth bathymetry. Is there nothing better at high resolution in that area?

hqz
Posts: 10
Joined: Tue Dec 23, 2014 7:06 pm
Location: Ho Hai university

Re: Model blow up

#8 Post by hqz » Mon Dec 26, 2016 7:04 am

I would try add wet_dry.The GEBOO i use is about 30 arc-second,and the other bathymetry i can get is ETOPO1 bathymery.

User avatar
kate
Posts: 3678
Joined: Wed Jul 02, 2003 5:29 pm
Location: IMS/UAF, USA

Re: Model blow up

#9 Post by kate » Mon Dec 26, 2016 4:04 pm

It's been a while since I looked at GEBCO. The problem with ETOPO1 is that it has inputs from satellite altimetry and it therefore has weird (unbelievable) highs and lows.

hqz
Posts: 10
Joined: Tue Dec 23, 2014 7:06 pm
Location: Ho Hai university

Re: Model blow up

#10 Post by hqz » Wed Dec 28, 2016 1:16 am

I turn on the WET_DRY(dcrit=0.1),and the model blow up again in the model time 712 day with speed v field in 12 sigma layer(total 20)(fig1),Can someone give me some suggestions.Thank you very much.
Attachments
fig1.png
fig1.png (6.38 KiB) Viewed 5830 times

User avatar
kate
Posts: 3678
Joined: Wed Jul 02, 2003 5:29 pm
Location: IMS/UAF, USA

Re: Model blow up

#11 Post by kate » Wed Dec 28, 2016 6:08 am

Reduce the timestep until you get past the blowup time?

hqz
Posts: 10
Joined: Tue Dec 23, 2014 7:06 pm
Location: Ho Hai university

Re: Model blow up

#12 Post by hqz » Thu Dec 29, 2016 12:54 am

Do you mean reduce dt and run from rst file?

User avatar
kate
Posts: 3678
Joined: Wed Jul 02, 2003 5:29 pm
Location: IMS/UAF, USA

Re: Model blow up

#13 Post by kate » Thu Dec 29, 2016 1:14 am

Yes. Try it and see if it helps.

Post Reply