THESIS ill l‘llllllllllllllllllllllllllllllll 3 1293 106995 .. u..;, v“ , 2%“-“g X”? This is to certify that the thesis entitled "Moisture Stress And Mid-Tropospheric Circulation Patterns Influencing The Forest/Prairie Transition In South Central United States" presented by William T. Corcoran has been accepted towards fulfillment of the requirements for Ph-D- ' degree in (38°92'61th 4 a r professor Date May 15, 1981 0-7 639 A OVERDUE FINES: 25¢ per day per ital RN! I RARY HATER ALS: Place in book return to move charge from circulation records Hull- 5/ u,» r "‘7‘ #61 207. t l €61 E: MOISTURE STRESS AND MID-TROPOSPHERIC CIRCULATION PAITERNS INFLUENCING THE FOREST/PRAIRIE TRANSITION IN SOUTH CENTRAL UNITED STATES By William Thomas Corcoran A DISSERTATION Submitted to Michigan State University in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY Department of Geography 1981 ABSTRACT MOISTURE STRESS AND MID-TROPOSPHERIC CIRCULATION PATTERNS INFLUENCING THE FOREST/PRAIRIE TRANSITION IN SOUTH CENTRAL UNITED STATES BY William Thomas Corcoran The climatic causes of the forest/prairie transition in the south central United States are not fully understood. Westward from Missouri, Arkansas, and Texas, oak, hickory, and pine forest gives way to grassland as the dominant upland vegetation type. Previous researchers have investigated many climatic variables as possible controls on the ecotone: growing season rainfall, rainfall intensity, evaporation, wind, drought, and even winter precipitation, but none of these variables has been shown to be critical in determining the regional vegetation distribution. Mid-tropospheric winds are responsible for the steering of airmasses and storms and thus determine the distribution of climatic variables such as temperature and precipitation. The characteristics and frequency of mid-tropospheric wind patterns controlling the distribution of climate in the south central states have not been adequately investigated. Two climatic variables of particular importance to plant growth are vapor pressure deficit (VPD) and evapotranspiration. VPD is a measure of water stress induced by high temperatures and low relative humidities. Evapotranspiration is a measure of total water loss induced by climate. The major hypotheses of this research were: 1) Mean VPD and evapotranspiration gradients correspond to the forest/prairie transition; 2) Strong VPD gradients across the ecotone are induced by the modal mid-tropospheric windflow pattern; 3) Drought on the Great Plains is a result of increased frequency of the modal mid—tropospheric flow pattern. VPD and Penman evaporation estimates in the central states were mapped. Three T-mode principal components analyses were run on daily 500 mb pressure heights from North America. The first T—mode analysis abstracted 500 mb flow patterns on days having a strong VPD gradient across the study area (called stress days). The second T-mode analysis abstracted 500 mb flow patterns for the whole data set (1961- 1977), and the third T—mode analysis abstracted 500 mb flow patterns for pre-determined drought years during the 1970's. The major findings were: 1) A strong mean VPD gradient exists coincident with the southern portion of the ecotone; 2) Days having the strongest VPD gradients across the southern U.S. had 500 mb flow patterns similar to the modal 500 mb pattern; 3) Drought years during the 1970's did not experience an increased frequency of the modal 500 mb pattern, but did experience increased frequency of the second most frequent pattern; drought years of the 1950's were associated with greater frequency of the modal 500 mb pattern; 4) the T-mode principal components analysis was cumbersome to use with daily data; use of raw pressure height data rather than zonal anomaly values retained excess variance not contributed by wave forms in the 500 mb pressure field, and therefore complicated determination of frequency of the 500 mb flow patterns. ACKNOWLEDGEMENTS Completion of a graduate career requires the cooperation of many individuals, and all deserve thanks. In particular, thanks are due Drs. P.G. Murphy, H.A. Winters and R.I. Wittick, each of whom provided me an opportunity to learn about subjects hitherto unknown to me. Dr. J.R. Harman served as my advisor for four years of study, and provided inumerable insights and opportunities for my growth in climatology and biogeography; without his help my graduate studies would not have been possible. As a teacher, administrator, and as a friend, Dr. Lawrence Sommers contributed significantly to my experience at Michigan State University. Finally, for helping with the frustrating labor of a Idissertation, and for putting up with the ravages of graduate school and an insecure profession, my geographer/wife, Debbie, deserves a special note of appreciation. ii TABLE OF CONTENTS LIST OF TABLES vi LIST OF FIGURES vii Chapter I INTRODUCTION 1 Climate and Vegetation 1 Climate and Circulation Pattern 3 Summary g£_Maior Hypotheses and Expected Findings 8 Chapter II BACKGROUND AND NATURE OF THE PROBLEM 10 The Study Area 10 Introduction 10 Geology, Soils and Physiography 10 Regional Vegetation 15 Generalized Climate 17 Some Explanations gf_the Ecotone 30 Nature of the Problem 35 Chapter III PLANT/CLIMATE VARIABLES 38 Introduction 33 Tree Response £9_Weather 39 Introduction 39 Light and Temperature 39 Water Stress 41 iii Estimation.gf Water Stress 42 Summary g£_Water Stress and Atmospheric Variables 45 Chapter IV ASSUMPTIONS, HYPOTHESES, AND DATA 47 Introduction 47 Assumptions 48 Hypotheses and Criteria for Evaluation 52 .Data 53 Chapter V METHOD OF ANALYSIS 59 Surface Data 59 Stress Day Circulation Pattern Selection 60 Circulation Patterns and Principle Components Procedure 62 Drought Pattern Analysis 69 Chapter VI RESULTS 71 Vapor Pressure Deficit and Evaporation 71 Stress-Day Circulation Patterns 82 LongeTerm Circulation Patterns 90 Drought Circulation 124 Chapter VII DISCUSSION AND CONCLUSIONS 132 Introduction 132 Evaluation of Hypotheses 133 VPD Gradient and the Ecotone 133 Flow Pattern on Stress Days 137 Drought and 500 mb Circulations 139 Mean Summer Circulations 141 iv Principal Components Analysis and 500 mb Patterns Introduction Reality of the Patterns Interpretation of First Components Combined Analysis and Frequency Conclusions Chapter VIII SUMMARY AND SUGGESTIONS FOR FUTURE RESEARCH Summary of Problem, Hypotheses, Methods and Statistical Results Summary of Major Conclusions Suggestions for Future Research APPENDIX BIBLIOGRAPHY 149 149 149 152 156 158 161 161 163 164 169 170 Table LIST OF TABLES Title Moisture Budget for Central Missouri and Central Oklahoma Stress Day Principal Component Statistics Summary Statistics for Combined Principal Component Analysis Summary Statistics for Drought Principal Component Analysis July, 1980 Mean Dewpoint and Maximum Temperatures vi Page 81 89 95 126 145 Figure 10 11 12 13 14 15 16 17 LIST OF FIGURES Title Mean 500 mb Heights in July (1961-1977) Mean 500 mb Heights in a Drought July (1954) Mean 500 mb Heights in a Wet July (1961) Pre-settlement Vegetation of the Study Area Mean Growing Season Precipitation in the Study Area (1931-1960) June-August Precipitation Variables for an East—West Transect June-August Precipitation Variables for a NW—SE Transect July Mean Dewpoint and Precipitable Water in the Eastern U.S. 30 Minute—2 Year Precipitation Intensity Frequency of Precipitation in Four Size Classes, East West Transect Frequency of Precipitation in Four Size Classes, NW>SE Transect Location of Weather Stations Used for Surface Data Location of Stations Used for 500 mb Data (1961-1971) Location of Stations Used for 500 mb Data (1972—1977) June Daily Vapor Pressure Deficit (1957-1977) July Daily Vapor Pressure Deficit (1957-1977) August Daily Vapor Pressure Deficit (1957—1977) vii Page 12 19 21 22 24 26 28 29 54 55 57 72 73 74 Figure 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 Title Summer (June-August) Daily Vapor Pressure Deficit (1957-1977) June Daily Penman Evaporation (1957-1977) July Daily Penman Evaporation (1957-1977) August Daily Penman Evaporation (1957-1977) Summer (June-August) Daily Penman Evaporation (1957-1977) June Stress-Day 500 mb Height Deviation of June Stress-Day 500 mb Height from 1961-1977 Mean July Stress-Day Mean 500 mb Height Deviation of July Stress-Day 500 mb Height from 1961-1977 Mean August Stress-Day Mean 500 mb Height Deviation of August Stress-Day 500 mb Height from 1961-1977 Mean June Stress-Day Component Scores July Stress-Day Component Scores August Stress-Day Component Scores (First Component) August Stress-Day Component Scores (Second Component) Component Scores of First Combined Component Component Scores of August, 1969-1971, First Component Component Scores of July, 1969-1971, First Component Component Scores of June, 1963-1965, First Component June-August Mean 500 mb Heights {1961-1977) Component Scores of Second Combined Component July, 1976 Mean 700 mb Height July, 1976 Temperature and Precipitation Departures viii Page 75 76 77 78 79 83 84 85 86 87 88 91 92 93 94 96 98 99 100 101 103 104 105 Figure 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 Title Component Scores of Third Combined Component Component Scores of July 1961-1962, First Component Component Scores of July 1963-1965, Second Component Component Scores of June, 1969-1971, Second Component June, 1970 Mean 500 mb Height June, 1970 Temperature and Precipitation Departures Component Scores of Fourth Combined Component August, 1962 Mean 500 mb Height August, 1962 Temperature and Precipitation Departures Component Scores of Fifth Combined Component August, 1972 Mean 700 mb Height August, 1972 Temperature and Precipitation Departures Component Scores of Sixth Combined Component June, 1976 Mean 700 mb Heights June, 1976 Temperature and Precipitation Departures Component Scores of First Combined Drought Component Component Scores of Second Combined Drought Component Component Scores of Third Combined Drought Component Component Scores of Fourth Combined Drought Component July, 1980 Mean 700 mb Height Palmer Drought Severity Index for August 30, 1980 Typical August Stress-Day First Factor 500 mb Pattern Typical August Stress-Day Second Factor 500 mb Pattern ix Page 107 109 110 111 112 113 114 115 116 118 119 120 121 122 123 127 128 129 130 143 147 150 151 Chapter I INTRODUCTION Climate and Vegetation Climate is the primary control of world vegetation patterns. Factors such as lithology, soil texture, relief, slope-aspect or competition may govern the local distribution of individual plants and species, but on the world-wide scale climate determines vegetation type. Thus, the terms "Arctic" or "tropical" refer to either the plants or the climate of particular areas, as they are inseparable concepts. Geographers often made use of this relationship by employing vegetation types to derive climatic classification systems (Thornthwaite, 1931; Kendall, 1935; Koeppen, 1936; Ackerman, 1941). A climatic classification system based on vegetation distribution cannot be used to explain the reason for the plant distributions; therefore a further understanding of the relationship between plants and climate is needed, and the elements of a climate that limit the geographic range of a species must be identified. Biogeographers have mapped and correlated the limits of plant species with mean temperature and precipitation (Salisbury, 1926), airmass dominance (Bryson, 1966), bioclimate (Sawyer and Lindsay, 1963), and a moisture index (Thornthwaite, 1948; Mather and Yoshioka, 1968), to name just a few of the climatic variables. Sometimes the relationships derived from this type of correlation/mapping approach have been misleading (Curtis and 2 McIntosh, 1951). In many instances the question remains unsolved: How does climate govern broad-scale vegetation distributions? Plants respond to and are limited by many climatic and meteorologic factors. The separation of plant species on the earth is governed by the distribution of these climatic factors and the tolerances of particular species. Length of growing season, heat accumulation, soil moisture availability, evapotranspiration, relative humidity, and many other meteorologic variables have been shown to be important to plant growth. Different variables may be limiting for different species, and a single species may be limited in different areas by different variables. Given the permutations and combinations possible with thousands of species and hundreds of climatic variables it is almost impossible to identify one small set of factors which are universally limiting in the geographic distribution of vegetation. In any given region, though, some climatic variables are more limiting to plants than others. Spatially consistent variables would not be expected to influence plant distributions, but, on the other hand, merely because a climatic index varies across a region does not necessarily mean that it is important to plant distributions. For a climate- vegetation correlation to be significant the climatic variable must have some physiologic importance to plants. A theoretical or demonstrated plant-climate relationship, studied in a geographical framework, holds promise for explaining vegetation patterns through climate. 3 Climate and Circulation Patterns The long-term means and extremes which are generally thought of as defining the climate of an area are summaries of daily meteorologic conditions such as temperature, relative humidity, or precipitation. Global, long-term (seasonal or annual) meteorologic conditions reflect the state of the atmosphere and the influence of the forcing functions, primarily the receipt and redistribution of solar radiation. Within a more limited focus, the mid-latitudes (30-55 degrees latitude) are a transition zone of energy redistribution between the tropics and Arctic, and experience frequent advection of airmasses from both regions, including airmasses of maritime and continental origin. Conditions are highly dependent on synoptic (continent-wide) circulation patterns, and it is the synoptic airflow patterns at the "steering Level," commonly taken to be 500 millibars (mb), that govern the type, intensity, and duration of airmasses advected over a mid-latitude location. Thus, daily atmospheric conditions such as temperature, relative humidity or precipitation respond to short-term variations of flow at the 500 millibar level as it controls the progression of airmasses and storms over a region. For example, the mean 500 mb height configuration over the United States in July (Figure 1) displays a ridge over the west- central states. The mean 500 mb height pattern during a July with growing season drought on the Great Plains (Figure 2) shows a ridge of higher amplitude displaced farther east than the long-term mean condition. Surface pressure systems are steered by the winds flowing along this middle atmospheric gradient and were thus deflected farther north and east during this dry year. Subsidence beneath with the 'K . Miles kJ r I 0 1000 Heights in Meters Source: NCAR Data Figure 1 Mean 500 mb Heights in July (1961-1977) 5500 5600 5700 5800 Miles I r o 1150 " Heights in Meters Source: Mbnthly Weather Review, 1954 Figure 2 Mean 500 mb Heights in a Drought July (1954) 6 ridge over the Plains dampened passing disturbances, and stabilized the middle layers of the atmosphere; moist air flowing north from the Gulf of Mexico was thus diluted by the subsiding air. As a result, storms were weak and precipitation deficient (Hawkins, 1954). During a wet July on the Great Plains (Figure 3), 500 mb trough formation was more frequent over the central states, deflecting storm tracks farther south, and disturbances frequently crossing the Plains had access to moist air from the Gulf of Mexico in addition to cooler, drier air brought in by the northwesterly flow aloft. The contrasting airmasses, coupled with a less stable atmosphere because of lower middle level temperatures and favorable vorticity advection, created a potential for much stronger storms than in the dry July (Figure 2), and precipitation was ZOO-300% above average over most of the Plains during this wet month (Green, 1961). The precipitation climates of these two Julys were quite different as a result of two different mean atmospheric circulation patterns. Mean atmospheric flow patterns dictate the location of areas of cyclogenesis, anticyclogenesis, airmass movements, and storm tracks. In this way the distribution of meteorologic and climatic variables important to plants can be controlled by the synoptic circulation patterns of the atmosphere. This research seeks to explain the forest/grassland transition across the south-central United States in terms of the regional synoptic climatology of North America. Oak-hickory forest gives way to grassland westward across Arkansas, Missouri, Kansas and Oklahoma. Growing season temperatures and precipitation are relatively homogeneous a . . i '» nu“, E \ 5700 o 5 .37' n “,1. u- I ; 5“"""° ;Ooo-OOo-- .. "-. ' i : ' . O 0 '- 5800 .: 9' 'Ivso...--0o.-..£..... I... .... ‘ '. 5850 1 _.~ 3 ...-...,1 __‘ I. . : z .' fl . $. 5900 ' .w . "I.“ "- ,.;'\‘,. ’"o. Miles I rf 0 1150 Heights in Meters Source: Climatic Data, National Summary Figure 3 Mean 500 mb Heights in a Wet July (1961) 8 across the transition, and previous climatic explanations of the vegetation transition (Borchert, 1950, 1971) only inadequately examined the relationship between stresses on plants and synoptic circulation patterns. The climatic control of this vegetation transition has not been clearly demonstrated. Summary of Major Hypotheses and Expected Findings The research hypotheses for this study are: 1) 2) 3) The 1) 2) 3) 4) 5) The location of the forest/grassland transition in the south central United States corresponds geographically to a gradient of moisture stresses induced by vapor pressure deficits and/or evapotranspiration demands on plants. The orientation and magnitude of the moisture stress gradient are governed by modal upper atmospheric circulation patterns. Drought in the grasslands is associated with increased frequency of the modal circulation patterns rather than with anomalous circulation types. expected results of this study include: Evidence that a vapor pressure deficit and/or evapotranspiration gradient coincides with the ecotone. Verification of the fact that atmospheric data representative of days on which a very strong vapor pressure deficit and/or evapotranspiration gradient occurs across the study area can be used to derive upper atmospheric circulation patterns associated with and causative of the stress gradient. Demonstration that these stress-day circulation types are the modal circulation types of all possible summer atmospheric circulation patterns. Evidence that these modal circulation types increased in frequency during droughts of the 1970's and 1950's. Evaluation of the use of T-mode principle components analysis to summarize the daily circulation of the atmosphere as being composed of particular, recurring flow patterns. 9 The results of this study will aid in furthering the understanding the relation of vegetation boundaries to climate. Individual surface climatic variables will be related to a major vegetation transition, and atmospheric circulation patterns will help indicate whether it is extreme or modal atmospheric conditions that this vegetation boundary responds to. In addition, this study will provide additional documen- tation of the atmospheric circulation patterns important during droughts. Chapter II BACKGROUND AND NATURE OF THE PROBLEM The Study Area Introduction The study area encompasses the transition from forest to western grassland in the south-central United States; parts of the ecotone occur in Missouri, Kansas, Oklahoma and Texas (Figure 4). Most of the main features of the transition can be summarized with reference to Missouri and Oklahoma, and these two states will be referred to frequently as representative of the area as a whole. The area is of interest because the climatic reasons causing this ecotone still are poorly understood, even tough the climate of the North American grassland has been studied intensively, and the northern portions of the ecotone have received some detailed study (Bryson, 1966; Knox, 1972; Davis, 1977; Harman and Braud, 1975; Harman and Harrington, 1977). Geology, Soils and Physiography The transition zone extends westward from the Interior Highlands and southwestern Coastal Plain through the Osage Plains section of the Central Lowland into the Great Plains Province, and includes the Ozark Plateaus. North of the Missouri River lies the substrate consists largely of pre-Wisconsinan age glacial and eolian deposits. Soils tend to be deeply weathered and contain high amounts of clay (Hunt, 1974). The Ozark Plateaus include the Salem and Springfield Plateaus in Missouri, Oklahoma, and Arkansas. Geology, soils, and physiography 10 11 Figure 4 Pre—settlement Vegetation of the Study Area 12 Figure 4 NATURAL VEGETATION OF SOUTH CENTRAL STATES Cour-uni In. Kieth! 1m and In“ 1980 LEGEND 0.0.0 .0. Gnu Transluc- v. f O N c ' an .I l av m v in M Forest IRIS .2. .22. t . t... .- . a... p» :2.» 1.... .. 1. .. 1...: 13 are more complex in this section, as the plateaus are formed mainly on Paleozoic carbonate rocks that dip southwestward from the center of a dome in east-central Missouri. In the center of the dome is exposed an area of exhumed, Precambrian crystalline hills, the Saint Francis (Francois) Mountains. Dome formation was gradual, and today's relief results from river erosion of valleys deep into the resistant limestones. Relief varies from 50-100 meters at the western border of Missouri to as much as 300 meters in the Saint Francis Mountains (Thornbury, 1965). Upland soils are frequently high in clay with large amounts of chert. In the Osage Plains area relief is more moderate, and bedrock is generally of upper Paleozoic age. Soils are variable throughout this area and depend greatly on parent material; sandstone, shale, and limestone are all common bedrocks. Soil type, physiography, and slope-aspect exert strong influences on local vegetation patterns, sometimes interacting in complex ways. For example, Read (1952) found that on north-facing slopes of the Boston Mountains (a rugged area immediately south of the study area) five major tree species (mockernut hickory, black hickory, blackgum, Ozark chinquapin, and red maple) grew only on cherty, sandy soils, while seven more mesophytic species (chinquapin oak, shagbark hickory, white ash, black walnut, sugar maple, hackberry, and redbud) grew only on fine- grained limestone soils. In this instance fine-grained supported more mesophytic species while coarse-grained soils supported more xerophytic Species. 0n south-facing outcrops of silicious limestones in the Ozark Plateaus a xerophytic glade type of vegetation occurs (Kucera and 14 Martin, 1957), while on cherty/sandy soils derived from other strata oak-hickory forest predominates. The silicious limestone soils are fine grained but less deeply weathered, shallower, and hold less water because of their shallower depth, while the cherty soils are more deeply weathered and thus provide a greater overall water supply (Kucera and Martin, 1957). The interplay of physiography, slope- aspect, and geology results in contrasting patterns of vegetation and soils. Greater weathering occurs on north-facing slopes of rugged areas, and deep, fine-grained limestone soils support the most mesophytic species. South-facing, droughty areas of fine-grained limestone soils are less deeply weathered, hold less water, and support the most xerophytic vegetation types. Parent material of soils also plays an important role in determin- ing local vegetation types farther west in the Osage Plains. The general pattern in this less rugged area best illustrates the influence of soils on vegetation patterns in the ecotone; sandstone beds outcrop in eastern Texas and central Oklahoma, and the sandy soils support oak and oak-hickory savannah, called the Cross Timbers (Dyksterhuis, 1948), whereas fine-grained soils in the areas generally support grassland. Soil permeability seems to be the important influence, as Bruner (1931) found the more permeable sandy soils to be wetted to a greater depth by rainfall, while clayey soils had greater runoff, less infiltration and a larger percentage of soil water held at tensions too great to be usable by plants. Physiography directly influences local vegetation patterns in two ways. First, within the ecotone and grassland, tree species are most frequent along streams. Typical floodplain forests extend far westward 15 along major stream channels, while grassland is predominant on the broad, flat uplands (Rice, 1965; Rice and Penfound, 1959). Second, in all areas slope/aspect plays an important role in determining species segregation. North-facing slopes have higher proportions of mesophytic species, greater vegetation densities, and overall less stressful habitats than south-facing slopes (Read, 1952; Bruner, 1931; Rice and Penfound, 1959; Johnson and Risser, 1972). These examples illustrate that soil texture as a critical factor in determining local vegetation type is influenced by slope-aspect, geology, and physiography (although the affinity of some species for calcareous soils and others for more acidic soils is undoubtedly important, too), and it is the interplay of these factors rather than a single factor that determines, along with climate, local species occurrence. Regional Vegetation According to Braun (1950; p. 177) the Forest-Prairie Transition, (Figure 4) "is an area occupied in part by oak-hickory forest, in part by oak and oak-hickory savannah and in part by prairie." The previous section described the influences of soils and physiography on local vegetation patterns. The following discussion will focus on the broad- scale species composition and transition in the study area. Throughout the southern section of the Forest-Prairie Transition (Braun, 1950), post oak (Quercus stellata) and blackjack oak (Quercus marilandica) are the most common_oak species. Post oak is the more numerous and important of the two in the east. Blackjack oak occurs along with post oak on most sites, but is dominant locally only on extremely dry, infertile soils (Brunet, 1931; Rice and Penfound, 1959; 16 Johnson and Risser, 1972). Black hickory (Carya buckleyi) is the usual hickory member of the association. Westward across the transition zone blackjack oak becomes the dominant member in the oak-savannah of Oklahoma and Texas. Bouteloua gracilis is important as a grass member of the understory in the savannah, and large expanses of Andropogon occur in between tree stands (Brunner, 1931; Dyksterhuis, 1948, 1957; Braun, 1950). Tree size declines from east to west in response to the increasingly stressful climate until in western Oklahoma upland trees are reduced to a shrubby cover dominated by shin oak (Q. havardii). Eastward from the Forest-Prairie Transition (according to Braun, 1950), Oak-Hickory Forest becomes dominant on drier, cherty slopes in the Ozarks, and black oak (Q, velutina), yellow (shortleaf) pine (Flaps echinata), winged elm (Ulmus alata) and persimmon (Diospyros virginiana) are frequent associates, with pine a dominant in the Oak- Pine Forest to the South. On more mesic slopes (generally north- facing) black oak, white oak (Q, alps) and red oak (g._£pp£g) are present. On the most mesic, limestone-derived, fine-grained soils, sugar maple (Acer saccharum) is present along with red oak, white oak, chinquapin oak (Q. Muhlenbergii), basswood (Tilia americana) and bitternut hickory (Carya cordiformis). Along the larger streams traversing the transition zone, tongues of flood-plain forest extend far west. Baldcypress (Taxodium distichum) sweet gum (Liquidambar styricaflua), river birch (Betula nigra), black gum (Nyssa sylvatica) and sycamore (Platanus occidentalis) are found in the eastern section. To the west many species drop out, and hackberry 17 (Celtis occidentalis), cottonwood (Populus deltoides), elm (Ulmus americana), and willow (Salix spp.) dominate. (Rice, 1965; Weaver, 1960). According to Bruner (1931; p. 110), broad areas of fine-grained soils within the transition zone (savannah) support tall grass prairie, dominated by species of Andropogon, Panicum virgatum, and Elymus canadensis. Farther west in the true prairie is the Stipa-Koleria association, dominated by species of Andropogon, Bouteloua racemosa, Agropyron smithii, and Sporobolus asper, as well as, "hosts of legumes, composites, evening primroses etc....." To the south, near the intersection of the Arkansas, Louisiana, and Texas borders, is the Oak-Pine Forest type, transitional between Oak-Hickory to the north and Southern Evergreen to the south (Braun 1950). The Oak-Pine region was delimited originally by the presence of shortleaf pine (Pinus echinata) as the common, most dominant pine, with loblloly pine (P, gasps) as a common forest constituent farther south. Longleaf pine (P. palustris) becomes common toward the southeast and south. Pines were originally dominant on the poorer, drier sites, and oak-hickory communities are common in the region. Braun's northern boundary of this forest type coincides with the Coastal Plain/Ouachita Mountains contact. Generalized Climate Court (1973) summarized approximately 300 separate studies of climate and weather in the United States. Except where otherwise noted the following brief description of climate in the study area will draw on this summary. 18 Mean annual temperature decreases northward in the southern United States, reflecting a cold season gradient. Mean January temperatures vary from 70 C in southern Arkansas to -40 C in northern Missouri. July mean temperatures are 25-270 C throughout the south-central states. The average frost-free period ranges from 180 days in Missouri to 210 days in Oklahoma, and to over 300 days in southern Texas. Mean growing season degree days vary from 3200 in northern Missouri to almost 5,000 in the southern Oklahoma/Arkansas area. Bruner (1931) noted that length of the moist growing season in Oklahoma varies from 8 months in the eastern part of the state to 5 months in the west. Summer rainfall has only a slight gradient, but late winter/early spring months are moister towards the east. The probability of extended periods of drought during the summer also increases westward. Mean annual precipitation varies westward across the study area, from over 135 cm in eastern Missouri to 60 cm in western Oklahoma. Summertime (June-August) rainfall varies from 30 cm in Missouri to 23 cm in western Oklahoma (Figure 5; ESSA, 1968). Increasing distance from the Gulf of Mexico generally causes decreasing atmospheric moisture and precipitation, and the greatest extent of the grassland lies north and west of the Gulf of Mexico, suggesting that, in the westerlies, water transport to the west is insufficient to support tree growth. The Gulf of Mexico is the main source of moisture for the eastern two-thirds of the United States (Court, 1973; p. 206); most of this moisture is contained in the lower layers of the atmosphere (below 850 mb) and is directed onto the continent by lower level pressure gradients resulting from almost i y. ! -ui Source: Climatological Data National _ summary ‘, 16 18 20 12 14 ' '0 JUNE-AUGUST (inches) 8 4 Figure 5 Mean Growing Season Precipitation in the Study Area (1931-1960) 20 continuous high pressure over the Gulf of Mexico and Caribbean Sea. Namias (1955) and Namias and Wagner (1971) illustrated the route that moist air takes into the center of North America on summer days when a large continental 500 mb ridge is present over the west central states. Although surface winds may be southerly, moist air flowing northward in the lower layers of the atmosphere is diluted by drier air subsiding beneath the ridge; moist, unstable currents of air are confined to moist tongues, the fringes of the ridge (generally to the eastern states), the immediate coastal area, and a zone of advection ahead of short-wave troughs passing across the country. The location of the upper atmospheric ridge and the strength of travelling continental anticyclones are critical to the invasion of moisture onto the continent (Harman and Harrington, 1978), and Figures 1-3 illustrated how the variation in position of these features influenced precipitation during dry and wet summers on the Great Plains. Summertime daily weather maps during a synoptic situation with a large continental anticyclone dominating the west central states frequently display dewpoint temperatures of 18-210 C (65-750 F) east of the ecotone, but only 7-130 C (45-550 F) on the Great Plains, with maximum temperatures approximately equal in both areas. The difference in "drying power" of air in such a situation is quite large. The number of days with rainfall generally declines from forest to grassland as does mean rain per event and total precipitation in summer (Figures 6 and 7). The relationship is less than simple, as it appears that the Ozark Highlands and Gulf Coast tropical storm rainfall cause a bit of variance in the relationship. The westward decline in rainfall frequency and magnitude could be caused by a westward decrease 21 Number f _. o 30 Number of Days with D ays Precip. With 25 — / Precip. ~‘FF‘-¢”"”" 20 _' and Total 15 '- Precip. (inches) Total Prec1p. 5 — .50 "' Precip. .a"" ‘—-~—.~__ per .25 '— Event 1 ch 8 (n 6) AM OK LR ME NA LSource: Climatic Atlas of U.S. (1931-1960 Data) Figure 6 June-August Precipitation Variables for an East-West Transect AM=Amarillo, Texas OK: Oklahoma City Oklahoma LR= Little Rock, Arkansas ME= Memphis, Tennessee NA: Nashville, Tennessee 22 Number Number of Days with Precip. of 35 __ Days With 30 )—- Precip. L.-"‘ 25 - and Total Total 20 __ Precip. Precip. (inches)15 P- 10 t— 5 i— l.— Precip.’SO per _ lv———-l Event '25 (inches) NP DC OK SH NO Source: Climatic Atlas of U.S. (1931-1960 Data) Figure 7 June-August Precipitation Variables for a NW-SE Transect NP: North Platte, Nebraska DC= Dodge City, Kansas OK= Oklahoma City Oklahoma SH= Shereveport, Louisiana NO= New Orleans, Louisiana 23 in strength of the dynamic uplift mechanisms causing rain, a decrease in the amount of water available to disturbances, or more rapid disturbance movement. No quantitative climatologic measure of dynamic uplift strength or disturbance velocities exist, but measures of atmospheric water vapor have been published (e.g., Dodd, 1965; Lott, 1976). Figure 8 presents a summary of some of these data for July as an illustration of atmospheric water vapor patterns over the central United States. A decrease in mean dewpoint temperature from 700 F (210 C) in central Arkansas to 650 F (180 C) in western Oklahoma corresponds to a decrease in atmospheric water vapor pressure from 24 mb to 18 mb or about 25 percent (Figure 8A). Precipitable water (Figure 8B) integrates the water vapor in the atmosphere over a vertical column from the surface to 500 mb (around 18,000 feet), and shows a 19 percent decrease (4.3 cm to 3.5 cm) in the same distance. These two estimates are in accord, since water vapor content of the atmosphere declines with height, as does the variability, and an integration of the content up to a higher level will be more conserva- tive than surface dewpoint temperatures. Comparison with other monthly maps in these two sources indicates that July is representative of growing season months. There is a difference in atmospheric water vapor content of approximately 20 percent across the study area on each summer day, which leads to lower relative humidity, decreasing precipitation for a given dynamic disturbance crossing the area, and ti decreasing chance of rainfall in the summer westward across the ecotone. One of the characteristics of precipitation in the study area long held to be important in controlling the prairie/forest border was 24 A. (Above) Dewpoint Temperature (0F); Data from Dodd, 1965 B. (Below) Precipitable Water (inches); Data from Lott, 1976 1.25 2.00 Figure 8 July Mean Dewpoint and Precipitable Water in the Eastern U.S. 25 the supposed "torrential" nature of rainfall in the grassland, which encouraged greater runoff and less infiltration than in the forests farther east. In reference to Oklahoma, Bruner stated (1931, p. 141) "westward there is a tendency for rains to become more and more torrential in nature...This results in high run-off." Rice and Penfound (1959; p. 595) echoed the thought: "Furthermore, the rainfall becomes more torrential westward... Neither source offered data or documentation for these statements. The ”torrential" nature of precipitation could imply two attributes of rainfall: intensity and magnitude. Intensity refers to the rate of rainfall or amount of rainfall in a given time period. Figure 9 presents data for the 2 year recurrence interval 30 minute precipitation event, which is the maximum 30 minute rainfall rate expected with a probability of occurrence of once in two years. This event is commonly understood to be a rather intense, relatively frequent storm. The pattern that this event displays across the study area is one of little variation; the 2 year 30 minute rainfall generally decreases to the north and west across the southern U.S. although there is a slight local increase in rate from Missouri to Oklahoma. Other intensity categories display similar patterns across the study area. (Data are available in U.S. Weather Bureau Technical Paper 27). At least at this scale of analysis there is not great cause to believe that grassland storms are more intense than storms in other areas. Westward across the study area the relative proportion of rainstorms in larger size categories (storms greater than 0.5, 1.0, and 2.0 inches in 24 hours) decreases while the relative proportion 26 2.5 Isohyets in inches I Source: U.S.W.B. Tech. Pap.27 (30 Figure 9 30 Minute-2 Year Precipitation Intensity This is the most intense 30 minute rainfall expected once in a 2 year period 27 of storms in smaller size categories (smaller than 0.5 inches in 24 hours) increases (Figures 10 and 11). Along the same transects the total number of days with rain decreases westward. The "torrential" nature of grassland rainfall apparently is not a result of larger magnitude rainstorms occurring westward across the study area. Clearly, there is a problem of perception and statistics associated with the "torrential" rainfall problem. The central United States, including Oklahoma, Missouri, and Kansas, is well known to be an area of frequent severe storms and tornadoes (e.g., Kelly, et a1, 1978). Based on penetration of storm cloud tops into the tropopause, Long (1966) found that the most intense thunderstorm cells in the United States occur in Kansas and Oklahoma, yet, at temporal scales of 24 hours, rainfall intensities and magnitudes appear to decrease westward (Figures 10 and 11). At temporal scales of 30 minutes there is a zone of slightly higher rainfall intensity extending into central Oklahoma (Figure 9). Perhaps at even finer temporal scales precipita— tion intensity may increase significantly westward across the ecotone. Research on the "torrential" nature of very short-lived precipitation events would help resolve this conundrum. The oft-mentioned "torrential" rainfall probably refers to the severe storms which are known to be most frequent in the Oklahoma/ Kansas/Missouri area, including grassland, ecotone, and forested areas. Severe storms are a characteristic of this area and decline in frequency and severity in all directions away from the core area. These storms and tornadoes are the result of a unique juxtaposition of moist and dry air sources as well as frequent dynamic disturbances (Barnes, 1976). The severe storm maximum of the south central states 28 Total Number of Days with 30 r" Precipitation ”’T”" 20 - 80 - Percent 70 _. ”,/”’ of «——-—‘ Precip. 60 .. Days in a 50 " Given Size 40 — Class _———""" '50 30— \ i ‘20.— /\ 21.0 \ 10 - 2:2.0 AM OK LR ME NA 1931-1960 Data Source: Climatic Atlas of U.S.; U.S.W.B Tech. Pap. 27 l Figure 10 Frequency of Precipitation in Four Size Classes, East-West Transect Size classes are inches of precipitation in 24 hours. Weather stations are the same as in Figure 6. 29 t“ - *1 Total Number of Days with 30 _ Precipitation r“"'—— / \ / 20 - Percent ‘ 50 70 r— ' 0f \/ Da 3 y 60 - in a Given 50 __ Size 2.50 Class 40 __ 30 - /\/ 10 -—___/ 32.0 NP DC OK LR SH NO 1931-1960 Data Source: Climatic Atlas of U.S.; U.S.W.B. Tech. Pap. 27 i; Figure 11 Frequency of Precipitation in Four Size Classes, NW-SE Transect Size classes are inches of precipitation in 24 hours. Weather stations are the same as in Figure 7. 30 is a result of the proximity of moist Gulf airmasses and drier airmasses westward, and these same atmospheric moisture variations are responsible for generally decreasing precipitation and relative humidity westward. It is the overall decrease in atmospheric moisture westward that is most likely responsible for the vegetational change. Severe storms are a dramatic attribute of the climate in the central United States, but neither the torrential rainfall from these storms nor its possible importance to the ecotone has been adequately assessed. These storms most probably are merely associated with the climatology of the area rather than causative of the ecotone. Some Explanations of the Ecotone Bruner (1931), in seeking to explain vegetation patterns of Oklahoma, attributed the forest/grassland transition to the following: 1) Eastern Oklahoma has 8 months (March-October) of very moist growing seasons, while western Oklahoma has only 5 months (April-September). 2) Periods of extended growing season drought are more frequent and severe in western Oklahoma than in the eastern part of the state. 3) Relative humidities during 2 representative growing seasons averaged 50 percent in the prairie and 63 percent in the savannah forest. 4) Increasing evaporation southward and westward across the state causes precipitation to be less efficient in providing water for plants. Weekly evaporation for 2 growing seasons averaged 16.7 cm in the savannah forest and 39.3 cm in the prairie. In addition, Bruner (1931) noted all of the other climatic gradients usually associated with the transition zone--winter temperatures, annual 'precipitation, wind, and sunlight, but chose to focus his explanation 31 of the transition on growing season moisture availability and demand. Although most of his meteorologic measurements encompassed only 2 years (1924-25), and his climatic statistics were from 1895-1914, they are consistent with currently published data both in trend and in magnitude. Transeau (1935) attributed the maintenance of the Prairie Peninsula, northeast of the study area, to lower annual precipitation, higher probability of drought, and lower relative humidities, probably with increased evaporation, over the grassland. He also recognized the importance of fire in maintaining prairie in an area where woody vegetation actively colonized grassland. These two relatively early studies summarized in numerical, graphical, and map form most of the climatic variables that have been hypothesized to be of importance in causing the forest/grassland transition, or that have been correlated with the transition in empirical attempts at determining the climatic cause. Borchert (1950) re-iterated the importance of drought and fire in the maintenance of grassland, and paid special attention to winter precipitation as a principle control for the southern margin of the grassland. In addition, Borchert added a climatic circulation variable--the dominance of "westerly" airmasses, derived from windflow over the western cordillera--as being the cause of summer drought and lack of winter precipitation in the grassland. Bryson (1966) illustrated the annual mix of airmasses over North America, and correlated the southern grassland boundary with a "mean frontal zone." Grassland to the west of the ecotone was dominated for 32 more than 50 percent of a year by mild Pacific and continental tropical airmasses, while forest to the east of the transition was dominated by maritime tropical air from the Gulf of Mexico. Borchert (1971) attempted to correlate drought years on the Great Plains with an index of climatic circulation type derived by Dzerdzeevskii (1969). Borchert's conclusion was that the 3 major droughts of the 1900's occurred during times of highly meridional circulation types, highly zonal circulation types, and during times of no predominance of zonal or meridional circulation types. Although drought is common in the grassland, it is not a simple phenomenon, easily explicable through gross circulation indexes. These airmass studies attempted to explain the ecotone as a result of airmass frequencies, where the vegetation zones were associated with their own particular dominating airmass originating over either the Gulf of Mexico (forest airmass) or the Pacific Ocean (grassland airmasses). Grassland airmasses lost their moisture over the western mountains and were therefore dry, with little resulting precipitation. Airmass studies of paleoclimatic and paleobotanical nature in other areas (Webb and Bryson, 1972; Kay, 1979) correlated airmass type with climatic characteristics, such as mean July rainfall and temperature in an attempt to derive climatic estimates along with airmass dominance estimates, but no such association has ever been published for the southern transition zone. The airmass studies remain descriptions of airmass dominance. In an even more complex type of analysis, Thornthwaite (1948) and Thornthwaite and Hare (1955) derived a moisture index and an overall 33 moisture/evaporation index that displayed a strong gradient coincident with the southern states ecotone. Physical interpretation of the indexes proved difficult, since they were trial and error combinations of rainfall, temperature, latitude and empirical constants. Mather and Yoshioka (1968) were able to quantitatively group the vegetation types of North America using a moisture Index (Thornthwaite's) and precipitation/evaporation ratios. Although the moisture index is, again, too complex to interpret, their results suggest the forest/grassland transition is chiefly due to a balance between precipitation and evaporation. All of these studies emphasized the role of climate in determin- ing the vegetation boundary. An opposing point of view was voiced by Wells (1970). He felt that virtually all of the pre-settlement grassland would support upland forests, and that the grassland owed its existence to frequent fires. Wells did not rule out climate as a determinant, but emphasized that they way in which a droughty climate wreaks destruction upon trees was through fires. In his view, the relative flatness of the Great Plains area could allow drought-induced prairie and forest fires to periodically sweep the plains clear of trees. Only areas of rugged topography or (much farther east) very moist climate would offer protection to trees. Without attempting to resolve a fire versus climate dichotomy, it should be noted that drought conditions are destructive to trees in the oak-hickory savannah (Albertson and Weaver, 1945), and need no help from fire to eliminate trees. And as Wells (1970) pointed out, a droughty climate is important ‘to the grassland because it encourages fires and provides conditions 34 unfavorable for tree competition with grasses. Schnell gp_al, (1977) analyzed the distribution of tree species and environmental variables in Oklahoma using principle components analysis. They hypothesized that moisture (precipitation and soil moisture) would be the environmental factors best explaining the distribution of trees. Their results, however, showed precipitation to be relatively unimportant, while soil texture and air temperature were most closely associated with tree distributions. Presumably, temperature relates to relative humidity and evaporation (these 2 variables were not included in the analysis). Their results summarize very well the long history of debate on reasons for the southern forest/ prairie transition. Many variables relating to moisture (precipita- tion amounts and distribution, relative humidity, evaporation, drought) have been proposed and examined as prime causes (along with fire), but none have been shown to be undeniably causative or critical in determining the vegetation distribution. Within the ecotone itself site conditions determine whether the local vegetation is forest, savannah, or grassland (Bruner, 1931; Rice and Penfound, 1959; Johnson and Risser, 1972), but its precise location is not a reflection of changing parent material. From east to west across Oklahoma tree height, basal area and frequency decline (Rice and Penfound, 1959). Local species segregation and growth-habit separation can be attributed to soil variation, but the general decline in tree size and frequency in forested stands indicates an overall control by climate, probably through moisture variables. 35 Nature of the Problem Two facets of the climate which cause the forest/grassland transition in the south-central United States remain unresolved: 1) Which climatic variables are primarily responsible for the transition? 2) What role do atmospheric circulation patterns play in determining the daily distribution of the important variables? Previous studies have been unable to identify one climatic variable as being critical to the transition, but perhaps more than one variable is important. Attempts to explain the ecotone in terms of composite indexes (e.g., Thornthwaite, 1948) have failed because the indexes are too cumbersome to interpret in physical terms. Correlation studies (e.g., Borchert, 1950) have provided a multi- plicity of variables as being of possible importance. Correlation studies can be misleading (e.g., Armstrong, 1967). Borchert (1950) identified winter precipitation as a variable having a strong correlation with the southern ecotone. He used this correlation for the basis of his theory on the importance of westerly airmasses, which are strongly dominant in winter. This season is the driest period in the grassland, and Borchert's analysis seemed to indicate that westerly airmasses were more important during drought years. There are two problems with this reasoning. First of all, Namias (1955) has illustrated that dry air dominating the Great Plains during drought years is the result of a large, semi-permanent anti- cyclone. The dry airmasses are the result of subsidence and warming of the lower and middle layers of the atmosphere. This subsidence 36 and stability collectively dampen both passing disturbances and convective overturning during the afternoon hours, thus decreasing precipitation, while the warming, drying air has very low relative humidity, which increases evaporation. The second problem with Borchert's (1950) rationale was in the selection of winter precipitation as a critical variable for the ecotone. Bruner (1931) pointed out that well developed oak-hickory forest occurs in eastern Oklahoma. Winter precipitation averages only 5—10 cm more in eastern Oklahoma than in the western part of the state, so it would be hard to attribute the transition from oak-hickory forest through oak-hickory and oak savannah to grassland in western Oklahoma to a winter precipitation gradient of this magnitude. It was precisely this problem which led Bruner (1931) to examine other climatic variables in seeking to explain the transition. If correlative studies have been misleading (at times), airmass and circulation studies (e.g., Byson, 1966; Borchert, 1971) have been disappointing. They have merely hinted at the possible controlling mechanisms in the general circulation and have perhaps incorrectly interpreted the circulation patterns that lead to dryness on the Plains. In addition, airmass studies have not demonstrated the relationship between airmass characteristics and their influence on plants. Whether one or many climatic variables determine the ecotone or whether there is a controlling atmospheric circulation pattern remains to be resolved. One way at least to begin the determination of important climatic variables is to examine meteorologic variables that have been demonstrated to be important to plant growth. Variables 37 which are critical to the physiologic functioning of plants can be examined in their spatial variability, and those that vary in a manner coincident with the ecotone can be examined further. Once the significant variables have been identified through physiology studies and mapping, then circulation patterns associated with them can be identified. This approach is still close to the correlative method, but relies on the work of plant physiologists to indicate which climatic/meteorologic phenomena to investigate. The approach is to start with established theory on the influence of weather on plants, and examine the spatial variability of conditions which are known to be physiologically limiting to plant growth. Chapter III PLANT/ CLIMATE VARIABLES Introduction Plant growth is influenced by many meteorologic factors. This section will review briefly some of the critical plant/atmosphere relationships, paying particular attention to those variables that are intimately controlled by daily weather patterns. Light, temperature, and water relations will be considered, with principle focus given to the effects of water stress and ways in which water stress is induced in plants. Finally, meteorologic variables important to water stress in the study area will be suggested, and methods of estimating or measuring these variables will be reviewed. Throughout the discussion emphasis will be placed on tree growth and response, since the ecotone is a result of westward increasing stress on life forms of all types. Thus, trees will be considered in terms of atmospherically induced stress, and grassland will be considered as a group of alternative life forms, better adapted to the stresses which cause the ecotone. Extensive literature is available on environmental control of plant growth (e.g., Evans, 1963). A recent summary of research with special emphasis on conifers is given by Helms (1976). Reviews of water/plant relationships of particular thoroughness are Slayter (1967), Kramer (1969) and Kozlowski (1968). Kozlowski (1979) summarized research on the influence of environmental stresses on tree growth. Most tree research concerning the physiology of water stress has been done on cxnnifers; relatively little research concerning water-stress relations of hardwoods is available (Federer and Gee, 1976). 38 39 Tree Response to Weather Introduction Most research on plant growth/environment relations focuses on photosynthetic performance, generally, photosynthetic rate and net primary productivity. Both of these quantities are strongly influenced by factors other than the environment, including species of plant, leaf area of the plant, stage in life cycle, and type of carbon fixation (grouped as C4 or C3 plants). In addition, other environmental variables such as soil type or fertility influence plants in complex ways that may inhibit or enhance photosynthetic performance. Thus, a perfect correlation of plant growth with meteorologic factors cannot be expected. But over such a wide area as the forest/grassland transition the gross pattern of species occurrence and plant productivity will be largely controlled by atmospheric variables. For example, Rosenzweig (1968) and Lieth and Box (1972) successfully modelled terrestrial ecosystem productivity using long-term climatic estimates as predictors. Light and Temperature Light and temperature are broadly limiting to photosynthetic capacities. Both are necessary in the proper amounts. In mature tree crowns where mutual shading of leaves occurs, photosynthesis may increase with light intensities all the way up to full sunlight, although exposed foliage may reach its saturation point much before (Kramer and Kozlowski, 1960). High light intensities are often associated with higher temperatures. Larcher (1969) found photosynthe- tic rates in conifers to be maximum at 19-250 C, with a range of -5 40 to 400 C being the effective range for photosynthesis to occur. Helms (1976) found Douglas fir to have net rates of photosynthesis at a given light intensity twice as high in early winter as in summer. Higher light intensities were coupled with higher temperatures and vapor pressure deficits in summer which lowered net photosynthesis due to both increased respiration and transpiration and closing of stomata during water stress. One confounding variable in separating light and/or temperature controls from other influences on plants is the type of carbon fixation utilized by the plant. Plants having the C4 type of carbon fixation as well as the C3 pathway are able to fix C02 at lower concentrations than CB plants, can use water more efficiently, have higher light saturation points, have higher optimal temperatures for photosynthesis, and lack photorespiration (which uses energy during daylight hours). C4 plants commonly grow in environments which have higher light intensities, temperatures, and water stress potentials. Thus, C4 plants seem to be better adapted to competition in stressful environments. All trees have only CB carbon fixation, and as a result are probably less well adapted to extremes of climate than typical grassland species (Black, 1971; Hatch, et a1, 1971; Dickman, 1973; Moore, 1974). On an annual basis, light intensity and duration are controlled by seasonal earth/sun relations. The seasonal distribution and magnitude of maximum temperatures are also largely controlled by earth/ sun relations. Within a shorter period of time such as a season or Inonth, however, weather influences light and temperature in a number of *ways, particularly through advection of airmasses from warmer or cooler areas and the generation of cloudiness by these large-scale atmospheric 41 movemen LS . Water Stress One of the main difficulties in determining the response of photosynthesis to temperature (and light) is that as plant and air temperatures increase, transpiration demands on plants also increase, frequently inducing water stress. Water stress occurs when transpirational losses through leaves exceed soil-water absorption by roots. Cell turgor decreases and the physical and chemical functioning of the plant may suffer. Net photosynthesis is affected because the stomata close during periods of water stress in order to reduce transpiration, and CO2 uptake is thereby limited. The main effect of water stress, then, is to limit assimilation and thus decrease growth, perhaps weakening the plant if photosynthesis cannot keep up with respiration demands. Severe water stress may make a plant susceptible to disease and/or limit recovery after a long drought, decreasing its competitive ability (Jarvis and Jarvis, 1963). Leaf-water potential is a common measure of plant water stress. The potential is a negative pressure resulting from excesses of transpirational demands over root absorption (part of the potential also comes from physical resistance to flow within the plant). Stomata have been shown to close completely at leaf-water potentials of -5 to -10 bar in some hardwoods, and at -19 to -25 bar in some conifers (Levitt, 1972; Lopushinsky, 1969). In comparison, Bouteloua gracilis, a common grassland species, experienced no decline in photosynthetic rates at leaf-water potentials of -44 bar, and assimilation was reduced by only 50 percent at -60 bar (Dyer and Trlica, 1972). In general, 42 needle-leaf trees are more resistant to water stress than deciduous trees, and grassland species are more tolerant than trees. Estimation of Water Stress A complete modelling of water stress conditions would include both soil and atmospheric components, as well as other factors such as resistance to water flow within the plant. The most difficult component to measure and model seems to be root absorption (Ritchie, 1974). Root absorption depends not only on transpirational demand but on soil water supply, which in turn depends on precipitation, evaporation, and soil texture. Zahner and Stage (1966) were able to account for 72-76 percent of tree growth in two pine species (Pippa resinosa and Pinus monticola) using a bookkeeping approach to calculating soil moisture based on soil texture, precipitation, and potential evapotranspiration demands. The soil moisture model they used requires estimates of soil textures and precipitation intensity/ duration, however, and is not amenable to regional, long-term applications. Soil water supply is very difficult to estimate for large areas over long periods of time. The driving force for transpiration is the vapor pressure gradient from leaf to air (Helms, 1976; p. 67). During periods of high temperature and low relative humidities, even trees in nearly saturated soil may undergo severe water stress because of excess transpiration (Kozlowski, 1979). Leaf shrinkage during the day indicates water stress caused by high transpiration losses and has been related to high vapor pressure deficits (Chaney and Kozlowski, 1969; Federer and Gee, 1976; ‘Natir and Kousalova, 1965). Vapor pressure deficit (VPD) is the 43 difference between saturation vapor pressure of the ambient temperature and saturation vapor pressure of the dewpoint temperature, and thus describes the "drying power" of air: VPD= (ea - ed) (1) where ea is saturation vapor pressure of air at a temperature Ta (ambient temperature), and ed is the saturation vapor pressure of air at a temperature Td (dewpoint temperature). VPD is the result of ambient temperatures higher than dewpoint temperature, and is inversely related to relative humidity. The relationship between air temperature and saturation vapor pressure is non-linear. Between 00 C and 400 C each 10° C rise in temperature leads to a doubling of water holding capacity. During mid-day maximum temperatures, high vapor pressure deficits can induce low leaf water potentials and cause stomatal closure. As a result, net photosynthesis declines linearly with increasing vapor pressure deficit (Helms, 1976). Plant water stress depends on a number of factors, including soil water availability, but vapor pressure gradient is the main atmospheric control on transpiration and water stress, even to the point of being a strongly limiting factor in net photosynthesis. To the extent that high light intensity and tempera- tures influence vapor pressure deficit, this variable includes some of the effects of temperature and light on plants. Federer and Gee (1976) found that 76 percent of the variation in leaf-water potential in unstressed hardwoods was related to vapor pressure deficits. Ritchie (1971) found that 56 percent of seasonal transpiration in western conifers could be explained by vapor pressure deficits alone. This ‘variable is of principle importance in inducing water stress in plants. 44 Other more integrative approaches have also been used to estimate atmospheric water demand on plants. These approaches are best used as large-scale, long-term estimates of atmospheric conditions. Calculation and measurement of water demand has focused on a number of quantities including evaporation, potential evaporation, evapo- transpiration, potential evapotranspiration and latent energy. (A good description of the history, derivation and application of the different approaches to calculating these quantities is available in Rosenberg et a1 (1968).) Two of the more common methods used to estimate atmospheric water demand were developed by Thornthwaite (1948) and Penman (1948). The Thornthwaite method uses an empirically derived set of relation- ships between temperature and daylength to derive potential evapo- transpiration estimates (e.g., Major, 1963; Daubenmire, 1954; Lieth and Box, 1972). It is somewhat inaccurate when applied to daily data (Pelton et a1, 1960). Rosenberg (1974) felt that the success of the Thornthwaite method resulted from the strong "autocorrelation" of evapotranspiration and temperature as similar functions of net radiation. The Penman equation combines vapor pressures, sunshine, net radiation, wind speed, and temperature measures to give estimates of actual evaporation or potential evapotranspiration. Lincare (1977) devised a simplified form of the Penman equation for evaporation, and found it to be accurate world-wide for estimating evaporation on a daily, monthly, or seasonal basis. Evaporation or evapotranspiration are of interest because they integrate the effects of meteorological variables that induce water 45 stress in plants. Vapor pressure deficits, light, temperature, and wind are all taken into account by evaporation estimates. Conditions at some time during a day or week may be stressful to plants, but the stress may not be long-term. Vapor pressure deficit is calculated for the maximum temperature during the day, but it is not integrated over the whole range of temperatures in a day, and thus indicates only momentary stress rathern than total environmental water demand. The Penman estimate of evaporation, though, gives an estimate of long-term atmospheric water demand, and acts as a check on short-term estimates of stress such as vapor pressure deficit. Together these variables provide a good indication of atmospheric water-stress on plants. Summagy of Water Stress and Atmospheric Variables Plant response to environment is complex, and depends on genetic, soil, and climatic factors. Of the climatic factors, water stress is one of the most seriously limiting conditions. Grasses are more tolerant of water stress than trees. The most important contributor to daily plant water stress is vapor pressure deficit. Atmospheric water decreases westward from the ecotone, principally as a result of subsiding air beneath a continental 500 mb ridge. Observation of daily weather maps suggests that on days having a 500 mb ridge located over the south central United States, a strong gradient of dewpoint temperature is present from west to east, inducing a vapor pressure deficit gradient across the ecotone, and since the use of dewpoint temperature in computing vapor pressure deficit directly addresses water content of the atmosphere, it is logical to choose vapor pressure deficit as the principle variable with which to analyze 46 water stress in a geographic context. Penman estimates will be used as supportive evidence, principally to describe the role of evaporation in influencing overall water budgets. Chapter IV ASSUMPTIONS, HYPOTHESES, AND DATA Introduction For analytic purposes it is necessary to consider time periods smaller than a year, even though partitioning a year into unrelated seasons is not entirely possible where plant growth is concerned; all seasons can in some way influence plants. Tree growth responds to many different and interrelated factors which are auto-correlated and dependent on previous conditions. Fritts, et a1, (1979) described climatic anomalies in the United States using tree ring chronologies of western conifers, and found that ring-width-derived maps of precipitation anomaly best approximated reality for a spring period of March—June; winter had the next best calibration, followed by summer and autumn; these are the months receiving greatest precipitation in that area. In the southern portion of the study area the growing season commences in early spring, and from a plant physiology point of view perhaps that period April-September would best approximate the total growing season, but maximum rainfall and peak temperatures occur during June-August. Standard meteorologic seasonal divisions were used in this research in order to limit the investigation to a single meteorologic season, and June was considered a summer month; atmospheric circulation patterns in June may not be identical to July and August, but June has more affinity with these months than with March, April and May (Namias, 1952). 47 48 Assumptions Three major assumptions will be made in order to simplify the plant/climate link. The first assumption is that growing season moisture stress is more important in limiting tree growth across the ecotone than stress or conditions during other seasons. The large winter precipitation gradient across the ecotone has been cited by Borchert (1950) as being an important variable influencing soil moisture, but it is the daily growing season weather that this study seeks to relate to plants and vegetation. It is not meant to imply that soil moisture budgets or antecedent precipitation conditions are unimportant to maintenance of the ecotone, but merely that summertime stress is critical. Summer was considered as the season important to the ecotone because of observation of summertime weather maps which frequently depicted a strong gradient in dewpoint temperatures across the study area. Dewpoint temperatures influence not only VPD but also indicate the amount of water available to precipitation processes; there is, in fact, a strong correlation between precipitation amounts and dewpoint temperature. The second assumption is that the pre-settlement ecotone as mapped by Kuchler (1964) and described by Braun (1950) can be considered generally reflective of climatic conditions prevailing both then and now. This assumption implies that climate has not changed radically since European settlement, and that the vegetation of pre-settlement times was in equilibrium with climate. That climate has changed radically since pre-European settlement is clearly untenable. Mean temperatures and precipitation have fluctuated since settlement (Wahl, 1968; Wahl and Lawson, 1970; Lawson, 49 1974), but the changes have been of relatively small magnitude. Some periods of 10 to 30 years show statistically significant difference of means (e.g., Wahl and Lawson, 1970; Diaz and Quayle, 1980), but the means appear to fluctuate about a longer-term mean in a quasi-periodic fashion, and when longer periods (50 years) are considered, statistical significance in difference of means declines (e.g., Skaggs, 1978). Wahl and Lawson (1970) argued that the current climatic regime in the United States is more representative of the climate during settlement of the study area (1830-1850) than was the period 1931-1960. Adequate tree-ring data from the area (Estes, 1970) and farther west (Weakly, 1965; Fritts, 1965) illustrate the major climatic variations as far back as 500 A.D., and do not indicate any reason to believe a radical change in climate has taken place. That vegetation is not uniformly in equilibrium with climate has been illustrated by Davis (1976) and Wright (1968, 1976) for the eastern United States, particularly that part covered by late-Pleistocene glaciers where migrations and soil development are incomplete. Braun (1950) considered the Oak-Hickory Forest of the study area to be quite stable. It is unlikely that the study area is experiencing migrations of large numbers of potentially dominant or important species on the scale proposed for glaciated areas because physiographically, geologically, and edaphically (in terms of parent material) the study area has been stable for hundreds of thousands of years (Thornbury, 1965; Hunt, 1974; Braun, 1950). Oak-hickory forest became established in the western Ozarks around 12,000 years BP; at about the same time the central Plains were becoming grassland (King, 1973; Wright, 1970), 50 and as early as 8,700 BP most of the current flora were present in southeast Missouri (King and Allen, 1977). Grassland probably expanded across Missouri during the Hypsithermal, but oak was still an important member of the upland vegetation. Around 5,000 BP the vegetation of southeastern Missouri assumed a mix similar to today's, and since then no major fluctuation in climate can be deduced from the botanical record (e.g., Wright, 1968, 1970, 1976). Thus, it appears that vegetation in the study area very early in the Holocene assumed a character much like that of today. Migrations into the area were probably quick, since most of the currently dominant species were either in the area or close by (Delcourt et al, 1980). The gross vegetational pattern of the study area can be considered relatively stable, but the nature and definition of an ecotone as a tension zone assures that its composition and location have been in constant flux. Within historical times Dyksterhuis (1948, 1957) and Albertson and Weaver (1945) and Weaver (1954) illustrated how land-use and climatic anomalies could influence composition of the ecotone. These changes must be considered part of the continual process of evolution of an ecotone, with forest advancing or multiplying at the expense of grassland, and vice versa depending on 20-30 year climatic fluctuations, droughts, fires or other catastrophes. Whatever the geographic instability of particular plant boundaries and species occurrence, the overall pattern of grassland to the west and forest to the east has not changed during the Holocene (Wright, 1968, 1970), and it is probably related to water vapor transport by the westerlies from the Gulf of Mexico. The gross location of the 51 continental landmass, Gulf of Mexico, and western cordillera have not changed significally in the Holocene, and the essential relationships between tropospheric waves in the westerlies, dry continental airmasses, moist gulf airmasses, and the role of the general circulation in determining the influence of each have probably changed little. The relative influence of cyclones, anticyclones, continental airmasses, Gulf airmasses and the mix of circulation patterns dominating this section of North America may have fluctuated during the Holocene, but as long as the basic sources of moisture and dryness (and solar radiation) have remained the same, and the general circulation has not undergone perturbation on the order of a glaciation, the orientation and general location of the forest/grassland transition may be expected to have remained the same. The ecotone is a transition zone, in effect an area of steep vegetational gradeent. In seeking to explain the transition using another environmental (climatic) gradient, I may find that the current areas of steepest climatic gradient will not coincide precisely with the mapped ecotone; errors in measurement, mapping, and the fact that vegetation responds to many aspects of climate would almost certainly blur the correspondence between climatic boundaries and vegetation communities. In fact, given the long-term history of fire on the Great Plains, the pre—settlement prairie border might lie farther east than a climatic explanation would suggest, since fire was important in inhibiting colonization of prairies by trees and would shift the ecotone toward the moister climate. At the scale of generalization of this study, some minor instability of vegetation or uncertainty in location of the ecotone can be tolerated. 52 The third, and last, assumption necessary for establishing the plant/climate link is that fire is a direct result of climate. Wells (1970) argued the case for fire as a prime variable in grassland maintenance, yet conceded that a climate that caused frequent, wide- spread dryness of plant matter is pre-requisite. A dry climate encourages fires, and any explanation for the climate is an explanation for the frequency of fires in a grassland. Hypotheses and Criteria for Evaluation There are three major hypotheses guiding this research. The first is that growing season (summer) vapor pressure deficit is important in determining the forest/grassland transition. To evaluate this hypothesis the distribution of vapor pressure deficits (and evaporation) will be examined, and, if a map analysis indicates that VPD measures do vary greatly across the ecotone, this hypothesis will be accepted. The second major hypothesis is that the modal growing season atmospheric circulation patterns are associated with the strongest vapor pressure deficit gradients across the study area. This hypothesis will be accepted if a subset of atmospheric circulation patterns, derived from days on which there was a strong VPD gradient across the ecotone, are the most frequent ones (modal or mean) in the long-term data set. The third hypothesis is that the modal (or mean) atmospheric patterns become more frequent during drought years. This hypothesis will be accepted if mean drought circulations of the 1950's can be shown to be similar to the modal atmospheric patterns, and if drought 53 circulation patterns for the 1970's are characterized by an increased frequency of the modal patterns. papa Surface data to evaluate precipitation, vapor pressure deficit, and evaporation estimates were collected for 20 stations across the study area (Figure 12). Daily 7 AM dewpoint temperatures and maximum temperatures for 5 years (1966-1970) were used, along with monthly mean temperatures (maximum, mdnimum and dewpoint) for 1957-1977. These data were collected from the Department of Commerce/NOAA publications Climatologic Data, Climatologic Data National Summary, and The Daily Weather Map. Monthly and annual mean precipitation data for 1931- 1960 and 1941-1970 were collected from Climatologic Data and Climatological Data National Summary. Mean precipitation data on intensity/duration statistics, number of days with rainfall, and probability of precipitation were collected from other published sources (ESSA, 1966, 1967, 1968; U.S. Weather Bureau, 1955; Paulhus and Miller, 1964). To evaluate upper atmospheric circulation patterns, the daily 500 mb height over North America was used. These data, which were obtained from the National Center for Atmospheric Research (NCAR) in Boulder, Colorado, are on magnetic tape for computer processing, and consist of once- or twice-daily observations of 6 variables at 3 levels in the atmosphere for over 200 weather stations across the continent during the period 1961-1977 (Figure 13). Missing data or partial records are a problem in this data set since a considerable 54 "II-0.... ;- — —. _._._.\ ._ ',Des Moines._ \ I . I H r ’." . ._ J.-. .Nor.th Platta‘ . - 1Chicago .",Fort /°""-:-":;-':ls \. :__'_______._.'_\- ''''' (.Spsingf'iield Waynf.’.t ".1 L\ iDodge City 'kansas\ '-. - r ‘- ; . . I .I \ . ! . ! City .‘ Jig .5 1111's.” ’.’.’.—" \ .' I )fi) va-"Jnsv_.__,g...-,—.-"" ,1 ..... £:I:..:u._.----—--:-"‘_ §E¥jpfiff C; ‘Nashvilfe l -{../"'—""'\. I i Oklahoma Litte zMelnphisz.-. K. ' C ! Clty . l I ‘. ‘- ! " ' R Ck“ ‘31, 1‘ b X ! ‘u'b- .’.~. . . . m 3 am . , Amarillo " “1 .fi ! “f E. I ...... . ! . f ‘ ‘ " \Sherélawort MO.Ei,EI_ _ _ ~ ' I " " I .. San Antonio Houston '*‘f" . . o « New Orleans Figure 12 Location of Weather Stations Used for Surface Data 55 Source: NCAR Data Figure 13 Location of Stations Used for 500 mb Data (1961-1971) 56 number of stations, particularly in Canada and oceanic locations, do not have complete records for 1961-1977; they either were established or went out of existence some time within this period. In addition, the Canadian portion of the data set contains no data later than 1974. Missing data are a problem because the statistical analysis of daily patterns necessitates that there be no missing values in a station's record. There are two ways to solve the missing data problem. Some authors (e.g., Kutzbach, 1967, 1970; Craddock and Flood, 1969) used a regular grid of values interpolated from the raw data (usually a mean monthly data set). This approach has intuitive appeal because the data points are not clustered, and the statistics derived, theoretically, will not be biased by groups of stations with similar observations. The other method, used in this research, is to use whatever complete station records are available. This choice was necessitated by the nature of the available data: to interpolate a grid for each day of the summer (June-August), 1961-1977 (1564 days), would have been beyond the scope of this study. Data coverage is definitely biased towards the United States, but only in the late 1970's was the problem acute (Figure 14). In general, there are data points in all portions or "corners" of the map, so interpretation of mathematical surfaces interpolated from an analysis of the raw data will not be without foundation. The problem of explaining the variance in a data set that is spatially clustered will be discussed more fully in the method of analysis chapter (Chapter V), 57 + + +‘ 219 “1 ~33?" l I A + + + A + + + + ( 7" +1 + + + ‘4 + + + + k + + + + + + + + + + + + + + + + + + + Miles , I— ' + + ' 0 1000 + 4- Source: NCAR Data Figure 14 Location of Stations Used for 500 mb Data (1972-1977) 58 since clustering influenced the choice of mathematical technique. Since areal coverage of Alaska and western Canada in addition to the United States is relatively complete, missing data stations were not considered a significant problem that would affect the analysis. Conclusions will not be drawn from analyses that have extrapolated functions into areas of missing data. To evaluate 500 mb circulations during drought years of the 1950's, monthly mean 500 mb heights for the United States were collected from Climatologic Data for 1952-1954; Namias (1955), Rice and Penfound (1959) and Borchert (1971) cite this as a major drought period on the Great Plains. Only mean monthly data are available for this period. Data for circulation patterns during the 1970's drought are derived from the previously mentioned daily 500 mb height data set. Bark (1978) and Riefler (1978) cited 1974, 1976, 1977 as drought years, although the areas affected and the severity of drought were apparently not the same as the 1930's or 1950's droughts. Nonetheless, these years will provide a basis for comparison of circulations during series of years characterized as major droughts. Chapter V METHOD OF ANALYSIS Surface Data The first portion of the analysis seeks to establish the existence of a vapor pressure deficit gradient across the ecotone, and then seeks to emphasize the importance of this gradient by corraborating atmospheric demand-induced water stress through the use of Penman evaporation estimates. Saturation vapor pressure at a given temperature was computed using a sixth order polynomial of the form: T2 + a3T3 + a4T4 E = + a T + a 5 6 3 a0 1 2 + aST + a6T (2) where E8 is saturation vapor pressure (millibars), T is temperature, (0C), and ac to a6 are constants given in Lowe (1977). Vapor pressure deficit was computed using equation (1) (Chapter III). A set of long-term values was desired, and since the relationship between saturation vapor pressure and temperature is non-linear, it was unclear as to whether long-term VPD could be computed using monthly mean temperatures in the formula for saturation vapor pressure, or whether daily temperatures would have to be used to compute daily saturation vapor pressures, which would then be averaged to yield monthly mean values. To solve this problem, for each station (Figure 12) vapor pressure deficit on each day during June, July and August, 1966-1970 was computed using daily maximum and daily 7 AM dewPoint temperatures, and then these values were averaged to yield monthly mean vapor pressure deficits. Monthly mean maximum and monthly mean 59 60 dewpoint temperatures were then used to compute a monthly mean vapor pressure deficit for each station for the same summer months. A simple bivariate correlation between these 2 measures of monthly vapor pressure deficit using each month's value at each station as observations yielded a Pearson r value of .98 (significant at .001). Monthly mean vapor pressure deficits for 1957-1977 were then computed using monthly mean temperature data in all further analyses. Evaporation estimates were computed using monthly data and a modified form of the Penman formula derived by Lincare (1977). The formula bases the estimate on temperature, latitude and elevation: E0 . 700Tm/(100-A) + 15(T-Td) (mm day-1) (3) (80-T) where Tm = T + 0.006H, H is the elevation (meters), T is the mean temperature, A is the latitude (degrees) and Td is the dewpoint (temperatures in OC). Values given by this formula typically differ from measured values by about 0.5 mm day-1 (about 2-52) for monthly means. (In reality the evaporation estimates are not actual values, since the Penman equation is calibrated for open water surfaces, and drying of surface soil inhibits evaporation from subsoil; on the other hand, transpiration from vegetation also depletes subsurface water, so the original estimates may be used as a valid approximation of water loss, if not the exact quantity). Stress DayyCirculation Pattern Selection Daily vapor pressure deficit data were used to designate 15-30 "stress days" in each summer month during 1968-1970. These days represent occurrences of a steep vapor pressure deficit gradient across 61 the study area, and were used to evaluate this first and second hypotheses, that is, that a strong VPD gradient is coincident with the ecotone, and that the circulation patterns which cause the gradient are the most frequent ones. Two groups of three stations each were selected based on their positions relative to the ecotone, called east group and west group, respectively. Daily VPD was computed for each station and then averaged by group to give a mean east group VPD and mean west group VPD for each summer day, 1968- 1970. West group VPD was subtracted from east group VPD, and the result was an estimate of the daily gradient in vapor pressure deficit from east to west across the ecotone. (Vapor pressure deficits are negative values; for ease of presentation only the absolute values of VPD will be used, except where this would be confusing). Stress days were designated as those which met the following criteria: 1) The VPD gradient exceeded 9 mb for July and August, or 6 mb for June. 2) VPD exceeded 15 mb for the west group. 3) Maximum temperature exceeded 24 0C (75 0F) in the west group. These criteria were chosen after inspection of the complete array of daily VPD estimates, and were established with two purposes: first, to extract only those days with greatest VPD-induced water stress in the west, and second, to select for each month a manageable-sized sample of days with a strong east-west gradient for comparison purposes. The first criterion was established with different values for June than for July and August because temperatures and gradients 62 were, on the whole, less pronounced in June. Criteria 2 and 3 were established in order to eliminate cool, dry days from the analysis. June had 15 stress days, July had 22, and August had 31 during the 3 year period (Appendix I lists the dates). These dates were used to derive 500 mb pressure height patterns for stress days, or what will be referred to as stress circulations. Circulation Patterns and Principal Components Procedure Atmospheric circulation patterns at the 500 mb level were derived two ways: using mean pressure height values and using principal components analysis. The principal component analysis was also used to derive long-term circulation patterns; in general the same procedure was followed in all analyses with only the time period of the data changed. The discussion of the method will be limited to a conceptual or geometric description of the technique; mathematical descriptions of the procedure are available in Harmon (1967) and Kutzbach (1967, 1970), and a particularly good description of the method as applied to the mapping of anomalies is given by Fritts (1976). The following discussion relies on Nie, et al (1970) for terminology, since that source is responsible for the computer program that generated the analyses. Principal components analysis (eigenvector analysis in some literature) is a mathematical technique that may be used to reduce a multivariate data set to a smaller group of uncorrelated or orthogonal components. The basic procedure of principal components analysis is to decompose a correlation matrix into separate components based on 63 the intercorrelations of the variables, so that the variables can be estimated from a smaller number of components, each of which is uncorrelated with the others (orthogonal). The purpose is to reduce as much as possible extraneous intercorrelations, and to eliminate non-significant correlations. A component is a linear combination of the original variables, and is a useful substitute when many variables are highly correlated, since it can be used to estimate the variables. An arithmetic average might do just as well as a principal components analysis at estimating a summary value for a large number of highly correlated variables, but when the same data set has two or more subgroups of intercorrelated variables which are not intercorrelated outside of their respective subgroups, the problem becomes complex, and principal components analysis can help. It is this situation that is often encountered in synoptic analysis; observable, discrete circulation patterns exist, but summarizing and comparing them quantitatively is difficult since they are map patterns. A number of statistics are important to interpret a principal components analysis. Starting with the original correlation matrix, the analysis seeks to derive a function (component or eigenvector) which combines the original variables in such a way that the variance that the function explains is maximized. Each variable is standard- ized, and thus contributes unit variance (1.0) to the analysis, and the strength of association or correlation of the variable with the component is termed its loading. The amount of variance of one variable explained by a component is found by squaring the loading, and thus the total data set variance explained by the function (its 64 eigenvalue) is found by summing the squares of all the loadings on the component. The loadings help interpret the function; they indicate which variables contribute large amounts of variance to the function, or, in other words, which variables the function most closely approximates. The eigenvalue gives a measure of how important the function is in the whole data set, and since there is as much variance in the data set as there are variables, the proportion of total variance explained by a component is found by dividing the eigenvalue by the number of variables in the analysis. Once the components and the associations between variables and components have been defined, scores are generated which are surrogate data observations substituted for observations on the variables. The component scores are derived (broadly) from the loadings of each variable on the component and the original data observations, and they can be interpreted as idealized observations, just as the component is an idealized variable. In practice, once the principal components are derived for a data set they are usually rotated or adjusted until each is associated with a different set of variables, while still maximizing explained variance. Blasing and Lofgren (1980) point out that the attempt to maximize variance sometimes leads to combinations on one component of map patterns that are in actuality the opposite of each other. Nonetheless, common practice is to attempt to explain a maximum amount of variance in a single component, and in accord with this dictate componentswith eigenvalues greater than 1.0 were rotated. The most frequent type of principal component analysis encountered in meteorolgic literature is termed R-mode analysis by Rummel (1970). 65 This method may use monthly mean sea-level pressure for a number of years (Kutzbach, 1967, 1970; Kidson, 1975) or perhaps daily 500 mb heights (Craddock and Flood, 1969) to derive a correlation matrix between grid points on a map. In the R—mode a regular spacing of data points is necessary, since the correlation matrix is derived between spatial variables, and each variable contributes unit variance to the analysis. If weather stations rather than interpolated grid points were used, the amount of variance in the analysis (and thus the principal components) would be weighted to areas where data stations were clustered. Interpolating the raw data field to a regular grid assures unit variance contribution from unit areas. Principal components derives the factors describing the greatest amount of variance in the grid based on the monthly (daily) values, or temporal replicates. Thus, the first component will describe the most important grid pattern, and succeeding components will describe less important grid patterns, while component scores will give an idealized estimate of the frequency of occurrence through time of each grid pattern. The current research will use a T-mode principal components model (Rummel, 1970), in which the correlation matrix is derived between the days in the analysis. Principal components derives the functions describing the greatest amount of covariance among the days based on the mean and variance for each day of the 500 mb observations at weather stations, or spatial replicates. Thus, the first component will describe the most frequent association of days, and succeeding components will describe less common daily patterns. Since each day 66 is also a discrete spatial pattern of the 500 mb pressure surface, the association of each day with a component (its loading) indicates how important that 500 mb pattern is in determining the component. This situation is very desirable because components can be directly interpreted from daily 500 mb height patterns, and component scores are mappable as the (standardized) station observation for the daily pattern extracted by the component. Interpretation of the components is done through visual examination of the component scores maps. The T-mode analysis has been used by Perry (1970) to map mean monthly sea-level pressure and precipitation as a way of identifying anomalies. He found the T-mode analysis appealing because it "does not require that the observation points describing the space field be equidistant from one another... (p. 55)." In other words, the procedure derives a function that maximizes the variance for a number of varia- bles, and in T-mode analysis each day contributes equally to the amount of variance in the analysis. Computation of the mean and variance of a day's mean 500 mb height may be skewed by clustering of data stations, but when the days (variables) are compared to each other it is over the same spatial distribution, and the components are not directly influenced by geographic inequalities in data coverage. As an example, the T-mode analysis was used to derive the circulation pattern associated with June stress days; 15 days were selected (1968—1970), and each of 145 stations had observed a 500 mb height on each of the 15 days, resulting in a 15 x 145 raw data matrix and a 15 x 15 correlation matrix. It is useful to think of the principal components model in a geometric sense; each day represents a dimension or axis in space; in this instance there are 15 dimensions, 67 and one wishes to see if fewer than 15 dimensions are adequate to represent the data, or, in other words, to see if there are recurring daily patterns in the height of the 500 mb surface. The 15 axes can be considered as calibrated in height values, although they are actually standardized, and an observation has a location in this 15 dimensional space defined by its value (height) on each one of the days. (In the simplest case of only 2 dimensions (days) a point would be located by its height on day 1 and day 2, or more familiarly, its 3 and y_values). In 15 dimensions, 15 height values are necessary to locate one point, and each of the 145 stations has a point in this space. If some of these 145 points occur in clusters, which would be the case if more than 1 weather station recorded similar 500 mb heights on each of the 15 days, it is possible to represent them with something less than 15 dimensions, and principal components identified one factor with an eigenvalue greater than 1.0 that explained 85 percent of the variance. Thus, the 15 days were effectively reduced to one major circulation type. Important for this analysis is the generation of component scores, for if the components represent new dimensions or days, some way is needed of representing the spatial pattern causing these new dimensions. In this example, component scores of the observations are obtained by projecting the points (as defined in 15 dimension space) onto the axis defined by the component. There is one score for each original observation, for these observations located the point in 15 dimension space, and now the scores locate the point in the new one dimensional (day) space. A score will represent an observation on that day. 68 Approximately 135 scores were generated on the June stress day component (some stations were beyond the map limits used). These scores were then mapped and the result is the main 500 mb circulation pattern associated with the 15 June stress days. The major technical problem with the T-mode analysis was caused by computation limitations. The computer program used to derive the principal components will allow approximately 100 variables, but since each day is considered a variable, a long-term principal components solution (1961-1977) for the summer season (June to August) would require the use of 92 days in each of 17 years, or over 1500 variables, and obviously was just as untenable as interpolating 1500 daily grids for use in an R-mode analysis. The problem was surmounted by treating 2 or 3 months (60-93 days) at a time in separate analyses. For instance, June 1961 and June 1962 together were subjected to the T-mode analysis, which generated 4 components with eigenvalues greater than 1.0, and component scores for 116 station observations. All other Junes were then processed in 3 year segments, and there were, thus, 6 separate principal component analyses for June, with a total of approximately 30 significant components, each with a set of scores for the weather stations. July and August data were processed the same way. With 18 separate principal components analyses, each deriving 4 to 7 components, approximately 100 components were available, many of which should identify similar or identical atmospheric flow patterns, each with a set of scores on the weather stations. To simplify this large number of components a second principal components analysis was run using the first 4 components from each original analysis as 69 variables and their scores as observations on the variables. The result is a set of "combined" components that group together the components from the original analyses. Each original component has a loading on the combined component, and loadings greater than 0.5 were taken to indicate that the original component and the combined component were identical. Both Fritts, et a1 (1979) and Blasing and Lofgren (1980) indicate that a correlation of 0.5 between two map patterns is sufficient to consider them identical. A measure of frequency or importance of each combined component may be had by summing the original eigenvalues of the original components loading over 0.5 on the combined component. For example, to get a measure of the importance of the first combined component, the original eigenvalues of the components that load on the combined first component are summed, yielding a "total combined eigenvalue." The eigenvalues derived by the combined principal component analysis are not themselves usable, because the variables in the analysis were components, each of which summarized an unequal portion of variance from the original analyses. Drought Pattern Analysis Drought patterns for the 1970's (1974, 1976, 1977) were subjected to principal components analysis also, in order to test whether or not long-term modal circulation patterns were more frequent during drought years. The method of analysis was, again, to derive a set of components for each month; the three Junes were analyzed as a separate set of variables, the three Julys as another set, and the three Augusts 70 as the third set. Each of the three individual principal components analyses yielded 5 to 7 significant components, called "original drought components," and these were then subjected to another principal components analysis to yield "combined drought components." Measures of significance of eigenvalues were treated in the same manner as in the long-term analysis. Drought patterns for the 1950's (1952-1954) can be treated only in a qualitative way, as daily 500 mb data were not available at the time of analysis at Michigan State University. Mean 500 mb maps will be used as well as other published accounts of these years to infer the important circulation patterns. This analysis alone will not be sufficient to reject or accept an hypothesis; rather it will be used as supportive evidence (dissenting evidence) for conclusions drawn from the 1970's data. Chapter VI RESULTS Vapor Pressure Deficit and Evaporation The mean and standard deviation of vapor pressure deficits were calculated for June, July, August and summer (1957-1977) (Figures 15- 18). The steepest gradients on these maps closely approximate the location of the ecotone through Texas and eastern Oklahoma, and then recurve westward through Kansas. The wedge shaped pattern of the gradient with its apex to the east and base to the west is similar in morphology to the Prairie Peninsula (Transeau, 1935) of grassland extending eastward across the United States. Although the southern margin of grassland closely corresponds to the strongest gradient, in the Missouri-Kansas region the ecotone, as mapped by Kuchler (1964), lies considerably to the east of strongest monthly and mean summer VPD gradient. The July and mean summer maps, in particular (Figures 16 and 18) illustrate the correspondence between the ecotone and the strongest VPD gradient, with summer VPD varying from approximately 24 mb in Missouri/Arkansas to 30 mb in Oklahoma. The magnitude of the gradient varies from 5 to 8 mb day"1 from Missouri to western Oklahoma, or approximately 21—34 percent difference in drying potential of the air (as expressed by VPD). Variability of the VPD as described by the standard deviation also increases westward (Figures 15-18 panel B). Penman estimates of evaporation (Figures 19-22) do not display as steep a gradient across the study area as the VPD maps do; potential 71 72 Miles A” Mean VPD (millibars) 1000 B. Standard Deviation (millibars) L Figure 15 June Daily Vapor Pressure Deficit (1957-1977) 73 Miles .A. Mean VPD (millibars) 1000 B. Standard Deviation (millibars) L Figure 16 July Daily Vapor Pressure Deficit (1957-1977) 74 a-.- . 3.! Ir. Miles A” Mean VPD (millibars) 1000 B. Standard Deviation (millibars) I Figure 17 August Daily Vapor Pressure Deficit (1957-1977) 75 Miles A” Mean VPD (millibars) r0 Standard Deviation (millibars) t Figure 18 (June—August) Daily Vapor Pressure Deficit (1957-1977) Summer Miles 76 A, Mean Penman Evaporation (millimeters) .‘.-.-.O.-.’-.‘ . is- J. .O...’”-.-. 8. Standard Deviation (tenths of millimeters) Figure 19 June Daily Penman Evaporation (1957-1977) Miles 77 A. Mean Penman Evaporation (millimeters) . . . . . ~ . --- t-.-o-n--\a. .-'-'-'--~.-_-_-‘- 1 -.-.J 5 1:1 '09 12 fi 11 ‘h--Ta.‘-‘“. --.- ' -' . mu..— B. Standard Deviation (tenths of millimeters) Figure 20 1977) July Daily Penman Evaporation (1957 78 Miles .A. Mean Penman Evaporation (millimeters) Standard Deviation (tenths of millimeters) Bo Figure 21 1977) August Daily Penman Evaporation (1957 1000 Miles l r 0 79 a-- . Ii 0 a . .O. 9r: .A. Mean Penman Evaporation (millimeters) of millimeters) B. Standard Deviation (tenths Figure 22 Summer (June-August) Daily Penman Evaporation (1957-1977) 80 evaporation differs less radically from east to west. The July, August, and summer mean maps (Figures 20-22) display a gradient stronger through western Missouri and Kansas than other areas, and the gradient there more closely follows the location of the ecotone in the area where VPD did not. The standard deviation or variability of the Penman estimates (Figures 19-22 panel B) increases westward by about 0.5 mm day—l across the study area. The magnitude of the Penman evaporation gradient is relatively 1 in Missouri or small, around 1.5 - 2.0 mm day-l, from 6.0 mm day— Arkansas to 8 mm clay.1 in western Oklahoma, or a 20 percent difference in daily evaporation. While the gradient may be of small magnitude, the effects of evaporation are cumulative, and small daily differences can lead to larger seasonal discrepancies. As a rough approximation on how important evaporation may be to the maintenance of the ecotone, the late spring and summer precipitation gradients were compared to the summer potential evaporation gradient from central Missouri to central Oklahoma (Table 1). Late spring and summer precipitation totals vary by about 3 inches per season across the ecotone, while summer potential evaporation (according to the Penman estimate) differs by about 5 inches, and as a result the potential water deficit for central Oklahoma in summer is almost twice as great as in central Missouri. When late spring rainfall totals are added in (assuming no evaporation in spring, which is an unrealistic but conservative assumption), central Missouri has an almost balanced water budget, while central Oklahoma still incurs a potential 14 inch water deficit. As Table 1 indicates, the main reason for the increase in potential 81 Table 1 Moisture Budget for Central Missouri and Central Oklahoma Central Oklahoma Central Missouri Spring Precipitation (April-May) 6.0" (15.2 cm) 9.4" (23.9 cm) Summer Precipitation (June-August) 9.0" (22.9 cm) 12.0" (30.5 cm) Total Precipitation 15.0" (38.1 cm) 21.4" (54.4 cm) Summer Penman Evaporation 29.0" (73.7 cm) 23.9" (60.7 cm) Summer Deficit (Precipi- tation minus Evaporation) -20.0" (~50.6 cm) -11.9" (-30.2 cm) Growing Season Deficit -14.0" (-35.6 cm) -2.5" ( -6.4 cm) Mean July Temperature 26°C 26°C Mean July Dewpoint o o ‘ Temperature 16 C 18 C Mean Summer VPD 30 mb 23 mb 82 evaporation westward is decreasing atmospheric humidity as expressed by the deWpoint temperature and reflected in vapor pressure deficit. Stress Day Circulation Patterns In order to evaluate the circulation patterns responsible for the VPD gradients (and their variation), stress days having strong gradients across the study area were selected (as described in Chapter V). The stress day mean 500 mb pressure height patterns are quite consistent for each month (Figures 23, 25, 27); the basic configuration shown on all maps is a ridge over the mid-continent, with lower pressure heights along the Pacific Coast and Great Lakes. The August pattern differs slightly from June and July in having an area of higher pressure heights over Hudson's Bay depicted by divergence of the 5600 and 5500 meter contours on Figure 27, possibly indicating more frequent high-latitude blocking in August. In all months the stress days are associated with negative 500 mb height anomalies in the Gulf of Alaska and positive anomalies over the central states (Figures 24, 26, 28). The Great Lakes area experienced lower than average heights in June, higher than average heights in July, and approximately average heights on August stress days. Troughing in the Gulf of Alaska and ridging over the mid-continent are the two most conspicuous deviations apparent for the stress day mean 500 mb circulation patterns. Principal components analysis derived one stress-day pattern with an eigenvalue greater than 1.0 in both June and July, and two patterns with significant eigenvalues in August (Table 2). Maps of 83 H Miles .___——4‘ o 1000 “v Heights in Meters Source: NCAR Data A”;' Figure 23 June Stress-Day 500 mb Height 84 Figure 24 Deviation of June Stress-Day 500 mb Height from 1961-1977 Mean 85 O 1000 Heights in Meters Source: NCAR Data Figure 25 July Stress-Day Mean 500 mb Height v Miles r_ O 1500 Deviation in Meters Source: NCAR Data Figure 26 Deviation of July Stress-Day 500 mb Height from 1961-1977 Mean 87 Miles I a '— I O 1000 Heights in Meters Source: NCAR Data Figure 27 August Stress-Day Mean 500 mb Height \Ny Stress Day Principal Component Statistics Number of Days First Component Eigenvalue Percent Variance Explained Second Component Eigenvalue Percent Variance Explained 89 Table 2 June 15 12.8 85.3 July 22 19.3 87.9 August 26 22.4 86.2 1.01 3.9 90 the component scores for June, July, and the first August component (Figures 29, 30, 31) depict a standardized 500 mb surface very similar to the mean 500 mb surface of Figures 23, 25, 27, and, at least for stress days, the first component may be interpreted as the dominant or mean 500 mb pressure height pattern. The second component derived for August (Figure 32) has an eigenvalue of only 1.01 and explains 3.9 percent of the variance. The pattern of scores for this component illustrates a tendency for deeper troughing in the Yukon area and greater ridging in the Hudson's Bay area on some August stress days. Long-Term Circulation Patterns Principal components analysis of the long—term data set provided over 100 components which have been summarized into 6 major patterns through a second principal components analysis on the first 4 components from each of the 18 original analyses. Table 3 presents the explained variance breakdown for the combined components. The 18 original first components all load over 0.90 on the combined first component (Figure 33), while no other original components load over 0.10 on it. All 18 original first components and the 3 stress day first components are highly intercorrelated, with £_values ranging from 0.92 to 0.98, and within each of the original analyses the first 'components explain 72-85 percent of the variance. Because of missing data the component scores and map grid used to derive the combined first component (Figure 33) are reliable only over the United States; the key features on this map are noticeable troughing along the West Coast, ridging over New Mexico to North Dakota and a trough over the 91 1000 Source: NCAR Data Figure 29 June Stress-Day Component Scores 92 .00 : Miles 4: O 1000 Source: NCAR Data Figure 30 July Stress-Day Component Scores 93 Miles ggj ‘r v 1 o 0 1000 (TN ‘\\ Source: NCAR Data Figure 31 August Stress-Day Component Scores (First Component) 94 ‘ ‘\ \o '5"- U . 1;? v ‘ ( ' O 00' ‘- / .. '1')!" 95 Table 3 Summary Statistics For Combined Principal Components Analysis Sum of Original Components Having Loadings _:_> 0.5 On Combined Components (Cn) Combined Analysis Accumulated Percent Percent Combined Total of Total Variance Component Variance yExplained Eigenvalue Explained C1 1275.4 81.5 18.3 25.4 02 37.1 2.4 14.9 20.7 C3 161.8 10.3 10.7 14.8 (16.6)* (1.1)* C4 19.2 1.2 8.3 11.5 C5 18.4 1.2 6.5 9.0 C6 6.5 0.4 5.4 7.4 1518.4 Total Variance f 1564 - number of days in data set Common to First 6 Combined Components = 1518.4 (97%) *Variance computed without contribution from original first components. See text for explanation. 96 I 'r 0 1000 a ‘ . Miles 1 ' ' Q Source: NCAR Data Figure 33 Component Scores of First Combined Component 97 Great Lakes. The trough/ridge/trough sequence over North America is visible on the all maps of original first components that load highly on the combined first component (e.g., Figure 34, 35, 36), but the amplitude and displacement of the features vary for the different time periods. An August and a July pattern (Figure 34, 35) depict the typical Gulf of Alaska trough and central states ridge, while a June pattern (Figure 36) depicts a similar sequence, but the southern portion of the pattern is shifted eastward so that the California/ Nevada area is dominated by a trough, and the mid-continent ridge has its axis from Texas to Minnesota. This original component has a loading of 0.90 on the first combined component, and the gross patterns are similar (this original component correlates strongly with other original first components, with p ranging from 0.91 to 0.97), but the actual synoptic situation and daily weather in the United States associated with a pattern such as Figure 36, would be slightly different from conditions under the previous two patterns. Figure 34 depicts a high amplitude wave pattern, while Figure 35 illustrates a more zonal pattern over the United States along with stronger ridging in the Yukon sector. The mean summer 500 mb pattern (Figure 37) is similar to the original first components (Figures 34-36) and the first combined component (Figure 33). Principal components analysis usually extracts as the first and most important component a general tendency of the data (Nie et a1, 1970, p. 482), and in this case the Gulf of Alaska trough/mid-continent ridge is well represented in all analyses, even though the exact position of the features varies in the different time 98 I I I— I 0 1000 Source: NCAR Data Figure 34 Component Scores of August, 1969-1971, First Component This component loads highly on the first combined component (Figure 33). . Miles *4 [— I Source: NCAR Data 0 Figure 35 Component Scores of July, 1969-1971, First Component This component loads highly on the first combined component (Figure 33). 100 Miles yJ o 1000 /\ Source: NCAR Data Figure 36 Component Scores of June, 1963—1965, First Component This component loads highly on the first combined component (Figure 33). 101 .1” 1 Miles J r 1 O 1000 Heights in Meters Source: NCAR Data Figure 37 June-August Mean 500 mb Heights (1961-1977) 102 periods and months. The first component score maps, thus, appear to be interpretable as representations of the long-term mean 500 mb pattern. The sun of original eigenvalues for components loading over 0.90 on the first combined component explains 81.5 percent of the total original variance (Table 3). The second combined component (Figure 38) is most frequent in July, as seven of the eleven original components loading significantly on it were from July analyses. All of the original components loading on the second combined component display the same amplified western ridge/eastern trough combination, and this component is interpretable as a discrete flow pattern associated with drought; it explains 2.4 percent of the total variance (Table 3). Under the highly amplified and extensive western ridge, precipitation is suppressed and vapor pressure deficits are high, while northerly flow in the eastern portion of the country causes continued advection of Canadian airmasses, depressing temperatures. July, 1976 (Figure 39), loaded heavily on this component, and temperature deviations (Figure 40A) were strongly positive for the western two-thirds of the country (exclusive of Texas and Oklahoma), while precipitation was deficient for the Great Plains and central Lakes states (Figure 40B).* The drought in the central Plains under this type of amplified ridge/trough pattern is remarkable. * Because the National Meteorological Center/NOAA changed data report- ing format in 1972, mean monthly 500 mb maps were not available for 3 particular months used to illustrate principal component patterns. The mean 700 mb chart has been substituted for these months, as the mean maps are provided as examples of flow patterns, and the 700 mb and 500 mb configurations are quite similar in appearance. Use of the 700 mb charts in no way weakens any of the arguments or conclusions drawn from the upper air circulation patterns. 103 0 lOOO\\\\\——“\“_‘) Source: NCAR Data Figure 38 Component Scores of Second Combined Component 104 2940 2900 2970 , 3000 3030 3060 2970 a 3090 J 0 <, 3030 3060 3090 3120 3150 V///// Miles , 1150 \ Heights in Meters 01- Source: Wagner, 1976 Figure 39 July, 1976 Mean 700 mb Height This month loads highly on the second combined component (Figure 38). 105 I. ‘Iv 'T-u B. Precipitation-Percent of 1941-1970 Normal Source: Climatological Data, National Summary we: ’I'( at KAI! 0‘ 94:0“ $0 :5 ‘w 50 X Figure 40 July, 1976 Temperature and Precipitation Departures Typical departures under a second combined component flow pattern (Figure 38). 106 Figure 39 more closely resembles the first factor pattern (Figure 33) than the flow pattern it is supposed to illustrate (Figure 38). This discrepancy can be ascribed to strong troughing along the west coast during the first part of the month and strong ridging over the Rockies and Great Plains coupled with a deep Great Lakes trough during the second part of the month (Wagner, 1976). The resulting mean map is a composite of different flow patterns. Since July, 1976, loads heavily on the second component it will be assumed that the balance of the days in the month most closely resembled the second component pattern (Figure 38). The third combined component (Figure 41) presents somewhat of a problem. The map pattern itself explained the third largest amount of variance in the combined analysis using the combined analysis eigenvalues, and the second largest amount of variance using the sum of original eigenvalues of components loading over 0.50 on it. But, of the original components loading on this factor, two were original first components, and the accumulated total variance for this component is extremely high as a result of the eigenvalues of the first components that load on it. Examination of other original component score maps that loaded highly on this component reveals that it is a combination components, that is, opposite map patterns are grouped on this component as a compromise method of maximizing variance in the principal components analysis. The standard first component pattern has a west-central states ridge and a Great Lakes trough, while 5 of the original components loading highly on the third combined component actually had a deeply amplified western trough/eastern ridge situation 107 Vfl l Miles , F I 0 1000 Source: NCAR Data Figure 41 Component Scores of Third Combined Component 108 (e.g., Figures 42-44). The resulting map (Figure 41) has a relatively gentle southwest/northeast gradient, approximating neither the first component pattern nor the other original component patterns loading on the combined component. If the eigenvalues of the two original first components loading highly on this combined component are subtracted from the total accumulated variance of this component, it explains 1.1 percent of the total variance, making it the sixth most important component (Table 3). Using the component score maps of the original non-first components loading on this component, (Figures 42-44), the pattern is one of a deep Gulf of Alaska trough/Hudson's Bay ridge. June, 1970, loaded highly on this component and illustrates the climatic conditions associated with this type of circulatory pattern (Figure 45, 46). The west coast was extremely wet and relatively warm, reflecting the prevalence of southerly flow off the Pacific. The south-central portions of the country experienced cooler than normal conditions, perhaps as a result of cloudiness associated with the advection of maritime tropical airmasses from the Gulf of Mexico by these southerly flows. Frequent outbreaks of Canadian airmasses associated with an east-coast trough kept the east relatively cool. Precipitation was near normal in the east and somewhat suppressed in the west central states as a result of the Hudson's Bay ridge. The fourth combined component (Figure 47) displays a Rocky Mountain trough and a Great Lakes ridge; it explains 1.2 percent of the total variance. Typical of the original components loading heavily on this one was August, 1962, (Figures 48 and 49), during which the 109 Miles " e -= \ 0 1000 /\ Source: NCAR Data Figure 42 Component Scores of July, 1961-1962, First Component This component loads highly on the third combined component (Figure 41). 110 Source: NCAR Data Figure 43 Component Scores of August, 1963—1965, Second Component This component loads highly on the third combined component (Figure 41). 111 Figure 44 Component Scores of June, 1969-1971, Second Component This component loads highly on the third combined component (Figure 41). 112 3500 5600 5500 sum \ . ‘ “5 sum 0 "x“_}: ‘ ‘ O I ....'J'. 56 00 :3. ' 9 a a a a» 5700 : : \’ kg '5, '\~ _, ”N...“ S" f u . 4' sum A ; x j x {~— I Miles __, I I 0 1150 5 i Heights in Meters Source: Climatic Data, National Summary Figure 45 June, 1970 Mean 500 mb Height This month loads highly on the third combined component (Figure 44). B. Precipitation—Percent of 1931-1960 Normal Source: Climatological Data, National Summary Figure 46 June, 1970 Temperature and Precipitation Departures Typical departures under a third combined component flow pattern (Figure 44). 114 Miles . 0' i, I 0.8 o 1000 / Source: NCAR Data Figure 47 Component Scores of Fourth Combined Component 115 3700 $000 \ " Miles I J r I 0 1150 Heights in Meters Source: Climatic Data, National Summary Figure 48 August, 1962 Mean 500 mb Height This month loads highly on the fourth combined component (Figure 47). 116 o A. Temperature Departures ( F) from 1931-1960 Normal B. Precipitation- Percent of 1931-1960 Normal Source: Climatological Data, National Summary Figure 49 August, 1962 Temperature and Precipitation Departures Typical departures under a fourth combined component flow pattern (Figure 47). 117 Pacific Northwest was much wetter than normal as a result of frequent storms associated with the mean trough position, and the rest of the country was generally drier than normal under the influence of the upper air ridge. Coolness in the North and warmth in the South reflected dominance by Pacific and subtropical airmasses respectively. All of the components loading significantly on this one display this same Pacific Northwest trough/eastern ridge pattern. All of the components associated with the fifth combined component (Figure 50) display an amplified Pacific-west coast ridge and amplified central states trough, and it also explains 1.2 percent of the total variance. August, 1972, loaded highly on this combined pattern and illustrates the climatic conditions under this regime (Figure 51 and 52); under the mean ridge, the west coast was relatively dry and warm, while under the mean trough the central and eastern portions of the country experienced amplifying short waves, frequent and enhanced frontal activity, and abundant precipitation. The central and eastern states also experienced below normal temperatures as a result of frequent outbreaks of Canadian airmasses, as did the south-central states. The sixth combined component explains 0.4 percent of the total variance and the two original components loading highly on this one display relatively zonal circulations, with a Gulf of Alaska trough, weak west coast ridge, and strong Hudson's Bay/Great Lakes trough, and these features are adequately represented by the combined sixth component scores (Figure 53). June, 1976, loaded highly on this component, and during this month the west coast was dry, with variable temperatures depending on location relative to the mean ridge (Figure 119 III 2080 3120 , i \ I l . 2‘. : .-' ‘ z"... . -. ;' ( I '- 3150 ”‘3. '9 . m... x \‘ . W'm'“ "" It s ,_.."‘A D / “my“... 1.. ...... ~.... “. g $ 1"“ , '“"" -. . “t "3. I! I ““4 ”3' r.“,,3;‘"""(.._ l: TNfii-v" a 0“ II..." I i .. “m 2". -" ~T"""'"“"“2'§::;‘. ....... . ,. .. J. I .25 I . "I \ 'l a. u-L‘. ( "i I 2’1- \ M“ . Mi les 4 ”8° If I 0 1150 \ Heights in Meters Source: Dickson, 1972 Figure 51 August, 1972 Mean 700 mb Height This month loads highly on the fifth combined component (Figure 50). 120 . .1“ (4111 15’ III I. \Jx-n B. Precipitation- Percent of 1941-1970 Normal Source: Climatological Data, National Summary Figure 52 August, 1972 Temperature and Precipitation Departures Typical departures under a fifth combined component flow pattern (Figure 50). 121 o 1000 ____.\ ‘ / Source: NCAR Data Figure 53 Component Scores of Sixth Combined Component 122 ‘\~ 3000 \ 3120 w/ 3150 / I Miles 4, If I 0 1150' Heights in Meters Source: Taubensee, 1976 Figure 54 June, 1976 Mean 700 mb Height This month loads highly on the sixth combined component (Figure 53). 123 Mwmme no» A. Temperature Departures from 1941-1970 Normal Una-SC "a" ”to” 79:0“? “9'! Owe-NR B. Precipitation- Percent of 1941-1970 Normal Source: Climatological Data, National Summary Figure 55 June, 1976 Temperature and Precipitation Departures Typical departures under a sixth combined component flow pattern (Figure 53). 124 (Figure 54 and 55). The interior Northwest was wetter than normal, apparently due to amplification of storms passing over the northern Rockies, while the eastern states were moist as a result of frequent southerly circulation advecting Gulf of Mexico airmasses northward. The southeastern states experienced below normal temperatures because of cloudiness associated with the maritime airmasses. In summary, the first long-term components appear to extract the mean 500 mb circulation or long-wave pattern from the data, and succeeding components are interpreted as actual synoptic circulation patterns. The second combined component depicts a strongly amplified 500 mb pattern associated with dryness in the Great Plains; this component is most frequent in July. The third combined component actually integrates contrasting circulation patterns; if the atypical original components are disregarded, this component becomes the sixth most important one; it is associated with dryness over the Great Plains. The fourth combined component is also associated with drier than normal conditions on the Great Plains, while the fifth and sixth combined components are associated with relatively wet conditions, at least for the eastern two-thirds of the country. Drought Circulations The summer months of the 1970's drought years were subjected to the T-Mode principal components analysis, and the first 4 factors for each month (called drought components) were subjected to another principal components analysis which yielded 4 significant combined drought components. Original drought components loading over 0.5 on the combined components were considered identical, and the original 125 eigenvalues of these components were summed to give a measure of variance contained in a combined component (Table 4). The map of first combined drought component scores (Figure 56), appears to be a mean circulation pattern, with an amplified central states ridge similar to the mean summer condition (Figure 37). The second combined drought component from the 1970's (Figure 57) is identical to the second long-term combined component (Figure 33). This circulation pattern, with a strongly amplified ridge covering the western half of the country, is associated with hot, dry (drought) conditions throughout the Great Plains (Figures 39 and 40). The third and fourth combined drought components (Figures 58 and 59) correspond to the fifth and sixth long-term combined components (Figures 50 and 53) respectively. Both of these patterns are associated with relatively moist conditions in the eastern two-thirds of the United States. Dryness on the west coast is induced by the ridging depicted in the fifth combined component (third combined drought factor). These components are not associated with drought on the Great Plains, and occur with the same relative frequency in the 1970's data as in the long-term data set. In summary, the drought components derived from the 1970's data are the same circulation patterns as the first two and last two long- term combined components. The mean pattern (first component) for the 1970's data displays an amplified central states ridge; the second 1970's (and long-term) component depicts a strongly amplified western ridge associated with July drought on the Great Plains. The remaining 1970's components are associated with normal to above normal rainfall in the eastern portion of the country, and thus it is the 2 126 Table 4 Summary Statistics for Drought Principal Components Analysis Total Variance (E) = Number of Days = 276 Sum of eigenvalues of original components Combined loading 3 0.50 on combined component Combined Component and percent of total variance Component Statistics Spm Percent Eigenvalue Percent Cl 214.5 78.0 3.1 25.9 C2 12.1 4.4 2.9 23.8 C3 11.5 4.2 2.5 21.0 c, 7.4 2.7 1.6 13.7 127 . Miles _, “'1‘ If 1 0 1000 \ Source: NCAR Data Figure 56 Component Scores of First Combined Drought Component 128 1000 Source: NCAR Data Figure 57 Component Scores of Second Combined Drought Component 129 l Miles . I T O 1000 Source: NCAR Data Figure 58 Component Scores of Third Combined Drought Component 130 ’ l Miles 1000 O Source: NCAR Data Figure 59 Component Scores of Fourth Combined Drought Component 131 most frequent circulation patterns which are associated with drought in the central states. Chapter VII DISCUSSION AND CONCLUSIONS Introduction The 500 mb flow patterns constructed in the attempt to evaluate the current research hypotheses raise further questions about the atmospheric circulations associated with the mean 500 mb flow patterns. Part of this chapter will deal with these questions, drawing greatly on the discussion of climate in Chapter II. First, the research hypotheses will be evaluated; second, summertime circulation patterns will be discussed, with special reference to the mean; third, data processing problems encountered in the principal components analysis wilI be elucidated; and fourth, conclusions will be drawn on the adjustment of plant boundaries to climatic stresses. Evaluation of two of the research hypotheses hinges on determining a single modal or mean 500 mb circulation pattern which has some explicit frequency of occurrence during the summer months. The combined principal components analysis attempted a synthesis of over 100 different patterns as a means of deriving the mean or modal pattern and its frequency. Two problems were encountered in this synthesis: 1) The combined component score map of the first component (presumably mean pattern) lacks definition and is unreliable beyond the borders of the United States; 2) Frequency of occurrence was not readily obtainable from the eigenvalues of the analyses. These problems do have bearing on how the research hypotheses were evaluated. It should be noted at the outset that principal components analysis usually 132 133 extracts a general tendency of the data as the first component (Nie et a1, 1970; p. 482), and therefore throughout this analysis the first component score maps are interpreted as the mean and modal 500 mb circulation pattern. Evaluation of Hypotheses VPD Gradient and the Ecotone From the Gulf coast of Texas to the Missouri/Arkansas border, the strongest VPD gradient generally coincides with or follows closely along the ecotone (Figure 4; Figures 15-18). Throughout this area the transition from forest to grassland takes place in a roughly east-to- west direction, as does the transition from lesser to greater VPD. Penman evaporation estimates (Figures 19-22) display a gradient similar to VPD throughout the southern portion of the study area. Northward from Arkansas and Oklahoma the isolines of VPD and Penman estimates recurve westward into Kansas and Nebraska, while the ecotone trends southwest-northeast across Missouri, indicating that the ecotone trends across or normal to the steepest VPD and evaporation gradients. Two interrelated explanations could account for this lack of agreement. First, north of Arkansas and Oklahoma other variables besides VPD and evaporation are important in determining species occurrence. For example, Kucera and McDermott (1955) presented evidence of a north-south species segregation within the Missouri ecotone, with temperature or temperature related phenomena (such as growing season) presumably being responsible for the species variation. At least in this case, variables other than water stress control species distribution in part of the study area. 134 Second, reducing the forest/grassland transition to a line on a map oversimplifies the direction and character of the vegetational transition. A comparison of pre-settlement vegetation maps (e.g. Kuchler, 1964; Braun, 1950) will reveal wide discrepancies in the placement and extent of deciduous forest, forest/grassland mixtures, and prairie in the state of Missouri. In addition, authors generalizing accepted source maps may completely eliminate a vegetation type. For example, King and Allen (1977; Figure 1, not reproduced herein) completely eliminated all tall grass prairie in Missouri on their map of vegetation, which was cited as a generalization of Kuchler's (1964). The problem exists for all researchers; for example, Kuchler (1964) needed to decide how extensive a "patch" of grass had to be before it was mapped as "Grass, medium height," rather than "Broadleaf deciduous trees. Grass, medium height in patches." Although subsequent biogeographers may see the problem of an indistinct or inconsistently mapped ecotone as an obscuration of the correspondence between climate and vegetation, the mapping problem is, instead, a symptom of underlying complexity. Missouri is in a climatic, geologic, and vegetational situation more complex than surrounding states in the study area. The northern third of Missouri was subject to Pleistocene glaciation and consists of rolling topography with soils that are relatively young and calcareous. This area supported mixed forest and prairie in pre- settlement times. The southern half of the state contains the steep, highly dissected hills of the Ozarks, with droughty, cherty, leached soils derived from limestone, dolomite, and sandstone. This area 135 supports oak-hickory forests. The west-central portion of the state lies in the Osage Plains, with flatter topography and more mesic limestone-derived soils than in the Ozarks. Prairie was best developed in pre—settlement times on the flat, limestone soils of the Osage Plains area, and Ozarkian oak-hickory forest today ends abruptly where flat, limestone topography of the Osage Plains begins. Within Missouri, then, the edaphic and physiographic diversity of habitats, along with multiple climatic gradients in north-south and east-west directions, contributed to a complex vegetational pattern, with large portions of the western and northern sections of the state being mixed forest and grassland. Within this ecotonal area, soil type and physiography were the major determinants of pre-settlement vegetation type. Some climatic variables other than VPD or evaporation apparently are important in the northern part of the state. I would speculate that precipitation/evaporation ratios are particularly important northward. The result is a vegetational distribution that is complex and apparently bears little resemblance to broad-scale climatic gradients used in this analysis. That the ecotone in this portion of the study area does not coincide with the steepest VPD and evaporation gradients is not too surprising. In the southern portion of the study area, however, a strong correspondence between steep VPD/evaporation gradients and the forest/grassland transition is apparent (see Figure 4). The monthly variability of VPD and Penman estimates also increases westward across the study area (Figures 15-18, panel B). Since monthly 136 VPD is highly correlated with daily VPD, a higher month-to-month variation also implies greater daily variability. Coupled with the overall increase in VPD westward, increasing variability suggests that the western portion of the study area is frequently dominated by dry airmasses but still occasionally experiences invasion by moist and/or cool airmasses, while the southeastern portion of the study area is only less frequently invaded by dry continental airmasses. Moist airmasses would almost certainly be derived from the Gulf of Mexico, while dry airmasses might be derived over the Rocky Mountains, Canada, the desert Southwest or in place over the Great Plains under subsident conditions. It is the pattern of domination across the study area by airmasses having radically different relative humidities that this analysis suggests was critical in determining the pre- settlement vegetation distribution. The correspondence between the ecotone and gradients of vapor pressure deficit and VPD variability (as well as in potential evaporation estimates, in which VPD is an important term) is evidence that a significant control on the vegeta- tion transition may have been generated by daily inequalities of vapor pressure deficit across the ecotone in the southern portion of the study area. Waring and Major (1964) discussed two methods of examining plant-environment relations: 1) correlative; and 2) operational investigations. Correlative studies assume a relationship between plant and environment and seek to measure any environmental variables possible in order to indicate which facets of the environment co-vary with plant behavior and distributions. Operational studies investigate 137 factors which are known to be physiologically or otherwise influential to plants. Although the emphasis of visual map correlations between VPD gradients and the prairie-forest border is somewhat of a correlative approach, Chapter III demonstrated that strong physiologic evidence exists for vapor pressure deficit as a limiting factor in plant growth and grassland/forest competition. The conclusions reached using VPD maps as evidence are operational in the sense that vapor pressure deficit is a strong influence on plant behavior, and the geographic variability of VPD in coincidence with the ecotone demonstrates the spatial controls placed on plant distributions by climate in at least part of the study area. Flow Pattern on Stress Days Principal component analysis of stress-day 500 mb data derived only one flow pattern as being of any consequence (Figures 29-31). (One other flow pattern explaining one percent of the August stress- day variance was also derived; Figure 32). The mean 500 mb pattern (Figure 37) and the stress-day patterns (Figures 23, 25 and 27) are similar in form, although the stress-day patterns depict a deeper Gulf of Alaska trough and higher central states ridge, as is illustrated by the stress-day deviations from the mean 500 mb surface (Figures 24, 26 and 28). The increased amplitude of the pattern on stress days could be caused in part by the occurrence of only a single flow pattern on stress days, in part perhaps by greater amplification of the flow pattern itself on stress-days (although amplification was not tested herein), and in part by the shorter time period used in the calculations. (Namias (1951) found that as the time period used in the 138 averaging process increased, the number and amplification of waves on mean upper-air maps decreased). The first components explained 85-88% of the stress-day variance and 81.5% of the long-term variance (Tables 2, 3): a difference of proportions test (Blalock, 1972; pp. 228-230) on the sun of the eigenvalues of the first stress-day components versus the sum of the eigenvalues of the long-term first components yields a 2 value of 2.04, indicating that significantly more variance is explained by the first component in the stress-day analysis (.05 significance level, one- tailed test); it can be concluded that the sample of circulation systems occurring on stress-days was more homogeneous than the long- term data. In other words, stress-days had significantly greater frequency of the mean 500 mb pattern (first component), and the greater amplitude of the stress-day 500 mb pattern (compared to the long-term mean) probably results from the more homogeneous sample of 500 mb flow patterns rather than from the occurrence of strongly amplified patterns on stress-days. The high explained variance of the first component (mean pattern) on stress days, and the similarity of stress-day mean and first component maps (Figures 23-31) with long-term mean and first component maps (Figures 33-37), suggests that the most frequent of the long-term circulation types is the one associated with the strong VPD gradient on stress days. And the second hypothesis, that the strong VPD gradient across the ecotone is caused by the most frequent circulation patterns, is affirmed. 139 Drought and 500 mb Circulations The patterns derived from principal components analysis and their eigenvalues can be used to test the third hypothesis, that the modal (mean) 500 mb circulation pattern increases in frequency in drought years. The drought analysis of the 1970's derived 4 components, the first two of which are identical to the first 2 components in the long-term analysis and are associated with dryness in the central states. The first drought component (Figure 56) is also the same as the typical mean 500 mb summer circulation pattern (Figure 37). The sum of the eigenvalues of the original components loading over 0.5 on the first combined drought component constitutes 78% of the variance of the drought data set (Table 4), while the sum of the eigenvalues of the first components of the long-term set explains 81.5% of the long-term variance (Table 3). The frequency of the mean circulation pattern appears stronger for the long-term set than for the analysis of drought years during the 1970's. A difference of proportions test on the sum of the first drought component eigenvalues versus the sum of the long-term first component eigenvalues yields a z of 1.36, indicating no significant difference in the proportion of variance explained. For the drought years of the 1970's the tendency toward the mean circulation pattern was no stronger or weaker than the long-term tendency; by this analysis drought cannot be attributed to increased frequency of the mean pattern during the 1970's. And, thus, the third hypothesis, that the mean circulation pattern increases in frequency during drought years, is not accepted. 140 Did any circulation pattern occur more frequently during the 1970's? The second combined drought component (Figure 57) is identical to the second combined long-term component (Figure 38), and therefore the second most important circulation types in both analyses are the same. A difference in proportions test between the sums of the eigenvalues of the original components loading on these combined components yields a 2 value of 1.9, indicating significantly greater variance explained by the second drought component (significance level of .05, one-tailed test). The second combined component occurred more frequently during the 1970's. This pattern is one of strongly amplified eastern ridge/eastern trough (e.g., Figure 39), and it is most frequent in July, which is also the month most likely to experience drought or rainfall deficiency on the Great Plains (Borchert, 1950). One of the key features of the second combined component score maps (Figures 38 and 57) is a deep trough in the Great Lakes region. Under this type of pattern, Canadian anticyclones frequently push south across the eastern two-thirds of the United States and drive Gulf of Mexico airmasses far south, thus reducing precipitation in the Great Lakes and Midwest states. Harman and Harrington (1978) ascribed the drought of August, 1976, in the Middle West to the frequency of dry airmasses advected southward by an amplified trough in the eastern portion of the country similar to the circulations depicted in Figure 39 for July, 1976. Both July and August, 1976, were relatively wet through portions of the southwest and southeast (Wagner, 1976; Harman and Harrington, 1978; Figure 40). Under these circulations (Figures 38 and 57), the Southwest is wetter than normal apparently due to influx of moisture from the Pacific and from the Gulf of Mexico. The 141 Southeast is wet because of frontal activity along the strong baroclinic zone associated with the deep eastern United States trough. Drought in this circulation pattern is primarily a mid-continent occurrence (Figure 40). Mean Summer Circulations All first component score 500 mb patterns have been interpreted as approximating the long-term mean 500 mb pattern (see the following section on Principal Components and 500 mb Patterns), and perhaps the most obvious question raised by this analysis is, "How does the mean 500 mb pattern affect the climate of the study area?" The pattern is one of a trough in the Gulf of Alaska, central Plains ridge, and eastern U.S. trough (e.g., Figures 1, 23, 25, 27, 33, 37). The key feature dominating the central U.S. is the 500 mb ridge. Under this circulation (according to Namias, 1955, 1963), anticyclonic upper air wind flow dominates the central states, causing general subsidence and warming of the middle layers of the atmosphere which induces inversions, enhances stability and dampens uplift mechanisms, thus reducing precipitation, while cyclonic storms are shunted north along the major baroclinic zone. Radiation heating of the continental interior induces a low-level southerly windflow. Along the Gulf coast the usual onshore surface pressure gradient and airflow caused by the western portion of the Azores subtropical high pressure cell may provide southerly windflow and moist surface airmasses, but over the study area mid-level stability is induced by the 500 mb ridge, and the southerly flow of surface airmasses frequently provides only continental tropical air from the desert Southwest. Moisture and resulting precipitation in the 142 eastern two-thirds of the country may stem from isentropic flow of "moist tongues" beneath the ridge, in which moist air parcels flowing along lines of constant potential temperature are entrained in the airflow and reach the surface as isolated, moist airmasses far north and west of the Gulf of Mexico (Wexler and Namias, 1936; Namias, 1955; 1963), or moisture and precipitation may invade the continent with the remnants of a tropical storm (Livezey, 1980). In either case the atmospheric moisture is not generally widespread, and precipitation, while perhaps intense in places, is scattered and rarely of long duration. The summers of 1980 and 1952-1954 provided an excellent illustra- tion of the importance of the mean circulation or first component pattern to summertime circulation characteristics and moisture deficiency across the study area. Daily 500 mb patterns during late June, July, and August, 1980, displayed a Gulf of Alaska trough, central states ridge, and east coast trough (Figure 60) that was more persistent and amplified than normal (Livezey, 1980; Wagner, 1980). Throughout the summer, 500 mb features constantly changed in response to moving shortwaves, but, between June 15 and August 31, distinct troughing in the Gulf of Alaska region was noticeable on 49 out of 77 days (Daily Weather Map, Weekly Series 500 mb chart), while more ill- defined troughing in the vicinity of Alaska and the West coast of the U.S. was visible on an additional 15-18 days. Central states ridging was apparent on virtually every day of this time period. The eastern trough centered over Hudson's Bay was deeper than normal in July, which caused the strengthening and displacement of the baroclinic zone and zone of maximum windspeeds farther south than normal over the Great 143 3000 3030 3mm Lao 3180 0.1- 1150 \ 3180 Heights in Meters Source: Livezey, 1980 Figure 60 July, 1980 Mean 700 mb Height This month had severe drought in the central Great Plains. 144 Lakes (Livezey, 1980). In August, the west coast trough broadened to include the Pacific Northwest, while the central states ridge shifted somewhat eastward, and the Hudson's Bay trough was replaced by higher 500 mb heights. Although the August pattern was slightly different from the mean, the central states ridge was still a noticeable and dominating feature (Wagner, 1980). During July, 1980, precipitation in the study area was deficient as the central states ridge and anticyclone dampened convection and shunted cyclonic storms and fronts far north (Livezey, 1980). Daily temperatures frequently exceeded 38°C (100°F) throughout the southern states as the subsidence beneath the ridge induced clear skies which enhanced solar heating of the surface. Dewpoint temperatures generally exceeded 16°C (60°F) throughout the southern states, indicating a good deal of moisture was present in the air, but with such extreme ambient temperatures, the relative humidity in the western portion of the study area was extremely low, and VPD was quite high. Table 5 summarizes the mean dewpoint and maximum temperatures as well as VPD along the western, eastern, and northern fringes of the study area (Oklahoma City, Oklahoma; Nashville, Tennessee; and Fort Wayne, Indiana, respectively). As a result of higher dewpoint temperatures to the east and higher ambient (maximum) temperatures to the west, the mean VPD was 53 mb in Oklahoma and 35 mb in Tennessee (Table 5). In Indiana, dewpoint temperatures were only marginally higher than in Oklahoma (16°C (60°F) in Oklahoma, 18°C (64°F) in Indiana), but maximum temperatures were much lower in Indiana, which led to a mean VPD of only 20 mb (Table 5). Thus, in the western portion of the study area 145 Table 5 July, 1980 Mean Dewpoint and Maximum Temperatures JULY MEAN TEMPERATURE MEAN VPD (0F, °C in parentheses) (mb) Dewpoint Maximum Oklahoma City, OK 60.1 (15.6) 102.4 (39.1) 53 Memphis, TN 71.2 (21.7) 98.2 (36.7) 35 Fort Wayne, IN 64.0 (17.8) 85.3 (29.4) 20 Source: Daily Weather Map Dewpoint temperatures are for 7 AM EST. 146 atmospheric moisture and relative humidity were relatively low, precipitation mechanisms were subdued, and VPD was extremely high. Maximum temperatures were lower farther north and east, and relative humidity was higher. The stronger upper air windspeeds and frequent frontal passages led to normal or above normal precipitation throughout much of the northeast portion of the country (Livezey, 1980). The 1980 climatic pattern of drought in the central Plains and normal-to-wet conditions in the eastern portion of the country (Figure 61) is a result of a synoptic circulation pattern that is the same as the long-term mean 500 mb configuration (Figures 1, 37, 60). During summer, 1980, the persistence of the 500 mb features was anomalous but not without precedent. In particular, the summers of 1952-1954 experienced a similar 500 mb pattern (Livezey, 1980; e.g., Figure 2), which persisted to the exclusion of almost all others during these summers (Mamias, 1955; Hawkins, 1954). The resulting drought during the 1950's has been documented by Borchert (1971), Warrick, et a1 (1977), and Felch (1978) among others. This same general pattern emerges as the long-term mean summer 500 mb pattern (Figure 37), the principal stress day pattern (Figures 23, 25, 27), the first combined component pattern (Figure 33), and the first combined drought pattern (Figure 56). When this 500 mb pattern is persistent and amplified as in July, 1980, or the summers of 1952-1954, drought is a mid-continent occurrence. Days with this synoptic pattern experience an extremely strong gradient of vapor pressure deficit in both east-west and north- south orientations across the study area, and this gradient in plant- water stress is present irrespective of soil moisture conditions. It 147 Above +4 Extreme Wetness +3 to +4 Severe Wetness +2 to +3 Moderate Wetness -2 to +2 Near Normal -2 to -3 Moderate Drought -3 to -4 Severe Drought Below -4 Extreme Drought Source: Weekly Crop and Weather Bull. Figure 61 Palmer Drought Severity Index for August 30,1980 148 is not necessary for drought to occur for there to be strong water stresses in the central states area; it is only necessary that the mean 500 mb pattern occur as a daily synoptic pattern. Daily circulation patterns do vary greatly from the mean, as is evidenced by the greater variability in VPD westward (Figures 15—18). As noted earlier, this gradient implies that the western Plains are at times invaded by moist airmasses and/or cool airmasses which reduce the vapor pressure deficit, while the eastern Gulf coast is much less frequently invaded by dry airmasses. Circulation patterns other than the mean account for most of the precipitation in the study area (e.g., the fifth combined component pattern, Figure 50). These circulation patterns provide synoptic-scale uplift mechanisms and strong Gulf moisture advection not present under the mean 500 mb pattern (Figure 37). How frequently the mean 500 mb pattern occurs as an actual daily synoptic pattern is not wholly answered by this analysis. It was hoped that the principal components analysis would alleviate this problem, but as is pointed out in a following section of this chapter, problems with data suitability prevented adequate estimation of absolute frequency of the mean 500 mb pattern. While this research indicates that a vapor pressure deficit gradient across the ecotone and drought are induced by daily 500 mb flow patterns similar or identical to the mean 500 mb pattern, the frequency of occurrence of these daily synoptic conditions cannot be adequately assessed through this analysis. 149 Principal Components Analysis and 500 mb Patterns Introduction Beyond the formal evaluation of research hypotheses, the most difficult issue raised by this research is the evaluation of the principal components analysis itself. The major problems in the principal components analysis are: 1) Reality of patterns. 2) Interpretation of the first components. 3) Determination of frequency of the patterns. Each of these topics will be treated in turn. Reality of the Patterns One check on the reality of the patterns derived through principal components analysis is to compare the patterns with actual daily weather maps. For example, the August stress-day analysis utilized 29 days, and provided a manageable data set for inspection; the analysis derived only two components with eigenvalues greater than 1.0 (Table 1, Figures 31 and 32). Examination of daily 500 mb charts for August stress days showed that the component score pattern depicted in Figure 32 did indeed occur on a number of days. A day typical of the 500 mb flow pattern prevalent on many August stress days, as well as June and July stress days, is illustrated in Figure 62, while a day having the less frequent pattern associated with the second August component is shown in Figure 63. Within this limited set of days, chosen because of the steep VPD gradient across the southern United States, only a small subset of all circulation patterns was possible. Principal components analysis described these patterns in terms of two components, and flow 150 10300 10700 . Miles y_j I'— 1 O 1150 Heights in Feet Source: Daily Weather Map, August 11, 1969 Figure 62 Typical August Stress-Day First Component 500 mb Pattern This day loads highly on the first component. 151 l Miles 4_j If 1 O 1150 Heights in Feet Source: Daily Weather Map, August 28, 1970 Figure 63 Typical August Stress-Day Second Component 500 mb Pattern This day loads highly on the second component. 152 patterns corresponding to the component score maps can be found in the actual daily 500 mb circulation patterns. Since examples of the principal components derived patterns can be found in the original daily data, and, as is pointed out in the following section, since the first component patterns strongly resemble the mean summer 500 mb flow pattern, the component score maps can be considered real 500 mb pressure height patterns. Other authors (e.g., Perry, 1970) have interpreted principal component patterns as real pressure patterns, or found that patterns depicted by the higher order eigenvectors resemble the most frequently occurring pressure patterns (Kutzbach, 1970). The first and second August stress-day components explained 86.2 and 3.9 percent of the variance respectively, which does not appear to be an accurate representation of their respective frequencies as daily flow patterns, since 25 out of 29 (86%) August stress days should resemble the first component pattern (Figure 32). It appeared that greater variability was present in daily 500 mb patterns than is implied by 86.2 percent explained variance. This problem will be discussed in a succeeding section. Interpretation of First Components Since principal components analysis disaggregates the daily data and tends to combine similar observations of phenomena onto a single component, days with similar flow patterns should be grouped on a single component, and the eigenvalues should provide the frequency of occurrence of the flow pattern. But the first components in the stress 153 day, long-term, or drought analysis consistently explained 72—88% of the variance, which seems rather high considering the daily progression of waves on any given sequence of daily 500 mb charts. Numerous meteorological studies using the eigenvector or principal component approach have found that the first component frequently explains the majority of variance in the data set. For example, Stidd (1967), in a T-mode analysis of mean monthly precipitation in Nevada, found that the first component explained 72% of the variance. Craddock and Flood (1969) found that the first eigenvector of daily 500 mb heights in the Northern Hemisphere explained over 50% of the variance in an R-mode analysis. An R-mode analysis of surface temperature over the Northern Hemisphere yielded a first eigenvector explaining 94.5% of the variance (Kidson, 1975), and the first component in an R—mode analysis of 700 mb temperature in the Northern Hemisphere explained 82.4% of the variance (Heddinghaus and Kung, 1980). None of these studies adequately addressed the significance of such a large proportion of variance explained by a single component. The problem with interpreting the principal components is that, at least for the first one, explained variance is not necessarily frequency of occurrence. Particularly in a single summer season the temporal variation of the 500 mb surface is relatively conservative; the gross pattern of higher pressure heights over the subtropics and lower eights poleward accounts for the majority of the variance in hemispheric 500 mb circulation patterns. Even days which have anomalous flow configurations (for example, Alaskan ridge and Great Plains trough during the summer) still have the overall equator-to-pole gradient of 154 500 mb heights. So much variance is contributed by this gradient rather than perturbations or wave positions that it is incorrect to consider eigenvalues of the original first components as describing absolute frequency of occurrence of their respective synoptic patterns. It is difficult to justify calling the first component pattern the mode based solely on eigenvalues, since information other than absolute frequency of the pattern is contained in the first components. A useful and justifiable interpretation of the first component score maps is that they are the mean flow pattern, and by inference the mode, since principal component analysis assumes a multivariate normal distribution. Principal component analysis frequently extracts as the first component a general tendency of the data (Nie, et al, 1970, p. 482). In addition, the summer mean 500 mb maps and maps of the original first component scores (and the first combined component) all illustrate essentially the same trough/ridge/trough sequence over North America (Figures 33-37), and the positioning of these circulation features agrees with other studies of summertime mean and modal trough/ ridge positions; in particular, Klein and Winston (1958), using 30 day mean 700 mb charts, determined that 30 percent of July mean charts displayed a Gulf of Alaska trough, and over 30 percent of the charts had a Great Lakes trough, while the central states area was prone to ridge conditions. The possibility exists that the original and combined first components do not, in fact, represent the actual mode of the data. Blasing and Lofgren (1980) point out that an eigenvector analysis of sea-level pressure anomalies combined two distinct and opposite synoptic patterns into one compromise eigenvector which approximated neither of 155 the original patterns. If, in the current research, the first components are deriving an efficient variance-reducing function that represents a compromise between two (or more) synoptic patterns, then original first components with the highest explained variance should represent the most efficient reduction of the data and probably the greatest compromise in pattern representation. Just the opposite was true; first components with a higher explained variance always depicted a more amplified wave pattern on the component score maps than first components having a lower explained variance (although statistical significance was lacking). It is inferred that simultaneously high first-component explained variance and amplified waves on first component score maps can only result from persistence of the depicted wave features in the original data. High first-component explained variance alone could be produced by persistent zonal 500 mb patterns or by a compromise solution grouping opposite amplified 500 mb patterns on the same component, but in either case the resulting component score map should depict zonal rather than amplified features. The fact that the first-component percent variance explained and amplification of waves on first component score maps consistently co-vary indicates that the eigenvalues or percent variance explained by first components are, in fact, a measure of relative frequency for the component score map patterns. Along with the fact that principal component analysis generally extracts the most important data pattern with the first component, and since the first component patterns apparently are real, the first component patterns can safely be interpreted as the modal 500 mb patterns. Absolute frequency of the mode cannot be calculated because of the additional variance of the equator-to-pole pressure 156 height gradient described by the first components, but the percent variance explained can be used as relative frequency measures to compare separate original analyses for their tendency toward the mean or relative frequency of the mode. The only assumption necessary to compare frequencies between analyses is that it is the same pattern in each analysis that is being used, and this can be determined by inspect- ing the component score maps. Combined Analysis and Frequency The combined first component is interpreted as the long-term mean and modal 500 mb pattern since all of the original first components load over 0.90 on it. The component score map for the combined first component (Figure 33) lacks definition, in part because, in the combined component analysis in which the scores of all original components were used as input, the original first components contribute only the same amount of variance as any other input component. Since these first components summarize the preponderance of variance in each original analysis, the relative weight or importance of the mean/modal pattern is reduced; in other words, after the persistence extracted by the original first components is largely eliminated, the closer a combined component pattern is to a straight north-south gradient (zonal pattern), the better it approximates all circulation types, including non-first components. As a result, zonal first component patterns loaded higher on the first combined component than did patterns with higher amplitude features (although statistical significance was lacking). Thus, the long-term first component pattern itself is better deduced from the original first components (e.g., Figures 34-36) and 157 the long-term mean 500 mb surface (Figure 37), rather than from the combined first component (Figure 33). The principal value of the combined first component is as a method of grouping original first components, since they all load highly on the combined first component and are all highly intercorrelated (with correlation coefficients greater than 0.92). Therefore, original first components may all be considered similar flow patterns, and the variations among them may be attributed to sampling variations caused by time-period variations in the 500 mb flow pattern. The relative frequency of the first component pattern in the long- term data can be computed from the sum of the eigenvalues of all the original first components (Table 3). In the long-term data 81.5 percent of the variance is explained by the first component patterns, and this value contrasts with the 86-88 percent explained by the stress-day first components and the 78 percent explained by the combined first drought component (Tables 2, 3, 4). The relative frequency of the combined components other than the first may be computed using the accumulated total variance of each component (Table 3), and thus the second combined component, a flow pattern associated with July drought on the Great Plains (Figure 38) is twice as frequent (in the long-term) as the fourth combined component (Figure 47). The fourth and fifth combined components (Figures 47, and 50) have very similar accumulated total variance and explain approxi- mately the same percentage of total variance; they are both about three times as frequent as the sixth combined component (Figure 53). Computation of relative frequency for the third combined component (Figure 41) is complicated by the problem of the original first 158 components loading highly on it. The best estimate of frequency is probably computed without the variance contribution of the first components, leaving the third combined component a relative frequency about the same as the fourth and fifth combined components. That the combined principal components analysis extracted only six components of consequence attests to the conservatism of the atmosphere in a single season; daily variations in mid-tropospheric wind fields are usually not large. The great proportion of variance explained by the mean or modal circulation illustrates how important this pattern is in summer; it may not be as important on an annual or seasonally transgressive basis. Conclusions Three substantive conclusions emerge from this study: 1) vapor pressure deficit appears to be an important variable controlling at least a portion of the pre-settlement forest/grassland transition; 2) the daily 500 mb pressure height pattern causing the strongest vapor pressure deficit gradient across the ecotone is identical to the mean 500 mb pattern; 3) although persistence of the mean 500 mb pattern may cause drought in the study area, other 500 mb circulation patterns are also associated with drought. From the northern borders of Oklahoma and Arkansas south to the Gulf of Mexico, the gradient of strongest summertime VPD matches the east-west vegetation transition from forest to grassland. Northward into Missouri and Kansas, there is no correspondence between VPD gradient and vegetation transition. It is my conclusion that in the southern portion of the study area the broad-scale pre-settlement 159 vegetation distribution was adjusted to climatic stresses (VPD) induced by the most frequently occurring mid-tropospheric windflow patterns; other climatic variables may be responsible for the transition in the northern portion of the study area. Summer days experiencing the strongest vapor pressure deficit gradients across the study area had 500 mb flow patterns similar to the long-term mean summer pattern, and the most frequent daily 500 mb flow pattern in the long-term data was one that strongly resembled the mean 500 mb pattern. Therefore it is the most frequently occurring 500 mb flow patterns that are responsible for creating the strongest VPD gradients across the study area. In addition, the vertical circulations associated with this 500 mb pattern (a ridge over the Great Plains) dampen precipitation mechanisms in the study area and thus reduce soil moisture availability. Continued persistence of this pattern may cause drought, but other circulations can also cause drought in the study area. One implication of these conclusions is that this particular life-form boundary is adjusted to the most frequent atmospheric conditions rather than to catastrophes. The influence of drought and fire in determining the location of the ecotone cannot be ignored; both phenomena enhance the competitive advantage of grassland over trees. But the occurrence of both drought and fire as influences on the forest/grassland transition is strongly controlled by frequent climatic conditions, not by infrequently occurring circulation patterns. Fire undoubtedly played an important role in determining location of the ecotone, and it is a biologic catastrophe, but the reason fire was so 160 influential along the forest/prairie border was because of the frequently dry climatic conditions induced by the most frequent atmospheric circulations. Drought and fire would be associated with the same circulation systems that cause the strong VPD gradient across the study area. Even though local or regional influences such as soils or physiography may obscure the relationship, on a continent- wide scale mean climatic conditions, drought, and fire are all related to mid-tropospheric windflow. Chapter VIII SUMMARY AND SUGGESTIONS FOR FUTURE RESEARCH Summary of Problem, Hypotheses, Methods, and Statistical Results 1) 2) 3) 4) 5) 6) The climatic causes of the pre-settlement forest/grassland transition in the south-central United States are poorly understood. Analysis of daily synoptic weather patterns can provide insight into the mechanism causing elements of weather and climate to vary significantly across the ecotone. Year-to-year climatic fluctuations can be caused by variations in the frequency of different daily 500 mb windflow patterns. Vapor pressure deficit (VPD) has been shown to be a meteorological elment which affects daily water stress in plants and thus photosyn- thesis. In general, grasses can photosynthesize at higher vapor pressure deficits than trees. Three hypotheses were tested: a) Growing season vapor pressure deficits vary geographically across the south central United States, and may thus be responsible for the pre-settlement ecotone. b) The mean (modal) 500 mb circulation pattern in summer is responsible for a strong vapor pressure deficit gradient across the study area. c) Droughts in the central United States may be associated with increased frequency of the mean 500 mb flow pattern. Mean monthly vapor pressure deficits computed using mean monthly values of maximum and dewpoint temperatures were found to be almost identical to mean monthly vapor pressure deficits computed using 161 7) 8) 9) 10) 11) 12) 162 daily vapor pressure deficits derived from daily maximum and dewpoint temperatures. Twenty years (1957-1977) of June, July, and August data were used to calculate and map monthly vapor pressure deficits for a grid of twenty weather stations across the ecotone. Summer mean VPD varied from 22 mb in Tennessee to 30 mb in Oklahoma, and the zone of strongest gradient corresponded to the ecotone in the southern portion of the study area. Stress days were selected on which a vapor pressure deficit gradient corresponded to the ecotone in the southern portion of the study area. Stress days were selected on which a vapor pressure deficit gradient greater than 6-9 mb was present across the ecotone, and mean 500 mb maps and maps of the stress-day deviation of the 500 mb surface from the mean were constructed. The daily 500 mb heights of stress days were subjected to a T-mode principal components analysis. One component was extracted that resembled the mean 500 mb flow pattern and explained approximately 85 percent of the stress day variance. The daily 500 mb pressure heights of June, July, and August, 1961- 1977 were subjected to a T-mode principal components analysis in groups of 60-93 days at a time. The component scores of the first four components from each of these "original" principal components analyses were then subjected to another T-mode analysis, and 6 "combined" components were extracted. The sum of the eigenvalues of the original components loading highly on the 6 combined 13) 14) 15) 1) 2) 163 components constituted over 95 percent of the original variance, and the first combined component accounted for 81.5 percent of the original variance. Maps of the component scores showed that the first combined component was the long-term mean 500 mb flow pattern. The second combined component explained approximately 6 percent of the total variance and depicted a strongly amplified western ridge/eastern trough pattern. It is associated with drought in the central states and is most frequent in July. Three years in which drought was present during the 1970's were used for an analysis of the daily 500 mb flow patterns during drought years. A T-mode principal components analysis was run on the daily pressure heights during the drought years, and the first two components extracted were identical to the first two components extracted by the combined principal components analysis of the long-term data. Summary of Major Conclusions Patterns of computed daily vapor pressure deficit displayed a strong gradient coincident with the southern part of the ecotone. The 500 mb circulation pattern inducing extremely strong gradients of VPD across the ecotone appears to be the same as the mean 500 mb pattern. Principal components analysis of stress days yielded only one 500 mb pattern of any consequence, and this pattern appears similar to the first component pattern of the long-term data set. 164 3) The first component of each principal components analysis was interpreted as the mean and modal long—wave 500 mb pattern. Absolute frequency of this pattern could not be determined because the component summarized the major equator-to-pole pressure height gradient as well as wave positions in the 500 mb surface. 4) Drought years in the 1970's did not experience increased tendency toward the mean 500 mb pattern, rather, the second long-term component pattern increased in frequency significantly. Apparently more than one 500 mb circulatory pattern is associated with drought in the central states. 5) As a result of this analysis I conclude that pre-settlement vegetation distribution in portions of the central states were responses to patterns of daily stress induced by vapor pressure deficits rather than to catastrophic events such as fire and drought, although both may be important adjuncts to the modal stress patterns across the study area. Suggestions for Future Research The two major problems encountered in this research were in the data processing and in the conceptualization of the hypothesis on the relationship between drought and circulation patterns. In the principal components analysis, determination of frequency measures is complicated, particularly for first components, by the use of raw pressure height data. Blasing and Lofgren (1980) used pressure anomaly values for a map correlation analysis, primarily because maps from 165 which the long-term mean and not been subtracted all had a high degree of intercorrelation, and anomaly maps made separation of patterns easier. The problem with using only anomaly data in the current research is that circulations corresponding to the long-term mean pattern would be eliminated (or show up as a zero deviation field, without height contours). Persistence of the mean pattern has important climatic repercussions, and an attempt to classify summer circulation types needs to retain information on the mean pattern as well as on anomalous flow patterns. The problem is that the eigenvalue of the first component is useful only as the broadest interpretation of tendency toward the mean, hot as a frequency measure. In retrospect, since a study such as this one seeks information on flow pattern or wave form, a useful approach might be to construct anomaly maps by subtracting the long-term mean equator-to-pole gradient from the 500 mb data before principal components analysis. A single value of mean 500 mb height would be used for any given latitude to construct the anomalies, and longitudinal variation of the 500 mb surface would not be eliminated. This procedure is analogous to fitting a simple plane or first-order trend surface to the data; the plane or trend surface would have an equator-to-pole slope based on the mean 500 mb equator-to-pole gradient. The long-term mean wave pattern would not be subtracted from the data (as in Blasing and Lofgren, 1980), merely an idealized trend surface describing the summertime equator-to-pole variation of the surface. This procedure would eliminate a major source of variance and allow the principal components to be computed for perturbations of the 500 mb surface, or wave forms, only. Each 166 wave form would have a frequency directly interpretable from eigenvalues, and the mean perturbation of Gulf of Alaska trough/central states ridge would be retained in the data (presumably as the most important factor). Further research in this area should be aimed at better defining summer 500 mb patterns and their frequencies of occurrence. Additional refinements of the principal components technique, such as subtracting a mean latitudinal value of the 500 mb surface from each data point, could make it possible to calculate expected seasonal frequencies of each circulation pattern. Additional map correlation methods (e.g., Blasing and Lofgren, 1980) could be used to calculate the frequency of flow patterns in each summer 1961-1977, thus providing a catalog of circulation type frequencies at a much finer resolution than currently exists (e.g., Dzerdzeevskii, 1968; Kalnicky, 1974). The T-mode princip components analysis is cumbersome and necessitated a second level of analysis to summarize the data, but interpolation of an evenly spaced grid for R-mode analysis is equally ungainly. This study could be improved by selecting 50-100 weather stations at relatively equal distances (non-clustered) for which a complete record exists and using an R-mode principal components analysis. Data for high latitudes might be sparse, and this might necessitate eliminating many U.S. stations to keep equal variance contribution from equal areas, but missing data in the T-mode analysis eventually reduced the usable number of data points to a minimum anyway. In the R-mode analysis, annual frequencies could be determined directly from the component scores, while in the T-mode analysis 167 individual year frequencies are unattainable because of the second level of analysis needed. (If a T-mode analysis could be run using 1500 variables all problems would be surmounted). One procedure not used in this analysis that would be useful in all further daily 500 mb analyses would be to substitute the report of the previous day for any missing data at a particular station. Obviously only a few days per month can be handled this way. The daily autocorrelation of the 500 mb heights at any given place makes this a practical means of handling small amounts of missing data. It was not until the current research was well underway that it was discovered that the computer program (SPSS) would not compute component scores in cases where missing data existed, even though stations with missing data were used to calculate principal components. The combined component score maps suffer from poor resolution because of this problem. The drought hypothesis presents a conceptual rather than a methodological problem, since multiple circulation patterns may cause drought in the study area. Research aimed at discovering 500 mb flow patterns associated with drought could be handled better in 2 ways. On a smaller time scale, individual summers could be treated in order to discover which particular circulation patterns are associated with dryness or drought in a given year (e.g., Harman and Harrington, 1978). On a larger time scale, a catalog of circulation patterns and their yearly frequencies could be developed using principal components analysis or other map correlation methods (e.g., Dzerdzeevskii, 1968; Kalnicky, 1974; Blasing and Lofgren, 1980), and the yearly variation of flow pattern frequencies could be related to some index of drought such 168 as the Palmer Drought Severity Index currently published by the National Weather Service/NOAA. In the first method, drought years would be studied singly for clues about the upper level circulation patterns associated with drought; this method is the classical synOptic study (e.g., Namias, 1955). In the second method, as many years as possible would be studied for the fluctuation of circulation pattern frequencies in the hope that a consistent relationship between drought and a small set of circulation types might be defined. APPENDIX APPENDIX I Days Used In Stress Day Analysis June 1968 13, 22, 27 1969 (none used) 1970 6, 7, 13, 15, 16, 23, 24, 25, 26, 27, 28, 29 July 1968 (none used) 1969 12, 13, 14, 15, 16, 17, 18, 23 1970 5, 6, 7, 16, 17, 18, 19, 22, 25, 26, 27, 29, 30, 31 August 1968 (none used) 1969 7, 8, 9, 11, 12, 13, 16, 17, 19, 20 1970 2, 3, 5, 6, 7, 8, 9, 11, 12, 13, 14, 15, 18, 19, 26, 27, 28, 29, 30 169 BIBLIOGRAPHY BIBLIOGRAPHY Ackerman, E.A., 1941; The Koeppen classification of climates in North America. Geogr. Rev.; 31: 105-111. Albertson, F.W. and Weaver, J.E., 1945; Injury and death or recovery of trees in prairie climate. Ecol. Mon.; 15: 395—433. Armstrong, J.S., 1967; Derivation of theory by means of factor analysis, or, Tom Swift and his electric factor analysis machine. Amer. Stat.; 21: 17-21. Bark, L.D., 1978; History of American droughts; in N.J. Rosenberg ed., North American Droughts, AAAS Symposium 15. Westview Press, Boulder. Barnes, S.L., 1976; Severe local storms: concepts and understanding. Bull. Am. Met. Soc.; 57: 412-419. Black, C.C., 1971; Ecological implications of dividing plants into groups with distinct photosynthetic productive capacity; in J.B. Cragg ed., Advances in Ecological Research, v. 7. Academic Press, London and New York, pp. 87-114. Blalock, H.M., 1972; Social Statistics, second ed. McGraw Hill, New York, 583 pp. . Blasing, T.J. and Lofgren, G.R., 1980; Seasonal climatic anomaly types for the North Pacific sector and western North America. Mon. Wea. Rev.; 108: 700-719. Borchert, J.R., 1950; The climate of the central North American grasslands. AAAG; 40: 1-39. Borchert, J.R., 1971; The dust bowl in the 1970's. AAAG; 61: 1-22. Braun, E.L., 1950; Deciduous Forests of Eastern North America. Hafner Publ. Co., New York. Bruner, W.E., 1931; The vegetation of Oklahoma. Ecol. Mon.; 1: 99-188. Bryson, R.A., 1966; Air masses, streamlines, and the Boreal forest. 170 171 Chaney, W.R. and Kozlowski, T.T., 1969; Seasonal and diurnal expansion and contraction of leaves and fruits of English Morello Cherry. Court, A., 1973; The climate of the coterminous United States; in R.A. Bryson and F.K. Hare eds., Climates of North America. Elsevier, Amsterdam. Craddock, J.M. and Flood, C.R., 1969; Eigenvectors for representing the 500 mb geopotential surface over the northern hemisphere. Quart. Journ. Roy. Met. Soc.; 95: 751-760. Curtis, J.T. and McIntosh, R.P., 1951; The misuse of climate as an indicator of vegetation and soils. Ecology; 32: 343-345. Daubenmire, R., 1956; Climate as a determinant of vegetation distri- bution in eastern Washington and northern Idaho. Ecol. Mon.; 26: 131-54. Davis, A.M., 1977; The prairie-deciduous forest ecotone in the upper Middle West. AAG; 67: 204-213. Davis, A.M., 1976; Pleistocene biogeography of temperate deciduous forests. Geoscience and Man; 13: 13-26. Delcourt, P.A., Delcourt, R.R., Brister, R.C. and Lackey, L.E., 1980; Quaternary vegetation history of the Mississippi embayment. Quat. Res.; 13: 111-132. Diaz, H.F. and Quayle, R.C., 1980; The climate of the United States since 1895: spatial and temporal changes. Mon. Wea. Rev. 108: 249-266. Dickman, D.I., 1973; Carbohydrate relations; in Tree Physiology Colloguim, Coop. Ext. U. of Wis. Dickson, R.R., 1972; Weather and circulation of August, 1972. Mon. Wea. Rev. 100: 819-824. Dodd, A.U., 1965; Dew point distribution in the United States. Mon. Dyer, A.J. and Trlica, M.J., 1972; Carbon dioxide exchange of short grass sods. Colorado-Wyoming Acad. Sci. Jour.; 7: 28—29. Dyksterhuis, E.J., 1948; The vegetation of the Western cross timbers. Ecol. Mon.; 18: 325-376. Dyksterhuis, E.J., 1957; The savannah concept and its use. Ecology; 38: 435-442. Dzerdzeevskii, B.L., 1968; Circulation Mechanisms in the Atmosphere of the Northern Hemisphere in the Twentieth Century. Trans. by R. Goedecke, B.F. Berryman ed., U. of Wis. Met. Dept., Madison. 172 Dzerdzeevskii, B.L., 1969; Climatic epochs in the twentieth century and some comments on the analysis of past climates; in H.E. Wright ed., Proceedings Seventh Cong;_of Int. Assoc. Quat. Res., pp. 49-60. ESSA, 1966; Normal monthly number of days with precipitation of .5, 1.0, 2.0 and 4.0 inches or more in the conterminous United States. Weather Bureau Tech. Paper 57. ESSA, 1967; Climatologic probability of precipitation for the conterminous United States. ESSA Tech. Report WB-S. ESSA, 1968; Climatic Atlas of the United States. U.S. Dept. of Commerce. Estes, E.T., 1970; Dendrochronology of Black oak (Quercus velutina), White oak (Quercus alba), and shortleaf pine (Pinus echinata) in the Central Mississippi Valley. Ecol. Mon.; 40: 295-316. Evans, L.T. ed., 1963; Environmental Control of Tree Growth. Academic Press, London and New York. Federer, C.A. and Gee, G.W., 1976; Diffusion resistance and xylem potential in stressed and unstressed northern hardwood trees. Ecology; 57: 975-984. Felch, R.E., 1978; Drought characteristics and assessment; in N.J. Rosenberg ed., North American Droughts, AAAS Symposium 15. Westview Press, Boulder. Fritts, R., 1965; Tree-ring evidence for climatic changes in western North America. Mon. Wea. Rev.; 93: 421—443. Fritts, H., 1976; Tree Rings and Climate. Academic Press, New York. Fritts, R.C., Lofgren, G.R., and Gordon, G.A., 1979; Variations in climate since 1602 as reconstructed from tree rings. Quat. Res.; Green, R.A., 1961; The weather and circulation of July 1961. Mon. Gregory, S., 1975; On the delimitation of regional patterns of recent climatic fluctuations. Weather; 30: 276—287. Gruger, E., 1972; Late Quaternary vegetation development in south- central Illinois. Quat. Res.; 2: 217-231. Harman, J.R., 1971; Tropospheric Waves, Jet Streams, and United States Weather Patterns. Comm. on Coll. Geo. Res. Paper 11, AAG, Washington, D.C. 173 Harman, J.R. and Braud, D.H., 1975; Flow patterns at the 500 mb level and the prairie-forest ecotone in Canada. Mich. Academician; 7: 347-356. Harman, J.R. and Harrington, J.A. Jr., 1977; In support of synoptic climatology; in H.A. Winters and M.K. Winters eds., Applications of Geographic Research. Michigan State University, East Lansing. Harman, J.R. and Harrington, J.A., 1978; Contrasting rainfall patterns in the Upper Middle West. AAAG; 68: 402-413. Harmon, H.H., 1967; Modern Factor Analysis. University of Chicago Press, Chicago. Hatch, M.0., Osmond, C.B., and Slayter, R.C., 1971; Photosynthesis and Photorespiration. Wiley-Interscience, New York. Hawkins, H.E. Jr., 1954; The weather and circulation of July 1954. Mon. Wea. Rev.; 82: 209-217. Heddinghaus, T.R. and Kung, E.C., 1980; An analysis of climatological patterns of the Northern Hemispheric circulation. Mon. Wea. Rev.; 108: 1-17. Helms, J.A., 1976; Factors influencing net photosynthesis in trees: an ecological approach; in M.G.R. Cannell and F.T. Last eds., Tree Physiology and Yield Improvement. Academic Press, New York. Hunt, C.B., 1974; Natural Regions of the United States and Canada. w,H,FTeeman, 725 pp. Jarvis, M.S., 1963; A comparison between the water relations of species with contrasting types of geographical distribution in the British Isles; in A.J. Rutter and F.H. Whitehead eds., The Water Relations of Plants. Blackwell, London. Jarvis, P.B. and Jarvis, M.S., 1963; The water relations of tree seedlings. 1. Growth and water use in relation to soil water potential. Phys. Plant; 16: 215-235. Johnson, F.L. and Risser, P.G., 1972; Some vegetation—environment relations in the upland forests of Oklahoma. J. of Ecology; 60: 655—663. Kalnicky, R.A., 1974; Climatic change since 1950. AAAG; 64: 100-113. Kay, P.A., 1979; Multivariate statistical estimates of Holocene vegetation and climate change, Forest-Tundra zone NWT Canada. Quat. Res.; 11: 125-140. Kelly, D.L., Schaefer, J.T., McNulty, R.P., Doswell, C.A., and Abbey, R.F. Jr., 1978; An augmented tornado climatology. Mon. Wea. Rev.; 106: 1172-1183. 174 Kendall, H.M., 1935; Notes on climatic boundaries in the Eastern United States. Geogr. Rev.; 25(1): 117-124. Kidson, J.W., 1975; Eigenvector analysis of monthly mean surface data. Mon Wea. Rev.; 103: 177-186. King, J.E., 1973; Late Pleistocene palynology and biogeography of the Western Missouri Ozarks. Ecol. Mon; 43: 539-565. King, J.E. and Allen, W.H., 1977; A Holocene vegetation record from the Mississippi River Valley, Southeastern Missouri. Quat. Res.; 8: 307-323. Klein, W.H. and Winston, J.S., 1958; Geographical frequency of troughs and ridges on mean 700 mb charts. Mon. Wea. Rev.; 86: 344-358. Knox, J.C., 1972; Valley alluviation in southwestern Wisconsin. AAAG; 62(3): 401-410. Koeppen, W., 1936; Das geographosche system der klimate; in W. Koeppen ed., Handbuch der Klimatology. Berlin. Kozlowski, T.T. ed., 1968; Water Deficits and Plant Growth. Academic Press, New York. Kozlowski, T.T., 1979; Tree Growth and Environmental Stress. U. Wash. Press, 192 pp. Kramer, P.J., 1969; Plant and Soil—Water Relationships. McGraw-Hill, New York. Kramer, P.J. and Kozlowski, T.T., 1960; Physiology of Trees. McGraw- Hill, New York. Kucera, C.L. and Martin, 8.0., 1957; Vegetation and soil relation- ships in the glade region of the southwestern Missouri Ozarks. Ecology; 38: 285-291. Kucera, C.L. and McDermott, R.E., 1955; Sugar maple-basswood studies in the forest/prairie transition of central Missouri, American Midland Naturalist 54: 495-503. Kuchler, A.W., 1964; Potential natural vegetation of the conterminous United States. Am. Geogr. Soc., Special Publication No. 36 (map). Kutzbach, J.E., 1967; Empirical eigenvectors of sea-level pressure, surface temperature and precipitation complexes over North America. J. App. Met.; 6: 791-802. Kutzbach, J.E., 1970; Long-scale features of monthly mean Northern Hemisphere anomaly maps of sea—level pressure. Mon. Wea. Rev.; 98: 708-716. 175 Larcher, W., 1969; The effect of environmental and physiological variables in relation to dry matter production by trees. Photosynthetica; 3: 150-166. Lawson, M.P., 1974; The Climate of the Great American Desert, Reconstruction of the Climate of Western Interior United States, 1800-1850. Univ. Neb. Stud. N.S. 46, Lincoln. Levitt, J., 1972; Responses of Plants to Environmental Stress. Academic Press, London and New York. Lieth, H. and Box, E., 1972; Evapotranspiration and primary productivity: C.W. Thornthwaite Memorial Model Publ. in Climat. 25: 3, C.W. Thonrthwaite Assoc. Lincare, E.T., 1977; A simple formula for estimating evaporation rates in various climates, using temperature data alone. Agri. Met.; 18: 409-424. Livezey, R.E., 1980; Weather and circulation of July, 1980. Mon. Wea. Rev.; 108: 1708-1716. Long, M.J., 1966; A preliminary climatology of thunderstorm penetrations of the tropopause in the United States. J. App. Met.; 5: 467-473. Lopushinsky, W., 1969; Stomatal closure in conifer seedlings in response to leaf moisture stresses. Bot. Gaz.; 130: 258-265. Lott, G.A., 1976; Precipitable Water Over the United States. NOAA Tech. Rept. NWS 20. Lowe, P.R., 1977; An approximating polynomial for the computation of saturation vapor pressure. J. App. Met.; 16: 100-103. Ludlam, D.M., 1980; Weatherwatch. Weatherwise; 33: 228-235. Major, J., 1963; A climatic index to vascular plant activity. Ecology; 44: 485-498. Mather, J.R. and Yoshioka, G.A., 1968; The role of climate in the distribution of vegetation. AAAG; 58: 29-41. Moor, P.D., 1974; Misunderstandings over C4 carbon fixation. Nature; 252: 438-439. Namias, J., 1950; The index cycle and its role in the general circulation. J. of Met.; 7: 130-139. Namias, J., 1951; The great Pacific anticyclone of winter 1949-50: a case study in the evolution of climatic anomalies. J. of Met.; 8: 251—261. 176 Namias, J., 1952; The annual course of month-to-month persistence. Bull. Am. Met. Soc.; 33: 279-285. Namias, J., 1954; Further aspects of month-to—month persistence in the mid-troposphere. Bull. Am. Met. Soc.; 35: 112-117. Namias, J., 1955; Some meteorological aspects of drought. Mon. Wea. Rev.; 83: 199-205. Namias, J., 1963; Surface-atmosphere interactions as fundamental causes of drought and other climatic fluctuations; in Changes of Climate, Arid Zone Research XX, UNESCO, pp. 345-359. Namias, J. and Wagner, J.A., 1971; Warm continental anticyclone with peripheral moist tongues--a recent example illustrated by satellite photographs. Mon. Wea. Rev.; 99: 162-164. Natir, L. and Kousalova, I., 1965; The effect of relative air humidity on the development of water saturation deficit in the cut leaves of two varieties of barley; in B. Slavik ed., Water Stress in Plants, Czech. Acad. Sci., D.W. Junk, The Hague. Nelson, J.G. and England, R.E., 1971; Some comments on the causes and effects of fire in the Northern Grasslands area of Canada and the nearby United States ca. 1750-1900. (kun Geographer; 15(4): 295-306. Nie, N.H., Hull, C.H., Jenkins, J.G., Steinbrenner, K., and Bent, D.H., 1970; SPSS Statistical Package for the Social Sciences, second ed., McGraw-Hill, New York. Palmen, E. and Newton, C.W., 1969; Atmospheric Circulation Systems. Academic Press, New York, 603 pp. Paulhus, J.L.H. and Miller, J.F., 1964; Average annual precipitation from daily amounts of 0.50 inch or greater. Mon. Wea. Rev.; 92: 181-186. Pelton, W.L., King, K.M., and Tanner, C.B., 1960; An evaluation of the Thornthwaite method for determining potential evapotranspiration. Agron. J.; 52: 387-395. Penman, B.L., 1948; Natural evaporation from open water, bare soil and grass. Proc. Roy. Soc. London (A); 193:. 120-145. Perry, A.H., 1970; Filtering climatic anomaly fields using principal component analysis. Trans. Inst. Brit. Geog.; 50: 55—72. Read, R.A., 1952; Tree species occurrence as influenced by geology and soil on an Ozark north slope. Ecology; 33: 239-246. Rice, E.L., 1965; Bottomland forests of north central Oklahoma. Ecology; 46: 708-714. 177 Rice, E.L. and Penfound, W.T., 1959; The upland forests of Oklahoma. Ecology; 40: 593-608. Riefler, R.F., 1978; Drought: an economic perspective; in N.J. Rosenberg ed., North American Droughts, AAAS Symposium. Westview Press, Boulder. Ritchie, G.A., 1971; Transpiration, water potential and stomatal activity in relation to microclimate in Abies amabilis and Abies procera in a natural environment. PhD Thesis, U. of Wash., Seattle, cited in Helms, J.A., 1976; Factors influencing net photosynthesis in trees: an ecological approach; in M.G.R. Cannell and F.T. Last eds., Tree Physiology and Yield Improvement. Academic Press, New York, pp. 55-78. Ritchie, J.T., 1974; Atmospheric and soil water influences on the plant water balance. Agri. Met.; 14: 183-198. Rosenberg, N.J., 1974; Microclimate: The Biological Environment. Wiley and Sons, New York, 315 pp. Rosenberg, N.J. ed., 1978; North American Droughts, AAAS Symposium 15. Westview Press, Boulder, 177 pp. Rosenberg, N.J., Hart, H.E., and Brown, K.W., 1968; Evapotranspiration-- Review of Research. Nebr. Agr. Exp. Sta. Misc. Bull. No. 20, 80 pp. Rosenzweig, M.L., 1968; Net primary productivity of terrestrial communities: prediction from climatological data. Am. Naturalist; 102: 67-74. Rummel, R., 1970; Applied Factor Analysis. Northwestern Univ. Press, Evanston, Ill. Salisbury, F.J., 1926; The geographical distribution of plants in relation to climatic factors. Geogr. Journ.; 67: 312-335. Sawyer, J.O. and Lindsey, A.A., 1963; The Holderidge bioclimatic formations of the eastern and central United States. Indiana Acad. Sci. Proc.; 73: 105-112. Schnell, G.D., Risser, P.G., and Helsel, J.F., 1977; Factor analysis of tree distribution patterns in Oklahoma. Ecology; 58: 1345- 1355. Skaggs, R.H., 1978; Climatic change and persistence in western Kansas. AAAG; 68: 73-80. Slayter, R.O., 1967; Plant-Water Relationships. Academic Press, New York, 366 pp. Stidd, C.K., 1967; The use of eigenvectors for climatic estimates. J. App. Met.; 6: 255-264. 178 Taubensee, R.E., 1976; Weather and circulation of June, 1976. Mon. Wea. Rev.; 104: 1200-1205. Thornbury, W.D., 1965; Regional Geomorphology of the United States. Wiley, New York. Thornthwaite, C.W., 1931; The climates of North America according to a new classification. Geogr. Rev.; 21: 633-655. Thornthwaite, C.W., 1948; Toward a rational classification of climate. Geogr. Rev.; 38: 55-94. Thornthwaite, C.W. and Hare, F.K., 1955; Climatic classification in forestry. Unasylva; 9(2): 50-59. Transeau, E.N., 1935; The prairie peninsula. Ecology; 16: 423-437. Wagner, A.J., 1976; Weather and circulation of July 1976. Mon. Wea. Rev.; 104: Wagner, A.J., 1980; Weather and circulation of August 1980. Mon Wea. Rev.; 108: 1924-1932. Wagner, A.J. and Namias, J., 1938; Thunderstorm forecasting with the aid of isentropic charts. Bull. Am. Met. Soc.; 19: 1-14. Wahl, E.W., 1968; A comparison of the climate of the eastern United States during the 1830's with the current normals. Mon. Wea. Wahl, E.W. and Lawson, T.L., 1970; The climate of the mid-nineteenth century United States compared to the current normals. Mon. Waring, R.H. and Major, J., 1964; Some vegetation of the California Coastal Redwood Region in relation to gradients of moisture, nutrients, light, and temperature. Ecol. MOn.; 34: 167-215. Warrick, R.A., Trainer, P.B., Baker, E.J., and Brinkman, W.A., 1975; Drought Hazard in the United States. Monograph, Inst. Behav. Sci., Univ. Col., Boulder, Col. Weakly, H.I., 1965; Agric. Eng.; 46: 85, cited in N.J. Rosenberg ed., North American Droughts, AAAS Symposium 15. Westview Press, Boulder. Weaver, J.E., 1954; North American Prairie. Johnson Pub. Co., Lincoln, Neb. Weaver, J.E., 1960; Flood plain vegetation of the central Missouri Valley and contacts of woodland with prairie. Ecol. Mon.; 30: 37-64 0 179 Webb, T. and Bryson, R.A., 1972; Late and postglacial climatic changes in the northern Midwest, USA. Quantitative estimates derived from fossil pollen spectra by multivariate statistical analysis. Quat. Res.; 2: 70-115. Wells, P.V., 1970; Postglacial vegetational history of the Great Plains. Science; 167: 1574-1581. Wright, H.E. Jr., 1968; History of the prairie peninsula; in R.E. Bergstrom ed., The Quaternary of Illinois. Special Publication No. 14, U. of Ill. Coll. of Agr., Urbana. Wright, H.E., 1970; Vegetational history of the Central Plains; in W. Dort and J.K. Jones Jr. eds., Pleistocene and Recent Environments of the Central Great Plains. U. of Kansas Special Publication No. 3, U. Press, Kansas. Wright, H.E., 1976; The dynamic nature of Holocene vegetation: a problem in paleoclimatology, biogeography, and stratigraphic nomenclature. Quat. Res.; 6: 581-596. U.S. Dept. of Commerce/NCAA; Climatological Data. U.S. Dept. of Commerce/NOAA; Climatological Data, National Summayy. U.S. Dept. of Commerce/NOAA; The Daily Weather Map (Weekly Series). U.S. Weather Bureau, 1955; Rainfall intensity-duration-frequency curves. U.S. Weather Bureau Tech. Paper 25. Zahner, R. and Stage, A.R., 1966; A procedure for calculating daily soil moisture stress and its utility in regressions of tree growth on weather. Ecology; 47: 64-74.