Unexpected Increase in SNOWC After MODIS-Based Snow Initialization in WRF/Noah-MP

S

seti

Member
Dear all,

I would like to ask for your advice regarding the snow initialization in my WRF/Noah-MP simulation.

I am using ERA5 as the meteorological forcing for my simulation. However, I found that ERA5 considerably overestimates the snow cover over my study area. For this reason, I used MODIS snow-cover observations to correct the initial snow conditions in the wrfinput files.

Following an elevation-based approach, I defined a snowline/baseline from MODIS and applied a gradual transition with elevation. Based on this, I modified the initial snow-cover field together with the related snow variables, including snow depth and snow water equivalent, so that the initial snow distribution became much closer to the MODIS observations.

At the initial model time, the corrected snow field looks reasonable and agrees quite well with MODIS. However, after the simulation starts, I see a very rapid increase in SNOWC. Already after about one hour, many grid cells reach values close to 1, although the changes in snow depth, SWE and the other snow-related variables are comparatively much smaller.

I understand that snow cover should naturally evolve during the simulation, but this increase in SNOWC seems too large and too rapid compared with the changes in the actual snow state.

Could you please advise me on how I should handle this? In particular, how can I initialize the snow variables so that the MODIS-based snow distribution remains physically consistent with Noah-MP after the model starts? Would it be appropriate to investigate the Noah-MP snow-cover formulation or parameters such as SCFFAC and MFSNO, or would you recommend a different approach?

My main concern is to avoid this unrealistic increase in SNOWC while still using the MODIS observations to obtain a realistic initial snow distribution.

Kind regards,
 

Attachments

  • Default.png
    Default.png
    823.4 KB · Views: 4
Hi Seti,
We are aware that ERA5 overestimates snow, especially snowh. In your case, did you only modify snowc, or you modified all snow-related variables (e.g., snowh, snowc, SWE, etc)? Can you check snowh in your initial condition?
 
Hi Seti,

Check if that SNOWC spike comes from a snow change or just Noah-MP recalculating snow cover from snow depth and density right after the first few physics timesteps.

In Noah-MP, the ground snow-cover fraction is diagnosed from snow depth and snow density, so even if your initial MODIS data looks fine, small shifts in depth, water equivalent, or density can drastically change your SNOWC output once the land-surface model runs.

I would therefore check the evolution of SNOWH, SNEQV/SWE, snow density, and SNOWC at every timestep during the first hour, rather than looking at SNOWC alone. In particular, it would be interesting to compare the values right before and after the first Noah-MP call.

I would also be a little cautious about changes in SCFFAC or MFSNO just to force SNOWC to match MODIS. These parameters affect how the model calculates snow cover and adjusting them might mask an underlying problem with your initial snow state. In the FSNO formulation, MFSNO is defined as the density-related melting factor, while SCFFAC is the snow cover factor.

Since you have already modified all the snow variables, I think the most useful next check should be to verify that the initial SWE, depth, and density align with that MODIS fraction. If they don't, Noah MP will calculate a different SNOWC once it recomputes the fraction.

It may also help to post a small table/plot of SNOWC, SNOWH, SWE, and snow density for the initial time, +10 min, +30 min, and +1 h. That would make it much easier to tell if it's a quick snow-state adjustment or a diagnostic response.
 
You must log in or register to reply here.
Back
Top