boundary for biogeochemical-ecological modeling

[chaoyu1985]

Hi, i am runing npzd. In my study area,alkalinity,NH4,LdetritusN,SDeN,LdetritusC,TIC data is absent. I can not find any other reanalysis data or simulation data from internet. So,I used a constent value for all grids and used as my boundary condition. the same value for all time series (12 months). the value was picked up from a literature. the model can not run sucessuflly. should I set the boundary condition changes with time?
I used climate data (monthly mean value)for NO3, O2, and so on (COADS 05 and WOA2009). I used modis chlorophyll monthly mean value in 2016 as the initial data and forcing condition. phytoplankton was set by chlorophyll.

now, the model blow up.
i was wondering if the forcing condition (U,V, pair, tair) are not match with NH4 (constants for all grids)? should they match with the changes of the chlorophyll and phytoplankton?

what should i do when the model blow up? should i try to track the fortran code and find which step is something wrong?
thank you so much for your help in advance, best wishes

Code: Select all

       326 0001-01-03 00:54:00.00  1.805362E-02  1.009324E+02  1.009504E+02  4.729840E+11
                     (004,028,18)  3.956102E-01  1.365641E-01  1.130226E+01  4.059272E+00
       327 0001-01-03 01:03:00.00  1.707953E-02  1.009146E+02  1.009317E+02  4.728986E+11
                     (004,027,19)  3.526031E-01  1.868201E-01  1.182530E+01  5.002697E+00
       328 0001-01-03 01:12:00.00  1.635206E-02  1.008860E+02  1.009024E+02  4.727603E+11
                     (006,025,18)  2.752542E-01  4.995576E-01  1.422893E+01  7.306116E+00
       329 0001-01-03 01:21:00.00  1.656034E-02  1.008469E+02  1.008635E+02  4.725701E+11
                     (006,024,19)  2.383016E-01  1.426992E+00  2.045112E+01  1.410777E+01
       330 0001-01-03 01:30:00.00  2.925909E-02  1.007976E+02  1.008268E+02  4.723289E+11
                     (006,023,20)  1.342662E-01  5.816475E+00  4.073108E+01  4.860259E+01

 Blowing-up: Saving latest model state into  RESTART file

Code: Select all

    GET_NGFLD   - ammonium southern boundary condition,            0001-01-16 00:00:00.00
                   (Grid= 01, Rec=0000001, Index=1, File: roms_bry.nc)
                   (Tmin=         15.0000 Tmax=        345.0000)      t =         15.0000
                   (Min =  5.00000000E+00 Max =  5.00000000E+00)
    GET_NGFLD   - chlorophyll southern boundary condition,         0001-01-16 00:00:00.00
                   (Grid= 01, Rec=0000001, Index=1, File: roms_bry.nc)
                   (Tmin=         15.0000 Tmax=        345.0000)      t =         15.0000
                   (Min =  2.69459943E-02 Max =  1.90680868E+00)

[cae]

Hi. In my experience, a blow-up like this usually results from a physical circulation problem and not from the ecosystem model. Have you configured this model without the biogeochemistry and ensured that it runs?

Also, with only 327 time-steps until a problem, you may be able to output the model every time-step to see exactly where in the domain and when the problem originates. Sometimes a problem is visible very quickly after the start of the run.

Good luck!

Chris

[chaoyu1985]

Hi. In my experience, a blow-up like this usually results from a physical circulation problem and not from the ecosystem model. Have you configured this model without the biogeochemistry and ensured that it runs?

Also, with only 327 time-steps until a problem, you may be able to output the model every time-step to see exactly where in the domain and when the problem originates. Sometimes a problem is visible very quickly after the start of the run.

Good luck!

Chris

Thank u so much for your reply and suggestion. It’s so helpful and I really appreciated it. I plotted the parameters in restart file (ocean_rst.nc) as your suggestion (i.e. NH4, NO3, Salt, temp, Zeta, Phytoplankton, zooplankton, chlorophyll, large detritus N - LdetritusN, SdetritusN, LdetritusC, SdetritusC , alkalinity，TIC,Hbbl,Hsbl,zeta). They look fine. There are no obvious weird abnormal in these plots. The ranges of the values of the parameter are fine. There is no obvious error in the edges of the study area.
I also checked the history file (ocean_his.nc) and average file(ocean_avg.nc). I found shflux, latent, sensible look weird in the history file (ocean_his.nc) and average file(ocean_avg.nc). There are errors in the edges of the study area(southen boundary). I attached them here. But lward, temp, sward look fine. In order to avoid the parameters are wrong from the start of the model running (I mean if I gave the weird values in the forcing file -roms_frc.nc). I checked the forcing file. They look fine. I also attached them here.
In this situation, which fortran code or fortran module I should check or follow.
In ur opinion, which kind of problem would be. What direction I should track?
Do u think the energy (KINETIC_ENRG POTEN_ENRG TOTAL_ENRG) is too big for the grid (006,023,20) in time step 330? (5.816475E+00 4.073108E+01 4.860259E+01)

Code: Select all

328 0001-01-03 01:12:00.00  1.635206E-02  1.008860E+02  1.009024E+02  4.727603E+11
                     (006,025,18)  2.752542E-01  4.995576E-01  1.422893E+01  7.306116E+00
       329 0001-01-03 01:21:00.00  1.656034E-02  1.008469E+02  1.008635E+02  4.725701E+11
                     (006,024,19)  2.383016E-01  1.426992E+00  2.045112E+01  1.410777E+01
       330 0001-01-03 01:30:00.00  2.925909E-02  1.007976E+02  1.008268E+02  4.723289E+11
                     (006,023,20)  1.342662E-01  5.816475E+00  4.073108E+01  4.860259E+01

[chaoyu1985]

Hi. In my experience, a blow-up like this usually results from a physical circulation problem and not from the ecosystem model.

the parameters in frc files look normal. i gave them form climate data (monthly averaged)

[ymamoutos]

Greetings

At the last record (328) - when your model is blowing up -
you have a maximum speed at point (i,j,k)(006,023,20)
of 48.6 m/s... It's clear that the problem is on your
physical model as already mentioned. Check your restart
file to see where this point is and what is going on.
Try to reduce your timestep (dt) and then rerun the model.

Giannis

[kate]

Also, you could make it blow up twice as fast by removing the northern half of your domain. It's all land, but the model computes over land points, so it is taking time.

(006,023,20) 1.342662E-01 5.816475E+00 4.073108E+01 4.860259E+01

These are CFL numbers for the x, y, z directions. The y-direction one is definitely unstable.