On the morning of June 10, 2014, a very small scale, vicious rain storm in northern Prince George’s County unloaded 4 to 5 inches of rain in two hours resulting in dramatic flash flooding. Numerous motorists were stranded in water and some houses had to be evacuated. Herein we recap the meteorology of the event, and describe why summertime flash floods remain very challenging to predict.
Capital Weather Gang forecasters knew that the day was primed for showers and thunderstorms, given the sultry air mass in place, and a stalled frontal zone would likely focus uplift of unstable air. On most summertime days, convective storms have a strong diurnal rhythm; that is, they develop during the time of peak afternoon heating, when the air mass is most unstable.
June 10 seemed no different, and all the morning high-resolution models portrayed a late-afternoon active with scattered to widespread convection. This was predicated on the heavy overcast thinning out, allowing the surface to warm and destabilize the air mass.
Around 9:30 a.m., we were surprised when a single, intense convective cell erupted over central Prince George’s (PG) County, remained stationary, and expanded in areal coverage. That small complex of cells began disgorging phenomenal amounts of rain, soaking the general region of College Park and Bladensburg. All this, from a storm complex barely topping 25,000 ft (most summertime thunderstorms rocket up to 45,000-50,000 ft) and utterly devoid of lightning.
Radar estimates portray up to 4-5 inches (or even a little more) of rain in just 2 hours. A flash flood warning was rushed out, covering northern PG county and eastern D.C. Figure 1 (below) captures the drama in situ – stranded cars, flooded basements, utility outages, water rescues, literally a state of panic for residents.
Why did this event unfold so early in the day? Why did it impact such a small region? Why can’t meteorologists do a better job anticipating these extreme events?
A back-building storm complex
Figure 2 shows a radar snapshot of the storm complex during peak intensity. The rain storm (circled in the figure) is a multicell convective complex – a persistent aggregate of smaller convective cells, some in the process of newly developing, some maturing, others decaying.
This multicell was a very efficient rain-manufacturing machine. New cells were continuously regenerating in the location shown by the magenta arrow. Winds aloft (green arrow) carried these cells toward the E-NE, where they quickly matured, dumped their rain, then decayed. But then a new cell popped up, where the first one had been. And so on…the storm cycled through this process several times.
You can do the simple “vector arithmetic” in your head: The wind moves cells away in one direction (green arrow), while the storm manufactures cells in the opposite direction at the same rate (blue arrow). The vector arrows (nearly equal and opposite) cancel. The larger storm complex, as a whole, remains stationary, anchored to one spot.
We call this a back-building mode because new cells develop backwards, or in a direction opposite the flow of prevailing wind. Given sufficiently high moisture, this type of storm guarantees flash flooding.
What were some of the larger-scale ingredients that allowed this rain-making machine to become established in the first place?
Subtle flash flood ingredients
Figure 3 shows the surface weather map at 8 a.m., just an hour before the event got underway. A stationary front was situated across the region, with a weak wave of low pressure right over D.C. A thick stratus cloud deck covered the region, and no organized precipitation was in progress anywhere along the front.
The final graphic, Figure 4, portrays one of the key elements for the flash flood. This is a busy graphic, with lots of contours and color shadings. Just focus on the colors. Light green represents 68-72 °F dewpoint air, and the small pocket of orange just east of D.C. is 72-76 °F dewpoint air. This is a sopping wet value, and the moisture extended upward through a deep layer, with up to 1.8” of precipitable water.
The weak area of low pressure provided enough convergence of low-level, moist airflow to trigger convective activity along the eastern side of the front. Super-moist air streamed west toward the front, sucked inward by low pressure, where it was forcibly lifted over PG Co. The frontal boundary itself served as an “anchor point” along which new storm cells were continuously regenerated. And there was just enough instability (although not large by any measure – at 9 a.m. with thick cloud, there was little surface heating) to sustain shallow convection, but not over a widespread region.
Why was the forecast of this event missed?
Understandably, it is very frustrating to the victims of a flash flood, when there is seemingly little or no advance notice of the event. Unfortunately, a large number of U.S. flash flood events unfold exactly in this manner – namely, without any sort of advance warning. The events are underway before they are recognized as being a continuing hazard.
The high resolution models yesterday morning gave no hint whatsoever about the formation of a small back-building cloud system. The models do not have the “grid spacing” (horizontal and vertical resolution) to simulate these relatively small cloud systems, nor to nail down the initiation point over a small portion of a county.
Nor did forecasters have sufficiently dense observations (neither in time nor in space), especially at the surface, to see the beginnings of the moisture convergence process setting up over PG Co. With very limited surface heating, we did not anticipate that an intense convective system would erupt just a few hours after sunrise.
It’s unlikely that forecasts of convective complexes this small and transient will demonstrate much skill increase in the near future. We do much better at predicting widespread, significant convective storm outbreaks (i.e. tornado super-outbreaks); the small-scale events remain very elusive. It’s a vexing problem with multiple facets – we need denser observations, better models, and an improved a priori conceptual understanding of how subtle atmospheric features will likely interact in very nonlinear ways.