Sometimes you will receive a message from CICE that says "departure points out of bounds" or "negative ice area in horizontal remapping". These are dynamic convergence errors and are akin to a CFL violation. Generally, this requires reducing the sea ice timestep. Here is a quick guide to help with this problem.
1. If this occurs right away on an initial/hybrid run, you can restart and change the coupling interval (ATM_NCPL == ICE_NCPL). You can try doubling this value. In most case set ups the default is 48 or 48 times a day (half-hourly), and for the one-degree resolution this should be more than sufficient so the error may indicate another issue (see below).
2. If you are part way in to a run and the atmospheric model will not let you change the coupling interval, you can subcycle the sea ice dynamics. This means you can increase the number of sea ice dynamics steps for each sea ice thermodynamic step. This reduces the timestep in the sea ice dynamics only. In CICE4 the namelist option is xndt_dyn, and after CICE5 the namelist option is ndtd. The default is 1 dynamic step per thermodynamic step. You can try doubling this which will cause there to be two dynamics steps per thermodynamic step. Again, for the one-degree resolution there is likely another issue going on that's the root cause of the problem.
3. For the high resolution CICE (0.1-degree), we set ATM_NCPL = ICE_NCPL = 144 or higher (coupling every 10 minutes). In these cases we set (ndtd (xndt_dyn) to 2 or 3. Increasing the coupling significantly for this resolution can really impact performance of the model. So, we balance the thermodynamic and dynamic timestep in this case.
4. For the low resolution (1-degree), there is likely some other issue as the root cause of the model crash. For example, if you are running a paleo configuration, a 4xCO2 that has no ice, or some other simple model configuration, then sometimes starting the model with no sea ice at all can lead to very small sea ice areas that accelerate very quickly. The above changes might help to alleviate this. Otherwise, some groups turn off the dynamics (kdyn = 0) in this configuration or run with stub sea ice (SICE).
5. The other thing that can happen is that the surface boundary layer and the the atmospheric PBL can start to "decouple". In this case, the winds grow to excessive speeds, particularly around steep topography like Greenland or Antarctica. In this case, changing the coupling interval should help. However, other adjustments in the atmospheric drag / PBL may be necessary.
Note that one can change the coupling interval or the dynamic subcycling back to the original values once your run has gone past the problem. However, often the crash will occur again several years later and I usually just leave it with the modified values.
1. If this occurs right away on an initial/hybrid run, you can restart and change the coupling interval (ATM_NCPL == ICE_NCPL). You can try doubling this value. In most case set ups the default is 48 or 48 times a day (half-hourly), and for the one-degree resolution this should be more than sufficient so the error may indicate another issue (see below).
2. If you are part way in to a run and the atmospheric model will not let you change the coupling interval, you can subcycle the sea ice dynamics. This means you can increase the number of sea ice dynamics steps for each sea ice thermodynamic step. This reduces the timestep in the sea ice dynamics only. In CICE4 the namelist option is xndt_dyn, and after CICE5 the namelist option is ndtd. The default is 1 dynamic step per thermodynamic step. You can try doubling this which will cause there to be two dynamics steps per thermodynamic step. Again, for the one-degree resolution there is likely another issue going on that's the root cause of the problem.
3. For the high resolution CICE (0.1-degree), we set ATM_NCPL = ICE_NCPL = 144 or higher (coupling every 10 minutes). In these cases we set (ndtd (xndt_dyn) to 2 or 3. Increasing the coupling significantly for this resolution can really impact performance of the model. So, we balance the thermodynamic and dynamic timestep in this case.
4. For the low resolution (1-degree), there is likely some other issue as the root cause of the model crash. For example, if you are running a paleo configuration, a 4xCO2 that has no ice, or some other simple model configuration, then sometimes starting the model with no sea ice at all can lead to very small sea ice areas that accelerate very quickly. The above changes might help to alleviate this. Otherwise, some groups turn off the dynamics (kdyn = 0) in this configuration or run with stub sea ice (SICE).
5. The other thing that can happen is that the surface boundary layer and the the atmospheric PBL can start to "decouple". In this case, the winds grow to excessive speeds, particularly around steep topography like Greenland or Antarctica. In this case, changing the coupling interval should help. However, other adjustments in the atmospheric drag / PBL may be necessary.
Note that one can change the coupling interval or the dynamic subcycling back to the original values once your run has gone past the problem. However, often the crash will occur again several years later and I usually just leave it with the modified values.
Last edited: