Superstorm ‟93 and Related Articles
This list provides a bibliography and summarization of articles about Superstorm ’93 and
directly related to the storm, forecasts of the storm, and the storm’s formation and
(1) Overview of the 12-14 March 1993 Superstorm
(Kocin, Shumacher, Morales, Uccellini)
Bulletin of the American Meteorological Society; Vol 76; 02/95; pp. 165-182
The paper gives background and some detailed information on Superstorm ’93, and
provides a basic understanding of the meteorological events of the storm. The paper goes
through three different looks at the storm, including effects and impacts, a synoptic
overview of the surface and upper level atmospheric conditions for each day, and a
mesoscale discussion of the formation of the cyclone over the Gulf of Mexico.
Effects and Impacts: pp. 165
- about 100 directly related deaths (media reports to 270)
- $2 billion in damage (1.6 in Florida)
- Record low sea level pressures
- Widespread snowfall, 90 million people with 1 in. + of snow
- Record low temperatures across much of southern U.S. (2 degrees F Birmingham)
- Most extensive distribution of heavy snow across the eastern U.S. in modern
Synoptic Overview: pp. 167-176
- 12 March: Thunderstorm development Texas/Mexico border
- Surface low pressure over northern Mexico
- Developing 850 mb circulation at 1200 UTC
- 30 m/s increase of upper level jet streak downstream (Ohio Valley)
- 13 March: Coastal flooding in Florida, Squall line moving through
- 971 mb pressure by 1200 UTC – 29 mb drop in 24 hours – not unheard of
- Massive circulation around cyclone (1/4 of U.S)
- Heavy snows up to Pennsylvania/New Jersey
- Circulation center moving northward into NC/DE
- 14 March: Lowest pressure at 0000 UTC (960 mb)
- Storm track west of urban corridor, less snow in biggest cities, some rain
- Blizzard condition west of low pressure center (New York State)
- Exits U.S. by 1800 UTC Mesoscale Discussion: pp. 176-181
- Two cold fronts, leading front small temp. gradient, trailing front brought coldest
- Squall line developed along two troughs in Gulf, moved over Florida then out to
sea, 11 tornadoes, 47 deaths.
- 1.5 mb/h deepening – much faster than expected by models (due to latent
(2) Performance of the NMC Global Medium-Range Model
Bulletin of the American Meteorological Society; Vol 76; 02/95; pp. 201-212
The second part of the papers from NMC gives an analysis of the model accuracy for the
storm. The paper goes into detail about the accuracy of the forecast 4-5 days ahead of the
development, and critiques what the MRF model did well and did not do well. The paper
explains that the forecasts along and to the north of the storm track were excellent in
terms of strength and position of the system, while the forecasts over the Gulf of Mexico
were consistently too weak. An outline of the good and bad parts of the models:
Forecast of Mean Sea-Level Pressure
- MRF consistently underestimated the rate of deepening from 0000 to 1200 UTC
13 March (storm over Gulf of Mexico)
- Later deepening was anticipated well and even overestimated for the period of
1200 UTC 13 March to 0000 UTC 14 March
- Storm position and intensity at 0000 UTC 14 March well predicted from 9 March
runs on. See maps pp. 202-203
- The European Centre for Medium Range Weather Forecasts (ECMWF) model
and UK Meteorological Office (UKMO) models (UKMET) were compared to the
MRF, ECMWF model showed lower central pressure starting in initial 8 March
runs for 1200 UTC 14 March timeframe, while MRF and UKMET were not as
- At 4 & 3 days out, ECMWF predicted 20 mb lower central pressure. WHY?
- POOR INITIAL CONDITIONS over the Gulf of Mexico were used in MRF
and UKMET models, and ECMWF tends to overintensify large cyclones.
- Poor initial cond: Also see pp. 201
- Ensemble prediction on 0000 UTC 10 March shows models in agreement for
major East Coast storm.
Middle Troposphere Forecast
- Rapid change in global jet stream pattern was predicted by mean sea level
pressure (MSLP), with a major low latitude trough over southeastern U.S. (500
- Change in circulation pattern was well predicted, but the speed of the trough and
the rate of intensification were underestimated, and the Interaction of two short
wave troughs was not resolved.
- The trough troughs were shows as a double low pressure center in the forecasts.
- There was high skill involved with the forecast (highest in 8 years) possibly due
to magnitude of anomaly.
- 72, 60, and 48 h forecasts produced center too weak and frontal trough too slow.
- No forecast of the 35-40 m/s winds that hit Florida.
- Models showed the 540 dam line well from southern Apps into Virginia, with
good rain/snow line, but may have greatly underestimated precipitation amounts
from Apps to the east.
Forecasting of the 12-14 March 1993 Superstorm (3)
(Uccellini, Kocin, Schneider, Stokols, Dorr)
Bulletin of the American Meteorological Society; Vol 76; 02/95; pp. 183-199
The third paper, this article refers in depth to the success and failure of the forecasts for
Superstorm ’93. The paper goes through the various time frames, and the accuracy of the
forecasts generated by the Meteorological Operations division of the NWS. The lack of
accurate forecast of the deepening of the central low pressure over the Gulf of Mexico is
examined, and the increasing accuracy as the development neared is noted. The article
then goes into the details of how the MOD forecasters were in contact with local NWS
offices, and how these offices issued statement/watches/warnings to the public. The
details of the manual forecasts based on the model guidance is critiqued throughout the
paper, although the authors are quicker to blame the models for forecasting problems than
the forecasters, possibly due to biases as forecasters themselves. The words success and
shortcomings are used, but not failure to represent the way the forecasts are examined.
- Early skepticism of major cyclogenesis due to cyclones earlier in season being
weaker than forecast by models, skepticism faded as MRF showed consistency in
forecasts over multiple runs.
- COMMENT: Forecasts by humans often depend on how well recent forecasts are,
i.e. forecasters are weary of predicting “big one”.
- 3 days before event, model differences were significant. The UKMET and
ECMWF showed the low pressure center further north at development than MRF.
- UKMET and ECMWF alternated between far inland (Lake Ontario) and off the
coast for low pressure center for later track, which concerned forecasters because
these models had correctly had cyclones west of Appalacians earlier in the year.
- MOD followed the MRF because it was consistent and also showed the system
being associated with increasing jet streaks over Ohio Valley and amplification of
upper-level trough/ridge pattern, both of which are precursors for major East
Coast Storms. (see pp. 188)
- MOD confidence in the track/intensity of the storm was assisted by an ensemble
forecast on 10 March of 14 MRF forecasts, and by a new MRF-based statistical
package that had the conditional probability of snow (CPOS).
- CPOS was high: 50% for cities N and E of Washington DC and 85% for West
- The failure of the forecasts from MOD was that they did not see the rapid
deepening in the Gulf of Mexico coming from early models.
- First short range models involving the system were initialized 1200 UTC 11
March. (LFM, NGM, ETA, AVN) Showed intense cyclone over East Coast,
- MOD issued storm summary statements about the storm every 6 hours, predicting
well that powerful storm would hit East Coast. Used words “unusually severe”
and “perhaps record breaking”. Issued first at 2200 UTC 11 March.
- The MOD and local forecasters used the “hurricane hotline” to communicate and
attempt to improve forecasts and keep in agreement: Concerns in snowfall
amounts and rain/snow line continued even as storm developed. - Forecasters faced most difficulties with initial formation over the Gulf of Mexico. - Valid 0000 UTC 13 March:
- 48-24 hrs out, forecast had low pressure at 1006 mb over Florida Panhandle (this
would be several hundred km NE and 22 mb too high of actual) - Pressure (intensity) forecasts were closer as 0000 UTC 13 March approached, but
12 hr forecast had 993 mb, still 9 mb too high.
- Forecasters disregarded LFM model showing much deeper cyclone moving out of
the Gulf of Mexico, citing the tendency of the model to overdeepen cyclones due
to “convective feedback”. The feedback of latent heat release on the cyclogenesis
was significant in this case. (How would this be predicted?)
- Valid 1200 UTC 13 March:
- At 36 hrs, 996 mb was predicted (by human forecasts), and the forecast pressure
reduced with every 12 hr period, down 20 mb to 976 by the time the 12 hr
forecast came out, still 5 mb higher than the recorded pressure of 971 mb. - Valid 0000 UTC 14 March:
- The MOD 48 hr forecast, after noticing that the models continued to
underdevelop and underdeepen the system, was released with a forecast central
pressure of 963 mb. This forecast was 3 mb below lowest model, a good
correction by MOD forecasters, the actual pressure was 960 mb. - When the cyclone began to develop on 12 March, rain/snow line was predicted
just west of urban corridor.
- The models changed significantly in the next 24 hrs, but forecasters decided that
they did not think that deepening would occur as models suggested over Virginia. - Snowfall of 20+ inches in Kentucky and Ohio was much more than forecast. - See pp. 193 for MOD snow forecasts.
- Very long Watch and Warning lead times, see chart pp. 194
- Winter Storm warning lead times up to 43 hours, with Blizzard Warnings 10-24
hours before snow started falling anywhere.
- State of Emergency in many states, all highways closed in Penn, national guard
activated for entire storm.
- Short term forecasts and WSR-88D radar details, see pp 196-197
Overall, somewhat biased article towards the forecasters, but makes clear that that
forecasts were very good in general, and even the specific short term forecasts were good.
The watch/warning lead times were unprecendented.
(4) A Snowfall Impact Scale Derived from Northeast Storm Snowfall
Bulletin of the American Meteorological Society; Vol ?; 02/04; pp. 177-194
The article presents a new scale for classifying winter snowstorms affecting the Northeast
U.S. Kocin and Uccellini have tried to create a scale that measures snowstorms based on
a mathematical formula taking into account amounts of snow over size of areas, and
amounts of snow over size of populations, and categorizing them. The scale takes into
account the formula, and then ranks storms from 1 to 5, with 5 being the most severe.
- In a scale created by Hart and Grumm, taking into account the normalized rddeparture from climatological means, the March 1993 superstorm ranked 3
among storms in the 53 yr time period studied.
- Kocin and Uccellini feel that storms should be characterized by their unique,
extensive distribution of snowfall in the NE urban corridor, because that is where
the greatest impact is.
- Have created the Northeast Snowfall Impact Scale (NESIS) to rate snowstorm
- The NESIS calculation is given on pp. 179-181
- March 1993 superstorm has the highest NESIS rating of 12.52, which gives it one
of only 2 storms with a category 5 ranking. The other is 6-8 Jan 1996. 70 storms
were studied in this, the 70 that Kocin and Uccellini feel were most significant in
last 100 yrs.
- NESIS values of 10+ are given a category 5 rating.
This paper by Kocin and Uccellini offers a new perspective possibly on how to categorize
a “superstorm.” Because the two are “winter weather experts,” can the scale they create
be used to classify a superstorm? Is a category 5 snowstorm now a superstorm because
they say it is? Does the algorithm being used take enough into account to be considered
an accurate ranking system? For example, 2 inches of ice on the roadways of the NE
population centers would not even show up on this scale as an extreme storm, even if a
great amount of snow fell over not as populated areas slightly inland. Something to think
(5) Natural Disaster Survey Report: Superstorm of March 1993
U.S. Department of Commerce; National Oceanic and Atmopheric
Administration; National Weather Service, Silver Spring, MD.
For the summary of useful information in the Natural Disaster Survey
Report, refer to Appendix A.
(6) The Blizzard of 12-15 March 1993 in the USA and Canada
“Weather” pub. by the Royal Meteorological Society;
March 1994 Vol. 49 No. 3; pp. 81-89
This article is a summary written in a journal of the Royal Meteorological Society. The
article recaps the major events and impacts of the storm, but does not go into any
extensive details. An overview of the accurate forecasts of the storm is given, but no new
information is given beyond that of sources 1,2, and 3.
- Some information is given about the public statements that were made to the
public by the NWS, and pp82-84 have a few of the winter weather statements.
- The synoptic development is given briefly day by day, followed by snowfall totals
and other impacts.
- No other new news is given, making this article relatively useless to us.
(7) A Classification Scheme for Winter Storms in the Eastern and
Central United States with Emphasis on Nor‟easters
Bulletin of the American Meteorological Sociey; January 2002. pp. 37-51
This article is a classification system developed before that of source 4, by Kocin and
Uccellini. The goal of the classification system is to summarize the potential impact of
future winter storms for given categories of Intensity and Duration, particularly as related
to snowfall amounts. The article discusses in length several storms, including some
mention of the March 1993 Superstorm, which ranks as a high category 4 storm.
- The first category used to evaluate storms is the Storm Intensity. The intensity
takes into effect three things: Central Low Pressure Difference, Rate of
Deepening of Central Low Pressure, and Maximum Pressure Gradient between
central low pressure and central high pressure of adjacent anticyclone.
- Storm intensity weights all of these relatively equally, and has a scale to divide
intensity into categories of 1-5, with 5 being the most intense.
- The second factor taken into account by system to determine impacts is the
Duration Factor. DF also ranges from 1-5, with 5 being slowest moving.
- Both factors taken into account when determining impacts, but Storm Intensity
used more than duration. Pp 48-49 give chart of many impacts (precip, winds,
coastal) based on intensities of storms.
- Superstorm 93 Reached only a high category 4 storm under this system, but
stayed at that status for more than 24 hours.
- SS93 was a relatively fast mover, only category 2 duration factor.
- The combination of filling and fast movement reduced snowfall amounts over
New England to lower than was expected given the history of the storm. The
storm dropped to a category 3 over this northern area.
This classification system creates a much different and more inclusive system than the
one offered by Kocin and Uccellini. The scale takes into account the intensity of the
storm at any moment on its path, making the scale useful in expected impacts, rather than
simply looking back on what has happened and rating it.
(8) ***RM to read
The March 1993 Superstorm Cyclogenesis: Incipient Phase Synoptic-
and Convective-Scale Flow Interaction and Model Performance.
(Dickinson, Bosart, Bracken, Hakim, Schultz, Bedrick, Tyle)
AMS; Monthly Weather Review; 1997 Vol. 125 pp. 3041-3072
This article focuses on the intense strengthening of Superstorm 93 over the Gulf Of
Mexico. Beginning with the fact that the 975 mb central low pressure while still over the
Gulf of Mexico was the deepest cool season storm to affect the Gulf of Mexico in the
1957-present period studied, the article goes on to look at the reasons why the low
strengthened so quickly (see figures 1 a,b pp. 3042). The paper is very technical in the
examination of the Potential Vorticity (PV) phenomena that came together over the Gulf
of Mexico with this storm. The model performance and representation of the initial
cyclogenesis is examined, with a look at the effects of latent heating and convection on
the rapid deepening of the storm.
- Unusual early intensification was not well forecast by the Regional Analysis and
Forecast System (RAFS) or the MRF.
- Large forecast underestimates of the central low pressure (Figure 1b) occurred
during a period of widespread convection near the storm’s center - pp. 3042
- Hypothesis of paper: Model’s inability to simulate properly the bulk effects of
cumulus convection contributed significantly to the poor forecasts.
- Coupling Potential (CP) examined with several PV anomalies moving across the
U.S. towards the Gulf of Mexico. CP is the difference between THETA-T and
THETA-E. Negative values indicate the atmosphere is susceptible to deep
convection given a suitable lifting mechanism and adequate moisture.
- The growth of low level PV occurs beneath decreasing Pt (and associated ridging
on the DT) consistent with the expected development of an area of anticyclonic
circulation above a region of extensive latent heat release.
- Failure to simulate properly cold surges east of the Rockies is a common
systematic failure of the NCEP operational models. Pp 3053
- Pp 3054: Results suggest that the MRF model adequately captured the mass and
winds fields associated with the precyclogenetic phase of SS93, but is inadequate
at later times as cyclogenesis proceeds in the presence of deep convection. This
suggests that inadequate parameterized model physics may contribute to the
poor forecasts of the incipient SS93 cyclogenesis.
- More model failures: Unforecast nonconservation of THETA-T prevents the MRF
model from capturing the significant advection at lower levels of THETA-T over
the rapidly intensifying storm. Unable to capture favorable cyclogenesis pattern.
- MRF also continues to fail to capture rapidly strengthening southwesterly jet
ahead of PV anomaly from Arkansas to Ohio. (JET STREAKS?)
- Diabatic heating effects associated with convection overnorthwestern Gulf of
Mexico was better forecast in the ECMWF model during development phase of
SS93. This and Influences of Convection explained of pp 3058-3060
- The heat flux (latent heat) discrepancy between the observations and model
forecasts over the Gulf of Mexico are attributed to 1) both weaker wind speed and
low-level thermal gradient forecasts in the models (especially in MRF) 2) in the
underestimate of the SST over the Gulf of Mexico in the models, especially
the MRF. See pp. 3060-3063
- Strong SST gradient from the Mississippi Delta out into the Gulf of Mexico, and
the gradient and baroclinic zone discussed pp. 3062-3063.
- For summary of why such deep cyclogenesis occurred, including the lateral and
vertical interaction of long-lived PV anomalies, exceptionally unstable airmass
over Gulf of Mexico, and Amplifying Trough/Ridge pattern, see summary a. pp
- MODEL ERROR SUMMARY: “The MRF failure to simulate properly the
incipient SS93 cyclogenesis over the Gulf of Mexico can be linked to the inability
of the model to replicate the observed widespread convective outbreak in the
storm environment triggered by the approach of PV anomaly “C” see pp 3064-
- This was shown to not have been based on initial conditions OF THE PV
ANOMALY because the anoma+ly was represented well the day before.
- The ECMWF had a much better forecast of the cyclogenesis over the Gulf of
Mexico based on several factors discussed on pp 3065. (Did forecasters have
access to the ECMWF model?)
- Author comment: A closer look is needed at how the MRF and ECMWF
- Medium Range: The article says that the ability of the models to forecast the east
coast storm four days in advance without correctly predicting the deepening in the
Gulf of Mexico suggests that the PV anomaly “A” (on the northern branch of the
jet) would have been enough to cause a major North American cyclone due to the
natural baroclinic zone from the Gulf Coast to the Atlantic Coast in winter.
- Pp.3070: Results: “Our results suggest that the exceptional intensity of the
incipient SS93 cyclogenesis over the Gulf of Mexico can be associated with 1) an
evolving large-scale flow pattern that favored the establishment of the positive
phase of the PNA pattern over North America with a deepening trough
downstream of the Rockies; 2) the presence of a strong PV anomaly in the STJ
that triggered the onset of widespread convection over Southeastern Texas and
offshore as it moved eastward across Mexico; 3) a positive feedback between
ascent, low-level convergence, cyclonic vorticity production, deep convection and
the advection of lower values of THETA-T and higher values of Pt in the presence
of a stronger than forecast barocyclinic zone and a stronger than analyzed SST
gradient; and 4) an increase in THETA-T and associated decrease in Pt, both
attributable to latent heat release downshear of the developing cyclone, that
facilitated self-development through enhanced downshear ridging and the
development of a strong outflow aloft.”
(9) ***RM to read
A Diagnostic Analysis of the Superstorm of March 1993
( Huo, Zhang, Gyakum, Staniforth)
AMS; Monthly Weather Review; 1995, Vol. 123; pp. 1740-1761
This article is a discussion about a different type of model that the team used to attempt
to better predict the rapid deepening SS93 over the Gulf of Mexico. This article first goes
through the atmospheric conditions at the time of the development, and the sources of
those atmospheric conditions. A very good synoptic overview is presented regarding
reason why the system strengthened so rapidly over the Gulf of Mexico. The paper then
goes into the details of the runs of the Canadian Regional Finite-Element (RFE) model,
and how that model predicts the deepening over the Gulf of Mexico.
Introduction pp. 1740-1741
- Necessary ingredients for explosive cyclogenesis are strong baroclinicity, short-
wave troughs, stratospheric extrusions, upper-level jet streaks and lower-level jets,
symmetric instability and weak static stability, latent heat release, and surface
sensible and latent heat fluxes
- Upper-level and low-level elements are often dynamically coupled trough vertical
and ageostrophic circulations that are aided by upright or slantwise convection.
- Inconsistencies in models are due to imperfect initial conditions, poor grid
resolution, or improper model physics representations.
Synoptic Overview pp. 1741-1745
- 1200 UTC 12 March
- SST was more than 3 degrees Celsius above the monthly normal (Gilhousen 1994)
- 850 mb: Weak warm advection ahead of the low, with a very cold air mass and
strong temperature gradients advected southward east of the Rockies. This led to a
rapid increase in baroclinicity along southern seaboard.
- Presence of strong southern flow is an indicator of ample moisture being
transported into baroclinic zone for subsequent latent heat release.
- Two short wave troughs in middle troposphere were associated with amplifying
- Upper troposphere had a “double jet” structure one ahead and one behind upper
level trough. This structure is known to have some part in the effect of tropopause
depression on extratropical cyclogenesis. - Westerly jet intensified rapidly in response to enhanced baroclinicity: The
presence of such strong jet streaks is an indication of strong baroclinicity which
would determine the subsequent explosive deepening of the storm.
- 0000 UTC 13 March
- Warm front moving north near east coast, while there is a squall line over central
Gulf of Mexico.
- High cloud tops at center of storm: 28000-34000 feet.
- Short-wave troughs intensifying and vorticity maxima increasing. The eastern
trough near the low pressure center caused positive vorticity advection toward the
low, which appears to have played an important part in the initial development of
- Low level jet formed ahead of the cold front, providing warm moist air to the
- The westerly jet had intensified rapidly, and the northwesterly jet had intensified
some, with the low pressure system beneath the right entrance of the jet streak
(see pp 1743) which is a position favorable for surface cyclogenesis (Uccellini
and Johnson 1979)
- 1200 UTC 13 March
- Warm front parallel to east coast had strengthened
- Low-level jet had increased in strength, significant moisture transport from
tropical ocean to warm sector of the cyclone. - 500 mb: Two short-wave troughs merge and produce negative tilt. Vorticity
centers strengthen and merge.
- Surface cyclone situated at right entrance of northern jet streak, and left exit of
southern jet streak, which is a more favorable position for cyclogenesis (Uccellini
and Kocin 1987)
- 0000 UTC 14 March
- Storm entered occlusion stage, began filling slowly
- Mid and upper-level disturbances reach maximum intensity
- Two jet streaks intensify further
- Figures of 850 hPa, 500 hPa, 250 hPa, and surface maps are available in this
Numerical Prediction System and Model Predictability pp. 1745-1751
- The details of the RFE model runs are presented in this section, making some note
of the importance of adding the latent heat release to the models. Without
including convection in the models, the vorticity of the system is grossly
Possible Deepening Mechanisms pp. 1751-1759
- Latent heat release: pp 1751-1753 describe the model run on the system with and
without the effect of latent heat release included. It is found in this paper that
without latent heat release, the path of the storm stays very similar to that with
latent heat release, but the storm is weaker in intensity and slower in displacement.