During the afternoon Thursday, a torrent washed over the heart of the D.C. area, unloading up to a half foot of rain in just a few hours. The deluge caused creeks and streams to rise up to eight feet in a single hour and overwhelmed drainage systems, bringing widespread flooding that inundated roads and stranded dozens of motorists.

The very narrow zone of heavy rainfall resulted from an extremely moist air mass, converging air and a weak disturbance over the area. These ingredients indicated the potential for heavy rainfall somewhere in the region, but predictive models did not show it occurring where and when it did.

The National Weather Service received more than 40 reports of flooding, concentrated in the zone from Alexandria through southern Montgomery County and extending east through the District and the northern and western parts of Prince George’s County. In this zone, two to six inches of rain were widespread.

The heaviest rainfall totals, exceeding four inches, focused in the area from Northeast D.C. through College Park to around Adelphi in Prince George’s County. Hyattsville received 6.35 inches, the highest amount recorded.

Reagan National Airport received 2.88 inches from this event, a record for the date and the heaviest single-day rainfall in 2020. Capital Weather Gang’s Ian Livingston noted it was the fourth time this year that at least two inches of rain fell, tied for the sixth most on record.

Dulles International Airport and Baltimore-Washington International Marshall Airport only received 0.11 and 0.47 inches from the event, showing the localized nature of the heaviest rainfall, which fell in a narrow corridor.

Extremely heavy rain triggered widespread flooding in the immediate Washington region on Sept. 10. (The Washington Post)

How it happened and the forecasting challenge

The National Weather Service had placed a long section of the Interstate 95 corridor from southeastern Virginia to Connecticut under a slight risk for flash flooding through the day and evening. Our area was also under a flash flood watch.

Our experience over the years is that somewhere within these broad threat zones, one relatively small area typically experiences the right combination of factors for a flash flood. The problem is a priori identifying that one small zone, which has proved notoriously difficult to do, even with the most sophisticated, high-resolution models.

Thursday was no exception to this rule. Much of the lengthy I-95 corridor had a similar setup for potential flash flooding: deep tropical moisture, a convergent axis of wind near the surface, enough instability to generate convective storms, and a weak upper-level disturbance. Nailing the small spot of extremely heavy rain — which found its way over northern Prince George’s County — is not within the reach of our best prediction capabilities.

Two of our more reliable short-term models showed no precipitation at all over the immediate area during the period of heaviest rainfall Thursday afternoon, complicating the short-term forecast midmorning, when weather radar showed only spotty, light showers over the region.

Let’s look at some of the elements that came together for the deluge in more detail. First, the tropical moisture plume.

As the image below shows, the region was broadly situated in a very deep corridor of high humidity, sourced in the deep tropics. This plume (outlined by yellow dashed lines) pinwheeled around the western limb of the Bermuda High (red “H”) and ahead of a cold front slowly approaching from the Ohio Valley.

Not shown is a weak tropical disturbance hovering along the Carolina coast for the past couple of days. That disturbance has introduced its own slug of juicy moisture into our realm, but the concentrated blob of deep convection along the North Carolina coast in the image above is a tipoff to its influence here.

The next figure shows a graphic dispatched by the National Weather Service around 12:40 p.m. Thursday. Note the purple scallop region, which shows the large area that would be targeted in the afternoon. This area coincides with a strong trough of surface low pressure — and convergent, low-level winds. Convergence of all that moist air leads to rising air, and in the presence of an unstable air mass, pockets of convection.

The problem with convective cells is that they are very localized. Intense cells can produce rain at rates of one-to-three inches per hour. The gradient in rain amounts can be an all-or-nothing prospect. But no forecasting model routinely used can accurately predict the location or timing of specific cells.

What happened is shown in the radar animation below. This reveals a very narrow corridor of cells repeatedly moving over the same location — called cell training — operating for several hours. The precise location of this corridor could not be accurately identified in advance. Why did the “rain train” set up with laserlike focus over northern and western Prince George’s County, the eastern part of the District and southern Montgomery County, and not farther south or farther north? We do not possess that understanding.

The training happened for one reason: The low-level “steering winds” for these convective cells were oriented along the axis of the convergent zone of low pressure. And when the deep atmospheric mass is sopping wet with tropical humidity, you are going to get torrents from each of those storm cells. The amount of available humidity for rain generation Thursday morning exceeded any value measured by routine weather balloon releases at Dulles Airport for the date and time.

We have been writing these explainers in the wake of flash floods time and time again across the region: Ellicott City in 2016 and 2018, Frederick, Md., in 2018, and in the District and Arlington in 2019, to name a few. Every time, the rain focuses in a tiny flood zone unfolding within a very large outlook area. Every meteorologist worth his or her salt wishes they could pinpoint these zones a few hours or more in advance; we all still have a lot to learn.

Several visuals of the flooding