For five days over the Mid-Atlantic, the rains kept coming, at times torrential.   Flood watches and warnings were issued across the Washington–Baltimore region.  To see an analysis of rain totals, and the science behind why so much fell, read on.

Waves of moderate to heavy rainfall quickly added up

Washingtonians were forewarned that Wednesday, October 9 through Sunday, October 13, would be gloomy, humid and wet – thanks to the stalled remnants of Tropical Storm Karen.   The prospect of a good, soaking rain was readily welcomed.   But as the adage states, “Be careful what you wish for!”   Our drought was busted…in spades!

Days in advance, the model forecasts painted a 3-4” region of rainfall just offshore, close to the coastal low’s center.   While heavy rain did indeed fall along coastal North Carolina and Virginia, over Delmarva, Maryland and Pennsylvania, the rainfall bulls eye shifted inland a considerable distance.   And much more fell than the models predicted.

The figure below shows four consecutive days of rain accumulation, from October 8 through 11.   These were the heaviest four days during the storm, and the totals are derived from weather radar estimates.

Daily rain totals for the Mid Atlantic, produced by the stubborn coastal low. (NOAA)

During October 9 and 10, significant amounts of rain (3-5”) fell along the North Carolina coast and Virginia Tidewater.  On October 10 and 11, however, notice the abrupt shift nearly 100 miles inland, to regions north and west – drowning locations such as Harrisburg, PA and the Washington metro region.  The rain finally came to an end, across most of the entire Mid-Atlantic, late on Sunday, October 13.

The rainfall total map, for the past seven days (October 6-12), is shown below:

Cumulative rain totals for the period October 6-12. Color scale same as shown in Figure 1. (NOAA)
Cumulative rain totals for the period October 6-12. Color scale same as shown in Figure 1. (NOAA)

There are some BIG numbers here – 10”+ inches over south central PA and along the Outer Banks.   In the immediate D.C. region, 5”-7” is the rule.   However, the values shown here are radar estimates.   Point estimates, collected by rain gauges, tend to be higher than radar, and are probably closer to “ground truth”.   (At the time of this writing, a detailed set of gauge estimates for this storm was not available from the Weather Service).

Post-mortem of the rain storm:  What exactly happened?

Fundamentally, this was a slow-moving coastal low sitting over the warm Gulf Stream waters.   The storm’s surface winds extracted heat and water vapor from the ocean surface.   Condensation in clouds released heat energy within the vortex, sustaining the low pressure center, and producing intermittent heavy rain.  Additional energy came from horizontal temperature gradients (fronts) straddling the storm.

Six factors contributed to unusually heavy rain production from this weather system.

1. Time of year.   This was an early-season Nor’easter.   Accordingly, Atlantic surface waters were very warm, having absorbed solar energy through the summer months.   The amount of water vapor evaporated from the ocean is a function not only of wind speed, but also water temperature.  The relationship with temperature is non-linear.  If you think of water vapor as the storm’s fuel, then copious amounts were available to feed the rain-making “engine” (much more than during the dead of winter, i.e. a late February Nor’easter).

The figure below shows the extremely high moisture content centered over the Washington region, a veritable bulls eye of precipitable water.   Precipitable water is the depth of rain if all humidity in the air column were to condense. A value of 1.6” is extremely high for this time of year.

Analysis of total precipitable water for the nor’easter. Notice the maximum (1.6”) centered on our region. (NOAA)

Precipitable water is an instantaneous estimate made from weather balloons and satellite measurements.   Since the humidity is drawn from a large region around the storm, local rain totals can greatly exceed any location’s specific precipitable water value.

2. Moisture jetting in off the Atlantic.   A meso-analysis of the storm revealed an impressive low-level “conveyor belt” of humidity streaming into the Nor’easter, on its northern side.   As shown below, the low-level (5,000 ft) easterly winds became focused into a narrow, arrow-straight jet, importing high Atlantic humidity into the storm’s core at speeds of up to 35 mph.   In essence, the jet replenished the moisture supply as fast as it rained out.

Another aspect of Washington’s coastal rain machine: A jet-like conveyor or “moisture feed” directly into the storm, courtesy of fast easterly winds in low levels of the atmosphere. (NOAA)

3. A lone storm, adrift.   Because this coastal low was cut off from the main jet stream flow (the jet stream was streaming across northern Canada), it meandered very, very slowly – and at times, became stationary.   Perhaps more so than any other factor, this aspect contributed most to high rain totals across the Mid Atlantic – there were simply plenty of hours, plenty of days for rain to pile up over the same locations.   Furthermore,  embedded pockets of heavy rain repeatedly swept across the same regions from the east – a process called echo training.

A storm without an agenda. Note the extreme, isolated nature of the coastal low’s vortex aloft, at the 18,000 foot level. (Unisys Weather)

4. A Stationary front focusing heavy rain.   This feature, shown below, contributed to high inland rain totals.   A stationary front, also called a Piedmont Front, developed between the westward push of warm, humid air and slightly cooler air sitting over the Appalachians.   The front set up shop over the Md. and Pa. Piedmont regions, assuming a north-south orientation.

 A stationary frontal boundary, oriented north-south (alternating blue and red line segments), which enhanced moisture convergence and uplift over the Mid Atlantic’s Piedmont.
A stationary frontal boundary, oriented north-south (alternating blue and red line segments), which enhanced moisture convergence and uplift over the Mid Atlantic’s Piedmont.

Note how the axis of heavy inland rain (top figure) coincides with this boundary.  Winds converged at low levels on either side of the front, lifting the air vigorously.  This contributed to bands and pockets of heavy rain along a line from northern Virginia to Harrisburg, PA.

5. Shortwave energy in the middle atmosphere

Weak waves of low pressure rotated counterclockwise around the stalled vortex in the upper levels.   These pockets of energy are called shortwaves  and they gave the coastal a periodic energy boost.  Predicting their timing, and even detecting their presence, was difficult.  But when they swept through, rain production was enhanced for several hours.   One of these shortwaves can be seen as a green-colored pocket two images up, located on the southwest side of the main upper-level vortex.

6. Mesolows

These small-scale gyres, just tens of miles across, were noted from time to time in the radar loops.  While the vortices are and transient shallow features, they enhanced the convergence of low-level moisture over regions for an hour or two – leading to local enhancement of rain rates.

How well did the model predictions do?

Suffice it to say we had good, 3-5 day lead time on a potentially long-duration storm system.    Accordingly, NOAA’s Weather Prediction Center (WPC) issued multi-day rain total predictions in the 3-4” range near the coast and 1-3″ inland, consistent with a slow-tracking Nor’easter moving up the coast (Figure 7, below).   The predictions are compared here with the observed storm total rain.

Side-by-side comparison of predicted vs. observed rain totals, for the duration of the multi-day event. (NOAA)

However, these predictions underestimated the inland penetration of heavy rainfall, and also the magnitude by a factor of 2-3.

In other words, timing and duration were well predicted, but not heavy rainfall amount and location.

Another way to capture poor aspects of the model performance is to compare a timeline of predicted vs. observed totals:

Time series of cumulative rain, predicted by various forecast models (shown in the key, upper right) on Friday (when some of the heaviest rain fall), compared with observations at Washington Reagan (black dots). The European is not shown, and it outperformed the U.S. models. (Courtesy of TerpWeather.)

There are many reasons for the abysmal under-prediction, but it seems likely that the models failed to adequately simulate important mesoscale processes.  These include the low-level moisture jet and the establishment of the Piedmont Front.  These two factors alone shifted the geographic focus of heavy rain, away from the coastline.

As a general rule, the models also do not simulate the heaviest rain cells, which are smaller than the model’s mathematical checkerboard of grids can resolve.  As these cells repeatedly tracked over the same regions, the training effect rapidly piled up rain over localized regions.