The June 19, 2007 Delaware County Flash Flood:
A Meteorological and Hydrological Analysis
National Weather Service Forecast Office, Binghamton, NY
A meteorological and hydrological analysis of an extreme flash flood event in central New York is presented.
The meteorological analysis indicated that the environment associated with this event contained many characteristics
previously found with convective flash floods. Important elements included unseasonably high atmospheric moisture
content, a strong low-level jet associated with significant moisture flux convergence over the flash flood zone,
a thermal structure characterized by a tall, skinny convective available potential energy profile, and a moist
environment favorable for high precipitation efficiency.
The hydrological analysis indicated that extreme flash flooding occurred in small, steep, heavily forested stream
basins with 3-hour rainfall that doubled the 24-hour 100 year rainfall extreme for the area. Minor flooding began
when estimated rainfall totals exceeded 2.00 inches, while major flooding began when estimated rainfall totals
exceeded 4.00 inches. Peak rainfall during this event eventually exceeded 10 inches. An estimation of the flow
along one of the streams involved in the flood was in excess of a 500-year return flow, consistent with extreme
rainfall rates. Brief, extremely rapid rates of rise on area streams were likely caused by debris pileups giving
way behind highway bridges.
This paper presents an extreme flash flood event that affected the upper Delaware River watershed of Delaware
County, New York on June 19-20, 2007. The setting of the event was several headwater basins that drain into
either the Beaver Kill or Pepacton Reservoir. Eyewitness reports of walls of water and scenes of houses being
washed away and bridges overtopped characterized the event.
Data sources for this study are provided in section 2. A meteorological analysis of the event, including
an examination of the synoptic and meso-scale environment, is given in section 3. Section 4 contains a
hydrological analysis of the event, including an examination of basin response to extreme rainfall, including
the initiation of flash flooding in headwater basins, and impacts downstream along mainstem rivers. Finally,
section 5 contains a summary and conclusion.
Meteorological data including WSR-88D radar reflectivity and precipitation products were obtained from
National Weather Service Advanced Weather Interactive Processing System (AWIPS) workstations. Rain gage
and bucket survey rainfall was obtained from local residents. Streamflow and reservoir elevation records
were obtained from the US Geological Survey (USGS), New York City Department of Environmental Protection (NYC DEP),
and the Mid Atlantic River Forecast Center (MARFC). Eyewitness reports and slope-conveyance indirect discharge
estimation were incorporated into the work.
3. Meteorological analysis
a. Previous research on convective flash flood environments
Junker (1999) examined several heavy convective rain producing systems over the Midwest U.S. in 1993, and found
that these systems typically occurred in areas where a veering, southerly to southwesterly low-level flow resulted
in a significant flux of moist, unstable air poleward across a low-level boundary aligned parallel to the mean flow.
The width of the axis of strong moisture flux correlated positively with the magnitude of the event. Upward vertical
motion was often enhanced over the flood area by divergence in the upper-troposphere associated with the right entrance
region of an upper-level jet-streak, however the heaviest rain typically fell south of the maxima of upper-level divergence.
These features match well with characteristics of convective systems identified as "frontal" or "meso-high" by
Maddox et al. (1979). Similar features were also identified by Moore (2003) in a study on heavy rain producing
elevated convective systems.
Several previous studies have also indicated that flash flood producing convective systems are often characterized
by slow system movement (Senesi et al. 1996, Petersen et al. 1999). Corfidi et al. (1996) examined 103 mesoscale
convective systems and determined that system movement can be approximated by the vector addition of the mean wind
and the opposite of the low-level jet. Another factor that can lead to slow system movement is anchoring of convection
by terrain (Maddox et al. 1978, Petersen et al., 1999, Nicosia et al. 1999).
Other studies have emphasized that flash flooding frequently occurs in environments that are favorable for efficient
rainfall processes. For example, the presence of high environmental relative humidity decreases the potential for
evaporation and dry-air entrainment within a convective storm (Doswell et al. 1996). Market et al. 2003 found a significant
correlation between precipitation efficiency and the relative humidity between the surface and the lifting condensation level.
Another factor that may be related to an environment's potential for producing convective flash flooding is the shape
of the vertical profile of convective available potential energy (CAPE; (HPC training module;
http://www.hpc.ncep.noaa.gov/research/mcs_web_test_test.htm). Tall, skinny CAPE profiles may be more favorable for
heavy rain than shorter, fat CAPE profiles, since storms that form in relatively skinny CAPE environments have relatively deep,
but weak, updrafts, which result in ample precipitation production, but less precipitation being propelled into the upper part
of the storm, where it may be carried away by strong winds aloft.
Local operational experience with flash flood-producing storms in central New York and northeast Pennsylvania has generally
confirmed the findings from the aforementioned studies. Several unpublished local case studies have shown that common features
associated with many significant flash floods in this area include a rapidly veering wind profile in the lowest 1 to 3 km of
the troposphere, capped by a low-level southerly or southwesterly jet, then a deep mid-tropospheric layer of small shear
above the jet. The rapidly veering profile in the lowest 1 to 3 km is indicative of lower-tropospheric warm advection,
and the south-southwesterly low-level jet has frequently proven to be an effective transporter of low-level moisture
northward, from the Gulf coast region. Finally, the pattern of a low-level jet, capped by a weakly sheared middle
troposphere is ideal for producing small system movement, based on the conceptual model introduced by Corfidi et al. (1996).
This pattern becomes particularly favorable for heavy convective precipitation when the environment includes favorable
conditions for efficient rain production.
b. 19 June 2007
The large-scale environment over the northeast U.S. on 19 June 2007 was characterized by ridging in the middle to upper
troposphere, with an eastward moving trough to the west over the Great Lakes region (Figure 1). The lower-troposphere
was characterized by southwesterly flow, and a southwest-northeast thermal ridge over the Appalachian Mountains (Figure 2).
A weak surface trough was forecast to move east across Pennsylvania through 00 UTC on 20 June, with the primary cold front
still located well to the west over Lake Erie. The southwesterly low level flow resulted in increasing deep-layer moisture
across the mid-Atlantic region on 19 June, with a broad axis of precipitable water values over Pennsylvania and southern
New York increasing to over 1.80 inches by 00 UTC on 20 June (Figures 3a-b; approximately 175 percent of normal for mid-June).
The axis of high precipitable water was associated with the development of a broad zone of moisture flux convergence across
Pennsylvania and New York (Figures 3c-d).
Figure 4a shows the Rapid Update Cycle (RUC) model forecast evolution of the wind speed at Avoca (AVP), in northeast
Pennsylvania (about 70 km southwest of the flash flood location) during the late afternoon and evening on 19 June. Wind
speeds were forecast to increase through a deep layer between 18 UTC and 00 UTC, as the trough over the Great Lakes moved
east toward the area, with a 30 knot low-level jet forecast to develop by 00 UTC.
In order to examine a reasonable approximation of the thermodynamic profile over the flood area at the onset of heavy rain,
a 2-hour RUC forecast sounding valid at 02 UTC June 20 at AVP is shown in Figure 4b. This forecast hour was selected since
this was the last RUC forecast available prior to the initiation of convection at AVP in the model. Copious amounts of
moisture are indicated, with a precipitable water value of 1.78 inches, and a mean relative humidity of 85 percent. The
lifted condensation level was below 0.5 km (1500 ft), and the mean relative humidity below the lifted condensation level was
greater than 80 percent. A tall, skinny CAPE profile is shown (indicated by a normalized CAPE value of 0.11). A strongly
veering low-level wind profile is also indicated, culminating in a speed maxima of 25 to 30 knots just above 900 hPa, with
weak speed and directional wind shear above 850 hPa. Nearby real-time surface observations also indicated high lower-tropospheric
moisture and a low lifted condensation level. For example, the 00 UTC observation at Monticello, about 30 km south of the flood
area, indicated a surface dew point of 20°C, and a dew point depression of 6°C, implying a lifted condensation level of about
0.8 km (approximately 2500 ft).
Figures 5a-d show reflectivity data from the Binghamton National Weather Service WSR-88D Doppler radar (KBGM) from 1830 UTC
through 2030 UTC 19 June 2007. The area of interest at this time is over central Pennsylvania near Williamsport, where isolated
storms appeared to re-generate across the same location for a couple of hours. Figures 6a-d show a similar evolution for
storms over central New York, just east of Binghamton, later in the afternoon. In this case, storms appeared to develop
along an outflow boundary that moved southeast across the area. The boundary was progressive, and the storms appeared to
dissipate once the boundary moved off to the south. (Note that that the storms over central Pennsylvania were too distant
from the radar for any associated low-level boundaries to be detected).
Thunderstorms in central New York and northeast Pennsylvania through 2200 UTC on 19 June were mainly isolated in nature, and
were associated with scattered occurrences of wind damage and large hail, but no significant flooding. After 2200 UTC, there
was a pronounced change in this tendency, as storms began to show a trend toward mergers and increased organization. It can
be hypothesized that this change was related to the increasingly strong wind field and vertical wind shear associated with the
approaching mid-to-upper level trough (Figure 4a). Figures 7a-d show that new storms began to develop just ahead of the
southward moving outflow boundary across southern Delaware County, New York around 2200 UTC. Once the boundary caught up
to these storms, additional storms developed and merged into a large, quasi-stationary cluster around 2300 UTC. Meanwhile,
a line of storms originating over northeast Pennsylvania merged into the cluster from the west around 00 UTC on 20 June, with
heavy rain continuing across the area until around 0100 UTC. It is unclear whether or not the low-level boundary continued to
progress southward during this time, as its increasing distance from the radar may have made it undetectable after 2300 UTC.
In summary, the atmosphere on this day appeared to transition from an environment supportive of isolated strong wind and hail
producing storms, to an environment that became more favorable for storm mergers and organization. Isolated stationary and
back-building storms occurred throughout the day, but it was not until the environment changed to being more supportive of
organized convection that stationary and back-building storms resulted in a major flash flood. Finally, it should be noted
that terrain likely played a role in determining where all of the back-building storms occurred on this day. Figure 8a
shows a plot of radar estimated precipitation ending at 00 UTC on the 20th. Figure 8b shows a map of the topography of
the area, with the locations of the back-building storms annotated. In all 3 instances, back-building storms appeared to
develop on the southwest slope of significant topographical features, in locations that were favorable for the moist
southwesterly flow on this day to attain a significant upslope component.
4. Hydrological analysis
a. Radar-indicated basin average rainfall
Rainfall developed across the flood zone around 5:00 PM. The bulk of the rain fell within a three-hour period from
5:30 PM to 8:30 PM. Radar-indicated rainfall totals ranged from 6.0 to 8.0 inches over the upstream half to two
thirds of the study watersheds. Radar-indicated rainfall totals ranged between 2.0 to 6.0 inches over the downstream
half to one third of the study watersheds (Figure 9). The maximum rainfall intensity, observed on a pixel-by-pixel
basis, was approximately an inch within 15 minutes.
Radar-indicated rainfall was averaged over each basin (Figure 10 and table 1) on a 15-minute sampling interval. Spring
Brook and Berry Brook had extremely consistent basin average rainfall throughout the event. Holliday Brook accumulated
the largest basin average rainfall totals and Cat Hollow the lowest. Holliday Brook held the highest intensity rainfall
for the longest time of any study watershed.
b. Ground truth rainfall reports
Rainfall reports were received from residents in the impacted watersheds (Table 2 and Figure 11). Most of the reports
were taken from buckets and other containers open to the air. The diameter of each bucket was not reported. Two reports
were in the 11.00 inch range. All rainfall reports from buckets were in excess of radar rainfall estimates. Rainfall
reports lead credence that the radar represented a minimum rainfall estimate and the radar underestimating has credibility.
c. Rainfall frequency
The Northeast Climate Center provides a 3-hour 100-year rainfall extreme as 2.5 to 3.0 inches and a 24-hour 100-year
rainfall extreme between 5.0 to 7.0 inches (http://www.nrcc.cornell.edu/pptext/). With radar-indicated basin average
rainfalls ranging from 4.0 to 6.5 inches, basin average rainfalls were about 2 times the 3-hour 100-year rainfall
extreme for this area and in the range of a 24-hour 100-year rainfall extreme. Ground truth bucket reports indicate
nearly double the 24-hour 100-year rainfall extreme.
d. Rainfall thresholds
The first report of minor flooding came in at 7:05 PM with water over a roadway. This corresponds to 2.6 inches
or 50 percent of total basin average rainfall for Spring Brook. Moderate flash flooding occurred in Spring Brook
with Route 206 impassible at 7:33 PM. This amounted to 3.85 inches or 73 of total basin average rainfall had fallen
by the time this report was received. Major flash flooding is said to have encompassed the entire Spring Brook
watershed at about 7:50 PM. At 7:50 PM, 4.40 inches of basin average rainfall or 84% of total basin average rainfall
had reached the ground. Minor flooding was reported to the NWS during the event by amateur radio operators. Moderate
and major flash flood times were discussed at the Town of Colchester meeting that NWS Binghamton attended on August 1, 2007.
A resident of lower Berry Brook, David Barnes, reported water began to rise at 6:30 PM and crested one foot over the
local bridge over Berry Brook between 8:30 and 8:45 PM. This amounts to 1.25 inches and 5.60 inches respectively or
22 and 100 percent of total basin average rainfall respectively.
The Town of Colchester supervisor believes that Holliday Brook and Berry Brook had major flash flooding about the same
time as Spring Brook. This amounts to between 72 and 84 percent of total basin average rainfall for Holliday Brook and
61 to 88 percent for Berry Brook.
The Mid Atlantic River Forecast Center had in effect a countywide 1- and 3-hour flash flood guidance value, for
Delaware County, of 3.3 and 4.5 inches respectively. Gridded 3-hour flash flood guidance for the basins flooded
was 4.0 to 6.0 inches (Figure 12).
e. Watershed characteristics
Study watersheds were relatively small at between 3.8 and 9.0 square miles (Figures 13a - d). Stream channels intersected
watershed divides between 2271 feet and 2767 feet. Holliday Brook and Berry Brook have the steepest channel slopes from
watershed divide to point of discharge. Basin characterizes were developed from 1:24,000 scale USGS topographic maps using
Maptech Terrain Navigator software. A list of characteristics for each basin can be found in table 3. Watershed land use
is generally state preserve forest. Development is confined to a narrow corridor in valley bottoms near each brook.
f. Watershed impacts
Damage throughout the study watersheds was indicative of a major flash flood. The flash flood was so massive that it
washed 4 houses away killing 3 occupants. Another person drowned as she apparently tried to find shelter. At least 30
people were evacuated from this area to Roscoe, NY that night.
Additional persons in the Spring Brook and Berry Brook watershed were caught in the flash flood while in their vehicles.
One couple in Spring Brook reported multiple logs hitting the side of their vehicle and having to spend the night in their
vehicle before rescue the following morning. New York State Trooper Joe Decker, of the Roscoe Substation, encountered the
flash flood in Spring Brook while responding and had to abandon his vehicle and swim for his life. Another couple was driving
along the Berry Brook Road and encountered a wall of water as high as their car hood.
A total of 37 homes were affected by the flash flood, with 30 homes sustaining severe damage and deemed unlivable with 4 homes
completely washed away. Roads and bridges in this area took on severe damage (Figure 14). Route 206 between Rockland, NY and
Downsville, NY was completely washed away in one section with a 25-foot high embankment formed by the floodwaters. Holiday Brook,
Spring Brook and Berry Brook roads were also heavily damaged. There were 10 other roads in this area that received flood
damage as well. Four bridges were completely washed out. Twenty-two transformers and 47 power poles were damaged by the
floods. This left 160 homes without power. Phone service was out in the disaster area.
Total damage estimates range from 25 to 30 million dollars in Delaware and Sullivan Counties. Both counties received
Presidential Disaster Declarations along with neighboring Ulster County.
g. Basin response
Eyewitness and news accounts portray an extremely rapidly responding event along lower Spring Brook. Joe Decker
reported encountering a 4-foot wall of water while driving his vehicle. Other witnesses report upwards to an 8-foot
wall of water traveling down the brook. Floodwaters rose about 2-feet per second according to another eyewitness account.
Scour and fill along Spring Brook varied widely. One area of Route 206 was replaced by a 25-foot gully. Along one
section of lower Spring Brook, channel capacity was reduced significantly by 12 feet of fill. Debris was observed in
enormous amounts in both Spring Brook and Holliday Brook. Residents along Spring Brook report hearing what was interpreted
as debris dams giving way at bridge locations along Route 206. An 18-foot high debris pile was reported by emergency
management near the top of the Spring Brook watershed along Route 206.
h. Slope-conveyance discharge estimate
None of the study watersheds contain stream gages. As a result, it was necessary to estimate discharge after the
fact. Berry Brook was selected to for this purpose. The above-mentioned scour, fill, and debris bulking make a
discharge estimate along either Spring Brook or Holliday Brook problematic.
A location was selected about a half of a mile downstream from the end of the old airstrip along Berry Brook. One cross
section was estimated at this point. Excellent high water marks were located on the right bank (Figure 15). These high
water marks were at a uniform height throughout the reach near the cross section indicating relatively steady state
channel conditions. High water elevations were also estimated using debris associated with trees within the floodplain.
This debris was consistent in height with high water marks on the banks (Figure 16). It was assumed that water depth in
the main channel was twice that observed along the roadway.
Discharge was calculated by multiplying the cross sectional area of the flood (A) at peak flow by the average flow
velocity (V) at peak flow:
Q = A * V
Cross sectional area was determined for the main channel and a right bank high-flow channel (Figure 17). Cross sectional
area was 213 ft2 for the main channel and 138 ft2 for the right bank high-flow channel.
Average flow velocity was calculated, for the main and high-flow channels using the Manning flow equation:
V = 1.49R0.66 S0.5 / n-value
Where R is the hydraulic radius, S is the channel slope, and n-value the Manning roughness coefficient. Hydraulic
radius of 3.05 ft was used for the main channel. Hydraulic radius of 2.56 ft was used for the right-bank high-flow
channel. Channel slope of 0.02 ft/ft was used for both the main channel and the high flow channel. The n-value was
determined in the field. An n-value of 0.050 was assigned to the main channel. The USGS had verified an n-value of
0.033 for the next downstream USGS stream gage located along the Beaver Kill at Cooks Falls. A higher n-value was
selected for this cross section due to the presence of vegetation, including trees, on the right bank of the main channel.
An n-value of 0.060 was assigned to the right-bank high-flow channel. This n-value represents an average of several
n-value extremes present including grass, road/eroded road surface material, and a wooded area.
Using the above values, an average velocity of 8.8 feet per second was calculated for the main channel and 6.5 feet per
second for the right-bank high-flow channel. The above velocities do not include the effect of obstruction / retardation
caused by vegetation other than the n-value itself. Using these velocities, a peak discharge of 1,874 cfs was obtained
for the main channel and 897 cfs for the right-bank high-flow channel. This yields a total discharge of 2771 cfs.
The above slope-conveyance discharge estimate should be considered a rough discharge estimate. Due to estimating cross
sectional area, experience-based selection of Manning roughness value, and an assumption that channel slope approximates
surface water elevation slope, the discharge estimate will have more error than that of a multiple cross section slope-area
indirect discharge estimate or a discharge estimate taken from a step-backwater model. As a result, the authors assigned
error of approximately 25 to 50 percent.
i. Flood frequency
The equations for ungaged basins in New York region 3 (USGS 2006) were used to assign flood frequency for Berry Brook.
A drainage area of 4.6 square miles was used (Table 4). Peak discharge for Berry Brook was in excess of the 500-year
flood frequency (less than a 0.2% chance of exceedance chance in a given year).
j. Probable maximum flood
The probable maximum flood (PMF) is an extreme flood based on the most severe hydrologic and meteorologic conditions
considered reasonable for the site. In general, exceedance probabilities are not assigned to PMFs. National Flood Frequency
program calculates the PMF using Chippen and Bue procedures (1977). PMF for Berry Brook was 13,100 cfs for region 4 and a
4.6 square mile watershed. As a result, Berry Brook’s peak discharge was less than 25% of the PMF.
k. Comparison to USGS gage on Beaver Kill
The USGS stream gage located along the Beaver Kill at Cooks Falls registered a rapid rise (Figure 18). In a 15-minute
time span, it rose 5.40 feet / 5,000 cfs. The USGS gage registered the crest at 8.64 feet / 6,840 cfs at 12:00 AM
local time, June 20th. The USGS using the crest-stage gage located at the site recorded a peak of 10.16 feet / 9,900 cfs.
This was defined as a provisional peak at the time of this technical attachment. Considering the flash flood nature of
this event, a rise to a crest of 10.16 feet was not inconceivable considering the 15-minute data resolution from the
USGS stream gage. It was hypothesized that the rate of rise of the river stage, in the Beaver Kill, was too rapid for
the water in the USGS stilling well to keep up with.
The USGS stream gage is located a short distance downstream from the outlet of Spring Brook and Berry Brook
(7.9 and 11.9 miles). While some attenuation may have occurred prior to reaching the stream gage, the crest was likely
representative of the sum of base flow at the time plus the individual flash flood peak discharges from Spring Brook,
Berry Brook, and Pelnor Hollow (a small watershed situated between Spring and Berry Brook). Due to the uniformity of
rainfall mentioned above between Spring and Berry Brook, it was assumed that a ratio of drainage area allows for the
estimation of peak discharge from Spring Brook and Pelnor Hollow. With Spring Brook at 9.0 square miles and Pelnor Hollow
at 2.0 square miles, peak discharge should be 196% and 43% of that of Berry Brook respectively or 5,431 cfs and 1,192 cfs.
With a base flow at the time of 120 cfs, a total of 9,514 cfs or 96% of the peak flow was accounted for using this method.
The Beaver Kill at Cooks Falls crest continued to travel downstream for a period of 48-hours and could be picked out
as far downstream as Montague as a significant rise (Figure 19).
l. Inflow into Pepacton Reservoir
Runoff from the Holliday Brook, Cat Hollow, and nearby north-facing watersheds flowed into Pepacton Reservoir. NYC DEP
reported inflow into Pepacton Reservoir peaked at about 16,000 cfs at 9:40 PM local time on June 19th (Figure 20). The
reservoir did not spill and stayed below the spillway crest throughout the multi-day period of the rise (Figure 21). The
presence of Downsville Dam prevented flooding in the community of Downsville downstream to Harvard (Figure 22).
5. Summary, discussion and conclusions
The extreme flash flood event of June 19, 2007 occurred in the type of meteorological environment that has previously
been found to be favorable for flash flooding in upstate New York. Key environmental factors included an unusually
moist atmosphere, and a wind field that evolved to promote increased organization of convection into a large cluster
of nearly stationary thunderstorms. The sudden change on this day from isolated convective storms to an organized,
stationary cluster presented warning meteorologists with a difficult challenge, as they were forced to “shift gears”
midway through the event, from warning for pulse severe convection, to warning for a major flash flood.
Regarding the hydrology of this event, we described an extreme rainfall event and resultant headwater flash flood that
translated into rapid rises on downstream rivers. Basin average rainfall greatly exceeded a 3-hour 100-year rainfall.
Results of the bucket survey point to upwards to twice the 24-hour 100-year rainfall extreme compressed into a three-hour
timeframe. Flood frequency for Berry Brook in excess of a 500-year return flow is consistent with extreme rainfall frequencies.
Rates of rise up to 2 feet per second, along lower Spring Brook, point to structural failure immediately upstream.
Reports of debris pileups giving way behind highway bridges were the likely cause. No evidence was found of dams or
Basin average rainfall of 2.00 to 2.50 inches initiated stream rises and reports of minor flooding. Moderate flooding
was reached by 4.00 to 5.00 inches. Major flooding was witnessed in the 5.00 to 8.00 inch range. Bucket survey reports
upward to 11.00 inches point to the radar underestimation of rainfall.
A timeline for the event can be viewed in table 5.
Jim Porter from NYC DEP for potential impacts on Downsville and Harvard and his useful review comments.
William B. Reed, Senior Hydrologist, NWS, for his review comments and input on slope-conveyance discharge
estimation for Berry Brook. Gary D Firda, of the USGS Troy New York office, for review comments on slope-conveyance
discharge estimation and flood frequency for Berry Brook. Many thanks to Robert Fenner from MARFC for his graphical
representation of the flash flood crest at Cooks Falls and down through the Delaware River system. To the many residents
of Berry Brook and the Town of Colchester who took the time to e-mail or be interviewed on the phone.
|Figure 1. NAM 00-hour and 12-hour forecast 500 hPa heights (dm) and vorticity (1x10-5s-1, positive values shaded) valid at (a) 12 UTC, June 19, 2007 and (b) 00 UTC June 20, 2007."
Figure 2. NAM 00-hour and 12-hour forecast sea-level pressure (mb, blue contours) and 850 hPa temperature (oC, yellow contours) valid at (a) 12 UTC, June 19, and (b) 00 UTC, June 20.
Figure 3. NAM 06 and 12 hour forecast precipitable water (in) and 850 hPa wind (kts) valid at
(a) 18 UTC, June 19, and (b) 00 UTC, June 20. RUC 00 hour and 6 hour forecast 850 hPa
moisture flux convergence (gkg-112 hr-1, values > 5 shaded) and 850 hPa wind (kts)
valid at (c) 18 UTC, June 19, and (d) 00 UTC, June 20.
Figure 4. (a) BUFKIT time-height display of 18 UTC, June 19 RUC forecast wind speed (kts) at Avoca, Pa (AVP) from 18 UTC through 00 UTC. (b) BUFKIT display of the 00 UTC, June 20 RUC forecast sounding at AVP valid at 02 UTC on June 20.
Figure 5. KBGM WSR-88D reflectivity at (a) 1830 UTC, June 19, (b) 1900 UTC, June 19, (c) 1930 UTC, June 19, and (d) 2000 UTC, June 19.
Figure 6. KBGM WSR-88D reflectivity at (a) 1930 UTC, June 19, (b) 2000 UTC, June 19, (c) 2030 UTC, June 19, and (d) 2100 UTC, June 19.
Figure 7. KBGM WSR-88D reflectivity at (a) 2100 UTC, June 19, (b) 2130 UTC, June 19, (c) 2200 UTC, June 19, (d) 2230 UTC, June 19, (e) 2300 UTC, June 19, (f) 2330 UTC, June 19.
Figure. 8 (a) KBGM WSR-88D radar estimated rainfall through 00 UTC, June 20. (b) Topography of central New York and northeast Pennsylvania (shaded; kft), with the location of the heavy rain events on June 19 annotated.
Figure 9. Storm total precipitation from June 19th overlaid with NWS Flash Flood and Monitoring Program (FFMP) small basin outlines. Basins are labeled for upper Spring Brook (A), middle to lower Spring Brook (B), lower Spring Brook near airfield (C), Berry Brook (D), Pelnor Hollow (E), Holliday Brook (F), and Cat Hollow (G).
Figure 10. Basin average rainfall derived from KBGM WSR-88D radar plotted against time on about a 15-minute time interval.
|Table 1 Basin Average Rainfall (inches)
|Time (EST)||Spring Brook||Berry Brook||Holiday Brook||Cat Hollow|
|5:34 PM ||0.37 ||0.10 ||0.56 ||0.45|
|5:47 PM ||0.86 ||0.25 ||0.78 ||0.62|
|5:59 PM ||1.11 ||0.73 ||1.25 ||1.03|
|6:16 PM ||1.42 ||1.11 ||2.06 ||1.55|
|6:32 PM ||1.66 ||1.27 ||2.74 ||2.09|
|6:44 PM ||1.94 ||1.44 ||3.46 ||2.42|
|7:01 PM ||2.59 ||2.12 ||3.81 ||2.44|
|7:13 PM ||3.13 ||2.57 ||4.09 ||2.46|
|7:30 PM ||3.85 ||3.44 ||4.65 ||2.59|
|7:42 PM ||4.32 ||4.17 ||5.11 ||2.71|
|8:00 PM ||4.56 ||4.92 ||5.49 ||3.13|
|8:16 PM ||5.12 ||5.49 ||6.11 ||3.75|
|8:37 PM ||5.25 ||5.61 ||6.50 ||4.00|
Table 1. Basin average rainfall derived from KBGM WSR-88D radar plotted against time on about a 15-minute time interval.
|Basin Location||Elevation (ft)||Source||Gage type||Rainfall (inches) |
|Cat Hollow||2200||Town of Colchester||Bucket||8.25|
|Lower Spring Brook||1600||Town of Colchester||Bucket||> 11.00|
|Upper Holliday Brook||1900||Town of Colchester||Bucket||9.00|
|Upper Spring Brook||2100||Edward Hamerstrom||Bucket||11.00|
|Upper Berry Brook along Henderson Road||2100||Edward Hamerstrom||Rain gage||6.00|
|Lower Berry Brook along Berry Brook Road||1600||Eric Hamerstrom||Rain gage||2.00|
|Pelnor Brook about 0.5 miles up from Beaver Kill||1600||Edward Hamerstrom||Swimming pool||9.50|
|Upper Beaver Kill near Lew Beach||1800||Edward Hamerstrom||Bucket||9.50|
Table 2. Rainfall reports throughout flash flood area.
Figure 11. Rain gage and bucket reports received from throughout the flash flood area.
Figure 12. MARFC 3-hour Flash Flood Guidance. Area outlined in black shows area of flash flooding of Delaware, Sullivan, and Ulster counties.
Figure 13. Watershed boundaries created from Maptech Terrain Navigator software (a) Spring Brook at landing strip, (b) Berry Brook half-mile below old airstrip, (c) Holliday Brook at Route 30 (d) Cat Hollow at Highway 30.
|Basin Name||Drainage Area (sq mi)||Linear profile slope (ft/ft)||Maximum channel flow length (miles)||Maximum elevation (ft)||Minimun elevation (ft)|
|Spring Brook at Landing Strip||9.0||0.034||4.97||2271||1391|
|Berry Brook below old Airstrip||4.6||0.069||0.02 at indirect discharge location||3.00 2767||1650|
|Holliday Brook at Route 30||4.8||0.076||3.51||2703||1299|
|Cat Hollow near Route 30||3.8||0.059||3.39||2409||1351|
Table 3. Basin characteristics derived from 1:24,000 scale USGS Topographic maps using Maptech Terrain Navigator software.
Figure 14. Binghamton Weather Forecast Office Senior Service Hydrologist standing at roads end along Berry Brook just downstream from old airstrip. Photo looking upstream.
Figure 15. Berry Brook indirect discharge location located about a half mile downstream from end of old landing strip. Photo taken looking upstream from right bank.
Figure 16. Berry Brook indirect discharge location. White line drawn in to show water surface elevation consistent between right bank high water mark and in-channel debris.
Figure 17. Channel cross section for Berry Brook looking in the downstream direction. Graphic is vertically exaggerated.
|Stream flow return frequency||Discharge|
Table 4. Stream flow return frequency, for Berry Brook.
Figure 18. Graph of observed river stages at USGS stream gage at Cooks Falls along the Beaver Kill. Rapid rise of 5.4 feet in 15-minutes can be seen. USGS crest stage gage at the site recorded a crest of 10.16 feet. National Weather Service flood stage of 10.00 feet shown on graph in red.
Figure 19. Graph shows flash flood crest at Cooks Falls in green and progression of the flood wave downstream along the Delaware River.
Figure 20. Graphic prepared by NYC DEP to show inflow into Pepacton Reservoir. Maximum inflow into reservoir of 15,969 cfs was reached at 9:40 PM on June 19, 2007. Holliday Brook, Cat Hollow, and several smaller north-facing watersheds are thought to have contributed the majority of this water.
Figure 21. Analysis conducted by NYC DEP to show mitigating effects of Downsville Dam on flood crest. Flood stage reached during flood of June 2006, hypothetical stage if dam was not present during June 19-20 flood, and actual stage reached from June 19-20 flood. Flood stage at USGS gage at Harvard is 10 feet. Hypothetical flood stage assumes the inflow hydrograph into Pepacton Reservoir coalesced into a flood peak of equivalent discharge in the Pepacton reach of the upper Delaware River and traveled downstream without significant attenuation as far as Harvard located approximately 12 miles downstream from the Downsville Dam. This hypothetical stage of 14 feet is a maximum possible stage. Attenuation of peak flow downstream of the site of Downsville Dam and/or peak flow hydrograph generation, which differs from the inflow hydrograph for Pepacton Reservoir, could lead to a lower crest at Harvard.
Figure 22. Graphic of pool water elevation at Pepacton Reservoir. It took a total of 4 full days for the rise in pool water elevation to level off after the evening of rainfall event of June 19, 2007.
|9/19||5:30 PM||Rain begins to fall in Spring Brook and Cat Hollow watersheds.|
|9/19||6:30 PM||Water began to rise along lower Berry Brook.|
|9/19||7:01 PM||Flash flood warning issued for south-central Delaware County|
|9/19||7:05 PM||First report of minor road flooding received.|
|9/19||7:10 PM||Approximate time that MARFC 3-hour gridded Flash Flood Guidance was exceeded.|
|9/19||7:33 PM||Moderate flash flooding with Route 206 impassible.|
|9/19||7:50 PM||Major flash flooding along lower Spring Brook.|
|9/19||8:30 PM||Major flash flooding along lower Berry Brook.|
|9/19||9:40 PM||Peak inflow into Pepacton Reservoir.|
|9/20||12:00 AM||Peak flow at USGS gage along Beaver Kill at Cooks Falls.|
Table 5. Event timeline.
Barnes, H. H. 1967: Roughness Characteristics of Natural Channels USGS Water Supply
Corfidi, S. F., J.H. Meritt, and J.M. Fritsch 1996: Predicting the Movement of Mesoscale
Convective Complexes, Wea. Forecasting, 11, pp. 41-46
Doswell, C. A., H. E. Brooks, and R. A. Maddox: 1996: Flash Flood Forecasting: An
Ingredients-Based Methodology, Wea. Forecasting, 11, 560-581
Junker, N.W., R. S. Schneider, and S. L. Fauver: 1999: A Study of Heavy Rainfall
Events during the Great Midwest Flood of 1993, Wea. Forecasting, 14, 701-712
Limia, R., D. A. Freehafer, and M. J. Smith: 2006: Magnitude and Frequency of Floods
in New York, USGS Scientific Investigations Report 2206-5112
Maddox, R. A., L. R. Hoxit, C. F. Chappell, and F. Caracena: 1978: Comparison of
Meteorological Aspects of the Big Thompson and Rapid City Flash Floods, Mon.
Wea. Rev., 106, 375-389
Maddox, R. A., C. F. Chappell, and L. R. Hoxit, 1979: Synoptic and meso-alpha scale
aspects of flash flood events, Bull. Amer. Meteor. Soc., 60, 115-123.
Market, P., S. Allen, R. Scofield, R. Kuligowski, and A. Gruber, 2003: Precipitation
Efficiency of Warm-Season Midwestern Mesoscale Convective Systems, Wea. Forecasting,
Moore, J. T., F. H. Glass, C. E. Graves, S. M. Rochette, and M. J. Singer, 2003: The
Environment of Warm-Season Elevated Thunderstorms Associated with Heavy
Rainfall over the Central United States, Wea. Forecasting, 18, 861-878