Inside look as model creates convection on grid scale

Model profiles of vertical motion, cloud water and cloud ice, and latent heating rates from the grid-scale and convective parameterizations are shown here for the 12 UTC 26 July 2001 operational Eta run at the same stations and times as the soundings on the previous page. It is recommended that you read through this entire page before viewing the loops. The plotting scale is exceedingly large! Under ordinary circumstances you will not see values this large in a 22-km model. At finer resolution, peak values in a grid box do get higher because the model can represent higher-amplitude small-scale features. RAP loop (26/1800-27/0600) PHP loop (27/0000-27/0600) SLN loop (27/2100-28/0600) P#G loop (28/0000-28/0600)

These plots show what the convective parameterization and grid-scale scheme are doing to the soundings! Convective and grid-scale heating profiles are in the BUFR fields so you should be able to plot them in real time when you make BUFR soundings. Likewise for vertical motion, cloud liquid, and cloud ice.

The convective parameterization usually causes heating in the upper troposphere and cooling in the lower troposphere. The convective parameterization over time may also moisten the middle and upper levels sufficiently to cause grid-scale condensation. However, most heating from the grid-scale scheme occurs at low levels where more water is available to condense since saturation mixing ratio drops off exponentially with temperature. The heating profiles not only shape the soundings, they also profoundly impact the dynamics and thus the vertical motion.

To help you interrelate all the fields and develop a picture of what's going on, annotated vertical profiles corresponding to the annotated soundings on the previous page are shown and discussed here.

Remember, the 03 UTC sounding at P#G indicated a growing cumulus cloud in the vicinity of 700 hPa with updraft source air from around 800 hPa. Here we see an 800 hPa vertical motion maximum of 50 microbars/s, or roughly 0.5 m/s! Sitting on top of that vertical motion maximum is a cloud, with peak water content and peak heating from condensation at around 700 hPa. The second cloud water maximum, at around 600 hPa, may have advected into this grid box from an adjacent taller cloud.

The sequence from RAP shows the build-up of the grid-scale cumulus tower discussed with the soundings.

• The low-level updraft peak strengthens and its location rises during the three-hour period from 21 UTC to 00 UTC
  
• The cloud water it supports and associated grid-scale latent heating also increase and ascend
  
  • But even after three hours of growth, the tower has only risen to 500 hPa
    
  • Grid scale precipitation (shown previously) doesn't reach the surface until almost 23 UTC
    
  • The peak water content tends to be located above the peak heating from condensation because condensate is getting advected upward
    
• The convective parameterization is active throughout the period, with upper level heating and lower level cooling increasing in strength
  
  • The roughly constant-with-height region of cloud ice above 500 hPa is probably a result of moistening from the convective parameterization.
    • At 23 UTC, the change at 500 hPa from sharp dropoff of combined cloud water and ice (imagine adding the green curves) to the flat distribution above suggests that the grid-scale ascending tower has only reached 500 hPa while the cloud ice above that is a grid-scale manifestation of the convective moistening. That's how the model is supposed to make an anvil.
      
  • Grid-scale cooling at 300 hPa is resulting from sublimation of this cloud ice or snow produced by the cloud ice
    
  • Convective cooling does not extend below cloud base — the BMJ scheme has no downdrafts to drop into the boundary layer
    
  • The grid-scale scheme is cooling the boundary layer as it evaporates falling rain
    

The grid-scale heating profile from the surface to 400 hPa resembles many published heating profiles for the convective portion of large squall convective systems with trailing stratiform regions. That's because the model is attempting to build a thunderstorm from the ground up, one level at a time, using the resolved vertical motion for its updraft! The heating profile from the convective parameterization more resembles many published heating profiles for the trailing stratiform region. That's because the convective parameterization is attempting to produce the thermodynamic effect that a whole convective system has on its environment!

The SLN profiles (shown in the loop) follow the same pattern until around 01 UTC, then evolve toward a more classic stratiform region signature with deep grid-scale cooling below the melting layer and vigorous grid-scale warming above. This is more like how mesoscale models should simulate an MCS, with the resolved motions simulating the mesoscale ascending and descending currents, sustaining a thick precipitating anvil with cooling below.

Now you can understand why the AVN forecast is so badly affected when it has one of these episodes of grid-scale convection: it's like making the convective region of a squall MCS 80 km wide! Or, if the band of grid-scale convection is two grid points wide, it's 160 km! Nobody yet knows exactly what kicks off such episodes, otherwise the model would be changed to reduce the number of spurious events. In the Eta, the model makes a system that in some ways resembles a real MCS, even if it goes about it in a backwards sort of way. At finer resolution it may be able to be more realistic, but the problem of getting the timing and placement of events correctly remains.