Wednesday, October 26, 2011

How did the thousand year rain in Washington DC happen?

In an earlier post, I documented an incredible three hour rainfall that fell on Ft. Belvoir, VA on 08 September that had a return interval of 1000 years.  Other places during the same time of day also experienced similar magnitudes.  All of these extreme totals were courtesy of the remains of tropical storm "Lee" as it slowly made its way up the east coast.  These rains occurred on the second day of a two day deluge.  But only during the late afternoon on the 8th did these extremely rare three hour rains occur.   Take a look at two rainfall maps (figure 1).  The rainfall 00 to 06 UTC on 08 September featured a widespread north to south swath with high amounts embedded while the rainfall from 21 to 03 UTC 08 to 09 September was confined to smaller but more intense areas.

Figure 1.  Six hourly rainfall amounts from two times showing the different character of the rainfall.  The latter rainfalls on the right showed two areas that exceeded 1000 year return intervals just south of Washington DC.  This image is courtesy of

Why did the 1000 year three hour rains occur on the second day?  Both the 7th and the 8th had the moisture,  forcing and the instability to produce heavy rain rates, at least the kind you'd expect from convection in a tropical-like airmass that was brought north by the remnants of Tropical Storm Lee interacting with a frontal zone.  In fact the 7th had even more precipitable water (figure 2) and and low-level horizontal winds to enhance moisture transport (figure 3).  So it's not surprising to see a much larger area of very heavy rainfalls that eventually led to the record flooding amongst the rivers of the Susquehanna river basin.  Grumm (2011) summarized the synoptic conditions that led to this event as being closer to a Maddox type synoptic flooding rain category.  But the 7th didn't produce quite the extreme three hourly rainfall that occurred on the 8th.  Instead these rains occurred on a day with weakened forcing aloft as the upper-level coupled jet structure weakened as the trough west closed off (figure 4) and the low-level jet weakened. With this in mind, we continue on understanding that synoptic-scale forcing cannot provide the full answer.  We have to drill down to the convective scale to explain this.

Figure 2.  Short Range Ensemble Forecasting (SREF) system analysis of precipitable water  and standardized anomalies for 21 UTC 07 Sept on the left column and then 21 UTC 08 Sept on the right columns.  The take home message here shows up in the bottom row where the shaded contours show a smaller area of 2 standard deviations above normal values on the 8th.  Mean precipitable water is also lower on the 8th.

Figure 3.  An SREF analysis of 850 mb mean winds for 21 UTC 07 September  in the left column and one day later in the right column.  The shading in the top row represents the 850 mb U-wind standardized anomalies.  The 850 mb V-wind standardized anomalies show up on the bottom row.  The strong southerly low-level jet on 07 September shows up as nearly four standard deviations above normal over the Mid Atlantic states. 

Figure 4.  Upper-level analysis showing the weakening of the upper-level dynamics from 00 UTC 08 September (left column) to 00 UTC 09 September (right column).

In order to have accumulated the three hourly rainfall with a thousand year return period in the DC area, a convective core dumping  around 2.5"/hr of rain must remain over a specific area for three hours.  This equates to roughly 45-50 dBZ of echo averaged over that three hours.  Either there was an incredible rainfall rate in a shorter period of time, say 6"/hr or there was persistent heavy rain over three hours.  It turns out the latter is true as Ft. Belvoir measured approximately 2.5"/hr for two hours and then almost 2" the final hour (23 - 00 UTC).   The DC area sees that kind of rainfall rate quite frequently when a core of a thunderstorm passes overhead.  But to have that kind of rain rate persist for more than 15 to 20 minutes is unusual and extremely rare for three hours.  

How did the thunderstorms dropping that kind of rainfall rate persist over one area for so long?  Either the convective rain area must be unusually large or was moving slowly.  In the video below two backbuilding multicell thunderstorms showed up, one on the western side of the Washington DC metro area south to Ft. Belvoir and another showed up in southern St. Charles county, MD.   The individual thunderstorm cells moved to the northwest while new cells continually form on the southeastern flanks of the two multicells.  The rate of formation on the southeastern flanks almost exactly matches the individual cell motion leading to anchored multicells (video 1).  These two anchored multicells produced outflows but they were so weak that they only served to sharpen the stationary front and reinforce the mechanism to produce or intensify upwind cells over the same places.  

Video 1.  A loop of radar 0.5 deg reflectivity and base velocity from KLWX 2011-09-08 21 - 23 UTC at 10 min intervals.

This anchored multicell behavior was largely absent on 07 September as the video 2 below highlights.  Instead, the individual multicells moved just a bit east of north along with the individual cells.  Some of the multicells appeared contain a dominant supercell such as the one that passed by Upper Marlboro, MD (east of DC).  Others appeared to be more linear with a long axis parallel to the convective layer steering flow.  But none of the multicells appeared to be anchored as well as the following day.

Video 2.  Same as video 1 except for 2011-09-08 01 - 03 UTC.

Despite the slightly lower precipitable water values of 09 September 00 UTC relative to 24 hours before, the sounding remained nearly saturated for the lowest 500 mb because of cooling above the ground (figure 5).  The nearly saturated atmosphere kept the downdraft production in check and thus the cold pool was too weak to clear the convection away from the western and southern DC area.

Figure 5.  The Sterling, VA sounding for 2011-09-08 - 00UTC and 2011-09-09 00 UTC. 
Both days thus had mechanisms to sustain heavy convection over the same areas for a long duration.  On the 7th to early 8th, the stationary front was parallel to the deep convective steering layer flow (0-6 km mean wind) despite the fact that the flow was strong.  Multiple small multicells and line segments formed in Virginia and traveled up the front eventually congealing into a large rain mass that flooded Pennsylvania (Fig. 6).  No one area sustained extremely heavy rain rates for over an hour but the multiple hits by these convective rains added up to huge totals over periods greater than six hours.  The lack of a significant convectively induced cold pool meant that new convection could could easily exist in the wake of previous convection.  It also meant the cold pool didn't propagate very far away from the stationary front.

Figure 6.  An analysis of key features involved with the heavy rains on September 7th through 06 UTC on the 8th September 2011.  

Late on the 8th, the same stationary front was there in the same general location to provide low-level forcing for new convection.  However, this time, the steering layer flow was much weaker.  Convection initiated on the front, moved to the northwest over the front while new convection would initiate on the front that was slightly enhanced by the weak cold pool (Fig. 7).   The cold pool sometimes surged to the east following a brief intensification of the convection but never very far (Fig. 8).  The forcing remained in the same area but this time the slow storm motion allowed more persistent extreme rains to remain in the same area for a longer period of time than in the previous day allowing for the three hour 1000 year recurrence interval to be achieved.  However the weaker forcing limited aerial extent of those rains.

Figure 7.  Similar to figure 6 except for the afternoon of 08 September 2011.

It certainly helped that the near saturated profile through a deep layer and fairly weak lapse rates fostered the low echo centroid type of convection implying significant warm rain processes aided by some graupel production in the cold layers. There was lightning observed but only occasionally.  This echo structure was observed in the devastating Ft. Collins convective flash flood on 30 July 1997 (Peterson et al. 1999) and the Madison County, VA flood of 27 June 1995 (Pontrelli et al. 1999).  Both of the referenced events occurred as topography helped anchor the main initiation region of the multicells rather than their cold pools.  In this case, it was the stationary front that served as the primary anchor.  Otherwise all cases had weak cold pools that remained in the vicinity of the convection.

Figure 8.  KLWX reflectivity  (upper left) and velocity (lower right) cross section for 2153 UTC 08 September 2011.  


Grumm, R. H., 2011:  Heavy rainfall associated with frontal interactions with Tropical Storm Lee 5-8 September 2011, [Available online at]

Peterson, W. A., and Coauthors, 1999: Mesoscale and radar observations of the Fort Collins flash flood of 28 July 1997. Bull. Amer. Meteor. Soc.,80,191–216.

Pontrelli, Michael D., George Bryan, J. M. Fritsch, 1999: The Madison County, Virginia, Flash Flood of 27 June 1995. Wea. Forecasting, 14, 384–404.