Content uploaded by Barry D. Keim
Author content
All content in this area was uploaded by Barry D. Keim on Apr 30, 2014
Content may be subject to copyright.
CLIMATE RESEARCH
Clim Res
Vol. 28: 143–154, 2005 Published March 16
1. INTRODUCTION
One of the primary objectives in synoptic climatol-
ogy is the classification of synoptic weather patterns
into discriminate groups (Yarnal 1993). Once these
groups, or ‘weather types’, are generated, they provide
a useful baseline for environmental analysis, with
widely ranging applications from studies of climate
change (Kalkstein et al. 1990) to human mortality
(Greene & Kalkstein 1996). Recent methods in synoptic
climatology have focused primarily on statistical
derivation of synoptic weather patterns through vari-
ous means including cluster analysis (e.g. Kidson 2000)
and/or principal component analysis (e.g. Serra et al.
1999). This study presents a manual synoptic weather
classification system for the East Coast of New Eng-
land, for use in environmental analyses in the region.
Manual synoptic climate classifications have one pri-
mary advantage over statistically generated systems in
that the weather class (or type) can be classified by the
user through simple visual examination of a weather
map. The primary disadvantage of manual systems is
that the classification of long time periods can be
prohibitively time consuming, although 2 long running
manual synoptic classifications exist for the British
Isles, dating back to 1861 (Lamb 1972), and New
Orleans, Louisiana (USA), commencing in 1961 (Muller
1977). Yarnal (1993) has shown that manual weather
typing is as effective a tool in environmental analysis
as those derived through statistical means.
© Inter-Research 2005 · www.int-res.com*Email: keim@lsu.edu
Manual synoptic climate classification for the East
Coast of New England (USA) with an application
to PM
2.5
concentration
Barry D. Keim
1,
*
, Loren David Meeker
2
, John F. Slater
3
1
Department of Geography and Anthropology, Louisiana State University, 327E Howe-Russell Science Complex,
Baton Rouge, Louisiana 70803, USA
2
Climate Change Research Center, University of New Hampshire, 56 College Road, Durham, New Hampshire 03824, USA
3
Division of Engineering, Math, and Science, Daniel Webster College, 20 University Drive, Nashua, New Hampshire 03063, USA
ABSTRACT: This study presents a manual synoptic climate classification for the East Coast of New
England with an application to regional pollution. New England weather was classified into 9
all-inclusive weather types: Canadian High, Modified High, Gulf of Maine Return, New England
High, Atlantic Return, Frontal Atlantic Return, Frontal Overrunning–Continental, Frontal Overrun-
ning–Marine, and Tropical Disturbance. Canadian High and Modified High weather are the domi-
nant weather patterns at Boston, Massachusetts, while Tropical Disturbance, Gulf of Maine Return,
and New England High weather types are the least frequent. Properties of the weather types were
determined at 07:00 h Local Standard Time (LST) each day in Boston. The coldest and driest weather
type is the Canadian High, while the hottest, most humid weather is generated by Frontal Atlantic
Return. The synoptic weather classification system, applied to airborne fine particle mass concentra-
tions with an aerodynamic diameter ≤ 2.5 µm (PM
2.5
), showed significant differences in concentra-
tions between weather types: transport from the north and northwest had low PM
2.5
, while transport
from the south and southwest had high PM
2.5
concentrations. This climate classification system also
has potential applications ranging from studies of insect migration to analyses of climate change.
KEY WORDS: Synoptic climatology · New England · USA · Boston · Weather types · Applied
climatology · PM
2.5
Resale or republication not permitted without written consent of the publisher
Clim Res 28: 143–154, 2005
An application to PM
2.5
(atmospheric particles with
an aerodynamic diameter ≤ 2.5 µm) at a rural site in
New Hampshire is demonstrated. By connecting
changes in synoptic weather conditions and changes in
the concentration of PM
2.5
, this work will help to
increase the understanding of some of the mechanisms
behind aerosol variability in the northeastern USA,
which in turn can reduce the uncertainty associated
with climate forcing by aerosols.
2. SCOPE AND METHODS
New England weather and climate patterns are
arguably among the most varied in the world. They
include extremes of both high and low temperatures,
large inter-diurnal changes in temperature, droughts,
heavy rainfall, and blizzards (Zielinski & Keim 2003).
These great variations are influenced by many factors
relating to the physical geographical setting, including
the region’s latitude near 45° N, its coastal orientation,
its continentality from its position within the westerlies,
and its mountains. Since the weather patterns in the
New England region are complex at fine scales, the
aim of this classification system is to find order and
structure at the synoptic scale, especially in the coastal
zone in eastern New England, where the major cities
of Boston, Massachusetts, and Portland, Maine, lie,
and where the University of New Hampshire has a
research program to understand transport of pollution.
Synoptic climatologies can be based on atmospheric
pressure patterns or air-mass properties partitioned
into classes based on manual, automated, or statistical
procedures. Here, we apply an atmospheric pressure-
based system using manual procedures similar to the
synoptic classification developed for New Orleans and
the central Gulf of Mexico coast by Muller (1977). This
circulation-to-environment classification relies on the
classic cyclone–anticyclone models with their centers
of low and high pressure, respectively, and associated
frontal boundaries, rain shields, wind vectors, etc. Any
point within New England will have specific weather
conditions because of its proximity to these cyclones
and anticyclones. Therefore, given the scale of New
England in proportion to these synoptic-scale weather
systems, there will be instances when the entire region
is experiencing the same weather type, or when the
western portion of the region may have weather of a
different type than the east, etc. Hence, applications of
the classification must be done on a point location
basis, rather than regionally as that of Davis & Kalk-
stein (1990).
Although this synoptic classification is devised for
the eastern coastal region of New England from Cape
Cod to the border with Canada, it has broad applicabil-
ity across the larger New England region. Boston, and
Durham, New Hampshire (approximately 100 km
north of Boston) were chosen as the foci for this analy-
sis (Fig. 1). Boston is the largest city in New England
and was chosen as the focal point for analysis of the
frequency of the derived synoptic weather types and
air-mass properties associated with them. Boston also
has abundant high quality data for numerous weather
variables, including temperature, dew point and wind
speed, which should be representative of the relative
frequencies and properties of the weather types.
Boston was also chosen for its location on the East
Coast, because this classification scheme will lend
itself for subsequent use in analysis of the atmospheric
chemistry in the rural region of Durham. The relative
frequencies and properties of the weather types prob-
ably do not vary much between these 2 locations,
given their proximity relative to the scale of cyclones
and anticyclones. However, there are situations when
a frontal boundary is located between the 2 sites, and
delineates different weather types with substantive
differences in air-mass properties. These differences
are even greater for more distant locations such as
Burlington, Vermont, but the general classification and
its nomenclature are nevertheless applicable through-
out New England.
144
Fig. 1. Map of New England, USA, showing locations
mentioned in the text
Keim et al.: Manual synoptic climate classification
The synoptic weather classification for the East
Coast of New England was devised through examina-
tion of weather maps, mostly from the Daily Weather
Maps series published by the National Oceanic and
Atmospheric Administration (NOAA). At Boston, the
time period and maps analyzed include the 07:00 h
LST (12:00 h Greenwich Mean Time) daily weather
map for the 5 yr period from January 1995 through
December 1999. This time period includes the El Niño
event of late 1997/early 1998 and the 1999 La Niña
event. However, the impacts of ENSO are minimal in
New England (Ropelewski & Halpert 1987, Halpert &
Ropelewski 1992, Bradbury et al. 2003) and these
occurrences should not affect the basic derivation of
the synoptic classification. This use of daily maps
should yield a reliable climatology, because the fre-
quencies of synoptic-scale pressure patterns are
largely unaffected by diurnal variability.
This dataset was produced for initial assessment of
weather types to analyze pollution transport into New
England. Subsequent to this analysis, a separate data-
set was derived for Durham, including analyses of 2
daily maps, at 07:00 and 19:00 h, for the summers of
2001 and 2002, to assess PM
2.5
.
3. SYNOPTIC WEATHER TYPES
Manual synoptic weather classification is formulated
over an extended period of time through continuous
examination of weather maps. Eventually, discernible
types of weather emerge that result from particular
recurring pressure patterns. From this process, 9 all-
inclusive weather types were generated for eastern
New England, reflecting regional transport pathways
into the region, potential for precipitation (which is not
directly analyzed here), as well as air-mass character-
istics, e.g. temperature, humidity, wind (Fig. 2). The
selection of 9 discernible weather types, partitions of
the classic cyclone/anticyclone model, closely follows
the classification developed by Muller (1977), which
included 8 weather types. The relative position of a
location to the centers of high or low pressure and/or to
the frontal boundaries determines the weather type.
For example, when Boston is located in the warm sec-
tor of a cyclone, yet close to the cold frontal boundary,
the weather type will be Frontal Atlantic Return. At the
same time, Concord may lie on the other side of the
cold front, and hence the classification may be either
Frontal Overrunning–Continental or Canadian High,
depending on the extent of cloud cover behind the cold
front; if Concord is positioned beyond the cloud shield,
it will be classified as Canadian High. There is vari-
ability within each weather type (Fig. 2), but the goal is
to create a meaningful number of classes that are use-
ful for environmental analysis. The geographical space
depicted in Fig. 2 can be parsed into virtually any num-
ber of classes, but the 9-class system follows the exam-
ple of Muller (1977), producing a system with poten-
tially meaningful applications in the region, especially
regarding air pollution.
Canadian High (CH) (Fig. 3). This weather type
occurs when the region falls on the eastern side of an
anticyclone, whereas the area of high pressure is usu-
ally situated northwest or west of New England (e.g.
over or just north of the Great Lakes). This pressure
pattern steers northerly or northwesterly winds into
the region, delivering Canadian air. Conditions in this
weather type are mostly fair, and in winter this
weather type delivers the lowest temperatures and dri-
est air to the region.
Modified High (MH) (Fig. 4). MH weather occurs on
the northern side of an anticyclone. As an anticyclone
of Canadian origin drifts south or southeastward, the
center of the high pressure complex becomes situated
in the American Midwest, bringing westerly winds into
New England. These anticyclones typically have had a
long trajectory into the central United States and are
therefore modified through slight increases in temper-
ature and humidity, when compared to CH. MH can
also be of Pacific origin. This is a fair weather type with
low temperatures in winter, but it is milder than CH.
Gulf of Maine Return (GMR) (Fig. 5). This weather
type occurs when a location falls in the southern quad-
rant of an anticyclone. When Canadian high pressure
moves north and east of New England, winds shift to
the northeast or east, advecting maritime polar (mP) air
from the Gulf of Maine into the region. Typically, the
weather is fair and humid, but can sometimes produce
ocean-effect snow in the coastal zone, especially over
Cape Cod.
New England High (NEH) (Fig. 6). Occasionally, a
dome of high pressure will track directly over New
England. When the high is centered over the area,
145
Fig. 2. Circulation-to-environment synoptic weather classifi-
cation for the East Coast of New England
Clim Res 28: 143–154, 2005
winds are weak and from changing directions across
the region. NEH are also classified when there is little
or no pressure gradient at all and the regional
response is the same as when the high pressure com-
plex is directly overhead. Fair weather is common in
these instances.
Atlantic Return (AR) (Fig. 7). AR weather occurs in
the return flow of an anticyclone (on the western side),
and/or within the warm sector of a cyclone. When high
pressure is positioned to the southeast of New Eng-
land, just offshore from the mid-Atlantic states, the
pressure pattern steers southwest, south, or even
southeasterly winds into the region. This weather type
typically produces high temperatures and humid
weather. AR is mostly a fair weather type, but scattered
afternoon thunderstorms are common in summer and
early fall.
Frontal Atlantic Return (FAR) (Fig. 8). As a mid-
latitude cyclone approaches New England from the
west, the region is positioned in the warm sector of the
cyclone, with south or southwesterly flow. Advection is
similar to the AR weather type, but frontal boundaries
and instability influence the region. This weather type
typically produces cloudiness and rain, even in winter.
Frontal Overrunning–Continental (FOR–C) (Fig. 9).
This weather type occurs in a narrow zone behind a
passing cold front. After the passage of a cold front,
high pressure typically builds in from the west or
northwest, driving continental winds into the area
from the northwest. Just after the surface front clears
the area, however, tropical air often overrides the con-
tinental polar air at the surface, causing clouds and
precipitation. In winter, the precipitation type is usu-
ally snow.
Frontal Overrunning–Marine (FOR–M) (Fig. 10).
FOR–M weather occurs in the northeastern sector of a
cyclone. When low pressure is positioned either off-
shore or south of New England, the region experiences
easterly winds (marine in origin) from the Gulf of
Maine, although maritime tropical (mT) air overrides
this relatively colder air from the south. This weather
type can produce rain, and prolonged snowfall is com-
mon in winter. This is the primary weather type for
coastal northeasters.
Tropical Disturbance (TD) (Fig. 11). Hurricanes and
tropical storms can occasionally make it to New Eng-
land, delivering heavy rainfall and high winds. Wind
direction is completely variable, depending on the
storm path. More often than not, New England
receives remnant tropical systems that produce rain,
but with gentle winds.
The basic classification system is the same as that
proposed by Muller (1977), even if there are obvious
differences in weather types between New England
and Louisiana. The differences between this synoptic
146
Fig. 4. Modified High (MH) weather type on October 23, 1998
Fig. 5. Gulf of Maine Return (GMR) weather type on June 10,
1999
Fig. 3. Canadian High (CH) weather type on March 16, 1998
Keim et al.: Manual synoptic climate classification
147
Fig. 6. New England High (NEH) weather type on August 8,
1998
Fig. 7. Atlantic Return (AR) weather type on June 1, 1999
Fig. 10. Frontal Overrunning—Marine (FOR–M) weather
type on May 2, 1998
Fig. 11. Tropical Disturbance (TD) weather type on August 19,
1991
Fig. 8. Frontal Atlantic Return (FAR) weather type on April 17,
1998
Fig. 9. Frontal Overrunning–Continental (FOR–C) weather
type on June 3, 1998
Clim Res 28: 143–154, 2005
climatology and the Muller system are subtle, yet these
subtleties may assist in understanding the varying
transport of pollution into the region. For example,
New England frequently experiences frontal overrun-
ning conditions with surface winds either from the
northwest (FOR–C with continental air at the surface)
or east (FOR–M with marine air at the surface), repre-
senting 2 completely different circumstances for sur-
face advection that may be important in environmental
analyses. The Muller system only includes 1 frontal
overrunning class, but 2 classes of high pressure sys-
tem (Continental High and Pacific High), a distinction
not considered important in New England. One other
notable difference with this system is the inclusion of
NEH weather, since anticyclones frequently migrate
directly over the region, a situation which is not com-
mon in Louisiana.
4. ANALYSIS AND DISCUSSION
Using the Daily Weather Map series, each day at
07:00 h LST from 1995 to 1999 was classified into 1 of
the 9 all-inclusive weather types, and air-mass proper-
ties, frequencies and seasonality, and transition proba-
bilities are examined.
4.1. Air-mass properties by synoptic weather type
To acquire a sense of the air-mass properties of
each weather type, meteorological data at 07:00 h
LST were extracted for Boston and averaged for each
weather type for January and July (Table 1). Some
sample sizes are rather limited, but their averages
are consistent with expectations. To demonstrate that
these air-mass properties represent differing popula-
tions, the data were tested using a 1-way MANOVA
(Rencher 1998). The 2 types with the lowest January
count (GMR and NEH), were eliminated from the
analysis described below. For each month (January
and July), the multivariate test indicated highly
significant differences among weather types (p <
0.0001). To further explore differences indicated by
MANOVA, pairwise multivariate tests for equality
between weather types were conducted. The signifi-
cance values for the pairwise tests shown in Table 2
indicate that, in most cases, air-mass properties are
significantly different between weather types. Sur-
prisingly, results for January are not as compelling as
July, although Table 1 shows smaller variances be-
tween weather types in both wind speed and visi-
bility in January. Other factors contributing to this
result may be related to the time of day (07:00 h
LST) under analysis and the propensity for snow
cover. In January, highly insignificant differences are
found between FAR and AR, both of which are
characterized by south or southwesterly advection
within the warm sector of a cyclone; hence the simi-
larity is expected. Air-mass properties for AR are also
not dissimilar from FOR–M, both of which are domi-
nated by advection off the ocean. Overall, these
results support the synoptic classification presented
here and the subsequent analysis of PM
2.5
will
further demonstrate differences in the synoptic
weather types, even when the air-mass properties do
148
NTemperature Dew point Relative Wind speed Visibility
(°C) (°C) humidity (%) (km h
–1
) (km)
January
Canadian High (CH) 36 –7 –12 64 19 18
Modified High (MH) 23 –4 –11 59 19 18
Gulf of Maine Return (GMR) 2 –1 –3 86 18 11
New England High (NEH) 4 0 –4 75 11 18
Atlantic Return (AR) 15 1 –3 78 16 14
Frontal Atlantic Return (FAR) 26 3 1 82 19 16
Frontal Overrunning–Continental (FOR–C) 33 –1 –4 80 16 11
Frontal Overrunning–Marine (FOR–M) 16 3 1 88 15 10
July
Canadian High (CH) 28 21 13 63 16 18
Modified High (MH) 33 23 15 63 13 18
Gulf of Maine Return (GMR) 7 18 16 86 10 16
New England High (NEH) 10 21 14 69 8 14
Atlantic Return (AR) 28 23 18 72 13 13
Frontal Atlantic Return (FAR) 24 23 19 82 13 11
Frontal Overrunning–Continental (FOR–C) 8 22 17 75 11 13
Frontal Overrunning–Marine (FOR–M) 16 20 18 91 10 6
Table 1. Average air-mass properties by synoptic weather type at Boston for 1995 to 1999 at 07:00 h LST
Keim et al.: Manual synoptic climate classification
not differ significantly, e.g. GMR from FOR–C and
FOR–M.
In January, the average 07:00 h LST temperature is
lowest during the CH and MH weather types, respec-
tively, with transport of continental polar (cP) or modi-
fied cP air masses from Canada. The highest tempera-
tures at 07:00 h LST in January are found with FAR,
FOR–M, and AR weather types, respectively. If the
afternoon was examined, AR weather would have
higher temperatures, since the 07:00 h LST tempera-
ture follows a night of radiational cooling under mostly
clear skies. Temperatures during FOR–M largely
reflect the offshore sea surface temperature, which
remains higher than the adjacent land temperature at
this time of year. Dew points and relative humidity are
low with the weather types controlled by high pressure
(CH, MH and NEH, respectively) and are humid with
advection from the east (FOR–M, GMR) or south (FAR,
AR). Wind speeds are highest with CH, MH and FAR
weather, while visibility is highest with the suite of
high pressure weather types (CH, MH and NEH).
In July, the lowest morning temperatures occur with
advection off the Gulf of Maine, including the FOR–M
and GMR weather types, which again reflect sea sur-
face temperatures that are lower than land tempera-
tures at this time of year. The highest temperatures
occur with FAR, AR and MH weather. The southerly
advection associated with FAR, and AR weather make
the higher temperatures unremarkable. However, the
presence of higher temperatures under MH conditions
probably implies that mT air from the Gulf of Mexico
and Caribbean is being transported northward into the
Great Plains of the United States and redirected east-
ward to New England by the pressure gradient associ-
ated with high pressure over the
American Midwest. Furthermore, the
air is warmed adiabatically as it
descends from the mountains in west-
ern New England to the coastal zone.
Some of the highest temperatures ever
recorded in New England have come
during times of westerly winds in sum-
mer. Dew points and relative humidity
are low with high pressure weather
(CH, MH and NEH) and are higher
with FAR, FOR–M and AR, respec-
tively. Wind speeds are generally
lower in summer than in winter, with
the greatest wind speeds again associ-
ated with CH weather. Visibility is
greatest under CH and MH condi-
tions, and lowest during FOR–M
weather.
4.2. Weather type frequencies
CH weather was the most frequent type at Boston
during the study period, with >25% of all days
falling into this class; MH weather was second at 16%
(Table 3). The relatively low percentage of TD days
(<1%) is not surprising (Louisiana, which is much more
vulnerable to tropical storms and hurricanes than New
England, had only 3% of all hours classified as ‘tropical
disturbance’ by Muller & Willis 1983). Weather type
frequencies by month also show seasonality. CH
weather was the most frequent weather type in 10 of
the 12 months, and MH was the most frequent in July
and October. CH weather also shows higher frequen-
cies in late winter and early spring, with slightly lower
frequencies in late spring and early summer. This is
consistent with the findings of Bradbury et al. (2002),
who noted an eastward migration of the 500 hPa
United States East Coast Trough from winter to early
spring, which would be associated with the CH
weather type. MH shows little seasonality. GMR
weather is relatively rare, with minimum frequencies
from November through January and higher occur-
rence rates from May through October. The low fre-
quencies in winter are presumably related to the
expansion and configuration of the circumpolar vortex.
After a cold air outbreak in New England, meridional
flow will tend to steer Canadian anticyclones to the
southeast, rather than due east (north of New Eng-
land). NEH shows little seasonality other than rela-
tively low occurrence rates in winter. AR and FAR
weather show a clearer signal with relatively high
occurrence rates in summer. The southerly advection
associated with these weather types is reinforced by
149
MH GMR NEH AR FAR FOR–C FOR–M
January
CH 0.0924 – – 0.0024 0.0000 0.0000 0.0000
MH – – 0.0048 0.0000 0.0002 0.0000
GMR – ––––
NEH ––––
AR 0.3626 0.8198 0.5121
FAR 0.0023 0.0757
FOR–C 0.0906
July
CH 0.0628 0.001 0.0421 0.0000 0.0000 0.0267 0.0000
MH 0.0000 0.0011 0.0004 0.0000 0.0470 0.0000
GMR 0.0312 0.0061 0.0030 0.3716 0.3000
NEH 0.0025 0.0000 0.0223 0.0003
AR 0.0421 0.7795 0.0000
FAR 0.0096 0.0089
FOR–C 0.0329
Table 2. Significance values of multivariate pairwise mean comparisons for
weather types at Boston. See Table 1 for abbreviations. Bold values are
significant at p ≤ 0.05. –: insufficient sample size for analysis
Clim Res 28: 143–154, 2005
strong Bermuda High circulation at that time of year.
FOR–C weather clearly declines in frequency during
summer. Finally, TD weather follows the pattern of the
hurricane season, with a gradual
increase in activity in July and August,
with a peak in September.
Weather type counts for each year
show large variations (Table 4). For
example, annual counts of CH
weather range from 72 in 1999 (20%)
to 118 in 1995 (32%), a difference of
12% in annual frequency. This and
other differences were tested for sig-
nificance using the χ
2
test. Results
imply that the differences among the
years are significant (χ
2
= 49.6, p <
0.05). Seasonal analysis of these data
shows that most of the variability
between years took place in summer
(Table 4) (χ
2
= 41.9, p < 0.05). Differ-
ences in weather type counts during
the other seasons were not significant.
These results suggest that between
1995 and 1999, summers were more
variable than the other seasons.
4.3. Transitions in synoptic weather
types
Transitions in weather types were
analyzed to determine the temporal
sequencing of weather patterns. Two-
step transition probabilities estimated
the probability of weather type occur-
rence on the day following a given
weather type (Table 5). In the case of
CH for example, the day after also tends to be CH (in
40% of the cases). The next highest transition proba-
bility for CH is MH, with TD being the least likely tran-
150
CH MH GMR NEH AR FAR FOR–C FOR–M TD
January 23.2 14.8 1.3 2.6 9.7 16.8 21.3 10.3 0.8
February 30.5 17.7 2.1 5.7 9.2 9.2 19.9 5.7 0.8
March 31.0 12.3 2.6 5.8 11.6 12.3 15.5 9.0 0.8
April 34.0 14.0 2.0 8.7 7.3 12.0 12.7 9.3 0.8
May 20.7 15.5 9.7 8.4 10.3 9.0 9.0 17.4 0.8
June 20.7 14.7 7.3 6.7 19.3 12.0 10.0 9.3 0.8
July 18.1 21.3 4.5 6.5 18.1 15.5 5.2 10.3 0.6
August 29.0 12.3 1.3 8.4 12.9 15.5 7.1 11.0 2.6
September 23.3 12.0 4.7 8.0 16.0 12.0 11.3 6.0 6.7
October 20.7 22.6 4.5 8.4 14.8 10.3 11.6 7.1 0.8
November 29.3 16.7 1.3 2.0 12.0 12.7 16.0 10.0 0.8
December 26.5 21.3 1.3 3.2 12.3 15.5 14.8 5.16 0.8
Annual mean 25.5 16.3 3.6 6.2 12.8 12.8 12.8 9.3 0.8
Table 3. Monthly and annual frequency (%) of each synoptic weather type at Boston. The weather type of greatest frequency for
each month is bold and the month of highest frequency for each weather type is italic. See Table 1 for abbreviations
CH MH GMR NEH AR FAR FOR–C FOR–M TD
Annual
1995 118 57 10 28 36 50 36 28 2
1996 82 54 17 29 50 46 46 38 4
1997 107 70 11 20 45 44 44 23 1
1998 87 57 13 17 44 51 50 43 3
1999 72 59 14 19 59 42 58 37 5
Summer only
1995 27 15 3 9 9 17 3 7 2
1996 17 13 5 8 20 13 2 14 0
1997 27 21 4 6 11 9 7 6 1
1998 18 10 4 6 14 17 12 9 2
1999 15 15 4 4 23 10 10 11 0
Table 4. Frequency (no. of days) of each weather type at Boston for each year,
and for summer (JJA) only. See Table 1 for abbreviations
CH MH GMR NEH AR FAR FOR–C FOR–M TD
CH 0.40 0.18 0.15 0.09 0.13 0.27 0.38 0.14 0.13
MH 0.13 0.30 0.00 0.09 0.14 0.20 0.16 0.11 0.13
GMR 0.04 0.01 0.15 0.08 0.04 0.01 0.03 0.03 0.07
NEH 0.11 0.09 0.06 0.09 0.02 0.03 0.03 0.03 0.13
AR 0.11 0.19 0.14 0.27 0.19 0.06 0.04 0.12 0.07
FAR 0.07 0.13 0.14 0.18 0.26 0.14 0.05 0.16 0.07
FOR–C 0.07 0.06 0.14 0.08 0.15 0.18 0.19 0.24 0.07
FOR–M 0.08 0.03 0.18 0.12 0.07 0.11 0.11 0.17 0.00
TD 0.01 0.00 0.03 0.01 0.00 0.00 0.01 0.00 0.33
Table 5. Estimated transition probabilities of each synoptic weather type [SWT
(n + 1) / SWT (n)] at Boston, based on 5 yr daily data from 1995 to 1999. The
highest probable transition from one weather type to another is denoted in bold,
and transition to the same weather type is denoted in italics. See Table 1 for
abbreviations
Keim et al.: Manual synoptic climate classification
sition weather type. MH also tends to persist for more
than 1 d, and most frequently the day following a MH
day is another MH. However, the next highest transi-
tion frequency is AR. In essence, these transition prob-
abilities build a common sequence of weather types as
follows: CH > MH > AR > FAR > CH, as the train of
cyclones and anticyclones moves from west to east.
FAR also frequently transitions to FOR–C, then back to
CH. These transition probabilities also demonstrate
that FOR–C and FOR–M tend to transition back and
forth. Note that these are only tendencies in the
weather type transitions. Table 5 demonstrates that—
barring TD—each weather type has been followed by
every other possible weather type at one time or
another, and that continuous predictable sequences
are not common. However, we tested the data with the
χ
2
test for independence and concluded that the data
are not randomly distributed; hence there is an over-
riding structure to the pattern that follows dynamical
considerations (p ≤ 0.05). Sequencing of weather types
in this manner is important for understanding the fre-
quencies and magnitudes of precipitation events (Hay
et al. 1991) and probably has other applications as
well, including interpretation of acute air pollution
episodes.
4.4. Weather type indices
To evaluate weather/climate conditions and envi-
ronmental responses, synoptic weather types can be
combined to form indices (Muller 1977). In this case,
we devised 3 indices: (1) the High Pressure Index
(HPI), which is a combination of CH, MH, GMR,
and NEH; (2) the (WWI) including AR and FAR; and
(3) the (SI), including FAR, FOR–C, FOR–M and TD
(Table 6). On average, Boston experiences HPI
weather on over half of all days (51.5%), which cor-
responds well with Boston’s normal number of clear
and partly cloudy days (55.3%). SI weather occurs on
about one-third of all days, which also corresponds
well with the normal number of rain days. Inter-
annual variations between the 5 years were tested
using the χ
2
test, which showed that the counts of
these 3 indices were significantly different between
years (χ
2
= 18.8, p < 0.05). The years 1995 and 1997
have more HPI and less SI than expected, while 1998
and 1999 have less HPI and more SI than expected.
Given the relatively low HPI (which are generally
cold weather types) for 1998 and 1999, it is not
surprising that these years had the highest average
temperatures. Also, the anomalously high SI in 1998
corresponds with the wettest of the 5 years.
5. APPLICATION TO PM
2.5
CONCENTRATION
Synoptic weather patterns often serve as transporta-
tion mechanisms for pollution and other chemical con-
stituents (e.g. Comrie 1996). One such type of pollution
that has received attention recently is fine airborne
particulate matter ≤ 2.5 µm (PM
2.5
). The scattering and
absorption of short-wavelength solar radiation by
atmospheric aerosol particles directly affects the radia-
tive balance of the Earth. Recent model estimates of
the direct industrial aerosol radiative forcing range
between –0.07 and –1.24 W m
–2
, compared to the esti-
mated range of +2.19 to +2.67 W m
–2
due to the radia-
tive forcing of anthropogenic greenhouse gases (IPCC
2001). One of the main causes of the large uncertainty
associated with direct aerosol radiative forcing is vari-
ations in aerosol properties in both time and space (e.g.
Schwartz & Andreae 1996). The northeastern USA is
one region where direct anthropogenic aerosol radia-
tive forcing is highly variable and typically exceeds
that of anthropogenic greenhouse gases (e.g. Grant et
al. 1999).
Along with its impact on climate by perturbing the
Earth’s radiation budget, the scattering and absorption
of short-wavelength solar radiation by fine particles is
the major cause for reduced visibility in many parts of
the USA (e.g. Malm et al. 1994). Additionally, epidemi-
ological evidence supports a link between adverse
human health effects and fine particles, primarily
because small particles have the ability to penetrate
into sensitive regions of the human respiratory tract
(Dockery et al. 1993).
For these reasons, we conducted a pilot study to
examine relationships between synoptic weather pat-
terns and PM
2.5
mass concentration during the summer
(JJA) of 2001 and 2002; we combined these 2 datasets
to increase our sample size. We chose summer months
because this is the time of year when aerosol concen-
trations are at their highest levels in New England (e.g.
Jordan et al. 2000). This study was set in Durham
(Fig. 1), in the coastal region of New England, 100 km
north of Boston.
151
HPI WWI SI
1995 58.4 23.6 31.8
1996 49.7 26.2 36.6
1997 57.0 24.4 30.7
1998 47.7 26.0 40.3
1999 44.9 27.7 38.9
Mean 51.5 25.6 35.7
Table 6. Synoptic weather type indices (%) at Boston. HPI:
High Pressure Index; WWI: Warm Weather Index; SI: Stormi-
ness Index
Clim Res 28: 143–154, 2005
To measure PM
2.5
, we used a Continuous Ambient
Mass Monitor (CAMM; Andersen Instruments); the
measurement principle is described in detail by Babich
et al. (2000). Different measurement methods of PM
2.5
are outlined in US EPA (1997a) and the standard of 24 h
PM
2.5
measurements is set forth in US EPA (1997b). The
CAMM has a detection limit of 3 µg m
–3
for PM
2.5
con-
centrations averaged over 1 h. To comply with US EPA
guidelines, we averaged 24 one-hour measurements,
as demonstrated by Babich et al. (2000), during each of
the days during the summers of 2001 and 2002 when
synoptic conditions were consistent at both 07:00 h and
19:00 h LST on the same calendar day. The goal of this
practice was to examine PM
2.5
data on days with a rela-
tively consistent synoptic weather type, when impacts
of the weather types are more clear.
This test case revealed differences in the average con-
centrations of PM
2.5
by synoptic weather types (Table 7).
Note that TD weather is not represented, which is an ar-
tifact of the brief sampling period. CH and MH weather
types, on average, contain lower PM
2.5
concentrations
than the other weather types. AR weather contains the
highest concentrations, with FOR–C, FOR–M, and FAR
following close behind, respectively. To test whether
these differences were significant, the Kruskal-Wallis
(K-W) test was used (Keller et al. 1988). Although the
sample sizes are small, the test revealed that there are
significant differences in PM
2.5
that are related to the
synoptic weather types (K-W statistic = 30.21; p < 0.001).
PM
2.5
are presumably advected from transportation
sources along the USA Eastern Seaboard, as well as from
industrial sources in the Ohio River Valley and the Ten-
nessee River Valley. This occurs during AR and FAR
weather types with south and southwesterly winds. Dur-
ing FOR–C and FOR–M, it is likely that a shallow
boundary layer traps PM
2.5
near the surface, leading to
relatively high concentrations. During CH and MH, the
wind is generally from the north and west, bringing
relatively cleaner Canadian air into the region, which is
attributed to the lower density of industrial sources in
southeastern Canada.
Similar results were found by Slater et al. (2002),
who used this classification system in relation to the
chemical composition of PM
2.5
at a rural site in north-
ern New Hampshire for a short sampling period. This
study revealed that during AR and FAR synoptic con-
ditions, PM
2.5
concentrations were higher by at least a
factor of 2, compared to CH conditions. During AR and
FAR transport, ammonium sulfate accounted for most
of the PM
2.5
. During times of air-mass transport from
the north (CH), ammonium sulfate and organic carbon
contributed equally to the PM
2.5
. This is an indication
of the predominance of sulfate in polluted air masses
arriving from the south or southwest of New England,
and the presence of naturally-formed organic aerosols
during time of transport from cleaner sectors to the
north.
6. CONCLUSIONS AND DISCUSSION
New England weather was classified into 9 all-inclu-
sive weather types, using a manual procedure. CH and
MH were the dominant weather patterns at Boston,
while TD, GMR, and NEH weather types were least
frequent, respectively. At 07:00 h LST, the coldest and
driest weather type overall was CH, while the hottest,
most humid weather was generated by AR and FAR.
There were significant differences in PM
2.5
concen-
trations resulting from different synoptic weather
types. This classification system has other applications
in applied climate research. For example, synoptic cli-
mate classifications have been used to analyze surface
ozone concentrations (Comrie & Yarnal 1992), insect
migration patterns (Muller & Tucker 1986), human
mortality (Greene & Kalkstein 1996), heavy rainfall
distributions (Keim & Muller 1992, Faiers et al. 1994,
Keim & Faiers 1995, Keim 1996), tornado outbreaks
(Davis et al. 1997), percent of hourly rainfall totals
(Faiers 1988), fluctuations in anticyclones (Rohli &
Henderson 1997), regional air-mass properties (Muller
& Willis 1983), coastal water level fluctuations
(Childers et al. 1990), evaporation rates (McCabe &
Muller 1987), impacts of ENSO (McCabe & Muller
2002) and climate change (Kalkstein et al. 1990). Fur-
thermore, these weather types have applications in
assessment of forest health and ecology, prediction of
high and low temperature episodes, and dry and wet
deposition of atmospheric aerosols. Jordan et al. (2000)
have found interesting relationships between atmo-
spheric chemistry and regional-scale atmospheric
transport into the New England coastal region. It is
therefore our hope that this synoptic climate classifica-
tion system may be as useful in the northeastern USA
as the Muller (1977) classification has been in the
southeastern USA.
152
Mean ± SD Median N
CH 9.8 ± 6.8 7.3 16
MH 10.7 ± 7.9 6.5 5
GMR 7.9 ± 4.7 7.9 2
NEH 14.7 ± 1.1 14.7 2
AR 27.5 ± 10.1 26.4 17
FAR 18.1 ± 8.2 17.5 5
FOR–C 22.0 ± 9.4 22.6 4
FOR–M 22.5 ± 9.9 19.2 11
Table 7. Summer (JJA) 2001–2002 daily average airborne fine
particle mass concentrations (PM
2.5
in µg m
–3
) in Durham,
New Hampshire, by synoptic weather type
Keim et al.: Manual synoptic climate classification
Acknowledgements. This research was funded by the
National Oceanic and Atmospheric Administration (NOAA),
Office of Oceanic and Atmospheric Research (OAR) Grant no.
NA97RP0309. J.F.S. was supported by the National Aeronau-
tics and Space Administration (NASA) under the Earth Sys-
tem Science Fellowship Program (Grant NGT5-30349). The
authors acknowledge Jane Fithian and Kristi Donahue for
assistance with graphic production.
LITERATURE CITED
Babich P, Wang P, Allen G, Sioutas C, Koutrakis P (2000)
Development and evaluation of a continuous ambient
PM
2.5
mass monitor. Aerosol Sci Technol 32:309–318
Bradbury JA, Keim BD, Wake CP (2002) U.S. East Coast
trough indices at 500 hPa and New England winter cli-
mate variability. J Clim 15:3509–3517
Bradbury JA, Keim BD, Wake CP (2003) The influence of
regional storm tracking and teleconnections on winter
precipitation in the northeastern United States. Ann Assoc
Am Geogr 93:544–546
Childers DL, Day JW, Muller RA (1990) Relating climatologi-
cal forcing to coastal water levels in Louisiana estuaries
and potential importance of El Niño–Southern Oscillation
events. Clim Res 1:13–42
Comrie AC (1996) An all-season synoptic climatology of air
pollution in the U.S.–Mexico border region. Prof Geogr 48:
237–251
Comrie AC, Yarnal B (1992) Relationships between synoptic-
scale atmospheric circulation and ozone concentrations
in metropolitan Pittsburgh, Pennsylvania. Atmos Environ
26B:301–312
Davis RE, Kalkstein LS (1990) Development of an automated
spatial synoptic climatological classification. Int J Climatol
10:769–794
Davis RE, Stanmeyer TM, Jones GV (1997) A synoptic
climatology of tornadoes in Virginia. Phys Geogr 18:
383–407
Dockery DW, Pope CA, Xu X, Spengler JD, Ware JH, Fay ME,
Ferris BG, Speizer FE (1993) An association between
air pollution and mortality in six U.S. cities. N Engl J Med
329:1753–1759
Faiers GE (1988) A synoptic weather type analysis of January
hourly precipitation at Lake Charles, Louisiana. Phys
Geogr 9:223–231
Faiers GE, Keim BD, Hirschboeck KK (1994) A synoptic
evaluation of frequencies and intensities of extreme
three- and 24-hour rainfall in Louisiana. Prof Geogr 46:
156–163
Grant KE, Chuang CC, Grossman AS, Penner JE (1999)
Modeling the spectral optical properties of ammonium
sulfate and biomass aerosols: parameterization of relative
humidity effects and model results. Atmos Environ 33:
2603–2620
Greene JS, Kalkstein LS (1996) Quantitative analysis of sum-
mer air masses in the eastern United States and an appli-
cation to human mortality. Clim Res 7:43–53
Halpert MS, Ropelewski CF (1992) Surface temperature pat-
terns associated with the Southern Oscillation. J Clim 5:
577–593
Hay LE, McCabe GJ, Wolock DM, Ayers MA (1991) Simula-
tion of precipitation by weather type analysis. Water
Resour Res 27:493–501
IPCC (2001) Climate change 2001: the scientific basis. In:
Houghton JT, Ding Y, Griggs DJ (eds) Contribution of
Working Group I to the Third Assessment Report of the
Intergovernmental Panel on Climate Change. Cambridge
University Press, Cambridge, p 289–348
Jordan CE, Talbot RW, Keim BD (2000) Water soluble nitro-
gen at the New Hampshire seacoast: HNO
3,
aerosols, pre-
cipitation, and fog. J Geophys Res 105:26403–26431
Kalkstein LS, Dunne PC, Vose RS (1990) Detection of cli-
matic change in the Arctic using a synoptic climatological
approach. J Clim 3:1153–1167
Keim BD (1996) Spatial, synoptic, and seasonal patterns of
heavy rainfall in the southeastern United States. Phys
Geogr 17:313–328
Keim BD, Faiers GE (1995) Heavy rainfall distributions by
season: synoptic interpretations and quantile estimates.
Water Resour Bull 32:117–124
Keim BD, Muller RA (1992) Temporal fluctuations of heavy
rainfall magnitudes in New Orleans, Louisiana: 1871–1992.
Water Resour Bull 28:721–730
Keller G, Warrack B, Bartel H (1988) Statistics for manage-
ment and economics. Wadsworth, Belmont, CA
Kidson JW (2000) An analysis of New Zealand synoptic types
and their use in defining weather regimes. Int J Climatol
20:299–316
Lamb HH (1972) British Isles weather types and a register
of the daily sequence of circulation patterns, 1861–1971.
Geophys Mem 116
Malm WC, Sisler JF, Huffman D, Eldred RA, Cahill TA (1994)
Spatial and seasonal trends in particle concentration and
optical extinction in the United States. J Geophys Res
99(D1):1347–1370
McCabe GJ, Muller RA (1987) Synoptic weather types: an
index of evaporation in southern Louisiana. Phys Geogr 8:
99–112
McCabe GJ, Muller RA (2002) Effects of ENSO on weather-
type frequencies and properties at New Orleans,
Louisiana, USA. Clim Res 20:95–105
Muller RA (1977) A synoptic climatology for environmen-
tal baseline analysis: New Orleans. J Appl Meteorol 16:
20–33
Muller RA, Tucker NL (1986) Climatic opportunities for long-
range migration of moths: In: Sparks AN (ed) Long-range
migration of moths of agronomic importance to the United
States and Canada: specific examples of occurrence and
synoptic patterns conducive to migration. ARS-43, Agri-
cultural Research Service, Department of Agriculture,
Washington, DC, p 61–83
Muller RA, Willis JE (1983) New Orleans weather 1961–1980:
a climatology by means of synoptic weather types. Mis-
cellaneous Publ 83–1, School of Geoscience, Baton Rouge,
LA
Rencher AC (1998) Multivariate statistical inference and
applications. Wiley, New York
Rohli RV, Henderson KG (1997) Winter anticyclone changes
over the Gulf Coast of the United States. Int J Climatol 17:
1183–1193
Ropelewski CF, Halpert MS (1987) Global and regional scale
precipitation patterns associated with the El Nino/South-
ern Oscillation. Mon Weather Rev 115:1606–1626
Schwartz SE, Andreae MO (1996) Uncertainty in climate
change caused by aerosols. Science 272:1121–1122
Serra C, Fernandez Mills G, Periago MC, Lana X (1999) Win-
ter synoptic weather types in Catalonia (NE Spain) and
their linkage with minimum temperature anamolies. Int
J Climatol 19:1675–1695
Slater JF, Dibb JE, Keim BD, Talbot RW (2002) Light extinc-
tion by fine atmospheric particles in the White Mountains
region of New Hampshire and its relationship to air mass
transport. Sci Total Environ 287:221–239
153
Clim Res 28: 143–154, 2005
US EPA (1997a) Guidance for network design and optimum
site exposure for PM
2.5
and PM
10
. EPA-454/R-99–022,
United States Environmental Protection Agency, Office
of Air Quality Planning and Standards, Research Triangle
Park, NC
US EPA (1997b) National ambient air quality standards
for particulate matter—final rule. United States Envi-
ronmental Protection Agency, Federal Register 62 (138):
38,651–38,760; 40 CFR Part 50
Yarnal B (1993) Synoptic climatology in environmental ana-
lysis. Belhaven Press, London
Zielinski GA, Keim BD (2003) New England weather New
England climate. University Press of New England,
Hanover, NH
154
Editorial responsibility: Robert Davis,
Charlottesville, Virginia, USA
Submitted: August 11, 2001; Accepted: January 4, 2005
Proofs received from author(s): February 15, 2005
