Burial of vegetation by snow in FATES

[katyarjay]

I've been running some single-point CLM-SP and FATES-SP experiments at Toolik (with only one pft - c3 arctic grass) and have noticed that there are large differences in snow depth between the CLM and FATES runs. Comparing with the NEON observations of net radiation at this site, the CLM-SP results look pretty good. It seems like this issue is likely due to a difference in how grasses are buried by snow in CLM.

Below is the FATES code describing the fraction exposed vegetation:

fates/biogeochem/FatesAllometryMod.F90

Line 3179 in e06e0df

fraction_exposed = 1._r8 - max(0._r8,(min(1._r8, (snow_depth-layer_bot_height)/(layer_top_height-layer_bot_height))))

And the parallel code in CLM, which allows for "snow burial fraction for short vegetation (e.g. grasses, crops) changes with vegetation height using a 20% bending factor, as used in Lombardozzi et al. (2018) GRL 45(18), 9889-9897"

        ! NOTE: The following snow burial code is duplicated in CNVegStructUpdateMod.
         ! Changes in one place should be accompanied by similar changes in the other.

         if (patch%itype(p) > noveg .and. patch%itype(p) <= nbrdlf_dcd_brl_shrub ) then
            ol = min( max(snow_depth(c)-hbot(p), 0._r8), htop(p)-hbot(p))
            fb = 1._r8 - ol / max(1.e-06_r8, htop(p)-hbot(p))
         else
            fb = 1._r8 - (max(min(snow_depth(c),max(0.05,htop(p)*0.8_r8)),0._r8)/(max(0.05,htop(p)*0.8_r8)))
         endif

Here is a time series and climatology of snow depth from a CLM-SP run, which looks as expected:

And corresponding results from the FATES-SP run:

Should FATES include a similar bending factor for grasses as is being used in CLM?

@wwieder @adrifoster @ckoven @jenniferholm

[rosiealice]

Hi @katyarjay

Exciting you are running at Toolik :)

I guess my question looking at these plots is 'why is the maximum snow depth so much lower in fates'? Is that all down to the differences in the snow burial algorithm? It would be helpful to see a couple more output variables- specifically the elai and tlai, and end albedo, and maybe some temperature variable (tveg?)?

Cheers,
Rosie

[rosiealice]

I guess it wouldn't be very hard to add a similar statement to the fates code. You would want to use the 'woody' pft parameter to distinguish whether things should be bending or not...

If anything, 20% seems like a very conservative estimate of how much grass bends under persistent snow. I guess danica's implementation was for her cover crop studies where snow might be more intermittent?

[rosiealice]

...although it might be easier to check the impact of the snow burial by commenting it out of the clm5-sp run instead?

[katyarjay]

Thanks @rosiealice,
Yes, we're wondering if these large differences in snow depth are due to the changes in snow burial in CLM, unless there's something else we're missing. I also forgot to mention that I'm running all of these experiments with the same LAI input streams (for CLM-SP and FATES-SP runs) so ELAI is the same in both cases. Here's a plot of albedo differences between the CLM and FATES runs:

I like your suggestion to try a CLM-SP run without the snow burial fix - I'll give that a try once derecho is back up!

[rosiealice]

Cool. So the differences are mostly in the winter, going with the snow burial theory. Are you sure elai is the same? Tlai will be the same but I have a feeling )and I haven't checked as I am multi tasking) that elai is still prognostic in sp mode?

[rgknox]

I wonder if some of the differences are due to CLM-SP having access to the new biomass heat storage module, which FATES doesn't use (but should!).

[wwieder]

I hadn't thought of that, Ryan, i guess the grass leaves would have a heat capacity, but without woody stems it seems like biomass heat storage would be a small term. Regardless we can turn BHS off too if turning off the snow burial term for grasses in CLM-SP runs doesn't results in more similar snow dynamics in CLM and CLM-FATES simulations. Now if they'd just turn Derecho back on...!

[rosiealice]

Hi @katyarjay did you end up figuring this out any further? Asking as we are running some couped simulations where we have too much snow!