Forecasting Challenges of Deep Convection

If you want to be meteorologically humbled, try forecasting deep convection in the Intermountain West.  I thought yesterday we would see some scattered deep convection in the Salt Lake area.  Although it came close, most of the action remained to our south and east and we were left with a few sprinkles from "cumulus pancakus."

Source: NCAR/RAL

As discussed by Doswell (1987) and Johns and Doswell (1992), there are three necessary ingredients for deep convection:

1. a moist layer of sufficient depth in the low-mid troposphere
2. steep enough lapse rates to allow for "positive area" (i.e., CAPE)
3. sufficient lifting for a parcel from the moist layer to reach it's level of free convection (LFC)

In short, this comes down to instability, moisture, and lift.  Instability and moisture are commonly called "thermodynamic factors" and are typically assessed using observed and model soundings, whereas lift can be assessed using model vertical velocities, quasi-geostrophic thinking (e.g., the omega equation), and conceptual models of processes such as thermally driven flows, etc. 

In the midwest, there is often plenty of thermodynamics, and the issue is whether or not there will be sufficient lift (i.e., dynamics) for parcels to reach their level of free convection (often referred to as "breaking the lid").  For example, in yesterday afternoon's sounding from Davenport, IA, there was a whopping 5048 J/kg of surface-based convective available potential energy (CAPE).  

Source: NOAA/NWS/SPC  

That's plenty of thermodynamics!  The problem, however, was that they were sitting right underneath the upper-level ridge.  There was no large-scale lift, and surface-based parcels could not break through the shallow stable layer at the top of the boundary layer and get to their level of free convection.  In other words, the convection was capped, and there was nothing going on in the area around KDVN.

Source: NCAR/RAL

In contrast, the KSLC sounding features limited CAPE, but also no capping layer.  Instead, there's an inverted-V with a deep convective boundary layer, which is surmounted in the middle troposphere by a near-saturated layer.  Some weak stable layers are evident above 500 mb.   Instead of being capped, as was the case at KDVN, parcels were able to reach their level of free convection, but the combination of a marginally stable midlevels and (presumably) entrainment led to the "cumulus pancakus" that we saw yesterday.  

Source: NOAA/NWS/SPC

Assessing moisture is a major challenge in our part of the world, and, unfortunately, there is sometimes great uncertainty in its observation.  Take yesterday afternoon for an example.  The precipitable water in the KSLC sounding above is 1.01 inches (26 mm).  However, precipitable water at KSLC is also measured quasi-continuously by ground-based GPS.  That system shows that the PW at the time of the sounding was 35 mm!  

Source: NOAA/ESRL

That is a remarkable difference.  In fact, if you look at the time series above, it is the largest error over the past 5 days, although there is clearly a systematic bias for the KSLC sounding to measure a lower precipitable water than the GPS system. 

What to believe?  It's hard to assess the moisture part of the deep-convection menu when dealing with such large uncertainties.