During the morning hours of Sunday, 07 February 1999 the precipitation ahead of a winter storm reached southwestern Pennsylvania. The precipitation started as snow or a mixture of snow and rain in most locations. By late morning, a band of moderate snow developed across north central Pennsylvania. This band produced 2-4 inches of snow across Columbia and Sullivan Counties during the afternoon hours. More snow and sleet arrived during the evening hours bring storm total snowfall to around 7 inches in the Millville-Muncy-Williamsport and LaPorte areas.

This was a marginal winter storm. The days preceding the storm were unseasonably warm. As early as Wednesday, 3 February, medium range forecast models began hinting that a storm would move across the Ohio Valley late Sunday into early Monday. These models suggested the atmosphere might be cold enough to produce snow across the northern third of Pennsylvania.

It was not until Friday evening that this storm began to be forecast by the short term forecast models (0-2 days). The morning run of the aviation model, which is the forecast model of the Global Forecast System which suggested snow, sleet and rain were likely late Sunday. This model produces forecasts out to 3 days.

By Saturday, all the available models were forecasting a storm to move through the Ohio Valley, across Pennsylvania and then re-develop along the Mid-Atlantic coastal plain. The models forecast marginal conditions with an intrusion of warm aloft and boundary layer temperatures barely cold enough to support snow. Worse, there were conflicting forecasts about the exact track and thermal structure of the atmosphere.

Ultimately, no single model properly forecast all aspects of this storm. The Eta models (32-km versions) properly forecast a strong frontal circulation over the State. This circulation probably helped to produce the low-level cold air that supported snow and sleet. However, a warm layer aloft extended farther north than indicated in the forecast model. This lead to more mixed precipitation farther north, which limited snowfall, amounts.

A more significant error was the timing error. All the models were too slow to move the fast moving surface cyclone eastward. At times, the model forecasts were 6 hours too slow. In addition to this, the models were too slow in forecasting the secondary low development to the lee of the Appalachian Mountains. This low developed 6-7 hours earlier than forecast and about 200 km farther to the south and west.

This page will look at some of the features and forecast problems associated with the winter storm of 7-8 February 199. Specially, we will show the forecasts available and some of the errors and low confidence factors they contributed to the forecasting of this event. Radar data will be used to explain the snowfall the problems associated with verifying winter storm warnings and advisories.

1. Precipitation

 

These images are under construction and should be available by 10 February.

 

Our first cut snowfall map is included. This map shows that the snowfall was confined to the northern two thirds of the State. Light snow amounts (yellow) extended from New Holland in Lancaster County to near Renovo in Clinton County. Light snow amounts were also observed in the mountains of southern Pennsylvania to include the Laurel Highlands and South Mountain.

 

The heaviest snow was observed over northeastern-central Pennsylvania.

 Radar Observations

 

The first radar image shows the precipitation moving across Pennsylvania around 1720 UTC (1200 Local). The intense returns (yellows) over southwestern Pennsylvania are the result of bright banding. In this region, the precipitation was falling primarily as sleet and freezing rain. Across northern Pennsylvania, smoother green echoes are indicative of light snow, which was being reported in these locations at this time.

 

The area of mixed precipitation continued to move slowly northward. By 1837 UTC, a small band of moderate snow had developed in the larger snowband in northern Pennsylvania. This band took on a double structure. By 1900 UTC, the southern band extended near Millville in northern Columbia County westward into Lycoming County. The northern band remained across Potter and Tioga Counties. These to bands persisted for several volume scans and could be seen at 1911 UTC, It is interesting to note that the southern precipitation area, which had moved northwestward, appeared to weaken. A portion of this weakening was the result of the band moving into colder air where most of the precipitation fell as snow. The lack of bright banding (melting) likely contributed to the weaker returns.

 

A series of zoomed in images from 1825, 1831, 1842, and 1848, show the evolution of the smaller bands into a narrow large band by 1929 UTC. These data suggest there should have been some locally heavy snow across northern Clinton and southern Potter Counties and probably southwestern Tioga County as well. Unfortunately, no reports were available from these regions. Recalling the snowfall map, there was actually a snowfall minimum in these data lacking regions.

 

2. Surface data

A manually analyzed surface chart valid at 2300 UTC 7 February is shown. This is one of the maps, which was analyzed during the event. At a later point, post event maps will be provided. This map shows the surface low over the panhandle of West Virginia and a secondary low developing along the coastal plain of North Carolina. The precipitation bands on radar were observed close to and to the left of the analyzed troughs.

By 0000 UTC 8 February two lows were apparent in the data along the coastal plain. Although unreadable, the observation site in northern West Virginia, in the narrowing warm slot was at 58 F.

Using these analyses, it will be shown that the models failed to forecast the rapid eastward progression of the primary low as well as the rapid development of the coastal low.

1. Aviation Model Forecasts- coming soon

 

Aviation model forecasts suggested the potential for a winter storm several days in advance. A series of forecasts of the surface low from valid at 08/0000 UTC is shown along with the 850-mb heights and isotherms. These forecasts, the first one from 05/0000 UTC, showed a surface low tracking across the Ohio Valley. Model 850-mb temperatures, as first approximation of the rain snow line, showed the potential for snow in north-central Pennsylvania.

Forecasts valid at 08/0600 UTC of both the surface pressure field and the 850 heights and temperatures also showed a consistent forecast. Again, all forecasts were selected to be valid at the same time, showing the relatively consistent forecast.

We hope to obtain MRF model forecasts to determine how far in advance the signals were present in the forecasts about this potential winter storm.

 

2. Eta Model forecasts

 The Eta model forecasts from 0000 UTC 06 February began to show the potential for wintry weather across Pennsylvania. The model runs from 0000, 0300, 1200, and 1800 UTC 6 February began converging on a potential winter storm across Pennsylvania and southern New York State. Unfortunately, detailed forecasts of the thermal profiles showed elevated intrusions of warm air, which placed the type of precipitation in doubt. It was not until the forecasts from the 7^th that the models began to show a potential for significant snowfall across northern Pennsylvania and southern New York. Model runs from as late as 1800 UTC on 7 February still showed an intrusion of warm air up to the New York border. Critical to these thermal profile forecasts were forecasts of the low-level surface cyclone.

The 07/0000 UTC Eta mean-sea-level pressure forecasts are shown. These data show 6-h forecasts beginning with the 12-h forecast valid at 07/12000 UTC. This forecast cycle placed the surface low over southern Ohio by 08/0000 UTC. The forecasts implied a redevelopment of the surface low over Maryland around 08/0600 UTC with a secondary low along the coast at this time. As already shown in the surface analysis, this surface low develop a full 7 hours earlier and nearly 500 km farther south.

The model quantitative precipitation forecasts (QPF's) from this forecast cycle are also shown. These panels show the 6-h forecasts valid from 07/1800 UTC through 08/1200 UTC. These forecasts show that the model forecast the precipitation to arrive in western Pennsylvania by 1800 UTC. The heaviest amounts were around 8 mm in southwestern sections of the State. The heaviest, Statewide, precipitation was forecast to occur between 0000 and 1200 UTC, overnight Sunday into early Monday.

Forecasts from the 07/0300 UTC forecast cycle are also shown. Due to the 3-h forecast information available, the forecast lengths are slightly different. Although the forecasts were quite similar, especially with the timing and placement of the secondary low over Maryland and Virginia. The intermediate 08/0300 UTC forecast showed no improvement. The associated QPF's are also shown. These forecasts are similar to the previous 07/0000 UTC forecasts with the exception of the slightly increased precipitation amounts forecast over southwestern Pennsylvania.

Forecasts from the 07/1200 UTC forecast cycle are also shown. These forecasts are quite consistent with the 07/0000 UTC run with respect to the track, intensity of the primary low moving across the Ohio Valley. The forecasts of the secondary low are also quite consistent. The associated QPF's are also shown. These forecasts are consistent with those from the 07/0000 UTC forecasts, especially with the placement of the 8 mm isohyet at 08/0600 UTC showing the implied "band" of heavier precipitation across northern Pennsylvania.

The last Eta model forecast available to be of use during this event was run at 08/1800 UTC. By this time, snow, sleet, freezing rain, and rain had been observed over most of western and central Pennsylvania. The guidance provided by this run, were of use to update the forecasts. All warnings and advisories were in effect by the time these data became available. The surface forecasts showed a slightly faster eastward position of the surface cyclone, over western Pennsylvania by 08/0000 UTC. However, it failed to show the secondary cyclone in North Carolina, which developed around 07/2300 UTC, developing this cyclone around 08/0600 UTC.

The associated QPF's with the 08/1800 UTC Eta were consistently consistent with previous forecasts. This failure to capture the secondary redevelopment and move the precipitation eastward affected the local forecasts and the forecasts to the east. A similar problem developed during the 14 January 1999 snow and ice storm. During that event, the models failed to capture the cold air damming and rapid cyclone development along the coastal plain. This cut-off the warm air and produce more snow than the anticipated sleet and freezing rain.

All available model profiles can be found by clicking here.

These data are from 06/0300 and 07/0300 UTC and show that an elevated mixed layer was forecast to remain above the boundary layer. This caused the model forecasts to produce a wintry mixture from Harrisburg (CXY) to Elmira, NY (ELM) in these forecasts. Similar profiles can be found by browsing the directory. A special thanks to Robert Hart of the PSU for providing these valuable data.

Eta model forecasts were converted to vis5D for display purposes. Using vis5D, the forecaster can clearly see where the model forecast rain snow line is. Unfortunately, in this case, the Eta forecasts had a slight warm bias in the boundary layer. This led to warm surface temperatures and a shallow surface based warm layer. This layer actually existed during the event as evidenced by the preponderance of reports of heavy wet clinging snows in northern Pennsylvania.

The vis5D 0^oC isosurface from the 07 Feb 0000 UTC forecast cycle is shown in the following images. The image valid at 0000 UTC 7 Feb (07/0000 UTC hereafter) shows the elevated warm layer and the more northerly surface based warm layer over Pennsylvania. The isosurface has been made semi-transparent to allow the user to see the map. The topography is turned off in most images. By 07/0600 UTC, the models strong frontal circulation led to significant cooling and below freezing temperatures across northern Pennsylvania. This cooling continued until around 07/1800 UTC. In this figure, it is interesting to note that the elevated warm layer is about 100 km south of the surface based warm layer and the intrusion of more warm air aloft from the southwest.

Snow was falling across northern Pennsylvania at this time, with freezing sleet and freezing rain across most of central Pennsylvania. Surface temperatures were several degrees below 0^oC over most of the region. However, warm air aloft was melting the snow. Clearly, the Eta forecasts were too warm in the lower levels. A view from the east of the 0^oC isosurface is also shown. At the surface, the 0^oC isosurface just cross the NY border. Aloft, the nose of the elevated warm layer aloft can be seen jutting into north central Pennsylvania. The isosurface is completely opaque in this image for clarity. The model did a relatively good job capturing the elevated warm layer, thus the problem with sleet and freezing rain across Pennsylvania. The surface based warm layer was a serious error.

The vertical velocity (0.7 cms-1) isosurface and the 305K isosurface valid at 1800 UTC are shown from above and from the east. These figures show the slope ascent up the isentropes and capture the sense of the elongated east-west front with its attendant cold wedge to the north. The N-S frontal slope is quite evident in this figure. At and around this time, the heaviest snow fell in northern Pennsylvania.

The first forecasts mentioning the potential for snow, rain, and sleet across central Pennsylvania on Sunday, 7 February were issued in the evening hours of Friday, 5 February. By 4 AM Saturday, 6 February, special weather statement began to be issued relating to the potential winter storm on Sunday into early Monday.

Around 4 AM Sunday morning, a winter storm warning for heavy snow was issued for the northern tier counties of central Pennsylvania. Early in day Sunday, several winter weather advisories for snow, sleet, and freezing rain were also issued.

During the event, spotter reports of snow and sleet were sent out in public information statements. These statements are of value to the broadcast weather and emergency management communities as well as weather enthusiasts.