Event Summary
     National Weather Service, Raleigh NC

July 30, 2007 Flash Flood Event in Northwestern Guilford County
Updated 2007/09/06

Event Headlines

...Heavy rain during the early morning hours of July 30, 2007 produced flash flooding in the Summerfield and Stokesdale areas of Guilford County...
...Radar estimates from the Raleigh, NC and Blacksburg, VA radars suggested that 10 to 12 inches of rain fell...
...At least 7.12 inches of rain was reported by one spotter in Stokesdale before their rain gage appeared to have malfunctioned and another spotter had at least 8 inches of rain in Stokesdale before the gage filled up...
...Several homes were evacuated and several cars were damaged by flood waters...


During the early morning hours of July 30, 2007, upper air data and RUC model analysis showed a deep upper level ridge located over the Midwest and an upper level trough located over the Northeast and Mid-Atlantic states. Water vapor imagery from 06Z 07/30 and RUC 500 MB height and vorticity fields showed several shortwaves and associated vortmaxes near central NC, one in central VA and another in eastern TN/KY. The region was in the nose of 75kt jet at 200 MB with upper level divergence increasing with time.

At the surface, a 06Z 07/30 analysis of observations showed a weak area of low pressure located over eastern TN with a trough axis extending north into eastern KY and WV and south into northern GA. A weak area of high pressure was located over southwest VA. This meso-high was the result of rain cooled air from showers and thunderstorms that had occurred several hours earlier during the evening of 07/29. Additionally, a weak inverted surface trough was located over central SC and NC. Dew points across central NC were in the lower 70s.

The 00Z/30th KGSO sounding depicted a low shear/high precipitable water environment. The air mass was slightly unstable with a lifted index of -1 to -2, with a fairly high K index of 36. The upper air analysis depicted a moist air mass through 700 MB then fairly dry aloft. The freezing level was abnormally high, around 15,000 ft.

Thunderstorms developed just after midnight on July, 30 2007 over the Northwest Piedmont of North Carolina. With very little flow in the atmosphere, the storms that formed were nearly stationary. One particular storm formed to the northwest of Greensboro, over northern Guilford and Rockingham Counties. At 3:42 AM EDT, the National Weather Service in Raleigh issued a Flash Flood Warning for Guilford County, because of the significant amount of rain that had already fallen, and the expectation that the thunderstorms will be very slow moving. Radar estimates at that time were that 5 inches of rain had fallen in the past 3 hours. At 4:30 AM EDT, Guilford County Emergency Management reported the following...
  • Three homes on Weitzel Drive in Summerfield had been evacuated due to rising flood waters.
  • Automobiles located on the property had been severely damaged.
  • Water levels were reported to have been as high as 4 feet.
  • Two feet of water covered Ellison Rd. near Highway 158.
  • U.S. Highway 220 was closed between Myers Fork Rd. and Rhondan Rd. in the Summerfield area.
The location that experienced the heaviest rain can best be described as a rather rural area with significant forestation and some rolling hills, typical of the northwestern Piedmont. Despite the fact that the location is rather remote, two rainfall reports of over 7 inches were received.

The image below is from the NWS Experimental Precipitation Analysis Product. This precipitation analysis or estimate is based off of radar estimates and rain gage data and it is reasonably accurate. Note the very localized nature of the heaviest rain with two pixels of estimated rainfall amounts in excess of 8 inches.

NWS experimental precipitation analysis product across Guilford County - Click to enlarge

Quality of the Rainfall Estimation

At least 7.12 inches of rain was reported by one spotter in Stokesdale before their rain gage appeared to have malfunctioned (their backup rain gauge holds up to 6.5 inches and it overflowed.) Another spotter had at least 8 inches of rain in Stokesdale before their electronic gage stopped reporting.

The storm total precipitation products (STP) from KRAX and KFCX were in very good agreement as to the amount (10-12 inches) and location of the heaviest rain. This is not really that surprising, since the storms location was very nearly equidistant from the radar sites, with KRAXs 0.5 degree scan interrogating the storm at around 7300 feet, and KFCXs 0.5 degree scan sampling the storm at around 6600 feet. There was some ice present in the storm as there was lightning and radar cross sections show that reflectivities about 40 dBZ exceed the 0 deg C isotherm. Ground-truth reports of rainfall exceeding 7 inches suggest that extreme rainfall did occur.

The image below is a radar rainfall estimate from the KRAX radar located near Raleigh, NC. Note that pixels with a red color are rainfall estimates of 5 inches or more and the pixels with purple color are rainfall estimates of 10 inches or more. A similar image for the KFCX radar located near Blacksburg, VA is shown just below the Raleigh, NC image. Note how similar the rainfall estimates are. Click on the image to enlarge and show a full legend.

KRAX radar precipitation estimate - Click to enlarge

The image below is a radar rainfall estimate from the KFCX radar located near Blacksburg, VA. Note that pixels with a red color are rainfall estimates of 5 inches or more and the pixels with purple color are rainfall estimates of 10 inches or more. A similar image for the KRAX radar located near Raleigh, NC is shown above. Note how similar the rainfall estimates are. Click on the image to enlarge and show a full legend.

KFCX radar precipitation estimate - Click to enlarge

Flood Survey Findings and Discussion

Ten inches of rain is a lot of rain, and it certainly alarmed the Service Hydrologist when he saw the data early in the morning of July 30th. There had already been flash flood verification from reports of homes being evacuated in Summerfield and a vehicle rescue on a bridge over Troublesome Creek in southern Rockingham County. The Service Hydrologist did not wait for additional information from Emergency Management, but departed for a flood survey of the area at 9:00 am that morning and arrived at the site of the home evacuations around 10:30 am.

The home evacuations (there were 2, perhaps 3 evacuations) were located in a fairly new, rather upscale development which had been built with very little disturbance to either topography or vegetation. The homes were at the end of Weitzell Drive which runs parallel to and about a quarter mile from Route 220. There was a fairly steep and heavily vegetated slope from Route 220 down to a small creek which bordered the yards of the flooded homes. This creek was well within its banks by the time it was visited, and it was estimated that it had risen 8 to 10 feet judging from high water marks on the homes. This point was directly under the rainfall maximum. There was no sign of debris on Weitzell Drive.

After traveling north on Route 220 and stopped on the bridge over the Haw River, which was high and fast, but there was no evidence of debris which would indicate that the bridge had flooded. There were no other reports of significant flooding in the Summerfield area, so the Service Hydrologist proceeded northwest to Stokesdale, where there was a report of residential flooding on Agatha Drive. I found the road quickly and determined that only one home at the end of the road had flooded, and it had been minor, with water confined to the homes crawl space.

The Service Hydrologist was initially puzzled by the minimal response downstream at the river gage (HAWN7) at Haw River (shown below). There are several factors that contributed to this minimal rise:
  1. The small areal distribution of the rainfall maximum. There was only a pixel or two of the ten-inch maxima, and 2 or 3 small basins that had a five inch basin- average rainfall.
  2. The runoff was distributed among 3 (you could argue 4) basins.
  3. There were 3 lakes on those basins which were able to absorb the runoff.
  4. Soil moisture was extremely low, so infiltration would be enhanced.
On the hydrograph below, note the difference in response from the heavy rain shower on July 29th where the rain fell largely south of the lakes (link to radar estimate), and the very heavy rain from the event on the 30th (link to radar estimate).

HAWN7 Hydrograph - Click to enlarge

Factors Contributing to Flash Flooding

Based on research conducted by Maddox (1979), there are several features and conditions that are common to many flash flood events. Many of these features were present across the Northwest Piedmont during the event...
  1. Flash floods are associated with convective storms.
  2. Storms occur in regions with high surface dew point temperatures.
  3. Relatively high moisture contents are present through a deep tropospheric layer.
  4. Weak to moderate vertical shear of the horizontal wind is present through the cloud depth.
  5. Convective storms and/or cells repeatedly form and move over the same area.
  6. A weak, mid-tropospheric, mesoscale trough helps to trigger and focus the storms.
  7. The storm area is very near the mid-tropospheric, large-scale ridge position.
  8. Storms often occur during night-time hours.

The Flash Flood Warning

The Flash Flood Warning for Northwest Guilford County was issued at 342 AM EDT with a 48 minute lead time and an accurate targeting of the Summerfield area, where evacuations were required at around 430 AM.

Forecasters at the NWS had a great deal of difficulty getting reports from contacts in Guilford County due to the early morning timing of the event. In fact, the NWS did not get word of the evacuations until around 530 AM. In addition, northwestern Guilford County is a rural area, so flooding was not nearly as widespread or severe (fortunately) as one might expect from such a heavy event. If the rain had been only 10 miles further south, Greensboro would have had a severe urban flash flood event. A
Flash Flood Statement issued about 40 minutes after the Flash Flood Warning added additional detail about roads and intersections that might flood.

Radar Analysis

A review of the 0.5 degree regional radar mosaic indicates that an outflow boundary originated from a large area of thunderstorm activity over south central Virginia between 03-04Z. While hard to discern in composite reflective due to a high amount of ground clutter and anomalous propagation (AP), the boundary is observed traveling toward the southwest during this time with isolated to scattered shower activity developing along this boundary as it moved toward the southwest. As the outflow boundary entered into Rockingham and Guilford Counties, it interacted with an area of scattered showers, likely initiated by the encroaching upper level jet dynamics. This shower activity quickly organized itself into deep convection with the passage of the outflow, first occurring over Rockingham County around 0530Z, developing downstream in NW Guilford County shortly thereafter. Little movement was exhibited by any of the convection; therefore these storms were very heavy rain producers. By 0630Z, approximately an hour after convective enhancement began, Doppler radar estimates from both KRAX and KFCX showed around 2 inches in NW Guilford County. Just after 07Z, the large thunderstorm that had been observed in Rockingham County collapsed. It appears the cold pool generated by the storm collapse helped to further enhance the convection over extreme NW Guilford County. With little to no movement noted with this activity, Doppler radar estimates quickly climbed to 10-12 inches by 0830Z, just under 3 hours from the beginning of the rainfall accumulation across the area.

The excessive rainfall and the resultant flash flooding appeared to be caused by a combination of factors. Some of the primary factors include...

1. A strong upper level jet (for this time of the year) crossed the area with the Northwest Piedmont being in the favored area for upper level divergence (left exit region of the jet). A mid level cold pool was also noted along with the strong upper level jet. At 00Z, the Greensboro, NC (GSO) upper air sounding showed a mid level warm nose (cap) around 600mb. A look at the upstream sounding from Roanoke, VA at 00Z does not show such a cap. By 12Z, the cap at the GSO had been eroded as the mid level cold pool moved through the area. This further enhanced the convective potential.

2. An outflow boundary from earlier convection in south central Virginia, which was seen on the looping 0.5 degree mosaic reflectivity, raced southwestward and interacted with the aforementioned upper level jet and mid level cold pool.

3. Weak steering flow in the 850 to 500 MB layer of the atmosphere was present from 00Z sounding data and model sounding data (averaging 5 kts). The weak steering flow allowed the thunderstorms in the area to pulse up and down and continue to fuel each other through a series of outflow boundaries each storm produced as they collapsed.

4. Precipitable water (PW) values based from the GSO soundings showed a moist atmosphere. The PW at 00Z was 1.57 (117% of normal) while the 12Z PWAT value climbed to 1.72 (129%) of normal. This suggested that any convective activity would be capable of producing heavy rainfall.

5. 0-6 km Bulk Shear was 20 kts. This indicates that there was not sufficient wind shear to sustain organized convection. Instead, the atmosphere was more favorable for pulse thunderstorm development, which greatly enhances the potential for storm collapses, resultant outflow boundaries, and cold pool interactions.

Sometimes isolated summertime flash flood events can be caused by low echo centroid storms. These storms exhibit most if not at all their reflectivity below the freezing level. In such a case, warm rain processes dominate and there is little to no lightning/hail production of the storm. These types of storms can appear very innocuous on infrared satellite. The storms that occurred in the early morning of July 30th do not appear to follow the description of a low echo centroid storm. Cross sections of reflectivity data show that the updrafts of the storms reaching well above the freezing layer with echo tops exceeding 40,000 feet. Over 50 cloud to ground lightning strikes were produced by these storms.

In summary, the reason why only this small area was affected was essentially timing and placement. The favored left exit region of the strong upper level jet was moving into the affected area at the same time as the outflow boundary was moving into the area. These two features coming together, coupled with a weak steering flow layer, a moist atmosphere, and further cold pool interactions, resulted in flash flooding for Northwest Guilford County.

Flash Flood Monitoring and Prediction Program - FFMP

The Flash Flood Monitoring and Prediction (FFMP) program differentiates basin rainfall amounts up to 5 inches using a color scale...with amounts exceeding 5 inches depicted as white. Lower amounts are to be expected since FFMPs depiction is basin-averaged rainfall rather than individual 4 kilometer bins. On the graphic below, keep in mind that FFMP cuts off basins which are outside the CWFA, so the area of rain exceeding 5 inches actually extends north a basin into the headwaters of Troublesome Creek in Rockingham County.

FFMP basin rainfall - Click to enlarge

In addition to allowing the user to quickly determine which basins are immediately affected, the program also allows the user to "trace" the rainfall into downstream basins. Routing the water downstream is, of course, crucial for determining the extent of warnings, advisories, or statements. The rainfall maximum from the case were considering affected 3 basins, which we consider below.

  1. Recall, the rainfall maximum extended north of the Summerfield area into the headwaters of Troublesome Creek in southern Rockingham County. Selecting a point in extreme northwest Guilford County, we see that the runoff will flow east northeast into a lake (Washburn) which will either capture (or thus end) the flood wave, or at least dampen and slow the wave resulting from that portion of the rainfall maximum. Water from the lake would continue downstream into the Haw River.

    FFMP basin trace - Click to enlarge

  2. Next we select a point around the actual maximum rainfall in Summerfield (just south of our previous point). Note that the affected basins are the Haw River and Meads Fork Creek. We see that the runoff from Meads Fork will merge with the runoff in the Haw River at the county line north of Greensboro and then continue slightly north of east into Rockingham County.

    FFMP basin trace - Click to enlarge

  3. The water then loops southeast into northwest Alamance County and proceeds southeast to Burlington and our river gage at Haw River.

    FFMP basin trace - Click to enlarge

  4. The southernmost portion of the rainfall maximum fell over Reedy Fork basin. Notice that this trace leads straight east into Lake Brandt and then immediately into Townsend Lake, which are water supply reservoirs for the city of Greensboro. This runoff is going to be absorbed by these reservoirs, which were below normal due to the insipient drought.

    FFMP basin trace - Click to enlarge

Radar Loop

Java Loop of KRAX radar reflectivity imagery at 15 minute intervals from 0432Z (1232 AM EDT) through 0958Z (558 AM EDT) Monday, July 30, 2006. Note - this loop includes 23 frames.

Java Loop of KRAX radar reflectivity imagery at approximately 5 minute intervals from 0530Z (130 AM EDT) through 0928Z (528 AM EDT) Monday, July 30, 2006. Note - this loop includes 51 frames.

Archived Text Data from the Flash Flood Event

Select the desired product along with the date and click "Get Archive Data."
Date and time should be selected based on issuance time in GMT (Greenwich Mean Time which equals EST time + 5 hours).

Product ID information for the most frequently used products...

RDUAFDRAH - Area Forecast Discussion
RDUZFPRAH - Zone Forecast Products
RDUAFMRAH - Area Forecast Matrices
RDUPFMRAH - Point Forecast Matrices
RDUHWORAH - Hazardous Weather Outlook
RDUNOWRAH - Short Term Forecast
RDULSRRAH - Local Storm Reports (reports of flooding or severe weather)
RDUFFWRAH - Flash Flood Warning
RDUFFSRAH - Flash Flood Statement
RDUSPSRAH - Special Weather Statement
RDUSPENES - Satellite Precipitation Estimates


Selected Photographs after the Flooding
Photos courtesy of Michael Moneypenny.
(Click the image to enlarge.)

Picture of the Haw River at Route 220 looking downstream - Click to enlarge           Picture of erosion on driveway on Weitzell Dr - Click to enlarge           Picture of Route 220 at the Haw River which flooded during the night - Click to enlarge

Picture of the nnamed creek that flooded the day before - Click to enlarge           Photo of shed that was lifted off its foundation and floated a few feet - Click to enlarge           Floodwaters reached the bottom of the siding of this home on Weitzell Drive - Click to enlarge

Final Thoughts

  • Comparing precipitation estimates from the KRAX radar with those being observed by the KRNK radar provided added confidence about the data and increased concern about the need to issue the Flash Flood Warning.

  • The basin trace function on FFMP was a big asset to have in this event. It correctly highlighted "Troublesome Creek" as a basin that would bear the brunt of the heavy rain. A warning from the NWS office in Blacksburg, VA indicated that a water rescue had occurred on this creek downstream from the heavy rain even though the rainfall that fell in vicinity of the rescue was much lighter than that observed upstream. This is an excellent case that illustrates the utility of the basin trace function on FFMP.

  • The Google Earth application was a beneficial tool. It indicated that the region of concern was not heavily developed or populated. Thus, we would likely need to exceed flash flood guidance to acquire a true flash flood event.

  • A media partner at FOX8 TV in High Point did a great job sharing the information she was able to gather and collect off the police scanner. This was a good example of the NWS and the media working as partners.

  • Case study team -
    Michael Moneypenny
    Scott Sharp
    Barrett Smith
    Brandon Vincent
    Jason Beaman
    Brandon Dunstan
    Jonathan Blaes

    For questions regarding the web site, please contact Jonathan Blaes.

  • NWS Disclaimer.