Forecasting the cold frontal passage of 12 February 1999
By
Richard H. Grumm
And
David J. Nicosia
National Weather Service
State College PA 16801
On 12 February 1999, a strong cold front raced across Pennsylvania. Temperatures fell between 10-20 degrees (Fahrenheit) with the frontal passage. Over most of western and central Pennsylvania, the frontal passage was accompanied by an abrupt wind shift and rapidly falling temperatures. Across eastern sections of the State, a narrow cold frontal rain band developed. Showers and thunderstorms marked the cold frontal passage. A few locations, mainly in the counties adjacent to the State of Maryland, experienced a few brief severe thunderstorms.
The impact on the weather of this front and the timing of frontal passage was critical to the sensible weather forecasts. Such products are provided by both private and government forecast agencies. Up through the evening hours of 11 February, the timing of frontal passage was still uncertain event. There were substantial disagreements between the operationally available forecast models produced by the National Centers for Environmental Prediction (NCEP). The impact was profound, in some locations, temperatures were forecast to remain warm all day with no threat of rain or colder temperatures until late Friday or early Saturday. Other forecasts suggest rain and falling temperatures during the day Friday.
Interestingly, as I write this, longer range numerical weather prediction models are forecasting a potential major East Coast storm on Friday. If this storm verifies, it would produce heavy snow in many locations from the mid-Atlantic region northward. As of Monday, 15 February 1999, there is agreement between all the major models that this storm will occur. Its track and intensity differ from model to model. These minor differences will have large impacts on the observed sensible weather. Many forecasters wanted to emphasize the potential of this storm in Special Weather Statements today. In light of the uncertainties with the frontal forecast of 12 February, waiting is probably the prudent position.
The goal of this paper is to show the model uncertainties associated with the cold frontal passage of 12 February 1999 and their impacts on the forecasts. The fallacy of picking the model of the day is accented herein. Unfortunately, in this one case, in terms of timing the frontal passage, there was a clear winner. It is doubtful this is true all the time and it is unlikely one could statistically pick the correct "winner" every time. WSR-88D data are used to verify frontal passage.
Model data shown include the aviation run (AVN) of the Global Spectral model (GSM) and the stepped terrain ETA (ETA).
Additional model data include the NCEP hourly model profile data. These data are used to pinpoint the timing of frontal passage in the model. Profiles from University Park, (UNV) and DuBois (DUJ) are used, due to their proximity to the central Pennsylvania WSR-88D to determine frontal passage time. Hopefully, radar data from KPBZ can be used to match to the PIT sounding.
Other data include operationally available WSR-88D archive II data collected during the event and displayed using WATADS, observational data, and all severe weather messages and statements issued and archived on AWIPS during the event.
The AWIPS data archives are made using Java programs, which extract all forecasts issued by the National Weather Service in State College and surrounding offices. These data are archived for verification and historical purposes. The Service Hydrologist and Scientific Operations Officer maintain these data and local applications software.
a. Long range forecasts
Before the event, the first model to forecast the frontal passage of 12 February was the ECMWF forecasts from 1200 UTC Sunday, 7 February. At that time, the MRF forecasts showed the frontal passage a full day earlier. Many northeast US forecasters were dealing with a minor winter storm on the 7th and may not have paid these forecasts much note. By the 0000 UTC 8 February run, the MRF moved toward the ECMWF solution with an early Friday frontal passage in western PA. The MRF and than the AVN never wavered from this forecast after the 0000 UTC 8 February forecast cycle. I only wish we had the data to show these forecasts. We do not archive MRF data.
The AVN forecasts of surface pressure, valid at 1200 UTC 12 February 1999 are shown in Figure 1. These data show the 60-h forecast from 0000 UTC 10 February, 48-h forecast from 1200 UTC 10 February, the 36-h forecast from 0000 UTC 11 February and the 24-h forecast from 1200 UTC February. The corresponding 850-mb equivalent potential temperatures and winds are shown in Figure 2. These data show how the AVN consistently placed the surface trough and the leading edge of the cold air at 850 mb in western Pennsylvania by 1200 UTC 12 February 1999. The key point is that the AVN consistently placed the front in western Pennsylvania in each successive forecast cycle.
Similar forecasts valid at 0000 UTC 13 February showed the cold front over eastern New Jersey or the western Atlantic and are not shown. However, the forecasts suggest a strong cold frontal passage during the day across Pennsylvania. Unfortunately, our data archives lack 1000 mb and surface fields such as temperatures and equivalent potential temperatures. We had them in AWIPS, and they showed the surface frontal timing close to the trough positions in Figure 1.
b. Short term model forecasts: the Eta
The ETA and NGM forecasts from 1200 UTC 10 February were the first to show the front approaching Pennsylvania. Each of the following Eta forecast cycles, valid at 1200 UTC 12 February, placed the front farther to the east as shown in Figure 3. This Figure shows the ETA 850-mb equivalent potential temperatures. Similar forecasts of the ETA surface pressure and 1000-mb equivalent potential temperatures show the trend toward speeding up the eastward progress of the surface front from to run. Clearly, using dProg/dt products would have clearly showed that the Eta was playing catch with its own forecasts. Although not shown, the NGM showed a very similar slow bias and it too played "catch-up" with its own forecasts.
The position of the front was well west of that shown in the AVN. Figure 3 shows the AVN 48 and 36-h 850 mb equivalent potential temperature forecasts (panels a & b) and the matching ETA forecasts (panels c & d). The contours are every 6K. The ETA kept the wind shift and surface front close to the 318K line. In panel C, the ETA forecast the surface front from the western shore of Lake Erie southward across west/central Ohio. Nearly 200 km west of the AVN forecast. Based on these forecasts, during the day Wednesday, a centrally prepared forecast stated "the AVN was too fast with the frontal position in western Pennsylvania. The ETA was the preferred solution." Clearly, the envelope of solutions suggested that frontal passage on Friday was well within the realm of possibilities. At this time, the NGM frontal positions lined up nicely with the ETA forecasts.
Similar forecasts as in Figure 3, but 12 hours later are shown in Figure 4. The ETA forecasts were still slower than the AVN forecasts, but faster than previous runs. By this time, the surface front in the ETA was closely aligned with the 312K contour, placing the front farther west than the leading edge of the equivalent potential temperature gradient. This was confirmed using with model sounding data from Columbus Ohio (CMH) and Pittsburgh (PIT) , Bradford (BFD) and l000 mb equivalent potential temperature plots. From these forecast, on Thursday, the same forecast center stated "prefer the faster AVN, which has been consistent run to run than the slower ETA". The "model-of-the-day" changed in 24 hours, despite the continuous consistency of the AVN prior to this time.
A list of all central Pennsylvania hourly profile data can be obtained here. Data from PIT, CLE, and ILN County warning areas (CWA’s) are not available on this server. These data can be used to determine the approximate time of frontal passage. Sample radar data, used to verify the frontal positions can be independently examined by clicking here.
An examination of Bradford (BFD) wind profile forecasts from 0000 UTC 11 February through 0000 UTC 12 February are presented. The 0000 UTC 11 February winds and temperature time sections show that the model forecast the colder air to reach BFD around 2300 UTC on 12 February. The forecast from 12-hours later showed an increase in the speed of the front as forecast by the wind and thermal field, with frontal passage forecast around 21 UTC. A mere 10 hours before frontal passage, the 0000 UTC 12 February forecast cycle still had a slow bias as shown by its 1600 UTC frontal passage forecast at BFD in the wind and temperature profiles. At BFD, this forecast was significant due to the potential snow.
The frontal passage sequence at State College and DuBois are important because these data can be compared to WSR-88D observations. However, from a weather perspective, the frontal passage in the southern part of Pennsylvania was critical. It was in this region on 12 February, the all-time record high in Harrisburg was crushed and severe weather was observed in the area as the front moved through.
The ETA wind and temperature profiles from the 0000 UTC 11 February 1999 forecast cycle showed no sign of the cold front during the day Friday. The model did forecast surface temperatures around 16C, about 7C lower than observed. By the 1200UTC, the wind and temperature profiles showed the frontal passage around 0300 UTC 13 February. As time drew near, the 0000 UTC 12 February run still kept the frontal timing too slow as seen in the wind and temperature time sections. The forecasts did show the convective rains arriving around 1900 UTC and the frontal passage around 2200 UTC. The model placed most of the convective rainfall in the warm, unstable air ahead of the surface cold front.
All available WSR-88D data can be viewed by clicking here. The frontal passage at Bradford was captured by the radar around 1433 UTC. The corresponding velocity data also shows the approximate frontal passage too. A similar sequence shows reflectivity and velocity data for the frontal passage near DuBois shortly before 1438 UTC.
The thunderstorms, which developed along the southern flank of the cold front, produced severe weather in southeastern Pennsylvania. The strongest cells along the line developed around 1754 UTC. The corresponding velocity data showed the surge of westerly winds, associated with the front, with local wind maximum near the individual thunderstorms along the line. These thunderstorms strengthened as they moved eastward, into the unseasonably warm air east of the mountains, at 1844 UTC. The strongest cell was over Adams County, northwest of Gettysburg as seen in the velocity data. As this line moved eastward, the narrow cold frontal rainband grew and a wide cold frontal rainband formed to the west as shown in the reflectivity data at 1935 UTC. The velocity data showed that the strong cold frontal winds were from the northwest. The large-scale view of the reflectivity data showed the narrow cold frontal rainband and several bands of more stratiform snow to the west in the cold air. This imagery shows a classic anafront type circulation, with the precipitation along and behind the surface frontal position.
