A historical-analog-based severe weather checklist for central New York and northeast
Pennsylvania
MICHAEL EVANS and RON MURPHY
NOAA/National Weather Service, Binghamton, New York
ABSTRACT
This presentation shows an operational severe weather checklist developed at the National
Weather Service Forecast Office in Binghamton, New York (WFO BGM). The checklist contains
two sections; a traditional severe weather parameter data entry section, and a historical analog
retrieval section. Parameters on the checklist are related to large-scale forcing, stability, wind
shear and moisture. The analog retrieval portion of the checklist returns information on five
historical events that are analogous to current conditions, based on data entered into the
checklist. An example of the utility of the checklist is shown for a major severe weather event
on 28 April, 2011, which featured numerous severe weather reports in the WFO BGM county
warning area, including seven tornadoes. Results from a verification study on the similarity
between observed events and corresponding analogs indicate that the system can help
forecasters to anticipate convective mode, and discriminate between major and null events.
1. Introduction
Previous research has demonstrated that a
thorough assessment of the pre-storm
environment is a critical step in the severe
weather warning process (Weisman and
Klemp 1984, Brooks et al. 1994, Lombardo
and Colle 2010, Dial et al. 2010, Thompson et
al. 2012). Severe weather checklists have
been developed and utilized in operational
forecasting in order to help forecasters to
digest and organize large quantities of
information related to this process.
Forecaster experience and pattern
recognition is another critical component in
the severe weather forecasting process
(Johns and Doswell 1992, Grumm and Hart
2001). Utilization of historical analogs can
help forecasters with this aspect of
forecasting. Recent applications developed to
help forecasters by identifying appropriate
historical analogs include Cooperative
Institute for Precipitation Systems (CIPS)
Analog Guidance (Gravelle et al. 2009) and
the sounding analog retrieval system
developed at the storm prediction center
(Jewell 2010).
Corresponding author address: Michael Evans, National Weather Service, 32 Dawes Drive,
Johnson City, NY 13790. E-mail: michael.evans@noaa.gov
This paper will demonstrate an application
developed at the National Weather Service
Forecast Office in Binghamton, NY (WFO
BGM), which combines many of the attributes
of a severe weather checklist with a historical
analog retrieval system. Section 2 introduces
the application by showing an example of
how the system could be utilized for a
potential severe weather event. Section 3
shows results from a verification study on
output from the application, and Section 4
contains a summary and conclusion.
2. The checklist on 28 April, 2011
A severe weather outbreak occurred
across the northern mid-Atlantic region
during the early morning hours on 28 April,
2011 as a cold front tracked east from the
Great Lakes toward the east coast (Fig. 1a).
The 00 UTC 28 April Pittsburgh (KPIT)
sounding indicated modest lapse rates and
minimal convective available potential energy
(CAPE) values across southwest Pennsylvania,
however winds were quite strong, with 40 to
50 kt values indicated through a deep layer
above 2000 ft (0.5 km, Fig. 1b).
Model
forecast soundings valid several hours later
for northeast Pennsylvania and central New
York (not shown) indicated that some
destabilization would occur within this air
mass as it was transported northeastward in
the strong southwesterly flow, with mixedlayer CAPE values approaching 1000 Jkg-1.
Winds were forecast to remain strong, with
values exceeding 60 kt at and above 850 hPa.
Figure 1. a) NCEP/HPC surface analysis at 0600 UTC on
28 April, 2011. b) Observed 0000 UTC 28 April, 2011,
KPIT skew-T diagram.
A completed severe weather checklist
based on proximity soundings and the largescale forcing described in this section is
shown on Fig. 2a.
The key characteristics
associated with this event were a progressive
cold front with moderately large midtropospheric geopotential height falls,
modest instability, and very strong winds with
large shear and storm relative helicity values
in the 0-1 km layer.
The historical analogs associated with this
case are shown in the table on Fig. 2b. Data
for all potential analogs are contained in a
locally-developed severe weather data base,
containing representative parameters from
267 events that occurred from 1998-2013.
Analogs are determined based on the degree
of similarity between values entered into the
checklist, and values stored in the historical
data base. The first, or “best” analog shown
on the table was 31 May 1998; a massive
severe weather event which included 17
tornadoes in the BGM county warning area
(CWA), or more tornadoes than any other
event in the historical data base. In addition,
the third analog was associated with multiple
tornadoes. Given that only 13 percent of all
of the events in the historical data base
include tornadoes, the fact that the analogs
for this case included cases with multiple
tornadoes would be an indication that
tornadoes would be an unusually large
concern on 28 April. The convective mode
associated with the analogs was mixture of
isolated supercells and linear modes,
indicating no strong preference for any
specific evolution.
Figure 2. a) A completed data entry portion of the checklist for the 28 April, 2011 severe weather event for
northern Pennsylvania and southern New York. b) Output page from the analog retrieval portion of the checklist
for the 28 April, 2011 severe weather event. The first five analogs are displayed in the table. A banner is also
displayed indicating that the parameters in this case were favorable for tornadic supercells (based on local
research).
The end result was a widespread severe
weather event on 28 April, including seven
tornadoes. Several flash floods were also
reported, despite the fact that no flash floods
were indicated on the analogs. Tornadoes
initially developed in association with isolated
supercells over central New York; however
the isolated cells were eventually overtaken
by a line of convection, which was associated
mainly with damaging straight-line winds and
flash flooding (Fig. 3).
Figure 3. a) Binghamton (KBGM) WSR-88D 0.5°
reflectivity at 0708 UTC on 28 April, 2011. Each of the
three annotated supercells produced a tornado at
some point during the event. b) KBGM WSR-88D 0.5°
reflectivity at 0836 UTC on 28 April, 2011.
4.
Verification
Results from a study to examine the
performance of the historical analog retrieval
portion of the BGM severe weather checklist
are shown in this section.
a.
Methodology
All 81 severe and convective flash flood
events occurring in the WFO BGM CWA from
2011-August, 2013 were examined. For each
case, data corresponding to the event were
entered into the checklist, and the checklist
was run to find associated historical analogs.
Each case was associated with four historical
analogs. Assessment of the quality of the
analogs returned by the checklist was
performed by comparing the convective
mode of the tested cases with the convective
mode of the corresponding analogs, and by
comparing quantitative characteristics of the
reports associated with the tested cases to
quantitative characteristics of the reports
associated with the analogs. Convective
modes were determined by viewing radar
animations of the test cases and analogs, and
assigning each case as linear, isolated or
multicellular. A tested event was considered
to be a match with its corresponding analog if
both the tested and analog events were
categorized as linear, or both were
categorized as isolated.
A non-match
occurred when an event categorized as
“isolated” was paired with a solid or broken
line. For all other types of pairings, the
inherent difficulty associated with easily
distinguishing between categories resulted in
a non-definitive classification.
b.
Results
Comparisons between the convective
mode of the tested cases in the study vs. the
mode of each case’s four corresponding
analogs indicated 146 matches, 37 nonmatches,
and
141
non-definitive
comparisons. Similar ratios between matches
vs.
non-matches
and
non-definitive
designations were found when only the top
analog was tested.
The observed number of severe reports for
each of the 81 test cases vs. the average
number of severe reports from each case’s
four corresponding analogues is plotted in
Fig. 4a. The cluster of data points in the
lower left part of the graph indicates a large
number of cases with a small number of
severe reports associated with analogs that
also had a small number of severe reports.
Cases with a large number of severe reports
were mostly associated with analogs
containing a large number of severe reports.
Correlations between the number of severe
reports with the tested cases vs. the number
of severe reports with their corresponding
analogs are shown in Fig. 4b. Correlations
were positive but rather modest for the best
analog and the second best analog, higher for
the average of the top two analogs, and
highest for the average of all four analogs. All
of the correlations were statistically
significant at the 0.95 level.
Figure 4. a) A scatter diagram showing the number of
observed severe weather reports for each of 81 tested
cases (x axis) vs. the average number of reports from
the top four analogs for each case (y axis). b) The
Pearson product-moment correlation coefficient
between the number of observed severe weather
reports in the tested cases vs. the number of severe
events in the top analog (correl analog 1), the second
analog (correl analog 2), the average number of events
in the first and second analogs (correl analog 1+2), and
the average number of analogs in the top four analogs
(correl analog all) for each event.
For the purposes of this study, “major
severe” events are defined as events
occurring since 2011, with at least 30 severe
reports. Eleven such events have occurred in
the BGM CWA since the beginning of 2011,
with a median number of 37 reports (Fig. 5a).
The median number of reports returned from
the four top analogs associated with those
eleven events was 17.25 (Fig. 5a).
By
contrast, a “null” event is defined as an event
with one or fewer severe reports. Twenty
three such events are included in the severe
weather data base since the beginning of
2011.
The median number of reports
returned from the corresponding four top
analogs was 3.25 (Fig. 5a). The difference
between the number of reports in the
analogs associated with major events and the
number of reports in the analogs associated
with null events is statistically significant at
the 0.95 level, using a Mann-WhitneyWilcoxon testing procedure (Gibbons, 1976).
When the checklist returns historical
analogs with large numbers of severe events,
a user could interpret the result as an
indication that the checklist is flagging the
current event as being a potentially major
event. For purposes of this study, a “checklist
flagged major” event is defined as such an
event, when the checklist returns analogs
with an average of at least 15 severe reports.
Twelve such events have occurred in the BGM
CWA since the beginning of 2011, with a
median of 19.63 reports per analog (Fig. 5b).
Those events were associated with a median
of 29 observed severe reports (Fig. 5b). By
contrast, a “checklist flagged null” event is
defined as an event when the checklist
returns analogs with an average number of
severe reports less than five. Thirty two such
events have occurred in the BGM CWA since
the beginning of 2011, with a median of 2.88
reports per analog (Fig. 5b). These events
were associated with a median of 2 observed
severe reports. The difference between the
number of observed severe reports
associated with the “checklist flagged major”
events and the number of observed severe
reports associated with the “checklist flagged
null” events was significant at the 0.95 level,
using a Mann-Whitney-Wilcoxon testing
procedure (Gibbons, 1976).
Figure 5. a) The median of the number of observed
severe reports in the 11 major severe events tested in
the verification study (blue) and the median of the
average number of reports from the corresponding
analogs (red). Also, a comparison between the median
of the number of observed severe reports from the 23
null events tested in the verification study (blue) and
the median of the average number of reports from the
corresponding analogs (red). b) The median of the
number of observed severe reports in the 12 “checklist
flagged major” severe events tested in the verification
study (blue) and the median of the average number of
reports from the corresponding analogs (red). Also, a
comparison between the median of the number of
observed severe reports from the 32 “checklist flagged
null” severe events tested in the verification study
(blue) and the median of the average number of
reports from the corresponding analogs (red).
5.
Summary
This paper describes an operational severe
weather checklist developed at WFO BGM.
The checklist contains two sections; a
traditional severe weather parameter data
entry section and a historical analog retrieval
section based on a comparison between
values of parameters entered into the
checklist in the data entry section, and
corresponding values of the same parameters
from a data base of historical severe weather
events.
Results from a verification study
indicate that the convective mode of tested
events was a match with the mode of their
corresponding analogs much more often than
a non-match, however in many cases the
comparison was non-definitive. Statistically
significant, positive correlations were found
between the number of severe reports in the
test cases and the number of severe reports
in corresponding analogs.
The highest
correlation to the number of severe reports in
the tested cases was found with the average
number of reports from all four
corresponding analogs, indicating the
importance of considering results from
multiple analogs when considering severe
weather potential for an event. Results from
the study also indicated that major events
typically returned analogs with many more
severe reports than null events, indicating
that the checklist can help forecasters
discriminate between major and null events.
Acknowledgements. The authors wish to
thank the forecasters at WFO BGM who have
been using and testing this application for the
past few years.
REFERENCES
Brooks, H. E., C. A. Doswell III, and J. Cooper, 1994: On
the environments of tornadic and non-tornadic
mesocyclones. Wea. Forecasting, 10, 606-618.
Gibbons, J.D., 1976: Nonparametric methods for
quantitative analysis. American Sciences Press, 463
pp.
Gravelle, C. M., C. E. Graves, J. P. Gagan, F. H. Glass,
and M. S. Evans, 2009: Winter Weather Guidance From
Regional Historical Analogs. Preprints, 23rd Conf. on
Weather Analysis and Forecasting, Omaha, NE, Amer.
Meteor. Soc., JP3.10. [pdf]
Jewell, R.E.D., 2010: A sounding analog system to
produce probabilistic forecasts of maximum hail size.
Preprints, 25th Conf. on Severe Local Storms, Denver,
CO, Amer. Meteor. Soc., 4B.3 [pdf].
Johns R.H., and C.A. Doswell, 1992: Severe local storms
forecasting. Wea. Forecasting, 7, 588-612.
Lombardo, K.A. and B.A. Colle, 2010: The spatial and
temporal distribution of organized convective
structures over the northeast and their ambient
conditions. Mon. Wea. Rev., 138, 4456-4474.
Thompson, R.L., B.T. Smith, J.S. Grams, A.R. Dean, and
C. Broyles, 2012: Convective modes for significant
severe thunderstorms in the contiguous United States.
Part II: supercell and QLCS tornado environments.
Wea. Forecasting, 27, 1136-1154.
Weismann, J.L. and J.B. Klemp, 1984: The structure and
classification of numerically simulated convective
storms in directionally varying wind shears. Mon. Wea.
Rev., 112, 2479-2498.