Dr. Kathleen V. Schreiber
Department of Geography
Millersville University
of Pennsylvania
Millersville, PA 17551
Introduction
Tropospheric ozone is a criteria
air pollutant that has long been noted for its negative impacts to human lung
functioning, forest health, agricultural yields, and man-made and natural
materials. Despite decades of multi-faceted urban air quality improvement
programs, ozone air pollution remains problematic for many areas of the United
States. Numerous metropolitan areas are plagued by summertime ozone exceedences,
but the largest continuous area of the eastern US impacted by excessive ozone
levels is well known as the ‘ozone transport region.’ This area includes the
major eastern seaboard cites from Washington, D.C. to Bangor, Maine and is
known for the high ozone levels that are produced as city after city adds
to the pollutant plume typically traveling from southwest to northeast across
the region. Part of the difficulty in controlling ozone is that the pollutant
is not directly emitted to the atmosphere. Rather, it is formed in the atmosphere
through a series of complex reactions involving volatile organic compounds
and nitrogen oxides, reaching high concentrations during specific meteorological
conditions.
Associations between meteorology and specific air pollutants such as ozone have been an important part of atmospheric research for decades. Precipitation events, predominant flow characteristics, and weather elements such as air temperature, humidity, and pressure have been related to the formation, transport, diffusion, and deposition of airborne pollutants (e.g., Mather, 1968; Altschuller, 1978; Hidy et al., 1978; Schreiber, 1996). General synoptic-level characteristics of ozone exceedences in the northeastern US typically include back-of-the high situations in which an anticyclone is centered somewhat to the east of the observation site. This situation usually results in clear skies facilitating photochemical formation of ozone and southwesterly winds which promote ozone transport across major northeastern metropolitan areas (Comrie, 1990).
While this generalized model provides some indication of weather/ozone relationships for the eastern US, it lacks the specificity needed to fully understand meteorological impacts at particular locations. In particular, the model neglects the impact of weather elements, such as air temperature, humidity, cloud cover, and wind speed, which can substantially impact ozone concentrations and may be quite variable within back-of-the high situations. The broad goal of this project is to create a surface climatology of ozone for Lancaster, Pennsylvania to develop a fuller understanding of specific meteorological conditions promoting excessive ozone levels at this location. A computer-automated, ozone-season, synoptic climatological categorization is developed which determines and describes relatively homogeneous meteorological categories based on surface weather elements.
Research
Methodology
Data Acquisition and
Preparation
Ten years (1989-1998) of
hourly surface meteorological observations (18 hours per day) were obtained
for Lancaster Airport from the National Climatic Data Center. These observations
were uploaded to the campus UNIX mainframe computer and quality analyzed.
Missing observations for hours used in the study (6am, 10am, 2pm, and 6pm
of each day) were interpolated from meteorological data at other times in
the day , but only when the period of missing data was six hours or less.
Otherwise, the particular day was discarded. Meteorological variables used
to represent the meteorological character of each day included surface air
temperature, dew point temperature, visibility, total cloud cover, sea level
pressure, and the east-west and north-south components of the wind vector
(measures of wind speed and direction), each four times per day.
Hourly averages of ozone concentrations in parts per billion (ppb) for Lancaster from 1989-1998 were obtained from the Division of Air Quality, Pennsylvania Department of Environmental Protection. As with the meteorological data, the pollutant data were uploaded to the UNIX mainframe and quality analyzed. Missing observations were interpolated over a 4-hour time period; days with greater than 4 hours of missing data were discarded. Finally, the 24 hourly observations were averaged to provide one daily mean ozone value.
Synoptic Categories
To develop a surface synoptic
categorization for Lancaster, principal components analysis and clustering
were applied to ozone season (April – September) meteorological observations.
Principal component analysis (PCA) was used as a data reduction technique,
ultimately reducing redundancy expressed across the 28 meteorological variables
representing each day (7 variables, each 4 times daily). This process rewrites
values of the original variables into “scores” on newly derived orthogonal
components, each of which explains successively less of the variance within
the original data set. Commonly, a number of components will exist which explain
relatively little of the original variance due to multicollinearity among
the original variables. These components are often removed to reduce computational
costs with little loss of information. Another benefit of the procedure is
that it prevents overweighting of highly correlated variables in the determination
of synoptic types. In this process, the 28 variables used to represent each
day were reduced to 11 principal components explaining 92 percent of the
variance (Table 1).
Following PCA, each day was placed into a synoptic category by applying clustering to each day’s component scores. Two clustering techniques were used—average link and k-means. Clustering is a means of partitioning like objects (such as days) to a similar group or cluster (in this case a synoptic category) according to similarities in the variables (meteorological variables) representing those objects. Average link is different from k-means in the method it uses to categorize objects. The average link procedure places objects into groups based upon the average dissimilarity between the object and all other objects in the category. Average link analysis is hierarchical, i.e., once an object is placed in a category, it remains there. K-means is a non-hierarchical technique that allows objects to switch groups as objects are added to the categories and the true data structure emerges. Objects are assigned to the cluster having the closest mean. Subsequently, new means for each cluster are recalculated, and objects which then become closer to a different cluster mean are repartitioned to that cluster. Whereas both of these techniques have been successfully applied in synoptic climatology, in particular applications one will usually perform better than another.
Once clustering was complete, an analysis of variance was performed on pollutant values associated with each of the k-means and average link categories to determine if the categories distinguished significantly different ozone levels, and which clustering method most successfully distinguished categories of poor and high air quality. Finally, the climatic and pollutant characteristics of the synoptic categories resulting from the clustering technique with the best analysis of variance were described.
Results
Clustering Results
and Analysis of Variance
Five warm-season synoptic
categories were identified by the average link cluster analysis (table 2).
However, one disadvantage of average link is its propensity to produce a
few large clusters and many small clusters. Of the 1789 days clustered, 1381
fell into cluster 1. A high proportion of summer days fell into this one
category since very little day-to-day variation exists in summer compared
to spring and fall. As a result, few summer and many spring and fall categories
were produced. This was of particular concern since a subset of this large
cluster contained particularly hot days associated with peak ozone levels.
To distinguish these high ozone days from the remaining days, cluster 1 was
nested, requiring clustering of that cluster. Seven nests of this cluster
were produced (Table 2).
The 11 total resulting cluster means that were produced by the average link procedure were used as initial ‘seeds’ in a k-means clustering analysis of the same meteorological data, and 11 k-means categories were produced. An analysis of variance showed that at least one category differed significantly from other categories in ozone concentrations, both for the k-means and average-link results. However, k-means categories did a superior job in discriminating ozone levels, as shown by its relatively high F value (Table 3). As a result, k-means categories were selected for meteorological analysis.
Climatic Characteristics
of K-means Categories
Means of category meteorological
variables (Table 4, Appendix
1) and Daily Weather Maps (Appendix 2) were
used to determine climatic characteristics of each of the categories. These
results were supplemented with monthly frequencies of each of the categories
(Table 5). Although synoptic scale features typically
associated with each category are presented below, note that the categorization
is based on values of meteorological variables only. Numerous synoptic situations
may potentially be associated with any set of meteorological variables. Category
mean ozone values appear in Table 3.
Category 1 represents a relatively weak, approaching high pressure system centered over the Midwest. A low pressure system is often found over New England, Quebec, or Labrador. The position of these systems together results in strong northwesterly winds over Pennsylvania and mild temperatures. Troughing activity aloft produces northwest upper level winds. A few days in this category are represented as the warm sector of a low pressure system. Although found throughout the warm season, this category occurs most frequently in spring or fall. This pattern results in moderately low mean ozone concentrations for this category.
Category 2 is a cold core polar high centered over the Midwest, Great Lakes, or mid-Atlantic region. It possesses higher pressure and lower temperatures than category 1, and is found only in spring or fall. Commonly showing moderately strong west-to-northwest surface winds and strong upper level troughing activity which produces northwest winds aloft, it is associated with relatively low ozone values. Cold temperatures, moderate radiation levels of spring and fall, and a trajectory over clean Canadian regions reduce ozone formation.
Category 3 is associated with an often deep low pressure system situated over Maine/New Brunswick or off the coast any place from New England (most common) to North Carolina. Cool, strong, surface winds range from northwest (most common) to south, and a deep upper level trough over the southern US brings southerly winds aloft. This spring and fall category of high cloud cover, northwest winds, and cool temperatures is associated with low ozone levels.
Category 4 is unique in that very cold temperatures early in the day are replaced by increasingly warmer temperatures throughout the day. Very strong easterly winds early in the day change into northwest winds that have been somewhat reduced in strength later in the day. This category is represented only by one day; daily weather maps were unavailable for this date.
Category 5 is characterized by weak anticyclonic activity over the southeastern US associated with the expansion of the Bermuda High. Additionally, a low pressure system to the north or northwest of Pennsylvania appears, and a cold front approaches from the west. The back-of-the-high circulation together with the frontal position act to channel west to southwesterly winds from over major metropolitan areas to the south. In combination with high air and dew point temperatures and moderate cloud cover, these pollutant-bearing winds result in the highest category ozone concentrations.
Category 6 represents a recent cold front passage with associated low pressure system to the northeast, often far off the coast of New England. Sometimes a High appears over New England itself. Relatively clean Canadian or Atlantic air is advected over the region as northeast to southeast winds are promoted by additional lows further to the south along the same front or to the west as part of a separate frontal system. A moderately deep trough aloft is centered over the central US, bringing southwesterly 500 mb winds. Warm temperatures, high dew points, and high cloud cover of this predominantly summer category are associated with the second lowest ozone concentrations.
Category 7, associated with anticyclonic domination, shows high pressure centers focused on the Great Lakes, mid-Atlantic, or southeast US. In this category, the High usually has not progressed as far east as in category 5. Small amplitude ridges aloft are associated with the High. For this category, pressures are higher and winds weaker than in category 1. It is found much more frequently in summer than categories 1 and 2. Low wind speeds of varying directions, low cloud cover, and warm temperatures are found with moderate average ozone levels.
Category 8 is a frontal passage, usually involving a north-south oriented cold front which is part of a low pressure center to the north. Although some days show recent passage, most are prefrontal. As expected, low pressure, warm temperatures, high dew point temperatures, and high cloud cover predominate. A moderate amplitude trough aloft brings upper level southwesterly winds. Similar to high ozone category 5, west-to-southwest surface winds and the approaching front may facilitate movement of precursor emissions and ozone from metropolitan areas of the south and into Lancaster County.
Category 9 shows the lowest mean ozone concentrations of all, and consists of a low pressure system often associated with the passage of a short wave aloft. The Low, positioned to the west or immediate south of Pennsylvania, induces an easterly wind flow from off the Atlantic. Cool temperatures, high relative humidity, moderate pressure, and high cloud cover of this predominantly spring category are associated with very low mean ozone levels.
Category 10 represents a departing polar high, centered just to the east of Pennsylvania and associated with small amplitude ridge aloft. This category occurs only in spring or fall. Cool temperatures, high pressure, moderate cloud cover, and light east-to-southeast winds are associated with relatively low ozone concentrations.
Category 11 consists of a high pressure system moving in from the west, and associated with the left side of moderately large trough aloft. Commonly, a cold front has passed, either recently or in the past day or two. A low pressure center is commonly positioned over New England or Quebec. Whereas similar category 1 occurs predominantly in spring or fall, this category more frequently occurs in summer and has higher air and dew point temperatures. It has lower pressure and higher wind speeds and cloud cover than category 7. Northwest winds, warm temperatures, building pressure, and clearing skies are associated moderate ozone levels.
Conclusions
In Lancaster County, a variety
of synoptic categories are associated with both high and low category ozone
concentrations. While some of the produced classes associated with high ozone
levels fit the conventional conceptual model of high ozone-causing meteorological
conditions, some do not.
As expected, conditions of high temperatures, radiation, low wind speeds, and the back-of-the-high southwest circulation (Category 5), providing the necessary energy and emissions critical to ozone formation, show comparatively high ozone concentrations. However, high ozone levels in Lancaster are also associated with cyclonic circulations in which southwest winds and an approaching cold front appear to promote ozone transport from urban centers to the south. Surprisingly, this situation (Category 8) is accompanied by relatively high sky cover, which substantially reduces ultraviolet radiation needed in ozone formation. This provides evidence that transport of previously formed ozone is taking place.
Low atmospheric ozone concentrations in Lancaster County are also promoted by a variety of meteorological conditions. The greatest suppression of ozone levels occurs in conjunction with low pressure systems to the west or immediate south of Lancaster which promote high cloud cover, easterly surface flow, and probable advection of clean Atlantic air over the area (Category 6 and Category 9). However, departing polar anticyclones, which also create easterly surface flow (Category 10), are associated with low category ozone levels.
As shown, conditions outside of the classic back-of-the High are associated with elevated ozone levels in Lancaster County. Additionally, both cyclones and anticyclones are associated with both good and poor ozone air quality. The critical factor for this region of the country appears to be wind direction. The three categories containing the lowest ozone levels all possessed easterly winds and were the only ones to possess easterly winds. The two highest ozone categories possessed southwesterly winds, and also were the only ones to possess predominantly southwesterly winds. This finding suggests that ozone levels in Lancaster County are highly controlled by pollutant transport from source regions to the southwest. Even under conditions of high pressure, stagnation, and low sky cover where precursor pollutants released from Lancaster County would be expected to accumulate and enhance ozone levels, concentrations remain low to moderate without southwest winds.
This study has also shown that individual assessment of specified locations is necessary to fully understand relationships between meteorology and tropospheric ozone levels at those locations. For example the ozone/synoptic category model produced in this paper would probably not work well for cities along the west coast of the US or to the west of major metropolitan areas. Southwesterly winds might promote clean conditions in these locales, while easterly winds would likely enhance ozone levels.
Future
Work
I plan to continue with this
research. I recently received upper level meteorological data, which is useful
in detecting modes of pollutant transport. Addition of it to this analysis
may substantially improve the ability of the categories to discriminate pollutant
concentrations. A consecutive-day analysis of the series of synoptic categories
preceeding high pollutant days may also be useful. Additionally, I will be
performing a discriminant analysis on the ozone data set to separate days
of ozone exceedances from other days based upon differences in the prevailing
meteorological conditions. Once the discriminant function is determined, it
can be used with weather forecast data to predict exceedances. I have also
received daily air quality data for a number of other pollutants, and would
like to extend the analysis to them.
Acknowledgment. The author gratefully acknowledges the Millersville University Environmental Institute and Lancaster Environmental Foundation for its funding and support of this project.
References
Altschuller, A.P. 1978. Association
of oxidant episodes with warm stagnating anticyclones. Journal of the Air
Pollution Control Association. 28:152-155.
Comrie, A.C. 1990. The climatology
of surface ozone in rural areas: A conceptual model. Progress in Physical
Geography. 14:295-316.
Hidy G.M., P.K. Mueller,
and E.Y. Tong. 1978. Spatial and temporal distributions of airborne sulfate
in the United States. Atmospheric Environment. 12:735-752.
Mather, J.R. 1968. Meteorology
and air pollution in the Delaware Valley. Publications in Climatology.
21:1-136.
Schreiber, K.V. 1996. A Synoptic
Climatological Approach to Assessment of Visibility and Pollutant Source Locations,
Grand Canyon National Park area. Doctoral Dissertation, University of Delaware.
Tables
Table
1. Diagnostics for the principal components. The ‘difference’ diagnostic
shows drops in the eigenvalue from component to component. A local maximum
indicates an ideal maximum number of components.
Eigenvalues of the Correlation Matrix: Total = 28 Average = 1Table 2. Clustering diagnostics for the average linkage main and nested clusters. A locally high semi-partial r-square indicates the relative difficulty in merging clusters and thus their unlikeness. The next largest number of clusters is optimal. All diagnostics support the five main cluster solution and seven nested cluster solution.
1 2 3 4 5
Eigenvalue 9.0810 5.0123 4.4300 1.9939 1.3098
Difference 4.0687 0.5823 2.4360 0.6841 0.1803
Proportion 0.3243 0.1790 0.1582 0.0712 0.0468
Cumulative 0.3243 0.5033 0.6615 0.7328 0.7795
6 7 8 9 10
Eigenvalue 1.1295 0.7627 0.6292 0.4869 0.4469
Difference 0.3668 0.1334 0.1424 0.0399 0.0348
Proportion 0.0403 0.0272 0.0225 0.0174 0.0160
Cumulative 0.8199 0.8471 0.8696 0.8870 0.9029
11 12 13 14 15
Eigenvalue 0.4122 0.3349 0.3279 0.2943 0.2348
Difference 0.0773 0.0070 0.0336 0.0595 0.0145
Proportion 0.0147 0.0120 0.0117 0.0105 0.0084
Cumulative 0.9177 0.9296 0.9413 0.9518 0.9602
16 17 18 19 20
Eigenvalue 0.2203 0.2046 0.1661 0.1536 0.1070
Difference 0.0157 0.0385 0.0125 0.0467 0.0319
Proportion 0.0079 0.0073 0.0059 0.0055 0.0038
Cumulative 0.9681 0.9754 0.9813 0.9868 0.9906
21 22 23 24 25
Eigenvalue 0.0751 0.0633 0.0382 0.0322 0.0231
Difference 0.0117 0.0251 0.0060 0.0091 0.0058
Proportion 0.0027 0.0023 0.0014 0.0012 0.0008
Cumulative 0.9933 0.9956 0.9969 0.9981 0.9989
26 27 28
Eigenvalue 0.0173 0.0091 0.0040
Difference 0.0082 0.0050
Proportion 0.0006 0.0003 0.0001
Cumulative 0.9995 0.9999 1.0000
Main clusters:Table 3. Non-Parametric Analyses of Variance of Ozone Concentrations for Synoptic Categories Derived by Average Link and K-means Methods. Categories over 100 in the average link analysis indicate nests of main cluster 1.
Number of Semi-partial Pseudo
Clusters Days R-square R-square Pseudo-F T-square
1 1789 0.000000 0.003958 . 7.100574
2 1788 0.003958 0.004741 7.100574 8.542371
3 1786 0.008699 0.006000 7.836457 10.864440
4 1782 0.014699 0.189085 8.876277 423.299477
5 1381 0.203784 0.001587 114.149495 3.862239
6 401 0.205371 0.043160 92.163031 93.778940
7 2 0.248532 0.000711 98.226147 .
8 89 0.249243 0.007005 84.467389 18.699525
9 312 0.256248 0.005982 76.658726 13.431181
10 1379 0.262229 0.019462 70.257543 49.032910
11 302 0.281691 0.000699 69.725910 1.557915
12 1334 0.282390 0.061581 63.570449 174.559740
13 301 0.343971 0.025643 77.599747 70.429411
14 116 0.369614 0.007713 80.056485 23.457675
15 160 0.377327 0.006087 76.786088 21.405484
Nests of cluster 1:
Number of Semi-partial Pseudo
Clusters Days R-square R-square Pseudo-F T-square
1 1381 0.000000 0.002793 . 3.862239
2 2 0.002793 0.001252 3.862239 .
3 1379 0.004045 0.034245 2.798023 49.032910
4 1334 0.038290 0.108356 18.274673 174.559740
5 160 0.146646 0.010710 59.115099 21.405484
6 1174 0.157355 0.136097 51.353480 265.415581
7 143 0.293453 0.011233 95.111285 26.659101
8 45 0.304685 0.005413 85.949366 11.022521
9 18 0.310098 0.003197 77.086155 9.942767
10 304 0.313296 0.037221 69.499193 75.611323
11 17 0.350517 0.000932 73.936898 1.869789
12 11 0.351449 0.001406 67.441651 3.725205
13 201 0.352855 0.017306 62.158294 39.178524
14 16 0.370161 0.002458 61.799707 6.858370
15 125 0.372619 0.000801 57.950464 1.959675
Average Link:Table 4. Meteorological averages of the k-means clusters.
CATEGORY Days Mean Ozone (ppb) Among MS Within MS
2165.53626 125.073267
2 390 29.6534615
102 139 25.1004317 F Value Prob > F
101 828 35.7939614 17.314 0.0001
103 285 30.9936140
105 16 36.0737500
104 44 35.4725000
4 4 22.2625000
106 1 29.0800000
5 1 31.4200000
3 2 39.2700000
107 1 18.7100000
K-means:
CATEGORY Days Mean Ozone (ppb) Among MS Within MS
5403.66732 106.025438
3 102 29.6920588
10 108 27.6703704 F Value Prob > F
9 154 26.1818831 50.966 0.0001
2 70 28.1468571
1 134 29.8950746
11 171 34.0695906
5 297 41.9929630
7 191 32.4606806
8 214 38.0284579
6 269 26.8276208
4 1 31.4200000
|-temperature-|---dewpoint---|--------pressure--------|--visibility---|----wind direction---|--wind speed--|-sky cover-|Table 5. Frequency of each category by month (first line of category entry) and monthly average category ozone concentrations in parts per billion (second line of category entry). The effect of increased insolation in early summer on ozone levels is apparent for many categories. A concentration of ‘0’ indicates no categories occurred in that month.
Fahrenheit Fahrenheit millibars statute miles degrees miles per hour tenths
time: 06 10 14 18 06 10 14 18 06 10 14 18 06 10 14 18 06 10 14 18 06 10 14 18 06 10 14 18
1 50. 64. 70. 68. 42. 42. 41. 41. 1016. 1017. 1016. 1016. 26. 33. 35. 35. 299. 318. 306. 304. 6. 11. 12. 8. 3. 3. 4. 4.
2 38. 50. 56. 55. 28. 27. 26. 25. 1022. 1023. 1022. 1022. 27. 33. 37. 38. 312. 326. 314. 318. 6. 10. 12. 8. 2. 2. 3. 3.
C 3 48. 54. 57. 54. 41. 40. 37. 34. 1008. 1009. 1009. 1010. 15. 20. 25. 27. 295. 302. 303. 308. 8. 14. 15. 13. 7. 8. 8. 6.
L 4 25. 30. 37. 39. 5. 2. 4. 5. 1019. 1020. 1018. 1018. 22. 30. 30. 30. 100. 320. 310. 280. 52. 23. 23. 18. 5. 3. 0. 3.
U 5 67. 80. 86. 82. 64. 66. 65. 66. 1018. 1018. 1017. 1016. 4. 8. 11. 11. 284. 276. 261. 233. 0. 4. 6. 5. 5. 4. 5. 5.
S 6 65. 71. 75. 73. 62. 64. 65. 65. 1019. 1020. 1018. 1017. 6. 8. 9. 9. 95. 110. 130. 128. 4. 5. 4. 5. 9. 9. 9. 8.
T 7 56. 72. 79. 76. 52. 53. 52. 53. 1022. 1023. 1021. 1020. 17. 22. 28. 28. 342. 22. 274. 205. 1. 2. 2. 3. 2. 3. 4. 3.
E 8 68. 77. 82. 77. 65. 67. 66. 65. 1011. 1011. 1009. 1009. 5. 8. 11. 11. 214. 255. 259. 270. 2. 6. 9. 6. 8. 8. 8. 8.
R 9 52. 58. 61. 59. 48. 50. 52. 52. 1014. 1014. 1012. 1011. 7. 8. 9. 10. 92. 104. 112. 106. 5. 6. 5. 3. 9. 9. 9. 9.
10 43. 57. 63. 60. 36. 39. 39. 41. 1025. 1025. 1023. 1022. 20. 22. 23. 22. 88. 141. 177. 169. 3. 4. 6. 6. 6. 6. 7. 8.
11 64. 74. 78. 75. 59. 59. 56. 55. 1013. 1014. 1013. 1013. 11. 19. 25. 27. 304. 318. 311. 313. 5. 10. 11. 8. 5. 6. 6. 5.
MONTHAppendix 1
CAT. 04 05 06 07 08 09
________________________________________
1 27 50 19 5 7 26
1 32. 33. 32. 28. 26. 21.
2 49 7 0 0 0 14
2 31. 35. 0. 0. 0. 16.
3 62 27 7 0 0 6
3 30. 30. 33. 0. 0. 23.
4 1 0 0 0 0 0
4 31. 0. 0. 0. 0. 0.
5 5 20 40 92 108 32
5 38. 47. 51. 45. 37. 35.
6 7 18 49 66 67 62
6 27. 27. 33. 29. 25. 21.
7 7 29 38 25 46 46
7 37. 38. 39. 34. 28. 25.
8 9 31 59 52 36 27
8 35. 39. 40. 41. 35. 32.
9 64 61 10 3 1 15
9 27. 28. 30. 20. 17. 15.
10 53 32 3 0 0 20
10 29. 32. 32. 0. 0. 17.
11 7 26 39 46 25 28
11 36. 41. 37. 36. 29. 25.
Simple Descriptive Statistics for K-means Categories. Abbreviations for row headings: ST= surface temperature (fahrenheit), STD= surface dew point temperature (fahrenheit), SP= sea level pressure (millibars), SV= surface visibility (statute miles), SUU= surface east-west component of the wind vector (miles per hour), SVV= surface north-south component of the wind vector (miles per hour), SCC= cloud cover (in tenths of sky covered). Numbers following these symbols refer to the time of day: 06= 6am, 10= 10am, 14= 2pm, 18= 6pm, local standard time.
Appendix 2
----------------------------------- Cluster=1 --------------------------------
Variable N Mean Std Dev Maximum Minimum
----------------------------------------------------------------------
ST06 138 50.2 6.9 66.0 43.0
ST10 138 63.6 5.7 78.0 51.0
ST14 138 69.8 5.8 84.0 56.0
ST18 138 67.6 6.6 84.0 54.0
STD06 138 42.4 7.2 56.0 24.0
STD10 138 42.2 7.1 59.0 25.0
STD14 138 41.3 6.8 58.0 25.0
STD18 138 41.3 6.8 58.0 25.0
SP06 138 1016.0 3.8 1024.1 1004.8
SP10 138 1016.9 3.7 1025.4 1006.1
SP14 138 1015.5 3.8 1024.1 1003.8
SP18 138 1015.5 3.8 1023.4 1005.5
SV06 138 26.3 11.8 50.0 0.0
SV10 138 32.7 10.9 50.0 7.0
SV14 138 35.3 10.1 55.0 15.0
SV18 138 34.9 11.2 50.0 15.0
SUU06 138 5.5 3.9 14.8 -5.1
SUU10 138 7.1 6.9 25.1 -10.3
SUU14 138 9.4 5.6 25.1 -4.5
SUU18 138 6.5 5.5 25.1 -11.5
SVV06 138 -3.1 3.4 5.0 -11.0
SVV10 138 -7.9 5.9 14.7 -20.7
SVV14 138 -6.9 5.9 8.0 -23.0
SVV18 138 -4.3 5.9 17.0 -19.9
SCC06 138 2.9 3.1 10.0 0.0
SCC10 138 3.3 2.9 10.0 0.0
SCC14 138 4.3 2.8 10.0 0.0
SCC18 138 4.0 3.0 10.0 0.0
----------------------------------------------------------------------
----------------------------------- Cluster=2 --------------------------------
Variable N Mean Std Dev Maximum Minimum
----------------------------------------------------------------------
ST06 72 38.1 6.4 52.0 27.0
ST10 72 49.6 6.2 63.0 36.0
ST14 72 56.2 6.4 69.0 41.0
ST18 72 54.7 6.6 70.0 37.0
STD06 72 28.1 8.4 49.0 11.0
STD10 72 27.4 8.2 47.0 7.0
STD14 72 26.5 7.6 45.0 8.0
STD18 72 25.4 8.5 46.0 7.0
SP06 72 1022.0 5.5 1035.8 1009.9
SP10 72 1023.4 5.6 1036.9 1014.3
SP14 72 1021.7 5.5 1034.9 1011.5
SP18 72 1021.6 5.5 1034.6 1010.5
SV06 72 27.0 11.0 50.0 10.0
SV10 72 32.7 11.2 50.0 12.0
SV14 72 37.0 11.1 60.0 20.0
SV18 72 37.8 11.3 60.0 15.0
SUU06 72 4.8 4.6 16.1 -8.4
SUU10 72 5.8 6.6 17.6 -9.5
SUU14 72 8.6 6.2 21.6 -5.9
SUU18 72 5.5 5.3 20.7 -9.0
SVV06 72 -4.3 4.5 5.6 -21.6
SVV10 72 -8.7 6.0 5.8 -22.7
SVV14 72 -8.3 5.7 2.3 -23.6
SVV18 72 -6.0 5.9 5.8 -22.7
SCC06 72 2.2 3.2 10.0 0.0
SCC10 72 2.5 2.8 10.0 0.0
SCC14 72 2.7 2.9 10.0 0.0
SCC18 72 2.6 3.1 10.0 0.0
----------------------------------------------------------------------
----------------------------------- Cluster=3 --------------------------------
Variable N Mean Std Dev Maximum Minimum
----------------------------------------------------------------------
ST06 109 47.6 8.8 73.0 28.0
ST10 109 53.8 8.4 71.0 34.0
ST14 109 57.1 8.2 75.0 37.0
ST18 109 53.7 7.1 70.0 36.0
STD06 109 40.8 9.8 63.0 18.0
STD10 109 39.5 8.7 61.0 15.0
STD14 109 37.0 8.2 60.0 15.0
STD18 109 34.4 8.4 53.0 16.0
SP06 109 1008.2 4.7 1018.3 995.3
SP10 109 1009.1 4.8 1019.3 995.3
SP14 109 1008.6 4.9 1020.4 994.6
SP18 109 1009.8 5.1 1023.4 995.0
SV06 109 15.1 9.6 40.0 0.0
SV10 109 20.0 10.3 50.0 1.0
SV14 109 25.2 9.6 50.0 5.0
SV18 109 26.9 10.0 50.0 4.0
SUU06 109 7.2 6.0 20.7 -7.1
SUU10 109 11.5 7.4 27.7 -13.5
SUU14 109 12.4 7.4 27.3 -8.7
SUU18 109 10.2 6.5 22.7 -5.6
SVV06 109 -3.3 5.6 10.3 -19.0
SVV10 109 -7.2 7.0 16.0 -19.7
SVV14 109 -8.0 6.9 11.0 -22.2
SVV18 109 -8.0 5.7 3.1 -25.1
SCC06 109 7.3 3.2 10.0 0.0
SCC10 109 7.8 2.7 10.0 0.0
SCC14 109 7.5 2.6 10.0 0.0
SCC18 109 6.1 3.3 10.0 0.0
----------------------------------------------------------------------
----------------------------------- Cluster=4 --------------------------------
Variable N Mean Std Dev Maximum Minimum
----------------------------------------------------------------------
ST06 1 25.0 . 25.0 25.0
ST10 1 30.0 . 30.0 30.0
ST14 1 37.0 . 37.0 37.0
ST18 1 39.0 . 39.0 39.0
STD06 1 5.0 . 5.0 5.0
STD10 1 2.0 . 2.0 2.0
STD14 1 4.0 . 4.0 4.0
STD18 1 5.0 . 5.0 5.0
SP06 1 1019.0 . 1019.0 1019.0
SP10 1 1019.7 . 1019.7 1019.7
SP14 1 1018.0 . 1018.0 1018.0
SP18 1 1017.6 . 1017.6 1017.6
SV06 1 22.5 . 22.5 22.5
SV10 1 30.0 . 30.0 30.0
SV14 1 30.0 . 30.0 30.0
SV18 1 30.0 . 30.0 30.0
SUU06 1 -51.2 . -51.2 -51.2
SUU10 1 14.8 . 14.8 14.8
SUU14 1 17.6 . 17.6 17.6
SUU18 1 17.7 . 17.7 17.7
SVV06 1 9.0 . 9.0 9.0
SVV10 1 -17.6 . -17.6 -17.6
SVV14 1 -14.8 . -14.8 -14.8
SVV18 1 -3.1 . -3.1 -3.1
SCC06 1 5.0 . 5.0 5.0
SCC10 1 3.0 . 3.0 3.0
SCC14 1 0.0 . 0.0 0.0
SCC18 1 3.0 . 3.0 3.0
----------------------------------------------------------------------
----------------------------------- Cluster=5 --------------------------------
Variable N Mean Std Dev Maximum Minimum
----------------------------------------------------------------------
ST06 317 66.7 5.2 79.0 51.0
ST10 317 80.0 5.0 93.0 57.0
ST14 317 86.0 5.0 98.0 70.0
ST18 317 82.3 5.3 97.0 65.0
STD06 317 63.8 5.2 76.0 46.0
STD10 317 66.4 5.0 78.0 51.0
STD14 317 65.0 5.5 81.0 53.0
STD18 317 65.5 5.3 82.0 50.0
SP06 317 1018.4 3.1 1028.1 1011.5
SP10 317 1018.5 3.1 1027.8 1012.2
SP14 317 1016.9 3.1 1025.8 1010.9
SP18 317 1015.9 3.2 1024.8 1009.2
SV06 317 4.3 4.6 30.0 0.0
SV10 317 7.5 5.0 30.0 1.0
SV14 317 11.0 5.4 30.0 1.0
SV18 317 11.0 5.4 35.0 2.0
SUU06 317 0.4 3.1 9.5 -10.3
SUU10 317 3.5 5.1 16.7 -10.0
SUU14 317 5.9 5.4 17.0 -13.9
SUU18 317 3.7 4.5 21.6 -10.3
SVV06 317 -0.1 1.7 8.5 -8.4
SVV10 317 -0.4 4.2 13.9 -13.9
SVV14 317 0.9 5.2 16.0 -14.0
SVV18 317 2.8 5.6 20.7 -10.0
SCC06 317 4.9 3.1 10.0 0.0
SCC10 317 4.4 3.0 10.0 0.0
SCC14 317 5.3 2.4 10.0 0.0
SCC18 317 4.7 3.1 10.0 0.0
----------------------------------------------------------------------
----------------------------------- Cluster=6 --------------------------------
Variable N Mean Std Dev Maximum Minimum
----------------------------------------------------------------------
ST06 278 64.6 5.4 77.0 49.0
ST10 278 71.3 6.6 89.0 54.0
ST14 278 75.2 6.8 88.0 56.0
ST18 278 73.0 5.7 85.0 57.0
STD06 278 61.9 5.7 74.0 45.0
STD10 278 63.6 5.6 76.0 47.0
STD14 278 64.7 5.3 76.0 50.0
STD18 278 64.9 5.1 76.0 45.0
SP06 278 1019.4 3.7 1030.2 1010.2
SP10 278 1019.7 3.8 1031.5 1010.2
SP14 278 1018.2 3.9 1030.9 1008.3
SP18 278 1017.4 4.1 1029.5 1004.3
SV06 278 5.5 5.6 30.0 0.0
SV10 278 8.0 6.3 30.0 0.0
SV14 278 9.4 6.8 50.0 0.0
SV18 278 9.4 6.2 40.0 1.0
SUU06 278 -3.7 3.8 6.9 -13.8
SUU10 278 -4.3 5.2 9.0 -23.0
SUU14 278 -3.4 5.7 15.0 -21.0
SUU18 278 -3.7 5.2 12.8 -22.7
SVV06 278 0.3 2.9 15.0 -9.0
SVV10 278 1.6 4.7 16.0 -10.3
SVV14 278 2.9 6.1 18.2 -13.9
SVV18 278 2.9 5.4 18.4 -12.1
SCC06 278 8.9 2.1 10.0 0.0
SCC10 278 9.1 1.5 10.0 3.0
SCC14 278 9.1 1.6 10.0 3.0
SCC18 278 8.5 2.2 10.0 0.0
----------------------------------------------------------------------
----------------------------------- Cluster=7 --------------------------------
Variable N Mean Std Dev Maximum Minimum
----------------------------------------------------------------------
ST06 199 56.4 6.4 70.0 39.0
ST10 199 72.3 5.7 86.0 57.0
ST14 199 78.6 5.3 89.0 64.0
ST18 199 75.6 5.6 89.0 61.0
STD06 199 51.5 5.7 67.0 37.0
STD10 199 53.3 5.3 66.0 38.0
STD14 199 52.3 5.2 64.0 36.0
STD18 199 53.1 5.7 68.0 37.0
SP06 199 1022.0 3.5 1031.2 1013.6
SP10 199 1022.6 3.5 1031.9 1012.9
SP14 199 1020.8 3.7 1031.2 1010.2
SP18 199 1020.0 3.8 1029.8 1006.8
SV06 199 17.2 11.5 50.0 0.0
SV10 199 21.5 10.4 50.0 4.0
SV14 199 27.9 9.7 50.0 8.0
SV18 199 27.5 10.7 50.0 4.0
SUU06 199 0.4 3.4 10.8 -9.5
SUU10 199 -0.6 5.8 16.0 -15.8
SUU14 199 1.6 6.2 15.8 -13.8
SUU18 199 1.2 4.7 12.1 -15.8
SVV06 199 -1.2 2.8 6.6 -19.7
SVV10 199 -1.5 4.2 9.0 -11.0
SVV14 199 -0.1 5.9 17.9 -13.2
SVV18 199 2.6 5.5 17.0 -8.9
SCC06 199 2.5 2.9 10.0 0.0
SCC10 199 3.0 2.8 10.0 0.0
SCC14 199 4.2 2.7 10.0 0.0
SCC18 199 3.2 2.8 10.0 0.0
----------------------------------------------------------------------
----------------------------------- Cluster=8 --------------------------------
Variable N Mean Std Dev Maximum Minimum
----------------------------------------------------------------------
ST06 224 67.7 5.7 81.0 51.0
ST10 224 77.1 6.6 93.0 58.0
ST14 224 81.6 7.0 101.0 57.0
ST18 224 77.0 7.3 95.0 55.0
STD06 224 64.7 6.0 76.0 35.0
STD10 224 66.6 5.3 76.0 41.0
STD14 224 66.1 5.3 76.0 50.0
STD18 224 65.4 6.4 76.0 37.0
SP06 224 1011.0 3.8 1020.4 999.7
SP10 224 1010.7 3.7 1018.0 1000.2
SP14 224 1008.9 3.9 1017.0 997.3
SP18 224 1008.8 3.9 1016.3 998.0
SV06 224 5.0 5.1 30.0 0.0
SV10 224 7.9 5.4 25.0 1.0
SV14 224 11.0 6.7 40.0 1.0
SV18 224 11.2 6.8 35.0 0.3
SUU06 224 1.2 4.1 19.2 -7.5
SUU10 224 6.3 5.4 20.7 -9.9
SUU14 224 8.9 6.3 25.1 -8.9
SUU18 224 5.8 5.9 26.0 -15.3
SVV06 224 1.8 4.0 17.0 -6.1
SVV10 224 1.7 6.2 21.0 -14.0
SVV14 224 1.7 7.1 17.9 -14.7
SVV18 224 -0.0 6.4 17.7 -17.0
SCC06 224 7.6 2.8 10.0 0.0
SCC10 224 7.8 2.7 10.0 0.0
SCC14 224 7.7 2.1 10.0 3.0
SCC18 224 7.5 2.6 10.0 0.0
----------------------------------------------------------------------
----------------------------------- Cluster=9 --------------------------------
Variable N Mean Std Dev Maximum Minimum
----------------------------------------------------------------------
ST06 159 51.5 6.3 69.0 34.0
ST10 159 57.5 7.1 73.0 33.0
ST14 159 60.9 7.7 79.0 35.0
ST18 159 59.2 7.3 77.0 34.0
STD06 159 47.9 7.0 66.0 27.0
STD10 159 50.2 6.8 66.0 30.0
STD14 159 51.7 7.1 65.0 28.0
STD18 159 51.9 6.9 65.0 31.0
SP06 159 1014.0 4.8 1026.1 1002.4
SP10 159 1013.7 4.8 1024.8 999.4
SP14 159 1011.8 4.9 1022.0 993.3
SP18 159 1011.0 4.9 1025.8 996.0
SV06 159 7.2 6.8 40.0 0.0
SV10 159 8.3 6.4 30.0 1.0
SV14 159 9.4 7.0 30.0 1.0
SV18 159 10.5 8.7 50.0 1.0
SUU06 159 -4.7 5.3 10.3 -25.0
SUU10 159 -5.4 6.9 16.1 -23.0
SUU14 159 -4.5 7.6 16.0 -21.0
SUU18 159 -3.2 6.8 13.2 -21.6
SVV06 159 0.2 3.9 14.7 -13.0
SVV10 159 1.4 5.3 16.7 -16.7
SVV14 159 1.8 7.6 21.0 -22.7
SVV18 159 0.9 7.4 18.2 -22.7
SCC06 159 9.1 2.0 10.0 0.0
SCC10 159 9.2 1.8 10.0 0.0
SCC14 159 9.1 2.0 10.0 0.0
SCC18 159 8.7 2.4 10.0 0.0
----------------------------------------------------------------------
----------------------------------- Cluster=10 -------------------------------
Variable N Mean Std Dev Maximum Minimum
----------------------------------------------------------------------
ST06 110 43.2 7.4 57.0 27.0
ST10 110 56.9 7.4 72.0 40.0
ST14 110 62.8 7.6 78.0 45.0
ST18 110 59.5 6.9 75.0 45.0
STD06 110 36.3 8.1 52.0 11.0
STD10 110 38.7 8.6 56.0 15.0
STD14 110 39.3 8.9 55.0 19.0
STD18 110 41.3 8.3 57.0 25.0
SP06 110 1025.1 4.6 1036.6 1014.6
SP10 110 1025.3 4.7 1039.0 1013.9
SP14 110 1022.9 4.8 1035.9 1011.5
SP18 110 1021.6 5.0 1034.6 1010.2
SV06 110 20.0 10.3 50.0 0.0
SV10 110 22.1 9.5 50.0 2.0
SV14 110 23.2 9.3 50.0 2.0
SV18 110 22.3 11.2 50.0 1.0
SUU06 110 -2.6 4.2 6.9 -14.0
SUU10 110 -2.3 7.2 15.8 -20.7
SUU14 110 -0.3 7.6 18.2 -18.8
SUU18 110 -1.2 5.5 15.4 -15.8
SVV06 110 -0.1 2.2 9.4 -9.5
SVV10 110 2.8 5.4 17.0 -8.5
SVV14 110 5.7 6.0 21.6 -8.4
SVV18 110 6.3 5.4 17.2 -9.0
SCC06 110 5.8 3.9 10.0 0.0
SCC10 110 6.2 3.5 10.0 0.0
SCC14 110 7.1 3.1 10.0 0.0
SCC18 110 7.8 2.9 10.0 0.0
----------------------------------------------------------------------
----------------------------------- Cluster=11 -------------------------------
Variable N Mean Std Dev Maximum Minimum
----------------------------------------------------------------------
ST06 182 64.4 6.1 79.0 46.0
ST10 182 74.1 6.1 90.0 60.0
ST14 182 78.5 6.1 97.0 57.0
ST18 182 75.1 6.4 89.0 54.0
STD06 182 59.4 6.3 72.0 41.0
STD10 182 58.6 5.8 73.0 44.0
STD14 182 56.2 5.4 75.0 40.0
STD18 182 54.9 5.6 68.0 32.0
SP06 182 1012.9 3.5 1020.0 1002.4
SP10 182 1013.7 3.6 1021.5 1000.7
SP14 182 1013.1 3.6 1021.7 1000.4
SP18 182 1013.3 3.5 1022.0 1002.4
SV06 182 10.6 8.6 30.0 0.0
SV10 182 19.1 9.4 50.0 3.0
SV14 182 25.4 9.9 50.0 7.0
SV18 182 27.2 10.9 50.0 10.0
SUU06 182 3.8 4.5 13.8 -13.4
SUU10 182 6.6 6.4 20.7 -9.0
SUU14 182 8.2 6.5 21.6 -9.0
SUU18 182 5.6 5.7 21.0 -8.5
SVV06 182 -2.6 3.4 5.0 -17.0
SVV10 182 -7.2 4.4 5.8 -21.0
SVV14 182 -7.2 4.6 7.0 -21.6
SVV18 182 -5.2 5.1 8.5 -18.0
SCC06 182 5.3 3.4 10.0 0.0
SCC10 182 5.7 2.8 10.0 0.0
SCC14 182 6.0 2.4 10.0 0.0
SCC18 182 4.8 3.0 10.0 0.0
----------------------------------------------------------------------
Example Daily Weather
Maps Corresponding to K-means Synoptic Categories
The following maps were duplicated,
with permission, from the Daily Weather Maps produced jointly by the National
Oceanic and Atmospheric Administration, the National Weather Service, and
the National Meteorological Center, Climate Analysis Center. Temperatures
are provided in degrees Fahrenheit, precipitation accumulations in inches,
surface isotherms in degrees Fahrenheit, upper-level isotherms in degrees
Celsius, and height contours in dekameters above sea level.
