Locally heavy rains produced flash flooding over northwestern Pennsylvania and over central Pennsylvania on 20 August 1999. Most of the heavy rain fell in the morning hours. The two distinct heavy rainfall areas over central Pennsylvania can clearly be seen the WSR-88D Storm Total Precipitation (STP) product valid at 21/0435 UTC. The area to the northwest occurred the Bradford area of McKean County and the southernmost area occurred over the Mifflenburg area.

Two distinctly different heavy rainfall events affected Bradford County, separated by several hours of no significant rainfall. The first event, which occurred mainly before 0900Z, was characterized by low topped convection characteristic of the Caracean and Fritsch (1983) model. Later in the day, a north-south oriented frontal zone approached. A line of deep convection developed ahead of this line and moved over the rain soaked areas mainly after 1500 UTC. This latter event produced and additional 3-5 inches of rain in the Bradford area.

To the east, low topped convection developed over Juniata and Snyder Counties between 0700 and 0900 UTC. The heaviest rain fell later that morning, ending before 1600 UTC. Once the convection was established, cell moved from south-southeast into the heavy rainfall area. North of here, cells moved from the south-southwest to the north-northeast.

This page will show radar data relating to the heavy rainfall aspects of this event.

Archive II data were used to examine these storms from our RIDDS tapes. Storm reports came from locally collected data. A list of all available radar images is available here. Our radar archive begins at 0549 UTC, shortly after the initiation of heavy rains.

We plan to re-run the WSR-88D precipitation algorithm for comparison purposes in the near future.

Model data from our mm5 and NCEP models is archived is being made available. Our mm5 run using the Kain-Fritsch Convective parameterization scheme (KF-CPS) is data is available along with the mm5 run using the Betts-Miller convective parameterization scheme (BM-CPS). The KF-CPS is used in all of our operational model runs.

Flood damage survey pictures from the Bradford area, mainly from along Boliver creek, can be accessed here.

1. Bradford area rainfall

The nature of the convection can be seen in the 0600 UTC composite reflectivity (CR) over McKean County. The white line shows the orientation of the cross section taken through the convection. The cross section shows the low-topped character of the convection over the region at this time. The strongest echoes were confined to below 12 kft in the storms over the area. This pattern of low-topped convection continual moving over the Bradford area continued for several hours as shown in the 0716 UTC CR and reflectivity cross section (RCX). This pattern continued through 0800 UTC as shown in the CR and RCX products. The convection rapidly diminished after 0900 UTC and had ceased by 0930 UTC as shown in the 0930 UTC CR product. Note there are no significant radar echoes over McKean County by 0930 UTC.

Loops of the radar data showed (not shown) new cells forming to the south and moving northeastward over the Bradford area. The WSR-88D storm track data (shown in all the images) did a good job showing the basic track of the cells depicted by the loops.

The resulting rainfall during this time period is shown in the 0839 and the 0959 UTC one hour precipitation products. Note that over 1.4 inches of rain fell between 0739 and 0839 UTC. The corresponding STP at 0704, 0802, 0838, and 0901 UTC are shown. Please note the scales were changed to capture the heaviest rainfall amounts. These figures show that heavy rain began to accumulate (around 2 inches) southwest of Bradford around 0800 UTC. 0901 estimated UTC 2.9 inches of rain to have fallen just west of Bradford. These estimates are slightly lower than reports of 4 inches from a spotter at 7 AM and 3.42 inches at the firehouse in Bradford at the end of the first phase of heavy rain.

The Bradford are and most of McKean County remained manly rain free after 0930 UTC until the occluded front began to approach from the west. By 1859 UTC a new line of showers and thunderstorms was moving into the area again. The RCX at 1859 UTC revealed that these thunderstorms were stronger and more upright than the storm which moved over the region early in the morning. Deep cores of 50-60 dBZ echoes reached over 20 kft in these afternoon storms, suggesting conditionally unstable conditions favoring deep convection were present. Over the next 2 hours, the line made little progress eastward as new cells moved north-northeast along the line and a second line developed to the west of the first line as shown in the 2003 UTC CR product. The 2003 UTC RCX showed that storms along the second line were also quite deep.

WSR-88D precipitation estimates suggested that up to 2.8 inches of rain may have fallen just west of Bradford between 1800 and 1900 UTC as seen in the 1859 UTC OHP. An additional 2.3 inches was estimated just to the north during the next hour. The STP valid at 2003 UTC estimated 8.0 inches of rain just west of Bradford as the heavy rain wound down. An event total product, valid at 21/0435 UTC showed a similar result with the heaviest rainfall just west of Bradford.

Rain gage data suggests that around 8.8 inches of rain fell in the Bradford area. Slightly more rain probably fell to the west. The radar estimates look reasonable for the event. However, it appears that the radar estimates captured about only 75% of the total rainfall during the low-topped event (underestimates) but overestimated the rainfall during the deep convective portion of the event. This lead to very reasonable STP but for the wrong reason. Radar algorithms and spotter reports suggest hail contamination lead to the overestimation error during the deep convective phase of this event.

2. Middle Susquehanna Valley rainfall

The storm total estimated rainfall over the Middle Susquehanna Valley shows that the heaviest rain fell over Juniata and Snyder Counties. Radar imagery showed that low-topped convection training over the area produced the high rainfall amounts. Fortunately, most of the rain fell over sparsely populated portions of both counties.

Radar images showed light rain showers over the Middle Susquehanna Valley around 0700 UTC. The intensity of the rain steadily increased as seen in the 0907 UTC reflectivity image. The strongest cells were over eastern Juniata County. The corresponding OHP showed that nearly 1.5 inches of rain had fallen along the Juniata and Perry County border. Cells moving northward over the area through 1005 UTC leading to continued large hourly radar rainfall estimates. A few locations in Juniata and Snyder County were estimated to have received 1.8 inches of rain from 0907-1007 UTC.

By 1102 UTC, the rain had moved north of the flood area, however, another northbound area of rain was moving northward from the lower Susquehanna Valley. Very little rain fell over the flood area between 1002 and 1102 UTC as shown in the 1102 UTC OHP.

Around 1200 UTC, another line of showers and thunderstorms began to move into the region from the south. Between 1200 and 1300 UTC cells moved from south-southeast most of these storms and attendant rainfall remained west of the flood area. However, between 1300 and 1400 UTC, storms began to move over the rain soaked area, producing heavy rains during the hour over the flood area. Over 2 and 1.5 inches of rain fell over portions of Snyder and Juniata Counties.

Cells continued to train up the line, and move slightly to the west of north between 1400 and 1500 UTC. This produced heavy rain during the subsequent hour with over 2.4 inches of rain falling in one hour over northeaster Juniata County. Heavy rain continued for the next 2 hour as seen in the 1605 UTC OHP. Note one location received in excess of 3.2 inches of estimated rainfall (note the scale changes to capture the maximum). Fortunately, the 1605 UTC reflectivity images showed that this was the last batch of heavy rain that would move over the area. A reflectivity cross section, taken one volume scan earlier, revealed that the convection over Juniata county was not overly deep with 35 dBZ cores below 24 kft, 50 dBZ cores remaining below 14 kft, and the LEC remaining below about 10 kft in the storm.

The zoomed in STP shows in excess of 8.8 inches of rain fell over portions of Juniata County. The overall product showed a single 9.6-inch maximum in Juniata County, which was greater than the operationally available 8.4, maximum in the same region displayed on the PUP during the event. This is a curious problem and may be related to data resolution issues or precision of floating point numbers on different computers.

3. The supercell thunderstorm in Adams County

A heavy precipitation supercell thunderstorm moved across Adams County between 615 and 730 PM. This supercell produced a tornado near Littlestown and locally heavy rain. Reflectivity images from every other volume scan are shown beginning at 2237 UTC.

2237 2247 2257 2307 2317 2257 cross section

Model QPF’s showed a variety of solutions for the rainfall expected over central PA. Both the KF-MM5 run and the BM-MM5 run forecast locally heavy rains across central Pennsylvania. Our MM5 runs produced more rainfall then the operational NCEP models. The total rainfall forecast by the KF-MM5 and BM-MM5 runs is shown Figures M1 and M2. Both runs produced an area of locally heavy rainfall oriented N-S across central Pennsylvania. The KF-MM5 run produced a small area of around 5 inches compared to around 3.5 inches in the BM-MM5 run. Neither model run correctly forecast the local maximum over NW Pennsylvania or over the middle Susquehanna Valley. However, both runs suggested potential heavy rains over central PA.

The hourly rainfall over Bradford PA from the KF-MM5 and BM-MM5 is shown in Figure M3 and M4 respectively. The character of the rainfall was different in Bradford from each run. The KF-MM5 run produced locally heavy rains between 06-0900 UTC and again after 1800 UTC. Most of the rain was explicitly forecast by the mode. In the BM-MM5 run, most of the rain fell between 12-1800 UTC. Like the KF-MM5 run, the BM-MM5 run produced most of the precipitation from grid scale forcing. The BM-MM5 run produced nearly twice that produced by the KF-MM5 run in the Bradford area.

No representative meteogram is available in the middle Susquehanna Valley. Both models showed heavy rainfall just west of this area with only light rain to the east as shown in M1 and M2. The meteogram from State College showed about 2.7 inches from the KF-MM5 and 1.2 inches from the BM-MM5 run during the time of the heavy rainfall (event totals were slightly larger). Both runs had a good handle on the timing of the heavy rainfall. The convective rainfall forecast to fall between 12-1500 UTC is for the two runs is shown in Figure M5 and M6 respectively. These plan view maps show that the heavy rains were forecast to be slightly west of the observed location and that most of the rainfall was forecast to be grid resolvable rainfall.

In the BM-MM5 run, the 0.5 inches convective rainfall area in southwest PA lead to the development of a low pressure center in this area as shown in Figure M7. The KF-MM5 run, where less convective rainfall was forecast, did not have this feature and surface pressures were several millibars higher over western PA and WV as shown in Figure M8. This low level circulation difference led to huge differences in the low level moisture fields as shown in Figures M9 and M10 for the KF-MM5 and BM-MM5 respectively. In the BM-MM5 run, note the long SE to NW fetch of low-level moisture across western Pennsylvania. In the KF-MM5 run, this plume is not present and the model seems to indicate a surge of low-level moisture coming up the west slopes of the Appalachian Mountains, a feature not present in the BM-MM5 run. Clearly, the choice of CPS used has an impact on the model mass fields as can be seen by comparing these four figures. Using grid diagnostics to explain why heavy rain would fall is a slave to the CPS employed. How and where the CPS triggers affects the model mass fields and moisture fields, which in turn affect how grid diagnostics will appear.

CONCLUSIONS:

1. Two phased event in McKean County near Bradford PA.

• Early phase was low topped convection which produced around 3-4 inches of rain locally
• Second phase was deep convection associated with the occluded front. It produced around 5 inches of rain. This rain fell from a N-S oriented band with cells moving to the NNE along the band.

1. Orographic low topped event in Juniata and Snyder Counties

• Low-topped mainly during early morning hours with NNW moving cells near and south of heavy rain area and NNE moving cells to N of heavy rain area.
• A lull in the rainfall was experienced in this event
• WSR-88D STP and WATADS STP’s showed slightly different values.

1. Supercell storm in Adams County produced locally heavy rainfall and produced a tornado.
2. This was a complicated event in terms of geographic distribution of the flooding; the multi phased events over each flood area, and the added problem with severe weather. It’s a credit to those who worked this event that it was handled so well.
3. CPS’s have a dramatic impact on model mass and moisture variables

• The BM-MM5 run produced less rainfall over PA than the KF-MM5 run
• The BM-MM5 run produced a spurious low which allowed too much low-level moisture into western PA
• The KF-MM5 run showed the second surge of moisture up the western sloped of the Appalachian Mountains. This probably explains the more realistic bi-modal rainfall maximum in the meteograms, matching or radar data.