CLM-FATES results in odd global forest patterns when hydro is on

[sdeherto]

Hi all,

I have been running CLM with FATES hydro and I encountered some strange issues. Specifically, forest only appear to grow in patchy regions which do not really correspond to where forest is supposed to be (see time slice of ELAI in year 16 of a global run).

When running CLM-FATES with hydro turned off these issues do not appear. Initially I was running into these issues for regional simulations, but for debugging purposes I set up a general set of simple global simulations using CTSM3 (release tag ctsm3.012) and the I2000Clm50Fates compset. For these runs, except for turning on hydro and adjusting the FATES parameter file to the two PFT file developed by Jenny I have not made any changes.

I have been discussing these odd results with Marcos Longo and Jenny Kowalczyk as well. One very odd thing is that Jenny can perform global runs with the exact same parameter file, meteo and CO2 conditions as I use and reproduce realistic patterns, indicating this might be a CLM-specific issue. One hypothesis we have is that it may have to do with soils getting very dry in the initial phases of the simulation causing C starvation everywhere except for the patchy regions where plants continue growing. I explored many variables and this spatial pattern seems to be mostly reproduced by Soil matric potential (SMP) in the early phases of the simulation, again indicating that it may have to do with the soil.

Any ideas and input is very welcome in how to debug/solve this!
Cheers
Steven

[jennykowalcz]

Thanks for documenting this here, Steven! Here's LAI from the ELM-FATES test that Steven mentions:

[sdeherto]

Small update, Ryan suggested to look into the soil pedo-transfer function with the potential cause being a mismatch in parameters getting passed along. However, this seems not to be the issue, several soil boundary conditions (BSW, WATSAT,SUCSAT ...) are the same in my default runs and hydro runs so this does not appear to explain the discrepancy.

On another note, when looking over this with Marcos, we noticed that soil temperatures ( TSOI_10cm) are quite extremely low in the first timesteps, being about 20 degrees lower over areas were the forest dies off (values around 270 K -> freezing conditions even in the tropics). At the same time soil water is much higher over the cold soils (H2OSOI) and evaporation is negative (EFLX_LH_TOT) meaning there is condensation ongoing. These conditions seem to stabilize as the first years pass and reach more normal values for all these variables from year 2 of the simulation onwards but by that point the plants have all died.

So there appears to be something off in my initial soils, I will start up two site scale simulations in nearby grid cells in Central Africa to compare a grid cell where forest grows against one where it doesn't to try to understand this a bit better.

[rosiealice]

Seems like a test of the init cold soil moisture condition would be useful... If the soils freeze, water potentials go way down and might cause all sorts of havoc...