1. INTRODUCTION
Historically, flooding
has been and continues to remain the greatest natural disaster to affect Vermont. The Great Flood of November 1927
claimed 84 lives, including the life of the Lieutenant Governor. Recently,
there have been 8 Federally-declared flood disasters in Vermont (1992 to 2002), resulting in
millions of dollars in damage as well as some fatalities.
Flood disasters in Vermont occur in both the cool season
(winter and spring) and warm season (summer and fall). Cool season events often have saturated or
impervious soils due to snow cover or frozen ground, and widespread duration
rainfall produced by upper jet dynamics. Warm season events typically have the
most intense rainfall produced by low level convergence or instability, and
large amounts of atmospheric moisture.
The greatest (severe)
flooding often results when long-duration heavy rainfall occurs across a
widespread area of saturated ground and steep terrain. The mountainous terrain
of Vermont’s Green Mountains contains numerous small river
basins with steep slopes. In these
basins, terrain relief reaches as high as 4000 feet, which contributes to rapid
runoff of heavy rainfall.
The 12
June 2002
flood event was a warm season flood event but also had some elements of a cool
season flood. It was focused over some
of Vermont’s steepest terrain, already
saturated by a previous rainfall event.
Unseasonably strong jet dynamics combined with warm season moisture and
stationary low level convergence produced
a wide band of moderate to heavy rainfall and a concentrated band of
flood-producing excessive rainfall.
*Corresponding author address:
Scott Whittier, NOAA/National
Weather Service, Burlington Intl. Airport, 1200 Airport Dr., S. Burlington, VT 05403.
email: scott.whittier@noaa.gov
Figure 1 shows the
24-hour rainfall from this event with widespread rainfall amounts exceeding
0.50-in (13 mm) across the northern third of New York and northern two-thirds of Vermont. Note the narrower, linear “banded”
structure of the greater than 2 in (51 mm) and the localized greater than 3 in
(76 mm) rainfall in mountainous locations along the Canadian border.
Operational Numerical Weather Prediction (NWP) model quantitative precipitation
forecast (QPF) fields adequately predicted the synoptic scale rainfall amounts
and location. However, these models may have lacked the mesoscale detail needed
to forecast localized rainfall amounts that exceeded twice the QPF values.
A locally developed
conceptual model of warm season excessive rainfall events in Vermont was used to effectively assess and
warn for the flooding on 12 June 2002. Section 2 will discuss the
conceptual model and Section 3 will show how this case is consistent with the
model. Section 4 will review the forecast and warning decisions made on 11-12
June 2002 and Section 5 will describe the observed rainfall and flooding. This
study demonstrates the value of developing conceptual models for use in the
forecast process.
2. CONCEPTUAL MODEL OF A WARM-SEASON EXCESSIVE RAINFALL EVENT IN NORTHERN VERMONT
The
continual flooding problem in Vermont, especially over the last decade,
has led NWS Burlington forecasters to the development of a conceptual model.
This model depicts atmospheric conditions most favorable to produce excessive
rainfall when superimposed on the topography of northern Vermont and serves as a “heads up” for
closer interrogation of a possible flood event.
Research and case
studies suggest the use of conceptual models and ingredient-based precipitation
forecasts (Doswell et al. 1996) are valuable in recognition of heavy rainfall
events. Conceptual model parameters
often linked to the occurrence of heavy rainfall include measures of absolute
moisture, instability, and vertical motion.
These parameters are useful in predicting rainfall intensity, but
experience repeatedly demonstrates that the stationary nature of these
parameters is often the key to heavy rainfall that results in flooding (Doswell
et al. 1996).
Figure 2 is a conceptual model of
meteorological ingredients conducive to heavy rainfall in northern New York, northern
Vermont, and
adjacent areas of southern Quebec. This conceptual model has been in
use for nearly 10 years at NWS Burlington to help identify heavy rainfall
events. Parameters shown (Fig. 2) produce vertical motion through low level
overrunning of tightening thermal gradients (frontogenesis) at 850 hPa,
directional and speed convergence at 850 hPa, topographic lift of low and mid
level westerly flow, and upper divergence and indirect circulation of jet
streaks (250 hPa). The conceptual model also includes abnormally high amounts
of absolute moisture (Lapenta et al. 1995) with precipitable water frequently
exceeding 1.5-in (38mm) and dew points at 850 hPa exceeding 12ºC.
Finally,
total rainfall is greater when parameters remain stationary over the same
area. In the model this generally occurs
when there is a stationary front at 850 hPa, balanced between the inflow at 850
hPa and the low level outflow of the rain-cooled air. In some cases, upper air
divergence and new areas of vertical motion redevelop upstream of existing
convective cells, resulting in a zero net displacement as individual cells move
downstream.
3. NWP FORECASTS AND OBSERVATIONS CONSISTENT WITH CONCEPTUAL MODEL
The 12
June 2002
event showed a strong correlation to the conceptual model, with low level
overrunning and convergence, upper divergence, and abnormally large amounts of
absolute moisture.
Figure 3 shows 0000 UTC 12
June 2002
MesoETA model analysis of frontogenesis at 850 hPa, and the tightening of
thermal gradients where the overrunning ascent is maximized. This axis of
frontogenesis over northern Vermont was crucial in forecasting where
the greatest upward vertical motion would be focused (Nicosia and Grumm 1999, as well as showing
the quasi-stationary nature of the boundary. Directional and speed convergence
helped enhance the vertical lift along this boundary.
Figure
4 shows 1200 UTC 12 June 2002 12-hr forecasted upper divergence
at 250 hPa in the right rear quadrant of an un-seasonably strong 120-140 knot
jet max (color shaded). This upper divergence, also centered over northern Vermont, promoted deeper upward vertical
motion above the area of 850 hPa frontogenesis during the period of forecasted
heaviest rainfall. Observed precipitable water values at 0000 UTC 12 June 2002
(Fig. 5) were approximately 200% of normal for June, with greater than 1.5-in
(38mm) over northern Vermont and available upstream.
4. FORECAST AND WARNING DECISIONS
QPF from all NWP models
was 1 to 2 inches with this event, and basically in the correct area. However,
forecasters recognized the similarity to the conceptual model and gained
confidence to forecast rainfall amounts more than twice as large as the model
QPF, and issue flood warnings more than 6 hours before flooding began.
Initial forecasts and
flood watches, on the afternoon on 11 June 2002, generally reflected the
placement and amounts of the model QPF, concentrating on northwestern Vermont
in the areas of greatest upslope, steepest terrain, and saturated soils caused
by 1 to 2 inches of rain in the preceding 10 days.
During the evening of 11
June 2002,
satellite, radar, and upper air observations continued to align with the conceptual
model. The accuracy of the MesoETA and ETA model forecasts of frontogenesis at
850 hPa and divergence at 250 hPa was becoming more apparent as areas of
heaviest rain reformed in southern Quebec and moved southeastward in a narrow
band across all of northern Vermont. At 0000 UTC 12 June 2002, flood watches
were expanded to include the northern half of Vermont, and public statements
were issued which mentioned greater than 4 inches of rainfall would be possible
in less than 18 hours.
At midnight (EDT) 12 June
2002,
locally more than 2 inches of rainfall was already estimated across
northwestern Vermont, using the Weather Surveillance
Radar (WSR-88D) at Colchester, Vermont (KCXX; Fig. 6). Rainfall rates of
about 0.25 in (6mm) per hour were being observed. Upstream, over southeastern Ontario and southwestern Quebec, satellite imagery showed an
expanding area of colder cloud tops (Fig. 7) indicative of deeper vertical
motion yet to arrive in Vermont.
Aware that several more hours of heavy rain was likely across northern
Vermont, and that monitored river gauges were beginning to show substantial
rises, forecasters issued county-wide flood warnings for Caledonia, Franklin,
Lamoille, and Orleans counties in northern Vermont and mentioned specific river
basins would be directly affected, including the Passumpsic and Missisquoi
basins.
5. OBSERVED RAINFALL AND FLOODING
Storm total rainfall
between 1200 UTC 11 June 2002 and 1400 UTC 12
June 2002
was 2.0 to 4.5in. (50 to 113mm) across the northern half of Vermont and northeastern New York.
Figure 8 shows a few observed rainfall totals plotted over the storm
total radar estimates from WSR-88D, KCXX. NWS cooperative observers in Vermont reported 4.50 in. (113mm) at Jay Peak, 4.30 in. (107mm) at Sutton and East Haven, and 4.20 in. (105mm) in Albany.
Substantial flooding
occurred across several small watersheds across Caledonia, Franklin, Lamoille and Orleans counties beginning around 1000 UTC
and continuing through the afternoon of 12 June 2002. These four counties were
eventually declared Federal Disaster areas (FEMA-1428-DR).
The flood damage was
greatest in the headwater regions of the Passumpsic and Missisquoi River basins. Figure 9 shows a time sequence of
river gauge readings from the United States Geological Survey (USGS) river
gauge on the East Branch of the Passumpsic River.
Note the increase in volumetric flow that occurred between late morning
on 11 June and late morning on 12 June, around a 4000% increase in flow and an
estimated 100-year recurrence event. This was the highest river stage ever
recorded on the East Branch of the Passumpsic River and on the Misissquoi River at North Troy, VT. Also during this event, the river
stage measured at the mouth of the Passumpsic River was the second highest ever reported.
6. SUMMARY
Forecasters used a
conceptual model to recognize the presence of critical ingredients for
excessive rainfall, permitting issuance of accurate countywide flood watches,
warnings and river flood warnings 6 hours before flooding developed.. NWP
analysis and forecast products described areas of vertical motion and were
helpful to refine location, intensity and duration of the rainfall, and to
predict rainfall amounts and expand flood watches. Finally, careful monitoring
of remote sensing tools like satellite, radar, and river gauges enabled
forecasters to issue and refine flood warnings.
ACKNOWLEDGEMENTS
The authors would like to thank Paul
Sisson, Science and Operations Officer at WFO Burlington, for his assistance in
preparation of this manuscript. Heather Hauser at the NWS Eastern Region
Headquarters, Scientific Services Division, also provided valuable review of
this manuscript.
REFERENCES
Doswell, C. A., H. Brooks, and R.
Maddox, 1996: Flash Flood Forecasting:
An Ingredients-Based Methodology. Wea. Forecasting, 11, 560-581.
Lapenta, K. D., B. J. McNaught, S.
J. Capriola, L. A. Giordano, C. D. Little, S. D. Hrebenach, G. M. Carter, M. D.
Valverde, D. S. Frey, 1995: The Challenge of Forecasting Heavy Rain and
Flooding throughout the Eastern Region of the National Weather Service. Part I:
Characteristics and Events. Wea.
Forecasting, 10, 78-90.
Nicosia, D. J., R. Grumm, 1999: Mesoscale Band Formation in Three Major Northeastern United States Snowstorms. Wea. Forecasting, 14,
346-368.
