Burlington Int’l Airport

South Burlington, VT 05403

August 26, 1997

Memorandum for: Robert Bell, MIC - NWSO BTV

Through: Stephen Hogan, WCM - NWSO BTV

From: Scott Whittier, Hydro Focal Point - NWSO BTV

Subject: Hydrological Report on July 15, 1997 Flood Event

Introduction

Devastating "flash" flooding occurred in north-central and northeast Vermont on July 15^th as a result of 3 to 6+ inches of rainfall over steep, complex terrain within a 6 to 8 hour period.

Preliminary storm damage estimates were $10+ million (personal property losses not included). Presidential disaster areas were declared for Caledonia, Franklin, Lamoille, Orleans and Washington counties in Vermont.

This report summarizes the following aspects of this hydrologic event; rainfall, river responses and operational tools and guidance available for this event.

History

NWSO Burlington’s Hydrologic Service Area (HSA) includes all of Vermont (except Windham and Bennington counties) and the northern New York counties of Clinton, Essex, Franklin and St. Lawrence.

The terrain within NWSO Burlington’s HSA is complex and diverse, consisting of two major mountain ranges with elevations exceeding 4000 feet (Adirondacks and Green Mountains), two major river valleys (St. Lawrence and Connecticut) and the fifth largest fresh water lake in the country (Lake Champlain). More importantly is the series of streams, brooks and rivers that must transport large quantities of water from mountain peaks (4000+ feet) to valley floors within short distances (20-30 nm), thus creating the potential for flooding.

In the not so distant past, flash flooding has occurred in portions of the same areas (Montgomery-Jay Peak 6/93, Lamoille river valley 8/95) affected by the flood event discussed in this report.

Since 1995, flooding has resulted in five Presidentially Declared Disaster areas in NWSO Burlington’s HSA (Vermont; 8/95, 1/96, 7/97 and New York; 10/95, 1/96, 11/96). These floods have resulted in at least three fatalities and monetary damage in excess of $50 million.

Other known previously declared presidential disasters due to flooding in Vermont included; June 1973, August 1976, June 1984, August 1989, July 1990 and March 1992. All of the above, except March 1992 were the result of "convective" flash flooding (due to convective weather).

Hydro-meteorology of the Event

An upper-level ridge was centered across western New York and extended north into the province of Ontario and Hudson Bay region of Canada. This pattern resulted in a upper-level northwest flow across northern New England. Water vapor imagery clearly showed a tropical connection from the western Carribean, northward along the backside of the ridge into Hudson Bay, then downstream into the province of Quebec (QB) and northern New England. Meanwhile, a closed upper-level low in the Canadian Maritimes resulted in the formation of a backdoor cold front at the surface. This cold front moved west across Maine and New Hampshire, underneath the upper-level ridge. The front reached a Montreal-Montpelier-Lebanon line by 00z on the 15^th.

The combination of a favorable upper-level jet stream, a weak short wave, a tropical moisture connection (precipitable water values greater than 1.5 inches) and convergence along the frontal boundary accounted for the development of numerous showers and thunderstorms in southern Quebec. These convective rains traveled southeast into north central and northeast Vermont between 03z and 12z.

Rainfall

Widespread rains of 3 to 5 inches with localized amounts in access of 6 inches occurred between 03z and 12z on the 15^th as observed by official NWS cooperative observers (Attachment 1). The heaviest rains fell within a 6 to 8 hour period from 04z to 12z. During this period, WSR-88D estimated rainfall rates of 1 to 1.5 inches per hour were noted at times.

Some observed rainfall totals included:

6.58 inches Jay Peak Franklin county

5.68 inches Albany Orleans county

5.22 inches Greensboro Orleans county

4.87 inches Groton Caledonia county

4.75 inches Westfield Orleans county

4.51 inches Eden Lamoille county

3.85 inches Marshfield Washington county

3.08 inches St. Johnsbury Caledonia county

3.01 inches Lake Elmore Lamoille county (2.74 inches; 06-11z)

3.00 inches Berkshire Franklin county

There was also an unofficial rainfall report from Montgomery Center, VT (eastern Franklin county) of 7.50 inches.

Observed rainfall reports from Environmental Canada indicated an elongated area of 2.5 to 4 inches from southeast of Montreal to Richford, Vermont (Franklin county) including 5.1 inches in Abercorn, QB (approximately 5 nm N of Richford, VT).

River Responses

Much of the rainfall occurred over the steep and complex terrain of north central and northeast Vermont. This geographical area includes dozens of smaller river basins, which comprise the tributaries to larger river basins such as the Black, Lamoille, Missisquoi and Passumpsic. Clearly, flash flooding of many brooks, streams and small rivers was the major flood event.

The greatest devastation occurred in and around Montgomery and Montgomery Center along Route 118 in Franklin county, Vermont (Attachment 1a). This area is located along the Trout River and its tributaries of the West Falls and Black Falls brooks. Also, road washouts were experienced along Route 242 from the Jay Brook between Montgomery Center and Jay. Other significant flooding occurred in Albany (Orleans county) and Lyndon Center (Caledonia county) along both branches of the Passumpsic river and Millers Run.

The inflows from these tributary sources was tremendous causing water levels of the larger rivers to rise up to 3 feet/hr at times (See Table 1). This caused flooding of the larger rivers, especially at the headwater and tributary points (Attachment 2).

Some of these larger rivers in Vermont that experienced flooding (mainly road and crop damage)included:

- the Missisquoi between Richford and Sheldon

- the Lamoille between Hardwick and Cambridge

- the Black between Albany and Newport

- the Passumpsic between East Burke and East Barnet

- the Wells between Groton and Wells River

Table 1: Observed rises in water levels at the following USGS gages(LST...add 5 hours to get UTC):

Greatest Rises Observed (ft)

Gage Location 15 min 30 min 1 hour 2 hour 3 hour

Missisquoi at 0015-0030 0015-0045 0015-0115 0015-0215 0015-0315

E. Berkshire 1.01 1.88 3.15 5.11 6.28

Lamoille at 0615-0630 0600-0630 0545-0645 0545-0745 0545-0845

Johnson 0.78 1.52 2.68 4.76 6.59

Passumpsic at 0615-0630 0615-0645 0600-0700 0530-0730 0530-0830

Passumpsic 0.66 1.28 2.43 4.46 6.01

Table 2: NWSO BTV Flood Stage Report - NWS E3 for July 1997

Dates Above Prior to

Flood Flood Stage Rainfall Crest

River and Station Stage From To Stage Stage Date/Time

Black River at

Coventry, VT 5 15/10z 17/12z 2.18 6.80 16/07z

Lamoille River at

Johnson, VT 11 15/13z 16/08z 2.50 16.46 15/21z

Missisquoi River at

East Berkshire 12 15/07z 15/22z 2.99 15.75 15/11z

Passumpsic River at

Passumpsic, VT 14 ------ ------ 2.63 13.28 15/20z

Hydrologic Tools and Guidance

WSR-88D (KCXX)

Strengths: NWSO Burlington’s WSR-88D (KCXX) was operational during this event and clearly showed the continued development of shower and thunderstorm activity over southern Quebec moving southeast into Vermont.

The WSR-88D One Hour Precipitation (OHP) product estimated at least 1 inch/hr rates for several volume scans during the event and Storm Total Precipitation products (STP) showed an accurate position axis of heavy rainfall in the flood ravaged area (Attachment 3).

Limitations: WSR-88D Storm Total Precipitation algorithms under-estimated actual observed rainfall by as much as 30 to 50 percent. This problem may have been the result of two factors; beam blockage and invoked Z/R relationship.

Much of the area impacted by heavy rainfall and flooding in Vermont occurred within KCXX’s beam blockage pattern at 0.5 degrees (>60 percent blockage - Attachments 4,5). This beam blockage of the lowest elevation angles impacted the accuracy of rainfall estimates within the flood area. WSR-88D storm totals estimated maximum rainfall of 3.5 to 4 inches around Montgomery Center with a broad area of 2-3 inches across north central Vermont. Actual rainfall totals of 3 to 5 inches with locally higher amounts were observed. Another beam blockage problem existed south of a Hyde Park to Glover, Vermont line where a beam blockage pattern at 1.5 degrees exist (>60 percent). Storm Total Precipitation estimates of 1-2 inches was depicted for Caledonia, southern Orleans, southeast Lamoille and northeast Washington counties. In reality, actual rainfall amounts of 2 to 5 inches were observed within this area.

Table 3: WSR-88D Storm Total Precipitation estimates vs. Observed Rainfall

Location STP estimates (Inches) Observed

Franklin 2-3, locally 4 3-6+

Montgomery 3.5-4.0 7.50

Berkshire 2.0-2.5 3.00

Lamoille 2-3, locally 3.5 3-5+

Eden 2.5-3.5 4.51

Lake Elmore 2.5-3.5 3.01

Orleans 2-3, locally 4 3-6+

Jay Peak 3.0-3.5 6.58

Albany 3.0-3.5 5.68

Greensboro 1.5-2.0 5.22

Caledonia <1-2, locally 2.5 2-5

Groton 1.5-2.0 4.87

St. Johnsbury 0.8-1.5 3.08

Eastern Washington 1-2 2-4

Marshfield 1.0-1.5 3.85

Abercorn, Quebec 3.0-3.5 5.12

Another possible consideration for precipitation under-estimation by the WSR-88D during this event was the maintaining of the default Z/R relationship (Z=300R**1.4)for the precipitation algorithms.

A WSR-88D Operations Support Facility (OSF) Adaptable Parameter Change authorization memorandum has allowed a change in the Z/R relationship from Z=300R**1.4 to the Rosenfield-derived Z=250R**1.2 for convective "maritime" rain events. The documentation referenced several case studies from coastal radar locations during a tropical "maritime" event (tropical storm/ Hurricane), in which rainfall estimates increased to more representative values.

In this case, a tropical air mass was present over NWSO Burlington’s forecast area with precipitable water values in excess of 1.5 inches and a tropical connection noted on water vapor imagery. However, there was no tropical maritime disturbance and the tropical connection was arriving on northwest flow from Canada (continental).

Currently, we are awaiting Archive II data from NCDC to compare Z-R relationships to see if utilization of the Rosenfield-derived relationship would have improved radar estimates.

At this time, I believe a change to the Rosenfield-derived relationship would have improved estimates, but that beam blockage problems represent a larger threat to the under-estimating of rainfall amounts by the WSR-88D.

For example, KCXX rainfall estimates performed well within the Lake Champlain/St. Lawrence river basins of southern Quebec and northwest Vermont where no beam blockage exists. KCXX storm total precipitation estimates for this area were 2 to 4 inches. Observed rainfall (from Environmental Canada) across this area was 2.5 to 4 inches. Meanwhile, KCXX rainfall estimates further east near Mt. Sutton, Quebec and north central Vermont were greatly under estimated due to beam blockage (at 0.5 degrees) from the Green Mountains of Vermont. KCXX radar estimated 3 to 3.5 inches for this area with actual rainfall of 5.1 inches in Abercorn, Quebec.

USGS River Gages

The USGS has 5 telemetered river gages (3 satellite, 1 phone modem, 1 dual satellite and phone) in the flood affected area:

- Black River at Coventry (CVYV1 - phone)

- Missisquoi River at East Berkshire (EBKV1 - satellite)

- Lamoille River at Johnson (JONV1 - satellite)

- Lamoille River at Georgia (GEOV1 - satellite)

- Passumpsic River at Passumpsic (PASV1 - phone/satellite)

As shown in Table 1, rises along some of these rivers were over 3 feet/hr and over 6 feet in three hours during the height of the flood event.

Routine satellite telemetry interrogation every 4 to 6 hours at headwater points such as EBKV1 and JONV1 can allow rapid rises to go undetected between routine HADS product issuances. Meanwhile, hourly interrogation of phone telemetry like PASV1 allows for earlier detection of significant rises in river levels.

Fortunately, numerous phone calls from a variety of resources (spotters, police and Vermont Emergency Management) notified our office of the magnitude of flooding upstream from these gages and eventual receipt of HADS reports assisted in our warning operations.

We (NWSO BTV) need to continue to coordinate with USGS about setting benchmarks (specific stage height and rates of rises) for each satellite telemetered DCP river gage for the generation of automatic random transmissions.

However, due to the priority of NESDIS products on the DCP channel and possible transmission delays, it’s critically important that a dual system with phone telemetry capabilities be implemented. This would better serve our (NWS) flood warning and public safety programs.

Another option, would be to review current USGS telemetry of all gages and possibly transfer the type of telemetry based on river responses.

For example, current or proposed phone telemetry on slow responding sites such as Lake Memphremagog, Lake Champlain and the Black river at Coventry could be switched with satellite DCPs on the quick responding Missisquoi river at East Berkshire and the Lamoille river at Johnson.

Ground Truth Rainfall Reports

Rainfall in this event occurred during the overnight period, lending to limited ground truth rainfall observations until the normal observation time for our cooperative observers (7 am EDT).

Although usually an active SKYWARN spotter network, the lack of any prior notification of a possible heavy rainfall event (such as a flood watch) and the time of the event (late night) likely prevented any heightened awareness for a more active network.

The only Automated Surface Observing System (ASOS) available near the general area of rainfall, but outside the observed axis of heavy rainfall, was at Morrisville (Lamoille county) with 1.50 inches.

Vermont’s Agency of Transportation has a few automated weather stations that include rainfall data, but this data was unavailable. The acquisition and transfer of such data to NWSO Burlington continues to be vigorously investigated.

Otherwise, we severely lack an automated rain gage network in the state of Vermont. Therefore, we need to spearhead a joint state and federal initiative to develop a real-time automated rain gage network across Vermont.

Northeast River Forecast Center (NERFC) Guidance

Flash Flood Guidance issued by the Northeast River Forecast Center (valid for the 24 hour period ending 1500 UTC July 15^th) appeared to have been representative of the existing antecedent conditions.

Table 4: NERFC County Flash Flood Guidance vs. Observed Rainfall

Average Rainfall (inches) Needed Observed Rainfall

County 3 hour 6 hour 12 hour within 6-12 hours

Caledonia 2.6 2.9 3.1 2.5 - 4.5

Franklin 3.1 3.5 3.7 3.0 - 6.0

Lamoille 2.8 3.0 3.2 3.0 - 5.0

Orleans 2.4 2.7 2.8 3.0 - 6.0

Washington 3.0 3.3 3.5 2.0 - 4.0

Note: Observed rainfall pertains to portions of counties affected by axis of heavy rainfall and experienced flooding. Also, County Flash Flood Guidance implies "average rainfall over entire county", yet values are often less in mountainous terrain.

The NERFC’s River Forecast products (RVF), missed the rapid changes on the Passumpsic river at Passumpsic (PASV1) due to inaccurate initialization.

The NERFC’s initialized stage for PASV1 at 12z was at 4.1 feet (according to GOES interrogation at 0930z). Meanwhile, phone interrogation by NWSO BTV at 12z showed an observed reading of 8.21 feet.

At 15z, upon receiving RVF product with improper initialization, NWSO BTV contacted NERFC with the updated river stage information and a special RVF run for PASV1 was conducted with improved results. However, just looking at the initialization column within the updated BOSRVFBTV, it appears that the 15z stage information was not utilized, as the forecast for 18z was only 0.2 feet above the 15z reading (see Table 5).

In the afternoon, another special run was conducted by NERFC with additional QPF input for the 18-06z time frame. The additional QPF forecast for the PASV1 drainage basin was doubled what had been observed earlier in the day. This additional rainfall never materialized and the subsequent crest forecast of 19+ feet was over forecasted by 6 feet.

Table 5 (below) illustrates the different initialization stages and subsequent forecast mentioned above. Also worth mentioning was the inability of the model to depict accurate crest time. The observed crest occurred at 20z/15 (13.3 feet) while the modeled crest time ranged between 00 and 12z/16 (omitting 1842z RVF).

Table 5: PASV1 Observed and Forecasted Values (feet)

15/12z Forecast

Initial 15z 18z 00z 06z 12z

NMCRRA(GOES at 0930) 4.1

NWS BTV(phone at 12z) 8.2

USGS Readings 8.2 12.2 12.7 12.7 9.9 8.5

BOSRVFBTV at 1444z 4.1 10.5 10.9 10.8 9.8

BOSRVFBTV at 1525z 4.1 12.4 12.6 12.9 11.8

BOSRVFBTV at 1842z 12.7 (18z) 13.0 13.6 19.0

Satellite Interpretation

There were no Satellite Precipitation Estimates (SPENES) issued for this event. This may be in part due to the nature of the shower/thunderstorm activity. Several spotters reported heavy rain with infrequent lightning or thunder, indicating tropical showers with only a few embedded thundershowers. Further, infra-red satellite imagery showed warm-topped convection with WSR-88D echo top estimates consistently under 25,000 feet.

Conclusions

Flood producing rain events have become a common occurrence and are clearly the leading cause of declared presidential disasters in Vermont. Vermont’s highly variable and complex terrain are the main reason that many of our flood events are true "flash" floods.

Most flash floods occur in small, diverse basins of variable terrain. These basins include streams, brooks and small rivers which are generally tributaries to larger rivers. The larger basins are monitored by a limited number of gages along some of the rivers, but no such network exists on the smaller "flashy" waterways.

The characteristics of a "flash" flood, including the quickness and origin, requires the use of timely and accurate observing and remote sensing tools, such as WSR-88D precipitation estimates and automated rain and river gages.

Therefore, to provide timely and accurate warnings for these smaller basins, an automated, real-time rain gage network needs to be seriously considered for development and installation in Vermont and northern New York.

In order to accomplish this goal, a cooperative effort between local, state and federal agencies must occur. Some of those involved include, but are not limited to; local towns, Agency of Natural Resources (ANR), VT Agency Of Transportation (AOT), Vermont Emergency Management (VEM), FEMA, USGS and NWS (including the Office of Hydrology).

Here are a few options on how a collaborative effort could work to accomplish this goal.

1) Local towns with "flashy" waterways (Montgomery, Ludlow, Lyndonville, etc.) would be wise to invest in and maintain local warning systems such as rain gages and/or audio river level alarms. These devices could be "hard-wired" into a law enforcement office (24 hour operations and preferably with the Vermont Law Enforcement Telecommunications System (VLETS).

2) Various state agencies (including AOT, VEM and ANR) procure and provide maintenance for automated rain gages and have information relayed via VLETS or other remote system terminal to NWSO BTV for real-time use in our warning and other public safety programs.

3) VEM/FEMA purchase automated rain gages (compatible with USGS telemetry platforms) for installation on current USGS river gage platforms. Phone telemetry for precipitation data would be preferred to best utilize data and provide NWSO BTV a warning tool. USGS has tentatively agreed to provide maintenance for current and future river gage telemetry using current funding formula (50/50 share between USGS and VT ANR).

4) National Weather Service’s Office of Hydrology (OH) and Operational Support Facility for the WSR-88D (OSF) should provide funding for acquisition and maintenance of automated rain gages to study and supplement WSR-88D radar estimates. A vision of the WSR-88D Precipitation post-processing system was to acquire real-time rainfall data to compare with WSR-88D estimates and make proper adjustments in various algorithms.

KCXX has a tremendous beam blockage problem (Attachments 4,5). In fact, a representative from OSF during KCXX acceptance commented that KCXX has one of largest beam blockage problems within the continental United States. NWSO BTV (KCXX) could be a "test" site for rain gage bias and WSR-88D precipitation algorithms modifications.

More likely, I believe it will probably be a fragmented collection of options that will assist NWSO BTV and the state of Vermont in providing an improved flood warning program.