md.timestepping.time_step and PDD

aisundong

My input is monthly precipitation data of climate state, and I use PDD to calculate surface mass balance.
At the same time, in order to accelerate the calculation speed, I changed the time steps to 10 years, 20 years, 50 years, and 100 years, and the results were very different. Is this because the PDD method requires that calculations can only be done once a year?

aisundong

440m-x06-ommassbalance.txt

2kB

smbpdd.m

15kB

runme-45.m

2kB

Perhaps your professionals can better identify the problem from it? Or can you give me some screening directions?

mathieumorlighem

I could be wrong but to me it looks like you are running into instability issues because your time steps are too large. Do you have stress balance and mass transport activated? If yes... you probably need much much smaller time steps to satisfy the CFL criterion.

aisundong

mathieumorlighem
I conducted a series of tests, and the problem seems to come from the choice of the dynamic equation. After changing the equation from SIA to HO, normal calculations can be performed.The results obtained are as follows.:

After using SIA, strange error results will be obtained, as shown below:

Is this because I made some incompatible settings for transient simulation? Should I first turn off some processes, such as groundingline?

mathieumorlighem

SIA and HO are 2 very different models. I don't know if this is speed or ice thickness. In any case, make sure that your time stepping is consistent with the CFL condition. You probably need time steps of 0.05 year or maybe even less.

joshcuzz

Like Mathieu recommended, a smaller time step is likely needed. You can try and see what time step your model converges to using the adaptive time stepping by setting a minimum and maximum bound. For example:

md.timestepping = timesteppingadaptive(md.timestepping);
md.timestepping.time_step_min = 0.01;
md.timestepping.time_step_max = 2;