The Use of High Resolution Hourly Forecast Soundings for the Prediction of Lake Effect Snow

Thomas A. Niziol* and Edward A. Mahoney
NOAA/National Weather Service, Buffalo, NY, 14225


In conjunction with the ongoing Lake-effect snow project being conducted by the National Weather Service, a user-friendly and highly interactive computer application has been developed that uses high resolution numerical weather prediction data to enhance the local forecasts of mesoscale snowstorms near the Great Lakes. The data sets are in the form of hourly forecast soundings from the National Centers for Environmental Prediction's Eta model. They are interpolated to many sites around the Great Lakes. An interactive software package called BUFKIT, developed at NWSFO Buffalo, streamlines the process of assimilating these data sets for operational use in the real-time forecast environment. The new data sets are extremely well tailored for the prediction of lake-effect snow because they provide the high degree of temporal and vertical resolution necessary to evaluate such mesoscale events.


The NWSFO at Buffalo, NY has been involved in research applications toward the improvement of lake-effect snow forecasting for some time. Lake effect snows are particularly important winter phenomena to a meteorologist because of the enormous amounts of snow that often occur over localized, and sometimes heavily populated, regions downwind of the Great Lakes (Sykes 1966, Niziol 1989). These severe winter storms develop as a result of very cold air passing across the relatively warm waters of the lakes during the late fall and winter months. The flux of heat and moisture from the lake creates convective instability in an otherwise stable airmass. The prevailing wind helps to channel the moisture plume and subsequent precipitation in narrow bands downwind of the lakes. Orographic features tend to enhance the precipitation process.

The storms are truly severe convective weather events that disrupt travel and commerce and at times can be life threatening. The scale of lake snows span the mesogamma (2-20 km) to mesobeta (20-200 km) as classified by Orlanski (1975). Although the storms are roughly on the horizontal scale of summertime convection, they are confined to a much shallower layer in the lower atmosphere than the typical summer thunderstorm. The mature storms frequently develop into steady-state features, often existing for prolonged periods of time over one location until a change in the overall synoptic pattern modifies the environment around the storm. For this reason the evolution of these mesoscale events can be forecast with some degree of accuracy, if the synoptic scale conditions are well predicted.

Over the years, Great Lakes forecasters have developed a series of rules and identified a number of parameters important to snowband evolution (Rothrock 1960, Hill 1971, Niziol et al. 1995). In the past only moderate success could be obtained with data sets that had neither the temporal or spatial resolution to predict short term evolution of the snowbands (Kolker 1977). More recently forecasters at Buffalo collaborated with John Jensenius at the Techniques Development Laboratory (TDL) to evaluate high resolution hourly model soundings for select points around the Great Lakes (Jensenius 1989). The data sets proved their worth almost immediately. In the first year of their use forecasters found that the data better delineated the timing and evolution of lake-effect snow (Niziol and Levan 1993). During the past two years these data sets have been ranked by forecasters as the top forecast tool in Eastern Region offices for the prediction of lake-effect snow (Carter et al. 1996).


BUFKIT is a software package that incorporates the most important forecast parameters to predict snowband location, duration and movement over very localized areas. The software is a critical step in the forecast procedure because it allows the forecaster to process and assimilate a large amount of data quickly and coherently. The parameters and analyses, specifically designed for lake-effect forecasters, display features such as inversion height, wind direction and shear, relative humidity, looping of hourly soundings, and maps that outline location of snowbands (Mahoney and Niziol 1997). The ability to display time sections of the parameters show the forecaster trends in the data. These trends often feed directly into the actual written forecast to the public. BUFKIT can also reformat and display a sounding with SHARP (Skewt/Hodograph Analysis and Research Program) to further evaluate the convective potential for lake-effect snow. The soundings provide an excellent overview of the thermal, moisture, and kinematic structure of the atmosphere within lake snow events. Finally, the size of the data files are very small and manageable, allowing for quick and easy data transfer. This is a very important and often overlooked requirement for operational forecast offices that, for the most part, have a limited bandwidth available for data transfer.


During 9-10 December 1995 a major lake-effect snow event occurred over the eastern Great Lakes. A record 39 in. (100 cm) of snow fell during a 24 hour period at the Buffalo airport (Fig 1). The metropolitan area was brought to a standstill, while to the east of Lake Ontario similar amounts of snow fell at many locations.

The overall synoptic scale pattern responsible  for the storm featured a deep 500mb trough over Hudson Bay that put the eastern Great Lakes in an area of strong cyclonic vorticity advection, very well aligned southwest winds, abundant low level moisture, and most importantly, a deep layer of arctic air.

The event began shortly after a sharp cold front crossed Western New York during the early afternoon of Saturday, 9 December. Narrow bands of squalls developed off Lake Erie on a 200o axis across the Niagara Peninsula of neighboring Ontario Province during the afternoon of 9 December, then consolidated into a single intense band of snow about 10 miles wide that was centered over metro Buffalo by that evening. The band continued to drift slowly south during the evening, settling across the suburbs immediately south of Buffalo. From an operational forecasting point of view it is extremely important to note that the band only moved about 10 miles during this time period, but the result was that an entire different segment of over a half million people was receiving heavy snowfall.

Overnight, the low level flow backed a bit to allow the band to drift north again into metro Buffalo and its northern suburbs. The band remained stationary through most of the morning and afternoon of Sunday 10 December (Fig 2). During a 12 to 14 hour period,snow fell at rates as great 4 in. (10 cm) an hour at times. The cutoff of heavy snow was dramatic with totals increasing from 1 to 30 in. (3 to 50 cm) over about a 10 mile area! Finally, during the evening of 10 December the SW/NE oriented band began to move south in response to strongly veering winds aloft. Throughout the rest of the night the snowband weakened considerably in response to the passing of the upper level trough and a shift in the upper winds to west and northwest. The significant directional shear between low and upper level winds eventually led to multiple bands of snow that continued through the night of the 10th and into the 11th across the area well south of Buffalo.


The Eta model soundings had just become available to the forecast office and were undergoing their maiden operational use. Until that time the NGM model soundings were used exclusively. This left forecasters a bit apprehensive about the data but they soon realized that the model data was verifying well. The data sets were used in conjunction with satellite imagery and the Doppler radar which had also just come on line the day before the event occurred. Real-time reports from the local snow spotter network detailed snowfall rates that rounded out the suite of new technology.

The Eta model soundings predicted the evolution of the lake-effect storm very well, within an hour or so of the actual development and subsequent weakening of the storm. Although the data sets were very new, forecasters recognized that the wind forecasts at select levels provided excellent information on short term location and movement of the snowbands. It took some modification of forecast techniques to adjust to the new data sets. Forecasters had been using techniques built on old data sets. Low level wind and temperature forecasts in the past were pretty much tied to the 850mb level, because that was all the information available to the forecast office (McVehil 1965). More recently NGM hourly soundings provided more highly detailed information but only 16 levels were available in the NGM vs. 33 levels in the Eta (Niziol and McLaughlin 1992). The new forecast products provide such an overwhelming amount of data that it took some time for the forecasters to decide which levels provided the best information. Once the proper levels were chosen, the wind forecasts in particular provided highly accurate trends in the location and movement of the mesoscale snowbands.

Figure 3 shows an example of how BUFKIT software displayed the critical model sounding data to the forecaster for this event. Time sections provided hourly wind directions, inversion height, and lake indices which were interpreted directly into the actual zone forecast. At the click of a button, each hourly wind direction also displayed a corresponding map that indicated where snow would be likely. The plotted variables correctly identified the beginning of the event, the drift south across the city of Buffalo on the evening of 9 December, the slight drift north back into the city late that night, and the extended period of 240-250o steering winds that led to a stationary band over Buffalo for a 12 to 14 hour period through 10 December. Finally, the combination of low level and mid level wind forecasts (surface through 700mb) indicated that the snowband would move south and weaken during the night of 10 December.

Forecast discussions provided highly detailed accounts of the mesoscale outlook for lake snow. One such account issued early in the morning of 10 December noted the heavy snowfall potential for Buffalo. It read "using level 8 wind direction from the Eta paints a grim picture for the city of Buffalo as winds remain 250o until this evening..."

Short term forecast or NOWCASTS also relied on the trends in the time sections from the hourly soundings to confirm what satellite, Doppler radar and the local snow spotter network indicated. As snowbands oscillated in one direction or the other, forecasters interrogated hourly wind direction trends from the latest` available model run to pinpoint snowband location and movement during the next few hours. When dealing with such extremes in snowfall over a very localized yet heavily populated region it is just as important to tell people where it will not be snowing as where the heavy snow will occur. The forecasters were able to provide highly detailed forecasts to the public without causing undue panic in areas where snow was not likely to fall.


The analysis of hourly forecast soundings from the Eta model with BUFKIT software is an outgrowth of years of experience and suggestions from the operational forecast staff at Buffalo. Lake-effect snows are unique mesoscale events that up to very recently have not been predicted well by synoptic scale models (Niziol 1987). However, because of the relatively steady-state nature of these mesoscale events, it has been shown that certain forecast parameters from the models can be used to provide valuable guidance about the evolution, location and movement of snowbands downwind of the Great Lakes. We have "stretched the envelope" about as far as possible with this type of forecast process. Current operational research efforts are concentrating on realtime mesoscale numerical models that will hopefully predict not only the location and movement of such features, but accurate snowfall amounts as well. Until those models are shown to improve on current forecast procedures, BUFKIT will be an integral part of the forecast process for lake-effect snow.


Carter, G. M. , Salem, T. L. Jr., and M. A. Wool, 1996: The use of new technology for enhanced snow forecasts in the vicinity of Lakes Erie and Ontario. Preprints, 15th Conf. On Wea. Anal. and Forecasting, Amer. Meteor. Soc., Norfolk, VA., 567-572.

Hill, J. D., 1971: Snow squalls in the lee of Lake Erie and Lake Ontario. NOAA Technical Memoranda NWS ER-43. NOAA/NWS, Bohemia, NY, 20 pp, [NTS No. COM-72-00959].

Jensenius, J. 1990: Personal communication. NOAA/NWS/TDL, Washington, D.C.

Mahoney, E. A., and T. A. Niziol: BUFKIT: A software application toolkit for predicting lake effect snow. Preprints, 13th Intl. Conf. On Interactive Info. and Processing Sys. (IIPS) for Meteorology, Oceanography, and Hydrology, Amer. Meteor. Soc., Long Beach, CA. (elsewhere in this volume).

McVehil, G. E., and R. L. Peace Jr. 1966: Project Lake Effect. A study of lake effect snowstorms. Final Report Contract CWB-11231, Cornell Aeronautical Laboratory, Inc., 53 pp. [Available from Calspan Corp., Genesee St., Buffalo, NY, 14225.

Niziol, T. A. 1989: Some synoptic and mesoscale interactions in a lake effect snowstorm. Postprints, Second NWS Winter Weather Workshop, Raleigh, NC., 26-30 September 1988, NOAA Technical Memorandum NWS ER-82, 260-269 NOAA/NWS, Bohemia, NY.

_____, and J. M. Levan, 1992: Forecast procedures for mesoscale lake induced snow events in the eastern U.S., Preprints, Symposium on Wea. Forecasting, Amer. Meteor. Soc., Atlanta, 202-209.

_____, and S. F. Mclaughlin, 1992: The use of high resolution data from the Nested grid model for the prediction of lake effect snows. Postprints, Third NWS Winter Weather Workshop, Portland, OR, 15-18 September 1992, NOAA Technical Memorandum NWS WR-218, 179-192, NOAA/NWS Salt Lake City.

_____, Snyder, W. R. And J. S. Waldstreicher, 1995: Winter weather forecasting throughout the Eastern United States. Part IV: Lake effect snow. Wea. Forecasting. 10, 61-77.

Orlanski, I., 1975. A rational subdivision of scales for atmospheric processes. Bull. Amer. Meteor. Soc., 56, 527-530.

Reinking, R. F., R. Caiazza, R. A. Kropfli, B. Orr, B. E. Martner, T. A. Niziol, G. P. Byrd, R. S. Penc, R. J. Zamora, J. B. Snider, R. J. Ballentine, A. J. Stamm, C. D. Bedford, P. Joe, A. J. Koscielny, 1993: Lake Ontario winter storms (LOWS) project. Bull. Amer. Meteor. Soc., 74, 1828-1849.

Rothrock, H. J., 1960: An aid in forecasting significant lake snows. ESSA Technical MemorandumWBTM CR-30, NOAA/NWS, Kansas City, MO 18 pp.

Sykes, R. B., The blizzard of '66 in central New York: Legend in its time: Weatherwise, 19, 241-247.

Back to NWS Buffalo Research Index | Back to NWS Buffalo Home Page