Early on January 13, 1964
January 13 of 1964 was a Monday. To help put the times in perspective, the previous evening, Sunday, was four weeks before the famous first appearance by The Beatles on The Ed Sullivan Show. Their song I Want to Hold Your Hand
was already climbing high on the charts. There were two big news stories in the northeast United States on that Monday morning. One story was the weather itself--a snowstorm, huge in areal extent and in amounts. The other story was weather related. A B-52 bomber with two nuclear weapons onboard had crashed during the very early morning hours in the mountains of western Maryland.
After completing its patrol mission(s), the B-52 had made an unplanned stop at Westover Air Force Base in Massachusetts. A fresh crew of five was dispatched on Sunday January 12 to ferry the bomber back to its home base in Georgia. The B-52 took off from Massachusetts shortly after midnight Eastern Standard Time (EST). The planned flight path (depicted on the first figure below) was northwest to Albany, New York, then southwest to Philipsburg, Pennsylvania, then south-southwest to Georgia. The cruise altitude was 31,000 feet. The path would take the plane briefly over Maryland, crossing the western panhandle about halfway between Grantsville and Frostburg. But near the Mason-Dixon line turbulence resulted in catastrophic damage to the tail of the plane. Four of the five crew members managed to bail out; two of the four survived. Much more information and memories are preserved at the Buzz 14 website (Buzz 14 being the flight's radio call sign).
I vaguely remember as a kid hearing about the B-52 crash. But on that Monday amidst the snow drifts (day off from school!) it seemed like the world ended a short distance beyond my neighbor's driveway. In the summer the mountains of Maryland were a little over an hour drive to the southeast. But in that winter wonderland the crash site in Maryland might as well have been in Siberia. I don't remember thinking about it again until I got into genealogy about 25 years ago. One branch of my family tree goes back to the early 1800s in Frostburg and points west (see my genealogy blog), and so I've repeatedly returned to explore that scenic historic area. Last May I stayed two nights at the Comfort Inn in Grantsville (actually a few miles east of the town) as a base for more exploring. Walking beyond the motel, I came upon Katie's Ice Cream stand. People travel to Katie's on Sunday evenings from far around to meet with friends. I invited myself to sit at a large picnic table by joking that it looked like the table for the elderly. The two couples sitting there had mostly finished their ice cream, so they did most of the talking while I ate my large strawberry sundae. It turned out that the one couple owns and lives on the land where the B-52 crashed. Though I missed an opportunity to ask to visit the crash site, I was more intent on getting a feel for where the four who bailed out had landed. Intriguing to me is that all four landed a few miles west of what appears to have been the final trajectory of the plane itself.
At the Buzz 14 website there is a somewhat redacted version of the Air Force accident report. The report includes weather information, but the weather charts are difficult to read since they appear to have been copied from the output of a facsimile machine. Also, the times of the charts are roughly six hours before and after the crash. I've accessed modern reanalyses of the old weather data, which provide a better perspective on conditions than what was available at the time. This particular reanalysis project has finer temporal and spatial resolution than previous reanalysis projects. The information from this one, the European Center for Medium Range Weather Forecasting ERA5 project, has become available only over the last few years for the earliest decades. Use of information from the project is governed by a Creative Commons license. Download of the gridded information is from the Copernicus Climate Data Store, specifically from their two datasets: ERA5 hourly data on pressure levels from 1940 to present; and ERA5 hourly data on single levels from 1940 to present (references at bottom of this post). I have used the python packages MetPy to process the information and Cartopy to display on maps. MetPy, in turn, relies on the python package Xarray to ingest the gridded information and Cartopy, in turn, depends in part on the python package Matplotlib to generate the plots. Before turning to the plots, I want to stress a few points by quoting from the ERA5 documentation (reference at bottom of this post; the emphasis here is mine). Data assimilation is a process whereby a model forecast is blended with observations to obtain the best fit to both the forecast and the observations, given the known uncertainties of both. The result is called an analysis (of the state of the atmosphere). ... The ERA5 data assimilation and forecasting system was used operationally for Numerical Weather Prediction in 2016.
The small table below is a reference for converting some of the available ERA5 pressure levels (millibars) to flight levels (feet).
| 250 mbar | 33,999 feet |
| 300 mbar | 30,065 feet |
| 350 mbar | 26,631 feet |
The 31,000 feet flight level is a bit above the 300 millibar pressure level. In transcripts of communications with Air Traffic Control (from the accident report), the pilot of the B-52 had reported passing Philipsburg at 0613 UTC (1:13 a.m. EST). The last intelligible transmission from Buzz 14 was at 0637 UTC, a little past halfway between 1 and 2 a.m. The two figures below are plots of wind speed at 300 millibars for the reanalysis times 0600 and 0700 UTC (1 and 2 a.m. EST). (A wind speed of 140 knots is approximately 160 miles per hour.) There are only subtle differences from the one hour to the next. On the first figure I have plotted the planned flight path. On the second figure I have omitted the flight path, but included streamlines of the wind flow.
Returning to the accident report from 61 years ago, in the weather section of the flight clearance form, in the block for turbulence the weather briefer had entered, "MDT CAT VA + N.C." That is, moderate clear air turbulence Virginia and North Carolina. Looking at the location of the flight path relative to the jet stream, it's understandable why that area was a concern. In retrospect though, there were additional areas to be concerned about turbulence. The weather analysis section of the accident report concludes with, "... A maximum wind band peaking at 145 knots existed in the vicinity of Washington, D. C. at 30,000 feet. Wind shear probably accounted for the turbulence encountered." When I first read this I imagined that the wind band referred to was one observed/analyzed 6 hours before or after the accident. But I now suspect that this statement was based on aircraft reports over DC near the time of the accident. Certainly the reanalysis, especially for 2 am EST, supports the existence of a wind band over DC. But it's not over south-central Pennsylvania, at least not at 30,000 feet, not 145 knots.
There was a snowstorm down below. Before continuing with what the reanalysis provides as a best guess for details in the vicinity of western Maryland, I want to look at the bigger picture by stepping back 62 years. The National Oceanographic and Atmospheric Administration has for many years produced a product for weather afficionados, not just meteorologists. The U.S. Daily Weather Map is actually 4 or 5 maps arranged as panels on a single sheet, with the largest panel being the national (continental US) surface map. The NOAA Library has an archive of Daily Weather Maps. The one for this Monday can be downloaded from that site by searching for the date group 19640113. It is a large pdf file, but very readable. I have cropped from the pdf a section of the surface map. The quality of this jpeg image is not as good as the pdf, but still mostly readable, at least if you click to see the image at the original size. It happens that in those years the surface map provided on the Daily Weather Map was for early in the day in the East, at 0600 UTC. So this is the surface map for 1 a.m. EST, near the time of the crash.
Not all available surface reporting stations can be plotted on a map at the national scale. But it happens that Philipsburg in central Pennnsylvania, a turning point for the path of the B-52, is plotted on this map. The temperature at Philipsburg was 8 degrees Fahrenheit. The accident report from 61 years ago provides observations for both 0600 and 0700 UTC from two surface reporting stations not on this map but closest to the crash site at about the same latitude: Morgantown, about 50 miles west, and Martinsburg, about 60 miles east, both in West Virginia. Morgantown's temperature was 15 degrees, with moderate snow at 1 a.m. increasing to heavy snow at 2 a.m. Martinsburg had 13 degrees with light snow, blowing snow and fog at 1 a.m., increasing to 14 degrees with moderate snow and fog at 2 a.m. At first glance this map might appear to be the beginning of a transition that happens often with snowstorms in the East. An initial low west of the Appalachians gives way to a rapidly intensifying low along the coast. But that happens when the surface coastal low couples with the upper air low. In this case the upper air low was still far to the west. Eventually there was a coupling, but not until about 6 hours after this time. At this time the two sea-level-pressure lows should be seen as marking the location in the lower atmosphere of a trough of low pressure extending east-west across most of the width of the map. To the north of this trough Arctic air was in place near the ground on both sides of the Appalachians.
The persistence of the east-west trough resulted in heavy snow over a broad area. There is a panel on the Daily Weather Map that summarizes precipitation, and on it snow cover is analyzed with contours for 1 inch and 6 inch depths. But the snow cover analysis is for 7 a.m. on the previous day; the snow cover analyzed on the 19640113 Daily Weather Map is actually for the previous morning. To see the consequences of this storm, it is necessary to view the 19640114 Daily Weather Map. There the snow cover analysis (for Jan 13 at 7 a.m. EST) has an area with depths greater than 6 inches covering most of Pennsylvania, and extending from there in two broad lobes, one southwest along the Appalachians and the other west to Saint Louis. But that analysis does not do justice to large areas that exceeded one foot of new snow. For example, I've focused on an area roughly inside a triangle defined by lines from Cumberland, Maryland, west to Morgantown, then north to Pittsburgh, and then southeast back to Cumberland. This area includes the crash site as well as where I grew up. I've downloaded for this area original reports from the Cooperative Observer Network. All of the stations that measured snow depth reported 12-16 inches of new snow from about noon on Sunday to about mid-morning on Monday. At lower elevations it was all new snow. At higher elevations it was on top of 6-8 inches already there Sunday morning. (The stations made their observations once a day, some in the morning, some at noon, some at 5 p.m. But they also indicated start and stop times for precipitation. The snow continued during the day on Monday, but became lighter and more off and on. The observer at Donora, south of Pittsburgh, helpfully added remarks for Sunday evening and Monday morning: 9 p.m. 6" snow, roads bad; 9 a.m. 16" snow, roads bad. At 5 p.m. Monday he measured 1.48 in. melted snow.) Such large amounts of snow over such a large area required considerable transport of moisture from relatively warm maritime areas. The Arctic air was too cold to hold much moisture, and the surface winds were too weak to transport much moisture to the west. There must have been something different happening above the surface.
Armed with that diagnosis of what had to have been, I'm now prepared to show and defend a vertical profile of the reanalysis data for a point a bit north of the crash site and near the flight path. The reanalysis information is provided on a grid with horizontal spacing of 0.25 degrees in both latitude and longitude. I've selected the point at 39.75 degrees north latitude, 79.0 degrees west longitude for this vertical profile (I'm using the phrase vertical profile to distinguish from a sounding observed by balloon-borne instruments.).
The diagram above was generated by the python package MetPy, and though the diagram is familiar to meteorologists, I'll explain since there may be non-meteorologist readers. Ignoring for the moment the red and green lines and the horizontal scale, the vertical scale on the left is pressure in millibars (aka hectoPascals). The vertical scale is logarithmic in pressure, because that makes height nearly linear. Jumping to the extreme right side, wind barbs are plotted at the pressure levels provided by the reanalysis. (I've omitted wind barbs for two pressure levels near the ground.) For now we'll focus on the winds in the upper atmosphere. The wind plotted at 300 mbar (hectoPascal) is southwest at 100 knots. Each solid pennant on a wind barb is 50 knots. (If you want to approximately locate 39.75N, 79.0W on the first map above, the one with the 300 mbar wind speeds, consider that this point is close to the flight path, and that the color bar for that map transitions from yellowish to orange at 100 knots.) According to the reanalysis there was in fact vertical wind shear at this point, from 60 knots at 350 mbar to 100 knots at 300 mbar, then to 130 knots at 250 mbar. So that's 40 knots of shear in the lower layer (100 minus 60), and 30 knots of shear in the upper layer (130 minus 100). But that is not the whole story when it comes to clear air turbulence. We also need to consider how temperature changes with height. And that requires dealing with the horizontal axis on the diagram.
The horizontal axis is labeled with temperatures in degrees Celsius. Lines of constant temperature are not vertical; instead those isotherms are skewed to the right. For that reason this diagram is called a Skew-T diagram. Temperatures in the troposphere ordinarily decrease with height quickly enough that the temperature profile still tilts significantly to the left (as it does in this profile for the layer from 700 mbar to 350 mbar; red is temperature and green is dewpoint temperature). When the temperature actually increases with height (as it does in this profile for the layer from 900 mb to 700 mb) the profile tilts significantly toward the right. The diagram being designed with skewed isotherms has the effect of fanning out
the contrast between tilts. The increased contrast extends to more subtle changes. In particular for this profile, in the upper atmosphere the tilt for the red line in the 350 to 300 mbar layer is slightly more to the left while the tilt in the 300 to 250 mb layer is slightly less to the left. These two layers are the ones discussed in the previous paragraph as the lower layer and the upper layer. Diagnosing let alone forecasting turbulence is problematic. Nevertheless, the lower layer might be more likely to produce clear air turbulence, while the upper layer, because the temperature does not cool as rapidly with height, might eventually tend to dampen whatever turbulence was propagating up from below. We know from the accident report that the pilot of the B-52, finding that the intensity of turbulence had increased to moderate at the cruise level of 31,000 feet, received permission and began a descent from 31,000 to 29,000 feet. But on passing 30,000 feet the pilot reported we're still in it
, and he transitioned to a climb, intending to go to 33,000 feet. (But it was too late.)
Continuing with the discussion of the vertical profile, and shifting attention to the lower part of the diagram, again I'll temporarily ignore the red and green lines by focusing on the winds plotted on the right. The reanalysis grids are provided at a vertical spacing of 25 mbar from 1000 to 750 mbar. To reduce the clutter I have omitted from the plot the winds at 925 and 875 mbar. These are the first and the third reanalysis pressure levels above the surface. (The model surface pressure is 932.9 mbar. The actual topography in the area ranges from about 2000 feet in the lowest valleys to a little over 2800 feet on most ridge tops.) Since the wind barbs are still difficult to read, here are the details as text. The model's surface wind is from the east at 13 knots, and basically the same at 925 mb (not plotted). The wind remains from the east at 900 mbar but starts to veer at 875 mbar (not plotted), at speeds of 42 and 47 knots. From 850 to 750 mbar the wind continues to slowly veer, overall from the southeast, at speeds of 47 to 48 knots (close to 55 miles per hour). Finally at 700 mb the speed is reduced to 43 knots (rounds up to 45 knots in the plot) and the direction has veered enough to have a slight component from the west. I know nothing about parachutes and descent rates, but I imagine that if the actual winds were something like these reanalysis winds a crewmember's landing point might be displaced a few miles northwest from where he ejected from the plane.
Before conjecturing about how the temperature and dewpoint profile near the middle of the atmosphere may have had an impact on turbulence at higher levels, I will first look at information from the ERA5 dataset titled hourly data on single levels.
One particular single level is the surface of the earth, where there are many model parameters (including the surface data plotted on the diagram above at 939.2 mbar). Some of the parameters at the surface are various forms of precipitation. The reanalysis model produces large scale
precipitation. But it is assumed there could be convective clouds as well, smaller than the model grid. Their presence is diagnosed from the large scale parameters on the 3D model grid, and then the effects of the convective clouds are fed back to the large scale. The ERA5 reanalysis provides for each hour a parameter Convective snowfall
, an accumulation over the previous hour. Below I have plotted the reanalysis convective snowfall for the hour in which the crash occurred.
This snowfall is in addition to the large scale snowfall, which in the reanalysis for this time was heavy to the northwest, and was light to moderate over a much broader area (roughly consistent with the Daily Weather Map surface analysis). In the model the convective band was moving northeast. I recognize that this is only one of many possible ways that convective bands of snow could have organized. But it is an indication from the reanalysis that the actual three-dimensional flow in the middle levels of the atmosphere might have been creating conditions for convection to thrive near the western Maryland panhandle. With that possibility in mind, it's time to turn back to the temperature and dewpoint profiles in the lower and middle parts of the vertical profile diagram.
On the vertical profile sometimes the plotted points of the green line (dewpoint) lie slightly to the right of the red line (temperature), indicating relative humidity exceeding 100 percent. Not knowing all the details of the parameterizations and adjustments and interpolations in the model, I don't know whether this is a feature or a flaw. Regardless, the message is that the reanalysis atmosphere is approximately saturated (100 percent relative humidity) from the ground to 350 mbar. In fact much of the atmosphere north, south, east and west of this point is approximately saturated through the same layer. Whatever complex 3D trajectory a particular air parcel had traveled over the previous day or two, it has now arrived at 700 mbar with a temperature of -5.6 degrees Celsius and a relative humidity of about 100 percent. We can by following lines I have omitted from the diagram (or by calculating) determine that this air parcel at 700 mb has the same bouyancy as if it had started at sea level with a temperature and dewpoint of about 53 degrees Fahrenheit. We can also determine that if this parcel were to be lifted to 350 mbar, it would follow a path close to but a little to the left of the red and green lines, and would arrive at the 350 mbar level with a temperature only about 2 degrees Celsius colder than the plotted temperature. Within the model's 0.25 by 0.25 Lat/Lon grid rectangle, pockets of air slightly warmer than the average at 700 mbar and/or slightly cooler than the average at 350 mbar would lead to convective clouds.
It would be nice to have a radar summary chart from 62 years ago. A radar site may have detected convective clouds and provided radar-diagnosed cloud tops. Lacking observations, my guess is only speculation, relying on the reanalysis scenario of the last figure. After making the turn at Philipsburg there would have been a few convective clouds below. Occasionally a convective updraft may have extended to 30,000 feet before dissipating. The updraft-generated turbulence might have propagated up to 31,000 feet. There would be a bump, then relatively smooth again. Approaching the Maryland panhandle the bumps would have become more frequent, and some of the updrafts below may have been more intense. The turbulence generated by the updrafts would have been augmented by that associated with the higher wind speed at 300 mbar. Descending toward the convective cloud tops would have been the wrong direction.
ERA5 References
Hersbach, H., Bell, B., Berrisford, P., Biavati, G., Horányi, A., Muñoz-Sabater, J., Nicolas, J., Peubey, C., Radu, R., Rozum, I., Schepers, D., Simmons, A., Soci, C., Dee, D., Thépaut, J-N. (2023): ERA5 hourly data on pressure levels from 1940 to present. Copernicus Climate Change Service (C3S) Climate Data Store (CDS), DOI: 10.24381/cds.bd0915c6 (Accessed December 2025)
Hersbach, H., Bell, B., Berrisford, P., Biavati, G., Horányi, A., Muñoz-Sabater, J., Nicolas, J., Peubey, C., Radu, R., Rozum, I., Schepers, D., Simmons, A., Soci, C., Dee, D., Thépaut, J-N. (2023): ERA5 hourly data on single levels from 1940 to present. Copernicus Climate Change Service (C3S) Climate Data Store (CDS), DOI: 10.24381/cds.adbb2d47 (Accessed January 2026)
Bell, B., Hersbach, H., Simmons, A., Berrisford, P., Dahlgren, P., Horányi, A., Muñoz-Sabater, J., Nicolas, J., Radu, R., Schepers, D., Soci, C., Villaume, S., Bidlot, J-R., Haimberger, L., Woolen, J., Buontempo, C., Thépaut, J-N. (2021): The ERA5 global reanalysis: Preliminary extension to 1950, Quarterly Journal of the Royal Meteorological Society, Volume 147, Issue 741, October 2021, Part B, Pages 4186-4227. Available from: https://doi.org/10.1002/qj.4174
Documentation web page, accessed December 2025.
Model Physical Processes web page, pdf downloaded December 2025.
